ijerph3

Med-MDPI/ijerph_3/ijerph-16-05-00766.txt

@@ -0,0 +1 @@
+These authors contributed equally to this work.Out-of-hospital cardiorespiratory arrest is one of the leading causes of death in the Western world. Early assistance with quality Cardiopulmonary Resuscitation (CPR) and the use of a defibrillator may increase the percentage of survival after this process. The objective of this study was to evaluate the effect of CPR training and the management of an Automatic External Defibrillator (AED). A descriptive, cross-sectional, observational study was carried out among students in the first year of a Nursing and Physiotherapy degree of the University of León. To achieve this goal, a theoretical-practical educational intervention of four hours’ duration which included training on CPR, AED and Basic Life Support (BLS) was carried out. A total of 112 students were included. The results showed an increase in theoretical knowledge on BLS as well as on CPR and AED, and practical skills in CPR and AED management. A theoretical exposition of fifteen minutes and the practical training of CPR wasenough for the students to acquire the necessary theoretical knowledge, although the participants failed to reach quality criteria in CPR. Only 35.6% of students reached the right depth in compressions. Also, ventilation was not performed properly. Based on the results, we cannot determine that the percentage of overall quality of CPR was appropriate, since 57.6% was obtained in this respect and experts establish a value higher than 70% for quality CPR. There was a clear relationship between sex, weight, height and body max index (BMI), and quality CPR performance, being determinant variables to achieve quality parameters. Currently, Basic Life Support training in most universities is based on training methods similar to those used in the action described. The results obtained suggest implementing other training methods that favour the acquisition of quality CPR skills. Out-of-hospital cardiorespiratory arrest (OHCA) is one of the leading causes of death in Europe, being potentially reversible if treated at an early stage [1]. The percentage of survival of patients in which CPR is conducted immediately after the cardiorespiratory arrest is two to four times higher than that of those who do not receive it, so the start of assistance by witnesses takes on special importance [2]. This survival rate increases when a defibrillator is used in addition to performing CPR, time being a key factor to maximise this percentage [1]. High-quality CPR is of vital importance and is the cornerstone of a series of actions that can reverse the state of OHCA in patients. Quality CPR consists of providing an appropriate frequency and depth of compressions, a correct chest decompression, minimised interruptions, and non-exceeded appropriate volume of ventilations [1]. However, several studies suggest that quality CPR performance is not optimal in most cases, even among healthcare staff that have previously been instructed in this kind of assistance [3,4]. The teaching of cardiopulmonary resuscitation is complex as it includes both theoretical knowledge and psychomotor skills. Acquired skills diminish in a span of 3 to 6 months, which implies a need to develop effective learning and retraining strategies [5,6,7,8,9,10].Among the various methods of CPR training for Health Sciences students, we find some that propose reading for the theoretical training, educational videos on CPR and the defibrillator [7], training in knowledge and skills with its subsequent feedback [5], theoretical lessons taught by an instructor for the acquisition of knowledge, manikins to learn skills [8], high-fidelity simulation, online theoretical lessons and simulation with manikins [11], self-directed teaching, and traditional teaching for retraining groups [12]. However, there is no universally accepted and proven gold standard efficiency method.There is no consensus about how long the training must be. In the works consulted by the authors, from one up to four hours of training are proposed. The same applies to the period of time after which knowledge and skills should be recycled, ranging from one to six months after the end of the course [2,5,7,8]. The objective of this study was to explore the effect of training in the acquisition of knowledge of Basic Life Support (BLS), CPR skills and the management of Automatic External Defibrillator (AED) in Health Sciences university students.A quasi-experimental, before-after type of study with a control group was conducted among students in their first year of the Nursing Degree and Physiotherapy Degree at the University of León (Spain). The Nursing degree is taught at the campuses of León and Ponferrada, while the Physiotherapy degree is only taught in Ponferrada. The study was developed during the academic year 2017–2018.For the procedure, 6 standard manikins Resusci Anne (Laerdal®, Stavanger, Norway), two torsos of CPR manikins without feedback, 3 AEDs, and 6 computers with the Resusci Anne Skill Reporter (Laerdal®, Stavanger, Norway) software were used as material resources. Human resources included 6 supervisors, two of them nurses and instructors in Basic Life Support (BLS) by the Spanish Society of Intensive Medicine, Critical Care and Coronary Units (SEMICYUC), two nurses and two nursing students with specific training in CPR. All of them had performed the correct BLS sequence and had obtained a QCPR higher than 95%.The knowledge questionnaire was used for the pre-test and the post-test. It consisted of 10 items and was prepared ad hoc. For its development, the recommendations on CPR made in 2015 by the European Resuscitation Council (ERC) were consulted. The questionnaire consisted of ten questions with four multiple-choice options, one being correct. The estimated time for completion was 7 minutes. After the questionnaire, a document was delivered with instructions to participants, where data confidentiality was guaranteed.A standard manikin Resusci Anne (Laerdal®, Stavanger, Norway) was used for the practice and for results measurement. Data were obtained from the CPR Resusci Anne Skill Reporter (Laerdal®, Stavanger, Norway) meter, already employed in similar studies [13,14,15]. The system was programmed using the CPR parameters proposed by the ERC in 2015: 50–60 mm depth of compression, 100–120 compressions per minute of frequency, and 500–600 cm3 of inspiratory volume [1]. The information provided by the meter was not shown to the participants during the evaluation. The obtained data were used to create an individual report of CPR of each participant. In addition, the exertion applied by the participants was analysed after their performance by means of the Borg rating of perceived exertion scale (RPE scale) [16].The dependent variables of the study were those derived from the questionnaire regarding the theoretical part and those linked to CPR performance—both the subjective and the objective evaluations—as for the practical part. Table 1 shows the relationship between the variables derived from the questionnaire and those derived from the CPR.The knowledge questionnaire included independent variables of the pre-test or post-test, date of birth, weight, height, body mass index, belonging to the Campus of León or Ponferrada, and university degree of the participants: Nursing or Physiotherapy.The study population was constituted by first-year students of Nursing and Physiotherapy (León and Ponferrada campuses) of the University of León as convenience sampling. The study population included students who were enrolled in the first course of the Nursing and Physiotherapy degree during the study period and who voluntarily decided to participate by signing the informed consent. Those participants who had done some BLS or CPR course in the two years prior to the completion of the action were included. The exclusion criteria were: incorrect, deficient, or incomplete filling of the questionnaire.A subjective sampling technique by reasoned decision was followed, according to the criteria of qualification and campus of the participants. The sample was considered relevant given its homogeneous nature with respect to the reference population for the criteria of belonging to the same branch of knowledge. It was composed of 3 groups of students (Nursing Degree, Campus of León, Nursing Degree, Campus of Ponferrada, and Physiotherapy Degree, Campus of Ponferrada), obtaining a total of 112 students.A theoretical and practical educational action took place, aimed at the training of Health Sciences students of the University of León in BLS and AED operating techniques. The action was dynamic and participatory and was developed from an approach that favoured the students’ motivation and awoke their interest. The training was conducted by following the recommendations of the ERC in its latest guidelines updated in 2015 [1].The educational action was structured in a single session of 4 h for each group, which was organised as follows:A: 20’. Informed verbal consent and pre-test knowledge questionnaire.B: 15’. Theoretical lesson.C: 60’. BLS sequence practical training (20’), AED practical training (20’), CPR practical training (20’).D: 15’. Break.E: 120’. Practical evaluation, exertion applied evaluation.F: 10’. Post-test knowledge questionnaire.A: 20’. Informed verbal consent and pre-test knowledge questionnaire.B: 15’. Theoretical lesson.C: 60’. BLS sequence practical training (20’), AED practical training (20’), CPR practical training (20’).D: 15’. Break.E: 120’. Practical evaluation, exertion applied evaluation.F: 10’. Post-test knowledge questionnaire.All students performed a minimum of 5 min of chest compressions and a BLS and AED sequence 5 times.Both the instructor/student and the manikin/student ratio were 1/6. The manikin/instructor ratio was 1/1.After the training, the students completed the initial questionnaire and simulated CPR on a manikin for two minutes. The clinical case was the same for all individual tests.In order to minimise bias, the evaluation was developed by other instructors who were not their coaches. Prior to the training, they were informed about the worksheet and the key points to consider.The information was recorded in a database created with the Epi Info™ software (Centers for Disease Control and Prevention, Atlanta, GA, USA), also used for statistical analysis. The qualitative variables were reported as relative frequencies and percentages, while the quantitative variables were presented as mean and standard deviation. The bivariate analysis was performed using Student’s T-distribution, Chi-squared test, using the ANOVA test or non-parametric Kruskall-Wallis test for the bivariate analysis according to the homogeneity or the lack of it in the variance. The Pearson correlation coefficient was used to analyse the linear relationship between two random variables. A difference was declared significant when the probability of type I error was equal to or less than 5%, which was assessed using the Pearson p value, with p ≤ 0.05 for that probability value. A linear regression analysis was performed to study the potential association of independent variables (weight, height, age, and BMI) and dependent variables.This study was approved by the Ethics Committee for Research with medicinal products of the Health Areas of León and Bierzo, with registration number: 1874 (24/04/2018). Principles of informed consent and confidentiality were observed during data collection. The students were assured that the fact that they did or did not participate would in no way affect their academic progress. Necessary permissions were received from the directorate of the Faculty. Information was gathered on 112 students. There was a 56% student participation (112/200): 31.3% (35/112) were from the Physiotherapy Degree; 36.6% (41/112) and 32.1% (36/112) were from the Nursing Degree of the campuses of Ponferrada and León, respectively. The age average among the Nursing (León and Ponferrada) and Physiotherapy students was 20.8 years (standard deviation (SD) = 3.2; 18 as a minimum and a maximum of 37). In total, 25.9% of the participants were men and 74.11% were women. They had an average BMI of 22.4 (SD = 2.9; 16.7 as a minimum and a maximum of 30.9) kg/m2. No statistical differences were found between the groups analysed except in their height, where the Physiotherapy students presented higher values (p < 0.01).Data were obtained from 112 students, of which 11.6% (13/112) had been trained in CPR in the two previous years for the completion of the action taken. Knowledge improvement was significant, with a mean of 5.9 out of 10 score (SD = 2) in the pre-test and of 9.6 (SD = 0.6) in the post-test (p < 0.001). The stated action improved the post-test scores in a statistically significant way, both by qualification (p < 0.01) and by university campus (p < 0.01).Figure 1 describes the increase of knowledge stratified by degree (Nursing and Physiotherapy) and by campus (León and Ponferrada).Figure 2 presents the mean global scores of both degrees in both campuses obtained by the students in each of the knowledge questions evaluated before and after the action. There were major differences regarding the questions ‘What is the appropriate depth for chest compressions?’ and ‘What is the correct frequency to perform compressions on victims of any age?’ with a difference of 7.5 and 6.3 points, respectively. There was a significant improvement in the results obtained in all questions.Table 2 shows the results obtained by participants in the subjective evaluation of the BLS sequence. The outcomes obtained in most of the items were higher than 75%. The completion of the head tilt/chin lift manoeuvre (89.3%; 100/112), the maintenance of the head tilt/chin lift manoeuvre (84.8%; 95/112), the assessment of the presence of breathing for 10 s (80.4%; 90/112), the proper rhythm of compressions (76.8%; 86/112), the correct opening of the airway (75%; 84/112), nose clamping (83.9%; 94/112), and the insufflation of air for a minute (86.6%; 97/112) were the items with scores lower than 90%. As for the AED, the values obtained were over 90% in all the studied parameters: turn AED on (98.2%; 100/112), follow instructions (94.6%; 106/112), properly placed pads (96.4%; 108/112), performs compressions while charging (95.5%; 205/112), performs minimal interruptions (90.2%; 101/112), and continues the CPR after the electrical shock (97.3%; 109/112).Information on 112 students from the Nursing and Physiotherapy degrees was gathered at the campuses of León and Ponferrada. The mean number of compressions (TNC) in two minutes was 170.7 (SD = 21.9), reaching mean values of 74.7% (SD = 6.8%) for continuous compression (ACC), with the correct position of the hands being 98.8% of the cases.The mean depth (MD) obtained was 44.9 mm, with a minimum of 22 mm and a maximum depth of 69 mm. The mean percentage of compressions (MPC) performed to the proper depth was 35.6% (SD = 27), while the mean percentage of correct decompression (PCD) was 78.7% (SD = 38.9). In total, 53.2% (SD = 34.1) of the compressions were performed at a proper rhythm (PCPR), obtaining 114.3 (SD = 12.7) as the mean value of compressions (mean rate) per minute performed by the participants.A significant correlation was found between the mean depth of compressions and the weight (r = 0.5), height (r = 0.4), and BMI (r = 0.4); between the percentage of correct compression and weight (r = 0.5) and BMI (r = 0.4); between the percentage of quality of CPR (r = 0.4) and weight and, finally, between the percentage of compressions at a proper rhythm and weight (r = 0.8) (p < 0.0001 in all cases).The percentage of CPR quality presented a statistical relationship with the weight (b = 0.9; p < 0.001), height (b = 0.9; p = 0.001), and age (b = 2.1; p = 0.008). Compression depth mean was also statistically related to weight (b = 0.4; p < 0.001), height (b = 0.4; p < 0.001), and age (b = 0.7; p < 0.001) and, in the same way, the correct depth percentage showed a relationship with weight (b = 1.7; p < 0.001), height (b = 1.5; p < 0.001), and age (b = 2.9; p = 0.013).Information was gathered on administered air volume of 91 participants. The mean volume (MV) reached was 454.9 (SD = 296.3) ml per ventilation. In total, 15.5% (SD = 15.7) ventilations were performed at an appropriate volume (AVV) and students had a mean of 3.2 (SD = 1.9) ventilations per minute (Mean Ratio).The percentage of students who carried out ventilations with quality volume (ranging between 500 and 600) was 39.6% (36/91). No statistically significant differences by sex and degree were found.The global percentage of quality CPR (QCPR) ranged from the 57.6% (SD = 26.3) on average, with 97% being the highest value obtained. The percentage of students who carried out quality compressions (over 70% in relation to the QCPR) was 31% (45/112). Statistically significant differences by sex and degree were found; male students performed more quality compressions than women (69% vs. 30.1%; p < 0.001) (Odds Ratio (OR): 5.2; 95% CI: 2.1–12.9). The same happened with the Physiotherapy students in relation to the Nursing students (54.3% vs. 33.8%; p = 0.04) (OR = 2.3; 95% CI: 1–5.3).Table 3 describes the values obtained in the objective evaluation for each variable analysed in relation to the compressions and ventilations, stratified by sex and degree, respectively. No statistically significant differences in relation to the university campuses were found.Men performed better than women in overall quality CPR, percentage of correct decompressions, in ventilation and ventilations per minute. Physiotherapy students achieved better results in ventilation, mean volume of ventilations and ratio of ventilations per minute than nursing students.The statistical differences between the items which coincided in the objective and subjective assessment were analysed. Regarding the position of the hands and the compression/decompression, no statistical differences were found (p = 0.4 and p = 0.7, respectively).Information was gathered on 112 participants. The mean overall exertion was described as ‘hard’ (mean 5.4; SD = 1.7). No statistical differences by degree, campus and sex were found. However, Nursing students valued the exertion as ‘hard’ (mean of 5.7; SD = 1.6), while Physiotherapy students described it as ‘somewhat hard’ (mean of 4.5; SD = 1.7).This paper has analysed the effect of an educational action in CPR on Health Sciences students of a Spanish University. For the determination of the duration of the action, the results presented by Hernández-Padilla et al. and Aqel et al. were considered, in which they stated this timing as adequate [8,12]. As for the duration of the theoretical presentation, the times used by the authors of the work consulted were taken into account, 42 minutes being the shortest duration used by Qi Li et al. [17]. Eventually, the proposal made by the authors of this work, who state that 15 min is enough to acquire the necessary knowledge of CPR, was put into practice. Five minutes per student were estimated as suitable for the acquisition of practical skills, as other studies have pointed out [18,19].As for knowledge, these results suggest that a theoretical exposure of 15 minutes using a presentation with slides and the subsequent practice of skills was enough to dramatically increase CPR knowledge, obtaining outstanding mean rates as opposed to the mean rate obtained prior to the action. This timing reveals that, with a brief theoretical exposure, good results could be achieved, as opposed to the time used by other authors to date [2,3,10,17], while bearing in mind the recommendation of the ERC on not including theoretical lessons [20]. Regarding the relationship between the data obtained in the objective and subjective evaluations, no statistical differences for the position of the hands or the relationship between compression and decompression were found, which suggests that the subjective evaluation by staff trained in CPR is meant to be a reliable measuring instrument.Nursing students rated the exertion after two minutes of CPR as ‘hard’, while Physiotherapy students qualified it as ‘somewhat hard’, according to the Borg Scale of Perceived Exertion. Our results are higher than those found in professionals who are assumed to be in better physical condition [14,21]. Although no statistical differences were found, this could be due to the fact that Physiotherapy students showed a higher average weight, height and BMI in relation to Nursing students, so their perception of the exertion was lower.In terms of the quality of compressions, a mean of 85.4 compressions per minute was obtained, which is slightly below the 100–120 compressions per minute recommended by ERC [1]. The same happened with the mean depth participants reached in the compressions (44.9 mm), slightly less than the 50–60 mm the same sources propose. Only 35.6% of students reached the right depth, and 78.7% of the same performed a complete decompression of the chest. Given that the depth of the compression is a value that depends on factors such as weight or height and cannot be completely modified, special importance was placed on the insistence on correct decompression of the victim’s chest as anyone can properly perform it and is one of the CPR quality parameters.The ERC establishes the appropriate volume to be administered in the ventilation between 500 and 600 cm [1]. The mean volume delivered was 454.9 cm3, so they cannot be considered as quality ventilations if we take into account that only 15.5% of them were performed at an appropriate volume. Note that only 75% of the students opened the airway to ventilate the victim, which means that one out of four students did not open the airway for ventilation, precluding the entrance of air in the lungs despite nose clamping and insufflating the right air volume.Based on the results, we cannot determine that the percentage of overall CPR quality was appropriate, since 57.6% was obtained and experts establish a value higher than 70% as a quality CPR [22].The results suggest that there is a relationship between weight, height and BMI and some of the studied variables, i.e., compression depth, percentage of correct compression, percentage of global CPR quality, and percentage of correct rhythm during the compressions, obtaining better results with higher levels of weight, height and BMI of the participants, as in previous studies [4,23,24,25].The sex variable was stated as an influential factor in certain parameters of CPR. Men performed better in mean depth, mean percentage of compressions, percentage of overall CPR quality, ventilation and ventilations per minute. We believe that this is due to men’s greater weight, height and BMI than women, establishing these characteristics as determinants for the performance of quality CPR. The action was effective regarding training in the use of the AED, with 90% of the participants obtaining the necessary knowledge and skills. Based on the results, it would be advisable to include this type of training in all courses on BLS and extend this knowledge to the greatest number of people possible, since its use significantly increases the percentage of survival after a CRA and these devices are progressively being implemented in public places.From the data obtained we can infer that the educational action has been positive in terms of knowledge but has not been effective regarding the practical skills achieved in CPR, as the results are significantly below the values that clinical practice guidelines recommend. We understand that this may be caused by the limited time for the practical training of CPR, the large number of students for each session, or the training method used. As the authors of the work consulted show, through other methods such as self-directed learning, high-fidelity simulation, feedback after surgery or video-guided practice, a more effective CPR training can be offered to Health Sciences university students [8,12,17].The participants’ performance was evaluated through a simulation with a manikin. Therefore, the findings may not be applicable to a real cardiorespiratory arrest, during which fear, or other psychological factors may limit the performance.On the other hand, the study participants were all Health Sciences university students who attended the course on a voluntary basis, so a high motivation and interest in the subject are assumed; as a consequence, the results may not be representative of the general population.These limitations may be controlled in future studies through the development of Randomised Controlled Trials in various university centres with different training methodologies.Training in CPR based on fifteen minutes of theoretical exposure along with practice oriented to the acquisition of skills with feedback was enough to obtain good results in relation to the acquisition of knowledge of BLS, use of AED and skills acquisition in CPR. This study has found a relationship between sex, weight, height, and BMI and the completion of quality CPR, these being variables that allow quality parameters. A training method based on a very brief training in CPR through the feedback offered by the manikin and that given by monitors is not sufficient to achieve quality values in the completion of the action. The authors of the work have, as a proposal for future lines of research, to investigate the time interval after which it is necessary to carry out knowledge and skills retraining in cardiopulmonary resuscitation. In addition, it is appropriate to carry out a randomised clinical trial in different populations of students with different CPR training methods. Conceptualization, J.G.-S., C.M.-M., S.M.-I. and D.F.-G.; Data curation, C.M.-M., M.G.-S., M.A.D.L.P.-R. and D.F.-G.; Formal analysis, C.M.-M., M.G.-S. and M.A.D.L.P.-R.; Funding acquisition, C.M.-M.; Investigation, J.G.-S., S.M.-I., M.G.-S., M.A.D.L.P.-R. and D.F.-G.; Methodology, J.G.-S., C.M.-M., S.M.-I., M.G.-S., M.A.D.L.P.-R. and D.F.-G.; Project administration, J.G.-S. and D.F.-G.; Resources, C.M.-M., S.M.-I. and M.A.D.L.P.-R.; Software, C.M.-M., M.G.-S. and M.A.D.L.P.-R.; Supervision, J.G.-S., S.M.-I., M.A.D.L.P.-R. and D.F.-G.; Validation, J.G.-S., S.M.-I., M.A.D.L.P.-R. and D.F.-G.; Visualization, C.M.-M., S.M.-I. and M.G.-S.; Writing—original draft, J.G.-S., C.M.-M. and D.F.-G.; Writing – review & editing, J.G.-S., S.M.-I. and D.F.-G.This research received no external funding.The authors declare no conflict of interest.Knowledge improvement by degree and campus.Pre-test and post-test global marks.List of variables resulting from the questionnaire and the CPR.CPR: Cardiopulmonary Resuscitation. OHCA: Out-of-hospital cardiorespiratory arrest.BLS sequence subjective evaluation.AED: Automatic External Defibrillator. CI: Confidence interval.Variables distribution of the Skill Reporter organised by Sex and Degree.QCPR: Global quality of CPR; CCP: Continuous compression percentage; HP: Hands position; TNC: Total number of compressions; MD: Mean depth; CDcP: Correct decompression percentage; CDP: Correct depth percentage; CRC: Correct rhythm compressions; RM: Rhythm of compressions per minute; Ratio; Compressions-ventilations base 1 relation; Ventilations: Total number of ventilations in 2 min; MV: Mean volume applied. VEV: Ventilations exceeding the maximum volume; AVV: Appropriate volume ventilations; NAVV: Non-appropriate volume ventilations; MEANRATIO: Number of ventilations per minute.

Med-MDPI/ijerph_3/ijerph-16-05-00767.txt

@@ -0,0 +1,25 @@
+Informing users of waterworks systems about the quality of tap water is an obligatory trend. It should be accompanied by studies on the influence of the risk of threats on public health. Waterworks systems, being included in a critical infrastructure of the city, should be subject to special protection in this respect. In the paper, the authors’ method of assessing threats to people and property from waterworks systems functioning in self-government units (SGUs), is proposed. Four categories of factors affecting the risk of threat to tap water consumers were assumed: the frequency or the probability of exposure—P, financial losses—C, damages to peoples’ health—HL, the degree of the security—S. Based on this, a four-parametric risk matrix was developed. It was assumed that risk is a function of the parameters mentioned above: r = f(P, C, HL, S). For every parameter the five-parametric weight scale was assumed. An example of applying the method is presented. The proposed method should be an important element of water safety plans. It can also be adopted for other municipal systems subject to SGU. Water supply systems belong to the so-called critical infrastructure of the state in accordance with the European guidelines. Therefore, they should be subject to special protection. Council Directive of the European Union 98/83/EC on the quality of water intended for human consumption, with subsequent amendments in 2003, 2009 and 2015, states that a balance should be struck between protection against risks caused by both chemical and microbiological agents and that the parameter values for water intended for human consumption, in the light of their future assessments, should be set on the basis of public health considerations using risk assessment methods [1]. Article 6 states that appropriate measures will be taken to reduce or eliminate the risk of non-compliance with the quality parameters of water, and consumers must be duly informed about any corrective actions related to the reduction or elimination of the above-mentioned risk. Article 8 imposes an obligation to analyse the risk to human health if water does not meet the quality parameters. Currently, work is underway on a new directive which introduces an obligation for entities responsible for supplying water to develop and implement the so-called water safety plans based on analyses and risk assessments [2]. Also the European Union recommend member countries to implement standards regarding drinking water supply security concerning guidelines for risk and crisis management through standards EN 15975-1 and -2 Part 1: Crisis management and Part 2: Risk management [3,4].The Commission pays particular attention to the following priorities in the water supply management process:application of a risk-based approach in the design, construction and operation of water supply systems,increasing the transparency of procedures and providing consumers with access to up-to-date information on water quality and potential threats,minimizing potential damages.application of a risk-based approach in the design, construction and operation of water supply systems,increasing the transparency of procedures and providing consumers with access to up-to-date information on water quality and potential threats,minimizing potential damages.According to regulations regarding the quality of water intended for human consumption, the competent, district or border sanitary inspectors issue periodic assessments of water quality. The authorities of the State Sanitary Inspectorate perform water quality and health risk assessment in the given area. The assessment includes a list of water producers supplying the population and providing water from individual intakes, as part of economic activity, to public buildings and residential buildings. In addition, information is provided about:the volume of water production supplied by individual producers and the method of water treatment,the number of people supplied with water,water quality, the way it is treated and disinfected, if this is used,exceeding the limit values of water quality parameters with an indication of their impact on the health of consumers,reported adverse reactions related to water consumption in the given area,administrative proceedings in the field of water quality,repair activities carried out by water and sewage companies.the volume of water production supplied by individual producers and the method of water treatment,the number of people supplied with water,water quality, the way it is treated and disinfected, if this is used,exceeding the limit values of water quality parameters with an indication of their impact on the health of consumers,reported adverse reactions related to water consumption in the given area,administrative proceedings in the field of water quality,repair activities carried out by water and sewage companies.Taking the right decisions for water treatment process or warning people of poor water quality always requires a certain time advance. A very important issue related to ensuring the safety of water supply operation includes the measurement of pressure and flow rate in the water supply network, as well as monitoring of water quality in treated water tanks and in selected points of the water supply network [5,6,7]. The sooner the system operator receives information about a threat, the greater the possibility to make the right decisions, as to restore the reliable water supply to recipients [8,9]. It should be noted that proper operation of a water network should include the following steps: conducting inspections of fittings of the water supply network through the maintenance or replacement, as well as current repairs of the water supply network and its renovation, reconstruction and replacement [10,11,12,13,14,15].The paper presents proposals for a risk analysis and assessment method based on a four-parameter matrix, including the assessment of risks for people and property from water supply systems operating in self-government units (SGU). The method should be applied in analysing risk of lack of water supply in crisis situations, taking into account the specificity of the local collective water supply system (CWSS).The risk associated with the operation of collective water supply systems means the possibility of an undesirable event having an impact on the achievement of the objective, which is the safe supply of water for consumption. According to the international ISO standard [15,16] and normative guidelines, the risk assessment consists of its identification, analysis, assessment and evaluation. People and Property Hazard Analysis (PPHA) method that can be used to estimate losses requires determining the upper limits of tolerable and controlled risk [17]. The goal of risk management is to bring its size to the level at least tolerable and, preferably, to the level of As Low As Reasonably Practicable (ALARP—which means “as low as reasonably feasible”) [18]. The novelty of the proposed method is the separation of material and human losses [17].In crisis management a proper risk assessment is the basis for taking actions to ensure safety efficiently and effectively. Efficient actions should be understood as full achievement of the set goals. Effective means achieving certain results. Risk analysis methods are being developed mainly to meet the needs of safety engineering [19,20,21,22,23,24,25,26,27,28]. This, in turn, implies the use of risk assessments in crisis management. The Act on Crisis Management obliges the assessment of risk in relation to, at least, human losses (fatalities, missing people, injured people requiring hospitalization and qualified medical help) and property losses [29,30]. The classic definition of risk shows that its assessment consists in multiplying the likelihood or frequency of occurrence of a threat by the projected losses [31,32,33]. Risk assessment requires determining the value of both factors. The joint consideration of human damages and material losses raises ethical concerns [34,35]. For this reason, we should categorize the risk associated with material losses and with human damage separately. People and Property Hazard Analysis (PPHA) method assumes the adoption of five-point scales for individual risk factors–very little, little, medium, large, very large) [17,36,37]. The ranges of values in the distinguished categories were derived from literature data from EU countries and expert experience of the authors of the article in the implementation of the WSP in water supply systems. The considered method is a practical implementation in the operation of small water supply systems. The risk analysis of water supply systems in big urban agglomerations should include uncertainty analysis, which must be strictly taken into account [38,39,40]. The infrastructure risk is defined as the product of the probability of system failure and associated costs of returning the system to service [41]. As to perform effective risk assessment and management, the analyst must understand the system and its interactions with its environment [42,43], and this understanding is requisite to modeling the behaviour of the state of the system under varied probabilistic conditions [44].The estimation of probability of threat can be made based on the modified Bernoulli distribution [17,19]. The classic Bernoulli formula for the probability of obtaining k of successes in n samples is calculated by the formula: (1)P(k)=(nk)pk·qn−k,where P is the probability of success, q = 1 − p is the probability of failure, k is the number of successes, n is the number of tries.Assuming that:P(A) = 1 − P(A1),(2)where P(A) is the probability of occurrence of a given threat, and P(A1) is the probability of no occurrence of a threat.In turn:(3)P(A1)=(nk)pk·qn−k,and for the event A1, n = k and p = 1 − q, then the formula (2) takes the form [17]:(4)P(A)=1−[(nk)pk·qn−k]=1−[1·(1−q)n·qo]=1−(1−q)n,where q is the frequency of occurrence of threat A, it is the value obtained from experience and can be identified with a posteriori probability.When determining the time perspective for which the probability of threat is calculated, not only the assumed time period of prospective N analysis should be taken into account, but also the time that has elapsed since the last year when threat occurred. Thus [17]:(5)n=N+(n1−n2),where n1 is the year in which the analysis is carried out, n2 is the year in which the last threat occurred, and N is time period of prospective analysis.Static probability is identified by a statistical estimate having a constant value for a given observation period, e.g., static probability determining the occurrence of risk once every 10 years equals P’ = 0.1. In contrast, dynamic probability is variable over time and dependent on the initial value, which is the static probability. Table 1 presents the scale of categories of frequency and static and dynamic probability of the occurrence of an undesirable event with the point weights for individual categories [17].Estimating material losses being a result of an undesirable event is a complex and multifaceted task [17]. The valuation of assets of people, enterprises, real estate, etc. is subject to current conjuncture. In the work as the measure of losses the interest on the SGU income was assumed.Table 2 presents the scale of the categories of material losses with the point weights for particular categories [17].The indicator used in analyses and assessments of accidents at work was adopted. The accident frequency rate indicates the number of undesirable events per 1000 people employed in a given sector of the economy. By analogy, the number of undesirable events per 1000 users of the public waterworks, was adopted. Three types of human losses were distinguished [17]:providing qualified medical assistance—HLpma,required hospitalization—HLrh,deadly descent—HLdd.providing qualified medical assistance—HLpma,required hospitalization—HLrh,deadly descent—HLdd.Table 3 presents the scale of the category of nuisances of human losses with the point weights for particular categories [17]. The loss category is always taken according to the largest value of estimated losses.The issue of estimating the risk associated with the estimation of the risk of human life threat is very sensitive and difficult in social perception, therefore we have presented, in addition to deadly descents, category of human losses as necessity of required hospitalization and providing qualified medical assistance. We are aware that comparing human and material losses is unethical, but the descriptions of spectacular water supply failures confirm that these types of losses are smaller. We have adopted the interpretation that the occurrence of losses related to death accidents should be considered as an unacceptable risk, independent of the final result obtained, and the transfer of the received risk to the unacceptable risk interval. It should be noted that the risk assessment requires each time performing a risk analysis.In the method the risk is determined according to the formula [23,24]:(6)r=P·C·HLS,where P is the probability of a threat, C are material losses, HL are human losses, and S is security category. Table 4 presents a four-parameter risk matrix, which, based on formula (6), allows one to estimate the amount of risk.Weights for parameters P, C and F are taken according to the Table 1, Table 2 and Table 3. For the parameter S, depending on the degree of water supply security, the following questionnaire can be used to estimate the category and point weight [23,24]. The proposed survey to pre-estimate the level of security of the CWSS [23,24]:
+		How often is the raw water quality monitored?
+			-every day—1 point,-periodically (once a month, once a quarter)—5 points,-randomly, if a threat is found—10 points,How often is the treated water monitored?
+			-every day—1 point,-periodically (once a week, once a month)—5 points,-randomly, if a threat is found—10 points,Does CWSS have a protection and warning station if it takes surface water?
+			 -yes—1 point,-no—3 points,Are the design requirements for the water intake protection zones implemented?
+       -in total—1 point,-with some exceptions—3 points,-there are difficulties, e.g., economic, legal, etc.—6 points,Is it possible to provide alternative water supply (emergency wells, two or more water supply sources)?
+			-yes—1 point,-partial—4 points,-no—10 points,Does the water supply company:
+			-have own specialized service for removing network failures—1 point,-have a contract with an economic entity that intervenes if necessary— 3 points,-search for a contractor to remove failures— 10 points,The emergency volume of treated water in water tanks is:
+			-0 ÷ 10% Qdmax—6 points,-10 ÷ 50% Qdmax—3 points,-over 50% Qmaxd—1 point,If the sum of points from the questionnaire is:
+			 -7 ÷ 10—high level of security—w = 3,-12 ÷ 34—medium security level—w = 2,-above 34—low level of security—w = 1,How often is the raw water quality monitored?
+			-every day—1 point,-periodically (once a month, once a quarter)—5 points,-randomly, if a threat is found—10 points,every day—1 point,periodically (once a month, once a quarter)—5 points,randomly, if a threat is found—10 points,How often is the treated water monitored?
+			-every day—1 point,-periodically (once a week, once a month)—5 points,-randomly, if a threat is found—10 points,every day—1 point,periodically (once a week, once a month)—5 points,randomly, if a threat is found—10 points,Does CWSS have a protection and warning station if it takes surface water?
+			 -yes—1 point,-no—3 points,yes—1 point,no—3 points,Are the design requirements for the water intake protection zones implemented?
+       -in total—1 point,-with some exceptions—3 points,-there are difficulties, e.g., economic, legal, etc.—6 points,in total—1 point,with some exceptions—3 points,there are difficulties, e.g., economic, legal, etc.—6 points,Is it possible to provide alternative water supply (emergency wells, two or more water supply sources)?
+			-yes—1 point,-partial—4 points,-no—10 points,yes—1 point,partial—4 points,no—10 points,Does the water supply company:
+			-have own specialized service for removing network failures—1 point,-have a contract with an economic entity that intervenes if necessary— 3 points,-search for a contractor to remove failures— 10 points,have own specialized service for removing network failures—1 point,have a contract with an economic entity that intervenes if necessary— 3 points,search for a contractor to remove failures— 10 points,The emergency volume of treated water in water tanks is:
+			-0 ÷ 10% Qdmax—6 points,-10 ÷ 50% Qdmax—3 points,-over 50% Qmaxd—1 point,0 ÷ 10% Qdmax—6 points,10 ÷ 50% Qdmax—3 points,over 50% Qmaxd—1 point,If the sum of points from the questionnaire is:
+			 -7 ÷ 10—high level of security—w = 3,-12 ÷ 34—medium security level—w = 2,-above 34—low level of security—w = 1,7 ÷ 10—high level of security—w = 3,12 ÷ 34—medium security level—w = 2,above 34—low level of security—w = 1,In the proposed method, due to the objectives of the SGU, taking into account the formula (6), three risk levels are distinguished. Determined levels of risk were result of a decision-making process, as to prioritize what action should be performed.
+        -tolerated risk—from 0.33 to 5,-controlled risk—from 5.33 to 15,-unacceptable risk—from 16 to 125.tolerated risk—from 0.33 to 5,controlled risk—from 5.33 to 15,unacceptable risk—from 16 to 125.The risk value criteria were assumed on the basis of literature data from EU countries and the authors’ experience in the implementation of the WSP in water supply systems. The three point scale is widely used in different technical systems. In the case of uncertain data fuzzy set theory may be used, which is presented in the work [38,39]. The obtained unacceptable risk means, that an immediate action should be taken to reduce this value of risk category. In case of tolerable risk no extra actions are required. Controlled risk means the intermediate level, which means that the system is permitted to function but under the condition that modernization or renovation will start, according to concept of As Low As Reasonably Practicable (ALARP) proposed by Health and Safety Executive.In 2015, the probability of occurrence of a threat of contamination of the source of surface water intake in the water intake, was estimated. During 30 years (since 1985), contamination occurred 5 times and the last time in 2014. What is the probability of contamination from a perspective of 5 years?
+        (7)n=5+(2015−2014)=6,
+        (8)q=530=0.1667,
+        (9)P(A)=1−(1−q)n=1−(1−0.1667)6=1−0.3341=0.6659,The probability of water contamination in 2017 was re-estimated, also with the perspective of the next 5 years. The last threat occurred unchanged, in 2014:(10)n=5+(2017−2014)=9,
+        (11)q=532=0.15625,
+        (12)P(A)=1−(1−0.15625)9=1−0.2167=0.7833,The probability of water contamination in 2015–2020 was 0.6659 and in 2017–2022 it increases to 0.7833. According to Table 1, this is category large with weight w = 4 (dynamic probability).As a result of the failure of the water supply network and the lack of power supply in the water pumping station, the average annual losses were estimated on the level of 0.45% of the SGU annual expenditures. For such a value, according to Table 2, the value of estimated material losses is in the category very little with the weight w = 1.The commune uses a group waterworks LM = 5000 people. For an undesirable event related to the secondary water pollution in the water supply network, it was estimated Fpl = 50, Fh = 5 and Fzs = 0.
+		the number of people who should be given qualified medical help is:(13)LM·Fpl1000=5000·501000=250 people,the number of people hospitalized is:(14)LM·Fh1000=5000·51000=25 people,the number of people who should be given qualified medical help is:(13)LM·Fpl1000=5000·501000=250 people,the number of people hospitalized is:(14)LM·Fh1000=5000·51000=25 people,For such values according to Table 3, the value of estimated human losses corresponds to category large with the weight w = 4.Using the data contained in examples 1-3 weights for the individual risk parameters are:P − w = 4,C − w = 1,HL − w = 4,S was estimated on the basis of the questionnaire for the security parameter, the sum of 21 points means w = 2.P − w = 4,C − w = 1,HL − w = 4,S was estimated on the basis of the questionnaire for the security parameter, the sum of 21 points means w = 2.Based on Equation (6), the estimated risk is r = 8. The analysed risk related to the risk of contamination of tap water within the analysed SGU is at the controlled level.Currently, a water supply company is obliged to inform about the deterioration of water quality a competent state sanitary inspector and a community head, mayor or city president, within no more than seven working days. The scope of information included in the application for consent to the derogation was expanded in the field of:reasons why water of the required quality cannot be delivered,justification along with an indication of actions aiming to ensure the right quality water,a study analysis prepared by a scientific institution dealing with public health, regarding the impact of the derogation (concentration and duration) on the health of tap water consumers.reasons why water of the required quality cannot be delivered,justification along with an indication of actions aiming to ensure the right quality water,a study analysis prepared by a scientific institution dealing with public health, regarding the impact of the derogation (concentration and duration) on the health of tap water consumers.In addition, an obligation to provide a systematic (every 3 months) detailed report on corrective actions taken and actions planned to be taken in the next reporting period, was introduced. Now, information for consumers about water quality also includes data on granted consents to deviation from the acceptable parameters:Standard information for residents about water quality should include:the area of the commune covered by water research,the area of the commune not covered by research with an indication of the reasons,threats resulting from the lack of water quality tests,defining activities that should be taken to protect health against contaminated water.the area of the commune covered by water research,the area of the commune not covered by research with an indication of the reasons,threats resulting from the lack of water quality tests,defining activities that should be taken to protect health against contaminated water.According to the authors, the method can be used in SGUs of up to 50,000 residents. For the probability of occurrence of an undesirable event: poor water quality or lack of supply (P), material losses (C), population health losses (HL), 5-stage scales were used and for the degree of security of the water system (S) a 3-stage scale was used. The adoption of a 3-stage scale in relation to security (S) results from the fact that for small and medium-sized CWSS such systems are not very expanded. To characterize the degree of security (S) we use the questionnaire with 7 questions.Water supply systems belonging to the critical infrastructure require a detailed risk analysis in view of the possibility of a crisis situation, taking into account rapid response plans to ensure the delivery of water to consumers.The method of risk analysis and assessment presented in the work can be used as part of the risk assessment for the needs of water safety plans. Its advantage is the possibility of adapting to the specifics of local CWSS. The method presents a detailed way of assessing the possibility of threats as well as a method of loss assessment.People and Property Hazard Analysis (PPHA) method is a kind of development of the Preliminary Hazard Analysis (PHA) method. It allows for the inclusion of human and material losses separately in the adopted planning perspective by the category of determining the probability of an undesirable event occurrence.The method also takes into account the degree of security of the water supply system against threats, including aspects of the multibarrier system. In this aspect the use of the authors’ questionnaire to estimate the security parameter is proposed.The method of analysis and risk assessment proposed in the work is based on the authors’ four-parameter risk matrix.J.R.R., B.T.-C. and K.P.-U. equally contributed to the development of this manuscript. This research was funded by Faculty of Civil and Environmental Engineering and Architecture, Rzeszow University of Technology, 35-959 Rzeszow, Poland.We thank the reviewers for their feedback, what helped to improve the manuscript quality.The authors declare no conflict of interest.Categories of frequency and probability of threat occurrence.Category of material losses.Category of human losses.* Notes: the occurrence of losses related to death accidents should be considered as an unacceptable risk, independent of the final result obtained, and the transfer of the received risk to the unacceptable risk interval.The four-parameter matrix.Notes:  tolerated risk,  controlled risk,  unacceptable risk. Ex. Obtained value r = 0.33, was calculated on the base of categories and values: P = 1 (frequency up to once every 50 years), C = 1 (material losses up to up to 0.5% of annual budget expenditure), HL = 1 (providing qualified medical assistance HLpma ≤ 5, no required hospitalization and none deadly descent), and S = 3 (high level of security estimated on the basis of the questionnaire).

Med-MDPI/ijerph_3/ijerph-16-05-00768.txt

@@ -0,0 +1 @@
+The acquisition of competencies in basic life support (BLS) among university students of health sciences requires specific and updated training; therefore, the aim of this review was to identify, evaluate, and synthesise the available scientific knowledge on the effect of training in cardiorespiratory resuscitation in this population. A comprehensive literature search was conducted in MEDLINE, CUIDEN, Web of Science, Wiley Online Library, CINAHL, and Cochrane, including all randomised clinical trials published in the last ten years that evaluated basic life support training methods among these students. We selected a total of 11 randomissed clinical trials that met the inclusion criteria. Participants were nursing and medicine students who received theoretical and practical training in basic life support. The studies showed a great heterogeneity in training methods and evaluators, as did the feedback devices used in the practical evaluations and in the measurement of quality of cardiorespiratory resuscitation. In spite of the variety of information resulting from the training methods in basic life support, we conclude that mannequins with voice-guided feedback proved to be more effective than the other resources analysed for learning.Cardiorespiratory arrest (CRA) has become a major public health problem and one of the leading causes of death in the Western world in recent years. Cardiopulmonary resuscitation (CPR) is the technique used in the cases of CRA. It consists of thoracic compressions (which are important for the perfusion of vital organs) and rescue breaths by means of artificial ventilation [1,2,3,4,5,6,7]. The quality of CPR is vitally important, and it depends on the level of knowledge and skills held by those who carry out the CPR. Even among healthcare professionals, that level can be inadequate. Therefore, an improvement in educating healthcare professionals in CPR techniques may increase survival rates in cases of CRA [2,8,9].Within a hospital, the nursing staff is usually the first group of professionals to identify CPR, so competence in basic life support (BLS) is a key factor in recognising cardiac arrest, activating emergency systems, initiating effective CPR, and safely using the defibrillator [10,11,12,13]. Roh and Issenberg concluded that technical skills in CPR among nursing students are very poor, and that despite efforts to improve the quality of psychomotor skills, the results obtained are still not encouraging [5].As has previously been described, BLS is a fundamental therapy for saving lives, and it requires a broad knowledge of cognitive and psychomotor skills. [13,14] In spite of this, several studies have shown that BLS education is difficult: learners’ retention of motor skills is poor (even immediately after they have completed the course), causing less-than-ideal performance of CPR [6,14,15]. In addition, if those who have been trained in CPR do not frequently perform it, their skills deteriorate over a period of between 3 and 6 months. Therefore, it is very important that in addition to developing different learning strategies, these should be combined with other recycling (retraining) measures during that period of time [10,16].Within CPR teaching, different methods have been proposed, such as simulation, classical instructor-led teaching, and self-directed mannequins with continuous verbal feedback, which have been shown to be much more effective for retaining knowledge and motor skills [9,12,17,18]. Other methods of learning may be based on interactive videos, high-fidelity 3D simulation scenarios, and partner-based training, in which very positive results have been obtained [1,19].With all this, there is a need for a systematic review that includes a comparison in the methods used (traditional versus alternative) trying to find the most effective for the teaching of BLS, CPR, and use of automatic external defibrillators (AEDs) in university health science students.Finally, the research question selected by the authors was what is the most effective method for teaching of BLS, CPR techniques, and use of AED for health science students?So that, the main objective of this systematic review was to identify, evaluate and synthesize what kind of method is more effective of training in basic life support, cardiopulmonary resuscitation techniques and use of automatic external defibrillator among health science students.We undertook a systematic review in line with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement [20].The search for articles was conducted during February and March 2017. The scientific databases searched were MEDLINE, CUIDEN, Web of Science, Wiley Online Library, CINAHL, and Cochrane.We used both English and Spanish descriptors that were located in the Medical Subject Headings (MeSH) and in Descriptores de Ciencias de la Salud (Health Sciences Descriptors; DeCS). These included “health science”, “students”, “cardiopulmonary resuscitation”, “training”, “traditional”, “new methods”, “motor skills���, “simulation”, and “evaluation of efficacy-effectiveness”. Descriptors that were synonymous with one another were combined in the search with the Boolean “OR” operator, while the “AND” operator was used to interrelate different concepts.As an example, one of the search strategies used in the Medline database was: “health science students” AND (“traditional cardiopulmonary resuscitation training” OR “new methods cardiopulmonary resuscitation training”).The studies that were selected for the systematic review met the following inclusion criteria.Year of publication: we included all articles published between 2007 and 2017, in order to obtain the most recent articles on training methods.Language: Spanish and/or English.Studies: we included full texts of randomised clinical trials (RCTs), because these epidemiological studies provide more evidence.Population: students of both sexes who were pursuing university degrees related to the health sciences.Intervention: any method used in the teaching of BLS and the acquisition of technical skills in CPR in adults.Results: we selected studies that contained information about the socio-demographic characteristics of participants, ones that analysed the effect of training in the acquisition of theoretical and practical knowledge, and ones that reported on measurement tools for skills relating to placement of the hands, number of compressions, average depth of compressions, number of ventilations, or volumes administered.Year of publication: we included all articles published between 2007 and 2017, in order to obtain the most recent articles on training methods.Language: Spanish and/or English.Studies: we included full texts of randomised clinical trials (RCTs), because these epidemiological studies provide more evidence.Population: students of both sexes who were pursuing university degrees related to the health sciences.Intervention: any method used in the teaching of BLS and the acquisition of technical skills in CPR in adults.Results: we selected studies that contained information about the socio-demographic characteristics of participants, ones that analysed the effect of training in the acquisition of theoretical and practical knowledge, and ones that reported on measurement tools for skills relating to placement of the hands, number of compressions, average depth of compressions, number of ventilations, or volumes administered.All articles that did not meet these criteria were excluded.Initially, two reviewers independently performed the article search in order to minimise selection bias. After deleting duplicates, an initial selection of articles was carried out following independent analysis of the titles.We then conducted a second review that included the reading of titles, abstracts and key words of the articles found by the two reviewers, who jointly proceeded to make the final selection of articles. At this point, a third independent reviewer intervened in the decision-making process in cases of disagreement. Finally, after obtaining the full-text articles, the third and final selection took place, in which the articles that were eventually used in the review were chosen.Because the studies reviewed were very heterogeneous and had different methods of intervention and assessment, it was not possible to undertake a meta-analysis.Last of all, and following the final selection of the articles included in the review, we extracted the following information from each article: first author, publication year, population, study groups, learning method, evaluation method, immediate results, and results after refreshers/recycling.The methodological quality of the randomised clinical trials included in the review was evaluated using the Jadad scale [21].The validity of this scale has been proven in the scientific literature, and it is simple and quick to use. In addition, the researchers were already trained in its use, having deployed it in other studies.The scale gives a score between 0 and 5 points, primarily to three aspects: randomisation, blinding (double-blind), and description of withdrawals and dropouts during follow-up. A score of 5 represents the highest possible methodological quality, while a score of under 3 means that the evaluated clinical trial is of a low methodological quality. Below, we provide the full scale with the items and their corresponding scores (Table 1).Through our search, we obtained a total of 522 articles that were potentially eligible for the review. Of them, 371 were eliminated on the basis that they were duplicates from across the different databases.After completing the first selection (reading of titles), 109 articles were excluded. We then analysed the abstracts of the 42 articles that were still potentially valid for inclusion, through which a total of 18 were excluded. Finally, and after obtaining the remaining 24 articles in full-text form, a total of 11 studies were included in the review [6,9,11,12,13,16,17,18,19,22,23].The excluded articles were those which did not meet the inclusion criteria for the study, which are shown in a flow diagram in Figure 1.Table 2 presents the final articles that were part of the systematic review based on their methodological quality.The scores obtained on the Jadad scale for the analysed articles ranged from 2 to 5 points, with an average of 2.81 points. Only one article with double blinding [22] obtained the maximum score, and four articles [9,12,19,23] (which were transversal and involved only one measure) scored 2 points. All the studies, with the exception of the one conducted by Isbye et al. [22], presented a high risk of bias, as they involved single blinding, making it impossible for there to be double-blinding for participants and researchers.All the data extracted from each article, with general and specific characteristics, are summarised in Table 3.The participants of the different studies were university students from different branches of the health sciences, including mainly nursing students [11,12,13,17,18,19] and medicine students [22,23]. Only one study did not distinguish the students’ degree titles [6].The analysed studies included a total of 2175 participants; that by Partiprajak et al. (n = 30) [13] had the fewest participants, while that by Oermann et al. (n = 606) [19] had the most.Information about previous knowledge of and technical skills in BLS and CPR was collected in seven articles [9,11,12,13,16,18,22]. This meant that some authors excluded a certain number of participants from studies [9,16], or conversely used this information as a basis when establishing prior training [11,12,13,18,22].The different teaching methods primarily related to two aspects: theoretical content aspects, and aspects derived from the acquisition of technical skills in CPR. To this end, the researchers were guided by the recommendations of the American Heart Association (AHA) [9,12,13,17,18,23] and of the European Resuscitation Council (ERC) [6,11,19,22]. The exception in this regard was the study by Mpotos et al. [16], which does not mention any recommendations. These guides were also employed later to carry out measurements in the evaluations.For the theoretical level of the training, the authors employed different techniques according to the different groups that they created within their studies. Accordingly, instructors taught lectures using visual media such as presentations or videos [6,9,11,12,17,23]. In other cases, the participants acquired knowledge independently through computer CDs or DVDs provided by the researchers [9,13] or through different tests with self-assessments [16,19]. In the rest of the studies, training methods were not specified [18,22].As for the teaching of skills, the most used method was traditional instructor-led teaching, which appeared in a total of 6 studies [9,11,12,17,22,23]. In the study by Li et al. [23], one of the group’s first underwent a pre-assessment (a practical scenario), which after being recorded and reviewed by the instructors, was subsequently used for training purposes as feedback.The second-most-used method was training with mannequins that had feedback systems (these are known as mannequins with a skill reporter). These featured in five studies [6,11,13,16,19]. In addition, in four studies the participants from some groups carried out self-directed learning with mannequins that, in addition to feedback, had voice prompts that corrected errors (the so-called voice advisory mannequin, VAM) [9,12,18,22]. It is also worth mentioning that in the studies by Aqel et al. [17] and Boada et al. [19] high-fidelity simulation programs were used to deliver the training.Finally, in two studies [6,9], the control groups did skills training without any kind of feedback or supervision from instructors, and in two others [16,18] no skills practice of any kind was performed.In terms of the duration of the training undertaken to acquire knowledge and skills, the most homogeneous approach was the traditional teaching method’s time frame of between four and five hours, except in the case of the study carried out by Spooner et al. [6], which took place over 8 h. There were large variations in the other studies.Evaluation methods were organised according to knowledge and skills. To measure knowledge levels, the methods used were pre-intervention [23], post-intervention [9,19] or pre- and post-intervention [11,13,17] questionnaires, or a subjective assessment by instructors of performance in the sequence of BLS steps [6,17]. In addition, two studies included a questionnaire about the confidence that participants had in executing the skills after the training [11,13]. The measurement of CPR-technique skills during the period in which they were being acquired was taken in 10 of the 11 studies through a skill reporter mannequin [6,9,11,12,13,16,18,19,22,23].The different results obtained after analysing the articles were divided into two groups for drafting purposes. We will first discuss the results of the studies in which a single measure was used following completion of the intervention, and we will then consider the other studies, in which more evaluations were carried out over time.Studies that involved a measurement that evaluated theoretical knowledge reported an improvement in all groups [19,23], with the exception of the study by Roppolo et al. [9], in which the control group that received theoretical training with an instructor obtained better results. In the study by Kardong Edgren et al., knowledge was not measured [12].As for the acquisition of technical CPR skills, the groups that acquired knowledge through a VAM obtained better results with statistically significant differences relative to the rest of the groups [9,12]. In the study that used high-fidelity simulation, no differences between the groups were established [19]. Finally, in the study by Li et al., in which a group carried out a practical scenario with feedback provided by instructors to participants, significant results were obtained in CPR technique for all aspects except for the placement of hands (both groups obtained 100%) [23].On the other hand, in the first assessment of the studies that applied several measurements [11,13], knowledge improved in all groups, with the exception of the study by Aqel et al. [17], in which the improvement was additionally statistically significant in the intervention group. In the study by Spooner et al. [6], the correct completion of the BLS algorithm was evaluated, with no differences between the groups. In addition, the studies by Hernández Padilla et al. [11] and by Partiprajak et al. [13] used questionnaires to analyse the confidence of participants in performing the CPR technique safely, and they noted an improvement in results after the intervention had been completed.With respect to skills in performing CPR, in three studies [11,13,16], no differences were found between the different groups, while in the studies by Spooner et al. and Aqel et al. [6,17], the intervention groups performed better in both studies. Finally, in the study by Isbye et al. [22], the instructor-led group obtained better results than the self-directed groups that used a VAM.Seven studies carried out a subsequent measurement after a refresher or simply by applying a knowledge retention period [6,11,13,16,17,18,22] over periods of time ranging from 6 weeks [6] to 1 year [18] after the intervention.In six studies, a measurement for knowledge retention was used [6,11,13,16,17,22]. In the study by Spooner et al., after 6 weeks the group that had undertaken practice obtained better results, while there were no differences when it came to correctly applying the BLS algorithm. [6] In the study by Hernández Padilla et al., the results were better for the self-directed group at three months. [11] In the study by Partiprajak et al., after three months, worse results were obtained in terms of knowledge, similar results were found in terms of confidence of participants when performing CPR, and better results were obtained in terms of acquisition of CPR skills [13]. The study by Mpotos et al. observed that the control group that had not received practical skills training obtained better results than the intervention group at 6 months [16]. In the fifth study in which retention of knowledge was evaluated, carried out by Aqel et al. [17], it was observed that after 3 months the improvement in results for knowledge and skills in the group that received the high-fidelity simulation remained. Finally, the study by Isbye et al. concluded that there were no differences between the groups after 3 months [22].In the study by Oermann et al. [18] there was no measurement of skills immediately after the intervention. Out of their control and intervention groups, they produced random subgroups at 3, 6, 9, and 12 months, which were the ones evaluated. Among these groups, no differences were established in terms of the number of compressions, volume: minute, or hand placements, but there were in relation to depth and volume administered, which decreased significantly as the measurements were taken over time. Finally, a fifth subgroup that was given a refresher was also established at 12 months, and statistically significant differences between the control and intervention groups were not obtained within it.To conduct this review, we drew on a total of 11 randomised clinical trials that were found in different databases and that aimed to assess the quality of training in CPR and BLS knowledge and technical skills among health sciences students. Most of the studies were conducted among nursing and medicine students, in line with the study by López Messa et al., which highlights that BLS training for future healthcare professionals should be reinforced at the undergraduate level, especially in nursing and medicine degrees [24].The studies included in the review were of a low methodological quality according to the Jadad scale. In view of these findings, one priority that emerges is the need to increase the number of RCTs with methodological rigour, which would make it possible to minimise biases and facilitate the identification of progress in scientific evidence regarding BLS training among health sciences students. To this end, the use of this same scale in other reviews or similar studies would facilitate this process.The articles are also characterised by the absence of homogeneity in establishing BLS training, technical CPR skills and use of AEDs. Despite this, the results have shown how studies that used VAM [9,11,12,18] improved all skills immediately or in the long term, though in the study by Isbye et al. [22] ventilations were not improved at first.Moreover, the realisation of practical cases through different simulation programs of high fidelity provided better results fundamentally in the acquisition of theoretical knowledge [17,19].Weidman et al. define learning through simulation as an essential part of training, whether it is high or low fidelity [25]. High-fidelity simulation is very useful when comparing the results obtained with real outcomes, even though it requires thorough intervention from the instructors [26,27]. In addition, this training provides realistic environments and is more student focused [28].The use of a skill reporter or VAM mannequins with feedback results in a remarkable rise in the improvement of the quality of CPR performed by nursing and medicine students, since it allows them to correct their mistakes or undertake knowledge refreshers independently, making it feasible to not have an instructor on an ongoing basis. Along this line, the study by Nielsen et al. concludes that this type of learning improves knowledge and skills [29].Finally, in the studies included in the review, the use of AED is scarcely mentioned. Although eight studies included AEDs as part of the theoretical and practical training [6,9,11,13,17,19,22,23], only Roppolo et al. [9] implemented a measure concerning the use of this device. They obtained unfavourable results that do not coincide with those of the study by Ahn et al. [30], the main finding of which was that students reduced intervention times as soon as they had an AED nearby. It has been shown that courses of between 2 and 4 h in the use of an AED may be enough to operate them safely [31].Therefore, and despite the fact that the use of AEDs is a priority when it comes to saving lives, there is a need for more studies that more comprehensively evaluate training in and handling and application of these devices in order for there to be fuller performance within BLS.One of the main limitations of the study is that in spite of BLS and CPR training for health science students, the number of randomised clinical trials is not very high, and studies that have appeared are very heterogeneous in terms of how they have been produced. Moreover, after reviewing the studies on a methodological level, we observed that it is necessary to increase their methodological rigour.Another of the limitations of the study is the fact that the recommendations issued by the AHA and ERC for BLS training evolve continuously, meaning that the inclusion of studies published over the last 10 years makes it very difficult to assess them in the same way.In relation to the use of AEDs, it was not possible to describe them because most of the selected studies did not include measurement results.Finally, researchers have not included students taking different degrees in their studies, so it has been impossible to establish differences between students, their degrees, and different training methods that it may have been possible to use.The studies included in this systematic review are characterised by a low methodological quality and heterogeneity in terms of their interventions.Findings have shown that the use of VAMs was more effective for learning CPR skills than the other resources analysed. With regard to the knowledge acquired, participants did not show differences between those who received a theoretical session with an instructor and participants who acquired knowledge independently through computer CDs or DVDs.Studies did not show results of the use of AEDs, so a comparison could not be made. Therefore, we would recommend future researchers to include in their research the use of AED, since we consider it necessary to increase information regarding its use and how students can face its use in a real case. Finally, we would recommend that future research have a high methodological quality so that studies can have greater relevance.Conceptualization, M.G.-S., C.M.-M., S.M.-I., and D.F.-G.; Data curation, M.G.-S., C.M.-M., S.M.-I., J.G.-S., and D.F.-G.; Formal analysis, M.G.-S., C.M.-M., S.M.-I., J.G.-S., and D.F.-G.; Investigation, M.G.-S., C.M.-M., S.M.-I., J.G.-S., and D.F.-G.; Methodology, M.G.-S., C.M.-M., S.M.-I., J.G.-S., and D.F.-G.; Project administration, M.G.-S.; Resources, M.G.-S.; Supervision, M.G.-S., and J.G.-S.; Validation, M.G.-S., C.M.-M., J.G.-S., and D.F.-G.; Visualization, M.G.-S., C.M.-M., S.M.-I., J.G.-S., and D.F.-G.; Writing—original draft, M.G.-S., C.M.-M., S.M.-I., J.G.-S., and D.F.-G.; Writing—review and editing, M.G.-S., C.M.-M., S.M.-I., J.G.-S., and D.F.-G.This research received no external funding.The authors declare no conflict of interest.Flowchart with selection of articles included in the review.Jadad scale.Methodological quality of studies, calculated with the Jadad scale. BLS basic life support; AED: automatic external defibrillator.Description of the studies included in the review. VAM: voice advisory mannequin.

Med-MDPI/ijerph_3/ijerph-16-05-00769.txt

@@ -0,0 +1 @@
+Medicine is a knowledge area continuously experiencing changes. Every day, discoveries and procedures are tested with the goal of providing improved service and quality of life to patients. With the evolution of computer science, multiple areas experienced an increase in productivity with the implementation of new technical solutions. Medicine is no exception. Providing healthcare services in the future will involve the storage and manipulation of large volumes of data (big data) from medical records, requiring the integration of different data sources, for a multitude of purposes, such as prediction, prevention, personalization, participation, and becoming digital. Data integration and data sharing will be essential to achieve these goals. Our work focuses on the development of a framework process for the integration of data from different sources to increase its usability potential. We integrated data from an internal hospital database, external data, and also structured data resulting from natural language processing (NPL) applied to electronic medical records. An extract-transform and load (ETL) process was used to merge different data sources into a single one, allowing more effective use of these data and, eventually, contributing to more efficient use of the available resources.Developments in medical research lead to an increase in the life expectancy of populations. Although this evolution leads to very positive outcomes, such as the end of some diseases (smallpox, plague, etc.) and the discovery of new approaches to several others, other challenges occur more often, such as dementia, cancer, etc. However, as medical knowledge grows and develops, other areas like computer science experience developments which can be used to support physicians in addressing these challenges. At present, healthcare providers produce and store large volumes of data, both medical and nonmedical. These data can regard drug prescriptions, treatment records, general check-up information, physician’s notes, medical information, or financial and administrative information. These data are essential not only to follow up on patients but also for management or research purposes. The knowledge that can be taken out of these data is therefore relevant for many purposes, such as service quality, by allowing the physician to access details regarding the patient that allow to improve diagnosis and decision-making. Service productivity can also benefit from this knowledge, as access to the required data is faster, letting resources increase the volume of their outcome and service control can be enhanced, by allowing disclosure of medical and performance data, among many other benefits. Data can be stored either in legacy systems or electronic medical records (EMR) [1]. EMRs are computerized medical information systems that collect, store, and display patient information [2].It is clear that analyzing data in an integrated way can drastically improve patient quality of care and both clinical and financial outcomes, but the way in which we collect, read, integrate, understand, and leverage data remains a broken process. Data integration and consequent analysis allows full approaches in healthcare diagnosis. A diverse range of prior studies focused their effort on analyzing data to predict waiting times, based on months, weekday, weather conditions, and external events like a football match or concert, among others [3]. Natural language processing (NPL) can extract information from electronic medical registers in a semistructured way and generate additional data, like the type and number of prescriptions, date of appointment, diseases, and need for surgery, among others [4]. Data and digitalization processes allow comparative data analysis, the use of past data events allows to predict new ones, and large volumes of data can be processed with data mining and machine learning algorithms.The format in which this information is recorded is essential because it has a direct impact on how it can be modeled to provide greater insight. Three different types of data can be found: Structured, semistructured, and unstructured [4]. Different data manipulation techniques, such as data mining (DM) and text mining (TM) or natural language processing (NLP) [5] are available, along with the associated data cleaning, data merging, and visualization approaches.Data, information and knowledge are three commonly misunderstood concept words that are occasionally used as equals, which may be the cause of some misunderstandings [6]. The three terms are related in a pyramid/chain-like relationship [7], yet they do not represent the same concept. The complexity and understanding increase from data to information and finally to knowledge. Data are a value, like a clinical measurement, such as heart rate (for instance, 50 beats per minute) [7], or just raw data, such as a clinical narrative [6]. Information, the next level in the hierarchy, is the result of data manipulation using processes such as referential, type, purpose, relevance, and interpretation [8]; in other words, data are put into context, acquiring some meaning. Continuing the example, in the context of a small child, a heart rate of 50 bpm gives some information to a doctor about the child, yet that same information could have different meanings if an adult presents the same values [7]. The third level in the scale, knowledge, is the most informative of the three. When information is structured and organized as a result of cognitive processing [7] to offer understanding, experience and accumulated learning [8] and validation, it becomes knowledge. Thus, to recapitulate, while data are raw values, information is what is achieved when those values are put into context, and, finally, knowledge is information that is structured, organized, and processed, and may be used to improve procedures or other processes.With the purpose of transforming data into knowledge so that medical professionals can improve the quality of the services they provide and the use of the time they have available, this work aims at developing a framework process to integrate data from different sources and make it available in a more effective way. In doing so, data from different medical records that result from a natural language processing (NPL) transformation are operated using an extract-transform and load (ETL) procedure to produce a single integrated file, allowing more effective use of that data. Addressing the goal of this research, this work starts with an analysis of the state-of-the-art in the area of data manipulation, considering the different types of data (structured, semistructured, and unstructured). Then, the methodology considered for the development of the framework is exposed. Three use cases are considered, two focused on structured data and a third one based on semistructured and unstructured data. The case studies that are presented concern the use of structured and unstructured data together and how the healthcare sector can gain value from the mixed use of these three types of data sets. The discussion is produced based on the findings from both use cases.Structured data are data that are stored in a fixed schema database, and out of the three data types, they are the easiest one to manage. The most common data in this scenario are demographic (e.g., race, ethnicity, birth date), admission and discharge dates, diagnosis codes (historic and current), procedure codes, laboratory results, medications, allergies, social information (e.g., tobacco usage), and some vital signs (blood pressure, pulse, weight, height) [9]. The fact that data are stored in a structured schema makes them easier and faster to access.There is no consensual definition of data mining (DM). For example, according to Reference [10], DM is described as “(…) the process of finding previously unknown patterns and trends in databases and using that information to build predictive models. Alternatively, it can be defined as the process of data selection and exploration and building models using vast data stores to uncover previously unknown patterns.” It can also be considered a methodology for discovering meaningful correlations, patterns and trends by sifting through massive amounts of data stored in repositories. Data mining uses a pattern recognition algorithm with a machine learning approach, as well as statistical and mathematical techniques. There is a wide range of different techniques, benefits, and areas that have been using DM, each for a specific situation. In the following sections, some case studies about DM in healthcare are reviewed to understand the diversity of techniques available, in which areas they are used, and the impact of each one of them. To facilitate the understanding on how structured data are managed and used in the healthcare sector, some cases are presented in Table 1, adapted from Reference [11,12], in which we divide the models into six areas: Classification, clustering, time series, regression algorithms, association, and hybrid model. For all of these DM types, many methods were used, showing the wide variety of techniques, applications, and possibilities. From cases studies [13,14,15,16], it is possible to access that there are many cases, in this scenario, nine out of 12, that use statistical methods as a base for classification models. On one hand, the majority of these case studies use DM techniques to train models and later apply them to predictions in many sectors, some more technical, such as cardiology [15,17,18], oncology [19], psychiatry [20], endocrinology [21], while others more focused on management, specifically quality of service [22], risk of patient rehospitalization [13], and forecasting daily bed needs [23]. On the other hand, as shown in Reference [16], predictions are not the only use of these structured data. Ravindranath explained that there are many ways to diagnose heart diseases and proposed a new decision support system for effect based on decision trees (DT) [16]. Analyzing Table 1, it can be perceived that there is a wide variety of methodologies or application of the studies that can be used. Nonetheless, it is possible to perceive that there are many studies that focus on prediction as previously stated.The unstructured data of an EMR are present in clinical notes, surgical records, discharge records, radiology reports, and pathology reports [6]. Clinical notes are documents written, in free text [6], by doctors, nurses, and staff providing care to a patient, and offer increase detail beyond what may be inferred from a patient’s diagnosis codes [26]. The information contained in clinical notes may concern a patient’s medical history (diseases, interventions, among others), family history of diseases, environmental exposures, and lifestyle data [4,6]. According to Reference [27], the knowledge of said notes is retrieved “by employing domain experts to curate such narratives”, which is not practical manually. Therefore, applying an automatic way of interpretation of these clinical notes and records is of the utmost importance. As explained, DM techniques that could be applied to structure data cannot be applied to this type of data without some previous structuring (preprocessing). Instead, TM is used to prepare and process the free text data. In the next chapters, we analyze how TM tools can be used, and case studies are presented. As this type of information is represented as free text, there is no common framework, and there may be improper grammatical use, spelling errors, local dialects [5,6], short phrases, and/or abbreviations [4,6]. Due to such difficulties, data processing and analysis becomes more difficult. A great deal of this difficulty is precisely the preprocessing of the so-called free text. Natural language processing (NLP) tools and techniques are used during the preprocessing phase and have proven very useful when it comes to extracting knowledge from ERM [4,6]. NLP technology involves the ability to turn text or audio speech into encoded, structured information, based on an appropriate ontology [6]. The structured data may be used solely to classify a document or ERM in a classified system or to identify findings, procedures, medications, allergies of patients, and others [6]. Therefore, with the help of methods like NLP, it becomes possible to structure the free text and apply DM techniques. Some examples of the types of operations that are done in preprocessing are the removal of digits, anonymization, and punctuation removal, among others. Table 2 shows some studies in the area of NLP and/or TM to extract structured information. This information can be later used in DM models for prediction [27] and disease identification [28,29,30].There are some TM techniques that are quite common, such as the removal of stop words, the lowercase treatment. Preprocessing depends on the data in hand and what that data are intended for. Table 2 shows examples of cases in which TM was used. Although the use of NLP methodologies and tools was different in all of them, the results were very satisfactory. Another significant aspect of these studies is that the unstructured data of the EMR that would not be used in some cases because it was too “wild” ended up being used and proved its usefulness.Until now, both DM and TM case studies have been shown and analyzed. While in some, only structured data were used, others only use the structured data extracted from the unstructured free text from de NLP processing used. Table 3 shows mixed approach case studies organized by year, application, reference, and methods. In these case studies, not only NLP techniques were applied, but DM techniques were also used:In Reference [31], topic modeling was used. It is a type of statistical modeling for discovering the abstract “topics” that occur in a collection of documents, and latent Dirichlet allocation (LDA) is an example of a topic model and is used to classify text in a document to a particular topic.In Reference [32], both techniques (Bayesian belief networks and decision trees (DT)) were used, and their results were compared. In other words, to detect early stages of dementia, the authors used two different techniques to assist specialists in the diagnosis of patients with clinical suspicion of dementia. It was possible to conclude that the model that used the structured data and the clustering of the texts written in free format by the physicians integrated, improved the accuracy of predictive models in all pathologies [32].In Reference [33], a prediction of patient admission was performed, applied to the logistic regression using five different iterations.In Reference [34], researchers had the objective of comparing the number of geriatric syndrome cases identified using structured claims and structured and unstructured EMR data to understand the added value of the latter. Conclusions were that results improved when combining both models. This type of fact led the authors to encourage “incorporating NLP methods to increase the sensitivity of identification of individuals with geriatric syndromes” [34].In Reference [31], topic modeling was used. It is a type of statistical modeling for discovering the abstract “topics” that occur in a collection of documents, and latent Dirichlet allocation (LDA) is an example of a topic model and is used to classify text in a document to a particular topic.In Reference [32], both techniques (Bayesian belief networks and decision trees (DT)) were used, and their results were compared. In other words, to detect early stages of dementia, the authors used two different techniques to assist specialists in the diagnosis of patients with clinical suspicion of dementia. It was possible to conclude that the model that used the structured data and the clustering of the texts written in free format by the physicians integrated, improved the accuracy of predictive models in all pathologies [32].In Reference [33], a prediction of patient admission was performed, applied to the logistic regression using five different iterations.In Reference [34], researchers had the objective of comparing the number of geriatric syndrome cases identified using structured claims and structured and unstructured EMR data to understand the added value of the latter. Conclusions were that results improved when combining both models. This type of fact led the authors to encourage “incorporating NLP methods to increase the sensitivity of identification of individuals with geriatric syndromes” [34].Recent research in the data processing [35] and biomedicine fields [36,37,38] underline the need for associating structure and information type with the datasets used in scientific projects. This information can be used to support standardization approaches and is most relevant regarding comma-separated values (CSV) or other tabular text formats where data types and structures are not enforced.The need for preprocessing procedures for data standardization has been recently explored, in particular when considering CSV or other tabular formatted text sources. Arenas et al. [35] pointed out that skewed tabular data—with missing rows and columns—is likely to occur, specifically considering CSV and other text tabular data formats, where typically there is no metadata information. The authors proposed a framework for automatically annotating CSV-like data.Aiming to facilitate data exchange and application interoperability through dataset reuse, this group was focused mainly on how metadata should be associated with a previously created dataset. The authors proposed the final recommendations based on the analysis of 25 case studies (http://w3c.github.io/csvw/ use-cases-and-requirements/). Regarding the scope of our work, some of the W3C recommendations should be underlined, in particular: (i) The relevance of associating syntactic and semantic information to each data column, and (ii) the relevance of defining only one data type for each column. The W3C work group suggested that the delivery of a dataset should be associated with metadata creation and sharing. Nonetheless, considering already available datasets, the workgroup stressed that the publication of metadata could benefit the user community and can be performed by an independent party, in addition to the dataset provider. The group argued that this possibility might enable the user community to benefit from each other’s efforts when it comes to dataset understanding and exchange.Ongoing research based on data collected during health care service delivery is increasing patient-centered knowledge [37]. However, the quality of EMRs is considered as not complying with research standards [36], which pushes research teams to overcome data quality problems using data preprocessing implementations. These implementations are frequently seen as near ad hoc solutions [35,37], not following any standardized data preprocessing approach. Wu et al. [39] published an extensive analysis of the impact of data preprocessing, modeling, and mining in the biomedicine context. These authors analyzed and described problems in structured and unstructured data, including data in EMRs. One of the work’s main conclusions was the need for data integration and interpretation guidelines as a factor that enables better prediction and therapeutics. Feder [37] focused mainly on structured information in EMRs, underlining the frequently missing information regarding data quality dimensions (like consistency or completeness, for example). The author recommended three general principles to be followed in EMRs research works: (i) The use of metadata definitions (referred to as data dictionary) for every dataset used, (ii) the use of statistical methods to deal with data quality issues (e.g., missing data) and finally, (iii) the generation of a report containing relevant data quality information (e.g., the proportion of missing data found, number of variables removal, and the number of transformations applied). Back in 2013, concerned with the lack of standardized data preprocessing approaches in data-driven biomedicine research, Weiskopf and Weng [40] analyzed 95 data-driven biomedicine research articles searching for the data quality dimensions most frequently referred by authors. This work was used by Reimer et al. [38] in their proposal of a six-step framework for data quality assessment in longitudinal registries (i.e., datasets built on heterogeneous data sources). Although the authors were not focused on analyzing data preprocessing methods, they concluded that associating metadata with datasets [38] is a useful factor when considering dataset reuse or replication.The main goal of this research was the definition of an ETL process for the creation of a single data set to improve the mining, prediction, and visualization process of data from several sources. The main idea is illustrated in Figure 1.Recommendations from Arenas et al. [35] were used in our research, as well as from the World Wide Web Consortium (W3C) [41].In order to extract clinical information from any EMRs, a whole system based on open-source modules is coupled together. Firstly, a translator is required to translate the EMRs’ clinical narratives from any language to the English language, with reasonable performance. Secondly, we need an open-source NLP system responsible for applying information extraction techniques and performing the clinical information extraction from the EMRs’ narratives (see [6]). An ontology unified medical language system (UMLS) to help biomedical vocabularies and standards is required to enable interoperability between computer systems. This UMLS is a repository of biomedical vocabularies developed by the US National Library of Medicine, containing more than 2 million concepts and 8 million concept names, some of them in different languages than English. A high-level depiction of the pipeline system we aim to build in this work is shown in Figure 2, and for more details and achieved results, see Reference [6].The output of this process is data extracted of diseases, symptoms, and clinical procedures, from the EMRs’ narratives, outputting a file with all of the extractions concerning those domains. These data are transformed into a CSV file to be able to be merged with structured data. Research teams commonly need to validate, clean or transform data contained in CSV input datasets so that they can be used in experimental setups. For that reason, researchers frequently implement ad-hoc preprocessing routines not completely specified in published papers, which can make work reproduction and validation difficult. The purpose of this work is to merge different source files, using as a validation approach the data in CSV formats. The use of heterogeneous data sources in data-driven scientific projects frequently requires a preprocessing phase where input data are interpreted, validated, transformed, and saved into an output data format that is typically different from the input data format. The results of each project seriously depend on the quality of data used in the experiments, regardless of their purpose (statistical analysis, knowledge discovery or other). The data preprocessing phase is crucial. Despite not having any data type information, CSV files are one of the simplest and most extensively used formats when it comes to data exchange [36].In order to address the lack of data preprocessing standardization in scientific projects, we propose to model the data preprocessing phase as an extract-transform and load (ETL) process. ETL derives from data warehousing and covers the process of how data are loaded from the source system to the data warehouse. A typical data preprocessing phase is thus composed of the following three phases: (i) Extract available data (extract), (ii) transform and clean data (transform), and (iii) store the output data in an output repository (load). These phases are preceded by the data source selection, typically a set of files or a database. Figure 3 illustrates the process. More detailed insight of each of the three phases is provided:1)The extract process is responsible for the data extraction from the source system or database and makes it accessible for further processing or management process. In health care, this process needs to deal with data privacy, and most of the extract process has an anonymization process associated. At this point, the researcher decides which data make sense to use.2)The transform is a process based on a set of rules to transform the data from the source to the target. In the current research, a new approach using semantics is proposed for this phase. This can be a complex process taking into account different dimensionalities cases, and it needs to assure that all variables are in the same units so that they can later be joined and a clean process can be conducted. The transformation step also requires joining data from several sources, generating aggregations, surrogate keys, sorting, deriving newly calculated values, and applying advanced validation rules.3)The load process merges all data into a target database (output data).The extract process is responsible for the data extraction from the source system or database and makes it accessible for further processing or management process. In health care, this process needs to deal with data privacy, and most of the extract process has an anonymization process associated. At this point, the researcher decides which data make sense to use.The transform is a process based on a set of rules to transform the data from the source to the target. In the current research, a new approach using semantics is proposed for this phase. This can be a complex process taking into account different dimensionalities cases, and it needs to assure that all variables are in the same units so that they can later be joined and a clean process can be conducted. The transformation step also requires joining data from several sources, generating aggregations, surrogate keys, sorting, deriving newly calculated values, and applying advanced validation rules.The load process merges all data into a target database (output data).The processed (output) data can then be used by data analysis and knowledge discovery algorithms. We structure our work in the transformation phase by dividing it into three sequential steps (E3TL). For each of these steps, there is an associated auxiliary configuration file. The first two steps are specific for each input file having its own specification and intermediary output file. The third and final step is unique and generates the final CSV output files. In more detail, the three E3TL steps are as follows (see Figure 4):(1)Split one row into many (1 : m row split). Input rows are split in a 1 : m transformation, in order to have just one row per patient examination (instead of one row containing multiple patient examinations). At this point, only the relevant input fields are kept. The specification of row splitting and field selection is made through an auxiliary file called split rules. In this file, each selected field of the input CSV file is associated to a type and eventually to a structure (e.g., a specific date format to be used such as ‘YYYY-MM-DD’, for ‘2019-01-23’).(2)Semantic validation. At this step, the resulting data files of 1:m row split are submitted to semantic validation, where the rules defined in the auxiliary file semantic rules are applied. These rules are based on domain knowledge and aim at validating specific domain constraints, like thresholds or non-empty fields in specific columns;(3)Data join. At this point, data is joint by applying the rules defined in the last auxiliary file, named join rules. This file describes which fields should be present in the output files and which will be used as keys when joining the rows from each file. In addition to the output data files, a file named transparency report is generated. It includes statistical information about operations applied during the transformation process.Split one row into many (1 : m row split). Input rows are split in a 1 : m transformation, in order to have just one row per patient examination (instead of one row containing multiple patient examinations). At this point, only the relevant input fields are kept. The specification of row splitting and field selection is made through an auxiliary file called split rules. In this file, each selected field of the input CSV file is associated to a type and eventually to a structure (e.g., a specific date format to be used such as ‘YYYY-MM-DD’, for ‘2019-01-23’).Semantic validation. At this step, the resulting data files of 1:m row split are submitted to semantic validation, where the rules defined in the auxiliary file semantic rules are applied. These rules are based on domain knowledge and aim at validating specific domain constraints, like thresholds or non-empty fields in specific columns;Data join. At this point, data is joint by applying the rules defined in the last auxiliary file, named join rules. This file describes which fields should be present in the output files and which will be used as keys when joining the rows from each file. In addition to the output data files, a file named transparency report is generated. It includes statistical information about operations applied during the transformation process.In order to validate our methodology, we applied it to a biomedical dataset, specifically to an amyotrophic lateral sclerosis (ALS) dataset produced by neurologists from Hospital de Santa Maria, in Lisbon, Portugal. This dataset contains information about ALS patients, including personal and medical examination information that was previously anonymized. In our experimental implementation, all transformations were coded in Python and the supporting files specified using Javascript notation (JSON) notation. Our experimental implementation of the E3TL methodology was also coded in Python. JSON was also used for auxiliary files specification. The implementation is available either as a Jupyter Python notebook and the equivalent Python source [https://github.com/ruifpmaia/e3tl]. The python file is imported as a module, and then each E3TL transformation can be invoked as a function.Under the scope of our work, we analyzed Amaral et al.’s [42] and Carreiro et al.’s [43] work on amyotrophic lateral sclerosis (ALS) disease evolution prediction, specifically on respiratory failure prediction in ALS patients. Our work focuses on proposing a new E3TL methodology based on the researcher’s data preprocessing approach, namely on row split (the authors refer it as execution split) and other frequently used data validations and transformations. For an extensive understanding of the preprocessing routines used by the authors, we also analyzed their code implementation in Java programming language and the obtained output files. The same ALS dataset was used in both experiments, made available for research purposes by Hospital de Santa Maria, a main Portuguese Hospital located in Lisbon. The model was applied to an ALS dataset produced initially as a Microsoft Excel spreadsheet format (XLS) by neurologists of Hospital de Santa Maria, containing information on 495 patients. The XLS file included 25 worksheets which were exported to 25 CSV files. Our approach generated, as expected, five output files matching the predefined output formats and a transparency report, statistically describing the E3TL process with information such as the number of input and output fields, the applied validations, and the number of input and out rows in each step. Data fields are delimited by one constant character and values can be simple (as integers, for example) or complex, like dates, floats or text comments. Each patient is uniquely identified by a number that is used across the different files as an identifier. Table 4 describes the data stored in the dataset in generic terms. In order to facilitate dataset processing, each worksheet was exported to a different CSV file. The dataset contains two classes of data: (a) Static, also referred to as administrative data, and (b) temporal data. Static data are independent of medical examinations and applied treatments and include personal information like age, gender or family history. Dynamic data are collected over multiple examinations and capture the results of the applied treatments. Each medical examination is composed of multiple clinical tests. These tests, associated with the same examination, may span several days or weeks. According to References [42,43], only 6 of the 25 input worksheets were used in the experiments, namely: Demographics, ALS-FRS, ALS-FRS1, NPO, Phrenic, and RFT. Demographics contains static patient information and family history, while ALS-FRS, ALS-FRS1, NPO, and Phrenic contain dynamic information obtained during medical examinations. Table 5 and Table 6 are examples of tabular data formats stored in the input worksheets. Table 6 contains the amyotrophic lateral sclerosis functional rating score (ALS-FRS) examination results. It is an example of dynamic data in which each line is associated with a patient, containing the results of multiple ALS-FRS examinations. The example shows how fields (Date, FeatureA, and FeatureB) are repeated in the same line, representing the values of two tests performed in different dates on the same patient.First, in 1:m row split, input rows are split. For example, each row of the ALS-FRS.csv file is analyzed, and if there is more than one medical examination per row, then it will be transformed into multiple rows according to the parameters in the auxiliary file named split rules. Each row field is parsed and formatted according to parameters also defined in the auxiliary file. For example, dates can be parsed and reformatted, or decimal conventions (decimal marks: ‘.’ or ‘,’, for example) can be normalized. This step will generate intermediary output files. The semantic validation step is applied to the intermediary output files produced by the first step. In this step, a significant number of rows might be removed due to validation of missing data, erroneous or unexpected values. These validations are parameterized by the semantic rules file. Mandatory fields found as missing or values under a required threshold, for example, might result in row deletion. Again, this step will produce new intermediary output files. The last step, data join towards the load process, consists of executing join procedures that will generate the final output files. All available fields from the intermediary output files (generated by the last step, semantic validation) can be used to form the final output file according to the parameters specified in the join rules auxiliary file. Addressing a double validation of E3TL, additional to a full verification against the descriptions in References [42,43], we checked the output files generated by our methodology with the ones received by authors implementations. The next three subsections detail each of the E3TL steps and the available parameters in each auxiliary configuration file.The first step of our methodology focuses on input files containing medical examinations. Each input file should have its own field mapping auxiliary file. This step splits each row of an input file into multiple rows whenever there is more than one medical examination per row. The resulting rows will only have the fields specified in the auxiliary configuration file metadata map (see Figure 5). Each field to be kept is associated with a data type and a specific structure (e.g., a date format ‘YYYY-MM-DD’). The 1:m row transformation depends on the knowledge about the structure of input files (Figure 5 exemplifies a row normalization transformation). Typically, it will not be applied for static data, such as a demographic information file. In the Santa Maria ALS dataset, each medical examination file contains one row for each patient, each one including multiple medical examinations. The same set of fields is repeated (as columns) through all exam iterations in the considered file. This step’s auxiliary file, identified as aplit rules, describes the field structure of the input and intermediary output file, specifying the fields that should be kept. Each field to be kept is associated with a metadata description, defining its type and format (when needed). This mapping operation is applicable whether there are multiple exams per row or just one exam per row. The definition of the split rules parameters requires knowledge about the field structure of the input data file. It specifies the fields to be present in the rows of the intermediary output file. This auxiliary file contains the following information:OutFile—The intermediary output file generated by the field mapping step;ExamFeatures—The number of fields per examination. This variable supports the indexing of each input field with the correct offset;ExamCountPerRow—The number of examinations per row. In the ALS-FRS.csv file, each row is associated with a specific patient. Therefore, it includes multiple examinations per row;OutputFeatures—The number of fields of the output table;FeatureMapping—The list of mappings between the column index of the selected input fields and the desired column output index;InputIdx—The index of a specified column in the input file;output is—The position of the field in the output intermediary file;StaticIdx—There are static and non-static based indexes. A static input column index is not dependent on the number of examinations per row. Differently, the “Date” field, with a non-static index, will be retrieved using the offset: InputColumnIndex = (exam number ∗ExamFeatures) + InputIdx where ExamNumber <= ExamCount;Name—The name to be used in the output file;Type—The data type to be used. Depending on the data type, different parsing routines will be used, either for multiple date formats, or different decimal notations, for example;Format—The output format of non-basic types to be used. For dates, for example, ‘YYYY-MM-DD’, for ‘2019-01-23’.OutFile—The intermediary output file generated by the field mapping step;ExamFeatures—The number of fields per examination. This variable supports the indexing of each input field with the correct offset;ExamCountPerRow—The number of examinations per row. In the ALS-FRS.csv file, each row is associated with a specific patient. Therefore, it includes multiple examinations per row;OutputFeatures—The number of fields of the output table;FeatureMapping—The list of mappings between the column index of the selected input fields and the desired column output index;InputIdx—The index of a specified column in the input file;output is—The position of the field in the output intermediary file;StaticIdx—There are static and non-static based indexes. A static input column index is not dependent on the number of examinations per row. Differently, the “Date” field, with a non-static index, will be retrieved using the offset: InputColumnIndex = (exam number ∗ExamFeatures) + InputIdx where ExamNumber <= ExamCount;Name—The name to be used in the output file;Type—The data type to be used. Depending on the data type, different parsing routines will be used, either for multiple date formats, or different decimal notations, for example;Format—The output format of non-basic types to be used. For dates, for example, ‘YYYY-MM-DD’, for ‘2019-01-23’.Regarding type and formatting, the 1:m row split can parse different multiple date input formats and decimal symbols in order to convert them to the selected output format. At the end of field mapping step, an intermediary file is generated having just one patient examination per row.The second step of the transformation phase applies data domain validations to an intermediary file generated by a 1 : m row split. Like in the previous step, there is different specification of the semantic validation step for each file, as well as a different validation rules auxiliary file. This step runs three types of validations: (1) Missing values to verify if a field is empty; (2) threshold validation, to check if a value is under or above a specified threshold; and (3) regular expression validation, to analyze a value against a regular expression (for example, to validate if a field contains only integer values: ^\d+$). These three types of validations were shown to cover the needs of the case study. Taking the ALS-FRS.csv file as an example, the validation and filtering of exam rows depended on the verification of patient ID, as the first output field, with the “InputIdx” : 0. It should be simultaneously non-null (“NotNull” : true) and a positive integer, represented through the regular expression "RegEx": "^[0–9]+$". An example of the validation rules of an intermediary file (ALSFRS.csv) obtained from the processing (by field mapping) of the ALS-FRS.csv input file. The available parameters for the validation rules auxiliary file are the following:OutFile—The intermediary output file generated by the semantic validation step;Semantic Rules—The list domain validations that should be applied to a specific field;InputIdx—The index of a specified column in the input file. NotNull requires a non-null value;RegEx—Verify if the value of the field maps to the specified regular expression;Threshold—Verify if the value of the field is over the specified threshold (only applicable to integer or double fields).OutFile—The intermediary output file generated by the semantic validation step;Semantic Rules—The list domain validations that should be applied to a specific field;InputIdx—The index of a specified column in the input file. NotNull requires a non-null value;RegEx—Verify if the value of the field maps to the specified regular expression;Threshold—Verify if the value of the field is over the specified threshold (only applicable to integer or double fields).The last step of the transformation phase is a join transformation where fields from the intermediary output files generated by the semantic validation step are possibly merged. According to the rules defined in the associated auxiliary file join rules, this step executes an Inner Join operation using the specified key. The result will be the generation of an output file with the specified fields. The available parameters of the join rules auxiliary file are the following:OutFile—The final output filename;FeatureMapping—The list of fields that should be present in each row of the output file and their relation to the input fields (from the intermediary files). Typically, each row includes dynamic (exams) and static information (like patient date of the birth, for example);InputIdx—The index of the input field on the auxiliary input file.InputFile—The auxiliary input filename;Name—The final output field name;Type—The output field data type;Format—The output field format for non-basic types (like dates, for example);JoinKey—Indicates if the field is part of the Join key.OutFile—The final output filename;FeatureMapping—The list of fields that should be present in each row of the output file and their relation to the input fields (from the intermediary files). Typically, each row includes dynamic (exams) and static information (like patient date of the birth, for example);InputIdx—The index of the input field on the auxiliary input file.InputFile—The auxiliary input filename;Name—The final output field name;Type—The output field data type;Format—The output field format for non-basic types (like dates, for example);JoinKey—Indicates if the field is part of the Join key.Table 7 presents an example of the ALSFRS file joined with the NIV (non-invasive venthilation) field retrieved from the Demographic file.Our methodology produced output files with the expected data and structure according to the specified E3TL process. The output files generated for the Santa Maria dataset include dynamic and static information. The dynamic information was retrieved from the original files containing data of medical examinations (i.e., NPO, ALSFRS, Phrenic, and RFT) and was later joined with static information from the demographics file. The output result was stored in five different files and represented in Table 8, Table 9, Table 10 and Table 11 (the demographics output table was split in three, due to its size). Results from the three E3TL steps were as follows:(a)1 : m row split execution example. The function receives two variables as arguments: One with the name of the CSV input data file and the second with the name of the JSON specification auxiliary file. As output, the function returns the name of the intermediary CSV output file.1 : m row split execution example. The function receives two variables as arguments: One with the name of the CSV input data file and the second with the name of the JSON specification auxiliary file. As output, the function returns the name of the intermediary CSV output file.(b) Semantic validation execution example. The function receives two variables as arguments: one with an intermediary CSV data filename generated by the previous step, and the second with the name of the JSON specification auxiliary file. As output, the function returns the name of the intermediary CSV output file.(c) Data join execution example. The function receives as arguments the JSON specification associated auxiliary file. As output, the function returns the name of the final CSV output file.The process generated a transparency report with the statistical description of input and output datasets and the number of operations executed in each phase of E3TL. The number of transformations applied and missing values found, for example, are registered in this file. Table 12 (a), (b), (c) presents the collected information.Some procedures with small changes on Python scripts were applied to urgency data from a hospital in Lisbon. These data include registers from 1 January 2013 to 31 December 2017. The original data were enriched with external data, like weather conditions and events. The ETL process was applied to generate a greater data file (see Figure 6) that allows improvements in the mining and prediction processes. This integration allowed the prediction of waiting times based on the day of the week, on the month, weather conditions and correlations with special seasons, and events, among others. For details, see Reference [43,44].In another use case, these data were integrated with the output of NLP over EMR [6]. A schema of all the components involved in the text processing is shown in Figure 7. This schema was applied to 5255 EMRs from a Portuguese hospital. Once all the extracted clinical information is structured and gathered in a CSV file, it is possible to apply the ETL process and create a single CSV file that can be loaded for the data mining process in open sources or commercial platforms. Results of this are available in Reference [6].The evaluation of the NPL system was performed based on standard metrics calculated for 75 of the 5255 EMRs. These standard metrics are precision, recall, and F1 score. Two healthcare practitioners from the hospital manually annotated the clinical terms present in 75 EMRs in order to establish a gold standard for this evaluation. The evaluation showed that the NPL system coupled together in this research has a precision of 0.75, recall of 0.61, and an F1 score of 0.67.The main drawback of the CSV format is the fact that it does not enclose data and information type for the stored fields. This problem exponentially escalates when dealing with large datasets, updated by multiple users or organizations and using different standards for information representation. Typically, research teams working on data-driven scientific projects focus on data analysis and knowledge discovery algorithms, trying to obtain, for example, the best prediction models and results. They implement ad hoc data preprocessing routines based on their domain knowledge for data validation and transformation operations. Preprocessing routines are commonly not detailed in published works that are mostly focused on reporting the implemented data analysis techniques and obtained results. Little attention is given to preprocessing standardization. Considering tabular formats like CSV, this lack of documentation or incomplete description can be a relevant issue hurdling work assessment and reproduction. Without ensuring reproducibility, it may not be possible to re-execute part of or an entire scientific work, thus restraining its validation or further development.We propose improvements in the area of tabular data integration and data preprocessing in the medical area. A new ETL methodology—E3TL—was proposed based on CSV file preprocessing in data-driven scientific projects aiming at standardized procedures for data integration, cleaning, and transformation.Our approach has the advantage of defining a comprehensible and systematic ETL process for CSV preprocessing. It also integrates three primary vectors concerning data processing and exchange (or reuse) transparency: (i) Metadata declarations as a basic structure for data preprocessing; (ii) systematic data validation definitions; (iii) integrated transformation, validation, and filtering metrics summarized in a transparency report. By including metadata information in E3TL, we aimed at facilitating interpretation and sharing of the original dataset knowledge, following W3C recommendations.The results from our research can improve the quality of the service provided to the patients at the hospital as data are integrated and knowledge is available to medical professionals to gain a broader and more complete picture of the patient they are treating. In parallel, by having the data available and transformed into knowledge, they can lead to faster and more accurate diagnosis, which allows’ medical personnel to become more productive. From a more managerial perspective, this type of integration and the knowledge it makes available can also be used, for instance, for the prediction of waiting times [43], allowing crossing it with external information, like weather conditions or big events.Applying the proposed approach to other use cases and datasets may lead to refinements and further developments of the E3TL model, including the incorporation of more detailed statistic processes. This approach can be generalized to different cases, as long as the baseline conditions are verified. Generalization to other cases will require adjustments to the specificities of those cases. Nonetheless, the overall framework is applicable.New achievements regarding the availability of knowledge from data can be performed with the merge of different data sources using as input structured and unstructured data. This is not only very positive for the hospitals and healthcare centers but, and most importantly, for the patients. Considering that the main objective of this field is offering healthcare to those in need or, in some scenarios, the best quality of life possible, by using structured and unstructured data separately, results can be obtained as shown. This mixed approach can be used not only in technical areas but also in operational ones. For example, through better disease prediction models, earlier diagnostics can be achieved, allowing a more adjusted medication to the patients and eventual avoidance of more acute health states. It can also be used to forecast seasonality and understand what sector of the hospital would be more requested in each season to improve the allocation of staff. Data integration and manipulation allow DM processes to be conducted using enormous data volumes to anticipate problems and/or solve them at their early stages, improve diagnosis and treatment because more data and cross-check functions are available and, in general, improve the quality of health services and patients’ quality of life.One paramount issue in DM is that data can be spread across many entities (hospital, government, and others). The integration that results from this approach allows to create new data sets with more information, and new knowledge can be generated to improve healthcare. Further, this integration of different data from the hospital database with ERM and external data also allows to improve management procedures based on the outputs from the data mining process. In this approach, the privacy issue was overcome with data anonymization. Data were used with the permission of the hospital.B.I.H. and A.L.M. provided an overview of the medical field application and wrote most of the final document. R.M. is a Ph.D. student who conducted all implementation and partly wrote the final document. J.C.F., as a data science expert, developed all the relations concerning data manipulation, data mining and text mining.This research was conducted without external funding.We are grateful of Hospital Santa Maria, in Lisbon, for providing the data.The authors declare no conflict of interest.Methodological paths of the research.A high-level view of the whole pipeline system.Research data preprocessing viewed as an extract-transform and load (ETL) process.Detail of the Transform phase of the proposed (E3TL) approach that is applied to comma-separated values (CSV) input files.1:m row split example.Data integrated from a hospital in Lisbon using the proposed ETL process.Schema of developed components involved in the text processing.Example of data mining (DM) case studies from 2013–2016.Examples of text mining (TM) case studies from 2015–2018.Example of mixed approach case studies 2015–2018.Dataset description.ALSFRS input table using a comma-separated values file.Demographics input table using a comma-separated values file.ALSFRS output table, generated using ASLFRS (dynamic) and demographics (static) file information.RFT output table; generated using RFT and demographics (NIV—non-invase venthilation) original input information, Forced Vital Capacity (FVC), Maximal Inspiratory and Expiratory pressures (MIP/MEP), partial gas concentrations (P0.1, PO2 and PCO2), values of mean (or under 90%) oxygen saturation (SpO2mean and SpO2 < 90%)NPO output table; generated using NPO and demographics (NIV) original input information.Phrenic output table; generated using phrenic and demographics (NIV) original input information.Demographics output table generated using demographics original input information. The table is partitioned in three parts due to its size.(a) Part one(b) Part 2(c) Part 3Transparency Report for the three phases of the proposed ETL framework. Field mapping, semantic validation, and join transformation are described from a data integration point of view through multiple information fields, including the number of applied transformations, missing values count, and type or format errors.(a) 1 : m Row Split(b) Sematic Validation(c) Data Join

Med-MDPI/ijerph_3/ijerph-16-05-00770.txt

@@ -0,0 +1,2 @@
+In October 2018, at Asia Pacific Conference for Disaster Medicine (APCDM), an expert meeting to identify key research needs was organized by the World Health Organization (WHO) Centre for Health Development (WHO Kobe Centre (WKC)), convening the leading experts from Asia Pacific region, WHO, WHO Thematic Platform for Health Emergency and Disaster Risk Management (Health-EDRM) Research Network (TPRN), World Association for Disaster and Emergency Medicine (WADEM), in collaboration with Asia Pacific Conference for Disaster Medicine (APCDM) and Japan International Cooperation Agency (JICA). International experts, who were pre-informed about the meeting, contributed experience-based priority issues in Health-EDRM research, ethics, and scientific publication. Two moderators, experienced in multi-disciplinary research interacted with discussants to transcribe practical issues into related methodological and ethical issues. Each issue was addressed in order to progress research and scientific evidence in Health-EDRM. Further analysis of interactive dialogues revealed priorities for action, proposed mechanism to address these and identified recommendations. Thematic discussion uncovered five priority areas: (1) the need to harmonize Health-EDRM research with universal terms and, definitions via a glossary; (2) mechanisms to facilitate and speed up ethical review process; (3) increased community participation and stakeholder involvement in generating research ideas and in assessing impact evaluation; (4) development of reference materials such as possible consensus statements; and (5) the urgent need for a research methods resource textbook for Health-EDRM addressing these issues.At the Asia Pacific Conference for Disaster Medicine (APCDM) [1], October 2018, an expert meeting to identify key research needs in major research areas was organized by the World Health Organization (WHO) Centre for Health Development (WHO Kobe Centre (WKC), convening the leading experts from Asia Pacific region, WHO, WHO Thematic Platform for Health Emergency and Disaster Risk Management (Health-EDRM) Research Network (TPRN), World Association for Disaster and Emergency Medicine (WADEM). An expert meeting was conducted along with aseries of progresses in scientific aspects of the implementation of the 2015 Sendai Framework on Disaster Risk Reduction (SFDRR) [2], the resulting document of the Third UN World Conference on Disaster Risk Reduction (WCDRR), included the establishment of TPRN [3,4] and following journal papers on recommended Health-EDRM research activities [5,6]. Through the expert meeting and related review of literature and existing projects and activities, key research needs in five major Health-EDRM research areas were identified.The Health-EDRM Network identified one major area of work that is important to address was clarity in relevant ‘Research Methods and Ethics’. The broad intersection of health and disaster risk reduction has resulted in an area of work now known as Health-EDRM which encompasses emergency and disaster medicine, disaster risk reduction, community health resilience, health system resilience, and impact of changing climate on health. Public health response during and after disasters has traditionally been focused on protecting populations from immediate threats [7]. Health-EDRM research involves the systematic analysis and management of health risks in emergencies and disasters by reducing the health risks and vulnerability. The complexity of undertaking research in disasters, and complying with ethical standards for these research, is critical but often much more difficult to ensure. This paper summarizes the outcome of the discussions and the proposed actions needed to support the delivery of Health-EDRM research.Prior to the meeting a range of international experts contributed experience-based priority issues in Health-EDRM research, ethics and scientific publication. It is of note that even in 1997, Stallings was able to state that “… it is the context of research not the methods that makes disaster research unique” [8]. The lead discussant, the rapporteur, and the other experts who participated to the discussion primarily aimed to identify priorities in scientific evidence on Research Methods and Ethics in Health Emergency and Disaster Risk Management Research focused on questions and issues to fill these gaps. In addition to identifying knowledge gaps, experts also aimed to assess knowledge-to-practice gaps in order to better integrate current expertise and research in this area into each phase of disaster risk management. Following a preliminary literature review and expert consultations, a series of questions that were thought to be very important to address for building better understanding of research methods and ethics in Health-EDRM included
+      (a)What are the definitions of research methods and technical terms for Health-EDRM?(b)How can impact evaluation methods for intervention and qualitative—quantitative mixed methods be standardized?(c)How can the publication process for Health-EDRM research become more systematic and effective?What are the definitions of research methods and technical terms for Health-EDRM?How can impact evaluation methods for intervention and qualitative—quantitative mixed methods be standardized?How can the publication process for Health-EDRM research become more systematic and effective?Those participated in discussion were leading experts in Health-EDRM as well as country experts in disaster and emergency, and came from multiple disciplines such as public health, emergency medicine, nursing, and health care management. They discussed major issues in Health-EDRM research from practitioner viewpoints and different regional perspectives. Each issue was addressed by active discussion of participants with the aim of addressing priorities and actions on how to progress research and scientific evidence in Health DRM. Interactive dialogues noted simultaneously into minutes of discussion were further analyzed by the moderators into the gaps, proposed mechanism to overcome the gaps and to provide a summary of recommendations.Experts from different parts of the globe participated in this thematic group discussion. The findings from this discussion were wide ranging. It was noted that research findings from disaster research is not easily translated into different contexts of the many countries around the world, and there were challenges and difficulties in implementing practices in addressing national health system, cultural and religious issues before, during, and after interventions. These issues were difficult to address without more complete and systematic evidence to inform Health-EDRM policy and practice. Although translating research findings into policy is the ambition of many researchers and practitioners in order to develop evidence-based policies, there are issues of how much researchers can communicate with policy makers in comparison to the opportunities to facilitate policy makers’ uptake of research findings. A key strategy to overcome this barrier is stakeholders’ involvement and community participation since the development of research ideas in designing phase of the research project. For example, research in disaster affected area might recruit participants who were disasters casualties and there should be rules and regulation especially listing ethical ‘don’t’s such as providing food as an incentive for participation in the research. For instance, food is sometimes used as incentive for the participation in the research, in some context. However, in the disaster and emergency situation, food is the basic need provided as the humanitarian aid regardless of participation in the research. Thus, if food is used as incentives for research recruiting such persons, it might be forcing someone to participate in the research, in the fear that they cannot receive food supply. Therefore, it is not appropriate to use food as incentives in Health-EDRM research.Thematic discussion in the meeting uncovered five priority areas: (1) the need to harmonize Health-EDRM research with universal terms and definitions via a glossary; (2) mechanisms to facilitate and speed up  ethical review process; (3) development of reference materials such as possible consensus statements; (4) increased community participation and stakeholder involvement in generating research ideas and in assessing impact evaluation; and (5) and the urgent need for a textbook for Health-EDRM research addressing these and other issues. Discussions also agreed that there was a need to harmonize definitions in Health-EDRM research with universal terms, and the development of a glossary of definitions. Such a glossary could promote common understanding and common usage of concepts, terms, and aims for Health-EDRM. If undertaken by WHO, the glossary might have the additional advantage of being translated into the WHO official languages. Even though the United Nations Office for Disaster Risk Reduction (UNISDR) convened an Open-ended Intergovernmental Expert Working Group (OIEWG) to report on indicators with recommended terminology relating to disaster risk reduction, which was delivered in 2016 [9], and adopted by the UN General Assembly on 2 February 2017 and updated the 2009 UNISDR Terminology on Disaster Risk Reduction [10]; however, not all the necessary Health-EDRM terms were included in these or other terminologies.The need to find mechanisms to facilitate and speed up ethical review process was addressed and the need to quicken the review process for disaster research ethics was noted as being complex and very dynamic. A basic requirement was to agree on methodological terms and should reflect research undertaken in the spectrum of disaster chronology such from preparedness and risk reduction, to emergency management, and to post disaster rehabilitation. One key suggestion was the requirement to reduce the review time for ethical approval with, wherever possible, with pre-agreed ethical approval. As Chan et al. (2019) pointed out their recent Lancet editorial that “research stakeholders have a responsibility to protect the interests of communities involved in research, achieving this is rarely straightforward in emergencies” [11]. From the discussion that it would require international consensus among the professionals and researchers, possibly using as good practice models the CONSORT (Consolidated Standards of Reporting Trials) statement for randomized controlled trials [12,13] and STROBE (STrengthening the Reporting of OBservational studies in Epidemiology) statement for cohort studies [14,15]. Additional models of good practice such as the Guidance for Managing Ethical Issues in Infectious Disease Outbreaks [16] and the Health Emergency and Disaster Risk Management fact sheet on ETHICS [17] were thought to be practical and helpful examples. More data on these topics to address Health-EDRM were considered to be of importance. Once a research proposal can fit in a standardized check list, the ethical committee should be able to agree the research proposed via an expedited channel.Such activities could be part of the development of reference materials such as possible consensus statements. Such consensus statement provide a checklist of methodology details which allows researcher to check the proposal themselves. Ethical committee can quickly assessthe quality assurance and safety through the list, speeding up the approval. Therefore, it is very essential step to develop a consensus for disaster research.The call for increased community participation and stakeholder involvement in generating research ideas and in assessing impact evaluation was considered and the need to listen to the voice of the affected and their community leaders and local representatives is increasingly critical. However, relatively few Health-EDRM reports on community participation and stakeholder involvement in generating research ideas were shared in the discussion. There are examples of where research to address stakeholder involvement in health research, such as the recent report from Kapiriri (2018) [18], but it is not from emergency or disaster research. Very little has been published definitively on this area.Much more is described in the quest of methodological rigour for impact evaluation. The experts brain-stormed how can impact evaluation apply quantitative methods in additional to commonly applied case-studies in disaster research.Impact evaluation is usually applied to see how a research programme works well in a particular setting. It is a term understood readily by multiple stakeholders whereas it dictates how to measure the outcome an intervention in basic research methodology. There are many methodological tools to measure how an intervention brought about changes in comparison to pre-existing situation or in comparison to naïve control group. A proposed intervention should be tested by efficacy and effectiveness trials before it comes into the guidelines and practice. It was considered that the synthesis of evidence would be primarily through the process of systematic reviewing and, if appropriate, modelling and cost effectiveness decision analysis [19]. The efficacy depends on how the intervention is planned to measure (design), how accurate are the measurement tools (validity and reliability) whereas the effectiveness will inform how robust is the intervention in the real world setting. Furthermore, the scalability of intervention will be challenged by the economic, social, and political context and the resource need. It would be important to follow process recommended where possible by organisations and their reports of activities that are already engaged in working in this area such as WHO [20,21], UK Medical Research Council [22], Organisation for Economic Co-operation and Development [23], and the Humanitarian Policy Group at the Overseas Development Institute and its Good Practice Review [24] with its chapter on Monitoring and Evaluation [25,26].Diverse research methods from case studies, natural experiment designs and randomized controlled trials can be applied in impact monitoring and evaluation. For example, clinical epidemiology is a discipline to determine clinical outcomes applying tools of epidemiology. Likewise, epidemiological tools can be selected to match where and how they would be applied in Health-EDRM research but are often less easy to monitor and evaluate as they are used in more difficult environments. Through the synergy of Health-EDRM experts and the discussion on research methods, designs, and tools, it might be possible to encourage selection and use of more appropriate mechanisms. Events to gather such multidisciplinary professionals such as international workshops are important opportunity to generate the list of strategic research methodology tools. Relatively few resources for disaster risk reduction and management research methods have been identified and some excellent examples are cited below [27,28,29,30,31,32]. However most of these do not specifically address the full range of Health-EDRM research domains. Therefore, it is considered that there is relatively little currently that reflects the wide needs of Health-EDRM researchers, practitioners, and policy makers. Therefore, a text book linked to a website for easy updating is needed and such a resource could be tentatively entitled “research methods for health emergency and disaster risk management”. This would be a rewarding endeavor and would be beneficial to the establishment of the WHO Thematic Platform for Health Emergency and Disaster Risk Management Research Network.It is hoped this paper on key issues in research methods and ethics in health emergency and disaster risk management will contribute to the identification and implementation of concrete solutions that foster the creation and the use of knowledge for research and ethics building resilience before, during, and after disasters. In the discussion, issues were raised on potential gap areas in the disaster research methodology:  impact evaluations; Consensus on the definitions; development of common research statements like the CONSORT or STROBE for disaster research; how research findings in disaster research can be translated into different contexts of the many countries around the world; and the need to develop the textbook in disaster research methodology to help guide international researchers.As for other major Health-EDRM research areas, research and ethics requires collaboration between experts, decision-makers, practitioners, and communities in order to facilitate coordinated response when and where it is most needed. It is critical a text book is created to provide a reference which would be fundamental and global contribution to the establishment of the WHO Thematic Platform for Health Emergency and Disaster Risk Management Research Network.Conceptualization, R.K. and V.M.; Methodology, M.N.A. and V.M.; Writing—Original Draft Preparation, M.N.A. and V.M.; Writing—Review & Editing, V.M., M.N.A. and R.K.; Project Administration, R.K.This research received no external funding.The authors thank all the meeting participants who contributed to achieving the results (alphabetical order of their last name): Jonathan Abrahams (WHO), Francis Archer (Monash University, Australia), Sarah Louise Barber (WHO), Emily YY Chan (The Chinese University of Hong Kong, China), Shinichi Egawa (Tohoku University, Japan), Paul Farrell (WADEM), Teodoro Herbosa (University of the Philippines, Philippines), Sonoe Mashino (University of Hyogo, Japan), Toshiyuki Ojima (Hamamatsu University School of Medicine, Japan), Philip Schluter (University of Canterbury, New Zealand), Shiori Usami (Kumamoto University, Japan).The authors declare no conflict of interest.

Med-MDPI/ijerph_3/ijerph-16-05-00771.txt

@@ -0,0 +1 @@
+These authors contributed equally to this work.Xerostomia (dry mouth) is the cardinal symptom of Sjögren’s syndrome (SS), which is an autoimmune disease involving the exocrine glands and other organs. Xerostomia may predispose patients to periodontal disease (PD) and an association between SS and PD has been reported. This association may be bidirectional; therefore, we conducted this study to investigate the risk of SS in patients with PD using data from the National Health Insurance Research Database of Taiwan. A total of 135,190 patients were enrolled in our analysis. In all, 27,041 patients with PD were matched by gender, age, insured region, urbanization and income, with cases and controls in a 1:4 ratio. Both groups were followed and the risks of SS were calculated by Cox proportional hazards regression. Finally, 3292 (2.4%) patients had newly diagnosed SS. Patients with PD had a significantly higher risk of subsequent SS (903 (3.3%) vs. 2389 (2.2%), adjusted hazard 1.47, 95% confidence interval: 1.36–1.59). In conclusion, patients with PD had an approximately 50% increased risk of subsequent SS. Physicians should be aware of the symptoms and signs of SS in patients with PD.Sjögren’s syndrome (SS) is an autoimmune disease characterized by exocrine gland dysfunction [1,2]. It is one of the most prevalent autoimmune diseases, with an estimated prevalence of 2–270 per 10,000 inhabitants [3,4,5]. The cardinal symptoms of SS are classical “sicca symptoms” of dry eyes (keratoconjunctivitis sicca) and dry mouth (xerostomia) but clinical manifestations may vary [2]. Usually, SS can be classified as either primary SS, which refers to involvement of solitary lacrimal and salivary glands or secondary SS, in which coexistence of other autoimmune diseases is observed, such as systemic lupus erythematosus and rheumatoid arthritis [1,2]. The heterogeneity of signs and symptoms often leads to a delay in diagnosis. Immune-mediated inflammatory processes are the main causes of SS and obvious lymphocytic infiltrates in salivary and lacrimal glands are observed [1,6,7,8]. Elevated autoantibodies, such as anti-Ro/SSA and anti-La/SSB, are common in patients with SS. Mucosal destruction and impairment of barrier function following immune-mediated inflammatory responses subsequently induce local inflammation and cause a vicious cycle. Extra glandular involvement of SS is common and increased risks of systemic diseases have been observed in patients with SS [9]. In patients with SS, increased risk of thyroid disease, cardiovascular diseases, gastrointestinal diseases, neuropsychiatric diseases and hematologic diseases were reported, and the systemic influences of SS raised our attention [10,11,12]. However, the underlying pathophysiology is complicated and not fully understood.Periodontal disease (PD) is a common disease experienced by an estimated 20–50% of the global population [13,14]. Poor oral hygiene is believed to contribute to PD and patients with poor oral habits have a 2–5-fold increased risk of PD [13,15]. Patients with dry mouth, a smoking habit and diabetes also are at higher risk of PD. Poor oral hygiene results in colonization by oral microorganisms and causes local invasion, breakdown of barrier function and destruction of structures surrounding the teeth, including the gums, periodontal ligament or alveolar bone [13,16]. Furthermore, chronic oral inflammation may cause systemic inflammatory responses and alteration of the cytokine expression profile in serum or saliva have been reported, including interleukin (IL)-1, IL-6, IL-10, IL-17A, IL-17F, IL-22, IL-25, IL-33, tumor necrosis factor-alpha and interferon-gamma [17,18,19]. Long-term, systemic inflammation increases the risk of systemic diseases in patients with PD. Patients with PD were found to have a higher risk of cardiovascular diseases, maternal infection, preterm birth, low birth weight, preeclampsia and ulcerative colitis [14,20,21,22,23]. The important impact of PD on extra-oral systems is worth our attention.Xerostomia is common in both diseases and immune alterations are evident in SS and PD. An association between SS and PD may exist. Several studies have investigated the risk of PD in patients with SS and they found that the mean plaque index, gingival index, bleeding on probing and mean gingival index were larger in patients with SS [24,25,26]. However, the evidence was not conclusive and the underlying pathophysiology was not clear [27,28]. The increased risk may be bidirectional and the risk of SS in patients with PD remains unclear. Therefore, we conducted this nationwide, retrospective cohort study to compare the risk of SS in patients with PD and a non-PD control group.This study was approved by the Institutional Review Board of the Tri-Service General Hospital (approval number: TSGHIRB NO B-104-21). We extracted patient data from the National Health Insurance program of Taiwan, which is a nationwide medical insurance system with high coverage—99.5% of Taiwan’s 23 million residents—and highly representative of nationwide medical data [29,30]. Medical information regarding patient diagnosis and treatment-related information are included in the National Health Insurance Research Database (NHIRD). The International Classification of Diseases, 9th revision, Clinical Modification (ICD-9-CM) coding system was used in NHIRD and 1 million randomly selected individuals from the NHIRD in 2000 (LHID2000) were analyzed for our study. Patients with PD were selected as the study group and a matched group without PD was selected as the control group. Both cohorts were tracked to investigate the incidences of SS.A total period of 13 years was investigated in our study and the flow chart of enrollment is shown in Figure 1. First, we identified patients with PD (ICD-9-CM: 523.3, 523.4 and 523.5) as the study group. The exclusion criteria included (1) history of PD before 2001, (2) incomplete medical records, (3) younger than 20 years and (4) previous history of autoimmune disease. The diagnosis was made by licensed dentists and mainly based on symptoms and local findings. The index date referred to the date of diagnosis of PD. Furthermore, a matched cohort was identified as the control group, with cases and controls in a 1:4 ratio. For each patient in the PD group, four matched controls were enrolled of the same gender, age, insured region, urbanization and income. Finally, we tracked both groups to identify patients newly diagnosed with SS (ICD-9-CM: 710.2) and compare the risk of SS in these two groups. SS was diagnosed mainly by rheumatologists according to clinical manifestations and laboratory tests. SS is a catastrophic illness in Taiwan and application for certification requires review by a second specialist. The index date referred to the date each individual was enrolled in the study. The censoring date referred to 7 years after the index date, death, the date of diagnosis of SS or lost-to-follow-up.Medical and demographic information for both cohorts was extracted from NHIRD and analyzed, including age, monthly income, geographic area of residence, urbanization level of residence and comorbidities. The primary outcome was newly diagnosed SS, a catastrophic illness in Taiwan. Age was divided into six groups based on 10-year intervals: 20 to 29, 30 to 39, 40 to 49, 50 to 59, 60 to 69 and ≥70 years. The monthly income of the study population was recorded in New Taiwan Dollars and categorized into four income levels. The geographic regions in Taiwan were divided into four areas: the northern region, central region, southern region and other region (eastern and outlying islands). The urbanized level of residence in Taiwan was classified into four categories. Finally, comorbid diseases included diabetes mellitus (DM, ICD-9-CM: 250), hypertension (ICD-9-CM: 401–405), hyperlipidemia (ICD-9-CM: 272), coronary artery disease (ICD-9-CM: 410–414), stroke (ICD-9-CM: 430–438), alcoholism (ICD-9-CM: 291, 303, 305.00–305.03, 571.1, 571.2, 571.3, 790.3, A215 and V11.3), obesity (ICD-9-CM: 278) and tobacco use disorder (smoking, ICD-9-CM: 305.1, 491.0, 491.2, 492.8, 496, 523.6, 649.0, 989.84 and V15.82). The incidences of newly diagnosed SS and the abovementioned covariates and comorbidities in both cohorts were investigated and analyzed.We used a Student’s t-test and a chi-square test to analyze and compare the categorical demographic characteristics and comorbidities of the two cohorts. Cox proportional hazards regressions were performed to evaluate the relationship between PD and subsequent SS. Moreover, the hazard ratio (HR) was calculated with the 95% confidence interval (CI) to compare the risk of PD. Further adjustment for potential confounders (age, gender, income, geographic area of residence, level of urbanization of residence and comorbidities) was performed in all models and the adjusted HR (aHR) was calculated. A two-sided p value < 0.05 was considered to indicate a statistically significant result. Statistical analyses were performed using SPSS software version 19.0 (SPSS Inc., Chicago, IL, USA) and data were managed with Microsoft® SQL Server® 2008 software (Microsoft Unternehmen, Redmond, DC, USA).As shown in Figure 1, 27,041 patients with PD were identified as the study cohort. We matched each individual in the study cohort for age and gender with controls at a 1:4 ratio and 108,149 patients without PD were enrolled in the control cohort. In total, 135,190 individuals were enrolled in our study and followed to investigate the incidence of SS. Table 1 lists the demographic data on both cohorts; there were no significant differences in age and gender. Most participants were in a low income bracket and resided in highly urbanized areas and in northern Taiwan. Patients with PD had higher rates of comorbidities other than alcoholism, including DM, hypertension, hyperlipidemia, coronary arterial disease, stroke, obesity and tobacco use disorder.The average follow-up duration was 6.91 ± 0.65 years for the PD cohort and 6.94 ± 0.57 years for the control cohort. Finally, 3292 (2.44%) patients had newly diagnosed SS and patients with PD had a higher incidence of SS (3.34% vs. 2.21%, crude HR = 1.52, 95% CI: 1.41–1.64, Table 2). After adjustment for confounding factors, Cox proportional hazards regression was performed to evaluate the independent risk of subsequent SS and the results are shown in Table 3. The aHR for PD was 1.47 (95% CI: 1.36–1.59, p < 0.05, Table 3). Additionally, males had a significantly lower risk of SS (aHR: 0.36% CI: 0.33–0.39, p < 0.05). Individuals with high income and those who resided in central Taiwan were at higher risk of SS. Patients with hypertension, hyperlipidemia, stroke, alcoholism and smoking were at lower risk of subsequent SS.Our study is the first to investigate the relationship between PD and subsequent SS. Xerostomia is common in both diseases and immune-mediated inflammatory processes are involved in both diseases. We found an approximately 50% increased risk of newly diagnosed SS in patients with PD. Physicians should be aware of symptoms and signs of SS in patients with PD, such as dry mouth and dry eyes.Xerostomia is a cardinal symptom of SS and is common in patients with PD. Several studies have investigated the incidences of PD in patients with SS. Patients with SS had a significantly higher risk of PD [24,25,31,32,33]. The observed risk may be bidirectional, and our study provided a scientific evidence supporting increased risk of SS in patients with PD. Several factors may contribute to the observed increase in risk. First, oral hygiene and overgrowth of microorganisms may contribute to the association. In patients with SS, destruction of the salivary glands results in dry mouth and bacterial overgrowth and is followed by PD. Periodontal treatment in patients with SS may increase salivary flow and improve clinical and immunological parameters and quality of life [34]. Furthermore, alterations of cytokine network may play important roles in the link between PD and SS. Alterations of the cytokine profile in serum or saliva have been reported in patients with PD, including IL-1, IL-6, IL-10, IL-17A, IL-17F, IL-22, IL-25, IL-33, tumor necrosis factor-alpha and interferon-gamma [17,18,35,36,37]. An increased risk of systemic diseases, such as cardiovascular diseases, diabetes, rheumatoid arthritis, preeclampsia, preterm birth and ulcerative colitis, has been reported in previous studies [14,20,21,22,23]. Similarly, obvious dysregulation of the cytokine network has been observed in patients with SS and cytokines may be considered as potential therapeutic targets [2,6,38,39]. In patients with SS, overexpression of pro-inflammatory cytokines, including interferon-γ, IL-12, IL-18, IL-6 and B-cell activating factor, was found. In contrast, anti-inflammatory cytokines, such as IL-4 and transforming growth factor-β, were downregulated. However, the complete mechanisms underpinning PD and SS are complicated and not fully understood and we found an obvious increased risk of SS in patients with PD. An association may exist between these two diseases and further studies are warranted to clarify the underlying pathophysiology.The cardinal symptoms of SS are dry mouth and dry eyes and are easily overlooked. Delayed diagnosis of SS is not uncommon [2]. The Sjögren’s Syndrome Foundation launched a 5-year Breakthrough Goal in January 2012 and its aim was to reduce the duration between the onset and diagnosis of Sjögren’s syndrome by 50% in 5 years [40]. They showed that the average period from symptom onset to diagnosis was approximately 6 years. The average follow-up duration in our study was 6.91 ± 0.65 years for the PD cohort, compatible with previous reports. Dry mouth in patients with PD may be an early presentation of SS. Reducing the time from onset to diagnosis is valuable and our study found a statistically increased risk of SS in patients with PD. Physicians should educate their PD patients and be aware of the symptoms of SS. If PD patients have one of the following symptoms: dry eyes, skin rash, joint pain, salivary gland swelling or any other symptom of SS, further investigation for SS is recommended.After adjustment for confounding factors, our study found a decreased risk of subsequent SS in patients with hypertension, hyperlipidemia, stroke, alcoholism and smoking (Table 3). The association between SS and other systemic disease is complicated. Chiang et al. investigated the risk of stroke in patients with SS and no obvious association was found. The relationship between smoking and SS was reported previously and anti-inflammatory effects of cigarettes may contribute to the observed reduction of risk [41,42]. However, the adverse effects of smoking are undoubtedly versatile and huge and smoking is not encouraged. Again, the underlying biological mechanisms of SS and related systemic diseases are complex and further studies are required.As the NHIRD is a nationwide database with broad coverage, the findings of our study are representative of the general population and clinically significant. However, our study is subject to some limitations. First, laboratory test results are not available in our database. Although SS is a catastrophic illness in Taiwan and the diagnosis is accurate and verified, it is valuable to explore inflammatory markers and disease subgroups and to investigate the possible mechanism behind the observed association. Second, patients with different severities of PD may have different risk levels of subsequent SS. Details regarding PD severity would contribute to further clarification of the increased risk. However, classification of PD severity relied on the detailed local findings, but related information was not complete in present database. Therefore, the severity of SS and treatment responses were not analyzed. Although a 13-year total study period is not short, long-term prognosis and treatment may be different. Similarly, edentulous disease (teeth loss) is common in patients with severe PD and overlaps the pathophysiology of PD [43]. Complete information of teeth loss in each individual was not available and we did not exclude patients with edentulous disease. It is valuable to compare the risks of SS in patients with different severity of PD and different edentulous conditions. Moreover, the entire mechanism of SS is not completely understood and further studies are required to elucidate the underlying pathophysiology linking PD and SS.In conclusion, this large-scale, nationwide, population-based study found that patients with PD have an approximately 50% increased risk of subsequent SS. An immune-mediated inflammatory response may contribute to the association. Physicians should be aware of the symptoms and signs of SS in patients with PD and appropriate investigations of SS may contribute to an early diagnosis of SS.All authors have read and approved the final manuscript. C.-Y.L., C.-F.T. and R.-J.H. conducted the methodology; J.-M.L., H.-C.C., W.-T.L. and L.Y.-M.L. collected the data; C.-L.Y., C.-F.T., J.-M.L., Y.-C.Y. and R.-J.H. designed the study, conducted the research, analyzed the data and performed the validation of the results; C.-Y.L. and C.-F.T. wrote the first draft; R.-J.H. had primary responsibility for the final content.This research received no external funding.The authors would thank Enago (www.enago.tw) for the English language review.The authors declare no conflict of interest.Flow chart illustrating the enrollment of study cohorts.Distribution of gender, age groups and comorbidities in individuals with and without periodontal disease (PD).Note: Abbreviations: CAD: coronary artery disease; DM: diabetes mellitus; PD: periodontal disease.Association between PD and Sjögren’s syndrome (SS) analyzed by employing a Cox regression model.‡ p < 0.001 for comparison between patients with two groups.Independent predictors of SS identified by Cox regression analysis.* Each variable was adjusted for every other variable. Abbreviations: CAD: coronary artery disease; DM: diabetes mellitus; HR: hazard ratio; PD: periodontal disease. † p < 0.05 for comparison between patients with two groups. ‡ p < 0.001 for comparison between patients with two groups.

Med-MDPI/ijerph_3/ijerph-16-05-00772.txt

@@ -0,0 +1 @@
+These authors contributed equally to this work.This cross-sectional study aimed to compare access to the nearest food stores with perceived access associated with intake frequencies of vegetables/fruits and meat/fish among older Japanese people. We used intake frequencies of vegetables/fruits and meat/fish from a self-administered questionnaire in the Japan Gerontological Evaluation Study among 83,384 adults aged over 65 years. We defined distance over 1 km as poor objective access in community level. We performed multilevel regression analysis to investigate the association of objective and perceived access with intake frequencies of vegetables/fruits and meat/fish, respectively. Participants who lived in poor objective access had a significantly higher intake frequency of vegetables/fruits than those who lived in good access. In contrast, residents with poor perceived access consumed lower frequent intake of vegetables/fruits (beta coefficient (standard error) 0.086 (0.021) for objective access; −0.093 (0.009) for perceived access). There was no significant association between objective access and intake frequency of meat/fish, but poor perceived access showed a significant association with lower intake frequency of meat/fish. There was inconsistency between objective and perceived measurement of access to food stores associated with dietary habits among older Japanese adults. Food access needs to be comprehensively assessed, while considering characteristics of measurements.Areas with poor food access, where it is relatively difficult to obtain healthy and affordable food, are referred to in Western countries as “food deserts” [1]. Previous reviews [1,2] have suggested that poor food access induces social disparities in diet-related health outcomes, such as obesity in relation to ethnic minorities and socioeconomically disadvantaged neighborhoods, especially in the United States and other developed Western countries. In Japan, it has been observed that approximately 35% of residents in a large city had poor access to fresh food as estimated using an objective information system (GIS) in line with a 2015 report [3]. Specifically, Japanese “food deserts” mainly affect older people residing in neighborhoods where smaller retail stores have closed because of the recent economic recession [4,5,6]. In Japanese urban or suburban areas, older adults often experience inconveniences when the stores were located more than 1 km away from their home [5]. The inconvenience was majorly induced by the physical burden of failing health and limited transportation [7]. Ikejima [3] reported approximately 35% of residents aged 65 years and older had poor access to fresh food in a large Japanese city. Therefore, the association between food access and dietary intake among older people has been given priority in public health research.This study was designed to investigate the limited evidence on the association of food access with dietary intake, such as vegetables/fruits and meat/fish intake. First, Asian studies that used both GIS-based (objective) and perceived access and compared the association with dietary intake are scarce, although some western studies have been reported [8,9,10,11]. Objective access is limited in its ability to measure store utilization or residents’ true access to stores [9]. Therefore, it is important to investigate whether objective access is associated with individual dietary habits as well as perceived access. Second, empirical evidence of the association between food access and vegetables/fruits and/or meat/fish intake among community-based older adults had not yet been reported. Previous articles have targeted younger and middle-aged individuals [8,9,10,11,12,13,14], mixed race/ethnic populations [8,9,10,11,12,14,15], and people with low socioeconomic status [9,11,12,13,14]. Third, seven studies performed in western settings [8,9,10,11,12,14,15] did not show a consistent association between food access and vegetables/fruits intake. One study [10] showed that individuals were more likely to increase their servings per day of vegetables/fruits with increasing distance from a primary food store. However, two studies showed that individuals who lived more remotely with decreased access to food stores consumed significantly lower vegetables [12,15] and fruits [15] than those who lived in close proximity. While, four of the studies [8,9,11,14] reported no significant association between objective food access and vegetables/fruits intake. Fourth, most studies [1,2] used only vegetables and fruits as a measure of healthy dietary habits. It is important to investigate meat and fish intake in relation to food access because these foods are one of the protein-rich foods and associated with frailty prevention among older people [16]. Fifth, most previous studies have investigated the association between food access and dietary intake in only urban/suburban areas [8,9,10,11,12,13] or rural areas [15], except for a study by Pearce et al. [14] that adjusted for urban and rural areas, as a higher number of food markets with fresh vegetables/fruits and meat/fish are generally located in urban areas rather than in rural areas [17]. Finally, although some Japanese studies have reported an association between objective and perceived access and health outcomes [18,19,20], more evidence is required. A recent study showed that lower availability of healthy food stores measured subjectively, but not objectively, was associated with mortality [20]. Regarding nutritional status, objective [18] and subjective [19] food access showed a culture-specific association with being obese or underweight among Japanese older adults compared to western settings [1,2]. As these two studies [18,19] investigated a limited region, studies involving a larger-scale region are needed.Our aim was to compare objective and perceived access associated with intake frequencies of vegetables/fruits and meat/fish among Japanese community-dwelling older people. As an additional investigation, we repeated the analyses stratifying by urban/suburban and rural areas. Furthermore, we analyzed the association of objective and perceived access with the prevalence of underweight and overweight individuals. In addition, to confirm the availability of healthy food, we investigated the association of intake frequencies of vegetables/fruits and meat/fish with body mass index (BMI) which is one of the measures of nutritional status.Respondents were identified from the Japan Gerontological Evaluation Study (JAGES), a prospective cohort study investigating the influence of the social determinants of health outcomes among older individuals aged 65 years or older in Japan [21,22]. For this study, we used cross-sectional data from the 2010–2011 survey (response rate: 66.3%). Participants who did not receive long-term care and resided in 31 municipalities in 12 of 47 prefectures in Japan were included. Of 102,869 potential participants, we excluded those with missing information on school districts (n = 4099), and those who lived in school districts with fewer than 50 residents (n = 5134) [23]. We used school districts as the smallest area unit available in the JAGES [22]. Historically, school districts were used to represent the former unit of “villages” before repeated municipality mergers took place in the last few decades in Japan [24].After excluding individuals with missing data on the self-reported frequencies of vegetables/fruits and meat/fish intake (n = 7004) and missing or “I don’t know” response to perceived food accessibility (n = 3248) in the questionnaire, a total of 83,384 participants (38,615 men and 44,769 women) who resided in 426 school districts across 29 municipalities were included in the subsequent analyses (Figure 1). Of those, the number of individuals in urban/suburban areas and rural areas were 60,576 (28,472 men and 32,104 women) and 22,808 (10,143 men, 12,665 women), respectively. This study was conducted in accordance with the Declaration of Helsinki. The JAGES protocol and informed consent procedures were approved by the Ethics Committee for Research of Human Subjects at Nihon Fukushi University (no. 10-05 and no. 13-14) and the Ethics Committee for Medical Research at the University of Tokyo (no. 10555).Food access was estimated using a GIS map created by the Policy Research Institute, Ministry of Agriculture, Forestry and Fisheries of Japan [5]. First, we created half-grid square data of the population from the 2010 national census obtained from the Statistical Information Institute for Consulting and Analysis [25] and the number of food stores, including large-scale department stores, giant or small supermarkets, and specialty shops (those selling vegetables, fruits, meat, and fish). Convenience stores were excluded based on wholesale or retail sales data from the Current Survey of Commerce, in 2007 [26]. Second, we calculated the probability of the population accessing the nearest food store at a specific distance from a residence, assuming that the population and stores were distributed uniformly within the half-grid square [5]. The probability of the population accessing stores was estimated every 100 m for a 0–1.9-km radius centered on the point of residence, every 1 km for a 2.0–19.9-km radius, and every 10 km for a 20–70-km radius. Third, we calculated the probabilities of the population accessing the nearest food store between each radii rage centered on the point of residence as described above. Next, we estimated the weighted average of the distance to the nearest food store within the half-grid square. Examples of the processed model include the following:the probability of the population accessing the nearest food stores located within a 50-m distance (i.e., midpoint of the radius between 0 and 100 m centered on the point of residence)over 0 m − over 100 m = 100% − 97.6% = 2.4%;the probability of the population accessing the nearest food stores located within 2.5 kmover 2 km − over 3 km = 12.0% − 9.4% = 2.6%;the probability of the population accessing the nearest food stores located within 65 kmover 60 km − over 70 km = 0.1% − 0.0% = 0.1%; and the weighted average of the distance to the nearest stores within the half-grid square= (50 m × 2.4% + 2.5 km × 2.6%+ 65 km × 0.1%)/(50 m + 2.5 km + 65 km).the probability of the population accessing the nearest food stores located within a 50-m distance (i.e., midpoint of the radius between 0 and 100 m centered on the point of residence)over 0 m − over 100 m = 100% − 97.6% = 2.4%;the probability of the population accessing the nearest food stores located within 2.5 kmover 2 km − over 3 km = 12.0% − 9.4% = 2.6%;the probability of the population accessing the nearest food stores located within 65 kmover 60 km − over 70 km = 0.1% − 0.0% = 0.1%; and the weighted average of the distance to the nearest stores within the half-grid square= (50 m × 2.4% + 2.5 km × 2.6%+ 65 km × 0.1%)/(50 m + 2.5 km + 65 km).Finally, the weighted average of the distance to the nearest food stores within the half-grid square was aggregated and averaged within the school districts as an indicator of food access at the community level (Figure 2). Areas with a population density greater than 4000 population/km2 were defined as urban/suburban areas, and those with a lower population density were defined as rural areas according to the Statistics Bureau [27]. The mean population densities of the school districts established by the JAGES [28,29], calculated in 486 school districts in urban/suburban areas and in 95 school districts in rural areas, were 7442 and 499 population/km2, respectively. This study used binominal food access to define groups of a distance of less than and over 1 km as good and poor food access, respectively [19].The measurement of perceived access in the JAGES study has been used previously [19,20]. The perceived availability of food was assessed using the question “How many stores or facilities selling fresh fruits and vegetables are located within 1 kilometer of your home?”. The following responses were given on a four-point Likert scale: “Many”, “Some”, “Few”, “None”, or “I don’t know”. Subjects who answered “Many” or “Some” were categorized as having high access, and respondents who answered “Few”, or “None”, were categorized as having low access. For a more a precise assessment, we excluded participants who responded “I don’t know” in this study.We used responses about average intake of vegetables/fruits and meat/fish over a one-month period among participants as previous study used [19]. The intake frequencies were stratified into the following categories: every day and over twice/day, every day and once/day, 4–6 times/week, 2–3 times/week, once-a-week, less than once-a-week, and almost never. In this analysis, we assigned scores of 2, 1, 0.7, 0.4, 0.1, 0.05, and 0 (times/day), respectively, to each of the categories.The number of convenience stores in each school district was used as a covariate, as data on current food access were not included due to technical issues. Further, the degree of land slope in the neighborhood (continuous value) and car use by individuals, family members, or friends driving (yes or no) were significant covariates in terms of food access [18,30]. The average land slope at a community level was calculated by using the national dataset from the Ministry of Land, Infrastructure, Transport and Tourism in Japan, based on the Digital Map 50 m Grid (Elevation) from the GIS [31]. To consider the area difference, we used urban/suburban and rural areas as a covariate. We further considered the following covariates in this study: age (65–69, 70–74, 75–79, or ≥80 years), sex (men or women), family structure (living alone, with a spouse, or with others), BMI (<18.5, 18.5–24.9, or ≥25 kg/m2), marital status (married, divorced, widowed, or never married), activities of daily living (ADLs) (<5 or 5 units), the number of remaining teeth (≥20 or <19), presence of comorbidities (yes or no), smoking status (current, past, or never), household income (<2.00, 2.00–3.99, or ≥4.00 million yen), and years of schooling (<9, 10–12, or ≥13 years). The unknown variables were treated as categorical data to examine any associations between food access and the frequency of vegetables/fruits and meat/fish intake. BMI was calculated as the body weight in kilograms divided by the square of the body height in meters. ADLs were assessed by five items: use of public transportation, shopping for daily necessities, preparing meals, paying bills, and managing bank deposits [32]. The annual normalized household income was determined from the total household income divided by the square root of the number of household members as an equivalent household income. In terms of comorbidities, respondents were asked if they were currently under medical treatment for any of the following conditions (all of which may confound vegetable/fruit and meat/fish intake): cancer, heart disease, stroke, hypertension, diabetes mellitus, obesity, hyperlipidemia, osteoporosis, gastrointestinal disease, mental disorders, or dysphagia [33].By using food access (poor vs. good) of objective measurement on a community level and perceived measurement on an individual level as binomial explanatory variables, we performed a multilevel Tobit model, adjusting for all covariates, to estimate the standardized beta (β) coefficient and standard error (SE) for the intake frequency of vegetables/fruits and meat/fish. Based on the random-effects variance, we calculated the intraclass correlation (ICC) to examine the proportion of the variance in dietary habits (i.e., intake frequencies of vegetables/fruits and meat/fish) that occurs at the neighborhood (i.e., school district) level [34]. An ICC equal to 1 would inform us that all the people in a neighborhood have identical dietary habits, and an ICC equal to 0 would indicate that the people do not share any neighborhood related those at a common level. As additional analyses, we repeated the analyses stratifying by urban/suburban and rural areas. Furthermore, a multilevel logistic regression model was performed to investigate the associations of objective and perceived access with the prevalence of underweight (BMI < 18.5 kg/m2) and overweight (BMI ≥ 25 kg/m2) among 80,012 residents without missing variable of BMI. We also investigated the association of intake frequencies of vegetables/fruits and meat/fish with BMI by using a multilevel regression model adjusted for age and sex. Statistical significance was set as a two-sided p-value of 0.05. All statistical analyses were conducted by using Stata (ver. 15.0; StataCorp, College Station, TX, USA).The proportion of individuals with poor food access was 36.4% (30,383 individuals) by objective measurement and 25.3% (21,105 individuals) by perceived measurement (Table 1). The average age was approximately 74 years old in all participants. Most of residents in poor objective and perceived access lived with others, and were likely to be married, BMI 18.5–24.9 kg/m2, <5 units of activity daily living, <20 of remaining teeth, having comorbidity, past or never smoking, <2.00 million yen/year of household income, ≤9 years of schooling, and using a car. The 36.6% residents in poor objective access and 61.8% residents in poor perceived access lived in urban/suburban areas. In community level, there were approximately 2–3 convenience stores in poor objective and perceived access. A steeper land slope in poor access was observed in objective and perceived access and was clearly observed in objective poor access.The associations of poor food access (vs. good) with the intake frequency of vegetables/fruits and meat/fish are shown in Table 2. Individuals who lived in poor objective access had significantly higher intake frequency of vegetables/fruits than those who lived in areas with good objective access (β = 0.086 (SE) 0.021). In contrast, individuals with poor perceived access had significantly lower intake frequency of vegetables/fruits than those living in good ones (β  = −0.093 (SE) 0.009). There was no significant association between poor objective access and the intake frequency of meat/fish, but a significant inverse association was observed for perceived access (β = −0.029 (SE) 0.004).When we performed stratification by urban/suburban and rural areas, associations of objective and perceived access with the intake frequency of vegetables/fruits and meat/fish did not change significantly. The association of objective and perceived access with the intake frequency of vegetables/fruits in urban/suburban areas was greater than those in rural areas (Tables S1–S4). As shown in Table S5, poor objective access compared to good food access was significantly associated with a lower prevalence of underweight individuals. However, poor perceived access was weakly associated with higher prevalence of underweight individuals compared to good food access. No significant association was observed between food access and the prevalence of obesity for both objective and perceived access. Regarding the association between intake frequencies of vegetables/fruits, meat/fish, and body mass index, a significant association (p-value < 0.001) was observed (Table S6).The present study found that the intake frequency of vegetables/fruits in poor objective access areas was significantly higher than that in good access areas among Japanese older people, on the contrary to the inverse association using perceived access. There was no significant association between objective access and intake frequency of meat/fish; however, poor perceived access was associated with lower intake frequency of meat/fish. To our knowledge, this is the first study to compare the measurement between objective and perceived access in relation to intake frequencies of vegetables/fruits and meat/fish among Japanese older people in a large-scale population-based study.Our finding is in line with another study [10] that evaluated urban senior citizens in the United States, that suggested an increase, which was not statistically significant, in servings per day of vegetables/fruits for every 10th of a mile in distance to a primary food store. However, two studies among rural seniors [15] and urban residents aged over 16 years [12] showed that individuals who lived in poor objective access areas consumed significantly lower vegetables [12,15] and fruits [15] than those who lived in good access areas. Several studies have reported no significant association between objective access and vegetables/fruits intake among low-income and/or urban residents including younger people [8,9,11,14]. The inconsistencies in the results of this study and those of previous western studies [8,9,11,12,14,15] could be partly explained by culture-specific food environments, and populations with comparably younger ages, low income, and minority race/ethnicity.This study found that, contrary to perceived access, those with poor objective access had significantly higher intake frequency of vegetables/fruits than those with good access. Objective access might not correctly reflect actual individual food purchasing behaviors better than perceived ones [9,11]. Especially in urban/suburban areas, residents with poor perceived access did not necessarily live in areas with poor objective access in this study. Urban residents are more likely to travel beyond their nearest supermarkets due to their demands, such as healthy foods [10] and low-cost foods [8]. Of those with poor objective access, 65% residents had good perceived access in this study. The residents with good perceived access in the poor objective access group might lead to higher intake frequency of vegetables/fruits. Some latent components that could not be incorporated with the objective access might result in the inconsistency of results between objective and perceived access in this study. Among latent components that could not be incorporated with the objective access, we suggest the possible components from Japanese specific food environment as follows. First, objective access might not appropriately capture that older adults may obtain vegetables/fruits at small local markets that were not identified by national data of food markets against perceived access. Second, there is a possibility that land uses for agriculture or fishing confounded the association between objective access and dietary intake. With the presence of land use for agriculture or fishing, residents might find it easier to obtain vegetables/fruits and fish through small farmer’s markets and/or food exchanges [35,36] than those who lived in other areas. Third, online shopping, home delivery service, a food vendor vehicle service, and small retail shop provided by the local government, nongovernmental organization, social organizations, or large retail companies (i.e., convenient stores) increased in number to support older residents who lived in areas lacking food stores since around 2010 [7]. These services may attenuate the inconvenience of food access due to the distance to the food stores.As previous studies did not focus on meat/fish intake in relation to objective and perceived access, the present findings could not be compared with other studies. The associations between objective and perceived access and intake frequency of meat/fish in this study indicated a similar trend but a weaker association compared to those of vegetables/fruits. We suggest that the association of food access with the intake frequency of meat/fish could be explained with similar reasons as those mentioned for vegetables/fruits above.This study showed poor objective access was significantly associated with lower prevalence of underweight. In contrast to our result, Hanibuchi et al. [18] found that the number of supermarkets on an individual level was negatively associated with being underweight, although this was not statistically significant. Furthermore, this study showed poor objective access was weakly associated with a higher prevalence of overweight compared to those with good access. However, the previous study [18] showed that there was a significant negative association between the distance to the nearest supermarket and overweight/obesity. The association between perceived poor food access and higher prevalence of underweight individuals in this study was consistent with the study by Nakamura et al. [19]; however, there was no significant association due to the limited regions. Although further prospective studies are warranted, it may be necessary to support older residents with poor perceived access to prevent underweight.Our study has the following strengths. First, we employed an accurate food access map from national census data compiled by the Policy Research Institute, Ministry of Agriculture, Forestry and Fisheries of Japan [3]. Second, we adjusted for the number of convenience stores and the grade of the land slope in the school districts as confounding factors in the association between food access and dietary intake. Third, our results have generalizability, since the data was collected from a large-scale investigation conducted in both urban/suburban and rural areas.Nevertheless, this study has some limitations. First, national data for assessing objective access might not be able to identify small food markets and non-market-based food access (e.g., exchanging with neighbors and making home gardens) [35,36]. This limitation may lead to the underestimation of objective access. Second, the perceived measurement we used was not validated. However, we assessed perceived access to be comparably accurate according to the hillier environments where residents with poor perceived access lived. Older adults with poor perceived access were likely to have a lower level health status than those who reported good access in this study. This characteristic was consistent to that reported in a previous study [7] which reported that older people who live in food desert areas suffer from the physical burden of failing health when trying to access groceries. Therefore, the perceived access can accurately describe the accessibility to food stores. Third, we measured dietary habits using only data regarding vegetables/fruits and meat/fish intake. We did not investigate the validity of self-reported intake frequencies of vegetables/fruits and meat/fish using a dietary record. However, population-based studies [19,33] have used simple measures to assess intake frequencies of vegetables/fruits and meat/fish, representing a limitation of this field of research. In addition, we confirmed that intake frequencies of vegetables/fruits and meat/fish were available for assessing healthy food by confirming a significant association with BMI. Nevertheless, Japanese people usually consume a variety of foods [37], which are purchased at food stores. Therefore, the association between food access and dietary habits could have been identified more clearly if we measured the dietary diversity among older adults [38]. Fourth, this study had no information as to whether older adults reported intake frequencies of vegetables/fruits and meat/fish purchased through take-out meals or through informal networks. As this study did not measure the amounts of vegetables/fruits and meat/fish in cooked meals, the energy intake from vegetables/fruits and meat/fish may be under- or over-estimated. Finally, due to the cross-sectional design, causality could not be evaluated.We found that there existed inconsistency between objective and perceived measurement of access to food stores associated with intake frequencies of vegetables/fruits and meat/fish among older Japanese adults. Food access should be comprehensively assessed, taking into account the characteristics of measurement of food access. In the future, we should perform prospective studies to investigate the association between food access and dietary habits affected by several factors including affordability, accommodation, and acceptability in addition to GIS-based measures [39]. Using these assessments, it is important to decide what dimension of food access we should support with priority for older adults.The supplementary materials can be found at https://www.mdpi.com/1660-4601/16/5/772/s1. Table S1: Characteristics of participants by objective and perceived access in urban/suburban areas, Table S2: Characteristics of participants by objective and perceived access in rural areas, Table S3: Intake frequencies of vegetables/fruits and meat/fish according to objective and perceived access in suburban/urban areas, Table S4: Intake frequencies of vegetables/fruits and meat/fish according to objective and perceived access in rural areas, Table S5: Evaluation of the association of food access with the prevalence of underweight and overweight individuals compared to those with a normal weight, Table S6: Evaluation of the association of intake frequencies of vegetables/fruits and meat/fish with body mass index.All authors approved the final version of the manuscript. M.Y. conceptualized and designed this study, performed the data analysis, and drafted the manuscript; K.T. conceptualized and designed this study, and made a dataset of food access; M.H. performed statistical analysis and supported the study design; N.S. contributed to the critical revision of the manuscript; K.K. and N.K. contributed to data collection and the critical revision of the manuscript.This study used data from the Japan Gerontological Evaluation Study (JAGES), which was supported by MEXT(Ministry of Education, Culture, Sports, Science and Technology-Japan)-Supported Program for the Strategic Research Foundation at Private Universities (2009–2013); JSPS (Japan Society for the Promotion of Science) KAKENHI Grant Numbers (22330172, 22390400, 23243070, 23590786, 23790710, 24390469, 24530698, 24683018, 25253052, 25870573, 25870881), 15K16181 (M.Y. received), 18K05856 (K.T. received), 15K18174 (M.H. received), 18K13885 (N.S. received), and 18H04071 (N.K. received); Health Labour Sciences Research Grants (H22-Choju-Shitei-008, H24-Junkanki (Seishu)-Ippan-007, H24-Chikyukibo-Ippan-009, H24-Choju-Wakate-009, H25-Kenki-Wakate-015, H26-Irryo-Shitei-003 (Fukkou), H25-Choju-Ippan-003, H26-Choju-Ippan-006) from the Ministry of Health, Labour and Welfare, Japan; the Research and Development Grants for Longevity Science from AMED (Japan Agency for Medical Research and development) (JP18dk0110027, JP18ls0110002, JP18le0110009, 17dk0110027h0001); and a grant from National Center for Geriatrics and Gerontology, Japan (24-17, 24-23). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.We would like to express our deep gratitude to all the participants of the study survey. We especially thank Tetsuro Yakushiji for their dedicated efforts to create a dataset of food access. We also thank the other members of the JAGES project for their constructive suggestions during the writing of this article.The authors declare no conflict of interest.Flow diagram of the study participants.Estimated distance to the nearest food store (food access (m)) by school district. Values on the maps represent food access in each school district in rural areas (left) and urban/suburban areas (right). As indicated, the school district was colored from light-gray to dark-gray as the distances to stores increased.Characteristics of 83,384 participants by objective and perceived access.SD = standard deviation; a Numbers or mean (SD) were indicated in Total.Intake frequencies of vegetables/fruits and meat/fish according to objective and perceived access.β  = beta coefficients; SE = standard error; Var RE = random effect variance in 426 school districts; ICC = intercorrelation between school districts; Associations were assessed using a multilevel Tobit model among 83,384 residents: Var RE (SE) and ICC (SE) in the null model was 0.016 (0.002) and 0.014 (0.002) for vegetables/fruits and 0.006 (0.001) and 0.025 (0.002) for meat/fish, respectively.

Med-MDPI/ijerph_3/ijerph-16-05-00773.txt

@@ -0,0 +1 @@
+Cycling has proven to be an important strategy in decreasing the risk of non-communicable diseases. This study aimed to discover barriers and enabling factors influencing satisfaction and safety perceptions towards the use of bicycle roads in the Seoul metropolitan area, South Korea. A cross-sectional survey of 190 youth and adult individuals was conducted. Sex, age, purpose of bicycle use, perceived safety, availability of facilities, road gradient, road width, and traffic on the bicycle road were associated with cycling regularity. Multivariate regression analysis showed that the sufficiency of bicycle parking space, moderate slope, and enough bicycle signs were significant enabling factors for satisfaction with the use of bicycle roads. Narrow bicycle roads were found to be a barrier to satisfaction with the use of bicycle roads. Moderate slope, enough bicycle signs, and enough maintenance facilities around bike roads were found to be enabling factors in the perceived safety of the use of bicycle roads, whereas traffic on the side of the bicycle road was found to be a barrier to perceived safety. Based on these findings, we conclude that healthy cities should promote cycling behavior encouraging enabling factors and initiating attempts to improve the factors that act as barriers through urban planning.Increasing regular physical activity is essential to improving public health. Previous research has found that active transportation is associated with reduced cardiovascular risk factor [1,2]. Cycling is an example of active transportation that has the potential to contribute to an increase in people’s physical activity levels [3]. In addition to health benefits, cycling is an environmentally sustainable mode of transportation [4]. Cycling daily can also provide significant economic benefits by substantially reducing household expenditure on transportation and providing a cost-effective method of exercise [5,6]. Planning for sustainable and healthy cities that include cycling has become increasingly important in the era of sustainable development.A study from the Netherlands showed that natural and built environmental characteristics contribute to cycling duration, as well as the differential effect of environmental characteristics on cycling duration by municipality size [7]. Previous studies have shown that cycling behavior, including duration and frequency, differ based on people’s socio-demographic characteristics such as age, sex, and education level [8,9,10,11,12], and that cycling behavior also differs based on geographic variability [13,14].Large variations exist in the use of bicycles between countries. Additionally, bicycle use varies between areas and municipalities within a country based on the geographical territory and built environment [7,15]. In 1995, South Korea enacted the ‘Promotion of the Use of Bicycles Act’ and has been pursuing a policy of steadily improving bicycle use [16]. In 2009, the South Korean government devised a national bike path master plan which included an infrastructure expansion scheme and covered a wide range of relevant topics such as educational programmes, publicity measures, and guidelines for building and managing bike paths. The expansion of cycling infrastructure has led to the number of bicycle users in South Korea steadily increasing, although most of them are leisure-oriented cyclists [17]. In 2016, in South Korea, the daily bicycling rate was 8.3%, and the rate of bicycling more than once a week was 25.6%. In Seoul City, the rate of daily bicycling was 9.3%, and the rate of bicycling more than once a week was 33.3%. This is higher than the overall average for South Korea [18]. According to another report, among the 23.8% of bicycle users in Seoul in 2016, 4.6% were used for transportation, and 19.2% were used for leisure [19]. To promote active transportation, Seoul City has built a total of 860.570 km of public bicycle roads with different associated facilities [19]. Many studies have been conducted on physical activity globally; however, little research has focused specifically on cycling. Cycling as a sustainable mode of transportation may have great potential in South Korea for dealing with non-communicable diseases. Thus, it is important to understand the environmental and socio-demographic factors that influence cycling behavior and the policy implications for the promotion of enabling environments to improve bicycling in metropolitan Seoul and other similar cities. Therefore, the current research aimed to uncover the barriers and enabling factors influencing satisfaction and safety perception with the use of bicycle roads in the northwestern part of metropolitan Seoul, South Korea.A cross-sectional study was conducted in 2018; data was collected from 193 adults and youths in Seoul metropolitan, South Korea. The information was collected on 2 and 3 June 2018, using a survey. Data from three individuals were excluded due to incomplete information provided for approximately 50% of the variables at the stage of analysis, resulting in a final sample size of n = 190. The study was conducted in the north-western part of the Seoul metropolitan area in three areas: Eunpyeong-gu, Mapo-gu, and Seodaemun-gu.The sample size was calculated using the G-Power program. The recommended total sample size was 177 with actual power estimated at 0.80. The information was collected from 193 individuals; and 190 participants were included in the analysis for the study.A self-reported questionnaire was provided to the participants to be filled by themselves. The questionnaire examined the barriers and enabling factors in the Seoul metropolitan area and consisted of questions regarding socio-demographic factors, cycling behavior, satisfaction, and safety perception in the city.Statistical Package for Social Science (SPSS) version 24.0 (SPSS Inc., Chicago, IL, USA) was used to analyze the data. Data were summarized using descriptive statistics; and a chi-square test was used to determine the association between type of cyclist and socio-demographic and influential factors. Multivariate regression models were computed with significance level of α = 0.05 to determine which factors influenced satisfaction and perceived safety with the use of a bicycle road among cyclists. Satisfaction with the bicycle road and perceived safety were measured using a continuous scale. Out of the 17 factors, two factors were excluded from the regression analysis because VIF values were > 3.To measure participants’ satisfaction with the use of bicycle roads, the following question was asked: ‘Are you generally satisfied with the use of bicycle roads in the northwestern part of Seoul (Eunpyeong-gu, Mapo-gu, and Seodaemun-gu)?’ Participants responded by selecting one of the following responses: very dissatisfied, dissatisfied, usual, satisfied, and very satisfied. To measure the perceived safety of using bicycle roads we asked: ‘How safe do you think it is to use a bicycle road?’ Participants responded by selecting one of the following responses: very low, low, usual, high, and very high. For both questions, the options corresponded to numbered options labelled one through five.Socio-demographic variables: Participants were asked to report on several socio-demographic variables, including sex, age, residence, type of residence, educational level, income level, number of bicycles at home, number of cars, and occupation.Cycling behavior: To determine what type of cyclist the participant was, and to determine the average amount of cycling per month, they were asked: ‘How often did you use your bicycle during the past month?’ The type of cyclist was categorized by how often the respondent rode a bicycle during the last month. Respondents who said, ‘I rarely ride a bicycle in the last month’, ‘very occasionally’, and ‘sometimes’ were grouped as a non-cyclist/irregular cyclist. Respondents who reported ‘frequently’ or ‘very often’ were categorized as a regular cyclist. Participants were also asked: ‘For how many minutes do you normally use a bicycle?’ to measure the average time spent cycling.Barriers and enabling factors: The following questions were asked to assess what barriers and enabling factors contributed to participants cycling habits as shown in Table 1. All factors were framed with the overarching question ‘How much do the following items affect you when you use your bicycle?’ Participants responded using a 5-point Likert-type scale ranging from 1 (very little effect) to 5 (very influential effect).The Institutional Review Board (IRB) provided approval for this study (IRB: 1041849-201806-SB-053-02). Informed consent was obtained from each respondent prior to data collection. The objective of the survey and research was made clear to participants before data was collected.Of the total respondents, 73.2% were males, 18.9% were youths under the age of 18, and 8.9% were above older than 65. Of the total respondents, 34.2% reported that they were regular cyclists and the rest of the respondents cycled rarely, occasionally, or sometimes within the last one month before taking the survey. On average, participants reported cycling 9 days (±8.7 days) in the month preceding the study, with 88.7 (±82.7 min) minutes per day being the average time spent cycling.In regard to satisfaction with the use of cycle roads in the metropolitan area, 26.3% of respondents were very dissatisfied or dissatisfied, 23.7% were satisfied, and 10.5% were very satisfied. Regarding perceptions of safety, 27.4% of respondents reported high levels of safety, and 3.2% reported very high levels of safety in using the bicycle roads. However, 7.4% and 20.5% mentioned feeling very low and low levels of safety when using the bicycle roads in the study area (Table 2).A total of 17 items were given to the respondents to rate and report their opinions on the factor affecting bicycling riding in the Seoul metropolitan area. The factors were presented based on their mean rank (factors are rank ordered based on their means). Conflict with pedestrians on the bicycle side of the road ranked as the top concern of cyclists, followed by the level of traffic on the bicycle road (Table 3).Both gender and age were significantly associated with the type of cyclist (i.e., regular, non-cyclist/irregular) which was based on cycling frequency over one month. However, the study area, household income level, and the number of cars available at home were not related with being an irregular or regular bicycle rider among study respondents. Regarding the various uses of cycling, among those who used it to commute, 75% used it regularly, while only 31.5% cycled regularly among those who used it for hobby and leisure time. There was a significant association between the purposes of cycling and the regularity of cycling. The number of bicycles available at home was also associated with the type of cyclist or the regularity with which individuals cycled over the period of one month (Table 4).Level of satisfaction was not significantly associated with the type of cyclists. There was a significant association between perceived safety with bicycle road use and regularity of cycling. The regular cycling rate was 28.3% among those who mentioned they felt themselves unsafe while cycling and it was 48.3% among those who reported they felt safe while cycling. Respondents mentioned that cycling was more common in places where more related facilities such as convenience stores, restaurants, restrooms, shelters, etc. were available. This shows availability of facilities was associated with the regularity of cycling in the study area. Respondents mentioned that road slope was a critical and influential factor for cycling. A significant association was found between road gradient and regularity of the cycling. Another important factor for cycling was road width, and there was a significant association between the regularity of the cycling and road width. A lot of traffic on a bicycle road was another significant factor influencing the regularity of cycling (Table 5).The regression analysis showed that 19% of the satisfaction with bicycle use is attributed to bicycle parking space. Another significant factor in bicycle use satisfaction is road gradient, for instance, one standard unit increase in the appropriateness of the road gradient accounted for a 24% increase in the satisfaction with bicycle use among the respondents. The current study found that 30% of bicycle use satisfaction was attributable to one standard unit increase or decrease in the sufficiency of bicycle signs. We also found that a one standard unit increase in the narrowness of the bicycle road corresponded to a 16% decrease in the satisfaction level. Additionally, one standard unit increase/decrease in the average time of bicycle use per day (in minutes) corresponded to a 23% change in the satisfaction level (Table 6).Perceived bicycle road use safety was found to be influenced by the appropriateness of the road gradient. One standard unit increase in the appropriateness of the road gradient corresponded to a 20% change in the perception of the safety level among respondents and a one unit change in the sufficiency of the bicycle signs accounted for 25.6% change in the level of safety perception. Additionally, a one standard unit change in the availability of maintenance facilities around the bike road correlated to a 25.6% change in the safety perception among the bicycle riders, while a one unit increase or decrease in traffic on the bicycle roadside corresponded to 21% fluctuation in the perception of safety (Table 6).This study revealed that on average the respondents engaged in cycling 9 days a month, and the average time spent cycling in a day was 88 minutes. Sex, age, purpose of bicycle use, perceived safety, availability of facilities, road gradient, road width, and traffic on the bicycle road were all associated with regular cycling.Sex was significantly associated with the frequency of cycling; specifically, females were less likely to cycle regularly. Similarly, a study conducted in Australia showed that men were more likely to cycle for recreation and transport than women; and men tended to cycle for longer time periods [10]. Similarly, age was also associated with the cycling frequency in the study. Among the socio-demographic variables, age and sex were the two important factors influencing cycling behavior. Encouraging women and adult population for cycling can be a way to overcome physical inactivity among them. Possession of bicycles at home was another factor which influenced the frequency of cycling among the respondents in this study. Similarly, Heesch et al. [4] reported that limited vehicle access was positively associated with utility cycling. Thus, it seems to be a better way to initiate efforts to improve access to cycles in the metropolitan area.In this study, the main purpose reported by participants for using their bicycle was as a hobby or for leisure time. This finding is supported by Shin et al. [17] in their report on Bicycle Transport Policy in Korea. In the current study there was a significant association between the purpose of bicycle use and type of cyclist based on the frequency of cycling, specifically, among those who used their bicycles for leisure and hobby, more than two-thirds were categorized as irregular users. The efforts to enhance bicycle use for leisure, as a hobby, and for cycling to work would be one of the most important public health measures for addressing the ever-increasing burden of non-communicable diseases in South Korea [20,21,22,23].The current research showed that the sufficiency of bicycle parking spaces was one of the significant factors affecting the satisfaction of cyclists. The study also showed that adequate bicycle signs were also a significant enabling factor influencing satisfaction among cyclists. Environmental factors such as moderate slopes were also found to be an enabling factor of bicycle road use satisfaction. A study conducted in Canada has also shown that the built environment and various spatial zones have a significant influence on healthy travel decisions [11]. The current study found that narrow bicycle roads were an important barrier to satisfaction with the use of bicycle roads. These findings are supported by a study conducted in Poland that reports the main perceived barrier to cycling was linked to a lack of good cycling infrastructure in the city [24]. A lack of bicycle-friendly design was found to be a considerable barrier to greater bicycle use in an Australian study [25]. These various findings illustrate the importance of the built environment in relation to cycling facilities and bicycle roads. To enhance metropolitan Seoul as a healthy city, city development policies and plans should consider the built environment and facilities that enables or hinder the cycling behavior of the population. In addition, it is clear that typical geographical factors such as gradients also have influence on cycling behavior. A systematic approach is recommended for urban planning to enhance health and sustainability through active transport, which promises to be a powerful strategy for improvements in population health on a permanent basis [26].For perceived safety, moderate slopes, enough bicycle signs, and enough maintenance facilities around bike roads were found to be enabling factors. At the same time, traffic on the bicycle roadside was found to be a significant barrier factor for perceived safety with the use of bicycle roads. A study from Poland also reported that the main perceived barriers to cycling were linked to feelings of insecurity related to the behavior of drivers, and to road maintenance during the winter [20]. In addition to this, Heesch et al. [4] found that perceived environmental factors (crime, nearby transport, and recreational destinations) were associated with utility cycling (p < 0.05). Similarly, the perception of safety was found to hinder bicycling in many areas of Australia [25]. Numerous previous studies have argued that it is necessary to separate bicycle roads from pedestrian roads and vehicle roads, and that related infrastructure should consider the matter when establishing new road or redeveloping the urban area [10,27,28].As the existing evidence supports the efforts to promote cycling as an important contributor for better population health [24], metropolitan Seoul may use cycling promotion as a strategy of population health addressing the barrier before mentioned. Cycling behavior has dual positive impacts on population health through both physical activity and eco-friendly transportation [1,2,3,4]. Based on these findings, the current study recommends improved policies and infrastructure improvements for bicycle-related facilities and transportation systems that foster feelings of safety among cyclists. Other than slopes of bicycle roads, which are determined by the geographical feature of the city, policy formulation and implementation are necessary to deal with the variables that affect the level of satisfaction among cyclists, including sufficient bicycle parking space, installed bicycle signs and other variables that affect perceptions of safety such as installed bicycle signs and sufficient maintenance facilities. As the study has been conducted in one metropolitan city in South Korea, this study has the limitations of being focused on that region alone. While there has been rapid innovation of the bicycle, including the e-bike, the study did not address this issue in the study. The study did not assess the types of bicycles being used. This study revealed that the average number of cycling days among respondents was 9 days based on cycling activity in the month prior to the survey, and 34.2% were categorized as regular cyclists. Sex, age, purpose of bicycle use, perceived safety, availability of facilities, road gradient, road width, and traffic on the bicycle road were all associated with the regularity of cycling among respondents. Multivariate regression analysis showed that sufficiency of bicycle parking space, moderate slopes, and enough bicycle signs were significant enabling factors, while narrow bicycle roads were perceived as a barrier to satisfaction in the use of bicycle roads. Moderate slopes, adequate bicycle signs, and enough maintenance facilities around bike roads were enabling factors, and traffic on the bicycle roadside was a barrier to the perceived safety of using the bicycle road. Based on these findings, concerned authorities should aim to maintain enabling factors while overcoming barriers to cycling and further encouraging cycling behavior in their cities. H.Y.K., B.S. and H.K.N. conceptualized the study. H.K.N. contributed to survey design and the data collection. B.S. analyzed the data and prepared the manuscript. W.Y. critically reviewed and revised the manuscript. All authors reviewed and finalized the present manuscript.No funding was provided for this study.We would like to expresses our sincere acknowledgement to the study participants of the study.The authors declare no conflict of interest. Factors that affect respondents’ use of bicycles road.Characteristics of the study population.Factors influencing bicycle road use in Seoul metropolitan.Association between the type of cyclist and socio-demographic factors.Association between the type of cyclist and influential factors for cycling.Regression analysis of the factors affecting satisfaction and safety perception with the use of bicycle roads in the northwestern part of Seoul.

Med-MDPI/ijerph_3/ijerph-16-05-00774.txt

@@ -0,0 +1 @@
+Physical housing environment and living arrangements are significant determinants of health, particularly in developing countries, although results are mixed. We conducted this study to examine the gender differences in geriatric depressive symptoms in rural China, and further explored the influence of housing environments and living arrangements on depressive symptoms. The data used for this study were from the third wave of the nationally representative China Health and Retirement Longitudinal Study (CHARLS) survey in 2015; a total of 2056 females and 2529 males were included in this study. According to the analysis findings, 46.15% of the respondents had depressive symptoms based on the CES-D, with a statistically significant gender difference of 54.32% in females and 39.50% in males. Logistic Regression findings identified that with regard to the items of physical housing environments, toilets without seats (OR = 1.349) and the unavailability of bathing facilities (OR = 1.469) were statistically associated with depressive symptoms among male participants, whereas for female participants the use of polluting fuels (OR = 1.248) and living arrangements (i.e., living with children, OR = 1.430) was statistically associated with depressive symptoms. Statistically significant gender differences were found for having shower or bath facilities and our findings underscored that physical housing environments and living arrangements were associated with depressive symptoms for both genders. Moreover, the study revealed that a slight gender difference exists in terms of geriatric depression in rural China. Females are more likely to become depressed than their male counterparts with the same characteristics.Ageing is one of the most conspicuous phenomena in the contemporary world. Between 2015 and 2050, the proportion of the world’s adults that is elderly is estimated to almost double from about 12% to 22% [1]. In absolute terms, there will be an expected increase from 900 million to 2 billion people over the age of 60 [2]. While ageing is a global phenomenon, it is progressing fastest in developing countries, especially in China. In 2016, the Chinese population aged 60 years and over was over 230 million, which will further increase to 480 million by 2050 [3].Older adults face special physical and mental health challenges that need to be recognized. According to a World Health Organization report published in 2017, over 20% of older adults suffer from a mental or neurological disorder (excluding headache disorders), and 6.6% of all disability (disability-adjusted life years—DALYs) among people over 60 years was attributed to mental and neurological disorders. Depression, one of the most common mental disorders, affects approximately 7% of the world’s older population and comprises a substantial proportion of the global burden of diseases [4]. Research on depression in the elderly began in the 1980s and has emerged as a growing field since the 1990s.The gender differences in the occurrence, development, and severity of geriatric depression have been well documented [5,6,7]. The higher prevalence of depression among female older adults has long been recognized. For example, previous studies suggested that female older adults are almost 1.3 to 2 times more likely than male older adults to meet the criteria for depressive symptoms [8,9]. However, the potential factors that lead to this gender difference vary from study to study and from population to population. Possible explanations of this gender difference in depression are complex social factors, such as social role and status, social support and participation, in addition to socioeconomic factors and health status [10,11]. Female older adults tend to have lower education levels and income than their male counterparts, and because of the difference in life expectancy they are more likely to lose their partner, which is an important risk factor for depressive symptoms.Housing environment, as an important social determinant of health, has a well-documented association with depressive symptoms [12,13,14]. Housing environments contain physical, social, and psychological attributes. For instance, the physical environment may encompass tangible and observable attributes such as physical structure, design, housing facility, etc. [15,16,17]. The social and psychological factors encompass the living arrangement and the experience of community living such as neighborhood environments, forging relationships with neighbors, security, and belongingness. The association between housing environments and depressive symptoms has been investigated in several studies, particularly in urban areas [18,19]. Some previous studies have shown the relationship between urban housing environments and mental health [20]. However, research on housing environments and depression among the rural population is remarkably underdeveloped. Few studies have discussed in more detail the impact of rural housing environments on depressive symptoms among the elderly, especially in China [21]. In rural areas of China, poor physical housing environments may lead to a greater possibility of developing depressive symptoms. With ongoing urbanization, it is necessary to further focus on the relationship between rural housing environments and mental health, which has a significant effect on the construction of age-friendly environments and can further reduce the health inequality between the urban and rural population [22]. Living arrangements, an important component of social support that elderly people receive from their partners or family members, has commonly been viewed as a risk factor for depression in the elderly [23]. Previous research has shown that living arrangements are an independent risk factor contributing to depression [24,25]. Gender differences in the association between living arrangements and depression have also been demonstrated. Previous studies reported that older adults living alone were more depressed compared with those living with other family members [26,27]. Considering cultural factors, it has been suggested that the relationship between living arrangement and depression may be more prevalent in certain ethnic groups, especially among the Chinese population. We conducted this study to examine the gender differences in geriatric depressive symptoms in rural China, and further explored the influence of housing environments and living arrangements on depressive symptoms. We believe our study will improve the understanding of the complex nature of the gender differences in geriatric depressive symptoms, and the association between gender, housing environments, living arrangement, and depressive symptoms among the rural elderly population in China.This study is based on the nationally representative China Health and Retirement Longitudinal Study (CHARLS) conducted by the China Centre for Economic Research of Peking University, which was designed after the Health and Retirement Study (HRS) in the USA as a broad purposed social science and health survey of the elderly in China. Face-to face interviews in respondents’ homes collected detailed information on their demographic characteristics, socioeconomic status, health-related behaviors and lifestyles, health status including health conditions, health insurance, and health services use [28]. More details on the CHARLS survey design are available from Zhao et al. Ethics approval for the baseline data collection was obtained from the Biomedical Ethics Review Committee of Peking University (IRB00001052-11015). Ethical approval for the use of CHARLS data was obtained from the University of Newcastle HREC (H-2015-0290). The data used for this study were from the third wave in 2015, and the inclusion criteria were: (1) age ≥ 60, (2) living in rural regions, (2) living at the survey site for at least six months, (3) being at home during the investigation period, and (4) able to participate in the study. Subjects with mental retardation and severe cognitive impairment had been excluded from the original research through a short screening form; only those respondents who needed little or no help with answering questions were allowed to advance. Thus, the number of respondents eligible for our analysis dropped to 4585 elderly.CHARLS utilized a multi-stage stratified probability-proportional-to-size sampling (PPS) technique to select participants. Face-to-face interviews were conducted to collect detailed information on demographic characteristics, socioeconomic status, health status including health conditions, health insurance, and health service use. The pilot survey was carried out in two provinces (Gansu and Zhejiang) in 2008, and was subsequently expanded to form the national baseline survey fielded in 2011. In the first wave 17,708 participants were surveyed; a total of 18,246 respondents aged 45 plus were followed up and surveyed in the second wave. In 2015, the number of participants increased to 21,095. CHARLS survey design has been described in more detail elsewhere [28,29].The 10-item Center for Epidemiologic Studies Depression Scale (CES-D) was used to screen depressive symptoms. The time frame for the CES-D questions referred to the week prior to the interview. Each item was rated on a 4-point Likert scale, with answers varying from ‘rarely or none of the time (0–1 day)’ to ‘most or all of the time (5–7 days)’. The total scores ranged from 0 to 30, with a lower score indicating a lower level of depressive symptoms. CES-D has shown good validity and reliability in the Chinese population. A previous validation study among older adults found that a cutoff point of 10 provides the optimal threshold to identify clinically significant depressive symptoms. Thus, a cutoff point of 10 was used in this study to generate the dichotomous depressive symptoms variable. Participants who scored lower than 11 were classified as non-depressed. The CESD-10 has been described in detail elsewhere [29].Two aspects of housing environment were assessed: the physical environment and the living arrangements.Five variables were used to measure physical environment: housing materials were dichotomized as ‘improved material’ (concrete and steel/bricks and wood) and ‘unimproved material’ (adobe, wood, cave dwelling, Mongolian yurt/woolen felt, stone); Cooking fuel was categorized as Clean fuel (2 = natural gas, 3 = marsh gas, 4 = liquefied petroleum gas, 5 = electric), Polluting fuel (1 = coal, 6 = crop residue/wood burning), and others (7 = other). Housing facility was assessed by the availability of a toilet, running water, and a shower or bath facility. These facilities were all categorized as “0 = yes” (available) and “1 = no” (not available), while toilets were categorized as “0 = toilet with seat” or “1 = toilet without seat.”The instrument of living arrangement contained three items: living with spouse (yes or no), living with children (yes or no), and living alone (yes or no).In this study, health status was assessed by “number of chronic diseases,” “disability status,” and “activities of daily living (ADL) limitations.”NCDS was measured as the cumulative number of diagnosed chronic conditions by a physician (hypertension, dyslipidemia, diabetes, cancer, chronic lung diseases, liver or gallbladder disease, heart disease, stroke, kidney disease, stomach or other digestive disease, emotional, nervous, or psychiatric problems, memory-related disease, rheumatism/arthritis, asthma) on three scales: 0, 1–2, ≥3 [22].Disability status was measured by asking participants: “Do you have one of the following disabilities?” Responses were categorized as “yes” when they had any disability of physical disabilities, brain damage/mental retardation, vision problem, hearing problem, or speech impediment.ADL limitations indicate any self-reported difficulty in any of the following six activities of daily living: eating, dressing, getting into or out of bed, using the toilet, bathing/showering, or controlling urination and defecation. The four response options were: 1 = “I don’t have any difficulty,” 2 = “I have difficulty but can still do it,” 3 = “I have difficulty and need help,” 4 = “I cannot do it.” It was dichotomized as “unimpaired” vs. “impaired.” “ADL-impaired” was defined as “have difficulty and need help” or “cannot do it” in any ADL item.Educational level and personal annual income were used to determine SES. Education levels were classified as “Primary school and below” or “High school and above.” Personal annual income was categorized to “≤1477 dollars,” “1477–2954 dollars,” or “≥2954 dollars.”Socio-demographic variables included age (year), gender (female, male), and marital status (married, unmarried).All statistical analyses and tests were conducted using Stata version 13 (Stata, College Station, TX, USA) [28]. Data are presented with percentages and proportions for categorical values. First, we conducted a χ2 test to assess the statistical differences between groups. Then, we conducted a binary logistic regression to assess the associations of depressive symptoms. Finally, we performed a t-test analysis to assess the interaction effects of gender on these associations. In all analyses, the criterion for statistical significance was p < 0.05.Table 1 reports descriptive statistics for the participants by genders. Out of 4585 rural participants 44.84% are females and 55.16% are males. Most respondents had lower level of education (elementary school and below), with significantly more men in the low-level group. Approximately 46.15% of the respondents had depressive symptoms based on the CES-D, with a statistically significant gender difference of 54.32% in females and 39.50% in males.Table 1 also shows depressive symptoms for both female and male participants were significantly associated with age, annual income, number of NCDS, disabilities status, and ADL limitations. The depressed participants tended to be older and have a lower personal income for both genders. Likewise, participants with depressive symptoms reported more chronic diseases, a greater proportion of physical disability, and ADL limitations. Among female older adults, the depressed were more often unmarried, whereas the depressed male participants tended to have a lower education level than those who were not reportedly depressed (Table 1).Table 2 shows the prevalence of depressive symptoms by genders according to physical housing environments and living arrangements. Depressive symptoms for both genders were only significantly associated with three items of physical housing environments (the type of cooking fuel and toilet; the availability of a shower or bath). In addition, depressive symptoms were significantly associated with living arrangements among female participants, whereas depressed male participants tended to have a lower education level than those who were not reportedly depressed.The final logistical regression results are shown in Table 3. Regression findings suggest lower personal income (≤10000: OR = 2.448, female; OR = 1.432, male), disability (OR = 1.318, female; OR = 1.786, male), ADL dysfunction (OR = 1.539, female; OR = 1.297, male), and more than two types of chronic diseases (OR = 2.448, female; OR = 1.432, male) are significantly and positively associated with geriatric depressive symptoms for both female and male participants. With regard to physical housing environments, a toilet without a seat (OR = 1.349) and a lack of bathing facilities (OR = 1.469) are statistically associated with depressive symptoms among female participants, whereas depressed male participants tend to use polluting fuel (OR = 1.248). Remarkably, the living arrangement (living with children, OR = 1.430) is only statistically associated with depressive symptoms among female participants.Statistically significant gender differences in these associations were found for having a physical disability and no shower or bath. Test statistics of the interactions for gender are shown in Table 4.Our study examined the gender differences in geriatric depressive symptoms in rural China, and further explored the influence of housing environments and living arrangements on depressive symptoms. In this study, a high prevalence of depressive symptoms was found in both genders, with a preponderance in females, which is in line with previous studies. Females were found to be more often depressed than males. Around 46.15% of the respondents had depressive symptoms according to the CES-D, with a statistically significant gender difference of 54.32% in men and 39.50% in women. Previous epidemiologic studies suggested that female older adults are 1.3–2 times more likely than male older adults to meet the screening criteria for depressive symptoms [8,10]. Numerous demographical, biological, social and psychological explanations for these gender differences have been proposed. Psychological factors include personality, coping styles and cognitive abilities. Previous studies reported that males report fewer psychological symptoms leading to difficulties in detecting depression. Females are also approximately twice as likely as males to be diagnosed with generalized anxiety disorder [11,30] and score more highly on self-reported scales measuring anxiety [31].In our study, using a polluting cooking fuel is statistically associated with depressive symptoms among male participants. Indoor air pollution (IAP), one of the major public health concerns in low- and middle-income countries, is mainly caused by the use of polluting cooking fuels such as coal, charcoal, crop residue, and wood burning. According to a WHO report in 2016, nearly 3 billion (40%) of the world’s people rely on solid fuels, including coal and biomass, for domestic cooking [32,33,34,35]. Evidence shows that IAP is associated with poor physical health. Prior researchers have identified IAP as the most important environmental risk factor globally associated with adverse health effects [36,37,38,39,40,41,42,43,44,45,46,47,48]. However, there is little empirical evidence of the relationship between IAP and depression based on nationally representative data, especially in developing countries [39,40]. In general, the negative effects of polluting cooking fuel on depressive symptoms have not been given enough attention because they are largely dependent on the population group [41,42].We found that using a toilet without a seat is statistically associated with depressive symptoms among female participants. However, very little research has explored the relationship between toilet type and depressive symptoms. Research on health, especially on microbiological pollution, has demonstrated that the species of microbes on squat pans are roughly the same as on toilets, but the number alone is much higher [43]. Some other previous investigations have suggested that the use of squat pans can over time cause anal fissure to a certain extent, raise the risk of dizziness, and even lead to falls [44,45]. Perhaps we can deduce from the existing research that the existence of chronic diseases resulting from using a toilet without a seat may lead to a greater possibility of depression among the elderly.In our study, we found that having no bathing facility is statistically associated with depressive symptoms among female participants. In terms of the relationship between shower/bath facilities and depressive symptoms, we found only a few articles, in the literature study, that involved the relationship between bath facilities and health outcome. A randomized clinical pilot trial conducted in Germany showed that hypothermic baths (HTB) do have generalized efficacy in depressed patients. The results of a non-controlled HTB study aimed at 20 depressive patients, also conducted in Germany, showed an improvement after hypothermic baths [46,47]. Furthermore, HTB, especially before bedtime, improved sleep quality in healthy, insomniac people and elderly patients with depression and vascular dementia. Whole-body hyperthermia, according to a further non-controlled study, showed a significant reduction in CES-D among 16 depressive patients. Our research further corroborated the previous research. The prior research suggests that HTB have antidepressant effects that are mediated through changes in circadian functioning and temperature physiology, although the underlying mechanisms remain unclear [48].Furthermore, the type of toilet and cooking fuel, and access to bathing facilities, as important symptoms of poverty, have associations with the process of developing depressive symptoms and thus reflect the impact of socioeconomic status (SES) on an individual’s mental health. It has been suggested that poverty/poor SES may lead to poor access to mental health services, and further affect the diagnosis and treatment of depression, as it is very difficult for low-income populations to meet regular health care needs and be screened for depressive symptoms [49,50,51,52]. How to approach these problems is outside the scope of this paper, but our findings highlight the importance of poverty or poor SES in the development of depressive symptoms in rural China.In the face of psychosocial housing environments, living with children was significantly associated with depression only in females, although these gender differences were not statistically significant. This is consistent with previous studies that social support from children might be recognized as more valuable than from other family members, especially for females [10,25]. Living arrangement, an important component of the social support that elderly people receive from family members, has commonly been viewed as a risk factor for depression in the elderly. Adequate social support and participation are generally regarded as an important protective factor against depressive symptoms in the elderly [50]. We surmise that the relationship between traditional culture and mental health may be more prevalent among the Chinese population. Chinese proverbs include “Raise children to provide against old age” and Confucius said, “While one’s parents are alive, one should not go far away from his or her parents.” This may explain the protective effect on older adults in the context of Chinese culture. One important explanation might be that adult children are important sources of psychological and financial support to elderly adults in China [17]. Previous research showed that more than 60% of older adults’ money comes from their children [51,52]. Moreover, the concept of filial piety from Chinese traditional culture is still critical to the social support system of modern Chinese families: children are still the main source of social support to their older parents. Not surprisingly, then, living arrangement might have a greater impact on mental health in aging Chinese population due to the cultural ideal of social support in later life.Our study adds to the existing knowledge highlighting the role of physical and psychological housing environments in exploring the development of depressive symptoms in rural China. Furthermore, the use of nationally representative data from all of China contributes to the implementation of cross-country comparisons and to the collection of much more information about the effect of physical housing environments and living arrangements on depressive symptoms. The main limitations are a cross-sectional study design and the use of self-reported data. Caution must be taken when generalizing the findings of this study to the rest of the population in China since it is a limited sample. There was no further clinical diagnosis or treatment for the participants who met the screening criteria according to CES-D, in the original research.Our findings underscore the importance of strong political commitment, inter-sectorial coordination, and adequate financing in order to prevent the development of depressive symptoms among the elderly in rural China. Overall, our findings reveal that more attention should be paid to the relationship between gender, housing environment, living arrangement, and depressive symptoms among the rural elderly population in developing countries.Statistically significant gender differences were found for having shower or bath facilities and our findings underscored that physical housing environments and living arrangements were associated with depressive symptoms for both genders. Moreover, the study revealed that a slight gender difference exists in terms of geriatric depression in rural China. Females are more likely to become depressed than their male counterparts with the same characteristics.X.M. conceived and designed the study; M.F. analyzed the data; J.C. and L.G. contributed materials/analysis tools; M.F. wrote the paper.This research was funded by China Medical Board (CMB): Public Health Education Reformation in China, grant number04-804.We would like to acknowledge the China Health and Retirement Longitudinal Study (CHARLS) team for providing data.The authors declare no conflict of interest.Distribution and prevalence of depressive symptoms by gender according to participant characteristics.Distribution and prevalence of depressive symptoms by gender according to physical housing environments and living arrangements.Logistics regression of physical housing environments and living arrangements associated with depressive symptoms by gender.Regression coefficients across models: interaction effects.

Med-MDPI/ijerph_3/ijerph-16-05-00775.txt

@@ -0,0 +1 @@
+Deceased.The role of social support in assisting youth in developed countries cope with their HIV diagnosis has been examined through a vast body of research; yet, there remains a gap in research around the effects of social support among youth living in sub-Saharan African countries including Kenya. This study aimed to examine the role of social support among Kenyan youth living with HIV, specifically with regard to the variations in influences of this social support. We conducted semi-structured focus group discussions with youth (ages 18 to 27) living in the informal urban settlement of Kibera in Nairobi, Kenya (n = 53). Data analysis followed a phenomenological inquiry framework, and seven major categories of perceived social support influences were identified: (1) linkage to services, (2) antiretroviral (ARV) adherence, (3) self-acceptance of HIV status, (4) healthy and positive living, (5) understanding of what it means to be living with HIV, (6) HIV status disclosure, and (7) family and occupational strengthening. The findings from this study suggest that Kenyan youth living with HIV can benefit from social support in a multitude of ways and can occur across several socio-ecological levels. Future research should further examine these influences, specifically regarding intervention development across socio-ecological levels.The global burden of HIV remains high: there were 36.9 million people living with human immunodeficiency virus (HIV) in 2017 [1]. Kenya has been more affected by HIV than have many other parts of the world, with prevalence rates within the top fifteen highest country-level rates; UNAIDS estimates that, in 2017, 1.5 million Kenyans were living with HIV [2,3]. That same year, HIV was the leading cause of disability adjusted life years (DALYs) and of death in Kenya [4]. At 6.1%, the adult HIV prevalence rate in Nairobi is higher than Kenya’s country-level rate of 4.9% and considerably higher than the global rate of 0.8% [1,5].Prevalence rates in Kenyan informal urban settlement areas, such as Kibera—Nairobi’s most populated urban informal settlement—are estimated to be even higher, at 12% [6]. Several factors, including poverty, unemployment, and substance abuse, make living in Kibera a high-risk environment for transmission of HIV [7]. Women may be at even higher risk; as a study by Amuyunzu-Nyamongo et al. (2017) found, women living with HIV in Nairobi informal settlements had commonly taken to survival strategies such as commercial sex work, which placed them at significant risk for HIV reinfection and infection with sexually transmitted infections (STIs) [8].HIV continues to disproportionately impact Kenyan youth, with incidence rates highest among those nearing adulthood [9]. A disproportionate share of incident adult HIV infections in Nairobi County occur among youth aged 15–24: in 2016, this population accounted for 46% of all new HIV infections in Nairobi County yet just 18% of the Nairobi County population [10]. In 2017, HIV was the leading cause of DALYs and of death among Kenyans aged 10–24 years old [4]. Combined with the general HIV rates in informal urban settlements, youth living in such areas represent a group highly vulnerable to the negative effects of HIV.In the initial period after receiving an HIV diagnosis, youth can undergo particularly high amounts of emotional stress. Those who test positive for HIV are not only burdened with adjusting to living with a highly stigmatized health condition that requires lifelong medical management but are also often struggling with typical adolescent developmental challenges [11,12]. This stress can primarily be attributed to adjusting to their diagnosis and disclosing their HIV status to others, which is often compounded by lack of accurate information and fears related to HIV medications and development of symptoms [11].As a result, youth living with HIV often experience reduced quality of life, psychological distress, social impairment, increased risk-taking behaviors, poorer antiretroviral (ARV) medication adherence, and increased exposure to additional psychosocial vulnerabilities [11,12,13]. Furthermore, the stigma associated with HIV may serve as a barrier to youth acquiring or maintaining social support [14]. As such, there is need for new, targeted interventions specifically for youth in the initial period after receiving an HIV positive diagnosis.Social support, based on pivotal work by House (1981), comprises providing emotional, instrumental, informational, or appraisal aid and assistance within social relationships [15]. Numerous studies from developed countries have found that social support positively impacts health and wellbeing, including with improving overall health, improving stress response, and improving coping related to HIV management [16,17,18]. Though diverse in their specific objectives and sample characteristics, there have been several qualitative studies among people living with HIV in sub-Saharan Africa that describe positive impacts from social support [19,20,21,22,23,24]. However, there remains a gap in such research among youth living in a large informal urban settlement like Kibera, as well as research specifically targeting the initial stage after being diagnosed with HIV. Moreover, none of these studies examined as deeply the diverse influences of social support received across informal and formal sources without some sort of scope or other limitation.Youth remain largely neglected when it comes to targeted service provision in Kenya [25,26,27]. To develop interventions that successfully address the unique barriers Kenyan youth may face in dealing with a recent HIV diagnosis, it is imperative to understand their lived experiences. Therefore, the purpose of this study is to better understand the diverse perceived influences of social support among youth newly diagnosed with HIV (i.e., within the past two years) living in urban informal settlements in Kenya, as it impacts their HIV management and other aspects of their lives.To understand the role of social support within the experiences of Kenyan youth newly diagnosed with HIV, six focus group discussions were conducted with youth residents of Kibera, the informal urban settlement in Nairobi. Qualitative data was collected through focus group discussion (FGD), as the aim of the study was to gain a phenomenological understanding of the lived experiences of youth newly diagnosed with HIV, and the group dynamic could provide potentially richer data than could individual interviews with regard to the phenomena of interest [28]. In a qualitative study by Hosek et al. (2008) in the United States with youth newly diagnosed with HIV, data were collected through both FGD and in-depth interview (IDI), and the researchers found that the group dynamics in the FGD helped participants to build off of each other’s comments and aided recall by triggering thoughts and memories that assisted participants in sharing greater detail [11].Given that the data presented in this paper are part of a larger study investigating risk and resiliency, Wallander and Varni’s (1992) Disability-Stress-Coping Model served as an initial framework for the focus group discussions to determine both the key sources of stress (i.e., psychological stress, functional independence, disease parameters) and coping mechanisms (i.e., intrapersonal competence, socio-ecological support, coping strategies) following each participant’s HIV diagnosis [29]. Semi-structured interview guides, based on a phenomenological framework, were developed with open-ended questions to investigate these key areas.‘Newly diagnosed’ was defined as having received an HIV diagnosis within the past two years and focused on the first six months following diagnosis. The time period was selected in consultation with HIV service providers and youth living with HIV in Kenya regarding how to capture initial challenges and opportunities the youth faced with managing their HIV. A chain-referral sampling technique was used to recruit a purposive sample of youth through existing peer networks, as peer-referrals allowed access to the predominantly-hidden group of youth living with HIV. Participants represented eight ethnic groups, as efforts were made to recruit youth from diverse ethnic groups. Focus group discussions held in Kibera lasted between two to three hours, with a total of 53 participants (aged 18–27) from ten villages in Kibera. Though youth is often defined as those aged 15–24, Kenyan policy often refers to youth as 15–30 [30]. The age range of study participants reflects the culturally appropriate definition of youth.Recruited individuals needed to meet the following inclusion criteria to participate: (1) be aged 18–27 years old, (2) have been infected with HIV through a behavioural mode of transmission, (3) have an initial HIV diagnosis date that falls within the past 24 months, (4) be willing to participate in focus group discussions, and (5) provide informed consent for study participation. Exclusionary criteria for participation included those individuals who (1) acquired HIV through perinatal infection, (2) demonstrated or reported any presence of serious psychiatric symptoms, (3) were visibly distraught, or (4) were intoxicated or under the influence of alcohol or other substances at the time of study enrolment. None of the originally recruited participants were excluded based on these criteria.Participants completed a demographic information survey prior to participation in the focus group. Bilingual local research team members facilitated focus groups in both English and Kiswahili, based on participant preference. Focus groups were gender-specific in order to promote more open dialog that was less restricted by gender norms and related social power dynamics, particularly given the stigma around the discussion topic. Each gender-specific focus group (three groups of women and three groups of men) consisted of 8–10 participants. A member of the local research team transcribed/translated into English the digitally-recorded focus group discussions.The local research team had prolonged engagement with the participant population through various programs in the field in order to assess possible sources of data distortion and to identify saliencies in the data. Through a continuous informal testing of information, participants’ reactions to data as they were being collected were solicited in order to assess the accuracy of the investigators’ reconstructions of what was being said. Our research team included individuals with various disciplinary backgrounds (e.g., Nursing, Public Health, Clinical Psychology, Community Psychology, and Social Work), as well as different countries of origin (Kenya and United States). These various disciplinary and geographic backgrounds allowed for investigator triangulation throughout the study.We conducted data analysis using a phenomenological inquiry framework. The composite descriptions discussed in this article describe both what and how phenomena were experienced by respondents [28,31]. We identified patterns related to the primary concepts being explored, specifically regarding the perceived influences of receiving social support by Kenyan youth newly diagnosed with HIV, and developed inductive codes through examining the data. Social support was operationalized as aid and assistance provided within social relationships, in accordance with the fundamental work by House [15]. We added subsequent codes during the iterative analysis process.We added supplemental content codes to the code list based on the lived experiences described by participants after reading and reviewing all transcripts and developed a codebook that included operational definitions of all codes. Pattern codes were then assigned to link related concepts together, after which we established consistent patterns in meaning, concepts, and themes across all focus group transcripts [28,32]. All codes were developed and independently assigned by the first three authors, who resolved any inconsistencies in data interpretation and pattern identification through discussions leading to unanimous agreement. Findings were shared during data analysis with professional peers who were not actively engaged in the study in order to receive external feedback on the analysis process, as well as to test and clarify interpretations of the data [33,34]. This article focuses on the identified theme of social support influences that emerged during the focus group discussions.Institutional review board approval was obtained from all participating institutions for the research protocol (GH071505PSY). Written informed consent was obtained from each participant prior to their involvement in a focus group. Youth received reimbursement for any transportation costs and a meal to compensate them for their time. Throughout the data collection and analysis process, the research team protected participant confidentiality. To protect patient identities, transcripts were de-identified for analysis and pseudonyms were assigned to participants.A total of 53 youth (49.1% women; 50.9% men) participated in this study’s focus group discussions. Table 1 summarizes demographic characteristics of study participants. Participants were aged 18 to 27 years old at the time of the study, and the majority were unmarried.Study participants reported having received social support within their first six months after diagnosis, which they perceived as having diverse influences. This social support fell into four major categories, aligning with the four major types of supportive actions categorized by House: instrumental support, defined as receiving something tangible that helps remove barriers youth may encounter; informational support, defined as receiving advice, factual knowledge, or suggestions that enables the adolescent to cope with adversity; emotional support, defined as receiving comfort, empathetic listening, or consolation that helps the adolescent cope with emotional distress; and appraisal support, defined as receiving constructive feedback, assurance, or validation that enables the adolescent to better adjust to his or her situation [15].While several participants described times of feeling isolated, many reported positive experiences related to having received social support during the initial six months after their HIV diagnosis. These perceived social support influences fell into seven domains: (1) linkage to services, (2) ARV adherence, (3) self-acceptance of HIV status, (4) healthy and positive living, (5) understanding of what it means to be living with HIV, (6) HIV status disclosure, and (7) family and occupational strengthening. There were also some instances in which respondents described social support influences as positive in too general of a sense to ascribe to one of the previously outlined categories. These seven main perceived influences of social support indicate that social support could diversely benefit Kenyan youth newly diagnosed with HIV. Below, we describe the details of each thematic influence of social support, and sub-themes that form the larger influences of support. The speaker’s gender and age follow any text quoted from participants.Youth participants recalled that the social support they received in their first six months post-diagnosis had the perceived effect of linking them to HIV services, including integrated psychosocial and clinical care. The internalization of the diagnosis and its long-term import in terms of medication and regular association with the health facility was of importance here. Two major sub-themes emerged under this theme: (a) connecting with a health center, counsellor, or support group and (b) connecting with ARV medication.Participants described how support they received aided them in connecting with a health center, counsellor, or support group. The initial support paved the way for a more nuanced socio-emotional mental health follow up with a specialist. Anna recalled her disbelief that she could have a positive HIV diagnosis, which prevented her from accepting her status and taking steps to manage the disease. She eventually visited another doctor who went above and beyond that of his or her duties:‘I felt sick but, at the back of my mind, I knew I had been told that I am HIV positive. During this time, I decided to go to the doctor; lucky enough, I found a good doctor who asked why I had allowed the disease to attack me to this extend [sic]. I told the doctor that, still, I don’t believe [I] am sick. The doctor took me to another doctor who tested me but, still, the results were the same’.(25, woman)‘I felt sick but, at the back of my mind, I knew I had been told that I am HIV positive. During this time, I decided to go to the doctor; lucky enough, I found a good doctor who asked why I had allowed the disease to attack me to this extend [sic]. I told the doctor that, still, I don’t believe [I] am sick. The doctor took me to another doctor who tested me but, still, the results were the same’.A 25-year-old man, Thabiti, described how the social support he received connected him with ARV medication. Thabiti affirmed the challenges in obtaining ARVs due to inconsistent supply volumes and fatigue from waiting in the long lines at the ARV clinic, as others in the focus group had described. Thabiti’s parents enacted support by picking up the drugs for him. He was able to obtain his ARV medication due to his parents’ support.Youth reported that social support they received helped them adhere to their ARV regimens by addressing challenges associated with ARV initiation or continued adherence. This influence type does not involve physically connecting the youth with ARVs. Two sub-themes emerged: (a) addressing challenges to adherence and (b) building adherence skills.A few participants perceived that external support alleviated difficulties they faced in remembering when to take their medications. These difficulties were assuaged through receiving reminders, receiving information regarding the importance of never missing a dose, or becoming motivated to adhere through encouragement. When participants reported receiving motivational and supportive care, they perceived that their illness burden reduced. Hadiya had difficulty remembering when to take her ARV medications:‘I had problems the first month, but I had a doctor who cared for me. He reminded me when to take drugs for the first three month. We had agreed when I should take my medicine, so when it was time, he call [sic] or sent me a please call me, then I know it’s my time to take my ARVs. I continued like that until I can remember on my own’.(25, woman)‘I had problems the first month, but I had a doctor who cared for me. He reminded me when to take drugs for the first three month. We had agreed when I should take my medicine, so when it was time, he call [sic] or sent me a please call me, then I know it’s my time to take my ARVs. I continued like that until I can remember on my own’.Another sub-theme discussed involved social support the youth received resulting in them learning a new skill that made adhering to their prescribed drug regimens more feasible in the face of HIV-related stigma. Through a support group, Faith learned to store her pills safely without fear of accidental disclosure:‘Disposing of the empty bottles of drugs was a challenge initially. That time I would go with a polythene bag and put all the drugs there and leave the container in the hospital. I later came to learn it was not good. The moment I joined the support group, I gained the knowledge of removing the labels on the bottles and disposing became a little bit easily [sic]’.(25, woman)‘Disposing of the empty bottles of drugs was a challenge initially. That time I would go with a polythene bag and put all the drugs there and leave the container in the hospital. I later came to learn it was not good. The moment I joined the support group, I gained the knowledge of removing the labels on the bottles and disposing became a little bit easily [sic]’.Many participants discussed gaining self-acceptance of one’s HIV status due to having received social support. Prior to this self-acceptance, participants described experiencing self-stigma and developing internalizing tendencies that led to self-doubt and poor self-efficacy to manage their HIV and live healthily. This social support influence often involved someone directly telling them that they should accept their status or providing examples of other young people living with HIV. Fadiya explained how an interaction with a close friend helped change her perception of what it meant to be a person living with HIV: ‘I really had stress. After that, she called me and told me that she was also a victim, but when you look at her, you can’t believe it, she is healthy and big. At that moment, that is when I began to regain myself’.(24, woman)‘I really had stress. After that, she called me and told me that she was also a victim, but when you look at her, you can’t believe it, she is healthy and big. At that moment, that is when I began to regain myself’.Similarly, Inaya shared that her husband’s parents aided her in accepting her HIV status:‘My husband’s parents who consoled me that I was not the only one suffering, I should accept my situation and, indeed, I accepted’.(21, woman)‘My husband’s parents who consoled me that I was not the only one suffering, I should accept my situation and, indeed, I accepted’.The participants described that social support they received also had the perceived effect of helping them to live positively, outside of making healthy choices directly related to managing their HIV diagnosis. We identified two sub-themes under this influence: (a) improving physical health and (b) improving emotional/psychological wellbeing.Several youth participants perceived social support they received or desired as helping them with their opportunistic infections or with another physical ailment. For example, Thabiti, age 25, had an accident and desired that his parents provide him with social support by donating blood.Another sub-theme associated with positive living was an improvement in participants’ emotional or psychological wellbeing as they received support. Participants spoke about various ways in which their emotional/psychological wellbeing was bettered. This included building new skills, releasing stress, practicing religious faith, and connecting with others. Azizi described the support he received from his friends:‘Emotionally, I was so down; like, I was finished, my life was coming to an end. I started drinking heavily, spending a lot of money anyhow, I had not [sic] future […] I came to realize that am not the only one in this situation, so I got some friends whom we attended therapy with together [sic], and that helped me heal my emotional problems’.(23, man)‘Emotionally, I was so down; like, I was finished, my life was coming to an end. I started drinking heavily, spending a lot of money anyhow, I had not [sic] future […] I came to realize that am not the only one in this situation, so I got some friends whom we attended therapy with together [sic], and that helped me heal my emotional problems’.Many youth participants revealed that the social support they received altered their understanding of what it means to be living with HIV in two distinct ways: a) acknowledging they are not alone in facing their HIV diagnosis and b) believing it is possible for people living with HIV to live a long and healthy life.Some youth recalled feeling a strong sense of isolation after receiving a positive HIV diagnosis. Specifically, they believed that they were alone in their struggle. Social support had an impact on altering this belief. In these cases, the social support was commonly from other youth living with HIV. For Kwasi, it was his peers who facilitated his change in understanding:‘You find one youth who is really sick but the morale they have just gives you psychic [sic] also to continue with life. You say, for sure, if this guy has made it in life, why not me. So that counselling sessions we have together encourages you at least to know that you are not alone’.(18, man)‘You find one youth who is really sick but the morale they have just gives you psychic [sic] also to continue with life. You say, for sure, if this guy has made it in life, why not me. So that counselling sessions we have together encourages you at least to know that you are not alone’.When first learning of one’s HIV diagnosis, some youth believed HIV was a death sentence. In these instances, youth often indicated feeling relieved, yet surprised, when they encountered another person living with HIV who appeared strong and healthy. Such encounters helped youth to better understand that one can be in good health while living with HIV. For Kafil, counselling helped him realize this:‘With time, I accepted myself. I went through counselling where I was informed that this disease was like any other disease in human life as long as you adhere to the rules given to make the disease sustainable’.(27, man)‘With time, I accepted myself. I went through counselling where I was informed that this disease was like any other disease in human life as long as you adhere to the rules given to make the disease sustainable’.Youth described how the social support they received allowed them to disclose their HIV status to someone, often a family member or friend. Two sub-themes were identified under this theme: (a) building confidence and/or skills to disclose and (b) receiving direct assistance with disclosure.For some participants, the social support they received provided them with the confidence and/or skills needed to disclose his or her status to another. This could involve providing encouragement, reassurance, a safe space, or skills to use in order to disclose. Kibibi described:‘Initially, I feared to tell him for a period of six months […] Through the teaching I received from the clinic, one day I decided I will tell [my husband] even if it means him sending me away’.(21, woman)‘Initially, I feared to tell him for a period of six months […] Through the teaching I received from the clinic, one day I decided I will tell [my husband] even if it means him sending me away’.Several of the youth participants shared experiences of another individual providing them with direct assistance around disclosure. A friend helped Inaya when the friend disclosed Inaya’s status to her husband’s parents, who in return, provided her with additional social support that helped her to accept her HIV status and adhere to her ARV medication:‘I don’t have parents except my husband’s parents […] I didn’t tell the parents of my husband. For a whole week, I didn’t eat; I was just quiet. A friend came who asked me what the problem was and I explained to her, she late [sic] went and told my husband parents’.(21, woman)‘I don’t have parents except my husband’s parents […] I didn’t tell the parents of my husband. For a whole week, I didn’t eat; I was just quiet. A friend came who asked me what the problem was and I explained to her, she late [sic] went and told my husband parents’.Participants also described that the social support they received helped them cope with aspects of their lives not directly related to HIV. Two main sub-themes were identified under this theme: (a) strengthening family/relationships and (b) improving job or economic stability.A few participants recounted how the social support they received strengthened their family and/or other relationships. In these instances, participants felt their romantic relationship or family stability—such as with a spouse or partner—was insecure and may soon end. However, receiving social support strengthened and stabilized the youth’s unsteady relationship.Participants also reported receiving support that allowed them to continue working at their place of employment or that validated their participation in income-generating activities. Kafil feared that his boss was going to fire him due to his HIV positive status. He thought this because his position required him to travel and his HIV status could prevent him from obtaining a passport:‘I went back to the office, informed my boss who took time to think over it, but later, he allowed me to continue with my job; this gave me morale and courage of moving on with life’.(27, man)‘I went back to the office, informed my boss who took time to think over it, but later, he allowed me to continue with my job; this gave me morale and courage of moving on with life’.This study explored the effects of social support perceived by Kenyan youth newly diagnosed with HIV. The phenomenological research framework allowed for an in-depth understanding of the diverse influences of social support experienced by these youth. This social support, provided in the forms of instrumental, informational, emotional, and appraisal support, impacted their HIV management and other realms of their lives. Study participants discussed experiencing seven main effects of social support. The support came from both formal and informal sources. Formal support sources included health facility staff, such as clinicians, counsellors, and support groups, as well as non-governmental organizations. Informal sources included family members, friends, spouses/romantic partners, participants’ beliefs in religious practices (e.g., praying, speaking with a pastor), neighbors, and bosses. For many participants, social support played an integral role in strengthening their sense of resolve to persist in life despite the diagnoses of HIV.Our data, in-line with previous scholarship, indicate that social support can positively influence Kenyan youth newly diagnosed with HIV [18,20,21,35,36]. These social support influences can be far-reaching and can occur across several socio-ecological levels, consistent with the findings of Hosek et al. (2008), which suggest youth recently diagnosed with HIV experience stressors within multiple social-ecological systems [11]. Participants in our study described what appeared to be factors enabling them at multiple tiers, including at the individual, interpersonal, and organizational levels. Physical health was improved at the individual and organizational levels, as described by the themes of linkage to services, ARV adherence, and healthy and positive living. Our participants shared that their mental health and wellbeing were also improved after receiving social support at these levels. This was achieved through self-acceptance of one’s HIV status, understanding of what it means to be living with HIV, and family and occupational strengthening. At the interpersonal level, social support aided participants in disclosing their HIV status to others or through strengthening relationships. Though the study focused on the direct influences of perceived social support, there are also many indirect influences. For example, for youth struggling with disclosing their HIV status to others, self-acceptance may facilitate a greater sense of self-efficacy for a successful disclosure event. Furthermore, participants did not describe any instances in which receiving social support had a negative impact. Our inquiry supports the notion that social support is a strong effect modifier and might significantly improve positive living and thinking among youth newly diagnosed with HIV.Similar studies of youth living in Tanzania and Botswana and adults living in South Africa, Swaziland, Uganda, and Kenya broadly describe influences of support that are similar to the themes elaborated by our participants during the focus groups—though, not necessarily using the same terminology nor as much detail as in this paper [19,20,22,23,24]. These studies largely focused on a few influences directly related to HIV, such as reporting decreases in felt stigma, better adherence to ARV regimens, and/or improvements in coping. The influences of receiving social support as described by participants in our study reinforce the positive social support effects described by South African youth in a 2010 study by Petersen et al. This study demonstrates that social support is important for coping and general well-being among adolescents living with HIV [21].This study is unique given that the influences of social support were operationalized from an unrestricted set of sources that, when coupled with the phenomenological research framework, elicited a wide range of influences not often considered. Other studies of this nature did not limit participant inclusion to those newly diagnosed with HIV. Exclusively examining the initial period after diagnosis provides insight into the role of social support during this period of high emotional stress experienced while adjusting to one’s diagnosis [11]. This study expands the current body of research by offering a deeper and broader understanding of the impact that received social support can have among youth newly diagnosed with HIV in Kenya and, potentially, in other sub-Saharan African countries. Furthermore, this study looks beyond the social support influences that directly impact the youth’s HIV management, recognizing that their HIV status does not define their complete lived experience and sense of self.The findings from this study emphasize the key role of incorporating social support into future secondary and tertiary prevention efforts designed to aid youth living with HIV in coping with their diagnosis. Not all study participants described having received social support during their first six months after diagnosis, while others were isolated for a shorter period after diagnosis. These avenues for intervention development involve taking a socio-ecological approach and including multiple stakeholders. This is particularly important given the barriers that individuals living with HIV can face in accessing these formal sources of support, such as barriers for men to participate in support groups as described in the 2012 study by Madiba and Canti-Sigaqa [37]. For instance, at the individual level, health providers may benefit from targeting social support sources in their interventions to improve ARV adherence and/or understanding of the disease. Kulzer et al. (2012) describe an innovative provider-initiated engagement strategy called the family model of care, a tool specifically designed to build upon the strengths of Kenyan families. In this model, providers use a family information table (FIT) to guide people living with HIV through the steps of identifying family members who may be at risk for HIV. At the same time, providers also address concerns related to status disclosure and facilitating family testing. Starting at the individual level, this model encourages each new patient to consider themselves as part of a group at risk for HIV and recognize that they do not have to face their diagnosis alone [38]. Comprehensive programs such as this could play a significant role in helping Kenyan youth better understand the implications of living with HIV. Such programs could also promote the family unit as a source of continued support from which additional influences of social support could arise.At the dyadic level, for example, an individual may better adjust to their partner’s new HIV diagnosis if interventions in which the sero-discordant couple participates demonstrate how to provide social support to one another. In a recent study of 469 HIV sero-discordant couples in Nairobi, incidence of relationship dissolution was high, with 24% of couples reporting separation during a two-year follow-up period. In the study, separation was most common among partners of low socioeconomic status and for women living with HIV [39]. For youth living in a high-risk environment such as Kibera, programs designed to maintain stability among sero-discordant partners could play an important role in both HIV prevention efforts and improving the health of the partner living with HIV. Moreover, social support and partner-focused interventions may lead to better engagement in treatment and care, utilization of counselling and testing services, and improved ARV adherence [23,39,40].At the organizational level, a social support intervention could initiate a paradigm shift in how an organization supports their members living with HIV. Such an intervention could potentially result in positive dissemination effects for other organization members, due to the connection of social support to general health and well-being [16,17,41]. For instance, religious organizations to which people living with HIV already belong could provide support to their members living with HIV, as described by Root (2009) and Watt et al. (2009). However, people living with HIV often reported feeling stigmatized during church participation [22,42]. Root (2009) found that, by defining personhood for people living with HIV, churches could assist members living with HIV with disclosure and help-seeking behaviors [22]. Utilizing faith-based organizations to address HIV stigma mitigation for young people living with HIV needs further exploration. Social support interventions among pre-existing organizational networks could counter stigma, encourage disclosure, promote HIV treatment, and impact wellbeing outside of HIV management.Given the complex relationships among factors influencing health status, lasting and widespread impacts could result from strategically harnessing social support at multiple socio-ecological levels in interventions for youth living with HIV in Kenya. One mechanism by which this may be accomplished is through the use of social media platforms, including social networking sites and mobile technology. As the use of social media becomes more widespread, particularly among adolescent populations worldwide, interventions designed to leverage such forums could prove to play a pivotal role in the way information about—and experiences of those affected by—HIV is shared among individuals, peer networks, and communities at large. A meta-analysis examining the role of social media in HIV communication cited the vast benefits of using social media for this purpose, primarily as perceived by the target populations. These benefits included access to information, strengthened ability to communicate, allowing for anonymity, enhanced social and emotional support, forming a virtual community, and geographical reach. Adolescents in these studies reported experiencing a sense of community from connecting with others through social media [43]. For Kibera youth newly diagnosed with HIV, use of social media platforms such as SMS messaging or Facebook could allow them to connect virtually with others who may not be in their existing peer networks and provide opportunities to discuss topics that may be uncomfortable to broach in-person. Furthermore, the option to retain one’s anonymity through the use of social media could help to alleviate stigma and fear around HIV and allow youth who may not otherwise seek social support a chance to share personal stories they may not feel able to in their offline environment.Social support may be particularly important early after diagnosis, allowing youth to receive such benefits while undergoing the often challenging and complicated process of adjusting to their new diagnosis. Based on the findings in this study, interventions aiming to achieve any of the seven influences discussed in this paper could incorporate social support into their models. It is important to note that youth living in Kibera may already be receiving some social support. Interventions should further foster this concept, to grow and diversify the realized social support effects. The desired influence(s) may differ depending on the needs of the participants and the support they are already receiving. Strengthening social support that youth receive may holistically impact the youth, as demonstrated by the described influences of healthy and positive living, as well as family and occupational strengthening. Though current interventions seek to promote social support at a specific socio-ecological level to directly impact HIV management, our findings indicate that influences of social support can be more widespread and the concept itself may be multi-layered. Recognizing the full potential of social support could address gaps in current interventions and lead to lasting impacts.This is one of first studies to investigate the influences of social support among Kenyan youth newly diagnosed with HIV. The sample for this study included representation of young men and women living in a high-risk, low-resource area in Kenya. Furthermore, the phenomenological approach allowed for a deep exploration of social support influences, due to the open-ended discussions with the youth that were not biased by closed-response options.There are some limitations with this study. Given that the transcripts only included the verbal discussions that occurred during the focus groups and the analysts for this article were not present during the discussions, experiential and situational cues could not be used to further inform and guide the analysis. Additionally, participants may have had similar experiences due to the chosen recruitment technique of chain referral sampling and due to the fact that all participants were currently receiving HIV-related services and living in Kibera. These youth participants may receive more frequent or impactful social support than those not receiving such health care. Finally, as the data collection occurred at a single time point without follow-up, changes in perceived social support over time may have been missed.Future studies should include additional populations, such as those from large urban cities, rural areas, and other sub-Saharan countries, to investigate if the findings in this study also apply, particularly as related research in such areas is lacking. Given the gender disparity within the HIV epidemic and the social and cultural differences experienced by Kenyan women, future research should explore the role of gender in how social support impacts individuals. As participants in the study described social support influences that fell outside of the direct realm of HIV management, future research should view HIV intervention development with a holistic lens through which it is recognized that an HIV diagnosis does not define a person’s complete identity and the interconnectedness of factors impacting health status.Mixed-method studies could provide a deeper exploration and more standardized and targeted information from study participants. Such studies could inform successful intervention development by illuminating key features of the support influences identified in this study. Additionally, a longitudinal study could explore how perceived support effects change over time. The order in which participants experienced different influences of support appeared to have varied by participant; however, this may have been due to reporting errors stemming from data collection occurring at a single time point.This qualitative study provides detailed insight into the diverse influences that Kenyan youth newly diagnosed with HIV can experience by receiving social support. Receiving social support may help youth cope on multiple levels, from a socio-ecological perspective. As such, it could complement a wide range of interventions. More research is needed to better understand the relationship between social support and the different influences, as well as how best to elicit desired effects through interventions. With the growth of social media popularity and potential of digital networks to expand access to social support, future research should examine the role of this media to provide social support influences within these marginalized populations.Conceptualization, N.M.L., K.L., F.S. and G.W.H.; methodology, G.W.H. and E.N.; validation, N.M.L., K.L., F.S., M.K., E.N. and G.W.H.; formal analysis, N.M.L., K.L. and F.S.; resources, G.W.H.; data curation, N.M.L., K.L., F.S. and G.W.H.; writing—original draft preparation, N.M.L., K.L., F.S., M.K., E.N. and G.W.H.; writing—review and editing, N.M.L., K.L., F.S., M.K. and G.W.H.; supervision, G.W.H.This research received no external funding.This work could not have been conducted without the contributions of members of the University of Nairobi’s Centre for HIV Prevention and Research under the guidance of Elizabeth Ngugi and Anne Gikuni, as well as members of DePaul University’s Adolescent Community Health Research Group under the guidance of Gary Harper and Andrew Riplinger. Finally, we acknowledge and thank the youth participants for their contributions to this study.The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.Demographic information of the study participants.

Med-MDPI/ijerph_3/ijerph-16-05-00776.txt

@@ -0,0 +1,59 @@
+Although research about the unintended consequences of paternal incarceration for family well-being has grown in recent years, there has been minimal exploration of food insecurity. Using qualitative methods, we aimed to understand the relationships between paternal incarceration and family food insecurity in Canada. An ethnographic study (24 months) was conducted that included naturalistic observation and in-depth interviews with formerly incarcerated fathers, their partners, and societal reintegration-focused stakeholders (n = 63). Interpretive thematic analysis based on family impact and intersectional theories, indicated that family food insecurity was elucidated by pre-incarceration, economic, social, health, and relationship factors; stigma and social/structural constraints; and intersections among individual, correctional system, community, and macro-level (i.e., economic, social, policy, physical contexts) factors. Participatory approaches and collaborative action among diverse stakeholders that include practitioners, policy makers, researchers, as well as health, social, and criminal justice agencies can guide best practices in creating supportive food environments for families impacted by adversities of incarceration. In particular, interventions aimed at prescriptive ethics, social justice, and meaningful rehabilitation show promise at mitigating the collateral consequences of incarceration-related food insecurity.In Canada, a substantial proportion of males who serve time in prison are fathers [1,2], and this exposes numerous families to the collateral consequences of incarceration. Paternal incarceration contributes to economic, social, and health inequalities [3,4] which are associated with physical and mental health [5,6] as well as, food insecurity [7]. However, there is limited understanding about family food insecurity in this context which impedes progress in programs and policies.Food insecurity—the tenuous access to sufficient, safe, and nutritious food to meet dietary needs—is among one of the inter-related economic, social, relational and familial impacts that occurs with paternal incarceration [8]. Incarceration contributes to economic instability, a known correlate of food insecurity [9], and to human and social capital deficits [3] which can incapacitate a family’s ability to consistently access appropriate food. Collateral consequences of paternal incarceration that can impact food security [10] include lower contribution of shared earnings with an intimate partner [11], destabilized family relationships, impaired parenting behaviors [12,13,14], increased mental health and behavioral problems [15], health care access challenges [14], barriers to reintegration [2,16,17] and the inter-generational transmission of inequality [2].When a father is incarcerated, the child’s mother tends to be the primary caregiver. While data on the global rates of paternal incarceration is lacking, it is estimated that about 5% of Canadian children are affected by paternal incarceration [2]. Although research about the repercussions of fathers being incarcerated has burgeoned in recent years, there has been limited exploration regarding the contributions of food insecurity and its familial impacts. For example, Turney [18] analyzed data from the Fragile Families and Child Well-Being Study using propensity score matching models and reported that paternal incarceration in the previous 2 years doubled the likelihood the father’s children (at 5 years of age) reported food insecurity. Two studies [19,20], that examined incarceration-related food insecurity and food-diet experiences during incarceration discussed links between maternal incarceration and their child’s later-life experiences with food insecurity and highlighted how diet modifications while incarcerated can be a stressor. While these investigations have provided important insights, most have been conducted in the U.S. and may not necessarily detect nuanced, historical, cultural, and dynamic experiences specific to Canada. To contribute to knowledge about paternal incarceration and family food insecurity, a qualitative study was undertaken that focused on fathers who had the experience of incarceration in the Canadian federal correctional system (i.e., serving or had served a custodial sentence of ≥ 2 years). Paternal incarceration was examined because its cumulative risk has been shown to be much greater than that of maternal imprisonment [21,22]. The goals of this investigation were to examine how: (1) a father’s experiences with food while incarcerated may impact on individual and family level food insecurity; (2) paternal incarceration impacts family food insecurity; (3) different factors contribute to family food insecurity in the context of paternal incarceration; and (4) stakeholders who work with individuals affected by paternal incarceration (e.g., from correctional, health, social organizations) perceive and address the issue of food insecurity.In Canada, responsibilities for corrections are shared by federal, provincial, and territorial governments [23]. Correctional Service Canada (CSC) is the federal agency and is headed by the Correctional Service Commissioner who reports to the Minister of Public Safety Canada and is supported by an executive committee of national and regional members. Federal corrections are concerned with offenders who have been sentenced for two years or more. Provincial and territorial corrections are responsible for young and adult offenders who have been sentenced for two years less a day (or less) or who have received community sentences such as fines, community service work, or probation. There are two types of court-imposed federal sentences. The first, a determinate sentence, has a specified completion date, after which CSC no longer has jurisdiction over the offender. The second, an indeterminate sentence, means that CSC has jurisdiction over the offender until the offender passes away. With an indeterminate sentence, the court imposes a minimum amount of time before the offender can apply to the Parole Board of Canada for conditional release. For example, a court-imposed sentence of life with no parole for 25 years would indicate that the offender would be incarcerated for at least 25 years prior to consideration for a potential conditional release to the community, under the supervision of a community parole officer. Once a sentence is imposed, a thorough intake assessment determines the offender’s risk level, needs, and initial security level placement. Maximum security institutions are the most restrictive. They house individuals who pose the greatest risk of escape and danger to society. The inmate’s daily schedule is structured and strict. In medium security institutions, officers are not armed and daily life is similar to that in maximum security facilities. The main functions of minimum security institutions are to foster successful societal reintegration. These facilities are usually organized as small communities where inmates reside in living units (houses) in small groups. There are no barbed wired fences or armed officers, as inmates in these institutions are deemed very low risk. Inmates can organize their schedule according to the activities they are required to participate in (e.g., work programs) and often are responsible for their own meals. While no recent data is available about fathers who are incarcerated, it is estimated that males account for 83% of adult admissions to correctional services in the provinces and territories. For federal corrections, males account for 92% of custody and community admissions [24].The study took place in the Fraser Valley of British Columbia where there are six federal correctional institutions for male offenders. The CSC-managed facilities in the region include all security levels (minimum to maximum; some are multi-level) and all are located within a 60 km radius. The investigation was a collaboration of the University of British Columbia’s School of Nursing and two community agencies, Long-term Inmates Now in Community (L.I.N.C.) and Hope Central. L.I.N.C. is operated by ex-offenders and provides services (including food provision) to people impacted by the criminal justice system (e.g., offenders, parolees, and ex-offenders, individuals who support them, victims). Members of L.I.N.C. come from a wide range of social and economic backgrounds. Hope Central is a centre operated by a community church that provides drop-ins and programming that focuses on relief (e.g., provision of basic supplies), rehabilitation, and life skills development to marginalized individuals including those who have been incarcerated.For this study the population of interest were families where: (1) at the time of study the father was currently or formerly incarcerated in a Canadian federal correctional facility; (2) at the time of the arrest the father either had custodial (i.e., living with their children) or non-custodial (i.e., had children with whom they were not living with) status; and (3) access to food was indicated as an issue during incarceration and/or reintegration by the participant and/or community partner(s) working with that participant.Candidates for stakeholder interviews were selected based on criteria that included: (1) providing CSC and/or affiliated services; (2) having expertise in working with individuals that have lived experience of being incarcerated in a CSC facility and/or their families; and/or (3) providing food related programs or services that include individuals that are currently or formerly incarcerated and/or their families. Although previous work has suggested that stakeholders’ insights can diverge from those with the lived experience food insecurity [25], including their perspectives was deemed to be important in ascertaining how they may be barriers in the perpetuation of inequality to food access.More than 570 h of naturalistic observation data was collected by two researchers over 16 months during drop-ins of Hope Central (62 drop-ins attended with an average number of 38 participants per drop-in), community garden program organized by L.I.N.C. that engaged current and former incarcerated individuals (31 visits with an average of 6 participants), and weekly support groups of L.I.N.C (43 visits with an average of 15 participants). In addition, visits to CSC facilities and other community-based programs were done to better understand the range of services that address food insecurity. Naturalistic observations were recorded as brief aides-memoire [26] with extensive field notes documented after site visits. The notes were organized into four categories: (1) details of visit; (2) sensory impressions and personal responses; (3) specific conversations and insider language; and (4) questions for future investigation. The notes were open coded to keep the inquiry broad and develop a list of exhaustive themes. Examples of preliminary codes included “control of dignified food access”, “expectation of community services”, and “need for advocacy”. Over time, the codes were refined to generate more meaningful categories.An ethnographic approach was used to learn about experiences of paternal incarceration and its culture, values, beliefs, behaviors, and practices [27]. The researchers spent time in the different settings to watch, listen, and engage with the individuals before theorizing the observed phenomena [26,28]. In particular, the researchers considered power relations, values [29], and influences on food insecurity as a means to support changes [30].Purposive and theoretical sampling was used to gather in-depth interview data on practices and perceptions related to food insecurity and paternal incarceration from a diversity of participants and stakeholders (Table 1). One set of data was collected from formerly incarcerated fathers and their family members. These interviews focused on gathering information about the father’s experiences of accessing food during incarceration and reintegration and the family’s experiences with accessing food in their community. Another set of data was collected from various stakeholders that focused on gathering information about factors related to appropriate food access during incarceration and reintegration and perspectives about how to address food insecurity in the context of paternal incarceration. Interviews took place at offices of the community agencies and in participants’ homes and ranged from 40 to 125 min (average 85 min). With permission, interviews were audiotaped and professionally transcribed.Focus groups were also conducted that included family members with the lived experience of food insecurity related to paternal incarceration and stakeholders. The purpose of the focus groups was to explore disparate views, enable point-counterpoint discussion and resolution, and to extract latent issues. The final sample size was determined as the point in data collection where redundancy in themes occurred.The individual and focus group interview guides are located in the Supplementary file.Participation in the interviews was entirely voluntary and had no effect on the receipt of any services. Written informed consent was obtained prior to each individual and focus group interview. Reimbursement for expenses related to interview attendance (e.g., parking, transportation) was provided. The study was approved by the University of British Columbia’s Behavioural Research Ethics Board and Memorandums of Understanding were signed between the community agencies and the research team (UBC BREB NUMBER: H 12-02095).Data was analyzed by six research team members using interpretative thematic analysis. Initially, transcripts were organized and coded as relevant passages of text. As data collection progressed, field and interview notes were read repeatedly to identify patterns and linkages to theory [31]. Transcript codes were compared to identify similarities and differences through discussions among team members to refine categories and themes. Using NVivo Version 10 [32], exemplars of coded text were extracted and compared within and across transcripts. Interpretations were reviewed by team members, stakeholders, and individuals with lived experience of incarceration to check for validity.To relate participants’ experiences with incarceration and food insecurity, the findings were interrogated based on intersectional [33,34] and family impact [35] theories. Like other scholars [36,37,38], we recognize that disparities arising from factors such as biological sex differences, gendered experience, ethnicity, and class affect food insecurity in ways that are independent and that intersect and compound one another. In addition, we acknowledge that this issue needs to consider how current policies and programs are either barriers or facilitators of family empowerment and stability [35]. In the final analysis, coded narratives were contextualized into a conceptual framework to better understand the multi-level individual, social, and environmental correlates influencing food insecurity and to help provide a clearer picture of the higher impact, upstream forces that may be used to address food insecurity in the context of paternal incarceration [39].For conceptual clarity, the definition of food insecurity adopted for this study was “whenever the availability of nutritionally adequate and safe foods or the ability to acquire acceptable foods in socially acceptable ways is limited or uncertain” [40]. This definition, supported by ethnographic research [41,42], means that food insecurity is experienced when there is uncertainty about future food availability and access, insufficiency in the amount and kind of food required for a healthy lifestyle, and/or the need to use socially unacceptable ways to acquire food. The term family food insecurity used in this study refers to food intake that is sufficient for family members to lead ‘active, healthy lives’ [43] and accounts for specific cultural relationships with foods [44].Although the data analysis was situated within the context of systemic inequities that impact individual and family levels of food security [45] which can inform social policy, where applicable the role of broader community levels of food insecurity was questioned. Community food insecurity occurs “when dominant food systems fall short in terms of social, economic and environmental sustainability” [46] and thus provides a foundation that can inform food policy.Several quality control measures were integrated in the analysis. As part of the study, two peer researchers (formerly incarcerated father and wife of an incarcerated father) were recruited to be part of a research team to help gather and interpret individual and focus group data. Triangulation of data from the fathers and family members, diverse stakeholders, and observational data also contributed to the rigor and trustworthiness of the analysis. Results are reported for areas where most (e.g., >75% of participants) shared similar views.The fathers (n = 13 who did individual or focus group interviews) who participated in the study had served either determinate or indeterminate prison sentences ranging from two to more than 20 years in different federal institutions (Alberta, British Columbia, Manitoba, Ontario, Nova Scotia). They ranged in ages from 33 to 64 years, had between 1–3 children, and identified themselves as Caucasian, Aboriginal, Asian, or Hispanic. Most (n = 10) had at least high school education; three had completed an undergraduate degree. All were custodial fathers prior to incarceration; at the time of the study four participants were no longer with their partners. Some had served more than one prison sentence (federal and/or provincial). Five of the 10 partners of currently or formerly incarcerated individuals who participated were connected with the fathers in the study. At the time of the study all participants were living in British Columbia. The stakeholders (n = 40) who participated in individual or focus group interviews were based in different locations across Canada. They included individuals who were employees of CSC (corrections officers, probation officers, food service and programming staff) and employees of agencies that provided integration services (e.g., halfway houses, vocational services, food programs).Analysis of the textual data clarified different pathways between paternal incarceration and family food insecurity that included: (1) pre-incarceration experiences/lifestyles and intensified family food insecurity; (2) changes in relationships and health status; and (3) the fathers’ experiences and parental relationship quality.Consistent with the literature about paternal incarceration, some participants reported histories of significant economic hardship and most made reference to their experiences of childhood adversity [47]. The described adversities have been reported elsewhere to be associated with food insecurity [19].
+              She (mom) made her choices. She had me when she was young. 16. I have no ill will towards her at all..I had no family support…My dad is doing time I think in the States. That is karma. Right? That is what I chalk that up to. Hopefully my kids won’t turn out the same.
+            
+              He came from a hard life and his family were involved with criminal activities in order to make ends meet. My son knows this and about his father being in jail. Even though dad is out we have problems….it worries me that [son’s name] will also think that crime will be an answer for money problems.
+            In addition to the economic hardships, the fathers and various stakeholders also described new or intensified incarceration-related experiences of food insecurity and how this perpetuated intergenerational effects of disadvantage and marginalization. For example, one of the Aboriginal spouses described how incarceration impacted their access to traditional foods.
+              We were generally able to eat whatever we wanted as a family but after (father’s name) went to jail, we had to plan meals more carefully and watch costs (before incarceration). We would eat fish which (father’s name) would catch. While he was is jail we ate no fish..(father’s name) wasn’t here to catch it or show the kids how to do this.
+            In other instances, spouses discussed how their mental health struggles worsened due to financial stresses which imposed constraints on eating. They described how these constraints led to patterns of dietary restriction followed by overconsumption that became cyclical and similar to binge-eating. Other research has reported significant associations between mental ill health and disordered eating [48] and respondent’s descriptions supported the “food stamp cycle” hypothesis which suggests that binge-eating behaviors occur when access to food increases (e.g., availability of food stamps) which is then followed by periods of involuntary restriction when food resources run low [49].
+              It’s like the welfare Wednesday situation where you binge spend after you get the cheque..you binge eat when you have the money.
+            In rare instances, a father’s incarceration had little impact on their family’s access to food, which was consistent with other research insights [19]. This was particularly the case if the father did not live with the family or did not substantially contribute to the family’s income prior to their incarceration. Finally, pre-incarceration factors such as education, social support, and prior food skills were discussed as resources that helped families be resilient to incarceration-related food insecurity.
+              ..They realized you’re very well educated. This is my diploma. They didn’t know what to do with me. What do you do with a highly educated inmate?.. I really don’t need your help..I’m fine.
+            Prior to incarceration, most fathers contributed economically to family life, were actively involved in parenting their children, and family food insecurity was not perceived to be an issue. Participants identified incarceration-related factors such as trying to maintain housing security, loss of social connections, loss of work, changes in social assistance, accumulation of legal and household debts, and inability to provide financially for their families which impacted on both economic and food security [50]. Feelings of shame, failure, and inadequacy and how these were motivational barriers in seeking healthy foods were often highlighted.
+              ..if finances are an issue then no doubt food insecurity becomes an issue. So as soon as that’s an issue then there is a whole host of problems… their social support system is a problem…they don’t leave all of a sudden and have a bunch of new friends that are going to allow for a new life.
+            
+              ..basically what you’re doing is that someone commits a crime in their home community and they get extracted from the community where they live, they get all their ties, their family, their friends, their social status gets severed and they get pulled out and get socialized with a large population of offenders...so they get basically taken from any pro-social healthy environment... for a long stage of time.
+            
+              While in prison I had no real sources of income. You can do work while in jail but they pay less than minimum wage. So yeah..I had nothing to contribute to family finances..[Wife’s name] was on her own..making sure the kids had a roof over their head and food on the table.
+            Adding to the economic burdens associated with paternal incarceration, the partners of the fathers reported experiences of significant social network disruptions with family, friends, and community that also placed constraints on food access. They described periods of time where foods consumed were insufficient, low quality, or undesirable. They also discussed how they worried about where to get food and felt forced into obtaining foods in socially unacceptable ways or consuming foods which did not meet personal standards of acceptability.
+              I can remember once when [husband’s name] was in prison...I was stone broke... I was on E.I...so I had to go to the food bank…so what happens is the stuff I get from the food bank was white sugar, canned fruit...crap that I couldn’t use. One of the things that I’m very conscious about in terms of prison is the lack of choice, the lack of empowerment around defining your own existence and that’s the same around food…
+            For partners that resided in smaller communities, they noted how they would try to avoid seeking out food from community-based sources such as food banks to avoid being highly visible and stigmatized. Instead, they would try to obtain alternative financial sources, including engaging in illegal activities, in order to obtain food.
+              So I started selling contraband smokes just so I could have some money for us to have a decent life...I didn’t like doing this...but your back is up against a wall...what are ya gonna do?
+            These comments suggest that the relationships and consequences of incarceration-related food insecurity are bi-directional which has been reported elsewhere [5,51]. Food insecurity either worsened or became a new issue families had to face when the father was incarcerated. Furthermore, these challenges forced families to resort to criminal activities as a means to address their food insecurity. Thus, the relationships between food insecurity and incarceration appeared to be shaped by systemic factors that facilitated mutual reinforcement [52].Several respondents discussed how the pursuit of meeting basic needs and despondency contributed to the neglect of care for both themselves and their families. Further to this, parents described how these obstacles contributed to the deterioration of their family’s health and how the added disadvantage of food insecurity worsened physical, mental, and social well-being [53,54].
+              When (partner’s name) was in prison I was overwhelmed...lawyer’s bills, losing friends..I became severely depressed. I didn’t want to cook...so for me and [son’s name] it was a lot of meals from a can...It was not surprising that problems started happening with [son’s name]. He was acting out at school..
+            These findings are consistent with other evidence that suggests paternal incarceration is associated with poor eating behaviors, academic and socioemotional skills deficits [55,56,57], antisocial and criminal behavior, internalizing symptoms, mental health problems [58,59,60] and drug use [61] among children whose parents have been incarcerated [52,62]. Furthermore, shared biological pathways among these behavioral (e.g., stress response), mental health (e.g., anxiety over having continued access to appropriate foods), and nutrition-related (e.g., inadequate intake of nutrients due to lack of access to health promoting foods) factors could exacerbate their food insecurity [63]. Macro-level drivers such as food-related policies (e.g., agricultural subsidies) and design of food environments (e.g., retail access to healthy foods) can enable and perpetuate a vicious cycle which relegates the family to society’s fringes and reinforce their struggles with food insecurity [64].In some instances, family members became formally diagnosed with chronic health conditions (e.g., diabetes, dyslipidemia), a common experience related to incarceration [65]. Often these conditions required adhering to therapeutic diets. Participants spoke about how experiences of food insecurity both seemed to lead to the diagnosis and impede their ability to manage their chronic condition.
+              (Daughter’s name) got diabetes and had to insulin. I think the bad diet we had to follow led to this…and now..now with the diabetes it makes things harder with diet …getting the right foods needed to keep her (blood) sugar under control is a huge problem..
+            It is widely known that health conditions, as described by respondents, arise through a complex web of interactions between genetic and environmental factors including food access [66]. This raises fundamental questions about how to shape social and health practices and policies that prevent gene-diet interactions that lead to chronic condition development. Although there is still much to be done to identify these interactions, this work is relevant to preventing and effectively managing chronic conditions and reducing health, social, and correction service-related costs.Participants shared reports of strained family relationships, including challenges with maternal parenting. These issues were heightened by factors such as the geographical proximity of the father to his family and the type and length of prison sentence. For example, a father who was institutionalized in a province where his family was not residing stated:
+              ..I never saw them…for almost a year. Which is critical for two little boys right? A whole year with no family. Not as hard on me as it was on those two boys. You know “where is daddy?” and they don’t really understand right? It was the distance. The cost. At the time I was in jail…I had no income… and they don’t understand.
+            Fathers described how these circumstances lead to few or no visits with their family and the opportunities that could occur with those visits to share experiences and connections around food. Furthermore, circumstances which lead to disruptions in family meals could impact on children’s biopsychosocial well-being [67]. Within the participant narratives, contextual factors such as facility differences which impacted on the father’s experiences and relationships with food were also highlighted.
+              When I was in high security access to food was dictated by the cafeteria. Healthwise I wasn’t doing great…very depressed. I had no control over what I could eat. When I moved into prisons where we could make our own food my relationship with food got better. I could choose what I wanted to buy from the prison’s store. I felt less like a degenerate.
+            As others have suggested, choices about food in correctional facilities are different than choices about food in the community [68]. Meal choices are planned by facility staff and those who are incarcerated typically do not provide input about foods offered or eating schedules. Similar to other findings [20], respondents discussed how access to appropriate foods would enable them to manage their health and to be physically and mentally prepared for reintegration. In extreme cases, some discussed how the system of food provision in correctional facilities negatively shaped their relationships in food. Examples were described of how limited food access during lock downs and witnessing violent fights over prized food items made them associate food with traumatic experiences. Many also discussed how correctional system fiscal constraints that often led to food budget cuts contributed to taking the pleasure out of eating and losing a sense of control over food choices. As others have noted [69], perceptions about inadequate food within correctional facilities can fuel frustration, humiliation, and deterioration in health status. This, in turn, can increase risk for rule violations, violence [70], and recidivism [71] which impact the management of correctional facilities.
+              So they gone regional [centralized food production—to save money. There is fewer staff..(The prisoners) are not happy. It is one of the few things that get them through the day—the food—and now they see that as being taken away from them. This, among other things that happen in prison, beats them down...
+            Although respondents acknowledged they were low on the social hierarchy, they discussed how being at this tier also had put them in positions of variable access to food while incarcerated and that this had carryover effects during reintegration.
+              …in prisons there is a hierarchy and food certainly played into this. At mealtimes certain inmates got better treatment...better food. Certain inmates would be chosen to work in the kitchen and had more access to food.
+            In some cases, social locations such as age, ethnicity, and incarceration experiences influenced their positioning within the carceral hierarchy and either facilitated or impeded food access.
+              Some were more likely to have less food available to them. Like the younger ones...they don’t get full. Others may have food brought in for them or be given money by family to use for canteen. Food was used as a commodity on the inside. Some would beg for food from others. Others would steal food.
+            
+              ..so you get a certain amount of respect for going through the maxes and what not. I couldn’t care a less. I didn’t care about that…And these guys right would like come and sit at the back table (at meals).and ask..Do you need this? I am just like no it is all good. I’ll just sit with the natives.
+            While most discussed food insecurity-related challenges, some noted that being in correctional facilities could provide opportunities to positively influence their eating experiences (e.g., learned food skills) and foster food security during reintegration:
+              Some of them have been involved in culinary arts programs. Some guys develop niches where they can maybe bake. So in exchange for baking the other guys will get the ingredients. ..Things like this can help them when they get out.
+            
+              While he was in there he actually learned how to cook. Even got his FoodSafe.
+            Many made reference to how knowledge about health, food, and CSC policies helped to mitigate issues related to appropriate food access. One spoke at length about how he studied policies and prisoner’s rights in depth in order to lobby for better foods.
+              We knew our rights..we had access to Stockwell Day’s..pathways to safety..it was some big 400 page manuscript he..that Stockwell Day wrote..in 2007. We got freedom of information on it and got a copy of it. It said things like medication and food proportion weren’t suppose to be messed with.
+            Various intersecting factors shaping experiences of food insecurity were apparent in the participants’ narratives. The main subthemes included associated stigmas of incarceration and the social and structural constraints/supports that affected one’s personal sense of control related to food.Stakeholders in particular discussed how various personal factors such as mental health status, race/ethnicity, and age intersected with problems of knowledge, attitudes (prejudice), and behavior (discrimination). Furthermore, these factors created additional barriers to food security during incarceration and reintegration [72,73]:
+              …there was this attempt to treat everyone the same and in some instances that just did not work. For example, the older inmates wouldn’t always get the things they needed...those not from Canada...same thing...they didn’t always get their cultural needs addressed. It would depend...If you’re Aboriginal, you may or may not get accommodated more. The thing is, all the staff would know these people’s histories, so in some ways they wouldn’t be given a chance...
+            
+              My experience was that, not from the inmates themselves, but rather from the people within the food system...not always a lot of...understanding or belief about the impact of food...I don’t feel there was a lot of credibility given to that area of life and potential positive that could come from it…I guess stigma...comes into that. There is the thought that…well you’re in jail so why should you have good stuff...
+            Instances where the stigmatization of incarceration impeded food access were also voiced by family members. For example, invitations to attend food-related social gatherings became less frequent for families when the father was incarcerated. Others described how participating in activities such as community gardens or community kitchens were impacted. For example, one mother described how she no longer went to weekly community kitchens because people would always be asking about her spouse and the reasons why he was incarcerated.
+              I thought going to a community kitchen while (spouse’s name) was in jail would help me make friends and bring home healthy food. But going to them became a chore..I thought I would find support there..instead I felt like I didn’t belong because my reasons for needing food didn’t seem legit to others.
+            Applying an intersectional approach to the analyses advanced understanding of how social and structural factors such as historical contexts and political will contributed to food insecurity. Respondents discussed how incarceration limited upward social mobility by preventing them from developing both human and social capital [52]. For example, one father’s conviction prevented him from returning to his work as an airline pilot. Furthermore, his highly skilled wife with post-secondary education had to leave her job to help raise their young children. After the father was released, he was forced to work in jobs that paid about one-quarter the wages he was accustomed to and the wife did not return to work. In this instance, both parents who were high income earners lost both individual and social capital as a result of the father’s incarceration. Further to this, participants discussed alternative approaches to incarceration that could facilitate the re-establishment of positive social ties after release as well as mitigate the impacts of food insecurity. Echoed in their narratives were descriptions of how various social and structural constraints blocked positive action.
+              ..on a completely theoretical and idealistic perspective we would do a deinstitutionalization type approach. So rather than institutionalizing our system it would be community based...If they provide a sentence of more than two years..institutionalization is mandatory. Ideally speaking, we would move away from that if rehabilitation is our goal. But we would need a social mandate to do that and I don’t think we’re anywhere close to that in our society and culture...society wants people who commit a crime to be locked away…So politicians..they want to appeal to voters...public safety is a concern and fear... from a reintegration and rehabilitation perspective that is unhealthy.
+            In addition to the losses in human and social capital, structural constraints specific to food environments were discussed as being barriers to food security [74] both in the carceral setting and during societal reintegration. As other research has shown, incarcerated individuals that reintegrate often live in areas with low access to healthy food retailers which can aggravate the inequity of food insecurity [75].
+              ..our food system…is an essential and critical part of our environment. And that environment that we create has major impacts on both our current…choices as well as our long term cause and effect stuff. So outside of the prison system, for example, the choices we have at...grocery stores has a major impact…in the environment that we create in the prison or outside the prison or the social circles and networks that are made are going to have major effects on the choices that are made.
+            When asked about resources that are needed to mitigate the effects of paternal incarceration and food insecurity, the fathers often discussed how their return to community was impacted by vulnerabilities such as age, illness, and social isolation and that meaningful rehabilitation was critical.
+              ..to foster rehabilitation..okay..It has to be within different categories. There is the hard core criminal. There’s the medium guy..For me, for example, I am rehabilitating myself. I am part of the community, working hard, providing a service that is important, hiring people and so on. This because..I have the ability, knowledge and education and so on – to say okay I have a dollar and make two dollarsv—then four dollars—I am smart enough to do this. When it comes to other people..I don’t believe there is enough support—financially—in order for that person to be establishing themselves..If there is no meaningful rehabilitation you don’t give the tools to rehabilitate.
+            Stakeholders tended to emphasize that the impacts of incarceration-related food insecurity also burdens communities, particularly those which had high crime and incarceration rates. As echoed by others [76], conditions such as family connection, residence, income, and basic food access provide the conditions to fully develop relationships with one’s community. Others spoke about the need for personalized approaches, collaborative action from diverse stakeholders, and integrating participatory strategies.
+              …there has to be a space where a person has a capacity to explore parts of themselves that are more in that belief and value level rather than just behaviour..it’s difficult to in my opinion to do this without doing a personal plan for every person to say in light of what your experience is, in light of the culture your part of, in light of your background and the trauma you’ve experienced and in light of your addiction, let’s have a conversation that tailors a plan to you as an individual...
+            
+              It needs to be people working together—not just the folks working in prison and with community programs. There needs to be a way of figuring out how a family is being affected by the dad being in prison and then coming up with solutions to make things better..especially for the kids.. any ideas for improving the situation needs to include those affected by the problem…
+            The thematic analysis of textual data from those experiencing food insecurity and the stakeholders that work with these individuals helped to shape a conceptual framework (Figure 1) that illustrates the dynamic interplay among micro-, meso- and macro-level factors that contribute to incarceration-related family food insecurity [77,78,79,80]. As depicted in the framework, food insecurity in the context of paternal incarceration results from continuous interactions between the present contexts and processes combined with individual, systemic, and historical factors [81]. For example, inter-related components of employment, income, education, race, gender experience, housing, and geography may intersect with systemic barriers (e.g., discrimination, limiting policies) via multiple socio-ecological influences that forces individuals to the margins of society, entrenches their food insecurity, and perpetuates inter-generational inequalities.The framework also helps to illustrate and explain how adversity resulting from paternal incarceration (micro-level factor) and community level factors (meso-level) such as proximity to family, cohesion, and disorganization influence family food insecurity [19,82,83]. Societal reintegration then becomes challenged and the access to and effective use of public and non-profit social services becomes critical [49]. It is acknowledged that programs and policies currently have an important role in the lives of families impacted by incarceration-related food insecurity including the provision of community supervision, health care, public transportation, and social services. Thus, policymakers and the resources they have access to, can be particularly influential in implementing interventions that foster the well-being of families impacted by incarceration-related food insecurity. Within the section of the framework “Opportunities for Prevention and Intervention”, various programmatic, investigative, and policy considerations that could intercept the effects of family food insecurity are outlined. The framework integrates existing literature with the findings of our study to suggest interventions to mitigate incarceration-related food insecurity.The primary goal of this study was to better understand the factors and consequences related to family food insecurity in the context of paternal incarceration. Using interpretative thematic analysis that applied family impact and intersectional theories, it was found that the stressors of family food insecurity and paternal incarceration coexist and mutually reinforce one another. The added burden of food insecurity when fathers were incarcerated further eroded the well-being of family members, which, subsequently, exacerbated the effects of food insecurity. At broader levels, stigmas and cultural norms that emphasize self-sufficiency contributed to the cyclical relationships that exist between paternal incarceration and food insecurity. Food insecurity, as a collateral consequence of incarceration, interacts with different economic, social, policy, and physical circumstances that converge on families and deepens their marginalization. The following discussion highlights potential mechanisms in which paternal incarceration compromises family food security and identifies components of the framework that can be leveraged as strategies to move forward.Consequences of food insecurity related to paternal incarceration occur for both the offender and their family members. Fathers living in a correctional facility experience food insufficiency [84] as a result of changes in the types, frequency, and variety of foods consumed. In addition, incarcerated fathers are excluded from the labor market which deprives their families of a source of income [3,85] and increases their vulnerability to food insecurity. Paternal incarceration also threatens the earning power of family members, who may sacrifice work time to perform tasks previously done by the father [86] or struggle to cover incarceration related expenses (e.g., legal representation, costs of maintaining contact through visits) [13]. A family’s financial instability can persist beyond the period of incarceration as ex-prisoners face labor market challenges during reintegration [16,17]. During incarceration, parental relationships tend to dissolve [87] which can increase the likelihood of food insecurity as earnings may no longer be shared [3]. These economic hardships [4] may elevate the family’s stress levels [21] and impede their ability to manage family food resources.Beyond financial influences are other disruptive factors such as chronic disease, family stresses, and poor function that increase risk for incarceration-related food insecurity [62]. As described by others [61], the lived experience of family food insecurity and paternal incarceration contributes to unfavorable outcomes such as household instability, impairments in mental and physical health, social exclusion, and behavioral challenges in children [5,21,88,89,90]. These effects are perpetuated by structural constraints such as Canada’s current social safety net which provides limited financial supports and policies that govern program assignments and payments for incarcerated individuals. Moreover, stigmas associated with both incarceration and food insecurity erode familial relationships and create motivational barriers to participate in society.This study’s results and proposed framework (Figure 1) shows that the conditions that lead to food insecurity in the context of paternal incarceration are dynamic, and that individual experiences depend on circumstances, history, and opportunities [91,92]. For example, poverty, as a determinant of health, interacts with and reinforces other determinants which then contribute to increased costs in health, social, and justice systems. To address food insecurity in the context of paternal incarceration, downstream and upstream reforms are needed such as screening for food insecurity, creating equities in income, social support, and health, and, where feasible, exploring incarceration and reintegration alternatives. Currently strategies to alleviate food insecurity in Canada include food-based (e.g., charitable programs, community-based interventions) and income responses. Of these two options, income-based responses such as policies that promote income security through employment policies, income transfers, tax exemptions and credits, are more effective at reducing food insecurity [5]. However, in the context of paternal incarceration, approaches to mitigate food insecurity must consider factors such as family justice and developing effective re-entry programs such as supporting access to appropriate living arrangements. Although current health, social, and justice systems are resource-constrained, continued neglect in addressing food insecurity will only perpetuate the human and economic costs that stem from this inequity.The proposed framework (Figure 1) highlights the need to focus on intersecting societal, policy, community, family, and individual factors to address family food insecurity in the context of paternal incarceration. Historically, the focus of most policies and research related to incarceration has not considered the socio-political contexts that may impact food insecurity. Attention must be directed to the influence of macro-level factors (e.g., housing policies), how incarceration intensifies inequities such as food insecurity, and long-term solutions that takes into account the multidimensional nature of family food insecurity. Examples of a re-entry approaches that would be beneficial for alleviating food insecurity include training farms that build food literacy and skills of incarcerated individuals. These programs foster food connections and improve employment potential of incarcerated individuals [93]. In some instances, these farms or gardens are located within communities and are sources of food provision for residents. These types of approaches align with goals of meaningful rehabilitation and restorative justice [94].Our results may be limited due to selection bias. We primarily recruited individuals who were associated with the agencies of our community partners and therefore may not have reached those experiencing extreme food insecurity. In our sampling, we did recruit people who had at some point encountered issues with access to food. In our experiences with the participants this meant many things such as having issues with the quality of the food offered in facilities to having the need to access emergency food relief services during reintegration. Participants that were interviewed had been exposed to varying levels of hardships and food insecurity. Secondly, part of our sample were stakeholders that were both directly and indirectly involved with corrections and societal reintegration services who could speak about the experiences of different individuals. Finally, we hired peer researchers who had experienced successful societal reintegration and overcame food insecurity issues to provide further insights. By integrating multiple perspectives we believe that the described mechanisms generalize to the scope of the study.The interviews were done with fathers who were formerly incarcerated and may not necessarily reflect current experiences with CSC services. For example, based on field observations and stakeholder interviews, current programs that build food skills and changes in some CSC food delivery systems were highlighted. The interview data from fathers and their partners were from those living in British Columbia and thus may not reflect the experiences of those residing in other Canadian provinces or countries. However, many of the respondents had been in CSC facilities across Canada and provided similar accounts related to experiences of food insecurity. Furthermore, some stakeholder interviews were with individuals residing out of province which helped strengthened the study’s representativeness. The purpose of the proposed framework is to help foster understanding about how food insecurity manifests in the context of paternal incarceration, however, a framework that includes specific evidence about food insecurity interventions in the context of incarceration in relation to health, social and justice outcomes, would provide more guidance on program and policy development.The use of ethnographic, family impact, and intersectional-based analytic approaches underscored the importance of understanding the multidimensional nature of food insecurity in the context of paternal incarceration. Furthermore, it clarified the need for interventions that reflect factors such as age, gender, socio-economic status, ethnicity/race, and marital status that shape food insecurity in the context of paternal incarceration. There are many opportunities to guide best practices that would foster supportive food environments for families affected by paternal incarceration and help reduce associated health, social, and justice system costs. Collaborative action among diverse stakeholders—practitioners, policy makers, health, social, and criminal justice-based agencies, researchers—that include participatory approaches could facilitate the change necessary to reduce inequities associated with paternal incarceration and food insecurity. In particular, interventions aimed at prescriptive ethics, social justice, health promotion, and meaningful rehabilitation show promise at mitigating the familial and intergenerational effects of paternal incarceration and food insecurity.The following are available online at https://www.mdpi.com/1660-4601/16/5/776/s1, S1: Individual and Focus Group Interview Guides.K.M.D. and V.L.S. with assistance from Annette Browne developed the proposal for grant funding. K.M.D. and V.L.S. formulated and implemented the analysis plan. All team members were involved in review of the transcripts and data analysis. K.D., C.D, and S.M. developed the manuscript draft that all team members reviewed and contributed to subsequent drafts.This study would not have been possible without the expertise and collaboration provided by Sherry-Edmunds Flett, executive director of L.I.N.C. Society, Greg Elford, pastor at New Heights Church located at Hope Central, and Carrie Prentice, Hope Central program coordinator. We would also like to acknowledge the staff and volunteers whose work facilitated this collaboration. We deeply appreciate the participants that shared their experiences that informed this study. To the peer researchers who worked on this project we are grateful for your insights and assistance with the study’s implementation and data interpretation. We thank Correctional Service Canada staff that enabled site visits and shared their insights which helped us to better understand the contextual factors of this study. We would also like to express our gratitude to Annette Browne for her help with proposal development and her valuable insights over the course of the project and to Helen Brown who reviewed the manuscript draft and provided valuable feedback.The authors declare no conflict of interest.Factors influencing family food insecurity related to paternal incarceration.Description of Participants.a Examples: L.I.N.C, Elizabeth Fry Society, The John Howard Society of Canada, St. Leonard’s Society, Correctional Service Canada, Hope Central, Salvation Army, Lookout Society.

Med-MDPI/ijerph_3/ijerph-16-05-00777.txt

@@ -0,0 +1 @@
+Cost-effective treatment of dyeing wastewater remains a challenge. In this study, a newly designed hydrolysis acidification flat-sheet ceramic membrane bioreactor (HA-CMBR) was used in treating high-strength dyeing wastewater. The start-up phase of the HA-CMBR was accomplished in 29 days by using cultivated seed sludge. Chemical oxygen demand (COD) removal rate reached about 62% with influent COD of 7800 mg/L and an organic loading rate of 7.80 kg-COD/(m3·d). Chromaticity removal exceeded 99%. The results show that the HA-CMBR has good removal performance in treating dyeing wastewater. The HA-CMBR could run with low energy consumption at trans-membrane pressure (TMP) <10 kPa due to the good water permeability of the flat-sheet ceramic membrane. New strains with 92%–96% similarity to Alkalibaculum bacchi, Pseudomonas sp., Desulfovibrio sp., and Halothiobacillaceae were identified in the HA-CMBR. Microbial population analysis indicated that Desulfovibrio sp., Deltaproteobacteria, Halothiobacillaceae, Alkalibaculum sp., Pseudomonas sp., Desulfomicrobium sp., and Chlorobaculum sp. dominated in the HA-CMBR.China is the world’s largest producer of dyes, at more than 1.15 × 109 kg annually [1]. Dye exports have brought some economic growth for China, but the dyeing production process generates large volumes of dyeing wastewater. The raw materials of dyeing production include benzene, naphthalene, and polycyclic aromatic hydrocarbons, but also contain heavy metals, salts, and other substances. Therefore, dyeing wastewater is characterized by complex compositions, deep color, and toxicity for living systems [2]. The discharge of untreated dyeing wastewater causes serious pollution to surrounding environments.The cost-efficient treatment of dyeing wastewater has attracted much research interest [3,4,5]. At present, physical and chemical methods are usually used to treat dyeing wastewater [6]. However, such methods have high operating costs. In comparison, biological methods have the advantages of low running costs and being more environmentally friendly, and are expected to be widely used in treating dyeing wastewater [7]. Li et al. [8] reinforced an anaerobic hydrolysis–denitrification coupling process, and the coupling process showed efficient removal of nitrogen and aromatics. Hayat et al. [9] achieved 87% chemical oxygen demand (COD) removal efficiency in an anaerobic internal circulation (IC) reactor treating real textile industry wastewater. Manavi et al. [10] used aerobic granules to treat real dyeing wastewater and achieved 73% color removal and 68% COD removal with a cycle time of 24 h and an anaerobic-to-aerobic period of 3:1. Therefore, cultivated anaerobic microorganisms have been shown to be efficient for the degradation of dyeing wastewater. However, anaerobic microorganisms are easily lost from the reactor due to changing conditions such as linear velocity, granules breaking, gas floatation, etc., resulting in lower treatment performance. Membrane separation was deemed as an effective method [11]. Membrane bioreactor (MBR) is a wastewater treatment technology that can achieve high load and good removal efficiency through efficient separation of microorganisms [11,12]. Aerobic MBR (AeMBR) and anaerobic MBR (AnMBR) are the usual forms used in treating wastewater. AeMBR usually uses organic membranes as a membrane separation module, but for AnMBR, these can be damaged during prolonged operation [13]. Compared to conventional organic membranes, the advantages of ceramic membranes include high permeability performance, energy savings and cost reductions, long lifespan, good durability, and easy maintenance and management [14]. A flat ceramic sheet can be used as the recycling module of an AeMBR in the treatment of industrial and municipal wastewater [14,15,16]. Based on the advantages mentioned above and practical experience, flat-sheet ceramic membranes are expected to perform well in AnMBR. In addition, subsequent aerobic biological treatment is needed to meet the discharge standard in treating dyeing wastewater [6,7,8]. Among them, the biodegradability of wastewater after anaerobic treatment is one of the main factors in determining the removal efficiency. Anaerobic processes include hydrolysis acidification and methanogenesis. The hydrolysis acidification stage can effectively improve the biodegradability of dyeing wastewater in a shorter hydraulic retention time (HRT) [12]. Combined with aerobic biological treatment process, it is possible to meet the discharge standard of China [7,8]. Meanwhile, compared with anaerobic processes, the hydrolysis acidification process has the advantages of a simple structure and less investment. Thus, hydrolysis acidification based on flat-sheet ceramic membranes is suggested to be evaluated in treating dyeing wastewater. Nevertheless, there are no previous reports on hydrolysis acidification based on flat-sheet ceramic membranes used for treating dyeing wastewater.The present study employs a flat ceramic sheet as the membrane module in a newly developed hydrolysis acidification flat-sheet ceramic membrane bioreactor (HA-CMBR) that was developed for treating dyeing wastewater. Anaerobic microorganisms were screened and seeded in the HA-CMBR. Treatment performance was evaluated using real dyeing wastewater, and 16S rRNA was employed to characterize changes in the microbial populations of the anaerobic sludge. The results might provide theoretical guidance and reference for engineering applications.The wastewater sample used in this study was obtained from Weifang Ruicheng company (Shandong, China) that produces disperse, reactive, and acid dyes. The compositions of the dyeing wastewater are shown in Table 1. Due to the lack of nitrogen and phosphorus nutrients in the dyeing wastewater, NH4HCO3 and KH2PO4 are added as nitrogen and phosphorus nutrients in the required proportion.A schematic diagram of an HA-CMBR system is described by Zhang et al. [12]. Anaerobic microorganisms from a running hydrolysis acidification tank [17] were seeded in the main reactor. The permeate was discharged into the effluent tank using a filtration pump. The feed flow rate of HA-CMBR was 16 L/d with an HRT of 12 h. The HA-CMBR consisted of an influent tank, main reactor, warm water control system, and effluent tank. The influent was adjusted for pH in the influent tank and pumped into the bottom of the main reactor. The temperature of the main reactor and membrane pool was controlled at 34 ± 1 °C. The HA-CMBR was seeded with 2.5 L of sludge that had a mixed liquor volatile suspended solids (MLVSS) content of 10,000 mg/L.Filtered COD (1 µm) based on dichromate was measured using the closed reflux colorimetric method [18]. Biochemical oxygen demand over a five-day (BOD5) and chroma were measured according to the standard methods [19]. Total nitrogen (TN) was determined with the persulfate method [20] using the ultraviolet spectrophotometric screening method for quantification of TN as NO3–N (i.e., oxidization product of persulfate digestion). Total phosphorus (TP) was measured according to Yue et al. [21,22]. The pH was measured using a pH meter (9010; Jenco, San Diego, CA, USA), and dissolved oxygen (DO) was measured using a DO meter (6010; Jenco, San Diego, CA, USA). Trans-membrane pressure (TMP) was recorded by a digital pressure sensor (SHANG YI, Foshan, China). Extracellular polymeric substances (EPS) and soluble microbial products (SMP) were analyzed according to Ramesh et al. [23].The HA-CMBR was fitted with a flat-sheet ceramic membrane of normal pore size 0.1 μm and effective area 0.05 m2. The hydraulic residence time was set to 8 h (according to the main reactor). The main reactor was seeded with 3 L sludge that had a mixed liquor volatile suspended solids (MLVSS) content of 8000 mg/L.Eight sludge samples were collected during the study: Two samples for seed sludge (No. 1 and 2); and six sludge samples, one (No. 3) collected at the end of the start-up period, and five (Nos. 4–8) collected during the organic loading rates (OLR) (kg-COD/m3/d) of 1.30–1.45, 2.5–2.84, 3.73–4.10, 5.83–6.18, and 7.70–7.84, respectively. The samples were analyzed according to Zhang et al. [12]. DNA was extracted using the E.Z.N.A. Soil DNA Kit (OMEGA Biotec. D5625-01, Norcross, GA, USA) according to the manufacturer’s instructions. Partial 16S rRNA gene amplicons were generated using TransGen AP221-02 (TransStart Fastpfu DNA Polymerase, Axygen, New York, NY, USA) and ABI GeneAmp® 9700 (ABI, Carlsbad, CA, USA). Duplicate PCR products were pooled and purified using the AXYGEN gel extraction kit (Axygen, New York, NY, USA). Sequencing was performed using the 454 GS FLX+ instrument (Roche, Branford, FL, USA), and the sequencing method manual XLR70 kit. PCR products were sequenced by the Shanghai Shenggong Company (Shanghai, China).The sequenced gene was compared with the National Center for Biotechnology Information (NCBI) website, and the denaturing gradient gel electrophoresis (DGGE) imaging was analyzed using Quantity One software (Manufacture, City, State abbrev., Country). Raw 454-pyrosequencing data were analyzed using Mothur v.1.40.0 (Ann Arbor, MI, USA). Operational taxonomic units (OTU) with 97% confidence were used for the construction of OTU rank–abundance curves.Data analysis was carried out using Origin 2017 software (OriginLab, Northampton, MA, USA). To explore the correlation between microbial diversity and environmental factors, detrended correlation analysis was performed on the species.Diluted dyeing wastewater with COD concentration of 1300–1349 mg/L was used for the start-up of the HA-CMBR. The OLR was set at 1.3–1.5 kg-COD/m3/d. During the first three days of start-up, COD removal efficiency was poor with an effluent COD concentration of about 900 mg/L, after which the effluent COD gradually decreased during the following 22 days. A COD removal rate of 65% was achieved on day 23. From day 23, the COD removal rate remained stable in the following six days. Thus, the startup of HA-CMBR was deemed to be accomplished in 29 days. During the start-up period, a good chroma removal rate of 99% was achieved, which indicates that the microorganisms in the seed sludge are competent in decolorizing the dyeing wastewater.After the start-up period, the effects of OLRs on the reactor were investigated. A total of six OLRs were set during the study. Influent COD concentrations and total removal rates were 1300–1450 mg/L and 57.5%–80.8%, respectively, in stage I; 2590–2840 mg/L and 58.7%–73.8% in stage II; 3730–4100 mg/L and 57.4%–69.1% in stage III; 5830–6180 mg/L and 55.7%–68.1% in stage IV; 7700–7840 mg/L and 60.2%–63.3% in stage V; and 9080–9110 mg/L and 36.3%–37.4% in stage VI.After treatment, the average BOD5/COD of stages I–VI increased from 0.04 to 0.39, 0.27, 0.24, 0.27, 0.23, and 0.18, respectively (Figure 1). The experimental data indicate that the biodegradability of the dyeing wastewater improved significantly following treatment by the HA-CMBR. Similar to the COD removal performance, at stage V and VI, average BOD5/COD decreased by approximately 50%. The results indicate that anaerobic microorganisms were inhibited during stages V and VI.Influent chroma and the total chroma removal rate were approximately 1000 and 99%, respectively, in stage I; 2000–2500 and 99.6% in stage II; 2000–2500 and 99.6% in stage III; 3500 and 99.7% in stage IV; 4000 and 99.8% in stage V; and 5000 and 99.5% in stage VI (Figure 2). Chroma removal rates exceeded 99%. During the first four stages, the influent color changed from navy blue to almost white. After filtration through the flat-sheet ceramic membrane, the effluent turned clear, and no obvious color could be identified. Although the chroma removal rates were slightly decreased in stages V and VI, the effluent chroma still met the relevant discharge standards. The results show that the HA-CMBR has good chroma removal performance in treating the dyeing wastewater used in this study. The average TN and TP removal rates were 20% and 6% in the HA-CMBR, which are similar to the results reported by Jin et al. [24]. DO was controlled below 0.5 mg/L. pH was maintained at 7.0–8.0.During the study, the membrane flux was set at 5 L/m2/h. As mentioned previously, the test was divided into six stages. At the end of each phase, the flat-sheet ceramic membranes were chemically cleaned by pumping 1000 mg/L NaOCl into the inner space of the membranes. The membranes were soaked for 1–2 h to recover filtration performance.Figure 3 shows the TMP, SMP, and EPS changes during the study. TMP increased from 1.7 to 8.3 kPa in stage I (488%); 1.8 to 7.2 kPa in stage II (400%); 1.8 to 5.9 kPa in stage III (328%); 1.7 to 8.5 kPa in stage IV (500%); 1.8 to 5.9 kPa in stage V (328%); and from 1.6 to 7.4 kPa in stage VI (462%). The results show that the TMP of the flat-sheet ceramic membrane was restored to the initial value after chemical cleaning.EPS and SMP easily adhere to the surface of the membrane, thereby clogging the membrane pores [12,14]. As shown in Figure 3, the EPS and SMP concentrations were closely related to TMP. From stage I to stage III, there were decreasing concentrations of EPS and SMP. Correspondingly, the TMP at the end of each phase decreased from 8.3 kPa to 5.9 kPa. In stage IV, the concentrations of EPS and SMP increased, as did TMP. Thus, TMP increased with decreasing membrane flux. Zhang et al. [12] also found that the decrease in membrane flux was mainly caused by a combination of EPS, SMP, and fine sludge particles attached to the membrane pores and surface. Beyond that, there were subsequent decreases in EPS and SMP concentration during stages V and VI. EPS and SMP were found to be positively related to microbial activity [14]. During stages V and VI, microbial inhibition led to poor COD removal rates and therefore decreasing EPS and SMP. The results show that membrane fouling in this study occurred mainly due to EPS and SMP.At the end of the start-up period, microbial diversity showed an obvious change. The seed sludge showed high microbial diversity, however, after treating the dyeing wastewater (see Figure 4), bands 1, 3, 4, 6, 7, and 10 disappeared, and new bands 2, 12, 14, and 15 dominated in the HA-CMBR. Bands 5, 9, 11, 13, and 16 were further enriched with increasing NLRs (Nos. 4–8), while bands 2, 8, 15, and 17 decreased. The present findings show that bands 5, 9, 11, 13, 14, and 16 can dominate in treating dyeing wastewater.To evaluate the relationships between the bacteria in the sludge, a microbial phylogenetic tree was built as shown in Figure 5. The identified bands can be clustered into eight genera: Desulfovibrio sp., Desulfomicrobium sp., Chlorobaculum sp., Chlorobium sp., Pseudomonas sp., Halothiobacillaceae, Clostridium sp., and Alkalibaculum sp. Desulfovibrio sp. was reported to have the potential to degrade both Reactive Black 5 and Remazol Brilliant Blue R under hydrolysis acidification conditions, to deal with simulated dyeing wastewater [25]. Li et al. [26] found that Desulfomicrobium sp. was abundant in anaerobic sludge used for synergetic decolorization of Reactive Blue 13 effluent.Chlorobium sp. was reported to be dominant in mature biofilm when treating real dyeing wastewater [27]. Pseudomonas sp. was identified in a down flow microaerophilic fixed film bioreactor treating reactive azo dyes [28]. Halothiobacillaceae was detected in a sequencing batch reactor for cloth printing and dyeing wastewater treatment [29]. Clostridium sp. was reported to achieve complete decolorization of Remazol reactive dyes after 24–72 h [30]. Alkalibaculum sp. was reported to participate in the degradation of complex organic matter, such as lignocellulosic wastes [31].The relative abundances of bacterial groups (at the phylum level) in the sludge samples are shown in Figure 6. Following the increasing NLRs, Desulfovibrio sp., Deltaproteobacteria, Halothiobacillaceae, Alkalibaculum sp., Pseudomonas sp., Desulfomicrobium sp., and Chlorobaculum sp. accounted for approximately 70% of the number of individuals, of which Desulfovibrio sp. occupied the largest share. In sludge sample No. 4, the share of Desulfovibrio sp. was about 50%. Although the share of Desulfovibrio sp. decreased in sludge samples 5 and 6, it still accounted for approximately 30% on average. In sludge samples 5–8, Deltaproteobacteria accounted for 20% on average. The results indicate that Desulfovibrio sp. and Deltaproteobacteria were the most important phyla in the HA-CMBR system used for treating dyeing wastewater in this study. Those phyla with a share of less than 1% are classified as other and together accounted for 30% on average, which means that the sludges were in abundance of biological diversity.The map clustering tree based on DGGE is shown in Figure 7. In this tree, if a species in a community becomes a species with high similarity in another community, it is considered a small change, and the distance between the two communities is short. On the other hand, if a species in a community becomes another species with a low relationship, there is a large distance between the two communities. There was only 37%–43% similarity between seed sludges and the cultivated sludge sampled during the start-up period, which means that almost 60% of species evolved with the dyeing wastewater in this study. Following the increasing NLRs, the similarity increased. There was 73% similarity between sample Nos. 6 and 8, which further increased to 77% between Nos. 5 and 7. The results indicate that the bacteria in the seed sludge were able to rapidly adapt to the changing feed.Microbial diversity was analyzed using the Shannon–Weaver index (see Table 2). The diversity indexes of sample Nos. 1 and 2 are 1.412 and 2.325, respectively. Nos. 3–8 have low diversity indexes of 1.935–1.999. No. 1 was collected from a running reactor in the lab, and No. 2 was collected from a real dyeing wastewater treatment plant. The results show that the seed sludge from the real wastewater treatment plant had high microbial diversity. However, following cultivation, microbial diversity decreased, indicating that certain species became dominant at the expense of overall diversity, as shown in Figure 6.In this study, good chroma removal is deduced to derive from the new strains developed in the HA-CMBR. There are many types of dyes, potentially requiring different types of microorganisms for decolorization. However, in many cases, microbial information is lacking, which is one of the difficulties faced in biodegradation of dyeing wastewater. In this study, microbes that were not contained in the seed gradually emerged and were further enriched after cultivation. This indicates that the bacterial population in the sludge can be changed and used for treating new types of dye wastewater through efficient cultivation. An HA-CMBR can reduce sludge loss by 100%, which is an efficient means of sludge enrichment. Therefore, HA-CMBR in this study is of practical value in treating dyeing wastewater.HA-CMBR was used in treating dyeing wastewater. After seven days of operation, the COD removal rate reached approximately 64%, and the chroma removal rate was as high as 99%. COD removal rate decreased sharply when influent COD concentration exceeded 9000 mg/L. The microbial community in the sludge samples showed obvious changes, and the main strains in the HA-CMBR were Desulfovibrio sp. (24%), Deltaproteobacteria (15%), and Turicibacter sp. (37%). The tested AnMBR demonstrates energy-efficient treatment of dye wastewater, and might further enable treatment of different effluent types (through rapidly evolving microbial community structure) and avoidance of biogas emissions/remediation costs.Conceptualization, D.W. and W.Z.; methodology, Y.J.; software, Y.J.; validation, Y.J., D.W. and W.Z.; formal analysis, Y.J.; investigation, W.Z.; resources, Y.J.; data curation, Y.J.; writing—original draft preparation, Y.J.; writing—review and editing, W.Z.; visualization, D.W.; supervision, W.Z.; project administration, W.Z.; funding acquisition, W.Z.This research was supported by the National Natural Science Foundation of China [grant numbers 51638006; 51668013]; Guangxi Science and Technology Planning Project under Grant No. GuiKe-AD18126018; Special Funding for Guangxi “BaGui Scholar” Construction Projects.The authors declare no conflict of interest.BOD5/chemical oxygen demand (COD) changes during the study (Inf., Influent; Eff., Effluent).Color changes during the study (left: Influent; middle: Treated by anaerobic microorganisms; right: Filtrate).Trans-membrane pressure (TMP), extracellular polymeric substances (EPS), and soluble microbial products (SMP) changes during the study.DGGE photos.Microbial phylogenetic tree (the ruler length represents 10% divergence. The number of nodes represents confidence).Microbial population relative concentrations during the study.Map clustering tree based on DGGE.Compositions of dying wastewater.Biodiversity change during the study.

Med-MDPI/ijerph_3/ijerph-16-05-00778.txt

@@ -0,0 +1 @@
+Background: Workplace or campus wellness/obesity-prevention policies and initiatives can improve health. Research tools to assess worksite or campus policies/initiatives are scarce. Thus, the aim of this research is to develop and validate the policies, opportunities, initiatives, and notable topics (POINTS) audit. Methods: POINTS was developed and refined via expert review, pilot-testing, and field testing. Trained researchers completed a web-based review from a student-focus or employee-focus regarding 34 health-promoting topics for colleges. Each topic was evaluated on a 0–2 scale: 0 = no policy/initiative, 1 = initiatives, 2 = written policy. When a written policy was detected, additional policy support questions (administered, monitored, reviewed) were completed. Results: Cronbach’s Alpha for the student-focused POINTS audit was α = 0.787 (34 items, possible points = 65), and for the employee-focused POINTS audit was α = 0.807 (26 items, possible points = 50). A total of 115 student-focused and 33 employee-focused audits were completed. Although there was little evidence of policy presence beyond stimulant standards (smoking and alcohol), there were extensive examples of health initiatives. The student-focused POINTS audit was validated using the Healthier Campus Initiative’s survey. Conclusions: POINTS is a web-based audit tool that is valid and useful for pre-assessment, advocacy, benchmarking, and tracking policies for health and well-being for students (campus) and employees (worksite).Over one-third of adults in the United States are obese [1]. Researchers have shown environmental factors influence weight status [2,3,4]. Unfortunately, much of the current evidence for the college environment suggests that both students and employees default to sedentary and poor dietary intake behaviors [5,6,7,8]. Further, other studies have found that healthy work environment initiatives can improve employee wellness and reduce employer health-related expenses [9,10]. Initiatives are interventions or programs intended to encourage healthy behaviors and decisions. Whereas, a policy is a written and published document outlining a definite course or method of action to determine and guide present and future decisions. Policies are most effective if they have defined goals and procedures for implementation, including a charged department or individual responsible for their implementation. Policies promoting a healthy workplace may improve dietary intake, decrease sedentary behavior, and increase overall health-promoting behaviors [10,11,12].Colleges as a workplace are required to mandate employee policies regarding overtime, medical leave, and occupational health and safety [13], and are audited by state and sometimes federal agencies. The Center for Disease Control and Prevention urges employers to implement health promotion policies in the workplace [14]. Numerous resources and best practice recommendations for workplace health and wellness policies are available for employers through federal, state, and other agencies [15,16,17,18]. However, nutrition and wellness policies are only mandated in the public-school system [19,20,21]. An extensive body of literature exists regarding the evaluation of school wellness policies [16,22,23,24,25,26,27] and many of these policy evaluation tools, although comprehensive [16,23,25,26], are tailored for the elementary or high school environment. Tools for assessing worksite or college policies/initiatives are lacking. Using a 21-question yes/no survey, one study assessed worksite healthy supports and policies [28]. However, only two of the 21 questions were about policies, the remaining items were environmental supports or interventions. On an international level, another study conducted semi-structured interviews with key stakeholders to evaluate the extent to which nutrition topics and policies were implemented [29]. The study authors developed and used a policy assessment tool based upon the “four Ps” marketing approach (price, product, place, and promotion) for health or nutrition policy. The researchers concluded that mandatory (policies/laws) versus voluntary initiatives were more effective for improving health yet less obvious.Likewise, the American College of Health Association encourages colleges to set and track effective health goals for their campuses [30]. The Partnership for a Healthier America, specifically the Healthier Campus Initiative (HCI) [31], established 41 guidelines, with the criteria that campuses meet at least 23 of the 41 guidelines to be designated a Healthier Campus. However, currently there is no tool designed to assess the extent to which wellness and obesity-prevention policies, in general, are implemented for college campuses. The HCI lists specific policies and initiatives with a yes/no evaluation, which fails to evaluate other examples of policies or the level of policy integration or support. The college campus is a unique environment in that it serves as both a learning institution and workplace. Some college campuses are similar in size to a city/village and are typically one of the largest employers in many communities. A policy audit tool appropriate for this environment may also be effective in a variety of similar educational/work settings. Well implemented wellness and obesity prevention policies and initiatives can greatly improve health habits of both students and employees [27,32].Tools exist to evaluate wellness policies in public schools and government entities; however, no tool exists to evaluate wellness policies or initiatives for educational workplaces or college settings. The purpose of this paper is to describe the development, field testing, and validation of the policies, opportunities, initiatives, and notable topics (POINTS) audit for college campuses and worksites.This paper describes the two phases used in the development of the POINTS. For phase one, instrument development, the audit was developed using a three-step process: (1) Inventory item development; (2) expert, pre- and pilot-testing, and audit revisions; and (3) field implementation. For phase two, instrument validation, POINTS was validated using the Healthier Campus Initiative’s (HCI’s) [31] 41 guidelines. Data were collected between 2015–2017 and analyzed in 2018. Syracuse University’s Institutional Review Board determined this research to be exempt because this was environmental, not human research.To develop the audit, the authors completed a thorough review of the literature and health expert policy recommendations. In addition to the peer-reviewed literature, the authors searched online and considered workplace and school wellness recommendations made by government agencies, non-profit health organizations, and health professionals. For purposes of this audit, the following definitions clarify policy, initiative, and a pledge:Policy: A written and published document outlining a definite course or method of action to determine and guide present and future decisions. Policies may have defined goals and procedures for implementation, including a charged department or individual responsible for their implementation.Initiative: A series of interventions or programs intended to encourage healthy behaviors and values. Initiatives may or may not contain any defined goals, procedures, or plans for implementation.Pledge/Commitment: A written and published agreement that is not specifically designated as being a policy. Pledges/commitments may or may not have defined goals or procedures for implementation.Policy: A written and published document outlining a definite course or method of action to determine and guide present and future decisions. Policies may have defined goals and procedures for implementation, including a charged department or individual responsible for their implementation.Initiative: A series of interventions or programs intended to encourage healthy behaviors and values. Initiatives may or may not contain any defined goals, procedures, or plans for implementation.Pledge/Commitment: A written and published agreement that is not specifically designated as being a policy. Pledges/commitments may or may not have defined goals or procedures for implementation.Thirty-four health, wellness, obesity-prevention, and sustainability topics were extracted from the literature review. The 34 topics (Table 1; to see the wording of each question, refer to Supplementary Materials S1 were content analyzed and grouped into the seven categories accordingly:Stimulant Standards: Smoking independently increases the risk of cardiovascular disease [33]. Excessive alcohol consumption increases the risk of weight gain [34], and other deleterious outcomes for students [35,36]. Randomized control trials found positive results for workplace wellness initiatives and environmental supports that encourage smoking cessation [37,38].Chronic Disease Management and Health Promotion: Numerous studies including health, nutrition, and/or physical activity education improved health outcomes [39,40,41]. Worksites with healthy environmental policies and initiatives reduced medical costs and increased savings [9].Healthy Student Course Requirement: Health and nutrition education programs effectively increased college students’ physical activity, fruit/vegetable intake [42,43], and their overall knowledge about nutrition [44].Health and Wellness Services: Workplaces that have wellness departments or professionals were more likely to have wellness programs and policies. [45]. Employees were more likely to participate in physical activity and make healthier choices if they were incentivized (i.e., rewards/prizes or lower health insurance premium rates) [46,47].Active Living: Employees were more likely to partake in non-work physical activity in safe and well-maintained environment (i.e., sidewalks and stairwells) [39,48,49,50,51]. Soler and colleagues encouraged workplaces to adopt numerous environmental policies which support safe environments and encourage physical activity [52]. The growing concern over carbon emissions has also motivated policy makers and key stakeholders to create and implement policies that discourage driving and encourage walking or biking to work [53,54].Nutritious and Sustainable Food Ways: Strong evidence exists regarding the relationship between the food environment and healthy eating patterns [55,56]. Behavioral economics and nutrition food policies can reduce obesity and positively influence dietary habits [57,58,59,60]. National nutrition policies for public schools receiving federal funding have been required for decades, and some states now have nutrition menu labeling policies. There are limited healthy choices in restaurants, stores, and vending machines on- and near-college campuses [61,62], so access to a farmer’s market [63,64] and food procurement policies on campus might positively effect non-communicable disease risk [65,66,67]. Sustainability is of growing concern on campuses and can have both health and environmental benefits [53,68,69].Healthy Student Living: Some environmental research suggests a health benefit to living on campus as compared to living off campus [70,71], so the policies for on-campus housing, dining hall contracts, and food security initiatives are important to review.Stimulant Standards: Smoking independently increases the risk of cardiovascular disease [33]. Excessive alcohol consumption increases the risk of weight gain [34], and other deleterious outcomes for students [35,36]. Randomized control trials found positive results for workplace wellness initiatives and environmental supports that encourage smoking cessation [37,38].Chronic Disease Management and Health Promotion: Numerous studies including health, nutrition, and/or physical activity education improved health outcomes [39,40,41]. Worksites with healthy environmental policies and initiatives reduced medical costs and increased savings [9].Healthy Student Course Requirement: Health and nutrition education programs effectively increased college students’ physical activity, fruit/vegetable intake [42,43], and their overall knowledge about nutrition [44].Health and Wellness Services: Workplaces that have wellness departments or professionals were more likely to have wellness programs and policies. [45]. Employees were more likely to participate in physical activity and make healthier choices if they were incentivized (i.e., rewards/prizes or lower health insurance premium rates) [46,47].Active Living: Employees were more likely to partake in non-work physical activity in safe and well-maintained environment (i.e., sidewalks and stairwells) [39,48,49,50,51]. Soler and colleagues encouraged workplaces to adopt numerous environmental policies which support safe environments and encourage physical activity [52]. The growing concern over carbon emissions has also motivated policy makers and key stakeholders to create and implement policies that discourage driving and encourage walking or biking to work [53,54].Nutritious and Sustainable Food Ways: Strong evidence exists regarding the relationship between the food environment and healthy eating patterns [55,56]. Behavioral economics and nutrition food policies can reduce obesity and positively influence dietary habits [57,58,59,60]. National nutrition policies for public schools receiving federal funding have been required for decades, and some states now have nutrition menu labeling policies. There are limited healthy choices in restaurants, stores, and vending machines on- and near-college campuses [61,62], so access to a farmer’s market [63,64] and food procurement policies on campus might positively effect non-communicable disease risk [65,66,67]. Sustainability is of growing concern on campuses and can have both health and environmental benefits [53,68,69].Healthy Student Living: Some environmental research suggests a health benefit to living on campus as compared to living off campus [70,71], so the policies for on-campus housing, dining hall contracts, and food security initiatives are important to review.Although the audit is tailored primarily for the student population, as we reviewed the literature we noted the topics unique to a student versus employee population. Using skip logic, the audit can be used to evaluate the existence and extensiveness of policies affecting the employee population. Table 1 indicates which topics are included in each audit version.Each audit question was scored on a three-point semantic differential scale to assess each policy topic (0 = no policy; 1 = initiative/interventions; 2 = written policy) [16,23]. When a policy was identified, the written policies were further assessed for the total comprehensiveness of the policy including: defined mission/goals, policy outcomes, implementation plan, a department charged with implementation, defined sanctions/fines for policy violation, monitoring/evaluation plan, and policy review plan. Additionally, when a policy was identified, the evaluator submitted a copy of the policy (via an URL link) on the Qualtrics survey. See Figure 1 for an example.POINTS was reviewed by ten experts in nutrition, health promotion, physical activity, and public health from various institutions to establish content validity. The POINTS audit was also cognitively tested with five research assistants to ensure the items were interpreted accurately. Cognitive testing and expert review resulted in improved wording of questions and semantic-differential response choices.For the pre-test, the lead authors interviewed three wellness and obesity-prevention professionals at a university located in the northeast regarding the campus health and wellness policies. Interviews were conducted in summer 2015. The professionals answered open-ended and non-leading questions such as “What policies exist on campus regarding food nutrient standards for the campus population?” The professionals were unable to identify the difference between a policy and initiatives. Monitoring and evaluating the outcomes of initiatives and policies were scarce.After refinements, POINTS was developed into the online survey for health promotion professionals. This survey was pilot-tested in fall 2015 at 15 US college campuses. The authors identified 51 wellness and obesity-prevention professionals within their universities to be contacted (one to four per campus). All 51 professionals were contacted via telephone by undergraduate/graduate student research assistants. The professionals were given a brief explanation of the survey and invited to partake in the research. If they agreed to participate, they were sent a website link to the survey via email. The professionals were directed to only complete questions that pertained to their job title and duties (e.g., foodservice manager—Nutrition and Sustainable Food Ways section) but often each professional completed the whole survey. When the discrepancy was detected, we reviewed their policies via a web-search to determine correctness of responses. Since professionals answered from their own perspective, without doing any additional research to verify their answers, there was very little agreement between different health promotion experts on a campus. Because of these limitations in the professional survey, the research team decided the audit should be completed as a web-review by trained research assistants.Data were collected in spring 2016 through spring 2017 at a total of 115 campuses. Campuses (n = 80) participating in the Get FRUVED [72] project collected the student-oriented POINTS audit as part of their participation in the social marketing and environmental intervention. Thirty-five additional campuses were identified and evaluated by the lead institution’s research assistants for both the student-focus and employee-focus audits. Researchers were trained to complete web searches to identify policy statements using the three-point semantic differential system. Data were collected through online survey software Qualtrics ™ (Qualtrics, Seattle, WA, USA) and proof of policy was the submission of the webpage URL links.Training and interrater reliability (IRR): Research assistants completed online video-based training that taught them how to: (1) Prepare for a successful audit; and (2) interpret and answer each audit question with respect to the varied web environments. Then, they practiced using the POINTS audit on two different school websites. Subsequently, they independently used the POINTS audit to evaluate two new university campuses, which were not included in practice sessions, to establish IRR. The data were compared to the standard set by the lead institution. POINTS audits were repeated until all data collectors on a campus achieved an IRR > 0.80, before they commenced with data collection. As more independent schools joined the data collection in early 2017, the IRR protocol was changed to an online quiz.In addition to a total POINTS score, sub-scores were created for each of the categories on the audit: Stimulant Standards, Chronic Disease and Health Promotion, Healthy Student Required Classes, Health and Wellness Services, Active Living, Nutritious and Sustainable Food Ways, and Healthy Student Living. The policy support score was the total comprehensiveness of the policies; the summation of the eight follow up questions when a policy existed—defined mission/goals, policy outcomes, implementation plan, a department charged with implementation, defined sanctions/fines for policy violation, monitoring/evaluation plan, and policy review plan. SPSS (version 24, IBM, Armonk, NY, USA) was used to run non-parametric statistics, t-tests, and ANOVA. Level of significance was set at p < 0.05.A total of 115 student-focused and 33 employee-focused audits were collected by trained student research assistants, who had satisfactory IRR (α = 0.783). The mean time to collect and enter the data was 3.75 ± 3.6 h; median 2.5 h. Cronbach’s Alpha for the student-focused POINTS audit was α = 0.787 (35 items, total potential points = 65), and for the employee-focused POINTS audit was α = 0.807 (26 items, total potential points = 50). More of the audits were collected from public institutions for both the student- and employee-focused audits (85% and 63.6%, respectively) (Table 2). For the student-focused audit, the highest percentage of audits were collected from the southeast region (40%), followed by the midwest and northeast; the least were collected from the southwest (3.5%). For the employee-focused audits, the geographic distribution was similar, whereas the smallest percentage of audits was collected from the northwest (6.1%).Average student population was 18,952 for the student-focused POINTS audit, with the employee-focused audits slightly larger at 21,297 students. Based on the distribution of campuses by size, schools were grouped by student population size. Very small campuses had a student population <4500 students. Small schools had 4501 to 12,500 students. Midsized schools were defined by a population of 12,501 to 17,500. Large schools had 17,501 to 29,000 students, and very-large schools had >29,001 students. School characteristics data are listed in Table 2.For the student-focused POINTS audit, almost all campuses (at least 90%) had smoking and alcohol/substance abuse policies (Table 3). Dining hall contracts and on-campus living policies were in place for a moderate percentage of schools (65.2% and 50.4%, respectively). Policy presence evidence was detected for health and wellness departments (37.4%), insurance premium incentives (29.6%), designated eating environments (17.4%), healthy campus fund raising (15.7%), and health education for credit (12.2%). The remaining 22 topics had less than 10% of schools with evidence of policy presence; however, at least 75% of the schools had intervention presence for 13 of the topics (non-credit health, nutrition or physical education; health screenings, environmental supports for active living, closed campus, sustainable transportation, healthy food options, local and sustainable food, organic waste reduction and disposal, farmer’s markets, campus garden, and an open campus).For the employee-focused POINTS audit, at least 93% of the campuses had smoking and alcohol/substance abuse policies (Table 3). Additional policy presence evidence was detected only for insurance premium incentives (27.3%) and health and wellness departments (12.1%). The remaining 20 topics had only one–two schools with evidence of policy presence; however, at least 75% of the schools had intervention presence for 15 of the topics (non-credit health, nutrition or physical education, health habit challenge, health and wellness services, environmental supports for active living, closed campus, sustainable transportation, healthy food options, designated eating environment, local and sustainable food, organic waste reduction and disposal, and farmer’s markets).The Total POINTS score for student-focused audits indicated more interventions than policies with a mean 30.13 ± 5.48, maximum possible 65; and low overall policy support with a mean of 23.2 ± 12.2, maximum possible 320) (Table 4). Sub-score means are an indication of the total number of interventions and policies present for each of the sub-categories. The Stimulant Standards had the highest evidence for policy/intervention presence (3.8 ± 0.5; max possible 4) and policy support (8.1 ± 3.6; max possible 16). Healthy Student Living followed closely behind (4.51 ± 1.01; max possible 7), but had significantly less policy support (4.0 ± 4.1, max possible 24). Health and wellness services on campuses were supported by a balance of policies and interventions (2.92 ± 1.31, maximum possible 6) and reasonable professional and policy support (7.4 ± 5.2, max possible 32). Although few policies existed for individual Nutritious and Sustainable Food Ways (Table 3), the mean score for campuses conveyed the high degree of interventions in place (7.46 ± 2.01, max possible 18) and a lack of policy support. The remaining three sub-categories: Chronic Disease and Health Promotion, Healthy Student Requirements, and Active Living had lower policy/intervention evidence and consequently very low policy support. A few significant differences were evident based upon demographic variables. Private schools as compared to public had significantly less evidence for policy presence for Active Living (2.57 ± 0.86 vs. 3.09 ± 1.10, p < 0.01); and Chronic Disease and Health Promotion (5.20 ± 2.35 vs. 6.46 ± 2.03; p < 0.01). Although there were no differences in total score or sub-score by geographic region, differences were detected by campus size. The smallest schools’ means were significantly lower than all school categories for Chronic Disease and Health Promotion (very Small schools: 4.09 ± 0.58 vs. (small: 6.0 ± 0.47, to 7.13 ± 0.30 for very large schools) p < 0.01. The smallest schools had scores significantly lower than all schools larger than 12,501 students on Active Living (very small: 2.24 ± 0.32 vs. (moderate sized schools: 3.10 ± 0.17 to 3.46 ± 0.17 for very large schools) p < 0.01 and on total POINTS, very small schools: 26.67 ± 1.68 vs. (Moderate sized schools: 30.76 ± 0.82 to 32.8 ± 0.85 for very large schools), p < 0.01).There was an equal split between the universities with contracted food service departments (n = 58) and those independently run by their campus (n = 57). Although there was no significant difference in the total Nutritious and Sustainable Food Ways subscale total, contracted departments had significantly more policy support than independently run systems (2.51 ± 3.9 vs. 0.94 ± 1.73 respectively; p = 0.006.The Total POINTS and sub-category scores for the employee-focused audits indicated very similar results to the student-focused audits (Table 4). The employee-focused POINTS audit was significantly lower for private institutions as compared to public, and had significantly less evidence for policy presence for Chronic Disease and Health Promotion (5.08 ± 2.50 vs. 7.00 ± 1.92, p < 0.05); Employee Health and Wellness (1.91 ± 0.51 vs. 2.48 ± 0.68; p < 0.05) and total POINTS (20.67 ± 4.92 vs. 25.10 ± 3.42; p < 0.05). There were no significant differences by campus size or geographic region.In collaboration with the Partnership for Healthier America’s Healthier Campus Initiative (HCI) [31], participating institutions (n = 60) in Get FRUVED [72] completed both POINTS and the HCI survey. The HCI survey was created for this study based upon the HCI guidelines and was chosen for validation, because it measures comparable concepts for the college campus. All audits completed were for the student-focused population.The HCI survey contained 41 questions, 15 regarding food and nutrition, 19 regarding physical activity, and seven regarding programming. Programming topics were similar to the POINTS Chronic Disease and Health Promotion category. Each question was a Yes/No check off to indicate if a campus had the initiative or policy. A summary of the topics assessed are listed in Table 5, with a more detailed listing of the questions included in Supplementary Materials S2.Each HCI category and the total scores were tallied. Spearman’s correlations were run on POINTS and HCI survey totals and sub-scores.Fifty-six of the 60 Get FRUVED intervention schools had matched POINTS and HCI data. Most of the schools were public institutions (76.8%) (see Table 6). The southeast represented 41.1% of the sample, while only 3% were from the southwest. Half the sample (50%) were from smaller schools with ≤ 12,500 students.There were significant correlations between total POINTS and the HCI total (r = 0.519, p < 0.01), and total Policy Support and the HCI total (r = 0.478, p < 0.001), as well as between Nutritious and Sustainable Food Ways (POINTS) and the Food and Nutrition HCI sub-score (r = 0.314, p = 0.019). There were no significant correlations between POINTS for Chronic Disease Health Promotion and Programming (r = 0.101, p = 0.46) or between both tools for Physical Activity—Active Environment (r = 0.210, p = 0.12).As a result of this research a new tool was produced to assess the extent of health and wellness policies and initiatives on college campuses supporting students’ and employees’ health. Through a web review of policies and interventions, student research assistants were able to reliably complete student-focused or employee-focused audits. The tool provides a simple total score and sub-scores for ease of interpretation. In general, college campuses lack policy support for health and wellness-related areas beyond smoking and alcohol. POINTS was designed to capture what health promotion/disease prevention policies should be on record, so one would not expect a significant number of institutions to have an extensive list of these policies implemented. There were some differences noted by campus size and public/private status. It was encouraging to note the extensiveness of the interventions in place, but to make them sustainable and enforceable, policy support must be implemented (mission, enforcement, and monitoring). POINTS was also validated by experts, and for the student-focused audit, by comparison to the HCI-recommended guidelines.Policy evaluation tools exist for schools, day care centers [16,23,73], and communities [25,26,29], but not for worksites/universities. The POINTS tool is easy to use and objective. It is conducted by searching for evidence of policy/initiative presence online. With training, students effectively implemented this protocol. The average time for data collection was 3.75 hours per institution, but the median was 2.5 hours. Differences in website design and evaluator approach to the task would lead to wide range effort.To the authors’ knowledge, comparisons in the literature for this study do not exist. Much of the literature regarding policy research has been with school districts, not college campuses or worksites. In those studies, reviews were completed to determine if school wellness policies existed [22,74]. There was one report of an intervention to improve the enforcement of a policy (i.e., after-school snack policy) [75], implying sometimes policies need intervention support to verify their need/effectiveness. Some studies found more policy support in places for where there was a higher need (school districts with higher free lunch participation and/or obesity) [76,77]. Many times, the papers focused on factors affecting their ability to implement policies [19,74,75,77].Policies can make a difference. For all Minnesota school districts, more comprehensive physical activity policies were associated with greater exercise [76]. States with stronger laws restricting advertising and competitive foods/beverages sales had reduced incidence of obesity [27]. In another study, restricting sugar-sweetened beverages was most effective for improving milk intake [78].Based upon a systematic review for worksites, there has been limited evidence for the effectiveness for worksite environmental interventions and very few policies on record to support healthful behaviors [51]. However, in more recent studies, healthy work environment interventions improved employee health and reduced employer health-related expenses [9,10]. Health risk assessments with feedback and education have proven to be effective for improving a variety of health parameters [41] and; therefore, should be considered as a policy for worksite environments.For either a student-focused or employee-focused environment, prior research provides some guidance regarding effective policy efforts. Dodson and colleagues [39] found the more structured and specific the worksite policies, the more likely the employees were to achieve exercise recommendations. Similarly, college campuses might require physical (or nutrition/health) education classes. In this study, there were less than 15% of the schools that had such policies. Historically these classes were required on more campuses but the trend for such requirements has been declining [79,80,81]. Other studies provide evidence that nutrition environment policies for catering and point of purchase labeling affect behavior [55] and taxing incentives (subsidy for healthy and tax for unhealthy) might be effective and affect predicted buying behavior and nutrient quality [56].The POINTS research found smaller campuses had lower policy evidence and supports. Similarly, Brissette and colleagues [28] found of the 832 worksites evaluated, smaller companies had less cardiovascular health policies. When a company indicated a wellness committee/coordinator, more policies were evident, indicating the importance of staffing supports for policy implementation.There is a need for more accuracy and consistent definitions regarding policies versus initiatives. Interventions and initiatives may be effective resulting in appropriate behavior change, but without policy support, they are often temporary and fleeting. Even with policy support, one study evaluating the effectiveness of school wellness policies and practice found very low agreement between the written policy and nutrition-related practices [24]. Having a policy on record is just the first step. The degree of policy support, specifically how well it is managed, enforced, monitored, and reviewed are important determinants of effectiveness. In a study of the Minnesota school districts [76], those with higher levels of poverty and obesity implemented higher quality school wellness policies in terms of strength and comprehensiveness. Reviewing the history for smoking policies, we can trace the successes and challenges with policy implementation. While Hopkin and colleagues [37] found effective policies reduced tobacco use, based on a systematic review, weak evidence was secured for the effectiveness of strategies for enforcing smoke-free policies [82]. Best practices for implementing policies include securing administrative buy-in, relationship building, conveying effectiveness, and conveying financial sustainability [83,84].The strength of the POINTS tool is that the policy items assess the health promotion and wellness concepts that should be implemented in worksites and on college campuses. In addition, a clear distinction is made between policies and interventions/initiatives, and it is specific yet flexible enough to be effective for a variety of student and employee populations. The POINTS audit and an extensive training with an IRR quiz are online. The user is provided with feedback and comparative results. Users can pre-assess their policies and interventions, advocate for changes, and track their progress over time.A limitation is the small sample size and the disproportionate representation by campus size/geographic location. Additionally, student research assistants implementing this internet-based audit might lack access to some policies which might be stored on a campus intra-net system. Finally, although the student-focused audit was validated against the HCI and had moderate correlations for the total and the food/nutrition scores, there were no correlations for the physical activity or health promotion categories. Future studies need to compare POINTS to student and employee health data and, validate the employee-focused POINTS audit against another tool, and with a diversity of campus and work environments.POINTS is a web-based audit tool that is valid and useful for pre-assessment, advocacy, benchmarking, and tracking policies for health and well-being for students and employees. The results of this study should act as motivation to implement high-quality health and wellness policies on campuses and worksites, as this tool provides a way to monitor progress and improvement.The following are available online at https://www.mdpi.com/1660-4601/16/5/778/s1, S1: POINTS audit questions, S2: healthier campus initiative survey questions.All the authors have made substantial contributions (a) to either conception and design, or acquisition of data, or analysis and interpretation of data, and (b) to drafting the article or revising it critically for important intellectual content, and (c) on final approval of the version to be published, and agree to its submission. Specifically, T.M.H. and M.S. designed the study. All authors pilot-tested the tool and acquired data, and T.M.H. and E.D.Y. analyzed and interpreted the data. T.M.H. drafted the article and all authors revised it. All authors provided final approval of the version to be reviewed and agreed to its submission.Funding provided by Agriculture and Food Research Initiative Grant no. 2014-67001-21851 from the USDA National Institute of Food and Agriculture, Get FRUVED: A peer-led, train-the-trainer social marketing intervention to increase fruit and vegetable intake and prevent young adult weight gain, A2101. Partial funding was also provided by South Dakota State Agriculture Experiment Station. The funders had no role in the design, analysis, or writing of this article.We would like to acknowledge (1) the technical support for data collection and training provided by Megan Mullin, Laura Brown, and Heather Brubacker; and (2) all of the research assistants at each institution who collected data.The authors declare no conflicts of interest.An example of the phases of identifying a policy with POINTS, regarding nutrient standards.Categories and topics assessed via the policies, opportunities, initiatives, and notable topics (POINTS) audit 1.SmokingAlcoholHealth education not for creditNutrition education not for creditPhysical education not for creditHealth promotion—all forms of mediaHealth fairsHealth screeningsChronic disease educationHealth habit challengesHealth education for creditNutrition education for creditPhysical education for creditCampus health and wellness departmentHealthy campus fundraising 2Healthy employee insurance premiumsPhysical activity during work hoursActive environments (i.e., bike lanes, stairs)Closed campusSustainable transportationHealthy food optionsNutrient standardsHealthy food labels and point-of-purchase nutrition infoFood taxes and subsidiesDesignated eating environmentsLocal and sustainable foodOrganic waste reduction and disposalFarmers marketsLocal food access on-campusCampus food gardensOn-campus housingOpen campusDining hall contractsFood security initiatives1 To see the wording of each question, refer to Supplementary Materials S1. 2 Questions factored only into the student version of the POINTS audit.Characteristics of schools with completed POINTS audits.Frequency of policy and intervention presence for the student-focused and employee-focused POINTS audits.1 Questions added only into the student-focused version of the POINTS audit. NA: This question is Not Applicable to the employee population.Mean sub-score, total, and policy support POINTS scores.1 Public institutions scored significantly higher than private institutions, p < 0.01. 2 Scores are significantly different by school size, p < 0.01; 2a Very small schools < 4500 students scored lower than all school size categories; 2b Very small schools < 4500 students scored lower than all schools with > 12,501 students. 3 For Health and Wellness Services: The number of health professionals indicated is added into policy support. 4 Public institutions scored significantly higher than private institutions, p < 0.05.Healthy Campus Initiative’s (HCI’s) survey topics.offer wellness mealssufficient whole foods (dining and catering)healthier dessertssufficient healthy beverages (dining and catering)plant-based foodstray-less systemprovide healthy food labelshealthier vending and cateringfree waterRegistered Dietitian Nutritionist assessments/counselinglimitations on fried foodsimplement local procurementoffer bike share/rentalsufficient fitness/intramural opportunitiesmonthly intro to movement classes,at least one 15 min physical activity break each dayfitness orientationssufficient outdoor activitiesrental for outdoor equipmentsufficient outdoor recreation clinics/tripsprovide marked walking routespedestrian crossingsufficient bicycle parking spaces sufficient free access to fitness/recreation centerdedicated physical activity spaceoutdoor running/walking track outdoor fitness systemcertified personal trainersimplement bicycle/pedestrian accommodation policypublic transportation incentivesimplement integrated/comprehensive wellness programmandatory health and wellness educationfood insecurity program/policybreastfeeding program/policyhealth/wellness service learning opportunitiesoffer rewards or rebates for insurance premiumshealthy cooking classesCharacteristics of schools participating in the validation study.

Med-MDPI/ijerph_3/ijerph-16-05-00779.txt

@@ -0,0 +1 @@
+There is growing research interest in emergency logistics within the operations research (OR) community. Different from normal business operations, emergency response for large scale disasters is very complex and there are many challenges to deal with. Research on emergency logistics is still in its infancy stage. Understanding the challenges and new research directions is very important. In this paper, we present a literature review of emergency logistics in the context of large-scale disasters. The main contributions of our study include three aspects: First, we identify key characteristics of large-scale disasters and assess their challenges to emergency logistics. Second, we analyze and summarize the current literature on how to deal with these challenges. Finally, we discuss existing gaps in the relevant research and suggest future research directions.Large scale disasters, such as Haiti’s earthquake in January 2010 and Japan’s biggest earthquake, tsunami and nuclear reactor meltdown in March 2011, often happen suddenly and cause large casualties and significant damages to society. In general, a disaster can be defined as “a shocking event that seriously disrupt the functioning of a community or society, by causing human, material, economic or environmental damage that cannot be handled by local agencies through standard procedures” [1]. When a large-scale disaster happens, immediate emergency responses are needed in order to save lives and relieve and control the damages [2]. As pointed out by Altay and Green III [3], “Disasters are large intractable problems that test the ability of communities and nations to effectively protect their populations and infrastructure, to reduce both human and property loss, and to rapidly recover. The seeming randomness of impacts and problems and uniqueness of incidents demand dynamic, real-time, effective and cost efficient solutions, thus making the topic very suitable for OR/MS research.” Emergency logistics is “the support function that ensures the timely delivery of emergency resources and rescue services into the affected regions so as to assist in rescue activities [4]” while humanitarian logistics is more focusing on aiding people in their survival during and after a disaster. However, in the research arena, the differences between emergency logistics and humanitarian logistics have been slight [5]. In this work, we will use the term emergency logistics as a general term, and do not emphasize the differences between them.Many scholars in the OR community have studied emergency logistics, especially after the 2001 9/11 terrorist attack in the U.S. [6,7]. We were able to find six survey papers on emergency logistics that are related to our study. Green and Kolesar [8] analyzed previous OR papers focused on urban emergency services and regular emergencies, published in the INFORMS journals from 1960 to 2004. Wright et al. [9] extended the literature scope into homeland security such as traffic and cyber space safety. Both of these studies focused on OR in routine emergency management, but not in the context of large-scale disasters. Altay and Green III [3] summarized the works of the OR community published from 1980 to 2006 under a broad umbrella of disaster operation management (DOM) in large-scale disasters. They provided a comprehensive literature classification in six dimensions: authors’ affiliation, disaster type, solution methodology, operational stage, research contribution type, and a discipline classification of management science, management engineering, and management consulting based on the extension of Denizel et al. For further research, they highlighted the gaps in multi-agency coordination, soft OR, the use of sensing technology, recovery planning, business continuity, and management engineering. Simpson and Hancock [10] reviewed the operational research (OR) foundation in emergency response from 1965 to 2007. They discussed the previous literature on both urban emergency service systems studied in earlier periods and of large-scale disasters in the 21st century, and they also identified a detailed literature citation network among those studies. Four major areas of opportunity were suggested: soft versus hard OR, information and DSSs, volunteers and temporary organizational structures, and performance metrics in the context of emergency response. Caunhye et al. [11] reviewed optimization models in emergency logistics in the pre-disaster operations phase and post-disaster operations phase. Using content analysis approach, they analyzed the current literature in detail through the perspectives of OR models, decisions, objectives, and constraints. They also suggested some future research problems such as post-disaster dynamic inventory modeling, combining aspects of transportation time, injury seriousness, on-field treatment, and medical center service load for casualty transportation, adding objective measures for coordination effectiveness and proper organizational structure, and taking human behavior uncertainty in disaster into consideration. Most recently, Galindo and Batta [1] reviewed the progress and research gap of OR/MS research in DOM after Altay and Green’s [3] review in 2006. They followed the same review framework of Altay and Green [3] but added an analysis about the most common assumptions in the field. They classified these assumptions into three categories: reasonable, limited, and unrealistic, and emphasized the importance of assumption validation. Finally, they suggested further research directions including: (i) improvement of the coordination among DOM actors; (ii) introduction of new technologies through more application studies; (iii) study of DOM problems using formal statistical analysis to establish realistic assumptions in DOM models; (iv) in-depth exploration of methodologies such as Soft OR and interdisciplinary techniques; and (v) measurement of the effectiveness of adopted strategies. The review papers and their contributions are summarized in Table 1.Generally, all of the survey papers have summarized and classified the existing literature in emergency logistics. Although they suggested further research areas in general, there is lack of detailed analysis on the gaps between what we have studied and what we should study in emergency logistics that directly address the challenges faced in real world disaster situations.Our research aims to identify the current research gaps in emergency logistics research in the context of large-scale disasters. Since emergency logistics is significantly different and facing more challenges than traditional business logistics, we take problem identification and solution approach rather than summery and classification approach to our literature review with the following steps. First, we identify the key characteristics of large-scale disasters and corresponding challenges posed to emergency logistics. Second, we analyze current OR efforts on how to deal with these challenges. Finally, we investigate the gaps in current research, and suggest future research directions.Emergency management activities are commonly described in four programmatic phases: mitigation, preparedness, response, and recovery [12]. “Mitigation is the application of measures that will either prevent the onset of a disaster or reduce the impacts should one occur. Preparedness activities prepare the community to respond when a disaster occurs. Response is the employment of resources and emergency procedures as guided by plans to preserve life, property, the environment, and the social, economic, and political structure of the community. Recovery involves the actions taken in the long term after the immediate impact of the disaster has passed to stabilize the community and to restore some semblance of normalcy” [3]. In our study we only concentrate on those OR literatures that investigate emergency logistics in the immediate response phrase and in the context of large-scale disasters.We searched for literature on emergency logistics published in academic peer-reviewed journals and book chapters, while conference proceedings or working papers are not included. We search the title, abstract and keyword of journal articles published in English only. The search keywords we used contain “disaster”, “large-scale disaster”, “catastrophe”, “emergency”, “emergency response”, “emergency logistics”, “humanitarian logistics”, “optimization”, as well as their combination and extensions such as “disastrous”, “catastrophic”, etc. As the result, we have identified 81 papers on emergency logistics from 42 journals, including Operations Research, Management Science, Transportation Research Part E, European Journal of Operational Research, Interfaces, Journal of the Operational Research Society, Interfaces, Computers & Operations Research, IIE Transactions, Annals of Operations Research, OR Spectrum, Naval Research Logistics, etc. In contrast with routine emergencies such as medical emergency or traffic accidents, large-scale disasters can result in severe impacts on large concentrations of people, activity, and wealth and last for a longer time. Some specific characteristics of large-scale disasters differ, depending on the type of disaster and the types of relief actors involved, and they pose certain challenges for emergency logistics in the aftermath of disasters [2]. In this section we first identify the common characteristics of large-scale disasters, and then investigate some potential challenges of practical emergency responses.As pointed out by Chen et al. [13], a typical emergency situation has characteristics such as, “great uncertainty; sudden and unexpected events; the risk of possible mass casualty; high amounts of time pressure and urgency; severe resource shortages; large-scale impact and damage; and the disruption of infrastructure support necessary for coordination like electricity, telecommunications, and transportation. This is complicated by factors such as infrastructure interdependencies; multi-authority and massive personal involvement; conflict of interest; and the high demand for timely information.” Recently, Holguin-Veras et al. [14] also analyzed the unique features of post-disaster humanitarian logistics. They further distinguished the term catastrophes from disaster where catastrophes were defined as “high-consequence events that generate widespread and crippling impacts, where the ability of the impacted society to respond is severely compromised”. It seems the concept of catastrophes is similar as large scale disaster. Based on Chen et al. [13] with small adjustment, the characteristics of a large-scale disaster can be categorized as: Large scale impact, severe consequences, multi-agency involvement, time pressure, demand surge and resource shortage, great uncertainty, and infrastructure damage, as described in Table 2.Chen et al. [13] studied the coordination support activities based on the challenges of disasters characteristics on emergency response. Based on the unique features of post-disaster humanitarian logistics, Holguin-Veras et al. [14] also analyzed the differences between commercial and humanitarian logistics and identified the research gaps along seven key components: the objectives being pursued, the origin of the commodity flows to be transported, knowledge of demand, the decision-making structure, periodicity and volume of logistic activities, and the state of the social networks and supporting systems. Following the same approach, we investigate the potential challenges for emergency logistics according to each emergency characteristic discussed in Section 3.1, with the results summarized in Table 3.Large-scale disasters may affect wide geographical areas and large populations with severe damage. Emergency logistics tasks therefore are very complex and complicated, involving overwhelming damage assessment and demand estimation, allocation of variety of resources, complicated resource distribution in short period of time, organizing rescue operation and mass evacuation, etc. Moreover, these tasks are interrelated to each other and cannot be solved individually without considering their mutual impact. There are also existing hard-to-measure factors like unanticipated surge of local demand, transportation infrastructure damage as well as possibility of secondary disaster occurrence, etc. These features making the problem structures in emergency logistics are inherently chaotic and complex. Some of them are intractable and even containing numerically unsolvable instances from mathematical point of view. Hence, these substantial complexities and obstacles pose a great challenge for emergency logistics [6].Large-scale disasters may cause large casualties and severe property damages. In this context, the objectives and decision criteria for emergency logistics should focus on saving lives, alleviating human suffering and reducing property damages, rather than the traditional objective of reducing operating costs and increasing profit for business [15,16]. Moreover, the objectives or decision criteria can be different even conflict among parties involved. For instance, lifesaving may conflict with damage control. The objective achievements are often ambiguous and hard to measure [17]. For example, “Imagine an organization whose mission is to alleviate human suffering. How can you measure such an abstract notion? How can an organization meaningfully assess its direct contribution to such a broadly stated mission?” [18].Large-scale disasters require multiple agencies such as police, firefighters, medical teams, red-cross, volunteers, etc, working together to carry out emergency response tasks [19]. With different authorities, functionalities and professions, multiparty coordination could be very complex. Not only do these enormous players have different incentives and motivations, but also may exacerbate the competition for limited resources. Therefore, good collaboration is needed for information exchange, resource sharing, and job dispatching among different parties [20]. Otherwise, a lack of collaboration can lead to disaster propagation and even higher numbers of casualties [21].Disasters usually happen suddenly and develop rapidly. Any delay of relief efforts may cause severe consequences and many failures. Hence, there is time pressure to make quick decisions and to provide quick responses [15,22,23]. Under the situation of strict time pressure, two challenges are imposed on emergency logistics: On one hand, we need to speed up the response operation such as quick transportation of humanitarian aid through better scheduling. Sometimes it is necessary to look for a quick feasible solution rather than an unrealistically sophisticated optimal solution, because time is critical for quick emergency response. On the other hand, we need to speed up the decision-making process in order to reduce unnecessary delay. Real-time information gathering and decision support therefore is very critical [10].Large-scale disasters may create a sudden huge demand for emergency resources which greatly exceeds resource availability [24]. In this situation, it becomes imperative to allocate these scarce resources for different demand areas and ensure their availability to those that need resources the most [25]. Emergency resource allocation needs to consider many factors simultaneously, such as damage scenarios, number of casualties, priority of demand fulfillment, urgency level of needs, as well as delay consequence of humanitarian aids. However, how to set up allocation principles and measure resource allocation performance is a subject of much debate [26], since besides efficiency and effectiveness, it unavoidably involves questions of justice and fairness. With urgent needs and insufficient resources, what are the justice and fairness for resource allocation? There is still lacking of consensus on what they should be [27], because it is a most controversial topic regarding of ethical issues [28].Large catastrophic disasters usually have little precursory features before their occurrence, which make them highly uncertain and difficult to predict. In the large-scale emergency response practice, it is usually hard to predict the scope and progress of a disaster situation, to assess the damages and to estimate the resource requirement accurately [29]. To cope with these uncertainties, it is necessary to establish stochastic or scenario-based emergency logistics models [30,31,32,33].Large-scale disasters may cause extensive damage to communications, power supplies, and transportation infrastructures, and make them unavailable for emergency relief operations. For example, disrupted transportation facilities including ports, airports, roads and bridges may limit the humanitarian aids access to disaster attacked regions; destroyed communication infrastructures such as telephone and radio towers can hamper information collection and transmission of the catastrophe so as to slow down the responsiveness. Hence, these additional constraints need to be taken into consideration for emergency logistics operations [34].Based on the challenges to emergency logistics discussed in Section 3.2, we further review the achievements and the gaps of current OR studies in responding to these challenges, and identify future research directions. The findings are summarized in Table 4, followed by a detailed discussion in each subsection.Emergency logistics can be viewed as a very complex dynamic process which consists of many interdependent tasks with complex objectives and constraints. For example, after an occurrence of a large-scale disaster, the primary task is collecting and distributing emergency resources to the affected areas. However, several interdependent tasks emerge, such as who holds the emergency resource, where can one get the emergency resource, who delivers the emergency resource to the affected areas, when are transport vehicles available, etc. Hence, emergency logistics is a complicated problem since various decision problems must be considered simultaneously.Currently, emergency logistics problems have been studied in four areas: demand assessment, resource allocation, resource distribution, and emergency evacuation. Few studies integrate two or more specific decision problems into one decision model.Damage assessment and demand estimation refer to the assessment of the damage from disaster affected areas and the estimation of possible resource requirements in these areas. Problems studied include damage assessment, disaster area grouping, demand requirement forecasting, and demand priority ranking. Moltchanova et al. [35] developed a stochastic model to evaluate the economic losses and loss of life to assist efficient earthquake response. Chang et al. [32] grouped the affected regions based on their geographic distribution and distance. Other works related to area grouping method can refer to Gong and Batta [36] and Jotshi et al. [37]. Sheu [7] investigated time-varying relief demand forecasting, disaster area grouping and information uncertainty evaluation. The gap on this topic is a need for demand assessment models that can provide information in multiple dimensions at different levels of aggregation and keep updating in a dynamic environment.Resource allocation refers to allocating limited resources to disaster affected areas with the guidance of allocation principle. Fiedrich et al. first pointed out the significance of optimal resource allocation to disaster affected areas during the initial search-and-rescue period after a large-scale earthquake happened. Sherali et al. [38] discussed the problem of allocating certainly available resources to mitigate risks that may arise after the occurrence of natural disaster. Gong and Batta [36] considered ambulance allocation problem in immediate disaster response operations. Felder and Brinkmann [15] addressed the emergency medical service allocation amongst urban and rural regions. Zhang et al. [39] took the possibility of secondary disaster into account in the multi-resource allocation model. Arora et al. [25] studied the antiviral allocation problem to cope with a large-scale pandemic flu. They discussed the tradeoff between maintaining local redundant capacity and relying on mutual aid of antiviral resource. In fact, resource allocation problem usually involves many different types of resources, with very different requirements, e.g., periodical need or one-time need. However, current researches seldom consider this difference in the decision models.Resource distribution discusses how to deliver the various relief resources to affected areas efficiently. To some extent, distribution activities play a central role in disaster response operations. We notice that it is important to identify specific features in disaster response, such as infrastructure damage, as well as availability and compatibility of various delivery tools. Without these features, it will be hard to distinguish the emergency distribution models from traditional distribution problems, e.g., Tzeng et al. [40,41]. Most studies formulate emergency resource distribution problem as vehicle routing problem. For instance, Haghani et al. [42] proposed a simulation model to assist emergency medical vehicle dispatching and routing decision through updating real-time travel information; Shen et al. [43] addressed a stochastic emergency vehicle routing problem in response to a large-scale bioterrorism emergency; Lin et al. [44] investigated a specific vehicle routing problem through taking prioritizing item delivery into account, and formulated it as a multiobjective integer programming model. Some scholars have studied route selection problem such as Yuan and Wang [45]. They developed a multi-objective route selection model with consideration of the travel speed on each route affected by disasters. Some other scholars have investigated road capacities. For example, Barbarosoglu and Arda [30] incorporated the randomness of transportation capacity into their model; Jotshi et al. [37] considered the data fusion of road condition in their proposed emergency vehicles dispatching and routing model. Another studying viewpoint is to choose a proper traffic modal according to road conditions and road capacities, e.g., distribution via helicopter in Barbarosoglu et al. [46] and Özdamar [47], multimodal transport routing optimization in Özdamar et al. [6].Emergency evacuation studies how to displace people from dangerous areas to safe places. In OR, the evacuation problem is mainly formulated as a network design and network flow control problem with the objective of improving the efficiency of emergency evacuation. The interested reader is referred to Abdelgawad and Abdulhai [48] and Hamacher and Tjandra [49] to get more detailed review. In the existing literatures, the necessity and importance of OR models have been established [48], in which the typical OR approaches contain multiobjective optimization model [50], static network flow model [51], dynamic network flow model [52,53,54,55,56,57], time-expanded network model [44,58], and fuzzy robust programming model [59]. Hamacher and Tjandra [49] provided a detailed classification on mathematical modeling of evacuation problems. From the mathematical solvable point of view, Kim et al. [60] discussed the design of heuristics to solve large scale evacuation network flow model. Liu et al. [61] presented an algorithm applicable to evacuation problem in a specific context after a flood disaster occurrence.Besides focusing on individual emergency logistic decision problems, some researchers also made efforts to integrate different specific decision problems into one decision model, such as integration of resource allocation and emergency distribution [62], combine vehicle routing with supply allocation [63], joint optimization of distribution network repair with relief distribution scheduling [34,64,65], coordination model between emergency distribution and evacuation [54,55], a transshipment model linked distribution with inventory relocation model [66], a combined stochastic model for the storage and distribution of medical supplies [67], and an integrated model of facility location, emergency resource delivery, and vehicle routing [68].In general, previous studies have mainly focused on decomposition-oriented methods, which deal with emergency logistics problems by simplifying the problems or decomposing large problems into multiple smaller problems. The advantage of this paradigm is to make the complicated problems more tractable and easy solvable, but the major disadvantage is that it omits the interrelationships amongst the emergency relief activities. On the other hand, although the idea of an integration-oriented method such as joint optimization model in the context of large-scale disasters has been raised recently, it is challenging to build integrated models that address the entire emergency logistics process for large scale problems. The integration may also be achieved through an emergency response coordination system in which different models for different problems can be connected to each other through a network of information flows. For instance, the output of demand assessment can be used as input of resource allocation and the output and the output of resource allocation can be the input of resource distribution etc.The objectives and decision criteria not only reflect the attitudes and principles of decision-makers towards decision problems, but also act as measurements to assess various schemes or schedules. Effective decision metrics can not only help practitioners make quick decisions and improve emergency responsiveness, but also can benefit in coordinating much interdependent tasks amongst various participators and smoothing the disaster relief operations. Owing to the central role of emergency logistics in relief operations, the decision objective and criteria is critical for responding to large-scale emergencies and control the consequences of disasters [22].In existing literature, objectives commonly used in emergency logistics can be classified into three groups: improvement of distribution performance, assessment of demand fulfillment, and reduction in human deaths or improvement in human survivability. Moreover, it is also possible to combine objectives from different groups as multiple objectives.The criteria for improving distribution performance include: minimizing the selection of shortest path [39,50,69], minimizing distribution time [34,37,40,44,47,53,57,67,60,70], minimizing the time span of task completion [34,36], minimizing evacuation time [59,71], minimizing delay time of distribution service [52,54,55], minimizing distribution cost [7,30,39,40,41,46,62,66,72,73,74,75], minimizing evacuation network construction cost [71], minimizing the vehicle utilization [76], maximizing the outgoing flow [58,77], maximizing vehicle tour duration [46], as well as minimizing the average travelled distance [78].The criteria to assess demand fulfillment include minimizing unsatisfied demand [43,44,52,54,55,68,76,79,80], maximizing demand fill-up rate [7,40,65], and minimizing the difference of demand satisfaction rates between different areas [44].The criteria of reducing human deaths or improving human survivability include minimizing the number of fatalities [81,82] or maximizing the expected number of saved people [83], maximizing weighted throughput of casualties [84], maximizing the human survival probability in ambulance service [23,85,86], and possible health outcomes (death, hospitalization, outpatient care) in evaluating different intervention policies for influenza pandemics [3].Unlike commercial logistics taking minimizing economic cost as primary performance measurement, emergency logistics metrics is more complicated and need to consider much complex factors [87]. Some researchers attempted to develop a performance measurement framework for emergency logistics decision. For example, Huang et al. [63] used equity, efficiency and efficacy as important indicators to measure emergency relief operations, and investigated the balance among the three metrics; Felder and Brinkmann [15] investigated the trade-off of equity and efficiency in emergency medical service; Davidson [4] used appeal coverage, distribution time, efficiency, and assessment accuracy to measure the performance of relief logistics; and Balcik and Beamon [22] suggested to apply the performance measurement framework of commercial supply chain to humanitarian relief chain, and developed a framework consisting of resource, output, and flexibility metrics through extending the previous work in Beamon [88]. Some literatures specially discussed the balance amongst different objectives in their developed models, such as the tradeoff of efficiency and equal service time in Chiou and Lai [64]. Holguín-Verasa et al. [89] argued that welfare economic principles must be incorporated in post-disaster humanitarian logistic models to ensure delivery strategies that lead to the greatest good for the greatest number of people. They suggested the use of social costs—the summation of logistic and deprivation costs—as the preferred objective function for post-disaster humanitarian logistic models where the deprivation cost was the economic valuation of the human suffering associated with a lack of access to a good or service. Recently, Eisenhandler and Tzur [90] investigated the tradeoff between effectiveness and equity for food allocation among welfare agencies, and formulated it as a routing resource allocation problem.In summary, most current research focuses on traditional logistics objectives (minimization of distribution time and distribution cost, and selection of shortest path). Typically, cost-based and time-based objective functions are often representative of current research efforts. Few studies have considered emergency related decision criteria such as minimizing the number of fatalities or maximizing demand fulfillment. However, the primary principle in emergency logistics is saving human lives and reducing property damage through various emergency relief activities. Thus, the decision objectives and criteria of future studies should be more directly linked to the end results such as life-saving, human suffering alleviation as well as damage reduction [91]. Moreover, although the performance measurement of emergency logistics has received greater attention by many academics [22], a uniform metric framework to guide emergency relief operation needs to be further investigated.Coordination can be defined as the management of interdependencies between activities to achieve a goal [92]. In emergency response tasks are often complex, uncertain, and interdependent to each other and need to be carried out jointly by multiple agents [93].Current OR has studied the coordination of multiple decision problems. Yi and Özdamar [55] and Yi and Kumar [54] investigated the decision coordination problem between emergency distribution and emergency evacuation. Chiou and Lai [64] and Yan and Shih [34] integrated road repairing scheduling with relief distribution to analyze the integrated schedules. Huang et al. [63] combined the vehicle routing with supply allocation, under the consideration of equitable service to all beneficiaries. Balcik et al. [62] developed a joint optimization model of resource allocation and emergency distribution. Moreover, Rottkemper et al. [66] investigated the integration of resource distribution and inventory relocation, and formulated it as an integrated transshipment model. Generally speaking, the integrated optimization and assessment method is a good way to coordinate different relief activities in emergency logistics schedules. The integration of relief activities can improve the effectiveness and efficiency of emergency relief operation and enhance the responsiveness to emergency events [94,95], whereas the development of a more complex emergency logistics model may increase the problem complexity and will put more stress on model solutions especially under time pressure of emergency relief decision making.Also, we notice that there are some OR literatures outside emergency logistics domain have made contributions to the topic of collaboration, such as collaborative transportation planning [96], shipper collaboration model [97], A detailed literature review of collaborative transportation is provided by Agarwal et al. [98]. We believe the basic principle of collaboration is common in different areas, and the collaboration approach in transportation domain may be applied to collaborative emergency relief operations.On the whole, the multiparty collaboration problem still remains at the top of the research agenda [21,99]. Most current literature investigates this topic from a single authority’s perspective, i.e., assume a single authority to deal with emergency response. Although some studies have considered the coordination of multiple decision problems such as the integration of resource allocation and distribution, they have not considered different objectives by different parties. In fact, emergency logistics almost always needs to simultaneously implement different sequential response tasks [100]. Hence, future studies in this field should focus on the investigation of the interdependency of tasks, resources, and workflows across and among different decision authorities, and possible modeling tools could learn from the knowledge in the fields of collaborative transportation planning and workflow technology. The need for multiparty coordination has also been emphasized by Altay and Green III [3,10], and Galindo and Batta [1]. As various groups involved in emergency response, organizational structures and relationships need to be built into multi-criteria, multi-objective decision making models.Time is life in emergency response. Any delay of decision and action may cause unnecessary casualty and human suffering that otherwise could be avoided [86,101]. The highest priority for emergency logistics is to save lives as soon as possible, and emergency response time has been identified as a critical indicator to measure the performance of emergency relief operation and survival possibility of injured people [23]. Hence, many countries have enacted the time threshold to respond in large disaster events. For instance, under the new Department of Homeland Security in the U.S., the Strategic National Stockpile (SNS) Program is required to maintain a stockpile of pharmaceutical agents, vaccines, medical supplies, and equipment to augment state and local resources during a large-scale disaster or bioterrorism event. Upon request, the SNS Program will deliver materials anywhere in the United States within 12 or fewer hours [102].To deal with critical time requirement, current OR studies tend to focus on making the minimum distribution time as the decision objective, or setting time windows as a constraint in the mathematical model. Those studies related to minimizing distribution time have been analyzed in Section 4.2. A typical study is conducted by Gong and Batta [36], who studied the problem of initial allocation and subsequent reallocation of ambulance amongst casualty clusters in consideration of round-trip service time. They developed a continuous function to depict the casualty growth in a cluster, and combined it with the criterion of minimax completion time that ambulances need to serve the casualty cluster. For studies of the time-window setting, Haghani and Oh [41] incorporated a time-window constraint into a time-space-based distribution network; Shen et al. [43] investigated an emergency vehicle routing problem in consideration of the time window constraint; and in Lin et al. [44], they took account of the soft time windows in their developed emergency relief planning model.For quick decision making, some papers have investigated the problem of real-time decision making through continuously updating information used in decision models. Thus, this requires the rapid data gathering and information processing, and continuous adjustment with the changes of disaster situations [103]. In current literatures, the widely used method for real-time information processing is data fusion, which is a process that refines its estimations and assessments of decision parameters continuously. This includes Sheu [104] who adopted data fusion methods to forecast relief demand in multiple areas, so as to support the emergency distribution, and Jotshi et al. [37] who estimated the number of casualties and road conditions in a post-disaster environment by using the data fusion method. In addition, combine efficient optimization technology with real-time decision support system is another interesting area of research to assist emergency response [42,105]. For example, Horner and Downs [106] and Chang et al. [32] incorporated the geographic information system (GIS) into the emergency rescue planning model.Some OR studies attempted to analyze the critical time requirement for life saving perspective and explored the quantitative relationship between critical response time and human being survivability. Felder and Brinkmann [15] pointed out that the response time can crucially determine the quantity and quality of life saving in an emergency event. Erkut et al. [23] and Knight et al. [86] investigated the patient survival possibility in the context of ambulance location models, and demonstrated that the probability of patient survival was a function of response time. Moreover, McLay and Mayorga [85] also discussed the performance evaluation of response time thresholds in terms of resulting patient survival rates.Although current literatures have taken time into consideration, most of them treated time as an objective or a constraint from the traditional logistics perspective, and did not reflect the time pressure feature in the aftermath of a large-scale disaster. Thus, quantitative metrics linking human survivability with response time is very much needed. This issue has been recognized and investigated by some scholars, but their studying efforts are all based on the statistical results from medical care field [23,86], and may inadequate in the case of general emergency logistics. We also need to make more efforts to address the issues of decision support user interface for real-time applications of decision models [10]. Finally, exploring time-varying process between disaster scenarios and time criticality is rare in current literatures, and need to be further investigated.During a disaster, resources need to be allocated to the affected people and places for effective rescue operation. A general issue in this problem domain is how to allocate emergency resources in order to relieve the consequences caused by a disaster [15,38]. In a large-scale disaster situation, resource allocation decisions are often affected by sudden demand surging and serious shortage of available resources. In the context of insufficient emergency resources, different criteria can result in different resource allocation schemes. Usually, priority setting is a key determining factor to allocate scarce emergency [107].To deal with resource shortage, we borrow the worth-oriented paradigm from social choice theory [108] and welfare economics [109]. According to their opinions, the competing ethical principles for scarce resource rationing are utilitarianism and egalitarianism. Utilitarianism focuses on maximizing total social worth or maximizing the greatest value of goods for the greatest number of people, which can also be interpreted as the most lives saved in the context of emergency response. The utilitarian principle is to ration the scarce emergency resources and determine the satisfaction of demand requirements through priority setting [110]. For example, the priority in emergency relief operations may be related to which affected region should be satisfied first, triage protocols in places [24,111], etc. However, this allocation principle may cause inequality amongst disaster victims at different affected regions. Caro et al. [28] argued that utilitarian efficiency should be tempered by the principle of equality in making decisions about providing life saving interventions and palliation. Although utilitarian efficiency is important, egalitarian criteria (i.e., equality or fairness) are also the key modifying ethical principle. Bertsimas et al. [112] even argued that fairness should be obtained at the expense of efficiency sacrifice in resource allocation. They also analyzed different price of fairness such as proportional fairness and max-min fairness in their discussion. According to Winslow [110], egalitarianism was based on equality of opportunity and fairness of demand satisfaction. In the immediate emergency relief operations, equality or fairness not only means the injuries have equal rights to have their needs met [28], but also refers to the variance in arriving times of resources should be as small as possible [63,91].The criteria for allocation decisions in the current literature can be grouped as two streams: one only focused on the utilitarian principle, and another considers both of utilitarian and egalitarian principles. For the stream of utilitarian principle, different utilitarian criteria have been embedded into the developed models, such as cost-effectiveness analysis [113], cost-benefit-based methods [25], deterministic priority setting [36,38,39,53], triage management policy [24,111], urgency level of disaster affected regions [7,68], injury classes ranking [81], humanitarian aids criticalities [62]. Although these criteria seem distinctive from each other and a bit dazzling, the essential of them is the same. That is, allocate the scarce resource or emergency service through prioritizing the affected regions or injuries. For the stream of following both utilitarian and egalitarian principles, Felder and Brinkmann [15] combined the efficiency (i.e., maximizing the total number of survivors) and equality (i.e., equal access to emergency medical service) in their developed emergency medical service allocation model, and find that the two objectives can lead to different deployment patterns. Jacobson et al. investigated the tradeoffs among demand urgency, rescue rewards and service times in allocating emergency resource to multi-categorization casualties, and formulated this problem as a priority assignment policies optimization model. De la Torre et al. [91] reviewed the allocation policies from the perspective of practitioners and academics. Besides that, there are three articles that are closely related to the latter stream, namely, Ferrer et al. [114], Huang et al. [63] and Mete and Zabinsky. Both of them investigated the metrics of equality and efficiency for emergency distribution.Generally, resource allocation criteria in current studies are focusing on utilitarian principle, and only few studies have considered the balance between efficiency and equality in their developed models. Current studies often consider the allocation of single type of resources. But in practice resource allocation usually need to deal with multiple types of resources that related to each other with different allocation principle. Current literatures usually rank affected regions with deterministic priorities and seldom consider the ranking priority changed over time. In fact, resource allocation is related to resource distribution capacity for delivery and cannot be decided separately in a sequential order, thus resource allocation needs to be adjusted frequently in order to respond to the changing situation such as surge of casualty. Hence, joint resource allocation and distribution needs to be studied.Future studies in this field can be presented from two aspects. On one hand, we need to develop more dynamical priority metrics to allocate emergency relief resources. From the utilitarian perspective, one can develop dynamic efficiency criteria in emergency resources allocation by introducing the law of diminishing returns, in which plenty of works in the economics area can be referred. From the egalitarian perspective, one can discuss the nonlinear consequence of humanitarian aids delay or of injuries’ waiting cost. And its model development can be learned from the problem formulation of delay and tardiness in the machine or job shop scheduling field. On the other hand, combine priority setting and demand fulfillment as the criteria for resource allocation, and achieve a balance between priority and equality. Although current studies have considered prioritization in resource allocation decisions, how to set priorities is still an open problem for emergency relief actors [115], and even need interdisciplinary studies devoted into this field [116,117]. In the context of demand requirement greatly exceeds resource supplies, emergency response should consider both resource shortages, human life equality, and need urgencies of disaster victims. Thus, future research in scarce resource allocation needs to take resource utilization efficiency, emergency relief equality and human life time into account simultaneously [28,107]. Some other studies on efficiency versus equality (or fairness) will be helpful to understand and formulate the tradeoff issues of allocating scarce relief resources, and interested readers can refer to these studying efforts in Hooker and Williams [118], Mandell [119], Golany and Tamir [120], Gloverand Ball [121], Bertsimas et al. [122], Jacobson et al. [107].In general, there are often resource shortages during emergency response. However, material convergence may also happen when hundreds of thousands of donors (governments, communities or individuals) sent massive amounts of supplies and equipment during a disaster. Although it brings in much-needed supplies, a significant portion of useless unsolicited donations creates major complications for the disaster response [14]. There is an urgent need to study the dynamics of the material convergence and the ways to control and reduce the negative impact of non- and low-priority material flows [14].There is great uncertainty in the context of large-scale disasters. The uncertainty exists in many aspects such as demand requirements, supplies availability, road conditions and damage levels, etc. There are also great uncertainties associated with human behavior in disaster situations [11]. The difficulty of information gathering and the damaged communication infrastructure during disaster make the degree of uncertainty even worse.Klibi et al. [123] defined uncertainty as the inability to determine the true state of the future business environment which may be partially known or completely unknown. They distinguished three types of uncertainties: randomness, hazard, and deep uncertainty. Randomness can be characterized by random variables related to business-as-usual operations. Hazard is characterized by low probability unusual situations with a high impact, and deep uncertainty is characterized by the lack of any information to access the probability of plausible future events. Therefore, the methods adopted to handle these different kinds of uncertainties should be adapted with their intrinsic attributes [124]. For the randomness problems, the common way is to model the uncertainty as random variables, and make “robust” decisions prior to uncertainty being realized, as in stochastic programming or robust optimization approaches. For hazards, it may be very difficult to obtain sufficient data to assess objective probabilities and subjective probabilities must often be used. While for the deep uncertainty problems, the tackling methods are to make prompt response after uncertainty became certain. Herein, we also follow their opinions in our discussion, because natural disasters especially the catastrophic emergencies are quite difficult to obtain sufficient information to predict their occurrence [125].Large-scale disasters usually have the intrinsic feature of deep uncertainties, and pose many hard-to-measure factors and stringent constraints on immediate emergency logistics decision making, but it does not mean those uncertainties cannot be predicted completely in practical emergency response operations. Indeed, some uncertain decision variables can be estimated approximately. For example, the uncertainty of supply amount can be handled by summing up the total resource delivered to the affected region until the decision making epoch. The uncertainty of demand requirement also can be estimated by the demographic information of the disaster attacked regions. Therefore, in emergency logistics for this kind of uncertainties, they can be characterized from previous experience or forecast. However, some other uncertainties have no laws and are difficult to predict or estimate. For example, a road is damaged after earthquake. We know the transportation on the road will be delayed, but it is very hard to use a random variable to characterize the randomness very well, since the road conditions dynamically varied with the possible earthquake aftershocks and emergency repairs. In this context, responsive decisions methods after the random events happen are more useful and practical than robust decisions before the random events happen. Therefore, in OR research works the modeling methods of disaster uncertainty should be in accordance with the problem characteristics, otherwise the conclusions and results may have not application values in practice [1].At present, the common methodology used in OR community to deal with uncertainty include stochastic programming [30], scenario-based modeling [32], robust optimization [76,126], rolling horizon approach [66,127], fuzzy programming [59], as well as simulation approach [37], among them stochastic and scenario-based methods are more popular in model development. Stochastic programming has good ability to cope with uncertainties of disaster development through incorporating probabilistic scenarios [128]. For scenario-based approach, the advantage is that it can make a complicated problem more tractable [129], while the disadvantage is that it is difficult to deal with infinite number of disaster scenarios [123]. For more information on optimization under uncertainty, please refer to a survey paper by Sahinidis [130].In the field of emergency logistics, the majority of current studies focus on the deterministic optimization models with the assumption of known data for given situations, and few works have investigated the stochastic models. Barbarosoglu and Arda [30] investigated the uncertainties of demand, supply and transportation capacity in emergency resource distribution scheduling, and developed a scenario-based two-stage stochastic programming to robustly disclose these uncertainties along with the progress of the emergency response. Reference [59] proposed a robust fuzzy programming to formulate the evacuation problem under uncertainty, and used a fuzzy lower and upper bound approach to handle the uncertainties of vehicle number, vehicle capacity, travel time, and number of waited trapped people, etc. Najafia et al. [76] developed a multi-objective robust optimization model for both of the disaster relief distribution and injured people evacuation in the aftermath of an earthquake happened. Shen et al. [43] considered the uncertain travel time and demand in the context of responding a large-scale bioterrorism emergency, and developed a two-stage stochastic vehicle routing model. In the first planning stage, they modeled the uncertainties as chance constraints, and coped with the chance constraints through revealing demand level and travel time in the second operational stage. Beraldi et al. [131] also used the chance constraint technique to hedge the emergency service reliability uncertainty in emergency service site locations and emergency vehicle assignments, and formulated the problem as a stochastic programming model. Mete and Zabinsky investigated the medical supply location and distribution to prepare and respond uncertain disaster scenarios, and formulated this problem as a two-stage scenario-based stochastic programming model. In their study, they adopted the scenario-based method to depict the plausible disaster scenarios and their emerged probabilities, and they then considered warehouse selection and inventory level decisions in the first preparedness stage, and medical resource distribution and demand satisfaction decisions in the second response state. The disaster-scenario-based modeling were also adopted by Chang et al. [32] and Li et al. [33], in which they respectively developed a flood-scenario-based mix integer programming model to assist the emergency preparations for floods and a hurricane-scenario-based bi-level programming model to prepare the attacks by possible hurricane events. Noyan et al. [132] investigated a post-disaster last mile distribution problem under uncertainty, and formulated it as a two-stage stochastic programming model. Balcik et al. [62] used a rolling horizon approach to real-time update the observed information and handle the uncertainties of resource supply and demand requirement. The information updating approach to handling uncertainties is also adopted by Chen and Miller-Hooks [83]. They studied a dynamical search and rescue team deployment problem over decision horizons. They considered the uncertainties of demand, service time and travel time, and formulated them as a multistage stochastic programming, in which they handled uncertainties through continuously updating the observed post-disaster information at each decision stage, so as to improve the robustness of emergency decision. In our understanding, the idea that prompt response after uncertainty became certain could be grouped into the umbrella of rolling horizon approach. Besides these formulations, another useful method for dealing with uncertainties is simulation, which was introduced in the emergency vehicle dispatching and routing [37], emergency evacuation [51] and resource allocation [105].In conclusion, current studies in this problem domain have developed some robust stochastic models to hedge the uncertain conditions in the context of immediate emergency response, but they have not carefully distinguished the differences among those uncertainties emerged in real emergency relief operations, and the inherent complexity of uncertainty therefore is not adequately captured.Further research on the uncertainty problem can be addressed from two aspects. On one hand, combine the disaster scenario generation with traditional optimization models. The big challenge might existed is that the plausible disaster scenario maybe infinite, especially in the context of high uncertainty and rapid disaster evolution. Under this circumstance, the appropriate alternative is adopting Monte Carlo method to generate the number of possible disaster scenarios [124,133]. On the other hand, deal with the difficulties in various kinds of uncertainties in an emergency situation, especially for those unprecedented emergency situations that no probability distributions are available. The major challenge of further studies on this aspect may come from modeling difficulty by using traditional OR methodologies, but catastrophe modeling techniques may provide a way to unlock this difficulty [125]. Catastrophe modeling is used to assess catastrophic risk and to improve risk management strategies. The modeling of catastrophe risk is a complex process that depends on scientific knowledge and subjective and objective inputs related to natural hazard such as hurricanes, floods, and earthquakes. Computer simulation is often used to estimate the risk of catastrophic events and assess the impact of possible actions. Catastrophe modeling tools are provided by software companies such as AIR Worldwide (www.air-worldwide.com), EQECAT (www.eqecat.com), and Risk Management Solutions (www.rms.com), and are widely used for risk analysis by insurance companies but it can have potential to be used in emergency logistics operations. Geographic Information Systems (GIS) with satellite image and remote sensor systems are also very useful for emergence management. GIS is capable to process, analyze and display spatial information for emergency response and have been used by emergency managers as a valuable decision support tool in emergency response in various kinds of disasters, including floods, bushfires, droughts, hailstorms, tsunamis, hurricanes, landslides, and earthquakes [134,135,136].Emergency logistics often needs to deal with last mile distribution which refers to delivery of relief supplies from local distribution centers to beneficiaries affected by disasters [62]. In practice, last-mile emergency logistics always plays a central role among all emergency relief actives [62], and its successful operations can improve the robustness and flexibility to respond major disasters. During a disaster such as an earthquake, transportation infrastructures were often damaged, resulting in many constraints imposed on last-mile emergency logistics and affecting feasibilities for emergency relief operations.Two approaches are used to handle the emergency logistics with damaged transportation infrastructure. One approach is to combine emergency logistics scheduling with damaged infrastructure repairing planning [34,64,65]. Another approach is to improve the robustness and flexibility for emergency relief operations. That is, consider possible uncertainties and additional constraints when making emergency logistics planning. Various circumstances from different perspectives have to be taken for consideration. Among them, capacity constraint is the most popular factor that to be considered in current studies, such as adding traffic capacity constraints into the distribution network [41], road capacity in emergency evacuation [53,59]. In addition, some other factors are also studied, such as road congestions, road complexities, etc. For example, Jotshi et al. [37] considered the road congestions in their emergency logistics network, and Han et al. [51] developed a one-destination evacuation model to avoid the road congestion or blockage; Yuan and Wang [45] investigated the shortest emergency distribution problem with the consideration of road uncertainty caused by disaster damage. Besides that, there are some other research streams related to damaged infrastructures, which include infrastructure vulnerability analysis [137], emergency recovery management [138].In summary, considering capacity constraint is the most common way used in emergency logistics models to cope with the impact of damaged infrastructure, which is usually regarded as a lower bound or an upper bound with given disaster scenario. However, how to determine the lower and upper limit is still a difficult issue, since the condition of infrastructure damage is dynamic and varied with the situation changes in large-scale disaster. Further research on this problem can be focused on alternative solutions, such as investigating the combinatorial choice of multi-mode transportation routing to cope with infrastructure damage/availability. On the other hand, improve the resilience capability of emergency logistics network to resist the disaster attack is another research direction. Here resilience refers to the robustness and flexibility against emergency events and quick recovery capability from disruptions [139], which has been paid increasing attention in the fields of transportation planning [140] and supply chain design [141]. Hence, building the resilience into emergency logistics network can not only improve its reliability under a deep uncertain environment, but also can reduce the restoration time from emergency event strikes [142].In this paper, we used the problem identification and solving approach to review existing emergency logistic literature. The main contributions and novel studies are summarized as follows. Firstly, we identified the key characteristics of large-scale disasters and assessed their challenges to emergency logistics. Secondly, we then analyzed and summarized the current literature that deals with these challenges. Finally, we discussed existing gaps in the current research and indicated future research directions.Our review approach is significantly different from simple summery or classification based on the years of publication, the algorithm used, the application area, the types of logistic problems, or the phases of emergency management. The main contribution of our work is to provide a new perspective to re-exam emergency logistic research. We cannot simply apply our knowledge of business logistics to emergency logistics without deep understanding of the unique characteristics and challenges of large-scale disasters. When we have understood these challenges, we can thus identify what we have accomplished and what are the gaps and further research directions.To deal with the challenges of large-scale disasters, we need to change many basic assumptions we usually use in traditional business logistics. For instance, we need to shift our focus from problem decomposition to multi-task integration, from time for operation efficiency to time for life saving, from single decision-making authority to multiparty authority, from unlimited resource availability to serious resource shortage, from uncertainty with randomness to deep uncertainty without previous knowledge, from perfect transportation infrastructure to damaged infrastructure. We believe that under those assumption changes we can significantly enrich OR in emergency logistics.A limitation of our work lies in that we only concern the phrase of immediate emergency response and survey current OR literatures contributed to this phrase. Actually, the framework of emergency logistics is broad, which could extend into preparedness operations before the occurrence of disasters as well post-disaster restoration activities. Another limitation is that we limit the scope of our review in the academic community, and haven’t included the works on disaster response from the practice viewpoint by some key agencies like Red Cross, WHO, UN, etc. Hence, these limitations provide a direction to enrich and improve our work in future research.Conceptualization, Y.J. and Y.Y.; methodology, Y.Y.; software, Y.J.; validation, Y.J. and Y.Y.; formal analysis, Y.J.; investigation, Y.J.; resources, Y.J.; data curation, Y.J.; writing—original draft preparation, Y.J.; writing—review and editing, Y.Y. and Y.J.; visualization, Y.J.; supervision, Y.Y.; project administration, Y.J.; funding acquisition, Y.J.This research was supported by Natural Science Foundation of Jiangsu Province, China (BK20160742), Humanity and Social Science Youth Foundation of Ministry of Education of China (17YJC630048), National Natural Science Foundation of China (71803084), Fundamental Research Funds for the Central Universities (NAU: KYZ201663; NAU: SKYC2017007; NAU: SKTS2016038; NAU: SKYZ2017025), and Project of Philosophy and Social Science Research in Colleges and Universities in Jiangsu (2017SJB0030).The authors declare no conflict of interest.Summary of review papers on emergency logistics.Key characteristics of large-scale disasters.Challenges for emergency logistics.Current studies and future research directions in emergency logistics.Most studies focused on decomposition to a specific or simplified decision problem, such as demand assessment, resource allocation, emergency distribution, and emergency evacuation; Few studies considered an integrated model.Develop integrated models that address the entire emergency logistics process for large scale problems.Most studies focused on traditional logistics objectives (minimization of distribution time, cost and shortest path selection, etc);Few studies considered emergency specific decision criteria such as minimizing number of fatalities, maximizing demand fulfillment, and minimizing unsatisfied demand.Make objectives more directly link to end results such as life- saving and damage reduction;Develop a uniform metric framework to guide emergency relief operation.Most studies assumed a single authority to deal with emergency response;Some studies considered the coordination of multiple decision problems such as the integration of resource allocation and distribution, but did not consider different objectives by different parties.Investigate task, resource, and workflow interdependency across different stakeholders.Current studies tended to focus on minimizing distribution time and setting time windows as a constraint;Some papers enabled real-time decision making through continuously updating the information used in decision models;Few researches linked human survival possibility with emergency response time;Lack of study on real-time decision-making implementation issues.Explore the measurement of critical time requirement;Develop more adequate quantitative metrics linking life- saving with response time; Make more efforts to address the decision support user interface issue for real-time application of decision models;Explore dynamic relationship between disaster scenarios and time criticality.Resource allocation was usually based on a given priority, and seldom consider the priority changed over time;Lack of consideration on multi-type resource allocation;Few studies considered the balance between efficiency and equality.Develop more dynamical priority metrics to allocate relief resources;Combine priority setting and demand fulfillment as the criteria for resource allocation;Achieve a balance between priority and equality.Most papers investigated deterministic models, with the assumption that data were known for the given situation;Few studies developed stochastic, fuzzy, and simulation models to tackle the uncertainties in disaster relief operations.Combine scenario technique with optimization model;Deal with the difficulties of unprecedented emergency situations (no probability distributions are available).The repair of damaged roads was incorporated into the distribution model;Traffic capacity constraints were added into distribution networks;Traffic capacity constraint was treated static, not dynamic during emergency response.Use combinatorial choice of multi-mode transportation to cope with infrastructure damage/availability;Improve resilience capability of emergency logistics network.

Med-MDPI/ijerph_3/ijerph-16-05-00780.txt

@@ -0,0 +1 @@
+Neurocysticercosis (NCC) significantly contributes to morbidity in developing countries. We recently published a study of prevalence and risk factors in school-aged children in three mountainous areas in Sichuan province of western China. Using structural equation modeling (SEM) on data from that study to guide intervention planning, here we examine risk factors grouped into three broad interventional categories: sociodemographics, human behavior, and sources of pork and pig husbandry. Because neuroimaging is not easily available, using SEM allows for the use of multiple observed variables (serological tests and symptoms) to represent probable NCC cases. Data collected from 2608 students was included in this analysis. Within this group, seroprevalence of cysticercosis IgG antibodies was 5.4%. SEM results showed that sociodemographic factors (β = 0.33, p < 0.05), sources of pork and pig husbandry (β = 0.26, p < 0.001), and behavioral factors (β = 0.33, p < 0.05) were all directly related to probable NCC in school-aged children. Sociodemographic factors affected probable NCC indirectly via sources of pork and pig husbandry factors (β = 0.07, p < 0.001) and behavioral variables (β = 0.07, p < 0.001). Both sociodemographic factors (β = 0.07, p < 0.05) and sources of pork and pig husbandry factors (β = 0.10, p < 0.01) affected probable NCC indirectly via behavioral variables. Because behavioral variables not only had a large direct effect but also served as a critical bridge to strengthen the effect of sociodemographics and sources of pork and pig husbandry on probable NCC, our findings suggest that interventions targeting behavioral factors may be the most effective in reducing disease.Caused by the zoonotic tapeworm Taenia solium, cysticercosis affects millions of people living in Asia, Africa, and Latin America [1]. There are two types of human infection: intestinal taeniasis and cysticercosis. Intestinal taenisasis is caused when humans consume uncooked or under-cooked pork containing cysticerci. The adult tapeworm develops and inhabits the gastrointestinal tract, with humans serving as the definitive host [2]. Cysticercosis is acquired by ingestion of oncospheres released in the feces of humans with intestinal taeniasis [3]. Hexacanths migrate to tissue throughout the body, forming cysticerci. In the case of cysticercosis, humans are a dead-end intermediate host for the tapeworm [4,5,6]. Human cysticercosis has been linked to poor sanitation, free-range pig raising, and limited pork inspection and control [7,8].Neurocysticercosis (NCC), the most severe form of cysticercosis, occurs when T. solium metacestodes invade the brain causing a wide range of neurologic morbidity [4,9,10] including seizures, chronic headaches, psychiatric disturbances, cognitive impairment, and focal neurologic deficits [2,11,12,13]. While the worldwide burden of NCC is not fully appreciated, it is a leading cause of death from food-borne disease resulting in 2.8 million disability-adjusted life years lost in 2010 [8].In China, NCC is prevalent in almost every province and autonomous region, with a high density in Southwestern China [14,15,16]. It is estimated that approximately 7 million people in China are affected by T. solium neurocysticercosis [1]; however, the risk factors and impact of NCC in rural areas of Southwest China remain poorly characterized [3,17]. A better understanding of the epidemiology and risk factors of the disease in these rural areas is needed to inform intervention strategies.The prevalence and effects of NCC among school-aged children have received little scientific attention. We found the seroprevalence of human cysticercosis IgG antibodies to be 6% in 5th and 6th graders overall and higher than 20% in some schools [16]. Owning pigs, feeding pigs human feces, and reporting proglottids in feces were found to be risk factors for the presence of antibodies [16].In the present manuscript, we use structural equation modeling (SEM) [18,19,20] to categorize important areas for intervention in pediatric populations. We use SEM to group factors into three latent variable groups, each representing potential intervention categories. Because neuroimaging is neither easily available in the study area nor financially feasible for large field surveys, using SEM allows for the use of multiple observed variables (serological tests and symptoms) to represent an unobserved construct: probable NCC cases.Detailed field and laboratory methodologies for this study have been described elsewhere [16], and are only briefly reviewed here.This study was conducted in three rural prefectures (Aba, Ganzi, Liangshan) in Sichuan province. Located at the eastern extremity of the Tibetan Plateau and with an average altitude of 2700 meters, these areas are largely inhabited by poor smallholder Tibetan farmers, many of whom raise pigs. We utilized a school-based sampling technique, focusing on fifth and sixth grade students [16].The study received ethical approval from the Stanford University institutional review board (Protocol ID 35415) and from the Sichuan University Medical Ethical Review Board (K2015031). The school principals, who are the children’s legal guardians while in school, provided consent for student participation and each student provided verbal assent prior to participating.Questionnaires including a student questionnaire and a take-home head of household questionnaire were used to collect information on sociodemographic factors, behavioral factors, sources of pork and pig husbandry, and neurologic symptoms [16].At each school, approximately 5 mL of blood was collected from each participating student via venous puncture. Sera were tested for T. solium human cysticerosis IgG antibodies using an enzyme-linked immunosorbent assay (ELISA) based on low-molecular-weight antigens (LMWAgs) of T. solium cysticerci collected from pigs in Chinese endemic areas. LMWAg-based assays have been shown to be highly sensitive and specific [21,22], have been used in previous field studies [23], and are especially attractive given their low cost, quantifiable results, and simplicity for use in low-resource areas [22]. Detailed assay methodology has been published previously [16,22].Because the timing of the survey conflicted with the class time of students, some schools in Ruoergai county did not complete the questionnaire, and 268 students in Ruoergai county were removed from the analysis because the questionnaire was not completed. In cases of missing data, we applied a conditional mode to fill in missing values [24].Univariate analysis was conducted to depict the distribution of all factors. The associations between sociodemographics, behavioral variables, sources of pork and pig husbandry, and probable NCC were examined by SEM. Two steps were taken when building our model. First, to assess whether observed variables (e.g., age, sex, and reported behaviors) were adequate indicators of latent variables (not directly observed but estimated from directly measured variables) we conducted confirmatory factor analysis to test associations. Second, the structural model was carried out to estimate the strengths of the associations among latent variables.In our SEM (Figure 1), we propose that risk of NCC is associated with three latent variables: sociodemographics, human behavior, and sources of pork and pig husbandry. The sociodemographics latent variable was estimated by age, sex, grade, boarding status (living and sleeping at school versus returning home to sleep), parental education level, and household asset score. A household asset score was developed using principal component analysis to aggregate all asset ownership variables. The behavioral latent variable was estimated by the reported presence of open defecation, overall frequency of pork consumption, and frequency of raw pork consumption. The sources of pork and pig husbandry latent variable was estimated by reports of purchasing pork at markets, pig ownership, herd husbandry practices (allowing free foraging, feeding pigs human feces, and if disposal methods left feces accessible to pigs), and if households reported noticing cysts consistent with cysticerci in their pork in the previous 5 years (photos of cysticerci in pork were presented to participants). For the latent variable representing probable NCC, we selected factors based on China’s national diagnostic standard [25] including the presence of IgG antibodies to T. solium cysticercosis, the presence of seizures, and the presence of frequent headaches (defined here as self-reported occurring 6 times a month or more).Multiple indicators were used to evaluate the fit of the model, including χ2/degrees of freedom (χ2/df), goodness-of-fit index (GFI), normed fit index (NFI), comparative fit index (CFI), parsimony goodness-of-fit index (PGFI), parsimony normed fit index (PNFI), parsimony comparative fit index (PCFI), root mean square residual (RMR), and root mean square error of approximation (RMSEA).The association was considered to be statistically significant if the 2-sided p value was less than 0.05. All analyses were performed using SPSS (IBM, Armonk, NY, USA) and AMOS 21.0 statistical software (IBM, Armonk, NY, USA) [26].A total of 2608 students from the data analyzed in our original prevalence manuscript [16] were included in the SEM analysis.Table 1 summarizes descriptive statistics for the variables utilized in the SEM.Student ages ranged from 8 to 20 years old (mean age of 12.1), and 48.4% of students were male. The majority boarded at their schools, with 61.7% of students reporting sleeping at school. Of the parents, 59.5% of mothers and 36.0% of fathers never attended school.Just under half of participants (47.5%) reported open defecation. Almost all (97.1%) reported consuming pork once or more in the month preceding the questionnaire administration, although 84.2% reported never consuming raw pork. Among students, 43.0% reported that their pigs freely foraged, and 43.8% indicated that their pigs consumed human feces. Only 27.5% of participants reported consuming market-purchased pork. Twenty-eight percent reported noting white spots or cysts in pork at some time in the five years preceding the survey.Overall, 5.4% of all participants had positive serological test result; 3.6% had seizures; while 8.7% complained of frequent headaches (6 or more episodes per month). The model (Figure 1) fit well with the data (RMSEA = 0.031, NFI = 0.951, GFI = 0.985, CFI = 0.964, and χ2/df = 3.473). Probable NCC could be well−represented by the serological test results (β = 0.29, p < 0.001), having seizures (β = 0.42, p < 0.001), and frequent headaches (β = 0.28, p < 0.001). Age (β = 0.25, p < 0.001), boarding (β = 0.40, p < 0.001), educational level of father (β = −0.18, p < 0.001), educational level of mother (β = −0.26, p < 0.001), and household fixed asset score (β = −0.74, p < 0.001) were good indicators of sociodemographics. For behavioral variables, open defecation (β = 0.54, p < 0.05), frequency of pork consumption (β = 0.23, p < 0.05), and frequency of raw pork consumption (β = 0.08, p < 0.05) were adequate indicators. Pork purchased at market (β = 0.40, p < 0.001), pig ownership (β = 0.56, p < 0.001), households allowing pigs to freely foraging (β = 0.85, p < 0.001), pigs consuming human feces (β = 0.88, p < 0.001), and noting white spots or cysts in pork (β = 0.09, p < 0.001) were used to represent sources of pork and pig husbandry variables.Sociodemographic variables (β = 0.33, p < 0.05), sources of pork and pig husbandry (β = 0.26, p < 0.001), and behavioral variables (β = 0.33, p < 0.05) were all direct risk factors for probable NCC among children.Sociodemographic factors could affect the likelihood of being categorized as probable NCC indirectly via sources of pork and pig husbandry (β = 0.27, p < 0.001) and behavioral variables (β = 0.20, p < 0.001). Sources of pork and pig husbandry (β = 0.26, p < 0.001) could affect the likelihood of being categorized as probable NCC indirectly via behavioral variables. The effect size depicted by squared multiple correlations (R2) was 0.46.Table 2 indicates the total effect of all latent variables on probable NCC. The total effect of a given variable group is the sum of its direct and indirect effects. Sociodemographics and behavior variables have the largest direct effect on probable NCC with the same standardized regression coefficient of 0.33. Indirect effect refers to the effect of one variable on another via a mediator (e.g., sociodemographics → behavioral variables → probable NCC), and is the product of the two individual direct effects. For example, the indirect effect of sociodemographics on probable NCC via behavioral variables was 0.07 (0.20 × 0.33), via sources of pork and pig husbandry was 0.07 (0.27 × 0.26), and via both sources of pork and pig husbandry and behavioral variables was 0.03 (0.27 × 0.30 × 0.33). The total indirect effect on probable NCC from sociodemographic variables is obtained by summing all indirect effects (0.07 + 0.07 + 0.03 = 0.13). Although sociodemographics had the largest total effect after adding the indirect effect, the behavior variables played a more key role in the model for large direct effect and strengthening all other variables with probable NCC.This study systematically examines the risk factors associated with NCC using SEM. Our analysis used a latent variable, consisting of serologic test results and children’s self-reported symptoms, to classify students as probable neurocysticercosis cases. All three of the latent variables representing general risk factor categories, which included sociodemographics, sources of pork and pig husbandry, and behavioral variables, contributed to children being characterized as probable NCC.Our results indicate that sociodemographics and behavior had the same direct effect on probable NCC. After adding the indirect effects via behavior, sociodemographics represented the largest total effect in our model. This was followed by the latent variable representing sources of pork and pig husbandry, which affected classification directly as well as indirectly via behavior. However, behavioral variables not only had a large direct effect, but also served as a bridge to strengthen the effect of sociodemographics and sources of pork and pig husbandry on probable NCC.Our sociodemographic latent variable included household income, parental education, age of the student, and school boarding status. Sociodemographic factors have been identified as risk factors for cysticercosis in the existing literature. Previous studies have identified lower levels of education as a significant risk factor for human cysticercosis [27]. Prior research in China also supports the importance of sociodemographic factors, with a study conducted in Gansu province showing that older age was associated with higher risk of cysticercosis among children [28] and other research demonstrating that poor households in China were more likely to report cysticercosis [28,29,30]. Our own work links the sociodemographic variables of parental education and boarding to the likelihood of a child receiving treatment for gastrointestinal worms, with children who do not board at school and have more educated parents being more likely to receive treatment for gastrointestinal worms [16]. The importance of sociodemographics in our SEM model suggests that interventions should be prioritized. Specifically, children in poor households, parents with low education levels, and students boarding at schools should receive more attention.The latent variable representing sources of pork and pig husbandry was significantly associated with probable NCC, suggesting that this would be an impactful area for intervention. Pig ownership, the consumption of home-raised pigs, the presence of freely foraging pigs, and pigs having access to human feces were identified as significant risk factors in our study. The link between pig ownership and human cysticercosis has also been widely reported in studies in Africa and South America [31,32]. Similar to our findings, other studies have demonstrated that consuming market-purchased pork decreases the risk of cysticercosis [31], suggesting that pork sold at markets is more likely to come from pigs which lack the risk factors for T. solium infection. Pigs being allowed to freely roam and having easy access to human feces have been widely recognized as a risk factor for pigs becoming infected [32,33,34,35]. Our SEM suggests interventions related to pig husbandry and meat sourcing such as porcine vaccination, the treatment of pigs with anthelminthic medications as well as improved meat inspection [36] would likely be impactful in our area of study.Behavioral variables not only had a large direct effect, but also served as a critical bridge to strengthen the effect of sociodemographics and sources of pork and pig husbandry on probable NCC. These findings suggest that interventions targeting behaviors may be the most effective in decreasing disease. Similar to our work, open defecation and consuming raw pork are widely recognized as risk factors for disease [3,7,8,29]. Our SEM suggests interventions related to such behaviors would have a large impact. Furthermore, compared with the sociodemographics and sources of pork and pig husbandry, behavioral variables may be easier to change in our population, and may be more easily targeted by school-based educational campaigns than the other areas of intervention. Therefore, health interventions could educate children to regularly use toilets and avoid consuming raw pork.This study has several limitations. The use of cross-sectional data does not allow us to infer causality. The use of self-report data can potentially introduce recall bias especially for variables based on participants’ long-term memory such as having seizures during the last year. In addition, since this study was conducted in extremely remote and poor mountainous areas with limited medical resources, we were not able to obtain neuroimaging allowing for a definite diagnosis of NCC, instead having to rely on the SEM to model this outcome.The strength of using SEM to better understand interventions directed at T. solium cysticercosis is the ability to create latent variables representing broad classes of intervention as well as the unobserved construct of probable NCC cases. Our results suggest that future intervention programs aimed at pediatric populations in our study area would be most effective if they are considering sociodemographic factors, sources of pork and pig husbandary, and especially behavioral factors. Our results suggest that an intervention package consisting of stopping open defecation and consuming raw pork, as well as treatment of pigs utilizing vaccination and anthelminthic medications, particularly in pediatric populations boarding at schools and in poor households may be the most impactful in our study area.Conceptualization, H.Z., S.R., S.F., A.M., and J.O.; Data curation, J.Z.; Formal analysis, Q.W. and J.Z.; Project administration A.M.; Supervision, H.Z., S.R., J.O.; Writing—original draft, J.Z.; Writing—review and editing, H.Z., J.O., Q.W., T.L., A.M., S.F., and S.R.The authors would like to acknowledge the Global Development and Poverty Initiative (GDP, https://www.gsb.staford.edu/seed/research/funding-opportunities) at Stanford University for providing funding to conduct this work.The authors declare no conflict of interest.Structural equation model of probable NCC in school-aged children in Tibetan rural farming areas of western China. Note: Values indicate standardized coefficients; * p < 0.05, *** p < 0.001.Descriptive statistics for the variables utilized in the structural equation modeling (SEM) (n = 2608).Effects of sociodemographics, behavioral variables and sources of pork and pig husbandry on probable NCC.

Med-MDPI/ijerph_3/ijerph-16-05-00781.txt

@@ -0,0 +1,2 @@
+These authors contributed equally to this work.Pregnancy in women with associated endocrine conditions is a therapeutic challenge for clinicians. These disorders may be common, such us thyroid disorders and diabetes, or rare, including adrenal and parathyroid disease and pituitary dysfunction. With the development of assisted reproductive techniques, the number of pregnancies with these conditions has increased. It is necessary to recognize symptoms and correct diagnosis for a proper pharmacotherapeutic management in order to avoid adverse side effects both in mother and fetus. This review summarizes the pharmacotherapy of these clinical situations in order to reduce maternal and fetal morbidity.Endocrine disorders and their treatment during pregnancy represent an important topic for clinicians, endocrinologists, obstetricians, gynecologists, and other medical specialties involved, due to their potential impact upon pregnancy and fetal development. Endocrine physiology for the mother and fetus is constantly changed across pregnancy and many endocrine events appear, partially explained by the development of the maternal–fetal unit in the placenta, a temporary gland. Both mother and fetus adapt using unique mechanisms, at this development during pregnancy, including changes of endocrine system in the mother and fetus, as well as related feedback changes. Initially, the endocrine function of the fetus is totally dependent on the mother because most of the endocrine glands start the hormonal production in the second trimester of pregnancy. Afterwards, the fetus is less dependent on the maternal endocrine function, but the fetal glands are continuously developing both in function and morphology until birth [1,2].Clinical situations needing endocrine treatment during pregnancy are related to two different conditions: to continue treatment for a chronic endocrine disorder diagnosed before pregnancy or for the treatment of symptoms or disorders newly diagnosed during pregnancy. Drug treatment during pregnancy exposes the mother and fetus to adverse effects and potential risks, which depend on the gestational age. Clinical trials during pregnancy are rarely accepted, risky and expensive, requiring very strict ethical regulations and long-term follow-up [3].In the first two weeks of pregnancy, the embryo is under “all-or-nothing” law, meaning that a drug could trigger embryonic death or it has no influence at all on the future development of pregnancy. In the next eight weeks until the end of the first trimester (the so-called major and minor organogenesis period) cellular differentiation and organogenesis occur. Any drug given during this period must be verified concerning the potential teratogenic risk and be free from causing congenital malformations. In the second and third trimester, the drugs could induce fetal toxicity, especially at the central nervous system level.In order to classify their safety during pregnancy, the drugs were divided by Food and Drug Administration (FDA) in five categories [4].Category A—Randomized control trials were not able to demonstrate any fetal risk in the first trimester of pregnancy, as well as no risk at all in the second and third trimester.Category B—Studies on experimental animals were not able to demonstrate a fetal risk but there are no well-controlled adequate studies in pregnant women. Most of the drugs are included in this category.Category C—Studies on experimental animals demonstrated an adverse effect upon fetus and there are no well-controlled adequate studies in pregnant women, but the potential benefits could justify the use of that drug in pregnant women, despite potential risks.Category D—There are important proofs showing the human fetal risks, starting from data of adverse reactions on experimental animals, or human studies, but the potential benefits could justify the use of drug in pregnant women despite potential risks.Category X—Studies on experimental animals and/or in humans from previously reported adverse effects or marketing studies demonstrate fetal anomalies and positive evidence on human-fetal risks. The risk of using the drug during pregnancy is far more than the benefits and women taking these drugs should avoid conception using contraceptive measures.Women whose pregnancies are complicated by endocrine disorders have a major risk of developing maternal and fetal complications that can be minimized by appropriate management and clinical observation [5]. Medical literature has been reviewed in order to obtain data regarding the correct therapy of these clinical situations, some of these being very rare. This paper highlights the pharmacotherapeutic management of these conditions in order to avoid maternal and fetal complications during the gestational and postpartum period.Most endocrinopathies associated with pregnancy can lead to maternal and fetal complications. In order to avoid them it is important to be treated properly. The initial diagnosis of many of them is often difficult due to the overlapping of symptoms occurring during normal pregnancy, such as hyperemesis gravidarum and those suggesting specific endocrine diseases, as well as changes in the baseline for common biochemical measurements resulting from physiological changes during pregnancy [5]. Hyperemesis gravidarum during pregnancy can be manifested by nausea, vomiting, weakness, drowsiness, poor health, irritability, and depression. It is caused by the increase of the human chorionic gonadotropin hormone (hCG) in the body of a pregnant woman. In the first trimester of pregnancy hCG stimulates the formation of hormones necessary for development and maintenance of pregnancy: progesterone and the estrogens estradiol and free estriol [5]. In the normal course of pregnancy, these hormones and human placental lactogen (hPL) will be secreted by the placenta. Placental lactogen hormone actively affects the metabolism, under its influence the intake of amino acids increases to “build” the baby’s tissues and cause nausea, headaches and fatigue to the mother [5].Thyroid function changes during pregnancy due to involvement of chorionic gonadotrophin in thyroid regulation (Figure 1). Thyroid balance is indispensable for the normal development of the fetus. In the first trimester, hCG acts as a strong stimulator for tyroid stimulator hormone (TSH) receptors, increasing the thyroid function and decreasing the TSH level. TSH and hCG changes mirror each other: when hCG increases, TSH decreases, a condition called gestational thyrotoxicosis [6]. After 12 weeks of gestation, when hCG decreases, TSH increases again. Total thyroxin increases above the upper limit of the reference range for nonpregnant women very early. This is related to the increased liver synthesis of thyroxin-binding globulin, being stimulated by high placental estradiol and therefore, the total thyroxin amount is increased [7].However, the free thyroxine is kept at the upper limit of reference range for nonpregnant women, and this should be a normal reference for healthy pregnant women. Small amounts of thyroid hormones are passing through placenta at 6 weeks of pregnancy, confirming their important role in embryogenesis [8]. T4 is inactivated by the placenta using type III deiodinase, which converts T4 into rT3, which is biologically inactive. This activity is increasing in the second part of pregnancy.Maternal TSH is not able to pass through the placenta, but thyrotropin-releasing hormone (TRH) seems to be synthesized in the placenta and contributes to fetal thyroid function, by stimulating fetal pituitary [9].Hyperthyroidism in pregnancy has a prevalence of approximately 0.1–0.4% [10]. Risks: Untreated or insufficiently controlled hyperthyroidism increases the risk of complications in pregnancy: abortion, premature birth, preeclampsia, retroplacental hematomas, and also maternal heart failure [11].Uncontrolled severe hyperthyroidism may cause fetal cardiovascular disorders such as tachycardia, cardiomyopathy, cardiac insufficiency, and fetal hydrops [12]. It is also associated with fetal complications: dead at birth, acute respiratory distress, and newborn with low weight according to his gestational age [13]. Management: The goal of treatment is to normalize thyroid function with a minimum amount of antithyroid medication. All classes of antithyroid drugs cross the placenta in small amounts, but there is little chance of producing fetal hypothyroidism. Patients should be monitored constantly and the amount of drug administered should be adjusted to maintain free thyroxine (FT4) at the upper limit of the reference range. Women diagnosed with small thyroid goiter can obtain the euthyroid status in two ways: (i) using minimal doses of antithyroid drugs and (ii) discontinuing of medication during the last few weeks of pregnancy [14].Antithyroid drugs are divided into two categories:Thiourea derivative propylthiouracil (PTU) classified by FDA in category D [4], inhibits thyroid peroxidase but also blocks the enzyme type I deiodinase, preventing conversion of T4 to T3, which is more biologically active.Imidazole derivatives methimazole (MMI) and carbimazole (CBZ), available in some countries, have the active metabolite methimazole. They inhibit thyroid peroxidase, reducing the synthesis of T4 and T3. Antithyroid drugs are the main pharmacological agents in the treatment of Graves’ disease in pregnant women [15]. Small amounts of PTU, MMI and CBZ cross the placenta and may decrease fetal thyroid function.Thiourea derivative propylthiouracil (PTU) classified by FDA in category D [4], inhibits thyroid peroxidase but also blocks the enzyme type I deiodinase, preventing conversion of T4 to T3, which is more biologically active.Imidazole derivatives methimazole (MMI) and carbimazole (CBZ), available in some countries, have the active metabolite methimazole. They inhibit thyroid peroxidase, reducing the synthesis of T4 and T3. Antithyroid drugs are the main pharmacological agents in the treatment of Graves’ disease in pregnant women [15]. Small amounts of PTU, MMI and CBZ cross the placenta and may decrease fetal thyroid function.Studies have shown that MMI and CBZ classified by FDA category D [4] taken during pregnancy are associated with aplasia cutis (rare disease characterized by growth deficit of skin and hair) and a rare embryopathy consisting of choanal atresia or esophageal and dysmorphic physiognomy [16]; PTU administration in the first trimester was also associated with birth defects, such as urinary tract, face, and neck malformations [17].Although both drug classes are effective in the treatment of hyperthyroidism, as soon as pregnancy is confirmed the treatment should be adjusted. The current guidelines of Endocrine Society [18] recommend discontinuing previous treatment with MMI/CBZ and change to PTU administration during the first trimester of pregnancy and to switch back to MMI/CBZ for second and third trimesters of pregnancy. This regimen is recommended because MMI is associated with an increased risk of congenital malformations while PTU with maternal hepatotoxicity. Doses of these drugs should be adjusted in order to maintain the levels of free-T4 at upper normal values during pregnancy. Serum TSH is primarily used for monitoring thyroid hormone replacement therapy, however, its levels become stable only after four weeks of therapy [19]. Thyroid function in pregnant women with antithyroid medication is recommended to be assessed after 30–40 days from initiation of the treatment and then every 4–6 weeks, according to Endocrine Society guidelines. As pregnancy progresses, drug dose can be reduced, finally discontinuing their medication in the third trimester. In some cases, women with hyperthyroidism may develop anxiety and palpitations. The Endocrine Society’s clinical practice guidelines for management of thyroid dysfunction during pregnancy indicate that the short-term Β-blocker medication propranolol attenuates these symptoms and is not associated with intrauterine fetal growth restriction of the second and third trimesters of pregnancy [18].If pregnant women with hyperthyroidism have associated asthma and beta-blockers are contraindicated, therapeutic alternatives for treating these symptoms are calcium blockers such as verapamil, although fetal safety data regarding its use during pregnancy are more limited [18]. Resistance to antithyroid treatment during pregnancy is unusual and appears due to reduced therapeutic compliance [20]. Side effects to antithyroid medication may occur in a small percentage of patients (3–5%) [21]. The most common side effect is itching with or without rash, which may occur 2–6 weeks after initiation of treatment. In this case, it is recommended that drug administration be stopped and to change to another antithyroid drug. Other less frequent side effects include migratory polyarthritis, lupus-like syndrome, and cholestatic jaundice [21].Agranulocytosis is an uncommon side effect and appears very rarely in patients receiving antithyroid medication—only 0.35–0.37% of cases—being determined by a secondary pharmacodynamic mechanism, either toxic or immunoallergic. Agranulocytosis may be asymptomatic, but in some cases clinical symptoms such as fever, chills, and sore throat may occur. It is diagnosed when granulocyte concentration (neutrophils, basophils, and eosinophils) drops below 100 cells/mm³, which represents only 5% of the normal range. Routine monitoring of granulocytes in women with antithyroid medication is not required [22].Radioiodine I131 therapy is contraindicated during pregnancy because it can cause fetal hypothyroidism, when administered after 10 gestational weeks [23]. Therefore in practice, prior to administering a therapeutic dose of I131 is mandatory a pregnancy test.Treatment options of hyperthyroidism in pregnancy are summarized in Figure 2.Primary maternal hypothyroidism is defined as the presence of elevated TSH levels during pregnancy in the absence of other causes such as TSH-secreting pituitary tumors.Its prevalence in pregnancy is only 0.3–0.5% [24]. The American Thyroid Association (ATA) defined it as elevated levels of TSH ≥ 10 mU/L regardless of serum-free-T4 levels [25]. The most common cause of hypothyroidism in pregnancy is developing autoimmune chronic thyroiditis (Hashimoto), thyroid surgery history or previous treatment with radioactive iodine for hyperthyroidism, goiter or thyroid cancer.Risks: Lack of adequate treatment causes maternal complications: premature birth, low birth weight, perinatal death, eclampsia and preeclampsia, anemia, and postpartum hemorrhage [26]. The child with fetal thyroid agenesis is normal at birth, although T4 values are below 1/2–1/3 of normal range [27]. Instead, children from mothers with hypothyroidism show a moderate but significant decrease in intelligence quotient (IQ) [27].Management: Levothyroxine, classified by the FDA in Group A [4] is the election therapy required in order to achieve a TSH level in the normal range according to gestational age [18,24]. If the pregnant woman has preexisting hypothyroidism and is already receiving L-thyroxine, dosage is determined by gestational age at the time of referral to an endocrinologist:Presentation before 12 gestational weeks: patients will require increasing of their levothyroxine dose by 25–50% [18,24]; doses will remain constant after 16 to 20 weeks of gestation until delivery [28,29]. It is recommended to increase the dose of L-thyroxine from 7 therapeutic doses/week to nine therapeutic doses/week, immediately after confirmation of the pregnancy [18,29].Presentation after 12 weeks gestation, will evaluate thyroid function tests, and L-thyroxine dosage is based on serum TSH levels [28,29]. Thus:
+			  If a TSH level is higher than normal for gestational age—the L-thyroxine dose should be increased;If a TSH level does not exceed the upper normal limit for gestational age—the L-thyroxine dose should not be changed;If the pregnant woman is diagnosed with hypothyroidism and is not being treated with L-thyroxine, thyroid function should be evaluated, and the dosage of L-thyroxine will be chosen based on TSH levels;TSH values >2.5 mIU/mL determine the administration of L-thyroxine in different doses. Therefore, TSH values between 2.5 and 10 mIU/L require 50 L-thyroxine µg daily. For TSH values > 10 mIU/L, the daily dose of L-thyroxine is 100 µg.Presentation before 12 gestational weeks: patients will require increasing of their levothyroxine dose by 25–50% [18,24]; doses will remain constant after 16 to 20 weeks of gestation until delivery [28,29]. It is recommended to increase the dose of L-thyroxine from 7 therapeutic doses/week to nine therapeutic doses/week, immediately after confirmation of the pregnancy [18,29].Presentation after 12 weeks gestation, will evaluate thyroid function tests, and L-thyroxine dosage is based on serum TSH levels [28,29]. Thus:If a TSH level is higher than normal for gestational age—the L-thyroxine dose should be increased;If a TSH level does not exceed the upper normal limit for gestational age—the L-thyroxine dose should not be changed;If the pregnant woman is diagnosed with hypothyroidism and is not being treated with L-thyroxine, thyroid function should be evaluated, and the dosage of L-thyroxine will be chosen based on TSH levels;TSH values >2.5 mIU/mL determine the administration of L-thyroxine in different doses. Therefore, TSH values between 2.5 and 10 mIU/L require 50 L-thyroxine µg daily. For TSH values > 10 mIU/L, the daily dose of L-thyroxine is 100 µg.Treatment options of hypothyroidism in pregnancy are summarized in Figure 3. Iodine supplementation is recommended to optimize the normal function of the thyroid gland during the perinatal period. An additional 150 micrograms of iodine daily is recommended to be administered during preconception, pregnancy, and lactation but, without exceeding the limit, excessive iodine >1100 µg per day could determine thyroid dysfunctions [18,24,25,30].It has a prevalence of 0.25–2.5% in pregnancies [31], being associated with an increase TSH level (TSH 2.5–10.0 mIU/L) and free thyroxine in normal range [25]. Risks: Maternal complications can occur: miscarriages, hypertension, preeclampsia, placental detachment, premature rupture of membranes, neonatal death, and gestational diabetes [32,33,34].Management: Medical therapy in subclinical hypothyroidism diagnosed in pregnant women is controversial. The American Thyroid Association (ATA) recommends L-thyroxine therapy in women with subclinical hypothyroidism and positive thyroid peroxidase antibody (TPOAb) [24].The European Thyroid Association (ETA) and The Endocrine Society recommended L-thyroxine to all pregnant women regardless of subclinical hypothyroidism and TPOAb values, until the TSH values reach typical gestational age, considering that the therapeutic benefits are far greater than the disadvantages [35].Some studies have shown that the number of miscarriages was less in euthyroid pregnant women TPOAb positive and the downward trend in free-T4 and TSH increase with advancing pregnancy [36,37]. During therapy for clinical and subclinical hypothyroidism is necessary to perform periodic thyroid function tests, especially in the first half of pregnancy and whenever there is a change in the treatment [35]. Well-controlled hypothyroidism during pregnancy does not justify additional fetal surveillance. In clinical practice, doses of L-thyroxine immediately after birth should be reduced to dosage prior to pregnancy [36,37]. Some guidelines state that these doses are reduced gradually to pre-pregnancy levels within two weeks postpartum, followed by evaluation [31].Biochemical gestational thyrotoxicosis may be due to increased secretion of hCG, hydatidiform mole, and hyperemesis gravidarum (characterized by severe nausea and vomiting with dehydration, losing 5% of body weight, and ketonuria). This most commonly occurs in multiple or twin pregnancies when serum hCG level is elevated [38,39].Risks: Pregnancy associated with gestational thyrotoxicosis can lead to miscarriages, birth of dead fetuses, premature birth, preeclampsia, low birth weight, intrauterine growth restriction, and congestive heart failure for the mother if the patient is not receiving treatment [40]. Sometimes, in the absence of a proper treatment, pregnant women with transient gestational thyrotoxicosis may develop severe complications such as thyrotoxic crisis, a real diagnostic challenge [41]. Recent studies have shown variable prevalence of this thyrotoxic crisis: in the US this varies from 0.2% to 0.7% [42], in Europe 2–3% [43], and is much higher in Asia—11% [44]. This is an emergency, consisting of a hypermetabolic condition due to the massive and sudden intoxication with thyroid hormones with severe consequences for mother, especially on the central nervous system (coma), cardiovascular system (shock–collapse) and adrenal glands (acute adrenocortical insufficiency) [45].Management: Since hCG values decrease after 10 to 12 weeks gestation, antithyroid medication is not necessary, only parenteral hydration and antiemetics [46].Postpartum thyroiditis is an inflammation accompanied by thyroid destruction, installed in the first year after birth, and it is a variant of the chronic autoimmune Hashimoto because this is characterized by the presence of antithyroid peroxidase (anti-TPO) antibodies [47,48]. This particular pathology occurs in patients with genetic predisposition and the presence of environmental factors (triggers) [49,50]. Pregnancy is a thyroid trigger postpartum by two mechanisms: cell exchange from mother to fetus [51] and the autoimmune rebound that occurs after birth [52].Clinical aspects: Generally, the evolution of postpartum thyroiditis is biphasic, beginning with transient hyperthyroidism, followed by hypothyroidism. Hyperthyroidism occurs at approximately 1–6 months postpartum, with duration of 2–3 months. Its manifestations are marked by fatigue, irritability, and palpitations; weight loss contrasts with the increased appetite. These are often mild and do not require treatment. In more severe cases, a beta blocker (propranolol or metoprolol) can be used to control the symptoms [53].Hypothyroidism occurs at 4–8 months postpartum, spontaneously remitting in approximately 4–6 months. Pregnant women with hypothyroidism can be asymptomatic or may accuse myalgia, arthralgia, fatigue, constipation, loss of concentration, weight gain, depression, and may be confused with postpartum depression [54].Management: The treatment of postpartum thyroiditis is symptomatic. Hyperthyroidism without Basedow’s disease does not require antithyroid medication. In hypothyroidism, substitution with antithyroid medication is required in 30% of cases [55]. Generally, the evolution of the disease is favorable with remittance of symptoms in the majority of cases. In 40% of cases there is a risk of relapse in the next pregnancies [56].Risks factors that lead to the appearance or increase in size of a thyroid nodule during pregnancy are hormone stimulation of the thyroid by hCG and TSH associated with relative immunosuppression and relative iodine deficiency in pregnancy, women who have had two births in the last five years through hormonal stimulation of thyroid in pregnancy and postpartum thyroiditis [57,58]. Diagnostic evaluation of a thyroid nodule discovered during pregnancy should be similar to that of nonpregnant patients, the pregnancy having specific problems only regarding the optimal moment of surgical treatment of a nodule suspected as being malignant [59]. Tests of thyroid function are performed in order to detect hypothyroidism or hyperthyroidism. Under uncertain circumstances, evaluation of thyroid autoantibodies or serum calcium (thyroid tumor marker) may be useful. Thyroid radionuclide scanning is contraindicated during pregnancy [60].Management: Diagnosis and decision-making on general management in the context of a nodular thyroid diagnosed during the pregnancy is primarily based on the results of thyroid ultrasound and fine needle aspiration cytology (FNAC) [60].There are no data to support the need to interrupt the pregnancy if the thyroid nodule that is suspected to be malignant (based on the cytology test or a rapid increase of nodule size) is diagnosed in the first trimester of pregnancy; surgery being indicated in the second trimester (Figure 4). In the case of thyroid nodule diagnosis from the third trimester of pregnancy, the surgery can be delayed after delivery [60].Studies have shown that postponing of the surgery until after birth, does not change the prognosis: newborn’s birth weight, neonatal morbidity, and mortality or fetal congenital malformations [61]. Treatment with radioactive iodine cannot be given during pregnancy and breastfeeding. A new pregnancy should be postponed one year after radioiodine administration [62].During pregnancy, calcium homeostasis is maintained by two hormones: parathyroid hormone (PTH) and 1,25-dihydroxyvitamin D (1,25-D). During pregnancy ~30 g of calcium is transferred to the fetus, especially in the last trimester of pregnancy, needed for bone mineralization. The upper limit of normal values for total serum calcium during pregnancy is about 9.5 mg/dL. Lactating women have lower bone density by up to 8% [63]. According to World Health Organization recommendations, from gestational week 20 until the end of pregnancy 1.5–2 g of elemental calcium is necessary daily [64]. It presents a very low prevalence in pregnancy of only 0.15%. In the literature, only 200 pregnant women were described with this diagnosis [65]. The first case was described in 1931 [66]. A few years later, the first case of neonatal hypocalcaemia producing tetany in a mother with undiagnosed hypercalcemia due to a hyperparathyroidism was described [67].Risks: Half of serum calcium circulates bound to plasma proteins, primarily albumin. During pregnancy, due to hypoalbuminemia, increased renal clearance, and placental transfer to the fetus, total serum calcium is slightly decreased. As a result, the parathyroid glands secrete parathyroid hormone in order to maintain calcium homeostasis. This explains why clinical hypercalcemia in pregnant women due to real hyperparathyroidism is diagnosed only in very severe cases, the final confirmation of this diagnosis being postpartum [68]. When serum calcium levels are very high the following disorders can occur: kidney stones, hypertension, cardiac arrhythmias, and pancreatitis.Management: The only effective treatment for primary hyperparathyroidism diagnosed in the first two trimesters of pregnancy when serum calcium exceeds 11 mg/dL is surgery [69]. For the last trimester diagnosis, treatment strategy is unclear and depends on the severity of hypercalcemia and the clinical condition of the pregnant woman. If surgical treatment was applied, calcium levels should be monitored every 6 h as transient hypocalcemia can occur postsurgery. In this case, calcium should be given intravenously, in the form of calcium gluconate (10%, 10 to 20 mL, for 5–10 min). Such intermittent administration may be repeated or calcium gluconate may be diluted in 5% dextrose or saline isotonic solution (saline) and continuously infused 1 mg/kg body weight/hour [69]. If pregnant women have associated bone disease, postsurgical hypocalcemia can be profound, and requires a more aggressive drug treatment; vitamin D should be supplemented as calcitriol 0.25–0.5 µg/day, classified according to the FDA as category C [4]. Drug treatment is reserved for pregnant women who have surgical contraindications and significant hypercalcaemia. It consists in the administration of oral phosphates 1.5–2.5 g/day, but the occurrence of side effects, such as nausea, vomiting, and hypokalemia, requires reducing the drug dose. Calcitonin and cinacalcet are recommended during pregnancy (category C, according to the FDA) and bisphosphonates should be used only when there is no alternative [69]. Other suitable pharmacotherapeutic measures include adequate parenteral hydration, avoidance of drugs which can cause an increase in serum calcium, such as vitamin D, vitamin A, thiazide diuretics, and early treatment of urinary tract infections [69].Secondary hyperparathyroidism is frequently encountered during pregnancy due to increased demand for vitamin D. Risks: In the absence of adequate amounts of vitamin D, PTH increases. PTH is not currently tested, only in certain cases as measurement of serum levels of calcium and vitamin D is considered enough for the diagnosis.Management: As it is frequent in women with poor intake of vitamin D in the diet, malabsorption syndromes, or pigmented or covered skin, supplementation with 25 hydroxy vitamin D (calcidiol) 400–1200 IU per day is sufficient [70]. Risks: The most frequent cause of hypoparathyroidism in pregnancy is thyroidectomy, and low calcium status can have negative effects on pregnancy and fetal bone development, with increased risk of rickets. It is rarely autoimmune. The first case of hypoparathyroidism in pregnancy was described in 1942, pregnant women may experience cramps, paresthesias, and seizures [71].Management: Drug treatment is 1,25 hydroxylated vitamin D (alphacalcidol), PTH is necessary for stimulated 1 alpha-hydroxylation. The dose will usually need to be increased during pregnancy because an increased amount of vitamin D is needed [65]. During pregnancy, the fetoplacental unit produces large amounts of steroid and peptide hormones: CRH (corticotropin releasing hormone) which increases after eight weeks of amenorrhea up to 100 to 1000 times compared with nonpregnancy values; proopiomelanocortin increases after eight gestational weeks, has a maximum concentration at 20 gestational weeks, and then remains stable until birth [72,73]. The plasmatic cortisol doubles its values, mainly through the increase of transport protein under the action of estrogens. The secretion of aldosterone, deoxycorticosterone, and renine is elevated by the secondary aldosteronism [74,75].Adrenal disorders in pregnancy are relatively rare during pregnancy; their pharmacotherapy is important because it is associated with maternal and fetal morbidity and mortality. Diagnosis of these conditions in pregnancy is often difficult, as pregnancy is marked by a series of changes in the endocrine system, including activation of the renin–angiotensin–aldosterone system and the hypothalamic–pituitary–adrenal axis [76].Congenital adrenal hyperplasia (CAH) refers to a group of autosomal recessive genetic disorders that arise from failure of steroidogenesis, resulting in the reduced production of cortisol and adrenocorticotropic hormone (ACTH) secondary to increased production [77]. Risks: There is very little data in the literature about pregnancies in women with CAH.Management: The goal of drug therapy in CAH in pregnant women is to correct the deficiency of cortisol and suppress ACTH overproduction. The treatment of choice is hydrocortisone 10–15 mg/m2/day divided in two or three doses per day, with a higher dose in the evening. Compared to dexamethasone, it is preferred because it is metabolized by the enzyme 11 beta-hydroxysteroid dehydrogenase-2 (11β-HSD2) in placenta and it does not affect the fetus [78]. It is recommended to check 17-OH-progesterone and androgens (testosterone and androstenedione) at least once per trimester. They are increased during pregnancy but normal levels for pregnancy have not been established. Prednisolone, or dexamethasone, which has a longer half-life, may be used if the control is not carried out only with hydrocortisone. They are associated with Cushingoid-like side effects: weight gain and stretch marks [79].Prednisone is not recommended, since conversion to prednisolone is insufficient used in small doses required for pregnant women with CAH [79]. If mineralocorticoid therapy is necessary, fludrocortisone is administered at 0.05–0.3 mg/day; the dose is adjusted to maintain plasma renin activity at lower levels, no dosage adjustment is necessary for drugs administered in pregnancy.Dexamethasone treatment in women with CAH starts before the 9th week of pregnancy, before the onset of adrenal androgen secretion and is designed to significantly reduce genital masculinization of women affected by suppression of excessive production of adrenal androgen. Dexamethasone, unlike hydrocortisone, escapes inactivating placental enzyme 11β-HSD2, has a longer half-life, and suppresses the secretion of ACTH. The optimal Dexamethasone dose is 20 µg/kg/day divided in three doses. It is recommended to start treatment as soon as pregnancy is confirmed, and no later than nine weeks after the last menstrual period [80,81].The prevalence of primary adrenal insufficiency (Addison’s disease) during pregnancy is very rare—~1:3000 pregnancies—most women being diagnosed before conception [82]. Addison’s disease (AD) is characterized by deficiency of adrenocortical hormones: androgenes, glucocorticoids, and mineralocorticoids. Glucocorticoid and mineralocorticoid deficiency symptoms are nonspecific: weight loss, vomiting, lethargy, and skin hyperpigmentation, which is due to increased ACTH stimulation of melanocytes.Because the symptoms of pregnancy resemble the clinical suspicion of AD, it must be considered in pregnant women with other associated autoimmune diseases [83]. Besides biochemical pregnancy: hyponatremia, hyperkalemia, increased blood urea and hypoglycemia, low serum cortisol at 9 am, and poor response to synthetic ACTH (Synacthen test). These tests are not as easy to interpret during pregnancy because the increased physiological cortisol levels may lead to “normal” results [83]. Risks: Placental unit autonomously produces steroids, and therefore maternal adrenal insufficiency causes no problems in the fetus [83].Management: The right treatment produces no maternal and fetal complications, especially after the synthesis of cortisone in 1950 [84]. However, there were reports of fetal growth restriction in babies born from mothers with untreated disease [85].Maintenance treatment in pregnancy includes replacement of glucocorticoid with hydrocortisone and mineralocorticoid with fludrocortisone. Hydrocortisone (category C—FDA) is the treatment of choice for glucocorticoid substitution; unlike other available glucocorticoids, it is degraded by the enzyme 11β-HSD2, it does not cross the placenta, and effects only occur in the mother’s body. The recommended dose is 12–15 mg/m2 body surface with 50–75% of the daily dose administered in the morning to mimic the physiological secretion of cortisol [86,87]. Because free cortisol increases gradually with advancing pregnancy, most women with AD require a daily dose of hydrocortisone increased by 20–40%, e.g., 5–10 mg in the third trimester of pregnancy [86,87]. In amniocentesis and caesarean section an initial dose of 100 mg of hydrocortisone is given intravenous (iv) or intramuscular (im) and then, every 6–8 h, the dose is repeated, with gradual reduction in the next 48 h [86]. Doses are increased in women with hyperemesis gravidarum that can be easily mistaken for an adrenal crisis. On the other hand, even hyperemesis can easily trigger an adrenal crisis. Treatment of acute adrenal crisis (acute adrenal insufficiency) is a medical emergency and consists in the immediate intravenous bolus administration of 100 mg of hydrocortisone, followed by injection of hydrocortisone 50–100 mg every 6–8 h (or as a continuous infusion of 200–300 mg/24 h) and intravenous saline (originally 1 L per hour, then 200 mL per hour), with regular monitoring of blood pressure, heart rate, and serum electrolytes [86,88].Fludrocortisone (category C according to FDA) is initiated at a dose of 0.1 mg (dosage ranges from 0.05 to 0.25 mg), and usually the same dose is continued for all the duration of pregnancy. During pregnancy the production of progesterone continuously increases, which exerts antimineralocorticoid activity in vitro and in vivo [89]. Hydrocortisone also acts on the mineralocorticoid receptor and by increasing the dose during the third trimester of pregnancy it can compensate for this; 40 mg hydrocortisone is equivalent to 0.1 mg fludrocortisone [89,90]. To monitor treatment with mineralocorticoid plasma renin may not be used because renin secretion increases physiologically during pregnancy; supine blood pressure is monitored along with serum electrolytes and urinary excretion of sodium [89].(A) Cushing’s SyndromePregnancy is very rare in women diagnosed with Cushing’s syndrome (CS) due to hypercortisolemia, hyperandrogenemia, and/or hyperprolactinemia that lower fertility [91]. Unlike nonpregnant women, the etiology of CS in pregnancy is represented by benign adrenal tumors produced by the aberrant development of adrenal receptors for LH and hCG [91]. Risks: In the absence of adequate treatment in pregnant women with CS maternal and fetal morbidity can occur due to hypertension, preeclampsia, eclampsia, gestational diabetes, congestive heart failure, and pulmonary edema causing miscarriages, stillbirths, and early neonatal deaths.Management: CS treatment in pregnancy should be individualized because improvements or exacerbations of CS may occur; irregular production of placental corticotropin-releasing hormone (CRH) exacerbates hypercortisolemia during pregnancy [92].If CS is diagnosed in the first trimester of pregnancy, drug therapy should be considered depending on the severity of hypercortisolemia. Tumor removal should be considered from the second trimester [93]. In the case of CS diagnosed in the third trimester of pregnancy, drug treatment is preferred, and surgery should be performed postpartum [92]. The recommended pharmacological agents are the steroidogenesis inhibitors metyrapone, ketoconazole, and mitotane [92,94].Metyrapone, category C according to FDA [4], is the drug of choice: in 69% of cases it has been shown to reduce hypercortisolemia, but it does not totally control the symptoms [94]. The most troubling adverse effects are determined by increasing the precursors such as 11-deoxycorticosterone, severe hypertension with risk of preeclampsia, intrauterine growth retardation, coarctation of the aorta, and adrenal insufficiency of the newborn [95]. Although animal studies have shown that Metyrapone easily crosses the placental barrier, no human fetal abnormalities were reported [95]. Ketoconazole, category C according to the FDA [4], has been used less during pregnancy because of adverse effects: antiandrogenic effect, transient neonatal hypoglycemia, and intense teratogenic effects (only demonstrated in animal studies) [96].Mitotane (category C—FDA) and other steroidogenesis blockers such as Aminoglutethimide (Category D—FDA) are rarely used and are highly teratogenic, causing fetal masculinization [97]. (B) Conn DiseaseRisks: Primary aldosteronism (Conn disease), characterized by hypertension and hypokalemia, is extremely rare in pregnancy and should be considered in a pregnant woman with uncontrollable, atypical hypertension, were other causes of preeclampsia were excluded. The association of hypokalemia and hypernatremia with hypertension requires further investigation for adrenal adenoma, adrenal hyperplasia, and adrenal carcinoma [98]. Because during pregnancy urine and plasma aldosterone increase, biochemical test results must be correlated with the normal values for pregnancy.Management: Drug treatment of primary aldosteronism is limited because in pregnancy are contraindicated aldosterone antagonists such as Spironolactone—a potassium-sparing diuretic (category C—FDA) [4]—angiotensin converting enzyme inhibitors, and angiotensin II receptor antagonists. Also, Spironolactone has antiandrogenic effects and feminizes the male fetus [90].Eplerenone, mineralocorticoid receptor antagonist (category B—FDA), has been reported to be successfully used in a single case [90], the same as Amiloride, a potassium-sparing diuretic [99]. Potassium salts can be used when blood pressure is controlled by methyldopa, nifedipine, or labetalol.Pheochromocytoma is a tumor of the adrenal medulla (90%) but could also be localized in other areas of the sympathetic nervous system. Risks: It produces adrenaline and noradrenaline and is a secondary cause of high blood pressure in pregnancy [100]. The prevalence in pregnancy is very low (0.007%) and can be fatal for the pregnant woman and the fetus because of extreme paroxysms of blood pressure, often imitating preeclampsia, but with much higher fetal mortality [100]. Typical symptoms include palpitations, excessive sweating, labile blood pressure, causing orthostastic hypotension, and significant hypertension. There are increased catecholamines in the urine and specific metabolite vanilmandelic acid, due to the excessive sporadic release of catecholamines from the tumor, which may be located in the adrenal medulla or anywhere along the sympathetic chain [101].Management: The goal of drug treatment is to normalize blood pressure and pulse. Among α-adrenergic-blocking pharmacological agents the most often used is phenoxybenzamine (category C according to FDA), which is safe in pregnancy and can be initiated at doses of 10 mg/day; the dose may be increased to 10 mg every 2 to 3 days until symptoms and blood pressure are controlled [102]. Subsequently, dosage adjustment may be necessary because phenoxybenzamine has a half-life of ~24 h and therefore the effect is cumulative. The usual dose is 40–80 mg/day in divided doses. The maximum dose of phenoxybenzamine is 1 mg/kg/day [102].Other selective alpha1-antagonists for use in pregnancy are Prazosin (category C—FDA) and Doxazosin (Category C—FDA); their α-adrenergic receptor blocking activity increases blood volume and can improve congestive heart failure and angina pectoris, if they are combined; it is recommended to administer them for 7 to 14 days if surgical treatment is indicated [102,103].Beta-blockers such as propranolol (category C—FDA) are recommended only after taking alpha blockers. Hypertensive crises can be treated with intravenous phentolamine (category C—FDA).All hormonal axes of the body are directed by the pituitary gland; most are activated and operational during pregnancy except for the ovarian hormone axis, which is inhibited due to high serum levels of estrogens and progesterone [104]. In pregnancy the secretion of prolactin from the anterior pituitary lactotroph cells increases;conversely, the pituitary gland grows in size up to two or three times. As opposed to lactotroph cells, the rest of the cells in the anterior compartment of the pituitary remain at pre-pregnancy size or even smaller [104]. Starting up with 16–17 weeks of pregnancy, the placenta secretes the placental growth hormone which in turn leads to increase of insulin-like growth factor (IGF). IGF inhibits pituitary somatotropic cells and the secretion of somatotropic hormone (STH) decreases [105]. The corticotropic axis is also modified because placenta secretes corticotropin releasing hormone (CRH) [105]. TSH (thyrotropic hormone) plasma levels decrease in parallel with the increase in hCG (human chorionic β-gonadotropin) levels, which has also affinity for the TSH receptor [106]. Even if the vasopressin concentration remains stable, in pregnancy it decreases the osmotic threshold of thirst and plasma arginine vasopressin secretion, consequently, the plasma osmolarity and natremia decrease [106].Pituitary disorders are uncommon in pregnancy, some occur during or after pregnancy, such as Sheehan’s syndrome (postpartum hypopituitarism), others are preexisting to the pregnancy, such as prolactinomas, Cushing’s disease, and acromegaly. Pituitary insufficiency prevalence is very low, as there are only about 42 diagnosed cases/million [107].Prolactinoma is the most common cause of persistent hyperprolactinemia and represents 50% of pituitary tumors [108]. These are predominantly benign tumors and classified as microprolactinomas (size < 10 mm) or macroprolactinomas (size ≥10 mm).Risks: Effect of prolactinoma on pregnancy. Hyperprolactinemia has no effect on pregnancy or fetal development, but there is a risk of estradiol receptor-related tumor expansion [109].Effect of pregnancy on prolactinoma: During pregnancy, the pituitary undergoes global hyperplasia due to increased serum levels of estrogenes leading to tumor growth and the possibility of visual loss [108]. Microadenomas of less than 10 mm rarely cause problems during pregnancy, while macroprolactinomas larger than 10 mm require increased attention regarding pharmacotherapeutic management [110]. Women with macroprolactinomas require visual field testing every trimester and in case of visual impairment treatment with dopamine agonists (DA) is initiated [111,112]. When there is a risk of optic chiasm compression DA are recommended throughout pregnancy.Management during pregnancy: The objective of pharmacological treatment depends on the size of the adenoma, the target being maintaining the prolactinoma away from the optic chiasm and levels of prolactin to be within the normal range for pregnancy [110]. Lack of efficacy of drug therapy requires transsphenoidal surgery for removal of adenomas, only in the second trimester [111].Prolactin secretion is under negative feedback control by dopamine, and consequently DA like Bromocriptine and, more recently, Cabergoline, is the mainstay of treatment in nonpregnant as well as pregnant women. Bromocriptine, category B according to the FDA [4], is the DA of choice for pharmacologic treatment of prolactinoma during pregnancy, some studies confirming no adverse effects in treating over 6000 cases [113]. The effective dosage is 2.5–5 mg/day, seldom being required at 7.5 mg/day or more. Bromocriptine is effective not only for the normalization of prolactin level, but also for reducing tumor size [113].Cabergoline, category B according to FDA [4], is the alternative DA, because it has a good safety profile without teratogenic effect; although it is not recommended for prolactinoma pharmacotherapy in pregnancy because long-term use (over 1 year) has been associated with fibrosis of the heart valves and constant supervision by echocardiography is necessary [114].Management after pregnancy: After birth, microadenoma presents a negligible risk and breastfeeding is permitted. In case of macroadenoma, breastfeeding is not recommended. It is also recommended to control prolactin and pituitary morphology three months after birth [115]. Sometimes it was found a total remission of a microprolactinoma (probably by vascular effect) after birth [116].The occurrence of pregnancy in women with acromegaly is very rare, because the hypersecretion of growth hormone from the pituitary somatotrophs leads to lack of ovulation. The simultaneous appearance of hyperprolactinemia in 40% of cases, with or without macroadenoma, also affects fertility [117]. The first case of normal pregnancy in a woman with acromegaly was reported in 1954 [93].Risks: Comorbidities in pregnant women are numerous: hypertension, preeclampsia, gestational diabetes due to insulin resistance resulting from the anti-insulin effect of growth hormone (GH), and heart disease [60,77]. There is also a potential risk of tumor expansion due to growth stimulation by estrogen, which leads to neurological complications and/or visual complications due to proximity to the optic chiasm, and transsphenoidal surgery is recommended [117].Management: Dopamine agonists can control pregnancy acromegaly in 10% of cases [118]. Bromocriptine (category B—FDA) is considered safe and is more efficient when there are cosecretant GH and prolactin (PRL) tumors. A higher prevalence of microsomic fetuses in pregnant women treated with bromocriptine was found however compared to untreated women [118]. Somatostatin analogs, octreotide (category B—FDA), or lanreotide (category C—FDA) cross the placental barrier;their safety in pregnancy has not been precisely established [119]. All five subtypes of receptors and somatostatin, including somatostatin receptor type 4 (SST4) which has a reduced affinity for Octreotide, are present in the placenta and umbilical cord, suggesting that the maternal–fetal barrier carries a weak functional response to somatostatin analoganalogs [119]. For the treatment of acromegaly, Octreotide is also administered before pregnancy, as a result, the drug is administered in the first few weeks after conception, but no fetal malformation was reported [119].Studies have demonstrated that administration of somatostatin analoganalogs in pregnant women with acromegaly has been correlated with intrauterine fetal growth retardation and birth of children with low weight (growth-restricted), compared to those who did not receive them [119]. A partial explanation would be the transient reduction of the uterine artery flow and of systolic velocity after the administration of short-acting Octreotide [120]. Pegvisomant (category B—FDA) has proven to be a safe and effective drug; fetal concentration is minimal suggesting a low or absent transplacenta passage [120,121]. Although medical treatment of pregnant women with acromegaly is not associated with major side effects of the mother or fetus, it should be discontinued during pregnancy. If medical treatment continues, close monitoring of fetal development is recommended.(A) Hypopituitarism during PregnancyRisks: Pregnant women with hypopituitarism (deficiency of hormones secreted by the anterior pituitary), usually secondary to pituitary tumors, have a high risk of cerebrovascular disease and, therefore, pharmacotherapy is very important during pregnancy [122].Management: Glucocorticoid of choice in pregnant women with ACTH deficiency is hydrocortisone 20–30 mg/day (two-thirds of the total dose in the morning and a third in the evening), (category C—FDA). In some clinical situations, the amount of hydrocortisone may be reduced by one third of the total dose, as the effects of cortisone are boosted by estrogen during pregnancy due to increasing in corticosteroid-binding globulin [123]. The therapeutic alternative to hydrocortisone is represented by synthetic corticosteroids: 5.0–7.5 mg prednisone daily and dexamethasone 0.5–0.75 mg per day, (category C—FDA), mentioning that they are not boosted by estradiol. Because these pregnant women have ACTH deficiency, aldosterone secretion is normal and there is no need of mineralocorticoid replacement therapy.For Levothyroxine (category A—FDA), the usual dose of 0.1–0.2 mg is adjusted so free-T4 levels are included in the upper normal range for gestational age [55] because small values of free-T4 especially in the first quarter have a negative impact on psychomotor child development [124]. TSH is not a useful marker to monitor treatment with Levothyroxine during pregnancy.Genetic recombinant GH treatment is not approved for administration during pregnancy, although some studies have shown women who became pregnant during treatment with GH and it has no longer been administered in the first trimester had pregnancies with good evolution and development of normal children [125]. Desmopressin (category B—FDA) administration is considered to be safe during pregnancy, nonteratogenic. Although it has similar chemical structure to vasopressin, it does not seem to increase the frequency or amplitude of uterine contractions although it has oxytocin-like properties [126].(B) Postpartum Hypopituitarism: Sheehan SyndromeSheehan’s syndrome (SS) is a very rare condition resulting from infarction and necrosis of the pituitary gland [127]. Risks: The pituitary gland, increased physiologically during pregnancy, causes upper pituitary artery compression and hypotension secondary to severe bleeding during childbirth and postpartum causes small blood vessel spasm, apoplexy and subsequent necrosis of the anterior pituitary. Management: Hormone replacement therapy is fundamental in SS. Glucocorticoid administration is done before treatment with L-thyroxine [127]. (C) Panhypopituitarism during PregnancyPanhypopituitarism is characterized by a global deficit in anterior and posterior pituitary hormones. A pregnancy in these cases is rare, with an unpredictable evolution as there are only a few reports of cases in the literature. Thus, Shinar et al. recently reported four cases of pregnant women, aged between 26 and 31 years, diagnosed with panhypopituitarism without any detectable pituitary function modification in repeated tests. They delivered to term with spontaneous labors. The pregnancies were obtained using in vitro fertilization and these have been treated all the pregnancy period with hormonal replacement therapy: glucocorticoids, vasopressin, and thyroxin. In none of the four cases the breastfeeding was possible due to the lack of prolactine [128].Tonda et al. presented the case of a young pregnant woman with severe clinical disturbances, such as visual disturbances, intense headache, insipid diabetes, polydipsia, and polyuria, which gave birth prematurely (gestational age 32 weeks). Blood tests revealed panhypopituitarism and imaging explorations showed a pituitary lesion. After six weeks of adequate hormone replacement therapy the imaging explorations showed a normal pituitary gland [129].Diabetes insipidus (DI) is a rare disorder characterized by antidiuretic hormone (ADH) deficiency. Produced in the hypothalamus and stored in the posterior pituitary, ADH acts directly on the distal tubules to allow water reabsorption in response to dehydration [121]. DI can be a result of ADH hyposecretion in the posterior pituitary gland (central DI, endocrine) or lack of response to its action in the kidney (nephrogenic DI, peripheral) [121]. As a result, there is an inability to maintain normal plasma osmolarity (275–295 mmol/L) and urine concentrations (>300 mOsmol/kg). Polyuria (>3 lit/24 h) stimulates increased thirst with compensatory polydipsia [121]. Plasma osmolality decreases by 5 to 10 mOsm/kg shortly after conception and throughout pregnancy remains low [130]. Accordingly, ADH level remains low throughout pregnancy, especially in the first and second trimester of pregnancy [131]. The ability to concentrate urine remains in the normal range [132]. The prevalence of DI during pregnancy is low, ~1:30,000 [133]. Transitional DI may occur in the last trimester of pregnancy, due to increased glomerular filtration rate, renal prostaglandins increase with ADH antagonism, and placental production of vasopressinase, an ADH degrading enzyme [134]. Risks: Morbidities may occur due to preeclampsia, acute fatty liver of pregnancy, and HELLP syndrome (hemolysis, elevated liver enzyme levels, and low platelet levels). The lack of therapeutic effect of vasopressin is due to the significant increase in plasma clearance of antidiuretic hormone (ADH) [134]. This explains also the good response to desamino-D arginine vasopressin (1-desamino-8-D-arginine vasopressin) DDAVP (Desmopressin, Minirin) [135], which is not metabolized by vasopresinase.Management: DI pharmacotherapy in pregnancy is approached differently depending on etiology.(A) Central and Transient Diabetes InsipidusAt the beginning of the eighteenth century, drug therapy in pregnancy DI was aggressive hydration and increased fluid intake. At the beginning of the nineteenth century posterior pituitary extract was first used in the treatment of pregnancy DI [136]. The extract has an irritating and sensitivity effect on the nasal mucosa and it induces formation of antibodies. Since this bovine or porcine extract contains oxytocin in addition to vasopressin, undesirable complications have occurred [136,137]. New drugs have been synthesized containing exogenous vasopressin: pitressin��an aqueous solution of vasopressin (vassopresin 20 units/mL); pitressin-tannate—oily solution (vasopressin 5 units/mL); and lypressin (lysin-8-vasopressin)—nasal spray (vasopressin 2 units per spray) but even they are rapidly catabolized by placental vasopressinase due to chemical structures similar to endogenous vasopressin [138]. Therefore, in women with DI receiving this medication, disease control was not achieved because as the pregnancy advances and placental vasopressinase level increased.Currently the elective drug treatment for central, preexisting DI and transitional pregnancy DI is 1-deamino-8-D-arginine-vasopressin (DDAVP) (desmopressin, minirin), category B—FDA. DDAVP is a selective activator of V2 receptors in the distal nephron inducing increased water retention with the emergence of hyponatremia [139]. DDAVP is an analog of vasopressin with a modified N-terminus and a change in the chemical structure of arginine at position 8, and is therefore resistant to the catabolic action of placental vasopressinase [140]. If central preexisting DI was controlled by DDAVP before pregnancy, treatment may be continued during pregnancy, sometimes requiring an increase in dose. Diagnosis of DI during pregnancy requires the same drug therapy to that outside of pregnancy [141]. DDAVP can be administered intranasally, subcutaneously, intravenously, or orally in a starting dose of 10 mcg intranasally before sleep to prevent nycturia. The intranasal or sublingual route is recommended during pregnancy, rarely parenteral [142]. Higher doses can sometimes be necessary: the effective dose depends on the severity of DI (partially or totally deficient vasopressin) and individual parameters, such as absorption, which affect the pharmacokinetics and pharmacodynamics of the drug. The therapeutic effect is long [141,142]. DDAVP standard doses are between 1 and 2 µg once or twice daily by injection; 5 to 20 micrograms twice or three times a day, nasally; and 60 to 120 µg two or three times a day, via the sublingual route [141,142]. For nasal administration, either a graduated tube is used (allowing a variable amount of product to be administered) or a nasal spray, containing about 10 µg per puff of the drug. Sublingual administration is simple, using desmopressin lyophilized and dissolved under the tongue [135]. Posology is adjusted to the symptoms: the therapeutic target is the elimination of polyurodipsic syndrome and the avoidance of drug overdose [141,142]. Usually the pregnant woman shall establish the optimal dose herself, increasing the evening dose gradually until she no longer drinks water at night and nycturia disappears, and then is advised to use the same procedure during the day.The main side effect of DDAVP is water intoxication due to overdose, the volume of water ingested is disproportionate to the removed volume because desmopressin blocks diuresis and causes hemodilution. This results in hyponatremia, oliguria, edema, headache, lethargy, nausea and vomiting [135]. Seizures and coma may occur due to hyponatremia. In order to avoid the risk of water intoxication, the patient should be advised to drink only when thirsty. It is also recommended to omit a dose of DDAVP once a week, for example. By producing transient polyuria any gradually accumulated fluid retention is eliminated, thus correcting fluid balance. The pregnant woman may then resume the usual dose of desmopressin [143]. Monitoring urine output, fluid intake restriction and monitoring of serum sodium osmolality and urine osmolality can minimize the risk of osmotic complications secondary to DDAVP administration in pregnant women with DI [135].In transient pregnancy DI, DDAVP can usually be stopped a few days or weeks before birth because postpartum, vasopressinase produced by the placenta disappears [143]. Although classified by the FDA as risk category B [4], there continues to be a lack of controlled studies for DDAVP to provide data on its safety for mother and fetus [126]. The great advantage of DDAVP is that being a selective V2 receptor activator, it avoids stimulating uterine contractions or blood pressure compared with exogenous vasopressin forms [126].(B) Nephrogenic Diabetes InsipidusManagement: Congenital nephrogenic DI therapy in pregnant women is very rare and requires the corrections of the abnormalities of calcium or potassium. Hydrochlorothiazide is used conventionally to treat nephrogenic DI [144]. Although included in category B by the FDA, the pregnant women need precautions to avoid hypovolemia and hyponatremia in mother and fetus. Hydrochlorothiazide can create difficult problems during labor due to the difficulty of balancing abundant polyuria by an equivalent fluid intake [144]. Pharmacotherapeutic management of endocrine diseases associated with pregnancy is complex and covers fetal, neonatal, and maternal issues that may occur during gestational period. Women with various antenatal diagnosed endocrine diseases should be counseled about the possible complications that may occur along the way. Medication and monitoring should be made by a multidisciplinary team; excessive doses should be avoided.The management is much more difficult if endocrine disorders are discovered during pregnancy; the symptoms are sometimes indistinguishable from those that occur physiologically during pregnancy. Thyroid disorders are the most common endocrine problems associated with pregnancy, but subclinical thyroid disease treatments generate controversy; the therapeutic target in these cases should be to maintain euthyroidism throughout pregnancy.The most common of pituitary tumors are microprolactinomas. If they have macroprolactinomas, pregnant women should be tested constantly for visual field. In both cases it is important that patients be counseled about the occurrence of intense headache requiring emergency medical presentation as tumor expansion may compromise the optic chiasm.Clinical suspicion for Cushing’s syndrome, pheochromocytoma, and Conn’s syndrome should be considered in women with high blood pressure who do not respond to classic, conventional treatment during pregnancy.D.C., A.O.D., and K.S.D. contributed to drafting and writing the manuscript and were responsible for the collection of relevant literatures. S.S., A.T., and A.M. contributed to the conception of the figure, interpreted the results, and revised the manuscript critically for important intellectual content. All the authors read and approved the final manuscript. This research received no external funding.This work was supported in part by the Internal Grant 162/2018 of University of Medicine and Pharmacy of Craiova.The authors declare no conflicts of interest.Pregnancy and hypothalamic pituitary thyroid axis. T4—Tetraiodothyronin, T3—Triiodothyronine, TRH—Thyrotropin-releasing hormone, TSH—Tyroid Stimulator Hormone, TRAb—TSH Receptor Auto Antibodies, hCG—human Chorionic Gonadotropin, D3—type 3 iodothyronine deiodinase.Pharmacotherapeutic management in maternal hyperthyroidism.Pharmacotherapeutic management in maternal hypothyroidism.Management algorithm of palpable thyroid nodule during pregnancy.

Med-MDPI/ijerph_3/ijerph-16-05-00782.txt

@@ -0,0 +1 @@
+Yang Liu and Shenzhi Song contributed equally to the paper.Changes in lifestyle and food environment have created a heavy burden of obesity and chronic disease in China. However, measurements of the food environment have been rarely reported in China or other countries with similar food cultures; this measurement shortage is partially due to the lack of valid and reliable measurement tools. The aim of the present study was to adapt and validate a Chinese version of the Nutritional Environment Measurement Survey for Stores (C-NEMS-S). Categories and items of the NEMS-S were culturally adapted to fit the Chinese population and included grains, dry beans, starchy tubers, vegetables, fruits, seafood, meat and poultry, dietary oils, milk, bread, instant noodles, and beverages. A scoring sheet for each food category was created to measure availability, quality, and pricing. Then, the C-NEMS-S was validated in 10 large-sized supermarkets and 10 convenience stores in Shenyang, China. Two trained raters performed their evaluations separately at the same store. The intra-class correlation coefficient (ICC) of the availability composite score was 0.98. All food measures had a moderate or good ICC (0.41 to 1.00). The kappa for each food measure ranged from 0.52 to 1.00. C-NEMS-S was able to show the difference in healthy food availability between large-sized supermarkets and convenience stores, as well as the price differences between healthier options and regular options. Large-sized supermarkets had a significantly higher total score (p < 0.001) and healthier option availability for all food measures (all items were statistically significant (p < 0.05), except sugar-free beverages). Healthier options cost more than regular options for grains, milk, bread, and instant noodles (from 4% to 153%). The adapted C-NEMS-S can be used to measure the consumer food environment in stores in China.The food environment, which could also be called a nutrition environment, refers to the physical presence of food that affects a person’s diet, a person’s proximity to food store locations, the distribution of food stores, food service, any physical entity by which food may be obtained, or a connected system that allows access to food [1]. The food environment plays an important role in food choice, eating patterns, and energy intake [2]. It is widely reported that the food environment is associated with the increasing epidemic of childhood and adult obesity [3,4,5,6]. Recent efforts on obesity prevention have partly focused on studying the role of environmental factors, and national level policies on the food environment have been implemented in many countries [7,8,9].It was reported that the food environment within stores, as it relates to healthy food availability, quality, and price, may contribute to the dietary intake disparities in many countries [10,11,12]. Specifically, higher quality food stores, such as specialized fruit and vegetable (F&V) markets, were associated with greater F&V intake after controlling for individual-level characteristics, according to a study conducted in a Brazilian city [10]. The proportion of in-store shelf space for skim, 1% fat, and 2% fat milk was reported by a study conducted in the United States to be associated with low-fat milk consumption [11]. Besides price and convenience, purity, freshness, association with specific places, and ‘Pakistani-ness’ were reported as the bases for making decisions about ‘good food’ in Pakistan [12].Over the past several decades, China has been undergoing economic development and urbanization at an accelerated rate [13]. In 1950, 13% of people in China lived in cities. By 2010, the urban share of the population had grown to 45%; this is projected to reach 60% by 2030 [14]. Meanwhile, the total energy intake of Chinese people increased by 80%, from 1635 kcal/capita/d in 1961 to 2943 kcal/capita/d in 2003. By 2006, less than 1 percent of people were getting 10 percent of their energy from fat, 44 percent were getting 30 percent of their energy from fat, and nearly two-thirds were getting more than 10 percent of their energy from animal products [15]. The prevalence of overweight and obese children increased from less than 2% in 1985 to 15% in 2010, and the overall rate of overweight and obese people in major cities is over 20% [16,17,18]. However, the food environment after such huge societal change in China, and its contribution to the obesity prevalence in China, has yet to be described. Answers these questions would help provide evidence for food environment improvement and potentially limit the obesity epidemic in China.Lack of valid, reliable measures of food environments hinders food environment studies in China and other Asian countries. Many food environment measurement tools exist for other countries [19,20]. Among these, the Nutritional Environment Measures Survey in Stores (NEMS-S) is an observational measure to assess the food environment within retail food stores [21,22]. The NEMS-S measures the availability of healthier options, price, and quality for 10 categories of food products. The NEMS-S is characterized by its relative ease of use and ability to adapt to different settings and populations [23,24,25,26,27]. The 10 categories of food products included in the NEMS-S were based on the types of food products that contributed the most fat and calories, which are different from those in China, to the American diet. So, before using the NEMS-S in China, food products were culturally adapted in this study to fit the Chinese food culture. This study aimed to adapt and validate a Chinese version of the Nutritional Environment Measurement Survey for Stores (C-NEMS-S) in order to provide a tool for measuring the food environment in China for future studies.The basic principle of the NEMS-S is to gather information on comparable categories and species of food across stores, which includes the availability, price, and quality of healthy options for a particular food. Three steps were conducted to adapt the NEMS-S to the C-NEMS-S: selecting food measures, defining healthy food or healthier options, and adapting scores for availability, price, and quality.Food measures were selected based on the foods that contribute the most energy to the Chinese diet, as listed in the China Food Composition handbook developed by The Institute of Nutrition and Food Safety at the National Centers for Disease Control and Prevention [28] and in the Chinese Dietary Guidelines developed by the Chinese Nutrition Society, Ministry of Health, Ministry of Agriculture, General Administration of Sport, and National Centers for Disease Control and Prevention [29].Chinese people divided formal meals into two general parts: “fan” and “cai” [2]. “Fan” is the staple food for every Chinese meal, accounts for a main part of Chinese dietary energy intake, and comprises grains (such as rice, wheat, maize, millet, or sorghum), dry beans, starchy tubers, or root vegetables (such as potatoes, cassava, yams, or taro) [3,29]. It was reported that, by 2015, annual consumption of grains in China had risen to 534 kg per person [30]. Hence, grains, dry beans, and starchy tubers were listed into the C-NEMS-S as the first three food categories. “Cai” could be made of various kinds of vegetables, meat and poultry, and dietary oil. Eight subcategories of vegetables (root vegetables; fresh beans; solanaceous fruit vegetables; bulb vegetables; tender stems, leaves and cauliflower vegetables; aquatic vegetables; wild vegetables; bacteria and algae vegetables) were included into the C-NEMS-S, according to the China Food Composition handbook, which covers all kinds of vegetables eaten by the Chinese people [28]. Meat and poultry that could be cooked as dishes for meals in China included poultry and livestock (pork, beef, mutton, chicken, duck, etc.) and seafood (fish, shrimp, crab, shellfish: freshwater or saltwater). Another important ingredient in Chinese meals was dietary oil, of which Chinese residents consume an average of 42.1 kg per year per person; dietary oil was reported to be related to nutrition-related non-communicable diseases (NR-NCDs) among the poor [31].Breads and instant noodles are the foods most often consumed as instant meals, according to the China Food Composition handbook and were included in the C-NEMS-S.Milk is being consumed more frequently in China, with the rise of household income [32]. Milk was recommended by the Chinese Food Guide as a healthy food item. Milk’s measurements and scoring in the C-NEMS-S were kept the same as in the original NEMS-S.Six subcategories of fruits (pip fruits, stone fruits, soft fruits, citrus fruits, subtropical and tropical fruits, and melons) were included into the C-NEMS-S and cover all kinds of fruits consumed in China [28].Six subcategories of beverages (lactic acid drinks, milk beverages, juices, teas, plant protein beverages, apple vinegar) were included into the C-NEMS-S and cover all kinds of beverages consumed in China [28].An important principle of the NEMS-S was to measure availability, price, and quality of “healthier” options (such as low-fat milk for the milk measure and diet coke for the beverage measure) or healthy foods (such as fruits or vegetables). Healthier options and healthy foods are defined and presented in Table 1.Raters evaluated and scored each food measure based on its availability, price, and quality. Firstly, raters looked for healthier options for each food measure, as listed in Table 1. Then, they judged whether the healthier options were available and compared the price with regular options. Scores were given to each food measure based on availability of their healthier options and price, as listed in Table 2. Quality was only evaluated for fresh food, which included vegetables, fruits, meat and poultry, and seafood. Raters judged whether the quality of these fresh foods were acceptable or unacceptable. “Acceptable” (A) referred to peak condition, top quality, good color, and being fresh, firm, and clean. “Unacceptable” (UA) referred to being bruised, old looking, mushy, dry, or overripe, having dark sunken spots in irregular patches or cracked or broken surfaces, and signs of shriveling, mold or excessive softening. The rating was based on the majority (>50%) of vegetables, fruits, meat and poultry, or seafood. If it was difficult to decide whether to mark “A” or “UA”, “UA” was used, and a description was included in the comments.The adapted C-NEMS-S was tested in environment audits in 10 large-sized supermarkets (independent or chain, with areas more than 6000 m2) and 10 convenience stores (areas no more than 6000 m2) in February 2017. To assess inter-rater reliability, two raters entered the same store at the same time, with the store owners’ permission, and did the evaluations separately. The Research Ethics Committee of China Medical University approved the study ((2017) 055).Two undergraduate students from China Medical University were recruited as raters. They were both junior students from preventive medicine and have taken courses on nutrition. Recruitment posters were posted on campus, and seven students signed up. These students took a test in nutrition, statistics, epidemiology, and environmental sciences, and the top two students were selected. Training for raters included taking the NEMS online courses [33], discussing with researchers about each food category and item in the C-NEMS-S, and studying a detailed operation guide on the C-NEMS-S.Both the intra-class correlation coefficient (ICC) and kappa coefficient were used to determine inter-rater reliability. The ICC was calculated based on composite score and availability score for each food category and item. An ICC less than 0.4 indicated poor inter-rater reliability and and ICC greater than 0.75 indicated good inter-rater reliability [34]. Kappa was calculated using the availability of the food item (yes/no). Kappa less than 0.4 indicated poor inter-rater reliability, a range between 0.4 to 0.6 represented middle inter-rater reliability, a range between 0.6 to 0.8 represented good inter-rater reliability, and greater than 0.8 indicated excellent inter-rater reliability [35].The known-groups comparison method was used to determine construct validity [26]. If the scale was valid, the scores of the two groups with known differences would differ significantly. According to previous studies and the original NEMS-S, supermarkets and convenience stores were two groups with known differences: supermarkets were healthier than convenience stores [21,26,36,37]; healthier options and regular options were two groups with known differences with regard to price [21]. Therefore, we used scores acquired from the C-NEMS-S to test whether this instrument could also identify these differences. The Student’s t test was used to compare the total score between supermarkets and convenience stores. A nonparametric test was used to compare the availability between supermarkets and convenience stores and to compare prices between healthier options and regular options for each food measure. All data analyses were completed using SPSS Statistics version 20.0 (IBM Corporation, New York, NY, USA).Twelve measures were included in the C-NEMS-S (Supplementary Table S1: C-NEMS-S Questionnaire): grains, dry beans, starchy tubers, vegetables, fruits, seafood, meat and poultry, dietary oils, milk, bread, instant noodles, and beverages. Table 3 shows the food measurements and score dimensions.Table 4 shows the inter-rater reliability for the availability of each food measure. The intra-class correlation coefficient of availability had a composite score of 0.98. All food measures obtained a moderate or good ICC (ranging from 0.41 to 1.00). The kappa for each food measure ranged from 0.52 to 1.00. For price score and quality score, agreement between the two raters was very high, as shown in Supplement Table S2 and Supplement Table S3. Due to the low variance between different stores, ICC statistics were either too low or could not be calculated for price and quality.As shown in Table 5, supermarkets scored significantly higher in total than convenience stores (p < 0.001). Table 6 shows the proportion of the food stores that provided healthier options for each food measure. Large-sized supermarkets had a significantly greater amount of healthier options available for all food measures (all 12 items were statistically significant (p < 0.05), except for sugar-free beverages). The results were consistent between rater 1 and rater 2.Table 7 shows the price comparison between healthier options and regular options for each food measure. Healthier options cost more than regular options for grains, milk, bread, and instant noodles. The results were consistent between rater 1 and rater 2.The present study adapted and tested the first tool to measure the retail food environment in China. The original NEMS-S was adapted to the Chinese culture. The C-NEMS-S adapted in this study had high inter-rater reliability and was able to display the differences in healthy food availability between large-sized supermarkets and convenience stores, as well as the price differences between healthier options and regular options.The NEMS-S is regarded as an observational measure or environmental auditing tool. The auditing ability of raters is critical to obtaining high reliability observations. For the original NEMS-S, the inter-rater kappa coefficient ranged from 0.83 to 1.00 [26]. For the Brazilian version of the NEMS-S, the inter-rater kappa coefficient ranged from 0.69 to 1.00, and the ICC raged from 0.75 to 1.00 [21]. The inter-rater reliability of the C-NEMS-S in the present study was only slightly lower than that of the original and the Brazilian version. Nonetheless, both the ICC and kappa coefficient were acceptable, ranging from moderate to high (0.41 to 1.00 for the ICC, 0.52 to 1.00 for the kappa coefficient) [34,35].Compared with other food categories, milk, bread, instant noodles, and beverages had a lower ICC or kappa. This may be because packaged food categories occupied larger proportions of shelf space. In order to determine healthier options for these food categories, careful reading of food labels and packaging information is needed. Some raters may, therefore, fail to find healthier options for these food categories during this process. Though the ICC and kappa for these food categories were acceptable, future training programs, to teach raters how to measure in-store food environments, should pay more attention to these packaged food categories.According to the results, large-sized supermarkets had a significantly greater amount of healthier options available for all food measures (all items were statistically significant (p < 0.05), except for sugar-free beverages) compared with convenience stores. The result was consistent with previous studies on differences between supermarkets and convenience stores. According to previous studies, supermarkets could provide healthier foods at reasonable prices [36,37]. For the original NEMS-S in the United States and the adapted NEMS-S in Brazil [21,26], supermarkets obtained higher scores on healthier food availability. Hence, supermarkets were hypothesized to score high in the C-NEMS-S, as well. According to the results in the present study, the C-NEMS-S obtained higher availability for all food measures in large-sized supermarkets. These findings in the present study support the ability of the C-NEMS-S to differentiate between supermarkets and convenience stores.Except for meat and poultry, all healthier options cost more than regular options. This result indicates the ability of the measurement scoring to identify the differences between healthier options and regular options. In addition, this result also suggests that access to various food outlets might not be a problem for the food environment in China, but more attention should be paid to the consumer environment. The price discrepancy between healthier options and regular options may become a target for future policies aimed at improving the food environment in China.There were several limitations in the present study. Firstly, dietary habit is different between northern China and southern China. The current validation survey of the C-NEMS-S was only conducted in Shenyang, a city in northern China. In addition, the field work happened in winter, so the seasonality of the food offerings could have played a role in influencing reliability testing and findings. However, the C-NEMS-S was developed based on guidelines suitable for the whole of China by referencing the Chinese Dietary Guidelines (2016) [29] and Chinese food categories in the China Food Composition handbook [28]. Food categories and species were selected based on a nation-wide perspective and were not limited to northern Chinese cuisine or limited by seasonality. With further validation surveys in cities in southern China and during different times of the year, the C-NEMS-S may be verified to be suitable for the whole of China. Secondly, a small sample size was used to conduct the environment audit in stores and to test the reliability of the C-NEMS-S. To cover more areas of China, a larger sample size is warranted in future validation studies. Thirdly, snacks were not included in the final version of the C-NEMS-S as a food measure. During evaluations, it was too difficult to determine low fat or low sugar healthier snacks from the large variety of snacks. This is a problem to be amended in updated versions of the C-NEMS-S in our future studies.A Chinese version of the NEMS-S, which allowed for assessment of the food environment in China and potentially other Asian countries with similar food culture, was adapted and evaluated for China in the present study. The present study also implies that supermarkets could provide healthier food options compared to convenience stores in China. However, healthier options cost more than regular options, which may hinder the choice to purchase healthy foods and should be taken into consideration in future intervention studies or government food environment policies.The following are available online at https://www.mdpi.com/1660-4601/16/5/782/s1, Table S1: Table s1. The C-NEMS-S. Table S2: Price scores by two raters. Table S3: Quality scores by two raters.Conceptualization, Y.L., S.S. and D.W.; Data curation, J.H.; Formal analysis, Y.L. and S.S.; Investigation, S.S.; Methodology, Y.L. and Y.M.; Validation, J.G. and N.J.; Writing – original draft, Y.L. and S.S.; Writing – review & editing, J.G., N.J., J.H., Y.M. and D.W.This research was funded by National Natural Science Foundation of China, grant number 71774173.We thank Karen Glanz and Margaret Clawson for permission to use and adapt the original NEMS-S for China and for kindly providing information on the development of the original NEMS-S.The authors declare no conflict of interest.Definition of healthy foods and healthier options.Scoring Sheet for the Chinese version of the Nutritional Environment Measurement Survey (NEMS) for Stores.POINT RANGES: Availability Subtotal: 0 to 48; Price Subtotal: −14 to 26; Quality Subtotal: 0 to 9; TOTAL NEMS SCORE RANGE: from −14 to 83.Food measures and score dimensions in the Chinese version of the Nutritional Environment Measurement Survey for Stores (C-NEMS-S).1 Healthier food; no need for listing of healthier options. X Score dimensions.Inter-rater reliability of each food category based on availability.a Intra-class correlation coefficient (ICC) was calculated based on availability score. b Kappa was calculated based on the availability of the food item (yes/no).C-NEMS-S total score by store type.a M means the mean. b SD means the standard deviation.Availability (%) of healthier options by store type.Price comparison between healthier option and regular option for each food measure.a M means the mean. b SD means the standard deviation. c The unit is RMB/0.5 kg d. The unit is RMB/250 mL. e The unit is RMB/ loaf. f The unit is RMB/ packet.

Med-MDPI/ijerph_3/ijerph-16-05-00783.txt

@@ -0,0 +1,5 @@
+Drawing on a new typology of intimate partner violence (IPV), this paper tests the relationship between indicators of totalitarian and anarchic IPV and child polyvictimization incidence and severity. The paper argues for and utilizes a quantitative approach to study polyvictimization severity. Polyvictimization is operationalized as a multiplicative relationship between physical abuse and neglect in a random sample of 204 children from Kyunggi province, South Korea. The indicator of totalitarian IPV significantly predicted polyvictimization severity and incidence even when a traditional measure of intimate terrorism was held constant. The indicator of anarchic IPV significantly predicted polyvictimization severity but not incidence when a traditional measure of intimate terrorism was held constant. Implications are discussed.Research on harm stemming from children’s exposure to multiple forms of victimization, conceptualized as polyvictimization, is relatively new in the research literature [1,2]. Not only does polyvictimization appear to be more harmful to children than single victimization, but low-frequency victimization across multiple domains (e.g., bullying at school and abuse at home) appears to produce more distress than high-frequency victimization in a single domain [1]. Child polyvictimization is associated with more psychopathology, depression, delinquency, low self-esteem, self-blame, suicidal phenomena, health risk behaviors, and anxiety among children, and higher unemployment, substance abuse and mental illness among parents [1,3,4,5,6,7]. Finkelhor et al. argued that there are four pathways to child polyvictimization: a) dangerous families, b) problem-beset families, c) dangerous communities, and d) children with pre-existing emotional problems that increase risky behaviour [1]. Chan extended the concept to incorporate co-occurrence of multiple forms of violence at the family level, labelling this family polyvictimization [2]. Family polyvictimization was indicated by child victimization, intimate partner violence (IPV), and elder abuse, and was associated with parental addiction and poor health in a Chinese household sample [2]. Despite Chan’s initiative, polyvictimization remains understudied in an East Asian context. Moreover, research on the dangerous families pathway [1] generally focuses on individual characteristics of family members rather than family systemic characteristics, little work has been done to link the concept of polyvictimization with its measurement, and the problem of how to incorporate high frequency single-victimization with polyvictimization has not been resolved. Specifically, despite a large literature on the co-occurrence of IPV and child maltreatment [8], little literature has examined the effects of types of IPV on child polyvictimization. Similarly, no existing measure integrates polyvictimization incidence and frequency into a common scale of polyvictimization severity. This paper examines the relationship between polyvictimization and a new typology of violent family organization in a probability proportional to size random cluster sample of rural South Korean children. We hypothesize anarchic and totalitarian family types are at higher risk for polyvictimization. The paper argues for and implements a measure of polyvictimization that allows single-victimization frequency and polyvictimization to be combined in the same scale that captures the severity of polyvictimization. We argue this measure is closer to the conceptual rationale for the study of polyvictimization.Although Chan [2] has stressed the importance of considering emergent qualities of the family as a unit in considering polyvictimization, and a 30-year research tradition documents conceptual and empirical links between intimate partner violence and child maltreatment [8,9,10,11], only a handful of studies searchable via google scholar explicitly examine intimate partner violence and child polyvictimization. Chan’s research examined exposure to parental IPV as one component of polyvictimization and found that polyvictimized children had lower self-esteem, higher rates of aggression, PTSD, addiction, and lower quality of life [2,5,12]. Pereda and Gallardo-Pujol similarly included IPV exposure as polyvictimization and found that it predicted re-victimization in adulthood [13]. Using a similar concept of polyvictimization, other research found that the relationship between child polyvictimization and PTSD were mediated by child attributions [14] and that polyvictimization was correlated with later disordered eating [15]. Although the foregoing literature on child polyvictimization has considered IPV exposure as a form of polyvictimization [5], it has insufficiently examined parental IPV in an etiological role with respect to polyvictimization via the dangerous families pathway [1], neither has it conceptualized IPV as having a profound effect on emergent family processes. Emery [16] argues that the introduction of physical violence into an intimate relationship shifts the foundation on which power rests. According to this argument, power dynamics between intimate partners in never-violent relationships rest on walk-away costs. However, when an act of physical violence occurs between partners the base of power may shift from walk-away costs to force. In that case power may rapidly shift to the partner who can command the most physical force [16], permanently changing family power dynamics. In Emery’s typology, IPV exists on dual continua of order and power, ranging from anarchic type (low order, chaotic, no consistent rules or legitimate power) at one extreme and totalitarian type (highly ordered, asymmetric power) at the other [16]. Totalitarian families are characterized by power asymmetry and elaborate systems of rules extending control to “mundane areas of everyday life …not normally thought of as norm- or rule-governed” [17]. Emery argues that the anarchic type is chaotic, unpredictable, and fits more with the social disorganization conceptualization of IPV common in criminology, while the totalitarian type reflects the traditional feminist conceptualization of IPV. IPV at both extremes, however, is likely to create a more dangerous family environment for children [16]. IPV at the extremes of these continua can put children at higher risk not only because the IPV may be more frequent and qualitatively severe [16], but also because maltreatment may be more likely to be legitimized as punishment for rule-breaking in the totalitarian type while family rules (norms) aimed at protecting children may be absent in the anarchic type. This suggests families with any history of IPV and characterized by anarchy or totalitarian style control may put children at particular risk for polyvictimization.Although some research [18,19] has begun to use the concept of the totalitarian type, no research has used a measurement specifically aimed at capturing anarchic or totalitarian type families, and no research has examined these types as risk factors for polyvictimization. Indeed, research on types of IPV in an etiological role for child polyvictimization broadly is lacking. Although arguably the most common typology of IPV is Johnson’s [20] intimate terrorism typology, a google scholar search of “intimate terrorism” and “polyvictimization” returns only 47 results, none of which feature intimate terrorism in a causal role for child polyvictimization. The intimate terrorism typology is defined on the basis of control motive [20], with the result that whether control attempts succeed or not (achieved control) is ignored. The intimate terrorism typology has been critiqued on this basis as ignoring power, for which reason this paper focuses on the anarchic/totalitarian IPV typology in etiological relation to child polyvictimization.The burgeoning literature on polyvictimization [1,2,3,4,5,6,7,8,9,10,11,12,13,14,15] rests on a largely implicit rationale: polyvictimization must be distinct in both etiology and impact from high frequency victimization of a single type. If this were not the case, multiple experiences of victimization of different forms could lumped together additively in a single measure without reference to type. For example, experience of acts of physical abuse, acts of neglect, and witnessing acts of IPV could simply be summed. The rationale for polyvictimization as a unique subject is empirically supported by findings that even low-frequency polyvictimization appears to have a more severe impact than high frequency single-victimization [1]. The problem for the polyvictimization field to date [1,2,3,4,5,6,7,8,9,10,11,12,13,14,15] is that researchers are forced to choose between examining the impact of frequency for single-victimization or sums of dichotomous indicators for various types of single-victimization that add up to polyvictimization. Continuous measures generally have more statistical sensitivity than dichotomous measures, and Emery et al. [21,22] have long argued for and implemented continuous measures of violent victimization weighted by the log-odds of injury. Although the continuous measure creates a right-skewed distribution, this problem can now easily be handled via monotonic transformation, robust standard errors, or both [23]. Moreover, a continuous measure can capture the severity, rather than simply the fact, of polyvictimization.We argue that the empirical logic of the rationale for the study of polyvictimization is one of interaction effects. That low frequency victimization in the context of a second form of victimization has a larger impact on child well-being than high frequency single-victimization [1] suggests that the effect of one form of victimization on well-being differs depending on whether a second form is present or absent. Interaction effects capture just such an effect, as when the effect of witnessing IPV on externalizing behavior problems depends on the child’s age [24]. The standard approach to handle this problem in linear models is to model multiplicative relationships between variables, (e.g., age X witnessing IPV) rather than only additive relationships [23,24]. In this paper, we present a continuous measure of polyvictimization which is additive when a single form of victimization is present but multiplicative when multiple forms are present. This extends the scale to higher levels when more than one kind of victimization is present. For example, if our measure of physical abuse severity is a 5 and only physical abuse is present, then the continuous polyvictimization scale will also be a 5. However, if the child has a physical abuse score of 5 and also a neglect score of 2, then the polyvictimization score will be 2 × 5 = 10. Our models compare the continuous severity measure with a more standard categorical measure of polyvictimization.Based on the IPV literature, two hypotheses were made: 
+            Polyvictimization and polyvictimization severity will be positively associated with totalitarian type IPV.
+          
+            Polyvictimization and polyvictimization severity will be positively associated with anarchic type IPV.
+          The aim of the current study was to extend the literature on polyvictimization by examining whether emergent family types (anarchic and totalitarian) from the IPV literature [16] are related to polyvictimization and comparing a new severity measure of polyvictimization with the standard approach. The current study also contributes to the IPV literature. The full models control for a measure typically used to capture a type of IPV known as intimate terrorism [20]. Intimate terrorism is commonly studied in the IPV literature, and is defined as violence in the context of systematic attempts to control the victim [20]. The intimate terrorism concept has been critiqued as inadvertently ignoring power [16]. If the anarchic and totalitarian family types predict polyvictimization when intimate terrorism is held constant this would suggest some support for the added value of the disorder to deviant order continuum theory (anarchic/totalitarian typology) [16]. Models also control for age and sex of the focal child, sex of the parent, and household income. Research participants were drawn from a probability proportional to size (PPS) cluster sample of parents of public-school children in Kyunggi province, South Korea. Eight rural public schools were selected using PPS sampling, after which grades were randomly selected via PPS sampling. Once a class within a school was randomly selected, the study team used connections with school officials, teachers, and parents to reach children’s parents, with the goal of obtaining all of the parents of children in the grade selected. Participation was incentivized with gift cards, and a refusal conversion protocol was used to keep the response rate high (81.5%). Refusal conversion attempts to convert a ‘soft no’ into an agreement to participate by using a different interviewer, a different time of approach, and a different incentive. When parents had more than one minor child, they were asked to select the child with the most recent birthday as the focal child. The total sample consisted of 107 girls and 128 boys with a mean age of 11.7 years old (SD = 4.2, n = 235). However, 13 parents were omitted from analyses for skipping or failing to answer correctly a quality control question (“If you read and understand this question circle the 5”). Interviewers underwent two days of rigorous training on written informed consent, sensitivity and study ethics, passing both written and oral exams prior to certification. The study protocol was approved by the IRB of Yonsei University.Two scales of polyvictimization were created, both based on underlying data from the Parent-Child Conflict Tactics Scale (PC-CTS) [25]. Responses to all items were once in the last year, twice in the last year, 3–5 times, 6–10 times, 11–20 times, more than 20 times, not in the last year, or never. Physical abuse and neglect of the child were measured by 8 violence items, 4 injury items, and 3 neglect items: (1) hit child on the bottom with a belt, hairbrush, or stick, (2) threw something at him/her, (3) hit some other part of the body with a belt, hairbrush, or stick, (4) pushed, grabbed, or shoved him or her, (5) kicked, bit, or hit him/her with fist, (6) choked or knocked him/her down, (7) beat up him/her, (8) intentionally burned or scalded him/her, (9) kid had a bruise from a conflict with parent, (10) kid had a cut from a fight with parent, (11) kid needed to see a doctor (MD) because of a fight with parent, (12) kid had to miss school because of a fight with parent, (13) had to leave child alone when she/he needed looking after, (14) could not give him/her the food she needed, and (15) child ran away and I did not know where he/she was (α = 0.86). Last year injuries were totaled and regressed on physical violence items, creating regression coefficients that indicated the average increase in injuries for each act of a particular type of violence. These coefficients were used to weight total acts of each type of violence and sum the weighted acts into a scale of physical abuse severity. The benefit of this scale is that an act of a particular type of physical violence is weighted by its associated injuries, avoiding false equivalence between low injury violence like hitting on the bottom with an object and high injury violence like intentional burning. Total last year acts of neglect were summed into a single scale. Prior to combining the two scales were standardized to have the same variance.Broadly speaking, the continuous polyvictimization measure can be understood as capturing the severity of polyvictimization. The continuous polyvictimization scale was created by setting the scale to zero when there was no abuse or neglect, equal to the physical abuse severity scale when there was no neglect, equal to the number of acts of neglect when there was no physical abuse, and neglect times physical abuse when both were present. A log transform was used to reduce right-skew. The conventional polyvictimization scale was 0 for no victimization, 1 for single victimization, and 2 for both kinds of victimization.The totalitarian and anarchic types are principally defined by order (a highly norm- and rule-governed family routine) and power (between intimate partners). These concepts were captured using previously published measures of family order [26] and an updated version of decision power [27]. IPV was measured using the Conflict Tactics Scale 2 (CTS2) [28]. Family order was captured by 9 Likert scale items (strongly agree–strongly disagree): In my family (1) we eat dinner together, (2) we eat dinner at the same time every night, (3) our house/apartment is neat and orderly, (4) our house/apartment is clean, (5) I know the daily schedules of everyone in our home, (6) I know what household chores it is my job to do, (7) I know what household chores other people in my family are supposed to do, (8) when one person in the family is sick or really busy other family members step in to do that person’s chores, (9) our family has many rules (α = 0.79). Power between intimate partners was measured using answers to who had more say in decisions about (1) buying a car, (2) buying a computer, (3) children’s education, (4) buying a house/apartment, and (5) where to live. Choices ranged on a scale from 1–5 (spouse only to self only) with 3 indicating self and spouse equally (α = 0.70). Equal power was coded ranging from 0% to 100% depending on how many items for which respondents indicated a ‘3’. Last year IPV items were answered for self and spouse, and consisted of (1) slapped, (2) hit partner with something, (3) had to see a doctor because of a fight with partner, (4) punched, kicked or bit, (5) used or threatened to use a knife or gun, (6) had a sprain, bruise, small cut, or felt pain the next day because of a fight with partner, (7) pushed, grabbed or shoved, (8) beat up, (9) choked (α = 0.97). Possible responses are the same as for the PC-CTS. Any IPV was indicated if any of these items had ever occurred by male or female partner. Relatively few families were characterized by low order, so low order was defined as below the median. This helps to meet the minimum theoretical characteristics of the anarchic type while ensuring a reasonable number of families are included in the category. Families low in order with ever any IPV were classified as anarchic type. High family order was defined as above the 75th percentile. Totalitarian type was classified if there was ever any IPV, family order was high, and equality in power was less than or equal to 83%. Similar to the case for anarchic type, this cut-off meets the minimum characteristics for the distinctly totalitarian type while ensuring a reasonable number of families meet the criteria. In theory, intimate terrorism is characterized by violence in the context of a high level of attempted control [20]. The original research on intimate terrorism operationalized the concept using multiple scales, one of which was control attempts: thinking about your husband would you say he (1) is jealous or possessive, (2) tries to limit your contact with family and friends, (3) insists on knowing who you are with at all times, (4) calls you names or puts you down in front of others, (5) makes you feel inadequate, (6) shouts or swears at you, (7) prevents you from knowing about or having access to the family income, even when you ask [20,29]. For female participants the questions were left as above and asked on a 5-item Likert scale (1. never – 5. always) (α = 0.87). For male participants, “my partner complains that I…” was added in front of each item so that the scale consistently captures attempts to control the female partner. For the male respondents, a 6th option on the Likert scale “my partner does not complain but I actually do this”. This was coded as the highest value on the scale (6) (α = 0.84). In the original research the items were used to create a dichotomous measure of intimate terrorism [20]. However, we keep a continuous (summed) measure because this allows the measure to have more explanatory power, which raises the empirical bar that the anarchic and totalitarian types must meet for significance. For the same reason, rather than combining control attempts with IPV, any IPV is introduced and controlled separately. The full model controlling for both any IPV and continuous control attempts can be said to control for intimate terrorism. The models of the continuous measure of polyvictimization were implemented using mixed effects regression models with a random effect for each cluster and robust standard errors. A Pearson’s goodness of fit test was not significant, (χ2 = 147.4, p = 0.997), which suggests a Poisson regression (rather than negative binomial) model can be used for categorical polyvictimization. Likewise, a Vuong test (p = 0.50) suggests no difference between zero-inflated and an ordinary Poisson model. Hence, Poisson models with robust standard errors were implemented to test the hypotheses using the conventional measure of polyvictimization. Item non-response for physical abuse and neglect narrowed the analysable sample to 215 for the Poisson models. Fewer cases (204) were available for the continuous measure of polyvictimization because a missing response for any abuse or neglect measure resulted in a missing value. For the conventional measure of polyvictimization, item non-response only resulted in a missing value of all other responses to abuse (or neglect) were ‘never’. Both models are presented for the 204 cases available for the continuous model, but the findings for the Poisson models using 215 cases were not noticeably different in sign, significance, or magnitude. Ordinal logits were used to estimate odds-ratios for Figure 2. A priori power analyses were difficult to estimate as this constitutes a first test of these two hypotheses. Post-hoc power analyses suggest the study was sufficiently powered for hypothesis 1 (λ = 0.96, power = 99%) but power to test hypothesis 2 was substantially weaker (λ = 0.45, power = 59%).Figure 1 shows 15.6% of children (34) in the sample suffered both physical abuse and neglect (polyvictimization) in the last year. 37.0% suffered from either abuse or neglect, but not both. The mean for the continuous polyvictimization measure was -.84 (SD = 1.49). The prevalence of any last-year physical abuse was 38.9% and any last-year neglect was 29.4%. The last year prevalence of very severe abuse (beat up, choked, intentionally burned) was 4.1%; abuse injury was 3.6%. The rate of very severe abuse was not significantly different from the 5.8% rate previous research found in Seoul (Z = 0.89, p = 0.37) but the rate of injuries was significantly lower than Seoul’s 8.5% rate (Z = 2.3, p = 0.02) [22]. With respect to children in the sample, 11% had experienced insufficient food in the past year; 24.3% had been left alone when parents thought they should not.Of children in the sample, 12.3% had been exposed to some form of IPV, 3.5% (8) lived in households characterized as totalitarian type, and 4.0% (9) lived in households characterized as anarchic type. The mean score for attempts to control the female partner was 13.4 (SD = 5.6). Previous research has characterized two standard deviations above the mean as ‘high control’ [18]; 4.7% of the sample met this threshold. Although this is not significantly different from a previous study of Korea (5.2%, Z = 0.2, p = 0.79) [18], power to detect a difference from the previous study is low for this comparison. There were more female than male caregivers (63.2%) in the sample. The average child’s age was 11.8 (SD = 4.2). There were slightly more male than female (54% vs 46%) children in the sample, and the median household income was 3.5 million won per month (3130 USD).Figure 2 shows increases in odds of polyvictimization associated with anarchic and totalitarian types (left side) from baseline ordinal logit models and increases in severity of polyvictimization (right side) from baseline mixed effects regression models. The baseline models control for child’s age, sex, participant sex, and household income. Figure 2 shows that when totalitarian type IPV is present odds of polyvictimization are 9 times higher, and when anarchic type is present odds of polyvictimization are 3.3 times higher. Arrows indicate 95% confidence intervals around odds-ratios. The right side of Figure 2 shows that totalitarian type is significantly associated with a 1.85 (1.2 standard deviation) increase on the polyvictimization severity scale, and anarchic type is significantly associated with a 1.08 (0.7 standard deviation) increase on the severity scale.Table 1 shows the results of Poisson regression for conventional polyvictimization (left side) and mixed effects regression models for polyvictimization severity (right side). The table allows readers to compare the baseline models with models controlling for attempts to control the female partner and any IPV. The final model controls for attempts to control the female partner and any IPV, and hence controls for intimate terrorism. In the final model, totalitarian type IPV is significantly associated with a 0.77 increase on the polyvictimization categorical scale (p < 0.001). Likewise, totalitarian type IPV is associated with a 1.8 unit increase in polyvictimization severity (p < 0.001). In the same model, anarchic type IPV is no longer significantly associated with polyvictimization categorically, but remains significantly associated with a 0.78 unit increase in polyvictimization severity (p < 0.01). This is the first study to operationalize the theoretical concepts of anarchic and totalitarian IPV and examine their relationship to child polyvictimization. As hypothesized, both totalitarian and anarchic types of IPV were significantly associated with the incidence and severity of child polyvictimization in models controlling for child’s age, sex, parent’s sex, and household income. The significant findings suggest that polyvictimization is both more likely and more severe when IPV occurs in the context of a highly ordered family with an extensive (and perhaps intrusive) set of rules for governing behavior and in which the couple is unequal in power. The findings also suggest that child polyvictimization is more likely and more severe when IPV occurs in chaotic families characterized by low order.The study simultaneously adds to our understanding of IPV. The relationship between child polyvictimization and totalitarian type IPV persisted even when intimate terrorism was controlled. Emery [16] argued that the classification of IPV by control attempts alone [20] was a theoretical liability. Rather, he argued that IPV must be classified based on underlying continua of order, power, and legitimacy. Our findings suggest that further investigation of this argument may bear empirical fruit. More research is needed to establish whether Emery’s [16] anarchic/totalitarian typology of IPV is consistently related to polyvictimization and other child well-being outcomes. This research should continue to compare that typology [16] with the effects of Johnson’s [20] intimate terrorism typology of IPV. There are many similarities between these two typologies. However, if the former continues to show stronger empirical relationships with child well-being than the latter, researchers on child well-being may eventually wish to adopt the typology with more explanatory power. Furthermore, if anarchic and totalitarian types place children at greater risk, policy makers may be able to efficiently mitigate risk by targeting interventions specifically at families with these types of IPV.This study is the first to use a measure of child polyvictimization severity as well as incidence. The severity scale was created by accounting for the effects of individual acts of physical abuse on injuries, the frequency of neglect, and the assumption of interaction effects between different types of victimization. The polyvictimization severity scale highlights the theoretical salience of polyvictimization and prior research: victimization across different domains may be worse for children than more victimization in a single domain [1]. As hypothesized, findings for the polyvictimization severity scale were similar to findings for the conventional polyvictimization incidence measure. As expected, there is evidence to believe the severity scale to be more statistically sensitive (anarchic type IPV remained a significant predictor of polyvictimization severity but not polyvictimization incidence when intimate terrorism was controlled). These findings suggest polyvictimization severity is worth further study.Further research is needed to develop the measure of polyvictimization severity. More types of child victimization could be included in such a measure in future. Moreover, conceptual and empirical work is needed to establish logical impact measures on which to base polyvictimization severity. The contribution of the perpetration of a particular act of violence to the increase in log odds of victim injury is a relatively logical criterion for severity of physical violence. Although this paper examined IPV at any time, severity of exposure to IPV could also be measured in terms of how much each act of IPV increased the likelihood of child injury (accidental or deliberate). However, reasonable indicators of neglect severity are more complex. Malnutrition might be a suitable outcome as it would catch problems stemming from both insufficient food and child consumption of unhealthy food due to lack of monitoring. However, determining the answer to this problem lies beyond the scope of this paper.The study found that more than 1 in 6 children in this rural sample of South Korean children had experienced polyvictimization in the previous year. More than half of the sample (52.6%) had experienced either physical abuse or neglect in the past year. This rate of child victimization is a reliable prevalence estimate based on a random sample of families with children in rural Kyunggi province. It is significantly higher (p < 0.01) than the 42.8% child victimization rate found in a representative sample of 7466 Chinese households [2] found in previous research. This finding suggests that child protection in rural South Korea is at least as serious an issue as in China, if not more so.Although the study is a representative random sample, the sample is small, and findings may not be generalizable beyond rural South Korean children. The measures of anarchic and totalitarian types of IPV are preliminary and the cut-off points should be replicated and examined in a larger study. Because the study is cross-sectional and non-experimental, statistical inferences are indicators of associations that are not necessarily causal. The findings are preliminary and require replication, and the child polyvictimization severity scale requires further conceptual and empirical development. Although anarchic type IPV was consistently associated with polyvictimization severity, it was not significantly associated with the categorical polyvictimization when intimate terrorism was controlled. This null finding may have occurred because there is no true relationship or because of the relatively small sample size. Only one parent was interviewed per family, and many types of polyvictimization (exposure to bullying and street crime) were not included.As the first study of the relationship between totalitarian and anarchic types of IPV and child polyvictimization the findings have important implications for understanding risk factors for polyvictimization. Important progress has been made in reducing IPV related homicide by using Domestic Violence High Risk Teams (DVHRTs) to assess for high risk cases and follow up with additional resources and increased monitoring [30]. These targeted intervention programs for IPV could be extended to monitor the risk for child polyvictimization. Future research is needed to replicate and extend these findings in diverse and larger samples. Polyvictimization poses a serious and unique threat to children. Should future research confirm our preliminary findings, policy makers should provide funds to pilot the extension of DVHRTs to monitor for child polyvictimization. Research is also needed to better understand the effects of totalitarian and anarchic types of IPV on victims, and particularly how these effects differ from or converge with the effects of intimate terrorism. The initial results suggest the anarchic/totalitarian IPV typology may be at least as effective (if not more so) as the intimate terrorism typology for understanding the relationship between high risk IPV and child polyvictimization. The mental, physical, and emotional costs of child polyvictimization are high [1,2,5,12]. We owe it to our children to continue to try to do better.Conceptualization, C.R.E.; methodology, C.R.E.; formal analysis, C.R.E.; investigation, H.Y., O.K., Y.K.; Writing—Original Draft preparation, C.R.E.; Writing—Review and Editing, C.R.E., H.Y., O.K., Y.K.; project administration, H.Y.; funding acquisition, C.R.E., H.Y., O.K., Y.K.“This research was funded by the Korea Research Foundation, grant number 2015” and “The APC was funded privately”. We would like to thank all of the hard work of the interviewers, facilitators, and research assistants in carrying out the study.The authors declare no conflict of interest. Prevalence of polyvictimization.IPV type and polyvictimization.Polyvictimization incidence and severity (n = 204).Notes: †p < 0.10, * p < 0.05, ** p < 0.01, *** p < 0.001.

Med-MDPI/ijerph_3/ijerph-16-05-00784.txt

@@ -0,0 +1,5 @@
+Increasingly stricter and wider official efforts have been made by multilevel Chinese governments for seeking the improvements of the environment and public health status. However, the contributions of these efforts to environmental changes and spatiotemporal variations in some environmental diseases have been seldom explored and evaluated. Gastric cancer mortality (GCM) data in two periods (I: 2004–2006 and II: 2012–2015) was collected for the analysis of its spatiotemporal variations on the grid scale across S County in Central China. Some environmental and socioeconomic factors, including river, farmlands, topographic condition, population density, and gross domestic products (GDP) were obtained for the exploration of their changes and their relationships with GCM’s spatiotemporal variations through a powerful tool (GeoDetector, GD). During 2004–2015, S County achieved environmental improvement and socioeconomic development, as well as a clear decline of the age-standardized mortality rate of gastric cancer from 35.66/105 to 23.44/105. Moreover, the GCM spatial patterns changed on the grid scale, which was spatially associated with the selected influencing factors. Due to the improvement of rivers’ water quality, the distance from rivers posed relatively larger but reversed impacts on the gridded GCM. In addition, higher population density and higher economic level (GDP) acted as important protective factors, whereas the percentage of farmlands tended to have adverse effects on the gridded GCM in period II. It can be concluded that the decline of GCM in S County was spatiotemporally associated with increasingly strengthened environmental managements and socioeconomic developments over the past decade. Additionally, we suggest that more attentions should be paid to the potential pollution caused by excessive pesticides and fertilizers on the farmlands in S County. This study provided a useful clue for local authorities adopting more targeted measures to improve environment and public health in the regions similar to S County.Rapid and huge economic growth has occurred in China during the four decades of reform and opening up. However, China is also facing a huge burden of disease during the process of economic and social development. More than 2.8 million people died of cancer in 2015, and gastric cancer is the second leading cause of cancer death in China, causing the death of nearly 500,000 people in 2015 [1], and it has been a serious threat to sustainable development due to its huge health and finical burden on the country and individuals in China [2].As a complex chronic non-communicable disease, gastric cancer is closely associated with many risk factors on the global, national, provincial, municipal, and county scale [3,4,5,6,7,8,9,10,11]. Personal lifestyles, such as high intake of salt, excessive alcohol consumption, and insufficient fruits and vegetables intake were strongly associated with gastric cancer [3,4,5,6]. Long-term exposure to water pollution caused by wastewater, sewage, and excessive application of fertilizers and pesticides from industrial and agricultural production also posed important impacts on the epidemiological status of gastric cancer [7,8,9]. In addition, positive family history of this disease and Helicobacter pylori (H. pylori) infection both increased the risk of developing gastric cancer [10,11].The influences of these risk factors on gastric cancer were deeply explored through various methods, such as meta-analysis [3], spatial correlation analysis [7,8], Delphi approach [11], and so on. Among these methods, GeoDetector (GD) is a powerful tool to analyze the effect of several independent variables on the spatial distribution of the dependent variable based on spatial variance analysis [12], since spatial heterogeneity is a major issue in geographical phenomena [13]. It was initially applied to analyze the effects of driving determinants on local disease such as neural tube defects [14]. Due to its good applicability, GD not only has an extensive application on public health issues [15,16,17], but is also an exploration in the fields of land use, ecology, regional economy, and so on [12].In recent years, public awareness about the increasingly serious environmental pollution has been greatly raised due to its important impacts on public health [18,19]. Although China has completed rapid development and huge socioeconomic achievements, the sustainable development mode, in particular the improvement of public health status, has been set as a basic requirement and national strategy [20]. As a result, research has been gradually focusing in on the evaluation of the influences of environmental factors, as well as their changes, on the incidences, and/or mortality of some chronic non-communicable diseases, since the environmental protection measures and hygienic interventions have been strengthened [21,22,23].In the Central China, S County is well-known for its environmental deterioration and a distinct increase of gastric cancer mortality (GCM) over the past decades [24]. Fortunately, the continuous development of the social economy, increasingly strict environmental managements, and strengthened comprehensive cancer interventions had been conducted in this county by the central, provincial, and local government since 2007 [25,26,27]. However, the evaluation on the contribution of these measures to the status of GCM was not conducted in time.Therefore, our study aimed to (1) investigate the spatial and temporal patterns of GCM in S County during 2004–2015, and to (2) explore the associations between GCM and environmental conditions and socioeconomic development, as well as their changes through the GD method. This study would provide some useful clues for local authorities realizing their efforts made for enhancing environment and improving public health.S County (‘S’ derives from the initial of the county name, we have to conceal the full name of some geographical vocabularies related to S County for some reasons), located in Central China, is a ‘famous’ region with both serious environmental pollution and high GCM [7,28]. S County includes 21 towns with a total area of 1080 km2 (Figure 1). S County is generally flat (altitude 35–50 m), and it is gently inclined and slightly descends from northwest to southeast. The region is dominated by warm temperatures and a continental monsoon climate with an annual average temperature of 14.5 °C and an annual rainfall of about 700 mm/m². There are two rivers, the SY and FQ rivers, flowing through this county from northwest to southeast.The death cases of gastric cancer (10th revision of the International Classification of Diseases code: C16) in 2004–2006 (Period I) and 2012–2015 (Period II) were derived from retrospective cause of death survey data and death surveillance data conducted by Chinese Center for Disease Control and Prevention. On the basis of village population in 2004 and the registered population growth of this county during 2004 to 2012 [29], the village population in 2012 was obtained. The age-standardized mortality rate was adjusted by Chinese standard population in 2000. To avoid the instability of GCM caused by irregular and non-uniformity of village administrative units [30,31], we used a 2 × 2 km2 grid as a basic statistical unit, which has been proven to be an appropriate analysis scale for S County [24]. In total, there were 260 grids across this county. For all geographical units, GCM data ware spatially smoothed using an empirical Bayesian smoothing method in GeoDa (GeoDa 1.12.1, April 2018, the Center for Spatial Data Science (The University of Chicago), Chicago, IL, USA) before conducting spatial and statistical analysis [32].As can be seen in Table 1 and Figure 2, the data of some proxy variables were collected for indicating these influencing factors during these two periods in S County, including human-made pollution, physical environment, and socioeconomic level [14,15]. Given that surface water pollution is an important environmental risk factor [7], the Euclidean distance between the grid center from the nearer river (SY or FQ river) was calculated for indicating the contact opportunity between residents and rivers. As S County is featured by agricultural production, the percentage of farmlands was derived from the land use data for indicating the potential pollution caused by excessive application of pesticides and fertilizers in this area [33]. As for the physical environment, topography plays an important role in the migration of carcinogens [34], and GCM may vary significantly at different altitudes [24]. Meanwhile, resident population density and gross domestic product (GDP) were used to measure the local socioeconomic level, which may affect family income, household sanitation, eating habits, and medical service level [35,36]. The gridded resident population density and GDP data were disaggregated from the county level based on multivariate statistical models [37,38].Spatial autocorrelation analysis is commonly used to explore the spatial patterns of disease incidence or mortality. In this study, we used Moran’s Indices to measure the degree of spatial autocorrelation [40], which is calculated as follows:(1)I=n∑i=in∑j=1nωij(xi−x¯)(xj−x¯)∑i=in∑j=1nωij∑i=1n(xi−x¯)2
+          where n is the number of grids; xi and xj stand for the GCM of grid i and grid j; x¯ is the average value of GCM; and ωij is the spatial weight between grid i and grid j that can be defined by the contiguity of grids. Moran’s I value falls between −1 and 1. A high positive Moran’s I indicated a spatial clustering, that is, adjacent grids had similar levels of GCM, whereas a low negative value implied a tendency toward dispersal. When Moran’s I is around zero, the value indicated spatial randomness.On the basis of the existence of spatial autocorrelation, hot spot analysis further captured the detailed spatial patterns by means of locating hot or cold clusters of GCM. We selected Getis-Ord Gi* to identify the locations of statistically significant hot and cold spots according to formula (2):(2)Gi*=∑j=1nωijxj−X¯∑j=1nωijS[n∑j=1nωij2−(∑j=1nωij)2]n−1
+          where xj is the GCM value for grid j; ωij stands for the spatial weight between grid i and grid j; n means the number of grids; and:(3)X¯=∑j=1nxjn
+          (4)S=∑j=1nxj2n−(X¯)2The Gi* statistic is a Z-score, and therefore, no further calculation is needed. For statistically significant positive Z-score, the larger the Z-score is, the more intense the clustering of high GCM (i.e., a hot spot). For statistically significant negative Z-score, the smaller the Z-score is, the more intense the clustering of low GCM (i.e., a cold spot). By conducting a hot spot analysis on GCM during the two specified periods, we were able to visually understand the spatiotemporal changes in GCM. We performed spatial autocorrelation and hot spot analysis on the ArcGIS 10.2 platform (ESRI, Redland, CA, USA).In this study, GD was used to explore the determinants of spatial variations of GCM in S County during two periods. The basic idea of the method was to compare the spatial consistency of the continuous response variable versus the categorical explanatory variables, and on this basis, to quantify the interpretation of the explanatory variables in relation to the response variable. Both categorical and discretized continuous explanatory variable can be analyzed by GD, while the classic regression model has limitations when it comes to dealing with categorical data [14]. Another advantage is that GD could be used to study the interaction impact of multi factors.In this study, the response variable is the gridded GCM, and the explanatory variables are the proxy variables as listed in Table 1. The correlation strength between factor X and attribute Y is measured by q-statistic, which is expressed by the following equation:(5)q=1−1Nσ2∑h=1LNhσh2
+          where h = 1, 2 … indicates that attribute Y is stratified by factor X (categorical variable or discretized continuous variable); N and Nh mean the number of total grids and grids in stratum h, respectively; σ2 and σh2 stand for the variance of attribute Y in all grids and grids in stratum h, respectively. The value of q ranges from 0 to 1. The higher the q value, the stronger the explanatory power of factor X to attribute Y. Moreover, the interaction detector could reveal the interactive influence of multi factors. A detailed introduction and computing tool are available at http://www.geodetector.org/.According to the surface water quality data derived from the state-control stations set for FQ and SY Rivers (Table S1 in Supplementary Files), the water environment of these two rivers has been obviously improved in the past years. As a typical region featured by agricultural production, S County has about 90500 and 85800 ha farmlands in 1995 and 2005, respectively, accounting for more than 70% of the total lands. From 1995 to 2005, about 5.2% farmlands were transformed into other land-use types (Figure 2c,d), which were mainly distributed in the region along the SY river and between the two rivers. The area of construction land increased by over 20 km2 in the region between the two rivers. There was a decline in the number of the resident population because of labor export in last decade [41], due to increasing urbanization of population as more and more rural people concentrated living in towns with better economic conditions for better living conditions. As a result, the distribution of resident population density presented clear spatial changes across this county (Figure 2e,f). Meanwhile, the economic level indexed by GDP possessed an outstanding increase from 10.9 billion (2005) to 21.4 billion (2015). Moreover, it was spatially differentiated across this county, the HD Town and BTX Town were the relatively developed areas across S County (Figure 2g,h). These results showed that S County achieved clear socioeconomic development and environmental improvement during 1995–2015.The age-standardized mortality rate of gastric cancer of S County displayed a decrease from 35.66/105 (period I) to 23.34/105 (period II). As illustrated in Figure 3, the gridded GCM was approximately described by a normal distribution in period I, and it turned into non-normal distribution and was right-skewed in period II. There were more than 60% points (the grids) below the 1:1 line (Figure 3), which indicates that the GCM of these grids declined while the others increased in 2004–2015. Meanwhile, the differences of the gridded GCM among these two periods were significant (p < 0.1) according to the results of the paired T-test. These results showed that the GCM of S County possessed a decline trend during 2004–2015.In S County, GCM was spatially featured on the grid scale in both periods. According to the Moran’s I as given in Figure 4, the gridded GCM was spatially clustered. Meanwhile, the grids with high GCM in the period I were mainly distributed along with two rivers, especially in the towns of XAJ and ZY around the SY River. In comparison, these grids tended to be away from these rivers in 2012–2015, resulting in a region with relatively lower gridded GCM between FQ and SY River. In addition, the grids with increasing GCM were mostly located in the towns away from the rivers including BLK Town in the northeastern corner, HS Town and LW Town in the east, and LF Town in the south as illustrated in Figure 4c. These results displayed that the spatial patterns of the gridded GCM had changed obviously.Moreover, the spatiotemporal variations in the gridded GCM were further confirmed through the hotspot analysis. As illustrated in Figure 5, the hotspot grids (Z score above 2.0) in both periods, as well as some persisting hotspot grids, were identified. In period II, the amount of hotspot grids increased from 42 to 46 and tended to be away from FQ and SY River. As a result, the persisting hotspot grids were respectively distributed in BYJ Town (9 grids) and LZD Town (1 grid), which deserved continuous attention in the future.By means of the GD method, the potential determinants were identified for the spatial variations in the gridded GCM across S County (Table 2). Not only the spatial variations of the factors across S County changed between the two periods (Figure 2), but also their effect on gridded GCM changed obviously. In these two periods, the variable indexed by the distance from the river was always an important factor, although its impacts decreased in the II period. In comparison, the elevation failed to impose impacts, and the other three variables indexed by the percentage of farmlands, population density, and GDP significantly affected the spatial variations in the gridded GCM across S County only in the II period. As a result, the powers of these factors could be ranked as following in period II: human-made pollution > socioeconomic level > physical environment. These results displayed that these selected variables had important effects to various degrees on the spatial variations in the gridded GCM across S County.In addition, the statistically significant difference in the average GCM between different amounts of strata was identified by the GD method (Table 3). In Period I, the closer to the river, the higher the GCM (Table 3). The effects of the distance from river on gastric cancer were reversed in Period II. There was no significant GCM difference at different elevation, which was fairly level with little variations across S County. The other variables showed spatial stratified heterogeneity only in Period II: GCM was high in the areas with a high percentage of farmlands; regions with more intensive population and GDP were accompanied by a lower GCM. Therefore, these variables’ influences on the spatial variations in the gridded GCM had distinctly changed over the past decades.At present, environmental changes and status of public health are two important public issues in China, especially in some regions (i.e., S County) with historic environmental pollution and serious public health. The spatiotemporal variations in the GCM and their relationships with environmental changes and socioeconomic status across S County were explored on a fine spatial scale in this study. Several interesting findings were achieved and would provide useful clues for local authorities to evaluate the measures of environmental protection and socioeconomic development timely.There was a remarkable decrease (37.4%) observed for the age-standardized mortality rate of gastric cancer in S County during 2004–2015. We think that the decline was probably related with the environmental improvement and socioeconomic development in this county. To our knowledge, quite a few small factories such as textile mills were shut down in the past decades, and similar policies were conducted in the upstream of SY and FQ River [23,25], which resulted in the environmental improvement, especially the surface water quality of SY and FQ River in S County (Table S1 in Supplementary Files). Meanwhile, the decline of resident population reduced the domestic pollution [42], and increasing urbanization of population and the development of economic guaranteed the implement of relevant interventions such as drinking water safety project [27,43]. In particular, the percentage of local population covered by the centralized water supply (i.e., the level of drinking water safety) in this county had been raised from less than 30% (2005) to more than 85% in 2015 [8,44]. Moreover, economic growth is always accompanied by the improvement of people’s living standards. Improvement of household sanitation, healthy eating habits, and the use of refrigerators all reduced the risk of developing gastric cancer [6,45,46]. In addition, the remarkable decrease was larger than the national level (20.6%) in the corresponding period, although S County still possessed a higher mortality rate (23.34/105) of gastric cancer than the national level (19.62/105) at present [7,47]. In a word, these results were encouraging for local authorities who had made great efforts, and then they could hand in a passable answer sheet for the public.Similarly, the spatial patterns of the gridded GCM changed obviously along with spatially differentiated socioeconomic development and environmental improvement in S County during 2004–2015. Among current proxy variables, the percentage of farmlands possessed an adverse effect on this disease, which was probably caused by excessive application of pesticides and fertilizers in this agricultural region [48]. Moreover, the distance to river made a relatively large contribution to the spatial variations in the gridded GCM, which was probably due to the historic status of both environment and public health. Before 2005, S County was well-known for its serious surface water pollution and some famous “cancer villages” distributed along with SY and FQ River [24], for which local authorities had implemented some strict environmental management (e.g., centralized water supply projects) in the towns along with these rivers [26]. However, the total contributions of current influencing factors were not large enough to fully interpreting the present spatial variations across S County. One reasonable explanation is that there may be one or more missing crucial variables (i.e., the town/village-level hygienic intervention) in this study. Another is that more information about the income level on the village/family scale was not adequately reflected by the economic level (GDP) employed in current study. In future investigations, more typical potential variables should be obtained for the evaluation. Nevertheless, these selected proxy variables have made clear contributions to the spatial patterns of GCM on the grid scale over the past decade.A preliminary trial was conducted in our study to characterize the spatiotemporal variations of GCM, and to evaluate their relations to some proxy variables indicating some environmental and socioeconomic changes in S County. To our knowledge, S County suffered from high mortality rates of other digestive tract cancers (e.g., liver, esophagus, etc.), which were closely associated with local environmental conditions over the past decades [7]. Moreover, there were several counties or districts possessing similar public health problems and environmental pollution, especially in Central China [28]. We think that this study exhibited a feasible paradigm for local authorities who would submit their answer sheets on environmental management and public health status.Several limitations of our study warrant mention. First, many more reliable variables, such as the village/household-level income conditions, and town/village-level hygienic interventions on this disease should be obtained and employed for more detailed analysis of spatiotemporal change of GCM in S County. Second, the contributions of selected variables to the spatial heterogeneities of GCM should be spatially characterized for each grid by using some spatiotemporal models (e.g., geographically weighted regression (GWR), temporal GWR, etc.) in the future, although the GD method has quantified these variables’ effects in the present study.The decline of GCM and its changes of spatial patterns in S County benefited from the environmental improvement and socioeconomic development. We suggest that the potential pollution derived from agricultural production should be effectively controlled, and that the local government should further promote the sustainable development mode in terms of coordinated the relationship between environment and the economy to maintain the positive situation. The study provided informative knowledge for other typical regions which were similar to S County.The following are available online at https://www.mdpi.com/1660-4601/16/5/784/s1. Table S1: Surface water quality in S County from 2004–2015.B.W. and H.R. conceived and designed the framework. C.C. carried out the experiments, and drafted the manuscript. H.R., Z.W., and B.W. revised and edited the manuscript.This research was supported by the National Key Research and Development Program of China (Grant No. 2016YFC1302602, 2016YFC1302603, and 2016YFC1302601).We would like to extend our special thanks to Professor Wu Jing at the Chinese Center for Disease Control and Prevention for her time and advice throughout the study. We would like to thank the editor and reviewers for their valuable comments and suggestions, which helped us to improve the quality of the article.The authors declare no conflict of interest.Illustration of S County.Five environment proxy variables of S County, including (a) distance from the river; (b) elevation (30 × 30 m2); (c)–(d) land-use type (1 × 1 km2); (e)–(f) resident population density (1 × 1 km2); (g)–(h) GDP density (1 × 1 km2). ’Median’ in (b)–(h) is the corresponding median value of proxy variables on grid scale.Gastric cancer mortality (GCM) in two periods. (a) Scatter plot of GCM; the blue dotted line is the 1:1 line. (b) Histogram of GCM in period I, the null hypothesis of normality was retained at the 0.05 level of significance. (c) Histogram of GCM in period II, the null hypothesis of normality was rejected at the 0.05 level of significance.Spatial variation of GCM. (a) GCM spatial pattern in period I; (b) GCM spatial pattern in period II, the Moran’s I is significant at 0.01 level; (c) the GCM difference between period I and period II (the grid without villages points were not taken into account).Result of hot spot analysis in terms of Gi Z score. (a) Hot/cold spot in Period I; (b) hot/cold spot in period II; (c) hot/cold spot both in Period I and Period II.Overview of proxy variables and data source.* Considering the lag effect of environmental factors on health [39], the percentage of farmlands in 1995 and 2005 are used as the corresponding environmental risk factor in period I and II, respectively.Determinants of GCM spatial heterogeneity.*** The result is statistically significant at 0.05 level.Average GCM in each region.The result is statistically significant at 0.05 level.

Med-MDPI/ijerph_3/ijerph-16-05-00785.txt

@@ -0,0 +1,6 @@
+Chronic kidney disease (CKD) has been redefined in the new millennium as any alteration of kidney morphology, function, blood, or urine composition lasting for at least 3 months. This broad definition also encompasses diseases or conditions that are associated with normal kidney function, such as a kidney scarring from an acute pyelonephritis episode or a single kidney, as a result of kidney donation. CKD is a relevant public health problem. According to the 2015 Global Burden of Disease Study, it was the 12th leading cause of death, leading to 1.1 million deaths, worldwide, each year. The role of CKD as a cause of death is evident where renal replacement therapy (RRT) is not available, however, its role in increasing death risk is not easily calculated. RRT consumes about 3–5% of the global healthcare budget where dialysis is available without restrictions. While the prevalence of CKD is increasing overall as lifespans extend, being linked to diabetes, hypertension, obesity, and atherosclerosis, CKD is at least partly preventable and its effects may be at least partly counterbalanced by early and appropriate care. We will welcome papers on all aspects of CKD, including organization, cost, and models of care. Papers from developing countries will be particularly welcomed.
+      
+        “In public health, we can’t do anything without surveillance. That’s where public health begins.”
+        —David Satcher, MD, PhD, U.S. Surgeon General, in 1998–2002
+      
+    “In public health, we can’t do anything without surveillance. That’s where public health begins.”Chronic kidney disease (CKD) is a major public health problem [1]. The 2015 Global Burden of Disease Study estimated that, in 2015, 1.2 million people died from CKD [2]. This umbrella term (which at the beginning of the new millennium replaced the previously used label, renal insufficiency) gathers all persistent alterations (those lasting at least three months) in morphology, urinary or blood composition, or any reduction in the glomerular filtration rate below 60 mL/min [3].The definition of CKD emphasizes the importance of the early phases of kidney damage, in which, due to the great functional reserve of kidney tissue, more than 50% of the kidney tissue has to be damaged before producing a rise in serum creatinine and a reduction in kidney function. In fact, this is the reason for the diffusion of living-donor kidney transplantation, a treatment aimed at restoring the health of a patient with end-stage kidney disease without producing detrimental effects on the health of the donor [4,5].The relationship between kidney tissue, kidney function, and health is complex. On one hand, as in the case of kidney donation, reduction of kidney tissue can be fully compensated without any effect on kidney function in standard situations. On the other hand, any reduction in kidney tissue, even in the presence of normal kidney function, can have a negative effect on conditions, for example during pregnancy, in which renal function is subject to stress [6]. It is only in the new millennium that the prevalence of CKD, the Cinderella of chronic diseases, has come to be fully appreciated. We now realize that in western countries CKD is present in 5–12% of the overall population, depending on the way its presence is detected and the features of the population. Its prevalence is therefore between that of diabetes mellitus and hypertension, much better known non-transmissible chronic diseases [1]. There seems to have been no decrease in the prevalence of CKD, which is not surprising given that it is more commonly found in older people and it is linked with the major non-transmissible diseases (hypertension, obesity, diabetes, and atherosclerosis). While in western countries the rise in the prevalence of CKD is generally in line with the ageing of the overall population, a sharp rise has been observed in emerging countries. This is more closely linked to a rise in diabetes and obesity, the price paid for adopting a “westernized” way of life, than to increased life expectancy [7,8].The relationship between kidney disease mortality and morbidity is complex. Where renal replacement therapy is available a loss of over 90% of kidney function is compatible with decades of survival on this therapy [9]. On the other hand, the presence of any degree of reduction in kidney function is associated with an increased risk of death, directly proportional to reduction in kidney function [10]. Although this complex relationship makes the estimation of the role of kidney disease as a cause of death difficult to quantify within these limits, CKD is considered the 12th cause of death worldwide.Conversely, the economic burden of end-stage kidney disease [ESRD] is more easily quantified. Renal replacement therapy [RRT], in the settings where it is available without restriction, consumes between 2% and 5% of the entire health care budget for a cohort that is approximately 1:1000 of the overall population. While kidney transplantation is the elective treatment for ESRD, with better overall survival results and lower costs, very old patients and those with multiple comorbidities are not considered eligible [11]. In most European countries the median age at the start of dialysis is about 75 and, while the conventional age limit for transplantation is now usually set at about 80, the prevalence of contraindications increases with age. As a result, the prevalence of kidney transplant candidates ranges from 40% to 70% according to the setting and the system of care [12].Dialysis is extremely expensive and while there is no single way to calculate costs, given significant differences in policies, health care systems, and reimbursement policies, in Europe one year of dialysis is calculated to cost between 50,000 and 100,000 euros [11].Dialysis is not available for all patients who need it. This is a tremendous health care problem affecting not only chronic patients but also patients with acute kidney diseases that require dialysis to allow kidney healing [13]. In 2010, an estimated 2.3–7.1 million people with end-stage kidney disease died without access to chronic dialysis [14]. The courageous program of the International Society of Nephrology, entitled ‘0 by 25’, is addressed to ensuring the availability of dialysis for all patients with an acute kidney injury [AKI], thus avoiding over 1.5 million deaths worldwide every year [13] (Figure 1).As is so often the case in this difficult world, the need is greatest in low-income countries, where, for instance, AKI is the main cause of death during and after pregnancy [15].On account of the high prevalence of CKD, the high cost, and unequal availability of treatment, ethical issues are an integral part of the history of nephrology and its present development. The first ethical committee known in the western world was set up in Seattle in the early 1960s to share the burden involved in deciding which patients should be put on dialysis, or “who should live”. In fact, in the early days of chronic dialysis the availability of an “artificial kidney” was so limited that choosing candidates was felt to require shared responsibility and the “God Committee” was therefore the first tragic example of the impact of economic and social factors on clinical choices (Figure 2) [16].While the involvement of politicians and the media led to “open access”, i.e., widening the availability of dialysis in North America and Europe, this tragic choice still confronts physicians elsewhere in the world in those regions which are, today, the most populated, poorest, and have highest incidence of kidney disease.From a public health perspective, the following are the main critical points:How to estimate the true prevalence and economic burden of CKDHow to implement efficient surveillance systemsHow to test new models of prevention and health care organizationHow to evaluate their effectivenessHow to ensure equity in access to preventive medicine and careHow to ensure the ecological sustainability of kidney treatmentsHow to estimate the true prevalence and economic burden of CKDHow to implement efficient surveillance systemsHow to test new models of prevention and health care organizationHow to evaluate their effectivenessHow to ensure equity in access to preventive medicine and careHow to ensure the ecological sustainability of kidney treatmentsThe concept of public health surveillance [PHS] is linked to all these critical points, as PHS is “the systematic and continuous collection, analysis, and interpretation of data, closely integrated with the timely and coherent dissemination of the results and assessment to those who have the right to know so that action can be taken” [17]. PHS is a complex system, collecting information from various sources and using it as the basis of decision making. Unfortunately, given a paucity of epidemiological data, lack of awareness, and limited access to testing, the actual burden of kidney disease is probably underestimated. PHS could contribute to increasing equity by underlining inequality, not only between social strata but also between areas and nations [18].The global burden of kidney disease is growing, driven by complex interactions, and treatment is fraught with environmental and socioeconomic disparities. Although CKD is costly, it is also at least partly preventable and adverse outcomes can be delayed, often by employing low cost treatments [19]. We need universal health coverage to ensure effective screening, prevention, and early treatment of CKD. Involvement of all relevant stakeholders and finding alternative financing strategies is necessary to promote equal access to care [20].There is need for action and action starts from awareness. This is the reason why we feel that this special issue will be welcomed as a means to strengthen the links between clinical nephrologists, economists, and policy makers.Conceptualization, E.V. and G.B.P.; writing—original draft preparation, E.V. and G.B.P.; writing—review and editing, E.V. and G.B.P.This research received no external funding.The authors declare no conflict of interest.One of the icons of the “O by 25” campaign of the International Society of Nephrology; Available online: https://www.theisn.org/all-articles/616-0by25 (accessed on 25 February 2019).The “God Committee” selecting patients for dialysis. West of the Rotunda—Difficult choices. Available online: http://danschmidtforsenate.com/blog/?p=291 (accessed on 23 January 2019).

Med-MDPI/ijerph_3/ijerph-16-05-00786.txt

@@ -0,0 +1 @@
+These authors contributed equally to this work.Coronary artery disease has become a major health concern over the past several decades. We aimed to explore the association of single nucleotide polymorphisms (SNPs) in the ATP-binding cassette subfamily A member 1 (ABCA1) and lifestyle factors with coronary artery disease (CAD) in dyslipidemia. This nested case-control study included 173 patients with CAD and 500 matched control individuals (1:3, case: control) from a district in southern China. We collected medical reports, lifestyle details, and blood samples of individuals with dyslipidemia and used the polymerase chain reaction-ligase detection reaction method to genotype the SNPs. The CC genotype of the additive and recessive models of rs4149339, together with regular intake of fried foods or dessert, increased the risk of CAD (adjusted odd ratio (OR) = 1.91, p = 0.030; adjusted OR = 1.97, p = 0.017; adjusted OR = 1.80, p = 0.002; adjusted OR = 1.98, p = 0.001). The AT + AA genotype of the dominant model of rs4743763 and moderate/heavy physical activity reduced the risk of CAD (adjusted OR = 0.66, p = 0.030; adjusted OR = 0.44, p = 0.001). The CT + CC genotype of the dominant model of rs2472386 reduced the risk of CAD only in males (adjusted OR = 0.36, p = 0.001). The interaction between rs4149339 and rs4743763 of ABCA1 and haplotype CTT (comprising rs4149339, rs4743763, and rs2472386) appeared to increase the risk of CAD (relative excess risk due to interaction (RERI) = 3.19, p = 0.045; OR = 1.49, p = 0.019). Polymorphisms of rs4149339, rs4743763 and rs2472386 in ABCA1 and three lifestyle factors (physical activity, fried food intake, and dessert intake) were associated with CAD in people with dyslipidemia in southern China. These results provide the theoretical basis for gene screening and the prevention of chronic cardiovascular diseases.In recent years, cardiovascular and cerebrovascular diseases, represented by coronary artery disease and stroke, have become the world’s top causes of mortality in humans [1]. According to the 2017 China Cardiovascular Disease Report, 11 million people in urban and rural areas suffer from coronary artery disease. The mortality rate of patients with coronary artery disease increased from 39.56 per 100,000 in 2002 to 110.91 per 100,000 in 2015, an increase of nearly three times [2]. Atherosclerosis is the main basis for the onset of coronary heart disease and it usually manifests as vascular endothelial cell damage and endothelial function decline caused by dyslipidemia, which are chronic inflammatory responses [3]. Dyslipidemia and atherosclerosis have been assessed by a large number of epidemiological studies [4,5,6,7]. Compared with individuals with normal blood lipid levels, individuals with abnormal lipid levels are more likely to develop atherosclerotic lesions, eventually leading to severe cardiovascular diseases such as coronary artery disease [8,9]. Coronary artery disease is a complex disease that is affected by genetic and environmental factors. Most genes can affect lipid metabolism in the body, followed by the development of atherosclerosis [10,11,12,13,14,15,16]. Some studies have found that cardiovascular diseases such as coronary heart disease have a strong genetic basis, demonstrated by family associations and degrees of genetic influence as high as 40–50% [17,18]. ATP-binding cassette transporter A1 (ABCA1) is a member of the ATP-binding cassette transport factor gene family, whose main function is to balance the cholesterol concentration inside and outside the cell to maintain normal levels among various blood lipids. It is also involved in the formation of the atherosclerotic inflammatory response, which may be directly related to the occurrence and development of atherosclerotic lesions [19]. In addition, lifestyle factors, such as lack of physical activity and a regular intake of fried foods and sweets, are also suggested to be risk factors for coronary artery disease (CAD) [20,21,22]. The present study was a community-based nested case-control study to analyze the association of three ABCA1 single nucleotide polymorphisms (SNPs, one in the 3’-untranslated region and two in the intron region) and certain lifestyle factors (dessert and fried food intake and physical activity intensity) with CAD in a Han Chinese dyslipidemic population in southern China.We randomly selected four townships under the jurisdiction of Ningbo, Zhejiang Province, as survey points by cluster random sampling. A total of 2349 people who underwent physical examinations at the community health service center from April 2013 to July 2013 and had dyslipidemia were included. All subjects were unrelated and >40 years of age. We excluded patients treated with antihypertensive drugs or hypolipidemic drugs, as well as patients with severe liver and kidney disease and malignant tumors. Lifestyle information and blood samples were collected from all subjects, and a visit was conducted in August 2016 to obtain data on the incidence of CAD in the subjects. The case group included 173 patients diagnosed with CAD between April 2013 and August 2016. Five hundred age- and sex-matched subjects were included as control individuals (1:3, case: control). All participants signed informed consent forms. The research protocol was approved by the Medical Ethics Committee of the Affiliated Hospital of Hangzhou Normal University.Dyslipidemia was defined according to the Guidelines for the Prevention and Treatment of Diabetes in Chinese Adults prepared by the Joint Committee of the Chinese Association for the Prevention and Treatment of Dyslipidemia. Total cholesterol (TC) > 5.18 mmol/L, triglycerides (TG) > 1.70 mmol/L, low-density lipoprotein cholesterol (LDL-C) > 3.37 mmol/L, and high-density lipoprotein cholesterol (HDL-C) < 1.04 mmol/L represent abnormal values. One or more of the above four blood lipid indicators suggest a diagnosis of dyslipidemia. Subjects diagnosed with dyslipidemia and taking hypolipidemic drugs were also considered under the dyslipidemia category. CAD is usually diagnosed by computed tomography, radiography, or coronary angiography. The diagnosis was based on the diagnostic criteria for coronary atherosclerotic heart disease issued by the China Health and Family Planning Commission in 2010 [23]. Clinically diagnosed coronary artery ischemia, asymptomatic coronary heart disease, myocardial infarction, angina pectoris, or ischemic cardiomyopathy are all covered under CAD. The field epidemiological investigation mainly included basic demographic criteria such as age, sex, occupation, and education level, as well as information on lifestyle such as diet and intensity of physical activity. Among them, the dietary behavior survey employed a semi-quantitative questionnaire, divided into food type, intake, and frequency. The main variables of lifestyle were defined as follows. (1) Diet: average weekly intake of fried food less than once was defined as “no fried food intake,” and average weekly dessert intake less than once was defined as “no dessert intake.” (2) Physical activity classification: “secretarial work, office work, etc.” was defined as “sedentary physical activity”; “sales, hotel service, chemical experimentation, lecturing, etc.” was defined as “light physical activity”; “daily student activities, motor vehicle driving, electrician work, metalworking, cutting, etc.” was defined as “moderate physical activity”; and “steelmaking, dance, sports, loading and unloading goods, construction work, etc.” was defined as “heavy physical activity.” Anthropometric data, including waist circumference; body mass index (BMI); systolic blood pressure (SBP); diastolic blood pressure (DBP); and TC, TG, HDL-C, and LDL-C levels, were evaluated by professional medical examination according to standard protocols.Three SNPs in ABCA1 were selected using the HapMap website and the Haploview 4.2 software (Broad Institute, Cambridge, MA, USA). The SNP rs4149339 resides in the 3′-UTR of ABCA1, and rs4743763 and rs2472386 reside in the intron of ABCA1. The minor allele frequencies (MAFs) of these three SNPs were greater than 5%, and the linkage disequilibrium r2 > 0.8.Five milliliters of whole blood from each fasted individual was anticoagulated with EDTA and stored in a refrigerator at −80 °C. DNA was extracted using Tiangen Blood Genomic DNA extraction kits (Tiangen Biotech, Beijing, China) and sent to Shanghai Jierui Biological Engineering Co., Ltd., for genotyping analysis using the polymerase chain reaction (PCR)-ligase detection reaction (LDR) method (Generay Biotech Company, Shanghai, China). The primer sequences of rs4149339 were 5′-TCTTGGCTTTTGCATTGTTG-3′ (forward) and 5′-CTGTGCCATGTTATTCAGCTC-3′ (reverse). The primer sequences of rs4743763 were 5′-TCTGTCATGTGGCTGCAACT-3′ (forward) and 5′-ATGCAACAGATGCCCTATCC-3′ (reverse). The primer sequences of rs2472386 were 5′-TTCCCCTGCATCAAGTTTTC-3′ (forward) and 5′-TTGATCTGCCCTTTGTTTCC-3′ (reverse). The PCR reaction volume contained 1 µL DNA, 1.5 µL 10X buffer, 1.5 µL MgCl2, 0.3 µL dNTPs, 0.15 μL of each primer, 0.2 µL Taq enzyme, and water to make the total volume 15 µl. The amplification conditions were as follows: 94 °C for 3 min; 35 cycles of denaturation at 94 °C for 15 s, annealing at 55 °C for 15 s, and extension at 72 °C for 30 s; and 72 °C for 3 min. The ligation reaction volume comprised 3 µL PCR product, 1 µL 10X Taq DNA ligase buffer, 0.125 µL Taq DNA ligase (40 U/µl), 0.01 μL of each discriminating probe, and water to make the total volume 10 µL. The reaction conditions were as follows: 30 cycles of 94 °C for 30 s and 56 °C for 3 min. To 1 µL of extension product, 8 µL of loading buffer was added, followed by denaturation at 95 °C for 3 min. Next, the product was immediately bathed in ice water. The mixture was analyzed using a sequencer (ABI 3730XL). For quality control, we randomly chose 10% of samples for re-genotyping, and the concordance was 100%.Statistical analysis was performed using SPSS 24.0 software (SPSS Inc., Chicago, IL, USA). Demographic characteristics and the SNP genotypes of ABCA1 were evaluated using the χ2 test, Fisher’s exact test (for categorical variables), Student’s t-test, and Wilcoxon’s rank sum test (for continuous variables). The χ2 test was used to test for Hardy–Weinberg equilibrium. The associations of genetic models and lifestyle factors with the risk of CAD were estimated by calculating odd ratios (ORs) and 95% confidence intervals (95% CIs) using logistic regression analysis. The forest map used R language (package “forestplot”). The relative excess risk due to interaction (RERI), OR, and 95% CI were determined using Microsoft Excel according to Knol et al. [24]. The associations between ABCA1 haplotypes and the risk of CAD were calculated using the R language package “haplo.stats.” In all analyses, p values less than 0.05 were considered statistically significant.The subjects included 173 cases and 500 controls; 42.3% of the subjects were female, and 57.7% were male. The mean BMI was 24.23 ± 3.23 kg/m2 in case subjects, which was significantly more than that in control subjects (23.11 ± 2.93 kg/m2) (p < 0.001). The medians of waist circumference, DBP, and LDL-C in the case group were significantly higher than those in the control group (p < 0.05) (Table 1).All three of the studied SNPs in the control subjects were in Hardy–Weinberg equilibrium (p > 0.05) (Table S1). Linkage disequilibrium (LD) analysis for the three SNPs showed obvious LD between two SNPs (Table S2). The χ2 test found significant differences for physical activity, fried food intake, dessert intake, and the rs4149339 genotype between the case group and the control group (p < 0.05) (Table 1).Stratified by sex, in the case group and control group, HDL-C, physical activity, derssert intake, fried food intake, and rs4743763 and rs2472386 genotypes were found to be statistically significant only in male subjects. The rs4149339 genotype was found to be statistically significant only in female subjects (Table S3). The other meaningful results are the same as in the previous table.The association between the ABCA1 gene and lifestyle factors with CAD in dyslipidemia was examined under each gene model. With or without adjustment for the confounding factors age, sex, waist circumference, smoking, and drinking, the rs4149339 additive and recessive models, the rs4743763 dominant model, physical activity, fried food intake, and dessert intake were found to be significantly associated with CAD in dyslipidemia. In rs4149339 additive model, people carrying the CC genotype were found to be 0.91 times more likely to develop CAD than those with the TT genotype (adjusted OR = 1.91, 95% CI = 1.06–3.41, p = 0.030). People carrying the CC genotype of the recessive model were at a higher risk of CAD than those with the TT + CT genotype (adjusted OR = 1.97, 95% CI = 1.13–3.44, p = 0.017). People carrying the AT + AA genotype of the dominant model of rs4743763 had a lower risk of CAD than those with the TT genotype (adjusted OR = 0.66, 95% CI = 0.45–0.96, p = 0.030). Compared with people performing sedentary/light physical activity, those performing moderate/heavy physical activity were less susceptible to CAD (adjusted: OR = 0.44, 95% CI = 0.27–0.71, p = 0.001). Regular intake of fried foods was 0.80 times more likely to cause CAD in dyslipidemia patients than in those who did not consume fried foods (adjusted: OR = 1.80, 95% CI = 1.24–2.61, p = 0.002). Similarly, regular intake of dessert posed a 0.98 times higher risk of CAD compared to no dessert intake (adjusted: OR = 1.98, 95% CI = 1.32–3.00, p = 0.001) (Figure 1).After stratification by sex, we found that carrying the AT + AA genotype of dominant model rs4743763 (adjusted OR = 0.28 , 95% CI = 0.15–0.55, p < 0.001), the CT+CC genotype of dominant model of rs2472386(adjusted OR = 0.36, 95%  CI = 0.19–0.67, p = 0.001) and moderate/heavy physical activity(adjusted OR = 0.24, 95% CI = 0.11–0.51, p < 0.001) are protective factors for CAD in males (Table S4). Regular Fried food intake (adjusted: OR = 2.48, 95% CI = 1.37–4.51, p = 0.003) and regular dessert intake (adjusted: OR = 2.69, 95% CI = 1.41–5.12, p = 0.003) are risk factors for CAD in males. rs4149339 CC genotype of recessive model (adjusted: OR = 2.11, 95% CI = 1.04–4.25, p = 0.037) is protective factor for CAD in females (Table S5).Table 2 shows the effect of the interaction of the ABCA1 rs4149339 polymorphism and lifestyle factors on coronary artery disease. After adjustment for age, sex, waist circumference, smoking, and drinking, compared with people carrying the CT + TT genotype who performed moderate/heavy physical activity, those with the CT + TT or CC genotype who performed sedentary/light physical activity were at a higher risk of CAD (OR = 2.45, 95% CI = 1.46–4.13, p = 0.001; OR = 4.10, 95% CI = 1.94–8.67, p < 0.001). Compared with people carrying the CT + TT genotype and having no fried food intake, those with the CT + TT or CC genotype and having regular fried food intake were at an increased risk of CAD (OR = 1.88, 95% CI = 1.26–2.79, p = 0.003; OR = 2.98, 95% CI = 1.47–6.02, p = 0.002). People carrying the CC genotype with or without dessert intake and those carrying the CT + TT genotype and with regular dessert intake were at a higher risk of CAD than individuals carrying the CT + TT genotype and having no dessert intake (OR = 5.26, 95% CI = 1.81–15.25, p = 0.002; OR = 3.04, 95% CI = 1.46–6.34, p = 0.003; OR = 2.29, 95% CI = 1.47–3.56, p < 0.001). The interaction between the rs4149339 polymorphism and lifestyle factors was not found in the additive model (p values of RERI > 0.05).Table 3 shows the effects of the interaction between the ABCA1 rs4743763 polymorphism and lifestyle factors on coronary artery disease. After adjustment for age, sex, waist circumference, smoking, and drinking, compared with people carrying the AT + AA genotype who performed moderate/heavy physical activity, those with the AT + AA genotype who performed sedentary/light activity, as well as those with the TT genotype and performing sedentary/light physical activity had a higher risk of CAD compared with those with the AT + AA genotype and performing moderate/heavy physical activity (OR = 2.80, 95% CI = 1.18–6.64, p = 0.020; OR = 3.88, 95% CI = 1.70–8.86), p = 0.001). Within the strata of TT, people performing sedentary/light physical activity had a higher risk of CAD than those performing moderate/heavy physical activity (OR = 2.07, 95% CI = 1.14–3.74, p = 0.016). People carrying the TT genotype with or without regular fried food intake and those carrying the AT + AA genotype and having regular fried food intake had an increased risk of CAD compared with those carrying the AT+AA genotype and having no fried food intake (OR = 2.38, 95% CI = 1.20–4.74, p = 0.013; OR = 3.50, 95% CI = 1.82–6.73, p < 0.001; OR = 2.92, 95% CI = 1.44–5.91, p = 0.003). People carrying the TT or AT + AA genotype and having regular dessert intake showed a higher risk of CAD than those carrying the AT + AA genotype and having no dessert intake (OR = 2.92, 95% CI = 1.55–5.50, p = 0.001; OR = 2.07, 95% CI = 1.04–4.16, p = 0.039). The risk of CAD in people with dyslipidemia having the TT genotype and regular dessert intake was 0.85 times higher than that of people with dyslipidemia having the TT genotype and no intake of dessert (OR = 1.85, 95% CI = 1.11–3.07, p = 0.018). The interaction between rs4743763 and lifestyle factors was not found in the additive model (p values of RERI > 0.05).Table 4 shows the effects of the interaction between the ABCA1 rs4149339 and rs4743763 polymorphisms on coronary artery disease. After adjustment for age, sex, waist circumference, smoking, and drinking, compared with people carrying the rs4149339 CT + TT and the rs4743763 AT + AA genotypes, those carrying the rs4149339 CC and rs4743763 TT genotypes demonstrated a higher risk of CAD in dyslipidemia (OR = 4.35, 95% CI = 2.07–9.15, p < 0.001). Within the strata of rs4149339 CC, dyslipidemic people with the rs4743763 TT genotype had a higher risk of CAD than those with the AT+AA genotype (OR = 4.66, 95% CI = 1.29–16.79, p = 0.019). Within the strata of rs4743763 TT, dyslipidemic people with the rs4149339 CC genotype had a higher risk of CAD than those with the CT + AA genotype (OR = 3.48, 95% CI = 1.71–7.10, p = 0.001). A positive interaction between the rs4149339 CC and rs4743763 TT genotypes was found in the additive model (RERI (95% CI) = 3.19 (0.07–6.30), p = 0.045).The haplotype frequencies of the three SNPs were compared between CAD cases and control subjects (Table 5). Five common haplotypes (frequency > 1%) derived from the three SNPs accounting for 95% of the haplotype variation were selected, and the remaining haplotypes were pooled into the rare group. The haplotype CTT was found to be associated with an increased risk of CAD (OR = 1.49, 95% CI = 1.07–2.08, p = 0.019).ABCA1 is located in the long arm of chromosome 9 (9q31) and is 149 kb long. It is a member of the ATP-binding cassette transporter superfamily, being highly expressed in a variety of tissues and cells such as the liver, intestines, lungs, white blood cells, and macrophages [25,26]. ABCA1 hydrolyzes ATP, and the energy released is used to transport various molecules across the cell membrane [27]. ABCA1 is involved in the first step in reversing cholesterol transport by regulating cholesterol and phospholipid efflux from peripheral cells to lipid-deficient apolipoprotein receptors [28,29]. According to both animal experiments and population studies, decreased ABCA1 activity might result in decreased lipid efflux in peripheral tissue cells. This increases inflammatory signaling factors in atherosclerotic plaques, thereby increasing the risk of coronary artery disease [19,20,22]. In addition, it affects the formation of HDL particles in the liver and intestine [30]. The results of this study showed that the CC genotype of the additive model of rs4149339 had a higher risk of CAD compared to the TT genotype; moreover, the recessive model was also significantly associated with CAD. After adjustment for confounding factors the correlation remained. This indicates that the C mutation site of ABCA1 rs4149339 is a risk factor for CAD in people with dyslipidemia, which is consistent with the results of a case-control study conducted by Lu et al. [31] in China in 2015. The dominant model of rs4743763 was found to be significantly associated with CAD. Even after adjustment for confounding factors, the association remained. This indicates that the T mutation site of ABCA1 rs4743763 is a risk factor for CAD in dyslipidemia, and it is a newly identified SNP for susceptibility to coronary heart disease in the Chinese population. Meanwhile, an association of ABCA1 rs2472386 with CAD was found in dyslipidemia only in male subjects. This suggests that the association may be influenced by sex.This study also analyzed the association of physical activity intensity, fried food intake, and dessert intake with CAD in dyslipidemia. These three factors were all found to be significant in the occurrence of CAD with dyslipidemia. Sedentary/light physical activity and the regular intake of fried and dessert increased the risk of CAD in people with dyslipidemia, consistent with most previous studies [32,33,34,35,36]. Physical activity can effectively protect the cardiovascular system from damage. Adequate exercise enhances the myocardial load capacity and increases left ventricular wall thickness and arterial cavity size [37]. Lack of physical activity has been found to result in up to twice the incidence of cardiovascular events, while moderate- to high-intensity physical activity can reduce cardiovascular risk by 10–50% [33]. Sweets and fried foods contain large amounts of saturated and trans fatty acids. In one study, the correlation between saturated fatty acid intake and TC in the blood was found to be 0.23 [38]. If foods with high saturated fatty acid content are used often in daily diets, the plasma levels of TC and LDL-c will increase considerably, and accordingly, the risk of coronary heart disease will increase [39]. After stratification by sex, the association between CAD and behavior in men was consistent with the above, and there was no significant association between the risk of CAD and three behaviors in women, suggesting that we should pay more attention to male subjects when conducting community behavior intervention.Epidemiological experts believe that quantitative interactions in the additive model are best suited to assess the importance of interaction. In case-control studies, the RERI caused by interaction is generally considered the standard measure of additive model interaction. Our study explored the interactions of ABCA1 gene-lifestyle factors, SNP–SNP, haplotypes, and certain lifestyle factors with the risk of CAD. RERI, as well as the p values and 95%CI of RERI, are reported herein. Although the interaction index of rs4149339 and rs4743763 with physical activity, fried food intake, and dessert intake was not statistically significant, people carrying risk alleles of rs4149339 and rs4743763 who also performed sedentary/light physical activity or had a regular intake of fried or dessert were also at a high risk for the disease. An interaction between rs4149339 and rs4743763 was found in the additive model. The RERI was 3.19 (95% CI: 0.07–6.30), suggesting that the estimated joint effect on the additive scale of rs4149339 CC and rs4743763 TT genotypes together was greater than the sum of the estimated effects of either genotype alone. Thus, there was a positive interaction on the additive scale. In addition, we found that the haplotype CTT, composed of rs4149339, rs4743763, and rs2472386, might also increase the risk of CAD.This study had some limitations. Owing to the time limitation of cohort observation, this study only collected cases within a 3 year period, resulting in the inclusion of fewer cases, which might have led to unstable research results. Second, this study was only carried out only in Ningbo. Due to geographical restrictions, the research conclusions might only be applicable to people in southern China. Whether the study findings can be applied to other groups remains to be explored.In conclusion, polymorphisms of rs4149339, rs4743763 and rs2472386 in ABCA1 and three lifestyle factors (physical activity, fried food intake, and dessert intake) were found to be associated with CAD in people with dyslipidemia in southern China. These results provide the theoretical basis for gene screening and the prevention of chronic cardiovascular diseases.The following are available online at https://www.mdpi.com/1660-4601/16/5/786/s1, Table S1: Distributions of allele and genotype in the subjects of case and control groups, Table S2: D’value of Linkage disequilibrium for three SNPs of ABCA1, Table S3: General characteristics, lifestyle factors and genotype distribution between case and control for male and female, Table S4: Associations of genetic models and lifestyles with risk of CAD for male, Table S5: Associations of genetic models and lifestyles with risk of CAD for female.Author Contributions: We thank all the individuals who participated in the present study. T.-Y. Z., L.-W.X. and L.Y. had the original idea for the study and, with all co-authors carried out the design. Y.-N.W., X.-N.W., P.-P.Z., X.-J.X. and L.Z. were responsible for recruitment and follow-up of study participants. T.-Y.Z. and S.L. drafted the manuscript, which was revised by all authors. L.H. provided advanced statistical methods. All authors read and approved the final manuscript.This study was funded by National Science Foundation of China (30860241). The authors are thankful to the team working with this study.The authors have no conflict of interest.Forest map for the risk of coronary artery disease (CAD) with different genetic models and lifestyle factors. Adjusted for age, sex, waist circumference, smoking, and drinking. p value < 0.05 was considered statistically significant and maintains significance using the Benjamini-Hochberg procedure with the false discovery rate at 0.11.General characteristics, Lifestyle factors and genotype distribution between case and control groups.IQR, interquartile range; SD, standard deviation; BMI, body mass index; SBP, systolic blood pressure; DBP, diastolic blood pressure; TC, total cholesterol; TG, triglyceride; HDL-C, high-density lipoprotein cholesterol; LDL-C, low-density lipoprotein cholesterol. Bold values are statistically significant with p value <0.05 and maintains significance using the Benjamini-Hochberg procedure with the false discovery rate at 0.11.Interaction between ABCA1 rs4149339 polymorphism and lifestyles for the risk of CAD.RERI, relative excess risk of interaction. Adjusted for age, sex, waist circumference, smoking and drinking. Bold values are statistically with p-value < 0.05 and maintains significance using Benjamini-Hochberg procedure with the false discovery rate at 0.11.Interaction between ABCA1 rs4743763 polymorphism and lifestyles for the risk of CAD.Adjusted for age, sex, waist circumference, smoke, drinking. Bold values are statistically with p-value < 0.05 and maintains significance using Benjamini-Hochberg procedure with the false discovery rate at 0.11.Interaction between ABCA1 rs4149339 and rs4743763 polymorphism for the risk of CAD.Measure of interaction on additive scale: relative excess risk due to interaction (RERI) (95% CI) =3.19 (0.07–6.30), p = 0.045. Adjusted for age, sex, waist circumference, smoking and drinking. Bold values are statistically with p-value < 0.05 and maintains significance using Benjamini-Hochberg procedure with the false discovery rate at 0.11.Frequencies of haplotypes among cases and controls and association with risk of CAD.a The alleles of haplotypes were arrayed as rs4149339–rs4743763–rs2472386; b Haplotypes with frequency <0.01 were pooled into the rare group; Adjusted for age, sex, waist circumference, smoke, drink; p-value < 0.05 and maintains significance using Benjamini-Hochberg procedure with the false discovery rate at 0.11.

Med-MDPI/ijerph_3/ijerph-16-05-00787.txt

@@ -0,0 +1 @@
+To examine the effects of short messaging service (SMS) frequency and timing on the efficacy of an SMS-intervention for Hong Kong Chinese adolescents, sixty nine students aged between 12 and 16 (mean age 13.75 ± 0.90) were recruited from five schools in Hong Kong. Participants were randomly assigned into one of five groups: high-frequency + self-selected timing (HST), low-frequency + self-selected timing (LST), high-frequency + assigned timing (HAT), low-frequency + assigned timing (LAT) and the control group. The total duration of the intervention was four weeks. No significant intervention effects were detected in adolescent’s PA among the five groups (F = 1.14, p = 0.346). No significant differences were observed in the stage movement among the five groups (χ2 = 6.18, p = 0.627). No significant differences appeared in the exercise benefits, barriers and benefits/barriers differential scores. However, a growth trend in the exercise benefits score in the LST and HAT groups was found in contrast to the downswing in the control group. The exercise barriers score in the HST group showed the largest reduction after intervention. The benefits/barriers differential score in all the intervention groups increased, whereas it decreased in the control group. Although an increase is demonstrated in the high dosage SMS frequency and timing, no significant intervention effects were found among the five groups in PA behavior, stage of change and exercise benefits and barriers among Hong Kong Chinese adolescents.Regular physical activity (PA) is associated with a reduced risk of obesity and type II diabetes mellitus in adolescents, as well as a reduced risk of diseases that develop later in life, such as breast cancer, hypertension, coronary heart disease and osteoporosis [1,2]. However, despite the well-established and documented benefits of PA, over half of the young population around the world does not engage in sufficient PA to achieve the potential significant health benefits [3]. The Boys’ and Girls’ Clubs Association of Hong Kong (BGCA) (2018) [4], has revealed that 59.1% of 6 to 17-year-olds in Hong Kong engaged in less than six hours of moderate to vigorous intensity activity weekly. The recommended level for this age group is 60 min or more of moderate-to-vigorous intensity PA on most days of the week. Another recent questionnaire survey revealed that less than 10% of 7 to 19-year-old self-reported participation in at least 60 min of moderate-to-vigorous physical activity (MVPA) a day. [5]. Therefore, innovative approaches to promote PA to adolescents are of paramount importance.Over the last decade, numerous physical activity (PA) interventions have been developed and implemented. However, many are ineffective [6], and traditional PA interventions, which are typically delivered face-to-face, may not be affordable for large-scale use. However, information communication technology (ICT), more specifically, Short Message Services (SMS) via mobile phones, is considered a promising intervention delivery method by which to influence and potentially increase behavioral change [7,8]. SMS is a text messaging service component of most telephone, internet, and mobile-device systems. It is a most widely used data application. The intervention effect of SMS on different health behaviors including physical and mental health has been supported in previous studies in which age, ethnicity and socio-economic status are considered [9,10]. Additionally, SMS is also fast, portable, convenient and inexpensive, and has a wider reach than other forms of communication [9]. Because the mHealth technologies are currently making considerable advances in PA behavioral change, SMS could play a more important role within PA behavioral change interventions if the functions are further explored and maximized.The reminder effect through SMS to the participants is the most obvious function in achieving a significant intervention effect in previous studies [11,12]. Conversely, SMS symbols could develop a closer relation, better outreach, and in-group culture when communicating with peers. Delivery of SMS could also reduce the barriers (i.e., data tracking and analysis) associated with treatments or services provided to the patients in medical settings [13]. SMS produced a more motivational, interactive and social support effect on the participants [14,15,16]. SMS is also an inexpensive and automated communication technology in which the message can be “pushed” to the receiver, and he/she would not consider rejecting it. This automated communication reaction enhances the probability of success of the intervention.Currently, among adolescent mobile phone users, SMS is an integral part of daily activities, enabling the maintenance of social networks and the expression of feelings [17,18,19]. More importantly, from an adolescent perspective, SMS is considered among young people as a fashionable means of reflecting individual identity and culture within peer groups, for example, through the increased use of personalized languages and emotional symbols, i.e., emoji [18,19].The use of mobile phone technology has been widely adopted in chronic disease management programs such as diabetes [20], asthma [21,22], sexual health [23], smoking cessation [24,25], alcoholism, drug abuse and weight loss [26] and has consistently demonstrated promising results in acting as an adjunct to existing treatment, increasing awareness, and improving compliance with medication and adherence to treatment programs.The use of mobile technology for improved PA levels has primarily focused on adult populations [27,28]. However, recently an increasing number of studies have implemented mobile to promote PA behaviors among adolescents [29,30]. A review by Lau et al. [31] found support for the use of SMS as a medium to improve PA among children and adolescents; however, six of the nine studies included in the review demonstrated satisfactory methodological quality.The trans-theoretical model (TTM) has been one of the major models for the stage of change in past decades when investigating behavioral change among children and adults in the literature [32,33]. This stage-matched concept demonstrated significant insights regarding both PA adoption and maintenance when designing interventions for adolescents in an intervention study [33,34]. A major value of the stage of change model is that it not only demonstrates the readiness of an individual’s intention to change his or her PA behavior but also provides stage-based and stage-matched guidelines to the PA researchers [35]. Because the TTM has demonstrated effects on the improvement of PA in the student population, it would be interesting to observe whether SMS may be able to produce a significant impact on the adolescents in this study, particularly on the transition, if any, between different stages of change in PA behavior [32].Perceived exercise benefits and barriers have been documented as significant mediators in the previous PA behavioral change studies [36,37]. Significant relations were demonstrated between PA benefits, PA barriers and the current PA habits of the participants [38]. A better understanding of the perceived benefits and barriers to PA could enhance the intervention effectiveness and efficiency [37]. The Exercise Benefits and Barriers Scale (EBBS) can help assess the changes in participant’s perceived benefits, barriers and any discrepancies. These scores would provide a better picture to understand and interpret the intervention findings in this study. It is also important to know the dynamic relations or priority between the benefits and barriers in PA when addressing adolescents’ PA behavioral change because these changes may be different when compared with different populations [39].Because SMS-based interventions are increasing, it is essential to establish practical guidelines, particularly with respect to content and frequency. Reviews of behavior change interventions delivered by SMS [9,40] indicated that successful SMS-based health behavior change interventions were delivered at high frequency (at least 5 SMS per week) with an individually tailored SMS timing. However, these studies have all been conducted in European and American populations, and no such investigations have been conducted in Chinese population or, more specifically, in Chinese adolescent population.In Hong Kong, a preliminary investigation supported the efficacy of promoting PA through SMS and Internet-based programs [41]. However, the effect of tailored timing and frequency has not yet been empirically examined. Existing behavior change theories have emphasized the “why” but relatively less the “how” (i.e., procedures), and behavior change interventions have not met expectations in terms of effectiveness [6]. This study’s objective was to address this issue by determining how best to integrate SMS into adolescent PA behavior. These findings may offer practical guidelines to inform the design of future SMS-based intervention for Chinese adolescent’s PA behavior. The specific purpose of this study was to examine the effects of SMS frequency and timing on the efficacy of a SMS-based intervention for Hong Kong Chinese adolescent’s PA.The research questions were:(1)What is the effect of the frequency and timing of the delivery of SMS impact on the quantity of adolescent’s PA?(2)Would the stages of motivational readiness for PA be affected as a result of the SMS intervention?(3)Would adolescents perceive a difference in benefits and barriers to PA as a result of the SMS intervention?What is the effect of the frequency and timing of the delivery of SMS impact on the quantity of adolescent’s PA?Would the stages of motivational readiness for PA be affected as a result of the SMS intervention?Would adolescents perceive a difference in benefits and barriers to PA as a result of the SMS intervention?Participants were randomly recruited from five of 469 government-subsidized schools in Hong Kong, China. Sixty nine students aged between 12 and 16 (33 girls, 36 boys, age 13.75 ± 0.90) were recruited by physical education (PE) teachers from the five schools in Hong Kong. Ethical approval was granted by the Senate Committee on the Use of Human and Animal Subjects in Teaching and Research, Hong Kong Baptist University (project identification code: FRG2/10-11/046). Screening occurred prior to the selection of participants, who were included if they: (1) were between 12–16 years old; (2) owned a personal mobile phone and (3) had a parent able to provide written informed consent.A fact sheet inviting participation was sent to parents three months prior to the intervention. Two weeks prior to the start of the intervention, all those who had provided written consent attended a briefing session and complete a baseline questionnaires which consisted of their socio-demographics, stage of change, perceived benefits and barriers of PA, and PA level. With this information, an algorithm of stage of change was used to identify their stage and messages matched with their respective stage of behavior were sent to them from the message bank which was developed in a pilot study [41]. Then, the tailored messages and phone numbers of each participant of the intervention group were entered to a commercial computerized texting system to set up a delivery schedule. The sender was set in the system by using their respective PE teachers’ name to provide a more solid support and motivation to the participants. All measures were assessed at baseline and 4-week (post intervention). The total duration of the intervention was four weeks.This study used a randomized controlled cluster design. Invitations to participate in the study were sent to all government-subsidized secondary schools in Hong Kong. Among the 469 schools invitations, 10 schools replied and five schools were selected based upon the district categorizations (Hong Kong Island, Kowloon and New Territories) in Hong Kong. Each school was allocated through a ballot in which the five group labels (HST, LST, HAT, LAT, C) were picked. Participants were school students randomly assigned into one of the five groups (Figure 1):(1)High-frequency + self-selected timing (HST): Participants received five SMS per week (on weekdays). Self-determined timing for the receipt of SMS (before school/afterschool/after dinner).(2)Low-frequency + self-selected timing (LST): Participants received three SMS per week. Self-determined timing for the receipt of SMS.(3)High-frequency + assigned timing (HAT): Participants received five SMS per week. Assigned timing for the receipt of SMS (after school).(4)Low-frequency + assigned timing (LAT). Participants received three SMS per week. Assigned timing for the receipt of SMS (after school).(5)Control group (C). No treatment.High-frequency + self-selected timing (HST): Participants received five SMS per week (on weekdays). Self-determined timing for the receipt of SMS (before school/afterschool/after dinner).Low-frequency + self-selected timing (LST): Participants received three SMS per week. Self-determined timing for the receipt of SMS.High-frequency + assigned timing (HAT): Participants received five SMS per week. Assigned timing for the receipt of SMS (after school).Low-frequency + assigned timing (LAT). Participants received three SMS per week. Assigned timing for the receipt of SMS (after school).Control group (C). No treatment.Measures included: demographic information, self-reported PA, stage of behavior change and perceived benefits and barriers to PA. PA was measured with the Physical Activity Questionnaire for Children (PAQ-C) [42]. The PAQ-C is a self-administered, 7-day recall questionnaire for children aged 8 to 14 years composed of ten items, nine of which are structured to discern moderate-to-vigorous PA (MVPA) using a 5-point Likert scale with higher scores indicating higher PA levels. The summarized PAQ-C score is the average of the nine items scores. The PAQ-C score was the total score calculated by summing all nine items to indicate participant’s overall PA levels. The validation of Chinese version of the PAQ-C has been documented [43].The stage of behavior change was measured using the Stage of Motivational Readiness Questionnaire [44], which was validated with the Seven-Day Physical Activity Recall [45]. This questionnaire contains four yes/no option items to determine participants’ readiness for PA and categorize them into different stages: pre-contemplation, contemplation, preparation, action, and maintenance. The questionnaire was translated into Chinese using the translation-back translation method and a pilot study was conducted. The Kappa index of reliability of 0.78 over a 2-week period was reported. To compare the movement status of groups related to the stage of motivational readiness, participants were divided into three categories (decreasing, stable and increasing). These category scores were determined by subtracting the pre-test from post-test scores.Perceived benefits and barriers to PA were measured using the adolescent version of the Exercise benefits and barriers scale (EBBS) [44]. The term “exercise” in the original questionnaire is replaced by “PA” in order to reflect the PA context of the adolescents for the present study. The revised questionnaire was conducted to the participants and they showed no troubles to understand and respond to the questions related to their benefits and barriers to PA. This scale contained nine benefit and ten barrier items using a 5-point scale. The scale could also be separately transited into the benefits subscale and barriers subscale. The final score were the PA benefits and barriers scores, which were summed directly based on the benefits subscale and the barriers subscale and benefits/barriers differential score, which were calculated by subtracting the mean barriers score from the mean benefits score. Higher scores indicated greater perceived benefits or barriers or a differential score. A Cronbach’s alpha of 0.84 was reported for both benefits and barriers subscale. The Kappa indexes of the two subscales were 0.72 and 0.76, respectively.The tailored SMS messages were developed in a published pilot study [41]. According to the pilot study, four types of messages were created: (1) greeting (e.g., Hi I’m Jackie, what’s your name? Do you do PA regularly?), (2) motivation (e.g., Congratulations! You met your target last week. Great! I need to learn from you!), (3) exercise tips (e.g., Listen to the music and dance. It’s easy and nice for you.), and (4) reminder (e.g., shopping in and walking in a mall is also an exercise, please do it).The content, language and tone of the SMS motivators were all tailored and matched according to the participant’s Stage of Motivational Readiness (SMR), PA benefit and PA barriers, as measured at the baseline assessment. The messages were written in colloquial language and a stage-appropriate tone. For the pre-contemplator, enabling them to understand the pros and cons of PA is the goal of this stage. The sample statement was “I can release stress and improve my health by doing simple PA like walking.” For the contemplators, the goal was to increase their likelihood to take action. A statement such as “Let’s try……” was used to stimulate the contemplators to think more and encourage them to do PA. The sample statement for those who want to improve his/her outlook (benefit), was “Let’s try to do some PA and improve my body posture and outlook”. Finally, for the participants in the preparation, action and maintenance stage, the goal was to increase their PA level to the recommended level. A directive statement was used to help these participants develop a PA plan. Therefore, “You can” was used in the sample statement for those who considered the distance of the PA venue excessively far (barrier), “If the PA venue is too far, you can stay home and do some sit-ups, rope skipping, and/or stretching. It’s better than doing nothing” (The President’s council on Physical Fitness and Sports, PCPFS, 2003) [44].In addition, participants were told that messages would be sent by PE teachers. This was done to enhance social support because PE teachers are a known motivational influence affecting adolescent’s leisure time PA [46]. A commercial database (PCCW Mobile HK Limited, 2007) was used to record the usage of the SMS (number of SMS sent by the researcher and replied to by the participants).Baseline characteristics of the four intervention groups and the control group were compared using chi-square tests for categorical variables and the analysis of variance (ANOVA) for continuous variables. Analysis of covariance (ANCOVA) was conducted to test the intervention effect on the outcome controlling for age, gender, BMI and the baseline assessment of variable. To compare stage movement among groups, participants were divided into three categories (stable, progression to a higher stage, and regression to a lower stage), and these data were analyzed with chi-square tests. The level of significance was defined as p < 0.05; all tests were 2-tailed. All analyses were conducted using SPSS (version 16.0, SPSS Inc., Chicago, IL USA).Participants’ baseline characteristics stratified by groups are presented in Table 1. Due to the small sample size, there were significant differences in gender (χ2 = 31.83, p < 0.001) and age (χ2 = 14.32, p = 0.006) distribution among groups. No significant differences were found in BMI among the five groups.The PAQ-C score at baseline and post-test in five groups are shown in Table 2. No significant intervention effects were detected among the five groups (F = 1.21, p = 0.421). However, an increasing trend was found in the HST and LAT groups. The PAQ-C score increased from 16.46 (SD: 6.54) at baseline to 18.75 (SD: 10.57) at post-test in the HST group as well as from 19.69 (SD: 8.59) at baseline to 21.85 (SD: 8.68) in the LAT group. However, this score decreased in the other four groups.After intervention, five (38.5%) students in the HST group progressed to a higher stage, seven (53.8%) children remained at the same stage, and one (7.7%) child regressed to a lower stage. In the control group, the number of children that regressed to a lower stage (five children, 31.3%) was higher than those that progressed to a higher stage (three children, 18.8%) (Table 2). However, no significant differences were observed in the stage movement among the five groups (χ2 = 6.18, p = 0.627).The exercise benefits, barriers and benefits/barriers differential scores are shown in Table 3. There were no significant differences among all the comparable groups (p > 0.05). However, we found a growth trend in the exercise benefits score in the LST and HAT groups compared with a downswing in the control group. The exercise barriers score in the HST group showed the largest reduction after intervention. The benefits/barriers differential score in all the intervention groups increased, whereas it decreased in the control group.When the SMS frequency and timing issues are treated separately, it is reasonable to assume that the SMS frequency has a positive relation with adolescents’ PA (Table 4). Orr and colleagues [9] stated that a higher frequency of SMS (multiple messages a day) had a better effect on the health behavior change than a lower frequency. Rosenbaum et al. [47] suggested that the frequency of feedback could be a vital key to learn and/or to change perceptual-motor skills and intellectual skills. In the specific area of SMS dosage, optimal frequency information on SMS delivery in motivational and learning contexts has been lacking. However, this study did not demonstrate that the HST group could significantly increase adolescents’ PA levels, even the results showed the possible trend and meaningful impact of higher frequency and the self-selected timing of SMS on behavioral change. Due to the small sample size in this study, a future study should further explore the higher frequency and self-selected timing in the intervention design. This finding may offer initial support that five weekly, tailored SMS had a greater impact on self-reported PA than three weekly SMS and the control group. This change trend is consistent with the systematic review [40] on behavior change interventions delivered by SMS. The researchers found that the frequency of SMS transmission reflected the frequency of the target behavior for all except for two (twelve of fourteen) studies in the review. Newton et al. [29] also indicated that the low frequency of SMS (one message per week) could be the reason that PA behavior change intervention (12-week SMS and pedometer monitoring) did not produce any significant increase in step counts.When considering the timing with the frequency, the self-selected (HST) and assigned timing (HAT) groups did not demonstrate any statistically significant change in PA levels (as represented by the PAQ-C score). However, an improvement trend was shown in the self-chosen time group when compared with the assigned time group and the control group. Crutzen, Cyr & de Vries [46] stated that, if participants had more control over their internet-based intervention, this may provide a positive effect to change their health behaviors. Tailored SMS are considered to be more effective in PA behavior change than untailored or generic bulk messages [40]. Autonomy of SMS timing could be part of the tailored SMS in which it could fit well into the modern daily life pattern or pace in different cities.No significant change occurred in the stage of motivational readiness over the course of the intervention among the five groups. De Bourdeaudhuij and colleagues [35] suggested that stages of PA behavior change are related to the frequency of encouragement from participants’ family and friends. The researchers revealed that the adolescents in the pre-contemplation stage reported the least social support from family and friends. This finding indicated that more social support and encouragement is needed to advance these physically inactive adolescents to the contemplation, preparation and action stages. In this case, SMS could be a desirable option to reach adolescents using a more convenient and cost-saving approach.Regarding the SMS delivery timing, this finding may imply that the participants did not observe the relevance between the stage of behavior change and the SMS delivery time. Alternatively, certain pre-requisites might need to be met before the timing could be meaningful to the participants. The finding could also relate to the different stages the participants were at. Because no previous research attempted to investigate the impact of the timing of SMS delivery on adolescents’ PA behavior, this question could be further examined in a future study.Despite insignificant results, the benefits/barriers differential score in all the intervention groups increased, although it decreased in the control group. This finding may suggest a potential value of SMS to the adolescents’ PA when considering the exercise benefits and barriers. Additionally, much larger differences were found in HST for perceived barriers when compared with the small decrease in the control group. More specifically, high frequency and self-selected timing had the largest impact on exercise barriers.Grubbs and Carter [48] found a significant association between exercise benefits and barriers to individual’s exercise behavior and suggested that a deeper understanding of exercise benefits and barriers would contribute to the exercise behavior establishment. In the study by Blanchard et al. [49], the researchers suggested that perceived barriers’ efficacy may produce a more significant impact to explain PA behavior compared with task efficacy. Therefore, the possible effect of high frequency and self-selected timing of SMS on exercise barriers deserves further study to examine whether it could produce significant behavior change in a longer term intervention. O’Dea [50] suggested providing more social support, education and information on environment restructuring, planning, time management, motivation and varieties of PA for the school children to overcome the exercise barriers. High frequency and self-selected timing of SMS may be the instruments to help achieve this objective through mHealth technology.To conclude, no significant intervention effects were found among the five groups in PA behavior, stage of change and exercise benefits and barriers among Hong Kong Chinese adolescents.According to a systematic review on the health behavioral change and mode of delivery by e-intervention, Webb and his colleagues [51] suggested that greater use of supplementary modes, such as SMS, email, telephone, CD-ROM, or videoconferencing, could produce a better intervention effect. Another review was conducted to investigate the intervention effect of SMS in health behavioral change studies [52]. The review demonstrated that 29 of 34 SMS studies (85%) found that SMS played a more significant role and produced a positive effect on the participants when combined with internet and other strategies (e.g., telephone and computer program). The results suggested that the combined usage of SMS with other behavioral change strategies would produce more significant and positive effects for future interventions. In this case, more combined usage of SMS and other e-intervention modes of delivery, including computer programs, internet website, and emails, may produce a better effect on PA behavioral change.The findings of this study should be interpreted with caution. First, this study used the self-reported data (PAQ-C) from the adolescents, and these data could be overestimated based upon the self-reported PA literature. Future studies should use the accelerometer to collect the objective PA data of the adolescents; thus, the analysis and findings could be more accurate. Second, it was a small-scaled study (n = 69); therefore, the generalizability of the findings is limited. Third, the relative short (4 week) intervention duration of this study may not allow for full conclusions to be drawn regarding the effects of the frequency and timing of SMS delivery on the adolescents’ PA behavioral change. Therefore, this issue should be addressed in the future. With the changes of PAQ-C in the HST and control groups after and before the intervention, power analysis result showed that the current sample size (n = 13 in the HST group and n = 16 in the control group) only provided the power of 0.71 at p set as 0.05. Due to the small sample size, this study was underpowered to detect a significant treatment effect. Further research should be conducted in a large representative sample in more schools and covering all grades to confirm our findings. Finally, future research may consider qualitative investigations to elaborate on the relations between the stages of change, perceived benefits and barriers, and the different frequency and timing of SMS delivery.P.W.C.L. and E.Y.L. designed the study and collected data. A.J.P., B.W.C.L. and J.-J.W. helped analyze the data. P.W.C.L. and J.-J.W. wrote the manuscript. All the authors reviewed and revised the manuscript.This research received no external funding.The authors declare no conflict of interest.A CONSORT statement of participants’ flow.Participants’ characteristics.Notes: HST, High-frequency + self-selected timing; LST, Low-frequency + self-selected timing; HAT, High-frequency + assigned timing; and LAT, Low-frequency + assigned timing.Participants’ PAQ-C score and stage movement stratified by group and time.Notes: HST, High-frequency + self-selected timing; LST, Low-frequency + self-selected timing; HAT, High-frequency + assigned timing; LAT, Low-frequency + assigned timing; and PAQ-C, Physical activity questionnaire for children.Mean (SD) for measures stratified by group and time.Notes: Data are presented as the mean (SD). HST, High-frequency + self-selected timing; LST, Low-frequency + self-selected timing; HAT, High-frequency + assigned timing; and LAT, Low-frequency + assigned timing.SMS Frequency and PAQ-C score.Notes: * Denotes p value analyzed using the Kruskal-Wallis h test for the post-test/pre-test differential score (calculated by subtracting the pre-test from the post-test score).

Med-MDPI/ijerph_3/ijerph-16-05-00788.txt

@@ -0,0 +1 @@
+Panel studies are an efficient means to assess short-term effects of air pollution and other time-varying environmental exposures. Repeated examinations of volunteers allow for an in-depth analysis of physiological responses supporting the biological interpretation of environmental impacts. Twenty-four healthy students walked for 1 h at a minimum of four separate occasions under each of the following four settings: along a busy road, along a busy road wearing ear plugs, in a park, and in a park but exposed to traffic noise (65 dB) through headphones. Particle mass (PM2.5, PM1), particle number, and noise levels were measured throughout each walk. Lung function and exhaled nitrogen oxide (NO) were measured before, immediately after, 1 h after, and approximately 24 h after each walk. Blood pressure and heart rate variability were measured every 15 min during each walk. Recorded air pollution levels were found to correlate with reduced lung function. The effects were clearly significant for end-expiratory flows and remained visible up to 24 h after exposure. While immediate increases in airway resistance could be interpreted as protective (muscular) responses to particulate air pollution, the persisting effects indicate an induced inflammatory reaction. Noise levels reduced systolic blood pressure and heart rate variability. Maybe due to the small sample size, no effects were visible per specific setting (road vs. park).Acute effects of air pollutants can be examined by time series studies [1,2,3,4,5,6] or panel studies [7,8]. Both approaches correlate temporal changes in exposure markers to temporal changes in health endpoints. The former type of studies focuses on population level effects and usually uses data obtained from health registries (e.g., hospital admission [1,3], outpatient disease events [6], or mortality data [1,2,4,5]). The latter type concentrates on individual participants. The disadvantage of a typically smaller number of participants in this type of studies is often compensated by more detailed information obtained by repeated measurements [9,10,11] or through detailed symptom diaries kept by the participants [12]. The more complicated the physiological measurements, the higher the effort per person. Therefore, panel studies with the requirement of more invasive or demanding examinations often rely on fewer participants. Previous panel studies on health effects of air pollution have relied on numbers of participants ranging from 13 to 163 [12,13,14,15].Bloemsma et al. [7] provide an overview on 25 panel studies on the acute effects of air pollution in patients with chronic obstructive pulmonary disease (COPD), published between 1993 and February 2016. Participant numbers varied between 16 and 459. In studies with fewer participants, repeated lung function tests or analyzed blood samples for inflammatory markers were performed, while the larger studies relied on symptom diaries, reported reasons for consultation of physicians, or reports about limitation in activities.Acute health effects of short-term air pollution episodes appear to be small when compared to the effects of prolonged or chronic exposure to similar concentrations of air pollution [16]. From a public health point of view, chronic exposure and longer averaging periods of exposure to pollution are more relevant than rare peak episodes. Nevertheless, the analysis of health effects of temporal variation in pollution concentration is important as it supports the causal interpretation of epidemiological findings in cohort studies. Confounding factors may differ between acute and chronic effects. If the same endpoints are demonstrated to be affected in both types of studies, this serves as a strong argument against a possible systematic confounding.Panel studies can be enhanced by randomly [17,18] or systematically [19,20,21,22] placing participants in various settings that differ in exposure levels. In that case, either natural settings with different pollution exposures are sought or experimental exposure settings are provided. Natural settings clearly have the benefit of a direct relevance for everyday life. Under realistic conditions, air pollution is not a binary (yes/no) variable, but there is always an exposure of a variable magnitude. Analyzing the effects using a simplified binary (high/low) exposure variable reduces study power, compared to the use of continuous variables where available [23]. Therefore, switching settings may be favorable for augmenting the exposure contrast, but not for solely defining the exposure variable.The effects of acute air pollution have been repeatedly demonstrated by panel studies. In these investigations, COPD patients [9,10,11,13,14,24], asthma patients [12,15,25,26,27,28,29], and patients with metabolic [30] or cardiovascular diseases [31,32] (e.g., patients with dual chamber implantable cardioverter-defibrillators [33] or with hypertension [34]) were examined. Such panel studies have been performed in the elderly [35,36,37,38] and in areas of high air pollution [39,40,41,42,43,44], but less so in young and healthy adults in settings with moderate to low air pollution levels [45,46,47,48,49,50,51]. Patients with respiratory diseases were usually examined for respiratory effects like lung function changes or related symptoms, while patients with cardiovascular and metabolic diseases were checked for cardiovascular and inflammatory parameters. The few studies in healthy adults did investigate very different endpoints, ranging from inflammatory markers [46,49] to heart-rate variability [45] and arterial stiffness [48,49]. Only a few focused on lung function [45,46,50,51], and these applied very different averaging times, ranging from one hour [50,51] to one day or even 2 weeks [46], and investigated different lag periods.Noise pollution is a major concern for the modern world, and in recent years, many studies emerged in order to protect people from its impact. In fact, it is already well-known that constant exposure to noise will lead to health effects, such as sleep disturbance [52,53], annoyance [54,55], cardiovascular effects [56], learning impairment [57,58], and hypertensive ischemic heart disease [59]. Prevention also depends on understanding temporal patterns of the local noise levels, and thus recent applications use wireless sensor networks for noise monitoring [60,61], representing a modern and cheap solution supporting and augmenting the mandatory noise maps and action plans [62]. In outdoor environments, acoustic barriers are the most widespread solution to mitigate the noise produced by the main sources: railway traffic [63,64], airports [65,66], and industrial plants [67,68]. Recent developments in the field are moving attention towards sonic crystals used as acoustic barriers [69].We recruited healthy students from Vienna to study the physiological reactions of both cardiovascular and respiratory systems to everyday urban air pollution exposure. We also controlled for noise exposure to disentangle the effects of the two main exposures from road traffic.The recruitment of healthy students was performed as part of the diploma thesis of five of the authors (J.P., L.U., E.U., B.E., S.P.). Each student approached friends and colleagues and organized the walks, for a total of 3 to 6 participants each (including him- or herself). These group leaders were responsible for the logistics and data collection. According to the principles and rules of diploma thesis projects at the Medical University of Vienna, each group leader had to apply separately for the ethical approval for his/her project. All ethical approvals were granted by the Ethics Committee of the Medical University of Vienna. Prior to the thesis projects, an overall ethical approval was also obtained from the Ethics Committee of the City of Vienna (EK 15-259-VK_NZ, November 19th, 2015).In total, twenty-four healthy students walked at least 4 times for one hour under each of the 4 settings: Along a busy road, along a busy road wearing ear plugs, in a park, and in a park but exposed to traffic noise (65 dB) through on-ear headphones with recorded road traffic noise. The road-walk was along the “Hernalser Gürtel”, a main road in Vienna with 400–500 cars/15 min throughout the day (6 a.m.–7 p.m.). The selected park was the “Augarten”, a large park in the center of Vienna. Walks were planned to be scheduled at fixed times of the day, but due to the time constraints of the participants, this was not always possible.The total mass of particles (PM) with an aerodynamic diameter above 2.5 µm (PM2.5), above 1 µm (PM1), particle number, and noise levels were measured throughout each walk. PM2.5 was measured using a Grimm Portable Laser Aerosol Spectrometer Model 1.108 (Grimm Aerosol Technik, Ainring, Germany). To measure particle numbers (PN), a miniature diffusion size classifier (miniDiSC, http://www.fierz.ch/minidisc/, Institut für Sensorik und Elektronik, Brugg-Windisch, Switzerland) was used. Concentrations were averaged at every 6 s intervals and stored. The noise was measured with a Brühl & Kjaer sound level meter, type 2236 (Brühl & Kjaer, Bremen, Germany). Every 15 min, the equivalent continuous sound level was recorded.Both spirometric lung function and exhaled nitrogen oxide (NO) were measured before, immediately after, one hour after, and approximately 24 h after each walk. Spirometry was performed using an EasyOne™ Spirometer (ndd Medizintechnik AG, Zürich, Switzerland) in an upright standing position and applying a nose clip following standard procedures [70,71]. NO in exhaled air [72] was measured using the portable instrument NObreath™ (Bedfont Technical Instruments Ltd., Harrietsham, UK).Blood pressure and heart rate variability (HRV) were measured every 15 min during each walk. For recording heart-rate variability the mobile ECG device eMotion Faros™ (Biomation, Almonte, Ontario, Canada) was used. ECG-files were analyzed over time windows of 15 min each with the Kubios software version 2.2. (Kubios, Kuopio, Finland)The temperature was obtained from a nearby stationary meteorological station. The data on fine particle (PM10) background concentration were also obtained from a nearby fixed monitor operated by the City of Vienna (Station near the General Hospital: AKH).In panel studies, each participant serves as his or her control. In repeated measurements, time-varying exposures and health indicators are assessed and correlated with each other. Theoretically a study could be performed with a single participant. The power of the study not only depends on the number of participants, but also on the number of observations (time points times participants). With a sufficient number of observations even in a single participant, the effects of exposure can be demonstrated. In this case, it would not be clear if the single participant is representative for a broader population or if, by chance, that participant is for example representative for a highly susceptible subgroup. With multiple participants it is possible to examine if effect estimates differ significantly between them. All statistical calculations were performed with STATA SE Vers. 13.1 (StataCorp LLC, College Station, Texas, USA). We applied the xtreg command that fits regression models to panel data. In particular, xtreg with the fe option fits fixed-effects models; and with the re option, it fits the random effects models by using the Generalized least squares (GLS) estimator. When fixed effect and random effect models provide similar estimates (checked with the Hausman test), significant variation in susceptibility can be ruled out.The date of respiratory and cardiovascular markers were recorded in two separate files for further statistical analysis. Respiratory parameters (immediately, 1 h and approximately 24 h after the exposure) were assessed for each participant with single air pollution markers serving as independent variables in a random effects GLS regression. Fixed effect models generally provided very similar results, although the Hausman test was significant in few but not in most models. The random effects option is the default setting in STATA and does provide broader confidence intervals than the fixed effects model. Therefore, for the sake of internal coherency, random effects were applied in all models. In two adjusted models, either the same respiratory marker or NO in exhaled air before the exposure were included to control for unmeasured influences prior to the walk.Cardiovascular markers were assessed every 15 min separately and applied to a random effects GLS regression. Exposures in the preceding 15 min (noise and dust) served as independent variables. The temperature was included in the models as a confounder.In a first run from December 2016 to May 2017, 20 students (11 male, 9 female) with an average age of 24 years (range 21–33) walked on average 12 times for one hour (range 8–20). All of them were non-smokers and reportedly healthy. In a second run, four additional (female) students were recruited, who performed their walks in May and June 2018.Air pollution and ambient noise were determined to be higher near the street but were temporally not correlated with each other. Individual exposure to noise was further de-coupled from ambient conditions by design (ear-plugs and head-phones respectively). Particle measures were found to be well correlated (R-values for all particle mass measures including PM10 at nearby fixed monitoring station >0.9). Personal exposure concentrations were actually found to be higher than concentrations at the fixed monitoring station, either indicating that the Aerosol Spectrometer overestimated particle mass concentrations systematically, or demonstrating higher personal exposure compared to the fixed station. As an example, hourly values of PM2.5 measured with the spectrometer and PM10 measured at the fixed station displayed a correlation coefficient of 0.96. A linear regression model with PM10 at the fixed station and the setting (road versus park) explained 92.5% of the variation of PM2.5. In this model, the difference between road and park was small (4.2 µg/m3) but significant (p = 0.016). The slope of the regression line (ß of PM10 at the fixed station) was 1.57 (p < 0.001). Table 1 describes the range of exposure for PM10 at the fixed station and the personal exposure measured as PM2.5, PM1, and PN. Because of the high correlation between the particle mass values, only PM1 from the personal monitoring and PM10 from the fixed site (controlling for setting—road vs. park) were further analyzed for health effects. In addition, effects of personal particle number concentrations (R with mass concentrations between 0.72 and 0.77) were investigated.After controlling for seasonal and daily trends, ambient noise levels at the road were approximately 10 dB louder than in the park. These ambient noise levels were partly overruled by earplugs and/or headphones. The average sound pressure level LA,eq was ~56 dB at the road. At the road wearing earplugs, noise was assumed to be 30 dB lower. In the park it was measured at about 46 dB and the headphones were set to 65 dB.Higher air pollution levels were found to be correlated with reduced lung function parameters. Measures of large airway resistance, especially peak-flow (PEF), were in the majority not significantly affected by air particle mass or count, while measures indicative of the small airways like mid-expiratory flow at 25% lung volume (MEF25) showed more consistent effects and remained low even 24 h after exposure. Figure 1 provides two exemplary figures for the generally observed pattern. A more complete overview of the results is provided in Appendix A (Table A1). Measured PM levels (either from the personal monitor or from the fixed site) had a stronger and clearer effect than the setting (road vs. park) and the effect of the setting (“road”) usually lasted less long than the effect of the pollutant concentration. This was expected, because the impact of the setting was restricted to a single hour, while the measured pollution concentration was fairly representative for the whole day and for the whole area.Exhaled NO levels were found to be significantly reduced immediately after and 1 after the walk with increased dust levels. After controlling for background dust levels, walking besides the road increased exhaled NO levels measured 24 h later (Figure 2). The background reflects the average exposure at that time for a longer period (and thus also affects NO before the setting) while the setting may add to the background exposure for a short and defined period of 1 h. Therefore the effects of pollutant concentration (Figure 2 depicts particle number as an example) were controlled for pre-walk NO levels.Noise levels were found to correlate with reduced blood pressure (stronger effects for systolic than for diastolic blood pressure) and lower heart rate variability. The effects on heart rate variability generally attenuated in the course of the walk, indicating some adaptive process. The temperature also had clear effects on cardiovascular parameters, but did not confound noise effects. Air pollution effects were less pronounced and not very consistent.Again, only examples of the recorded data are visually presented here (Figure 3) that are typical and representative for effects on other parameters as well. For a more complete overview of the results the reader is referred to Appendix A (Table A2).Total exhaled NO originates from different exogenous and endogenous sources [73,74,75]. NO is secreted by epithelial cells inducing relaxation of the smooth muscle cells of the bronchial walls. This secretion is inducible and a reaction after a reflective muscular narrowing of the airways: An irritant stimulus will at first lead to increased muscle tone and thus after a very short interval to an increased airway resistance. The increased tonus of airway smooth muscles is then typically quickly antagonized by an increased cellular level of NO, which thus induces a lower airway resistance. NO also serves as messenger molecule in inflammatory processes. Here, it is mostly involved in eosinophilic inflammation and signifies allergic asthma (and thus an increased bronchial reactivity as well as a likely increased airway resistance). Epithelial cells capable of producing NO may be compromised during (neutrophil) inflammation and by some toxic substances (e.g., in cigarette smoke). Therefore, smokers generally have lower NO levels, and after the smoking of cigarettes, NO is further reduced [76,77,78].NO might relax smooth muscles and thus reduces airway resistance. However, NO in the tissue may also act as an oxidative stressor and thus contributes to an inflammatory response after it is induced by a reflective muscular contraction elicited by an irritant. Therefore, after a certain delay, higher NO levels might predict increased airway resistance again.These different complex pathways make the interpretation of NO in a panel study a challenge. In this context, timing is essential and even with our protocol of repeated measures (before, immediately after, 1 h after, and 24 h after a defined exposure of one hour) we were not fully able to capture the complete time course. Considering both background pollution levels that are likely to represent the longer term average exposure and the effects per setting which by design lasted only for one hour, we could further disentangle the temporal variation. In this endeavor we were hindered by the finding that differences between settings were small compared to variation in background exposure over time.Nevertheless, we were able to demonstrate reduced NO levels due to higher background exposures, indicative of toxic damage to epithelial cells, as well as increased levels of NO with 24 h latency after a 1 h acute exposure, indicative of inflammatory response. A reduced lung function led to an increase in NO levels, and higher NO levels were indicative of increased lung function levels at the next measurement point (data not shown).We could not show consistent effects of the specific setting (road vs. park) on lung function. This could be due to a lack of power because a binary exposure variable provides much less information than a continuous one. We had designed our study following the London study [19] where a panel of 60 asthmatics walked for 2 h, alternatingly along Oxford Street or through the nearby Hyde Park. In that study, significant differences between the two settings could be demonstrated. Asthmatics might be more susceptible to air pollution than healthy adults. The London study examined substantially more participants. The authors also point out that Oxford Street is a street canyon with very high diesel traffic. In Vienna, the particle concentration did differ between the two settings. However, especially regarding particle mass, the daily variation was much more pronounced than differences between the settings. This might be another reason we did not observe consistent effects of the setting.Unexpectedly, the only clear and significant effect of the setting was a higher peak flow (PEF) after the walk on the road. The peak flow is sensitive of the participant’s cooperation and muscular efforts. Keeping in mind that the participants were not blinded to the setting, the higher PEF values after the walk along the road could be due to the participants’ subconscious ambition to try harder.Also with measured air pollutants effects on the larger airways, as signified by PEF, the expected indicative of a muscular reflex response could not be demonstrated consistently. Effects on the small airways signified (e.g., by MEF25) persisted for 24 h and likely represent inflammatory responses that might be clinically more relevant.Lung function values mostly remained in the normal physiological range. This is not surprising, considering the comparatively low exposures and the generally healthy state of the participants. We observed effects during everyday activities in everyday settings and therefore did not expect to observe severe detrimental health effects. However, it seems noteworthy that even under these conditions and even in young healthy adults with a comparatively small number of subjects and a moderate number of repeated observations, several relevant effects were reached. While acute effects shortly after the exposure are likely due to physiological protective reactions under nervous control, the longer term changes, as most clearly seen in the end expiratory flows, are indicative of an increase in resistance in the small airways and are likely caused by inflammatory tissue responses. This is a cause of concern, as even small effects, when cumulated over the years, will in the long run hasten the functional decline of the respiratory system.Cardiovascular effects were mostly observed in relation to noise levels. These effects remained even after controlling for temperature, which was the most important predictor of heart rate, heart rate variability, and blood pressure. Also these effects remained in the physiological range, which again was to be expected. Even subtle effects to everyday exposures during everyday activities still can be of concern in the long run.Health effects of environmental noise exposure are well established [79,80,81,82,83,84,85,86]. However, effects of noise on annoyance are usually assessed through cross-sectional studies [83] and effects on cardiovascular health through cohort and case-control studies [86]. Acute reversible physiological effects of noise have not been investigated that much lately. Effects of much higher occupational noise levels have been studied in relation to sleep [87] stress hormones [88,89,90], or blood pressure [91]. However, studies on acute effects of environmental noise exposure are rare [92]. The latter study was similar to ours as it also examined the effects of air pollution and noise in a panel of healthy young adults (33 men and 33 women). Air quality was assessed based on available stationary data while noise exposure was measured by personal monitoring. This study investigated blood pressure as the only outcome. Similar to our study air pollution led to higher blood pressure but contrary to our findings also noise increased blood pressure.Other acute effects of air pollution have already been demonstrated in some panel studies, notably on glucose metabolism [93,94] and inflammation [95]. However, we are confident that we have covered the most important endpoints, including respiratory [96] and cardiovascular [92,97] effects. The observed respiratory effects of this study are generally in line with those reported in the literature. The two studies most similar to our approach [50,51] investigated a small group of cyclists (15 and 12 respectively) riding on low- and high-traffic intensity routes. Both studies examined lung function only by comparing settings. The former did not find significant differences while the latter unexpectedly found a better lung function immediately after the ride on the high-intensity route, but poorer lung function 6 h later. Their results are reflected by our study, as we also failed to find consistent effects per setting and we could also demonstrate lasting effects.Cardiovascular effects of air pollutants are typically reported in patients and not in healthy young adults. Only one study [47] examined HRV in healthy adults, but this study monitored HRV (over 5 min intervals) during sleep only. A clearly vegetative regulation of cardiovascular function will differ between sleeping and waking hours. In our study, the only observed effects of air pollution consisted of a small increase in blood pressure.Our study design did not allow for the examination of lag periods for cardiovascular effects. This might also explain why cardiovascular effects were predominantly seen with noise and not with air pollution. Traffic noise is considered an important environmental stressor. Because of this, we would have expected a positive association between sound pressure level and blood pressure. Interestingly, the reverse was seen in our study while the reduction in HRV confirms our hypothesis.Exposures to air pollutants cause statistically significant respiratory reactions, even in healthy young adults. These effects were visible even with a relatively small number of participants and only few repeated measurements. The effects were reversible and generally not very severe, but nevertheless clear and consistent. This was especially true for measured air pollutants that were not only representative of the exposure of the single hour of the walk at the road or in the park, but also of the general air quality on that day and in the whole region. Noise had a clear effect on most cardiovascular parameters. For most of the endpoints, the noise effects displayed attenuation over the course of the walk. Contrary to expectations, higher noise levels lead to lower blood pressure.We cannot fully exclude the possibility that the latter unexpected findings are due to a lack of study power. Indeed, in a first analysis of the incomplete data set, noise effects on blood pressure were even stronger. However, unmeasured confounding seems to be a more likely explanation, although controlling for temperature as a surrogate of seasonal and diurnal variation did not strongly affect our findings. In addition, we were not able to show effects of setting (road compared to park), and this is likely due to poor power because of the small sample size.Epidemiological studies often focus on easy-to-measure lung function parameters like FVC, FEV1, FEV1/FVC, or PEF. However, higher resistance in the small airways is better reflected by end expiratory flows (MEF50, MEF25). These measures showed the clearest results and the longest lasting effects. While immediate reductions in lung function could be interpreted as signs of protective reflexes of the bronchial muscles, prolonged reductions in the flow values might be due to an inflammatory swelling of the mucosa, and thus would be clearly an indicator of an adverse effect.H.M. designed the study and wrote the paper. The other authors (J.P, L.U., E.U., B.E., and S.P.) carried out the fieldwork and collected the data.This research received no external funding.The authors declare no conflict of interest.For a list of abbreviations and their explanation please refer to Table A3!Overall, the study produced a lot of data. For example, regarding air pollution effects on lung function (Table A1), personal particle mass and the number as well as stationary particle mass data were examined in relation to seven lung function parameters at three points after each walk. Effect estimates could be displayed from the unadjusted model or adjusted for pre-walk conditions, either controlling for pre-walk lung-function level or exhaled NO. As demonstrated with the examples depicted in Figure 1, the adjustment for pre-walk conditions did not have a strong effect on the estimates. Therefore, in Table A1, only unadjusted estimates are presented for each lung function parameter, for each time point post-walk, for personal PM1 and particle number, and for PM10 from the central monitor. Because the individual exposure is affected by both the regional pollution level (measured at the central monitor) and the setting (road vs. park), the model with PM10 from the central monitor also setting is included.Unadjusted effect estimates (p-value): lung function change per 1 µg/m³ (personal PM1, central monitor PM10), per 1000 particles per cm³, and road vs. park. The latter independent parameter is combined in one model with PM10 from the central monitor. (bold: p < 0.05).Similarly, several cardiovascular parameters were assessed four times during each 1 h walk. In this regard, in addition to air pollution, noise exposure also acted as a determinant. Table A2 summarizes the findings. The temperature was only weakly correlated with noise (R between 0.25 and 0.30 for the 4 quarters of the walks) and therefore did not confound noise effects in a meaningful way. The temperature was strongly negatively correlated with particle mass (PM2.5) with R between −0.6 and −0.7. Controlling for temperature mostly slightly increased and strengthened effect estimates for particles. For consistency reasons, only unadjusted results are shown.Unadjusted effect estimates (p-value): change in cardiovascular parameters per 1 µg/m³ (personal PM2.5), per 1000 particles per cm³, and per dB. (bold: p < 0.05).The abbreviations used in Table A1 and Table A2 are explained in Table A3. The explanations for the HRV parameters follow the paper by Schaffer and Ginsberg [98].Abbreviations.* Power parameters of the HRV frequency domain are measured in ms².Examples for air pollution effects on MEF50* (in L/s): (a) Effect of the setting “road”, controlled for background pollution; (b) Effect of PM1 (µg/m³). Each triplet stands for: unadjusted, adjusted for lung function before exposure, adjusted for exhaled nitrogen oxide (NO) before exposure. The time points are (left to right): Before exposure (one value only), immediately after the walk, 1 h after the walk, and 24 h after the walk. * For Abbreviations please see Table A3!Effect of exposure on exhaled nitrogen oxide (NO) values (in ppb) at different time points (immediately after the walk, 1 h after the walk, and 24 h after the walk): (a) Exposure near the road compared to a park, controlled for background concentration; (b) Effect of particle number (per 1000 particles per cm³) controlled for setting.Effect of noise on cardiovascular parameters: (a) systolic blood pressure; (b) Standard deviation of intervals between two consecutive heartbeats. Noise levels are averaged over 15 min (4 periods per 1 h walk) each. Blood pressure was measured at the end of each period; electrocardiogram readings were taken over the whole 15 min period.Exposure to particles during walks.

Med-MDPI/ijerph_3/ijerph-16-05-00789.txt

@@ -0,0 +1 @@
+The mounting mental health issues faced by elderly urban residents increase the social and economic costs to society associated with dementia and depression. Therefore, it is necessary to identify the characteristics of elderly urban residents suffering from mental health issues, to address these issues more effectively. We used 2015 Community Health Survey data from the Korea Centers for Disease Control and Prevention to identify the demographic and social characteristics of 11,408 elderly urban residents in relation to stress levels and symptoms of depression in seven metropolitan areas in Korea, and to calculate the odds ratio for urban green space. We found that the prevalence of these mental health issues generally decreased in relation to the ratio of green space of an area. These findings suggest identifying elderly people who are vulnerable to certain mental health issues based on demographic and social characteristics and demonstrate that the ratio of urban green space within a community is an important component in improving mental health outcomes for elderly urban residents. These findings have policy implications for assisting elderly people vulnerable to certain mental health issues and for establishing a green welfare policy targeting this population.Increased life expectancy and falling fertility rates are accelerating the aging of populations, and the worldwide population aged over 65 is expected to increase from 524 million in 2010 to nearly 1.5 billion, representing 16% of the world’s population, in 2050 [1]. In Germany, the population aged 65 years or older is expected to increase from 20.7% in 2009 to 29% in 2030 and to 31% in 2050 [1]. However, the fastest increase in the aging population has occurred in East Asia [1]. Korea began to be considered an aging society in 2000, when 7.2% of its total population was aged over 65 years of age; in 2017, this percentage had reached 14.0% (7,257,288 people as at the end of October 2017) and is expected to increase to 24.5% in 2030 and to 28.7% in 2035 [2]. This shifting age structure requires more targeted forms of health and social care and changes in national infrastructures, particularly in relation to healthcare systems [3].With the development of urbanization in Korea, 91.8% of the population lived in cities in 2017 [4], the percentage of the elderly population living in urban areas rose from 56.4% in 1994 to 76.6% in 2014 [2]. Urban environments often lack access to green spaces due to the proliferation and density of buildings, and urban residents have been found to be more vulnerable to mental health issues such as stress and depression [5,6], due to deterioration in their social and economic status, as well as due to physical illness [1]. Mental health issues among elderly people can also exacerbate dementia [7,8] and increase the suicide rate [9], resulting in an increase in the social and economic costs to society [1,10]. Given that mental health issues occurring among elderly urban residents have implications for society extending beyond the individual level, it is even more necessary to focus on prevention and effective solutions.Urban green spaces have been shown to provide various health benefits. Recent studies have reported a relationship between mental health and parks and green spaces at the neighborhood level [9,11,12,13,14,15]. Mental health issues may arise because of genetic factors or the psychological state of the individual [16] but can be exacerbated by social and economic inequalities and the state of the surrounding environment [17,18]. Parks and green space can help improve mental health through encouraging physical activity [12,19,20], social interaction, and contact with nature while reducing psychological stress [9,21,22,23]. In addition, it has been reported that people living in environments with greenery, which provides physical, social, and economic benefits to residents, enjoy better mental health than those who do not have access to green space [24].Previous studies on this topic have mostly used demographic data from people of all ages; thus, little research on the characteristics of elderly households is available. Furthermore, research that is available is limited to national units or specific geographical areas that do not consider environmental characteristics such as urban and rural areas. However, it is important to identify the characteristics of elderly urban households (e.g., single-member household, elderly single-generational household and multi-generational family household) and factors (e.g., health status and health behaviors) affecting mental health to implement effective policies related to elderly urban residents in aging societies. This study thus addresses various gaps in the literature through examining the mental health of the urban elderly and the effects of demographic characteristics and access to urban green space.The purpose of this study was to investigate the demographic and social characteristics of the elderly who are vulnerable to mental health problems, such as depression and stress, in seven metropolitan areas in Korea. We also investigated the prevalence of mental health problems related to urban green spaces based on the findings of previous studies that urban parks and green spaces are associated with mental health benefits. Furthermore, this study is intended to assist in mental health policy-making and to promote mental health among the growing urban elderly population, and offer practical suggestions concerning welfare policy in relation to green space.We used 2015 Community Health Survey (CHS) data from the Korea Centers for Disease Control and Prevention (KCDC) [25]. The purpose of this survey is to assess the health status, health behavior, and health determinants of Koreans and produce community-based health statistics. The CHS is a survey conducted by trained researchers who visit households sampled nationwide using a multi-stage stratified cluster sampling procedure. The CHS’s target population comprises adults over 19 years of age living in their communities. It is a cross-sectional survey, where participants are sampled each year. CHS data is available to researchers upon request through the KCDC’s online site. Raw data has been provided annually, both at the national and regional levels (one national dataset, and 17 city and province datasets), since 2008. The data are available in text formats. Collected data include sociodemographic information, health behaviors (e.g., smoking, alcohol consumption, physical activity), health status, and subjective health indicators [9,26].We selected seven major Korean metropolitan cities to study the urban elderly population. Detailed geographic information derived from these locations is shown in Figure 1. Among the 11,720 participants aged 65 years and older included from the target areas, questionnaires with missing values among the survey variables were excluded from the analysis. In total, 11,408 people were finally included in the study. This study was approved for exemption by the Institutional Review Board of our university (IRB No. E1902/002-002).Sociodemographic variables were selected based on previous studies [9,11,14,26], including sex, age, educational level, labor market participation, being a basic livelihood social security recipient, monthly household income, household type, comorbidity, physical activity, cigarette smoking, alcohol consumption, and participation in social activity. The participants were divided according to age groups, as follows: those aged 65–69 years old, those aged 70–79 years old, and those aged 80 years old or older. Educational levels were classified as having completed primary school, middle school or high school, college or university, or having graduated with a Master or higher degree. Participation in the labor market was defined using two answer options (yes, no) to the question “Have you worked in the past week for income purposes?” Participants on temporary vacation were considered labor market participants. Comorbidity was categorized as a disease condition if a participant had been diagnosed by a doctor and was receiving treatment at the time of the survey for two or more diseases. Household type was classified into single-member households, elderly single-generational households (i.e., married couples), and multi-generational family households with children and grandchildren. Physical activity was defined as the number of days in which more than ten minutes of occupational activity or physical activity, such as moderate physical activities (e.g., slow swimming, doubles tennis, volleyball, badminton or table tennis) or athletic activities (e.g., running (jogging), climbing, cycling, rapid swimming or jumping rope) was undertaken during the preceding week. For cigarette smoking, we categorized the participants into current smokers, former smokers and those who had never smoked. In terms of alcohol consumption, participants were classified as alcohol consumers if they had consumed alcohol within the preceding year and non-alcohol consumers otherwise. Participation in social activity was categorized as regular participation in social activity (yes, no) at least once a month.The response variables for mental health included subjective stress levels and symptoms of depression. A person’s stress level was classified as high when their answer was “I feel stressed very often or a lot” to the question “How often do you feel stress every day?” However, the answer “I feel stressed a little or rarely” was classified as low. A person was classified as having exhibited symptoms of depression using two response options (yes, no) to the question “Have you ever felt sad or desperate for more than two weeks in the last year?” If the participants answered “yes”, they were classified as having had symptoms of depression.We used the proportion of urban green area per administrative area derived from CHS data to assess the degree of exposure to green space. Urban green areas include parks and green space (roadsides, road and riverside greenery, small parks, children’s parks, neighborhood parks, theme parks, amusement parks, and green spaces excluding cemeteries), which require little money or time to visit and which are easy to access and use in daily life. Therefore, we selected an existing variable of the ratio of urban green area per administrative area, including park area, to determine the likely amount of exposure to the green environment. We divided them into quartiles with natural breaks to compare the odds ratios of stress levels and symptoms of depression related to the proportion of green areas in various cities (Figure 1). The first quartile refers to the lowest proportion of green area and the fourth quartile refers to the highest proportion of green area per administrative area. In addition, interaction terms between the proportion of green areas and physical and social activities were generated and used as explanatory variables to identify the potential effects of subjective behaviors.We conducted a frequency analysis and a chi-square test using PASW Statistics 18.0 (SPSS Inc, Chicago, IL, USA) to analyze the demographic and sociological characteristics of the study sample. We conducted a binary logistic regression analysis, with reported symptoms of depression and stress levels as response variables for mental health indicators. We calculated odds ratios according to differences in relation to sociodemographic characteristics and urban green area.Table 1 shows the sociodemographic characteristics of the study sample by sex. A total of 11,408 people (4922 [43.1%] men and 6486 [56.9%] women) participated in this study, with those aged 65–69 comprising 3865 people (33.9%), those aged 70–79 comprising 5789 (50.7%) people, and those aged 80 or over comprising 1754 (15.4%) people. For household type, 2891 (58.7%) men and 2167 (33.4%) women reported living in elderly single-generational households, and 1608 men (32.7%) and 2587 (39.9%) women reported living in multi-generational family households. Single-member households were more common among women (1732 [26.7%]) than men (423 ([8.6%]). Furthermore, men were more educated than women (x2 = 1736.241, p < 0.0001). More women (518 [8.0%]) than men (252 [5.1%]) were social security recipients. Most women (557 [82.6%]) did not participate in the labor market, while a substantial number (143 [35.4%]) of men did. The most commonly reported disease was hypertension (total 558 [50.5%]), and the rate of arthritis was much higher for women (1661 [25.6%]) than for men (338 [6.9%]). Many more men (947 [19.2%]) than women (142 [2.2%]) were cigarette smokers. More men (3245 [65.9%]) than women (2216 [34.2%]) were consumers of alcohol. Most participants did not engage in physical activity (8335 [73.1%]). However, men had a higher rate of physical activity (x2 = 85.0, p < 0.0001) and regular social activity (x2 = 74.636, p < 0.0001) than women. Finally, 19.3% (2203) of all respondents reported high levels of stress, and 8.0% reported having experienced symptoms of depression.Table 2 shows the sociodemographic characteristics of the sampled population in relation to stress levels and symptoms of depression. The indicators for having a high risk of stress were: being female, being between the ages of 65 and 69, living in a multi-generational family household, having had a lower level of education, receiving social security payments, having a lower income, comorbidity, participating in the labor market, being a cigarette smoker, engaging in less than three days of physical activity per week, and not participating in regular social activity. The indicators for having symptoms of depression were very similar to the high stress level indicators but had slightly different characteristics. Multi-generational family households reported higher symptoms of depression than single-generational households. All variables except education level and alcohol consumption were statistically significant (p < 0.05) for participants who reported high stress levels and symptoms of depression.Table 3 shows the OR (95% CI) of stress levels and symptoms of depression among the sample population in quartiles according to the urban green area ratio. The fourth quartile was the respective reference category for the response variables. In Models 1 and 2, where the potential variables were adjusted relative to the unadjusted model, the OR for both stress levels and symptoms of depression tended to increase as the ratio of the green area decreased; from the fourth quartile with the highest green area ratio to the first quartile with the lowest green area ratio. After complete adjustment (Model 2a), prevalence of stress levels increased by 2.2% (OR: 1.022, CI: 0.892–1.171) for participants in the third quartile and by 18.3% (OR: 1.183, CI: 1.034–1.353) for participants in the second quartile compared to those in the fourth quartile, the highest green area ratio.In the case of symptoms of depression, there was a 26.9% (OR: 1.269, CI: 1.056–1.541) increase of participants in the third quartile and a 28.0% (OR: 1.280, CI: 1.047–1.540) increase of participants in the second quartile compared to the fourth quartile. However, both stress levels and symptoms of depression scarcely appeared in the first quartile, the lowest green area ratio. Apart from the first quartile, there was a clear tendency for stress levels and symptoms of depression to increase as the urban green area ratio decreased (p < 0.005).In this study, we first examined the relationship between stress levels and symptoms of depression and the sociodemographic characteristics of elderly urban residents. Elderly women were more likely to experience higher levels of stress and symptoms of depression than elderly men. Some studies suggest that the lower economic status of elderly women compared to elderly men might negatively affect their mental health [23,27]. Most elderly Korean women have been housewives, with no independent income. It has been reported that fewer opportunities for labor market participation can limit the possibility of forming social relationships, which can lead to depression [27]. Those elderly people who do participate in the labor market experienced more stress but were less likely to experience symptoms of depression. These results suggest that encouraging labor market participation as opportunities (e.g., volunteer, join social groups, etc.) for forming social networks among the elderly could be a promising way to reduce the occurrence of depression among this population. However, the stress caused by labor market participation would need to be mitigated through improvements in the working environment for elderly working people. Interestingly, multi-generational family households were found to be more stressful and more linked to symptoms of depression for elderly persons than single-generational households. Within multi-generational families, there is a greater likelihood of differences in political views or economic power among family members, and these differences might cause inter-generational conflict [28,29]. This finding suggests that comprehensive household welfare policies and services are needed to help families live together in multi-generational homes, and to provide support to families where needed to counter the stresses that can occur in such living arrangements.Current cigarette smokers were more likely to experience higher levels of stress and symptoms of depression. Because this study used a cross-sectional design, these results cannot be interpreted as causal relationships. However, current cigarette smokers may be more likely to experience mental health challenges, as there is a strong association between mental health and personal health behavior, such as smoking [9,26,30,31]. On the other hand, non-consumers of alcohol in our study reported higher rates of stress and symptoms of depression compared to previous studies suggesting that drinking alcohol is associated with poorer mental health outcomes [26,32,33,34]. In Korea, drinking alcohol among elderly people, but not heavy drinking of alcohol, has often been culturally regarded as a social lubricant for communication and alcoholic drinks are served at various social and family functions [35,36]. In this study, it was found that moderate drinking among elderly people had a positive effect on stress relief and in reducing the symptoms of depression. However, the criterion for alcohol consumption in this study did not concern how much was consumed but whether alcohol had been consumed in the preceding year. Although no relationship was found between stress levels and symptoms of depression in relation to physical activity, regular social activity was shown to be positively linked to reducing stress and symptoms of depression [9,26,27,37,38]. Previous studies have reported that social engagement at the neighborhood level could reduce the likelihood of depression caused by air pollution [39] and urbanization [40]. These results suggest the necessity of supporting and increasing opportunities to participate in social activity for the elderly. Overall, the results of this study provide further evidence concerning the importance of identifying relevant sociodemographic characteristics that are likely to affect mental health among elderly urban residents.In recent years, welfare policies for elderly people have been focused on promoting elderly-friendly environments involving increased opportunities for contact with nature [1,41,42]. Many studies have suggested that there is a positive association between providing green environments in urban areas and mental health benefits [5,43,44,45]. The physical environment affects mental health [44], and environmental improvement at the neighborhood level can contribute to the social integration of elderly people [45,46,47]. Previous studies have assessed the effects on mental health of the density or presence of green space at the neighborhood level, through using the Normalized Difference Vegetation Index (NDVI), a method of measuring the composition of residential greenness [48,49,50]. However, this study was unable to apply these methods to determine relevant environmental indicators because of limited data sources. Therefore, the ratio of urban green area including parks and all open space green areas was used as a quantitative exposure index to determine the extent of the green environment within the urban areas covered in this study.In this study, no significant relationship was found with stress levels and symptoms of depression in the quartile with the smallest urban green area ratio. Nevertheless, we found that the higher the rate of greenery in a city, the less stress and fewer symptoms of depression reported among its elderly residents. This result supports previous cross-sectional design studies reporting that exposure to the green environment in urban areas has positive relation to mental health among elderly people [51,52]. In addition, because of the interaction effects of the green areas and physical and social activities which may be associated with mental health outcomes were not significant, this study could still identify clear mental health benefits according to the extent of exposure to greener environments in cities.We assessed factors affecting the mental health of elderly urban residents in relation to stress levels and symptoms of depression, focusing on the effects of specific sociodemographic factors and varying exposure to green areas. Our results provide substantial empirical data on these aspects within Korea, with policy implications given the uneven extent of certain mental health issues among elderly people. The mental health issues identified in this study could be more effectively targeted with better information concerning the relevant individual characteristics involved and through extending the green environment. Therefore, welfare policies should be implemented for vulnerable groups in terms of mental health, based on relevant data and a commitment to establishing or extending green space to ensure a wider exposure. However, it is not easy to increase the amount of greenery within urban areas in a short period. As an alternative, we propose the development of nature-based activities that utilize urban parks or gardens in order to expose elderly people to green environments in their daily lives. For example, forest bathing (therapy) program is one of the nature-based activities that promote physical and mental health using various elements of forest environment such as landscape, sound, phytoncide and anion [53]. The mental health of elderly people is more related to social activity than physical activity, based on the results of this study. Forest bathing programs provide elderly people with an opportunity to casually form social groups in nature [53] and help them facilitate connectedness with self, neighbors and nature [54]. In addition, resting and engaging in nature reduce fatigue and stress [55] and social contact within nature in parks or gardens boosts health and well-being [56,57]. As levels of association increase within a group, people have been reported to become more open with one another [58]. As a result, positive psychological changes can begin to occur through mutual understanding and interests when social interactions are formed in a green environment [37]. Therefore, for green welfare policies that focus on vulnerable individuals, utilizing urban green spaces is likely to assist in reducing depression and stress for elderly people who lack the opportunities for physical activity and social interaction [53], and also help to reduce the financial burden of healthcare on governments. However, it should be noted that such a suggestion is beyond the scope of this study.This study has some limitations. First, many studies on the relationship between exposure to greenery and health have assessed the level of greenery using NDVI through setting neighborhood units linked to participants’ addresses. In this study, we obtained a green exposure level for the participants using the administrative level of a district (“Gu” in Korean), but addresses could not be obtained. Therefore, it was not possible to precisely define green exposure in terms of individual neighborhood units. In addition, the random effects that could occur at various green levels were not controlled because we reflected the green areas of the administrative districts to the participants. However, the reliability of this study was enhanced through the large sample size used and the quality of the official administrative data collected and released annually. Second, the quality of the green spaces involved was not investigated. However, considering the seasonal characteristics of plants in Korea, and that the CHS is usually carried out from August to October, outcomes in relation to the mental health issues focused on in this study were deemed unlikely to be affected by the seasons and state of the greenery, because the quality of greenery undergoes little change during the survey period. Third, the sample was limited to the residents of seven metropolitan cities. Although populations in all cities were not included, the reliability of the sample was increased through selecting the cities with the largest populations. Fourth, the amount of time participants spent in their residential areas was not considered. If the participants had spent more time in other areas, this could have led to misclassification, due to diverse exposure levels to a green environment with differing durations of exposure. Fifth, one question was used to assess depression among the response variables for mental health outcomes, and only two response options (yes, no) were used. In addition, it is possible that more biased response results were obtained because of the subjective nature of the responses, rather than what might have been obtained using medical diagnoses.This study investigated certain sociodemographic characteristics of elderly urban residents, such as their socioeconomic status and health behavior, and the effect of exposure to urban green area on mental health in relation to stress and symptoms of depression. We have identified the characteristics of urban elderly people who are vulnerable to mental health issues and found that the proportion of green areas within a community is an important component in improving their mental health outcomes. Therefore, to ensure ongoing improvements in mental health and maintain the mental health of elderly urban residents, it is critically important to give more attention to identifying vulnerable elderly groups and to either construct new urban green spaces or develop suitable nature-based activities that utilize existing resources.H.J.L. conceptualized the study and further developed the study’s design with D.K.L. H.J.L. conducted data collection and analysis and drafted the manuscript. D.K.L. revised the manuscript.This research is supported by Korea Ministry of Environment (MOE, Project No. E416-00021-0604-0) as “Public Technology Program based on Environmental Policy”.Data used in this study were derived from the Korea Centers for Disease Control and Prevention (KCDC).The authors declare no conflict of interest.Prevalence of self-rated experiences of stress and/or depression among Koreans aged 65 and over within quartile divisions of urban green areas in seven Korean cities.Descriptive characteristics of the study sample: 2015 Korean community health survey (n = 11,408).Odds ratios (95% Confidence Interval (CI)) of self-rated stress levels and symptoms of depression in relation to sociodemographic characteristics: 2015 Korean community health survey (n = 11,408).* p < 0.05; ** p < 0.001.Odds ratios (95% CI) of self-rated stress levels and symptoms of depression in quartiles of urban green area ratio.Model 1: adjusted for demographic factors (i.e., sex, age, household type, education, monthly income, financial aid, labor market participation, and comorbidity prevalence); Model 2 (a): adjusted for model 1 + individual behavioral factors (i.e., cigarette smoking, alcohol consumption, moderate physical activity, and regular social activity); Model 2 (b): adjusted for model 2 (a) + interaction of the urban green area ratio with physical activity and regular social activity; * p < 0.01, ** p < 0.005, *** p < 0.001.

Med-MDPI/ijerph_3/ijerph-16-05-00790.txt

@@ -0,0 +1 @@
+The influence of perceived unfairness at the workplace (organizational injustice) on employee health is well established. Several theories explain the unpleasant and stressful nature of the experience of injustice, using trust as a central element. This study examines the effect of trust to supervisor on the association of perceived injustice with vagal tone—an objective marker for stress experience. Questionnaires assessed organizational justice and trust. Vagal tone was measured by indictors of heart rate variability (HRV), which captured parasympathetic (pNN50, RMSSD, and HF) and parasympathetic and sympathetic (SDNN, and LF) regulation. Synergistic effects were tested by linear regressions with interaction terms between organizational justice and trust to supervisor in 38 managers. Organizational justice was related to HRV indicators that reflect in particular the parasympathetic branch (βpNN50 = 0.32, p < 0.05; βRMSSD = 0.27, p < 0.1), and interaction effects with trust to supervisor were also most pronounced there (interaction βpNN50 = −0.41, p < 0.01; βRMSSD = −0.47, p < 0.01). In conclusion, the combination of low perceived justice and trust to supervisor appears substantial to the physiological stress threat of employees. Promoting fairness at the workplace might reduce stress; if not possible, trust to supervisor should be enhanced.The perception of unfairness at the workplace (organizational injustice) has consistently been linked to various job-related factors, such as commitment, turn-over intention, motivation and productivity, as well as to adverse mental and physical health [1,2,3,4]. The negative health consequences are seen to be the result of the negative emotions and stress associated with the perception of injustice [3,5]. Several theoretical concepts have been developed to explain why people are interested in fairness and why their perception of unfairness might be stressful to them. The fairness heuristic theory argues, for example, that perceived justice at work is important to employees because of facing the “fundamental social dilemma”, which is expressed in the question of whether to trust and cooperate with authorities [6]. Therefore, “perceptions of fairness will be used as a shortcut to deciding whether to accept the authority’s decision or reject it” [7]. Accordingly, justice can be seen as an employee’s proxy for trust to supervisor and a prerequisite for the willingness to cooperate. In this context, fairness is used as a heuristic because the trustworthiness of an authority is difficult to judge since it depends on not directly observable concepts, such as integrity, benevolence, and ability [8]. In consequence, the absence of fairness signals a lack of trustworthiness of an authority, accompanied by higher levels of uncertainty. Related to this reasoning and building upon the fairness heuristic theory, the uncertainty management theory argues fairness to be a heuristic to deal with uncertainty at work in general [9]. Justice and subsequent trust thus help employees to manage general work uncertainty [10]. The concept of trust is also central to the leading theoretical framework for the importance and health effects of organizational justice: the social exchange theory [11,12,13]. According to this theory, social relationships are understood as long-term exchanges of resources (or favors) with a diffuse obligation to reciprocate [14,15]. Perceived organizational fairness is seen as a prerequisite for such a social exchange, while trust is an essential part because it can reduce uncertainty about the others’ reciprocity and obligations [10,14,16]. All these theories establish a close link between justice and trust, a link, which is also empirically confirmed in the field of organizational psychology [1,2,17]. Increasing evidence indicates that trust transmits the effects of justice [10,14,18]. In this context, trust might have the potential to reduce the uncertainty of injustice, which enhances predictability and lowers levels of distress [10,16,19,20].The physiological manifestation of prolonged high stress levels can be indexed by vagal tone. Persistent stress is accompanied by a dysregulation of the autonomous nervous system (ANS) in terms of a shift of the relation of its two branches: the sympathetic nervous system (SNS) and parasympathetic nervous system (PNS). The SNS ensures maximum performance, for example, by increasing heart rates and blood glucose levels. The PNS, by contrast, provides relaxation and recovery through reduced heart rates. Chronic stress can lead to chronic over-activation of the SNS (the parasympathetic is not sufficiently employed), leading to cardiac disease, including myocardial infarction, cardiac sudden death, and cardiovascular disease (CVD) (e.g., [21,22,23,24]). Dysregulation of the ANS can be indexed by heart rate variability (HRV) through indicating PNS activity, also called vagal tone [25]. Evidence shows HRV to be related to adverse psychosocial working conditions, including perceived organizational injustice [16,26,27].Based on the reasoning above, this empirically psychophysiological study examined the hypothesis that associating organizational justice with vagal tone, as indexed by HRV, depends on the degree of trust to supervisor in difficult situations. We hypothesize that HRV is lowest if perceived justice and trust to supervisor are low, while high trust (or justice) can compensate for low justice (or trust) reflected in HRV levels comparable to the condition in which both are high. This study uses data of a cohort followed up after a preceding stress management intervention. Employees of a manufacturing plant in Southern Germany were invited in July 2006 to participate in a stress management training program. This program was designed to improve the ability to identify and cope with workplace stressors [28]. Participants were followed up after this training in three waves: 2008, 2015, and 2016. In the last wave in 2016 organizational justice, trust to supervisor, and the heart rate (HR) were measured. Of the 63 persons participating in this wave, 49 persons volunteered HR measurements. After excluding persons with missing or invalid data on questionnaire scales (organizational justice, trust to supervisor (n = 3)) and HRV measurements (artifact ratio > 5% (n = 5) and measurement quality < 50% (n = 2)) and the only female, 38 male middle managers remained for analyses. All participants provided written informed consent, and the ethical committee of the Heinrich-Heine-University of Düsseldorf approved the study (no. 5684).Organizational justice was measured by the validated German organizational justice questionnaire (G-OJQ) [29]. This 11-item scale (seven items capture the procedural justice dimension and four items are related to the interactional justice one) asked participants to indicate on a 5-point Likert scale to what extent each statement applies to their work situations (ranging from ‘‘does not apply at all’’ to ‘‘applies completely’’). Example items are “The supervisor makes decisions that are free of personal biases” or “Everyone has the opportunity to question decisions that are made”. A mean score was calculated with lower values indicating lower organizational justice perceptions and higher values denoting higher organizational justice perceptions (Cronbachs’ α = 0.90).The trust to supervisor was assessed by one item taken from the Short Questionnaire for Work Analysis [30]. The item asks: “I can trust my direct supervisor when things get difficult at work” (1 = does not apply at all, 2 = does rather not apply, 3 = partially applies, 4 = largely applies, 5 = fully applies). The HR was measured by Faros devices, which were attached by means of chest strap to the participants at the same day they received the questionnaire. On the next day, participants returned the device. HRV indices were calculated by relevant software, which has been applied to recent research [31,32,33]. An overview of the indices used is presented in Table 1. Three indices measure HRV in the time domain (pNN50, RMSSD, and SDNN) and two in the frequency domain (LH and HF); pNN50, RMSSD, and HF primarily reflect the PNS, while SDNN and LF indices reflect the SNS and PNS.As potential confounders age (in years), job position, height and weight (to calculate body mass index (BMI)), and smoking behavior (yes/no) were assessed by questionnaire. Job position was classified as segment leader vs. other positions.To approach normal distributions, HRV indicators were logarithmically transformed, outliers (standard deviations of ±3.5) were removed, and variables were z-transformed for analyses. Linear regression models estimated associations of organizational justice with HRV indices in three sequential steps. The first step estimated the association of organizational justice with the different HRV indices, while the second step also included trust to supervisor. In the last step, a multiplicative interaction term between organizational justice and trust to supervisor was included to assess synergistic effects. Two models of adjustment with potential a priori defined confounders were calculated [26]. The first model (Model 1) controlled for age, while the second model (Model 2) additionally adjusted for job position, smoking behavior, and BMI. Analyses were performed using SPSS 25 (IBM Corp. Released 2017. IBM SPSS Statistics for Macintosh, Version 25.0. Armonk, NY, USA).Characteristics of the study population and mean values for organizational justice, trust to supervisor, and the HRV indices are displayed in Table 2. In linear regression models (Table 3), a significant interaction term between organizational justice and trust to supervisor was observed for pNN50, RMSSD, and HF power, which were also independent of level of adjustment (all betas ≥ −0.37; p-values ≤ 0.05). As indicated in Figure 1, HRV was lowest in the case of low perceived justice and low trust to supervisor. In all other conditions, HRV indices were higher on a comparable level. These findings confirm our hypothesis.Our study attempted to provide empirical evidence examining the synergistic effects of low perceived organizational justice and trust to supervisor on reduced vagal regulation. We found that HRV was lowest (indicating most stress-related vagal dysregulation) when both organizational justice and trust to supervisor were low. HRV indices were on a higher and comparable level when organizational justice and/or trust to supervisor were high. This indicates that trust to supervisor can buffer the negative stress effects of low organizational justice. Several studies have established the influence of trust on the association of organizational justice with organizational outcomes, such as job satisfaction, turnover intentions, commitment, task performance, and citizenship behavior [14,34,35]. This study extends the literature to include stress-related physiological regulation: lowest HRV values were found when perceived justice and trust to supervisor were low. With fairness heuristic theory and uncertainty management theory as our basis, we have argued that perceived justice might be used as a shortcut for trust to supervisor, which appears hard to determine. The findings of our study suggest that increased trust to supervisor might buffer stress levels which result from low organizational justice. On the other hand, if persons rate their supervisor to be less trustworthy, high organizational justice might also compensate for this experience and reduce the stress load.The reciprocal compensation of low justice and trust to reduce employee stress levels has relevant practical implications. Organizational justice perceptions might be improved by occupational interventions. For this purpose, Greenberg [36] identified three essential aspects. The first aspect refers to a dignified and respectful explanation to employees of the resource allocation in the company, while the second aspect states that employees should be given a voice and this voice should also be heard. The third aspect relates to the procedures in the company, which should be accurate, unbiased, and implemented transparently. However, occupational interventions to improve organizational justice face several challenges. Mangers are mostly unaware of injustice as a problem and tend not to address this topic [37]. Moreover, the acceptance and trust of the company, as well as the willingness to learn and change, are often lacking. These factors are essential both to change practices and to implement procedural justice rules (i.e., procedures should be consistent, without bias, accurate, correctable, representative, and ethical) to assure fairness of the decision-making process and a fair allocation of outcomes [37,38]. If it is not possible to implement such large-scale changes within a company, a remedy to buffer stress threat might be the improvement of trust to supervisor. Trust to supervisor and organizational justice were related to time-domain HRV indices reflecting PNS activation in particular (i.e., pNN50, and RMSSD). Regarding the frequency domain, an association with LF power, which reflects PNS and SNS activity, was also observed. However, significant interactions between justice and trust were found exclusively for PNS reflecting HRV indicators (pNN50, RMSSD, and HF power). This indicates that lacks of perceived organizational justice and of trust to supervisor might be related to reduced bodily recovery, potentially linking low perceived organizational justice to physical and mental ill-health [3,4,39]. The limited strength of associations between organizational justice, trust and HRV indices might raise the question about its clinical relevance. A recent dose–response meta-regression revealed that a 1% increase in HRV is associated with an approximately 1% lower risk of CVD [21]. Thus, even small changes in HRV indicators can have relevant effects and efforts to change the psychosocial work environment in terms of enhanced perceived justice and trust to supervisor have pertinence.In this study, we admit that the assessment of trust was rather simplistic, which has been criticized in the past [40]. In the context of organizational justice, different types of trust appear relevant: trust in supervisor, trust in organizations, cognition-based trust (i.e., confidence in dependability, reliability, and professionalism) and affect-based trust (i.e., confidence in emotional investments, expressions of genuine care and concern, and an understanding of reciprocated sentiments) [10,14]. Further studies might consider how these (and potentially other) types of trust relate to vagal tone.Several more limitations must also be reported. The cross-sectional design of the study prohibits causal inferences. The temporal reciprocal effects of justice and trust on stress load have to be disentangled in further longitudinal studies. In addition, due to the relatively small sample size restricted to men of a specific company in a specific position (sandwich position between higher management and production), this study must be considered as preliminary and further studies should include both women and other employee levels. Only one woman was available for our analyses; however, she was excluded because of known gender differences in justice perception (e.g., [41]). Another limitation refers to potential confounding factors, which were not included in analyses, like other CVD markers (such as blood pressure), or specific lifestyle factors (i.e., alcohol consumption and physical activity). Furthermore, HRV measures are only indirect assessments of autonomic activity. In addition, this cohort is a follow-up study of a preceding stress management intervention, potentially resulting in decreased injustice perceptions and a general improved health, leading to a selection bias. This might have restricted the measurement range and potentially led to an underestimation of the true associations. However, a comparison of the study sample with the persons who could not be followed up (attrition analyses) revealed no significant differences in key variables measured at baseline: demographics (age and education), professional variables (leadership responsibility, shift work, hours of overtime per month, daily break time, self-reported sick leave days), health status and behavior (BMI, intensive sports, and smoking), sleep quality, stress reactivity, effort–reward ratio, and mental health (anxiety and depression) (p-values > 0.10). The analytic sample and the drop-outs only differed significantly regarding the professional status (Χ2 = 9.7, p = 0.044); in the analytic sample were more segment leaders (51.4% vs. 32.1%) and less group leaders (2.7% vs. 23.4%). This overall suggests no strong selection bias. Lastly, multiple regression analyses might have inflated the Type I error rate. Bonferroni correction of the p-value for multiple testing (α/number of tests) [42] results in a p-value of 0.01 for an α of 0.05 and five HRV indicator tests. In consequence, especially the associations with p-values of ≤ 0.01 (i.e., two asterisks in Table 3) should be considered to be relevant.In conclusion, low levels of perceived justice at the workplace combined with low trust to supervisor manifest in lowest vagal tone, indicating highest physiological stress load. The findings of our empirical investigation confirm theoretical considerations that fairness might be a heuristic for trust, and unfairness is especially stressful in the absence of trust to supervisor. Potential workplace interventions have two starting points to reduce the stress levels of employees: promoting organizational justice and/or enhancing trust to supervisor.conceptualization, P.A.; methodology, R.H. and J.L.; software, R.H.; validation, R.H., J.L. and P.A.; formal analysis, R.H.; investigation, P.A.; resources, P.A.; data curation, R.H.; writing of original draft preparation, R.H; writing of review and editing, J.L. and P.A.; visualization, R.H; supervision, J.L. and P.A.; project administration, P.A.; funding acquisition, P.A.This research was funded by the German Federal Ministry of Education and Research (BMBF) (grant number: 01EL1409B).The company’s medical services supported the research team in conducting the study. We are indebted to all participants of the study and to the hosting company for supporting the stress management workshops.The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.Synergistic effects of organizational justice and trust to supervisor on HRV (pNN50; adjusted for age (Model 1)).Denotation of applied heart rate variability (HRV) indices.ms = milliseconds; ANS = autonomic nervous system; SNS = sympathetic nervous system; PNS = parasympathetic nervous system; Hz = hertz.Characteristics of the study population.Linear regression models for the association of organizational justice, trust to supervisor, and their multiplicative interaction with HRV indices.** p ≤ 0.01, * p ≤ 0.05. Model 1 adjusted for age. Model 2 adjusted for age, job position, BMI, and smoking behavior. S.E. = standard error.

Med-MDPI/ijerph_3/ijerph-16-05-00791.txt

@@ -0,0 +1 @@
+Background: Social campaigns focusing on health are commonly used within an attempt to change behavior. To date, there has not been a targeted analysis of societies’ general perception about social campaigns. The aim of this study is to assess citizens’ opinions on the effectiveness of health-related social campaigns. Methods: The data set used in the analysis was obtained from Poland’s nationwide “Social Diagnosis” study. The determinants of public opinion were assessed using a multivariate logistic regression. The independent variables included socio-demographic characteristics, lifestyle, social participation, and the use of digital media. Results: The logistic regression model was developed using 23,593 cases. Opinions about the effectiveness of campaigns depended on all the predictors included in the socio-demographic cluster, smoking, self-declared excessive alcohol consumption, physical activity, the use of mobile phones, and watching TV. A significant impact was found in all but one variable related to social participation. Conclusions: The analysis revealed that opinions about social campaigns present in the media “landscape” are influenced by many factors. Interestingly, persons exhibiting unhealthy behaviors are more resistant to health-related campaigns and surprisingly the need to make use of healthcare resources is not accompanied by an acceptance of the interventions.Today, social marketing is perceived as a promising strategy that can lead to health improvement through behavioral change. In 1971, Kotler and Zaltman defined social marketing as “an application of marketing to the solution of social and health problems” [1]. Andreasen also emphasized the importance of social change accomplished through social marketing [2]. It is expected that social marketing employs key elements originating from commercial marketing including consumer research, segmentation, targeting, and an appropriate balance of marketing mix attributes [3]. However, the analysis of the secondary evidence on the use of social marketing strategies in relation to health challenges shows that these requirements are not always strictly addressed [4].Behavior change was indicated as one of key benchmarks enabling identification of the success of social marketing programs [5]. According to this approach, genuine social marketing programs should use behavior change in both the phase of design and evaluation. Behavior change is also perceived as one of the key aims of health promotion interventions. The epidemiological transformation that occurred in the 20th century resulted in new challenges for public health with non-communicable or chronic diseases becoming one of the main problems. According to a World Health Organization (WHO) Report, eight leading risk factors causing deaths in 2004 included: High blood pressure, smoking, elevated glycaemia, insufficient physical activity, overweight and obesity, high cholesterol, risky sexual behavior, and alcohol consumption [6]. These were responsible for 50% of deaths worldwide [6]. Most of these risks are strongly linked to individual behavior. Therefore, the modification of health behaviors is an essential outcome for a successful health promotion campaign, especially in the context of non-communicable diseases. However, despite many theoretical frameworks for modifying health behaviors [7] and the efforts made to develop effective interventions, the area remains a major challenge in public health [8,9]. Some authors stress that interventions focused on changing health behaviors, undertaken in order to reduce alcohol consumption, to prevent obesity or to increase physical activity, have only limited impact on the targeted populations [9,10]. A search for new approaches to influence health-related behaviors results from the apparent disillusionment with the existing strategies.Social marketing, despite many reservations, is a possible approach. The interest in social marketing is based on the recognition that today, lifestyle and health related behaviors are strongly influenced by the overwhelming influence of consumerism and an exposure to commercial marketing. There is growing evidence that the prevalence of chronic diseases is strongly dependent on the consumption of unhealthy products aggressively marketed by the industry [11]. Social marketing, employing similar techniques to those used for commercial marketing, is perceived as an adequate countermeasure.Furthermore, social marketing is an attractive option for public health campaigns, as it can be used to address many potential objectives; not only shaping individual health behaviors, but also changing attitudes of policy makers and stakeholders who influence the legal environment for health.The evidence accumulated so far indicates that social marketing techniques may be effective in many key areas for health promotion. Their feasibility was demonstrated in relation to items such as: Youth obesity [12], increasing physical activity among adults [13], addressing global health issues [14], preventing and controlling sexually transmitted diseases [15], distributing health-related products [16], promoting influenza vaccination [17], healthful eating [18], and cessation of smoking [19]. On the other hand, there are also areas of health promotion in which the effectiveness of social marketing techniques was not confirmed [20].While social marketing programs are usually expected to assume a more comprehensive approach to behavior change, they are frequently limited to campaigns based only on communication interventions. Campaigns utilizing modern communication channels, which are an inherent element of “the landscape” of traditional and digital media have become a popular strategy for delivering health-related messages to society. There are numerous examples of the robust assessment of social campaigns addressing health-related issues such as smoking [21], however, many social campaigns broadcasted by the mass media are not followed by an appropriate evaluation. While techniques of social marketing may be feasible for many health challenges, their use has rarely been assessed in terms of effectiveness or acceptance by the general public.The primary aim of this study is the assessment of determinants of opinion about the effectiveness of health-related social campaigns by the public. Social marketing has become an important tool for health promotion and disease prevention in Poland. The increased trend of using social marketing accelerated after Poland joined the European Union in 2004, by the ability to use structural funds designed for social change in new member states. Both public institutions and non-governmental organizations began exploring social marketing techniques as a means to promote change in public behaviors and attitudes.To clarify the perception of health campaigns in Polish society, the dataset of a nationwide study entitled “Social Diagnosis” was explored and relevant data were extracted [22]. A multivariate logistic regression model was developed to determine factors shaping public opinion about the effectiveness of health-related social campaigns. Potential predictor variables were derived from items asking about socio-demographic features, lifestyle, and health status, the use of television (TV), modern technologies, and social participation.The analysis of factors determining public opinion about the effectiveness of health-related social campaigns was carried out from data collected during the wave of the Social Diagnosis study performed in 2011. The “Social Diagnosis” is a panel study carried out every two years on a national sample representative of all Poland [22]. The study is focused primarily on the assessment of objective and subjective quality of life. A two-stage households sampling procedure, based on territorial units and categories of the area of residence was used in the study. Respondents of 16 years and over were asked to fill individual questionnaires in the presence of interviewers employed by the study. The data set from the Social Diagnosis study may be accessed from the website maintained by the research team [23]. The details of the sampling procedure and generation of calibrated weights are specified in the report from the study [24].During the 2011 study, an item asking about the respondent’s opinion of the effectiveness of health-related social campaigns was included in the individual questionnaire (“Do you think that social campaigns and other actions aimed at improving health, e.g., antismoking campaigns, campaigns against drugs and promotion of vaccinations, are effective and change people’s behaviour in Poland?”). A dependent variable used for logistic regression model development was derived from this item. It was coded as follows: The response options “decidedly yes” and “rather yes” were treated as confirmation of conviction that social campaigns are effective (coded as ‘1’), and options “decidedly no”, “rather no”, and “difficult to say” as a lack of conviction about the campaign’s effectiveness (coded as ‘0’).Four types of independent variables were used in the multivariate logistic regression model. They included sociodemographic characteristics, variables related to lifestyle and health behaviors, variables related to social participation, and finally, variables reflecting the use of media and information technologies (IT).The cluster of sociodemographic variables consisted of gender, age category, education level, place of residence, marital status, occupational status, and net income. The age of respondent was included as a categorized variable with 6 intervals assumed in the “Social Diagnosis” study: 18–24, 25–34, 35–44, 45–54, 55–64, and 65 years or more. Place of residence also was characterized by 6 values: urban areas with >500,000 inhabitants, 200,000–500,000, 100,000–200,000, 20,000–100,000, <20,000 and rural area. Education level was assigned with four options: primary or lower education (level 1 or lower, as classified by the International Standard Classification of Education (ISCED)) [25], lower secondary (vocational education or middle school; level 2 according to ISCED), upper secondary (level 3 according to ISCED), and post-secondary non-tertiary education or university education (levels 4–8 according to ISCED). Marital status assumed four options: ‘married’, ‘unmarried’, ‘widower/widow’, and ‘divorced or remaining in separation’ (formal or actual). Six categories were used for occupational status: ‘employee’, ‘self-employed or entrepreneur’, ‘retired or on disability pension’, ‘university or school student’, ‘unemployed or occupationally passive’, and ‘not provided’. The initial continuous variable providing information about the level of personal net monthly income (average from three preceding months) was categorized into four categories: ≤1000 Polish zlotys (PLN), >1000 to 2000, >2000 PLN per month, and ‘not provided’.Use of health care services during the last year was a dichotomous variable coded as ‘0’, if the respondent used neither public nor private health care services in the preceding year, and as ‘1’ in cases of at least one episode of use. The variable disability status was derived from the relevant item asking about respondent’s being disabled (either formal status when disability was established by decision of a relevant institution in Poland, or informal, when such a confirmation was not sought).Respondents were asked if they undertake any form of physical activity or sport (no activity vs. at least one type of physical activity or sport). Items asking about current smoking and excessive alcohol consumption were also dichotomous (‘no’ vs. ‘yes’). Finally, the body mass index (BMI) was calculated from weight and height values provided by respondents in the individual questionnaires and categorized by four intervals: <18.5 (underweight), ≥18.5 to <25.0 (normal weight), ≥25 to <30.0 (overweight), and ≥30.0 (obesity).In this group, three variables were included: computer use, mobile phone use and TV viewing. Computer use and mobile phone use were dichotomous variables (‘no’ vs. ‘yes’) The variable related to daily TV viewing assumed four values: Not watching TV, watching TV for less than 1 h, for 1–2 h, and for more than 2 h daily.There were six variables related to social participation used in the model: The number of friends with whom a respondent meets regularly, involvement in activities for the local community, membership in social organizations, participation in religious practices, participation in the last election before the survey, and the ability to indicate a political option. The variable related to number of friends declared assumed 4 values: not more than 2, 3–4, 5–9, and more than 9 friends. The variables indicating involvement in activities for benefit of the local community (district, village or city, neighborhood) and membership to social organizations were dichotomous (lack of involvement or membership vs. confirming activity). Participation in religious services or meetings was assessed based on the number of events attended per month. A relevant variable could assume three values: ‘0’ for those participating in religious services less than once monthly or not participating; ‘1’ for once monthly and ‘2’ for at least twice monthly. Political activity was determined from the response asking about active participation in local government elections, which took place in 2010 (the last elections taking place before the 2011 Social Diagnosis study). The variable could assume three values: not participating in elections, declaring participation, and response not provided. Finally, an independent variable was derived from the response asking for indication of a political party that is closest to the respondent’s views. It assumed three values; ‘no political sympathy’, ‘difficult to say’, and ‘clear political sympathy’. The analysis was performed with SPSS v.21 for Windows (IBM Corp., Armonk, NY, USA). Only data of respondents of 18 years or over were retrieved for the analysis. The frequencies reported in the paper were provided after exclusion of missing values. The factors that could have impact on the opinion about the effectiveness of health-related social campaigns were included as independent variables in the multivariate logistic regression model. The forward method available in SPSS v.21 package was applied. For each independent variable included in the model, odds ratio (OR), and 95% confidence interval (95% CI) were provided. The logistic regression modelling was preceded with assessment of missing values. As the percentage of missing values was lower than 1% for all but one independent variable; in the case of computer use, it was 1.3%, and it was decided that all records with at least one missing value would be removed from the model. Thus, the multivariate logistic regression model was calculated, after adjusting for standardized weights. The analysis of multicollinearity demonstrated that VIF values for independent variables remained in the interval 1.07–2.07. Therefore, all of them were included in the multivariate logistic regression model.Weighted number of respondents was 23.593 persons with women composing 55.92%. The percentage of persons with university education or higher was 19.15%, and those with primary education 18.48%. The number of persons working as employees was 35.29%, self-employed 3.86%, farmers 6.42%, and retired or receiving disability pension 33.29% of all respondents.Nearly 82% of all respondents used private or public health care services at least once in the preceding year. In the study population 14.26% had disabilities. Only 34.20% declared some type of physical activity, 6.14% reported excessive use of alcohol, and 26.28% smoking. 54.91% of the respondents were overweight or obese. Computer use was confirmed by 53.35% of respondents and the use of mobile phone by 81.62%. Only 2.56% reported that they did not watch TV.Respondents who had no more than two friends were the most numerous group (32.47%). Only 15.91% confirmed involvement in activities for their local communities and 11.62% reported participation in some type of social organizations. 67.93% of respondents reportedly participated in local government elections in 2010 and 45.17% of them indicated political party views close to their own. Finally, 53.35% of respondents expressed the opinion that health-related social campaigns are effective, while the remaining respondents were not convinced about the effectiveness. Detailed characteristics of the study group are provided in Table 1 (the numbers and frequencies are provided for weighted values).Multivariate logistic regression modelling was performed for 22 variables deemed to have impact on the opinion of respondents about the effectiveness of health-related social campaigns (Table 2). The model demonstrated adequate goodness-of-fit (Hosmer and Lemeshow test chi2 = 14.02, df = 8, p = 0.081). However, overall, the predicting value of the model was rather limited (the increase of proper classification from 53.4% to 58.7%; Nagelkerke R2 = 0.052).Men were more skeptical about the effectiveness of social campaigns (OR 0.88, 95% CI = 1.04–1.33) than women. As far as age groups, a significant difference was found only for comparison of the youngest group of respondents (18–24) and those in the category of 25–34 years of age (OR 0.88, 95% CI = 1.04–1.34). There were no significant differences between the referential category (18–24) and older age categories. The respondents who achieved higher levels of education were more prone to believe in the effectiveness of health-related social campaigns. The odds that a person with post-secondary or higher education is convinced about their effectiveness were 1.40 times higher than a person with primary education. Respondents living in urban areas (OR 0.89, 95% CI = 0.72–0.87) and small cities <20,000 inhabitants (OR 0.84, 95% CI = 0.76–0.94) were less prone to express belief in the effectiveness of campaigns than those living in urban areas with >500,000 inhabitants. Persons with net income surpassing 2000 PLN were more often convinced about campaigns effectiveness than persons having a net income not surpassing 1000 PLN (OR 1.11; 95% CI = 1.01–1.23). Finally, married persons demonstrated a higher belief in the effectiveness of campaigns than unmarried (OR 0.88, 95% CI = 0.81–0.96), widows/widowers (OR 0.89, 95% CI = 0.80–0.98), and divorced or in separation (OR 0.86, 95% CI = 0.76–0.97).The opinion about effectiveness of health-related campaigns did not depend on disability status, the use of health care services during the preceding year or BMI. Persons declaring at least some form of physical activity were more prone to believe in campaign’s effectiveness (OR 1.40, 95% CI = 1.07–1.21). Smokers and persons drinking alcohol excessively less frequently showed belief in campaign’s effectiveness (respectively, OR = 0.88, 95% CI = 0.83–0.94, and OR = 0.84, and 95% CI = 0.76–0.93). There was no difference between computer users and nonusers (OR 1.00, 95% CI = 0.93–1.08). Mobile phone users were more prone to believe in campaign’s effectiveness (OR 1.17, 95% CI = 1.07–1.28). The odds that person watching TV will be convinced about the effectiveness of social campaigns was at least 1.14 times higher than person not watching TV at all. The highest difference was seen between non-watchers and persons watching TV for 1–2 h daily (OR 1.27, 95% CI = 1.09–1.48).All but one variable related to social participation had a statistically significant influence on the opinion about the effectiveness of health campaigns. The comparison of persons who either undertake or do not undertake activities for local community did not show a statistically significant difference (p = 0.53, OR = 1.08, 95% CI = 1.00–1.17). Respondents with a higher number of friends with whom regular meetings or contacts were maintained, more frequently believed in the effectiveness of health campaigns. Results for comparisons between persons with not more than two friends were (OR (95% CI)), for 3–4 friends, 1.14 (1.05–1.23), for 5–9 friends, 1.09 (1.01–1.18), and for more than 9 friends, 1.08 (1.01–1.17). Furthermore, the odds that a person who declared membership to at least one social organization were 1.10 (95% CI = 1.01–1.20) times higher when compared with a person who did not belong to such organizations. The conviction about the effectiveness of social campaigns was also higher among respondents who participated in elections of local governments (OR = 1.21, 95% CI = 1.14–1.29) and who could indicate a political party which would be the closest to their views (OR = 1.41; 95% CI = 1.33–1.49). Persons attending religious services or meetings once or twice monthly more frequently believed in the effectiveness than those not practicing or practicing less frequently than once monthly (OR and 95% CI, respectively; 1.17, 1.06–1.30, and 1.13, 1.06–1.20).The analysis performed in this paper was based on the large dataset retrieved from the 2011 wave of nationwide Social Diagnosis study. Among the respondents, about 53% were convinced that health-related campaigns are effective. After exclusion of cases with at least one missing value, a multivariate logistic regression model was developed with 23,593 cases. The independent variables used in the model represented four main groups. They included socio-demographic and economic variables, health care utilization and lifestyle, TV and IT use, and finally, social participation. Only four of the independent variables did not exert any influence on the opinion about effectiveness of health-related social campaigns. The level of influence of independent variables that reached statistical significance was not high. The differences between reference categories and tested categories expressed as OR, rarely reached values below 0.80 or above 1.20. The highest differences were seen for the comparison of persons without political sympathies and those with a clear connection with their closest political party (OR = 1.41), for persons who did not participate and those who participated in the last election (OR = 1.21), for persons with the lowest and the highest level of education (OR = 1.39), or those with upper secondary education (OR = 1.28), and finally for the comparison of persons not watching TV at all with those watching it for 1–2 h daily (OR = 1.27).The people more frequently convinced of the effectiveness of social campaigns, considering sociodemographic characteristics, are women, the persons who reached a higher education level, those living in larger urban areas, persons employed in private or public organizations, people with the highest income, and those living as married couples.Utilization of health care services during the previous year and disability status did not have a significant impact on the opinion about such effectiveness. It is surprising that utilization of health care services did not have an impact on the opinion about health campaigns. One might expect that persons having more intensive contact with health care systems would also receive some advice on general health risks and therefore should appreciate the importance of health promotion interventions on a societal level. Persons who smoked or declared excessive alcohol drinking expressed a belief about campaign’s effectiveness less frequently. Such an effect was not seen in case of persons with obesity or who were overweight.Computer usage also did not predict an opinion about the effectiveness of health campaigns. This finding seems to remain in line with the results of the study reported by Redmond et al. [26]. The analysis of the data from Health Information National Trends Survey carried out in the USA in 2005 and 2007, revealed that the use of print media and interpersonal sources of health information exerted a stronger effect on self-reported health behaviors than the use of other sources, including TV and the Internet.On the other hand, the model developed in our study showed that persons who watched TV at least one hour daily were more in favor of the effectiveness of health campaigns than those who did not watch TV at all. A similar effect was seen among those who used a mobile phone.Those who were involved in religious participation, engaged in social organizations, had defined political views and declared participation in voting in local government elections preceding the Social Diagnosis study favored the effectiveness of social campaigns.Unfortunately, it seems that there is a research gap as to the perception of social campaigns by the general public. It is also obvious that more efforts are directed toward the assessment of the needs of potential target groups for specific campaigns. Therefore, a discussion of the results obtained in this study in comparison to other populations is hardly possible. However, opinion about the effectiveness of health-related social campaigns may be treated as an indicator of attitude to public health measures and readiness to accept such communications on key health problems in the society. Assuming this understanding, one can refer to the effects reported of social participation variables used in our study on health behaviors and related outcomes.The findings reported by other authors tend to agree with the results of our study. Bender et al. demonstrated that higher social capital was associated with a higher probability of participating in a health check-up [27]. In their study, informal socializing and voter turnout was used as measures of social capital. The effect of informal socializing was maintained after inclusion of neighborhood deprivation in the model, and voting turnout became non-significant. Informal socializing was assessed based on responses to two items asking about frequency of contacts with family members and with friends or acquaintances. Neighborhood (census districts) voter turnout was based on the neighborhood participation in the Danish parliament election [27]. Inequality in voter participation may be related to self-rated health. According to Blakely et al., individuals living in the states (USA) with the highest voting inequality had an odds ratio of fair/poor self-rated health status of 1.43 (95% CI: 1.22–1.68) [28].The beneficial influence of involvement in religious participation/practices on health outcomes, morbidity, adult mortality risk, and longevity was reported earlier [29,30,31]. This effect is attributed to the impact of religious involvement on lifestyle factors and social behaviors [29,32,33,34]. Some authors also emphasize the importance of social support which can be received in religious communities [29,32,33,34,35]. The impact of involvement in religious practices on the perception of the effectiveness of health-related social campaigns seen in our study may be linked to an acceptant attitude toward society level communications encouraging healthy behaviors related to religious involvement.Interestingly, people demonstrating harmful health-related behaviors are less prone to participate in social activities, e.g., in elections. In the study of Albright et al., daily smokers were less than half as likely as non-smokers to report having voted in the election [36]. The authors suggested that their findings could indicate that there was a link between risky health behaviors and political mistrust. In our study, smokers were also less likely to believe in the effectiveness of social campaigns. It seems that this mistrust may be extended to other domains. It is also important to note, that, although smokers are a key target group for antismoking campaigns, they may be more resistant to this form of health communication than non-smokers.The number of close friends is treated as one of the indicators of social support. It is commonly accepted that loneliness leads to unfavorable health outcomes and social support is linked to better health, partly because it prevents loneliness [37,38]. Our study revealed that persons with a higher number of friends were more prone to believe in the effectiveness of social campaigns. It is likely that the protective health effect of social support may to some extent be explained by an increase of an acceptant attitude linked to social interaction by health promotion interventions.The study reported in this paper suffers from several limitations. First, it is a secondary analysis of data collected during a study which was not focused on the analysis of determinants of societal approaches to social campaigns. Second, the selection of the variables used as predictors for logistic regression modelling was based on the assumption that they may have an impact on respondents’ views about the effectiveness of campaigns. The Social Diagnosis study encompassed only one item related to social campaigns and a more thorough analysis of relations between respondents’ views and variables established based on the items available in the individual questionnaire was not possible. Among the most important deficiencies, one should enlist the lack of feedback on understanding and being able to properly identify examples of health-related social campaigns. Therefore, it is not fully clear if the selection of the responses showing the lack of belief in the effectiveness of campaigns or inability to express an opinion originates from problems with understanding the concept of social campaigns or an actual opinion about their effectiveness. This weakness was compensated to some degree by the way the item asking about social campaigns was formulated. Two examples of social campaigns were included in the content of the item when asking about their effectiveness.Finally, the dataset available from each wave of the Social Diagnosis study is very broad and some potentially relevant items could have been overlooked as to their importance for the opinion about social campaigns.The perception of the effectiveness of social campaigns by the general public has rarely been addressed so far. It is also not clear how such perception translates into the effectiveness of campaigns measured with health-related outcomes, e.g., actual change of harmful behaviors.The analysis performed in this paper indicates that the highest conviction about effectiveness of social campaigns was demonstrated by females, persons with higher education and achieving higher income, remaining in marital relation and living in more urbanized areas. Unhealthy behaviors declared by the respondents (higher alcohol consumption, and tobacco smoking) were related to lower belief in the effectiveness of social campaigns. It is not clear if such attitude precedes the beginning of risky behaviors, or it is developed in association with such behaviors. Nonetheless, it seems that people undertaking risky behaviors could be less prone to accept public communication related to health challenges. On the other hand, respondents presenting healthier lifestyles, as confirmed by involvement in physical activities, showed more positive attitudes towards campaigns.Interestingly, the factors related to social participation which favor the conviction about social campaigns being effective, are also associated with beneficial health effects. This was seen in both the case of election participation and involvement in religious practices.The finding that recent utilization of health care services does not influence opinions about the campaigns’ effectiveness is unexpected. One could expect that personal experience of searching for medical help would result in a more open attitude toward health-related communication present in the public domain.The study was performed in the framework of statutory project No K/ZDS/006112 carried out in the Department of Health Promotion, Institute of Public Health, Faculty of Health Sciences, Jagiellonian University Medical College, Kraków, Poland.The author would like to thank Alex Solsbery, MA, for proofreading of the manuscript.The author declares no conflict of interest.Characteristics of the study population (weighted values).a education categories used in the survey were mapped by the levels distinguished by the International Standard Classification of Education (ISCED) classification from 2011 [25]. * In 2011, according to the Polish National Bank, the 1 USD to PLN exchange rate ranged between 2.9822 and 3.4174 and the mean gross salary in Poland was 3400 PLN.Multivariate logistic regression results for opinion about the effectiveness of health-related social campaigns.BMI: body mass index.

Med-MDPI/ijerph_3/ijerph-16-05-00792.txt

@@ -0,0 +1 @@
+Human exposure to carbon nanotubes (CNTs) can cause health issues due to their chemical-physical features and biological interactions. These nanostructures cause oxidative stress, also due to endogenous reactive oxygen species (ROS) production, which increases following mitochondrial impairment. The aim of this in vitro study was to assess the health effects, due to mitochondrial dysfunction, caused by a sub-chronic exposure to a non-acutely toxic dose of multi walled CNTs (raw and functionalised). The A549 cells were exposed to multi-walled carbon nanotubes (MWCNTs) (2 µg mL−1) for 36 days. Periodically, cellular dehydrogenases, pyruvate dehydrogenase kinase 1 (PDK1), cytochrome c release, permeability transition pore (mPTP) opening, transmembrane potential (Δψ m), apoptotic cells, and intracellular ROS were measured. The results, compared to untreated cells and to positive control formed by cells treated with MWCNTs (20 µg mL−1), highlighted the efficiency of homeostasis to counteract ROS overproduction, but a restitutio ad integrum of mitochondrial functionality was not observed. Despite the tendency to restore, the mitochondrial impairment persisted. Overall, the results underlined the tissue damage that can arise following sub-chronic exposure to MWCNTs.With the growing trend in the production and applications of carbon nanotubes (CNTs), the increasing use in composite materials [1] and their exploration as innovative solutions for biomedical applications [2,3,4], there will be a corresponding increase in potential human exposures [5]. Such exposures can cause substantial health issues as a result of the chemical–physical features and biological interactions of CNTs. Numerous efforts have been made over the past two decades to investigate the biocompatibility and toxicological effects of CNTs, which are still poorly understood and controversial [6,7,8,9].The harmful effects of CNTs have been shown to be highly dependent on the type of cells used for in vitro studies [10] and on the heterogeneity of the produced CNTs. The length, diameter, structural defects, surface area, tendency to agglomerate dispersibility in water solution, presence and nature of metal catalyst residues [11,12] and surface chemistry [13,14,15,16] greatly influence the biological reactivity of CNTs.Oxidative stress is a common mechanism involved in the cytotoxicity of these nanostructures [17]; following inhalation, the CNTs penetrate deeply in the respiratory tract and cause a strong pro-oxidant effect. This effect, which could trigger carcinogenic asbestos-like mechanisms, was previously observed by our research group in an in vitro cell model of alveolar epithelium and neuronal-like cells after short-term exposures to multi-walled CNTs synthesized in our laboratory [18,19]. Consistent with an overproduction of reactive oxygen species (ROS), we detected significant time- and dose-dependent increases in lipid peroxidation and mitochondrial impairment.Polyunsaturated fatty acids (PUFAs), present in the mitochondrial membrane, are particularly susceptible to free radical-initiated oxidation, which determines the decrease in the mitochondrial transmembrane potential and metabolic impairment because of the shift from oxidative phosphorylation to anaerobic glycolysis. Mitochondrial damage, by releasing caspase-activating proteins, can trigger the intrinsic apoptotic pathway, as observed in the alveolar cell line exposed to airborne particles [20,21], asbestos [22], metals and oil fly ash [23,24,25].Mitochondria are the powerhouses of cells and, at the same time, the suicidal weapon store because they are the primary sources of intracellular ROS production. It has been estimated that physiologically, during oxidative phosphorylation, 2–4% of the oxygen is converted to O2− and, then, to H2O2. Dozens of lethal signal transduction pathways converge on mitochondria and cause the permeabilisation of the mitochondrial outer membrane, leading to the cytosolic release of pro-apoptotic proteins and to the impairment of the bioenergetic functions of mitochondria [26].The increased residence time of electrons in complexes I and III of the respiratory chain [27] causes a depolarisation of the inner mitochondrial membrane (∆ψm), which can further amplify the redox imbalance MWCNT-induced through the production of endogenous ROS. Several studies confirmed the destruction of mitochondrial membrane potential (∆ψm) and mitochondrial swelling by CNTs; both events were produced as a result of the increased formation of ROS, the release of cytochrome c (cyt c), the disturbance of the mitochondrial electron transfer chain (mtETC) complexes and the collapse of the mitochondria [6,28,29].ROS are determinants for the release of the pro-apoptotic cyt c because cyt c is bound to the inner membrane by anionic phospholipids, specifically cardiolipin (CL), that are highly susceptible to peroxidation [30]. The proteins, Bax and Bak, increasing mitochondrial outer membrane permeabilisation, promote cyt c release and apoptosis [31,32].Mitochondrial permeability transition pore (mPTP) is another key participant in mitochondrial apoptosis leading to the mitochondrial depletion of Ca2+ and to the release of cyt c that trigger caspase activation [33].Most of the in vivo and in vitro studies reported in the scientific literature [3,6,10] were performed to evaluate acute effects of CNTs using higher doses compared to the doses at which the population may be exposed. Conversely, the aim of the present study was to increase knowledge on the health effects, focusing on the effects on mitochondria function, of sub-chronic exposure to a non-acutely toxic dose, which is the most realistic dose, especially for people occupationally exposed.We examined homemade raw (i.e., pristine) MWCNTs (which are called pMWCNT) and functionalised MWCNTs (that is, MWCNT-COOH, named fMWCNTs); these two MWCNTs have been widely studied for the assessment of short-term toxicity in two different cell lines [18,19]. pMWCNTs were synthesised by catalytic chemical vapour deposition (CCVD) and subsequently purified from both free metals and carbonaceous particles, by products of the synthesis process [34,35]. The covalent insertion of a carboxylic group was obtained by strong acidic oxidation. Both MWCNTs were carefully characterised by thermogravimetric analysis (TGA), UV (ultraviolet) spectra, scanning electron microscopy and high-resolution transmission electron microscopy, as previously reported [18]. In addition, abiotic and in vitro experiments were performed to assess the bioavailability of iron that was used as a catalyst in the MWCNT synthesis. Iron was detected by atomic absorption spectroscopy analysis and was equal to 2.5–2.8%. This was almost fully comprised of Fe2O3 and was not bioavailable [12].Due to the strong hydrophobicity, common to all CNTs, and to the van der Waals forces occurring at the surface, the concentrated MWCNT suspensions (100 × in PBS (Phosphate buffered saline)) were sonicated for 20 min in an ice bath (frequency 40 kHz). Moreover, just before all in vitro experiments, the MWCNT work suspensions were, further, sonicated for 3 min in the culture medium containing 10% fetal bovine serum (FBS) to enhance their dispersibility by the protein content of the cell medium.The lung alveolar region is the prime site of deposition for inhaled particles, including engineered nanomaterials, which, due to their size, fall within the breathable fraction. Therefore, we used the human alveolar cell line A549 as a pulmonary-like cell system. Cells were cultured in RPMI medium with 2 mM L-glutamine, 10% (v/v) FBS, 100 IU mL−1 penicillin and 100 µg mL−1 streptomycin at 37 °C in a humidified 5% CO2 atmosphere.Based on data previously obtained using the same cell model and from the literature regarding possible exposures in the workplace [6,17,18], we assessed the effects of sub-chronic exposure to 2 µg mL−1 of CNTs suspensions. In parallel, cell monolayers were exposed to pMWCNTs and fMWCNTs. The cells were maintained in the same medium, with the addition of PBS instead of the CNT suspensions, to serve as a negative control, and cells treated with MWCNT suspensions at 20 µg mL−1 were used as a positive control. The 20 µg mL−1 concentration was chosen to guarantee a sufficient level of survival of the exposed cells for the times established by the experimental protocol. Every three days, cell medium with and without MWCNTs was changed and treated and non-treated monolayers were sub-cultured weekly. At 1, 7, 21, 28 and 36 days, aliquots were sampled to perform the analyses. In parallel, cell viability was evaluated by the trypan blue exclusion test using a 0.4% dye solution. To confirm MWCNTs–cell interaction, previously observed in short time examinations, we carried out a qualitative analysis, and A549 semiconfluent monolayers were observed by transmission microscopy.To assess the mitochondrial function, we used western blot analysis to detect the expression of pyruvate dehydrogenase kinase 1 (PDK1) and the release of cytochrome c in the cytosol fraction devoid of mitochondria.To measure PDK1 protein expression, cells were lysed for 10 min on ice with RIPA buffer (1% NP-40 or Triton X-100, 1% sodium deoxycholate, 0.1% SDS (Sodium Dodecyl Sulphate), 150 mM NaCl, 50 mM Tris-HCl, pH 7.8, 1 mM EDTA (Ethylenediaminetetraacetic acid)) supplemented with a protease inhibitor cocktail. Moreover, to evaluate the mitochondrial release of cytochrome we resuspended the cells in 200 μL of STM buffer (250 mM sucrose, 50 mM Tris-HCl pH 7.4, 5 mM MgCl2, protease and phosphatase inhibitor cocktails) and homogenized for 1 min on ice using a tight-fitting Teflon pestle. The homogenate was decanted into a centrifuge tube and maintained on ice for 30 min, vortexed at maximum speed for 15 s, and then centrifuged at 800× g for 15 min to obtain cytosolic and mitochondrial fractions in the supernatant. Then, by a further centrifugation of the supernatant at 11,000× g for 10 min, a cytosol fraction devoid of mitochondria was obtained, and 30 µg of this latter supernatant was fractionated on SDS-PAGE. Subsequently, they were electrically transferred to a nitrocellulose membrane (Millipore, Rodano, Italy) and were blocked with 5% non-fat dry milk in TBS-T buffer (10 mM Tris-base, 10 mM NaCl, and 0.1% Tween-20) overnight at 4 °C. Then the membranes were probed with mouse anti-PDK1 monoclonal antibody (diluted 1:500 in TBS-T), anti-cytochrome c (diluted 1:100 in TBS-T), and β-actin (diluted 1:3.000 in TBS-T) for 2 h at room temperature followed by incubation with horseradish peroxidase-conjugated anti-mouse secondary antibodies (respectively diluted 1:1.500, 1:1.000 and 1:10.000 in TBS-T) (Sigma-Aldrich, Milan, Italy). Immunoblots were developed with an ECL kit on Kodak film. After normalization against β-actin, blots were scanned and quantified by densitometric analysis with Image J 1.47 (http://imagej.nih.gov/ij/).Moreover, cellular dehydrogenases, including succinate dehydrogenases (SDH in mitochondrial complex II), were detected by measuring the reduction of 3-(4,5-dimethylthiazol-2yl)-2,5-diphenyltetrazolium bromide (MTT). Analyses were performed on aliquots of each sample. Briefly, cell suspensions were normalised to a final concentration of 1 × 105 mL−1, transferred to microplates (100 µL/well), and 0.4% of MTT was added before incubation at 37 °C for 3 h. Then, a mixture composed of 50 mM HEPES (4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid), pH 8.0, and ethanol (1:9, v/v) was added to solubilise the resulting violet coloured formazan crystals. Absorbance of the dye was measured at 540 nm using a microtiter plate reader (Tecan Italia, Milan, Italy). The enzymatic activity was obtained by calculating the product between the absorbance values and the dilution factors. The values obtained were compared with the negative control (100%).Mitochondrial permeability transition pore (mPTP) opening was evaluated by using the calcein-AM/cobalt method [36]. After treatment with MWCNTs, the cell aliquots, prepared as described above, were seeded in 96-well plates (100 µL/well) and loaded with 5 µM calcein-AM and 0.5 mM CoCl2 (cytosolic calcein quencher) in PBS for 15 min at 37 °C. The cells were analysed by a microplate reader with an excitation wavelength of 488 nm and an emission wavelength of 525 nm. Decreases in fluorescence were indicative of the loss of calcein due to the mPTP opening.The effect of MWCNT–cell interactions on mitochondrial function was further evaluated by analysing mitochondrial transmembrane potential. This was accomplished by measuring the incorporation of the fluorescent probe rhodamine 123 (R123, 10 µM) (Invitrogen Molecular Probes, Milan, Italy). Based on its chemical properties, this cationic fluorochrome crosses the mitochondrial membrane and it is stored in the matrix of functional mitochondria with a transmembrane potential (Δψm) that is indicative of an active proton gradient during oxidative phosphorylation.Intracellular ROS were detected using the fluorochrome 2′,7′-dichlorofluorescein-diacetate (DCF-DA) (1 µM). For the analyses, treated cells were washed three times with PBS containing 10 mM D-glucose at pH 7.4. Then, aliquots of cell suspensions (1 × 105 mL−1) were prepared in the same buffer and the emission values were read both before and after the separately addition of the respective probe, in order to subtract emission values that may have been due to auto-florescence. The probe-loaded cells were incubated for the times indicated and at the temperatures established in the study protocol. Fluorescence intensity was measured using a microplate reader (Tecan Italia), and the wavelengths of excitation and emission were 535–595 for R123 and 485–535 for DCF-DA, respectively. For each analysis, the obtained values were used to calculate the percentage changes (%Δ) compared to untreated cells.Apoptotic cells were assessed by fluorimetric detection using AnnexinV-FITC (Sigma-Aldrich, Milan, Italy). The analysis is based on the changes of phosphatidylserine (PS) position in the cell membrane and in its outsourcing in apoptotic cells. Cell suspensions were obtained by combining the cells that were recovered using 0.25% trypsin and 1 mM EDTA with the ones suspended in medium. Briefly, cells from all subcultures were harvested by centrifugation at 1000× g for 5 min and were resuspended (approximately 1 × 106 cells mL−1) in Annexin Binding Buffer 1× (100 mM HEPES/NaOH, pH 7.5, 1.4 M NaCl and 25 mM CaCl2) containing Annexin V-FITC (0.5 μg mL−1). After 20 min of incubation at 37 °C, cell suspensions were centrifugated washed twice and resuspended in 100 μL, using the same buffer, and then transferred into 96 well microplates. In a microplate reader (Tecan Italia) and by using 485 nm as wavelength of excitation and 535 nm for emission, the emission values were measured to calculate the percentage changes (%Δ) of apoptotic cells compared to untreated cells.An overview of the effects of MWCNT was obtained by determining the cellular proliferation index. Starting from the same number of cells for each treatment, the index was obtained by the cell count that was periodically carried out on the sample aliquots. The average values at the intervals that were assayed were compared to the values recorded in the control cells.All data are presented as mean ± the standard error of the mean (SEM) based on at least three independent experiments. Data were analysed by one-way analysis of variance (ANOVA), and multiple comparisons of the means were performed by the Tukey–Kramer test (GraphPAD Software for Science, San Diego, CA, USA). The relationships between different parameters were assessed by the Pearson correlation coefficient. Significance was accepted at p < 0.05.To evaluate the effects of sub-chronic exposures at occupationally realistic doses of p- and f-MWCNTs, we preliminarily checked the interactions of these two MWCNTs with lung epithelial cells. Our qualitative assessment confirmed what was obtained previously in the same acutely exposed cell model, and it highlighted the concentration and time-dependent effects of MWCNTs. As shown by the microscopic observation (Figure 1), the presence of dark aggregates in or on the cells was clearly visible in A549 monolayers treated with both types of MWCNTs at the highest dose (positive controls). Moreover, at this dose, some spaces were observed, indicating the detachment of dead cells that were more numerous in pMWCNTs-treated cells. Instead, at 2 μg mL−1, the effects of the nanotube in the monolayers were barely detectable, especially in f-MWCNTs- treated cells, and the cellular morphology was superimposable to that of the control cells. In these experiments, cell viability was assessed measuring the percentage of dead cells. The table in Figure 1 reports the results obtained by the trypan blue exclusion test. Compared to untreated cells, significantly higher percentages of dead cells were observed after 1 day in positive controls (20 µg mL−1) for both p- and f-MWCNTs (p < 0.01). Despite the fact that the number of dead cells was almost halved after 36 days of exposure, the differences in comparison to control cells were maintained and were very significant (p < 0.01). Even at the tested dose, the percentage of dead cells in the monolayers treated for 1 day were significantly higher (p < 0.05). However, treatment for 36 days decreased these percentages that, in comparison to untreated cells, were not significant. The trend of cellular mortality showed an adaptability inversely related to MWCNT sub-chronic exposure dose in the sub-cultured cells.A wide spectrum of analyses was performed to assess the mitochondrial compartment and the MWCN-induced apoptosis. In particular, the expression of PDK1, the release of cytochrome c, the transmembrane potential, the mPTP opening, and the outsourcing of phosphatidylserine were evaluated. The MWCNTs-induced inhibition of mitochondrial activity was detected by assessing the expression of PDK1 that regulates glucose metabolism by phosphorylation of the E1α subunit of mitochondrial pyruvate dehydrogenases. Figure 2A shows a representative result of a Western blot analysis that highlights the dose dependent higher levels of PDK1, in MWCNTs-treated cells in the short time. Despite the fact that PDK1 levels decreased over time, limiting the effects on TCA cycle a full metabolic recovery was never completed in the examined time interval. Furthermore, the Western blot results of cytochrome c in the cytosolic fraction devoid of mitochondria are reported in Figure 2A. The higher levels of cytochrome c, observed after short exposure, confirmed the MWCNTs-induced mitochondrial impairment that was partially counteracted over time. Mitochondrial impairment was also assayed by measuring the transmembrane potential, which is strongly related to the active proton gradient that is maintained during oxidative phosphorylation in metabolically active cells. The measurement of R123 emission underlined the MWCNTs-induced mitochondrial dysfunction, and Δψm values were almost halved at 1 and 7 days (p < 0.05) in positive controls of both nanotube types, highlighting the organelle collapse (Figure 2B). Although not significant, even at the occupationally realistic dose tested, Δψm decreased and the values were within 65% of the control cell values. Analogously to what was observed with the analysis of the enzymatic activity, the transmembrane potential also increased progressively over time, confirming a tendential but not complete recovery of mitochondrial function in cell progeny that were steadily exposed to MWCNTs. In fact, compared to control cells, Δψm values in cells treated for 36 days were 75 and 83% for p- and f-MWCNTs, respectively.The results obtained by mPTP opening further confirmed MWCNTs-induced mitochondrial impairment (Figure 2C). The emission values of calcein were decreased in cells treated for 1d, underling mitochondrial release of calcein in the citosol, where its fluorescence was quenched by the Co which is unable to cross mitochondrial membrane. However, in sub cultured cells, a recovery was observed and, at the end of the examined interval, the emission values recorded in cells treated with 2 μg mL−1 were 10% and 5% below the control cells for cells treated with p- and f-MWCNTs respectively. As expected, the results of apoptotic cells obtained by using the outsourcing of phosphatidylserine, as a marker of apoptosis, were strongly related to all assayed mitochondrial parameters. Figure 2D reports that the %Δ of apoptotic cells compared to untreated cells was higher after exposure for 1d to both types of MWCNTs and for both doses. In the following subcultures, apoptotic cells decreased over time, especially in cells treated at the tested dose of MWCNTs. This confirms the adaptability of alveolar cells to the exposure conditions. However, they were not totally restored to baseline conditions. The results of the Pearson test showed highly significant correlations between apoptotic cells and cytochrome c release, Δψm, and mPTP opening, respectively (p < 0.01).To assess cellular dehydrogenases, included succinate dehydrogenases (SDH in mitochondrial complex II), the time course of MTT reduction was measured in cells exposed to MWCNTs (Figure 3A). In the positive controls exposed for 1 day to both MWCNTs, the activity of cellular dehydrogenases was 60% below (p < 0.01) that of untreated cells. Instead, the absorbance values in cells exposed for 1 day at 2 μg mL−1 was not significantly different from the negative control, and dehydrogenase activity was equal to 67.1% and 74.5% for pristine and oxidised nanotubes, respectively. Over time, we observed a progressive increase of the absorbance values, highlighting an adaptation of the sub cultured cells to the steady exposure. Despite the increased absorbance values at 36 days, which was higher in the f-MWCNTs-treated cells, the activity of cellular dehydrogenases did not overlap the control cells. This emphasizes that MWCNT exposure, even at a low dose, caused cellular damage, albeit of a limited extent. Considering that the MTT values were strongly and inversely related to apoptotic cells and to Trypan blue values (p < 0.01), it is highly plausible to believe that the reduction of dehydrogenase activity was not due to a mere enzymatic inhibition, but rather to a reduction in the number of viable cells.Figure 3B reports the ROS levels in exposed cells compared to control cells. As previously detected in the same cell model, a stronger pro-oxidant effect was caused by pMWCNTs than fMWCNTs, confirming what was observed in the short-term exposures to high doses. In comparison to positive controls after 7 and 21 days, the A549 cells exposed at 2 µg mL−1 were able to almost completely counteract the oxidative stress induced by f-and p-MWCNTs, respectively.Conversely, ROS production remained higher in the positive control sample (20 µg mL−1) of pristine nanotubes than in untreated cells, and the observed values were on average 25% higher than the control after 36 days.The cellular proliferation index was used for an overview of the MWCNT-cell interactions. In comparison to untreated cells, whose average values of proliferation index in the interval assayed was arbitrarily set at 100; the values at 2 µg mL−1 were 90.9 (±5.4) and 93.8 (±6.2) for p-and f-MWCNTs, respectively. The values in positive controls were further decreased and equal to 75.1 (±6.5) and 78.4 (±5.9), underlining the harmful effects of the engineered nanoparticles.The production of safer nanofibers is extremely important considering the ease with which lightweight nano-sized MWCNTs aerosolize, increasing the risk of developing pulmonary disorders after their inhalation [3,8,15]. Once inhaled, the high hydrophobicity of the CNTs promotes their interaction with the plasma membrane surface. The internalisation pathways of nanotubes include the energy dependent endocytosis (i.e., pinocytosis or, only in specialised cells, phagocytosis) and passive diffusion. Due to the high length to diameter ratio, the latter pathway allows MWCNTs to efficiently cross the phospholipid bilayer of cell membranes. As established by several studies, both mechanisms do not alter membrane integrity [37,38,39,40]. Unlike the endocytosis-mediated internalisation of CNT tangles, passive diffusion reduces cytotoxicity because it does not cause lysosomial content leakage by overloading endosomes [12]. However, regardless of the internalization process, nanotubes cause oxidative damage in cell compartments.The results were obtained by simulating sub-chronic exposure in a work environment and they highlighted the efficiency of homeostatic mechanisms to counteract over time MWCNTs-induced ROS overproduction. However, the mitochondria were not entirely exempt from the harmful effects of the engineered nanoparticles, which caused apoptosis in a portion of exposed cells. Mitochondria are dynamic bioenergetic semiautonomous organelles that execute a myriad of functions pertaining to cellular metabolism and homoeostasis. In addition to cellular energy generation via oxidative phosphorylation (OXPHOS), they play a central role in calcium homoeostasis, initiation of caspase-dependent apoptosis, cellular stress response, sulphur metabolism, and biosynthetic processes [41,42,43,44,45,46,47]. The paramount importance of mitochondrial integrity for the proper functioning of the OXPHOS system is well known. It comprises five multimeric enzymes (complexes I to V) that are incorporated into super-complexes to reduce ROS levels and two mobile electron carriers (coenzyme Q 10 and cyt c) [48]. Despite the clear tendency to restore full functionality, our results showed that the MWCNTs-induced mitochondrial impairment persisted over time not only in the positive controls but also at the dose to which workers could conceivably be exposed. Considering the strong association between depolarization and electron transport impairment, this effect was highlighted by the transmembrane potential values.The superimposable effects of both CNTs could be imputable to excessive lengths with regard to the p-MWCNTs (10–20 μm vs. 200–1000 nm) and to higher surface reactivity with regard to the f-MWCNTs [12,19]. Even if the presence of carboxyl groups enhances water dispersibility and causes a reduction in the length to diameter ratio of CNTs, making them more biocompatible, these effects are nullified by the acid-induced erosion in the graphene external layers. This increases the surface reactivity and, consequently, cellular toxicity [18].Mitochondria are signalling organelles that constantly regulate the production of energy according to the needs of the cell. At the same time, the mitochondria are able to manage cell behaviour by retrograde signalling to other cell compartments. These signals include mitochondrial ROS, transmembrane potential and calcium fluxes across the mitochondrial membrane as well as AMP/ATP and NAD+/NADH ratios as bioenergetic and redox signals, respectively. While mitochondrial ROS (mtROS) overproduction and/or hiROS (high ROS, such as hydroxyl radicals or peroxynitrites) are liable for cellular pathology by directly damaging biomolecules, physiological levels of mitochondrial loROS (low ROS, such as superoxide or hydrogen peroxide) modulate gene expression and cell survival. Despite the positive recovery results observed in the present study, sub-chronic exposures to our homemade MWCNTs did not allow a restitutio ad integrum of mitochondrial functionality. Considering the functions of the mitochondria as summarised briefly above, the values of Δψm, equal to 75% and 83% for p- and f-MWCNTS, suggested a metabolic impairment in the exposed cells. Those findings were confirmed by the results of cytochrome c, which highlighted the way in which mitochondrial dysfunction unavoidably affected cells by triggering a mitochondrial apoptotic pathway. The dissociation of cyt c from the inner mitochondrial membrane, followed by the release of cyt c from the intermembrane space to the cytosol through the outer membrane, is regulated by a mechanism involving two pro-apoptotic Bcl-2 family proteins, Bax and Bak, which are able to increase mitochondrial outer membrane permeabilization [31,32]. This pathway was clearly underlined by the release of cyt c that was recorded in the cytosolic fraction devoid of the mitochondria of MWCNTs–treated cells and, downstream, by the increase of apoptotic cells as shown by the outsourcing of phosphatidylserine, which is a marker of apoptosis.The pathogenic mechanism was further supported by the results of the calcein test performed to assess mPTP opening. mPTP is a key participant in mitochondrial apoptosis that has catastrophic consequences on the fate of cells, leading to the release of Ca2+ from the mitochondrial matrix and causing energetic collapse by depletion of adenosine triphosphate (ATP). Furthermore, mPTP, formed by voltage-dependent anion channel (VDAC), adenine nucleotide translocator 1 (ANT1), and cyclophilin-D (CypD), a peptidyl–prolyl–cis, trans-isomerase [48], causes the release of cytochrome c that triggers an apoptotic pathway via caspases activation. During stress conditions, such as exposure to engineered nanoparticles, the bond in the inner mitochondrial membrane (IMM) between CypD and ANT1 causes mPTP opening [49]. Even if we observed a partial recovery over time, the MWCNTs-induced mitochondrial impairment persisted and prevented the restoration of physiological conditions.Several mitochondrial quality control mechanisms (QC) form a hierarchical network of interacting pathways that act at the molecular, organelle, and cell levels to maintain a “healthy” mitochondrial population in each cell. At the molecular level, ROS scavenging detoxifies superoxide and H2O2 through nuclear encoded Mn superoxide dismutase (MnSod) and glutathione peroxidase or peroxiredoxin. Considering the dual role of ROS reported above, a fine-tuned regulation of cellular ROS level is required in order to hinder damage (by abstraction of electrons) to proteins, lipids, and nucleic acids when threshold levels are overwhelmed. In the integrate network of mitochondrial quality control, this first protective line formed by ROS scavenging would seem sufficiently efficient after sub-chronic exposure to MWCNTs. The decreased levels of ROS suggest that the enzymes and the small molecules required in the scavenging process, such as glutathione and thioredoxin, were effectively regulated, counteracting the over production of ROS that boosts the pro-oxidant effect of engineered NPs.Conversely, our data show how the MWCNTs-induced damage exceed the threshold within which downstream mechanisms of mitochondrial QC are able to maintain only functioning mitochondria. These QC mechanisms include repair of damaged components to restore their function and/or mitophagy (i.e., autophagy of whole dysfunctional organelles), and they control the balance (i.e., mitohormesis) between mitochondrial fusion and fission from which, thanks to biogenesis, a fully functional population of mitochondria can be generated, ensuring cell survival [50,51,52]. As suggested by the mitochondrial depolarization, the impairment of enzymatic activity and the increased mPTP opening, sub-chronic exposure to our highly purified MWCNTs causes a saturation of mitochondrial quality control mechanisms in alveolar cells. The presence of dysfunctional mitochondria affected the cellular proliferation index underling the potential pathogenic effects of MWCNTs-cell interaction in the onset of chronic obstructive pulmonary disease in the occupationally exposed subjects.These results, obtained by simulating sub-chronic exposure to MWCNTs in a work environment, reduce the alarming picture that has emerged from many short-term studies that used unrealistic CNTs concentrations. Over time, ROS overproduction in alveolar epithelial cells was minimized due to the excellent mechanisms of homeostasis. However, even if the oxidative stress induced by the MWCNTs tested dose was temporary, the mitochondria were not entirely exempt from the harmful effects of the engineered nanoparticles. Considering the key role of mitochondria in managing the vital functions of cells, it is useful to underline the tissue damage that can arise following exposure to these synthetic nanoparticles.Conceptualization and Methodology, A.D.P., D.I. and A.P.; Validation Formal Analysis and Investigation G.V., A.F. and M.C.; Resources, P.L. and V.L.F.; Data Curation, A.D.P., G.V. and A.F.; Writing-Original Draft Preparation, A.D.P.; Supervision, A.D.P. This research received no external funding.The authors declare no conflict of interest.Representative phase contrast microscopy images of A549 semiconfluent monolayers to assess MWCNT-cell interactions. (A–C) control cells and cells exposed for 1 day to p- and fMWCNTs (20 µg mL−1). (D–F) control cells and cells exposed for 36 days to p- and fMWCNTs (20 µg mL−1). The images of cells treated with 2 μg mL−1 are not shown because the interactions were barely detectable and the cellular morphology was superimposable to that of the control cells. Table reports the percentages of dead cells, as assessed by the trypan blue exclusion test.Assessment of mitochondrial function and apoptosis on MWCNT treated cells over time. (A) Representative result of western blot analysis performed to evaluate the expression of PDK 1 and to assess cyt c release (ctr: control cells). In the table, the results are expressed as relative fold changes. (B) Time course of R123 emission values that were used to assess the Δψ m in A549 cells. The results are expressed as relative fold change in comparison to control cells. (C) Results of the calcein-AM/cobalt method that was used to assess MWCNTs-induced mPTP opening during the examined interval. (D) Assessment of MWCNT-induced apoptosis by analysis of the outsourcing of phosphatidylserine over time. The results are expressed as percent change in comparison to control cells. All data are presented as mean ± SEM based on at least three independent experiments.Time course of cellular enzymatic activity and ROS production in MWCNT- treated cells. (A) Formazan absorbance during the assayed interval for the assessment of cellular dehydrogenases. The results are expressed as percent change in comparison to control cells. (B) DCF emission values during the assayed interval to asses ROS levels. All data are presented as mean ± SEM based on at least three independent experiments.

Med-MDPI/ijerph_3/ijerph-16-05-00793.txt

@@ -0,0 +1,2 @@
+The analysis of the postural attitude of workers during the interaction with workstation’s elements and working environment is essential in the evaluation and prevention of biomechanical overload risk in workplaces. RULA (Rapid Upper Limb Assessment) and REBA (Rapid Entire Body Assessment) are the two easiest methods for postural risk assessment in the workplace. Few studies investigated postural risk in forestry sector with regard to human–machine interaction, in particular manually fed wood-chippers. The aim of this study was to evaluate the postures assumed by an operator during the manual feeding of a wood-chipper, and to compare RULA and REBA, in order to identify the more effective and appropriate method for the assessment of the risk of biomechanical postural overload. The results pointed out several postural issues of the upper limbs, and showed that RULA is a more precautionary method to protect operator’s health during the targeted tasks. Implications to improve the human–wood-chipper interaction are discussed.The analysis of the postural attitude of the worker during the interaction with workstation’s elements and working environment is essential in the evaluation and prevention of biomechanical overload risks in workplaces [1]. Awkward working postures may decrease the workers’ concentration and increase accidents frequency and biomechanical overload [2,3,4,5,6], giving rise to musculoskeletal disorders in the different body regions involved, as at the main limb joints level and the vertebral column [7].Some standards [8,9,10] dealing with the biomechanical overload caused by incongruous static and dynamic postures have been developed to define the risk assessment methods to evaluate postural load referable to activity and workstation characteristics and to human–machine interaction. They are referenced in international [11] and national [12] legislations aimed at protecting workers’ health and safety.RULA (Rapid Upper Limb Assessment) [13] and REBA (Rapid Entire Body Assessment) [14] are two easy methods for occupational postural risk assessment. Indeed, previous studies [15,16] showed that observational methods are considered effective in the assessment of biomechanical work-related overload, having the advantage of being more versatile and less expensive in terms of time and resources required compared to objective laboratory measures. Both RULA and REBA allow to obtain a numerical index that represents the quantitative value of the risk at which the worker is exposed during the targeted work activity and to derive the priority level of intervention and the actions needed. The RULA method is suggested for the identification of postural disorders of the upper limbs, of the neck and of the back in relation to the muscular action and external loads applied to the body. The REBA method is applied to identify postural disorders of the whole body, in relation to the muscular action, to the external loads applied to the body and to the type of grip. They are referenced in the international standard for occupational risk assessment [9] and cited among the selected tools for Work-related Musculoskeletal Disorders (WMSDs) prevention according to International Ergonomics Association (IEA) and World Health Organization (WHO) [17]. These methods are also widely applied in several working contexts, mainly industrial work activities (secondary sector) and those producing services and goods (tertiary sector), characterized by a precise standardization of tasks, geometries, gestures and relative execution frequencies that allow a systematic and controlled forecasting and quantification of the biomechanical overload risk. Furthermore, the two methods have been combined and compared to assess postural risk in industry [18,19,20], construction [21], supermarkets [22], hospital and dental sector [23], work with video terminals [24], waste collection activity [25], for firefighters and emergency medical technicians [26], and artisans [27,28,29] and sawmill activities [30].RULA and REBA have been adopted a few times in the agricultural [31,32] and forestry sectors [33,34] because the evaluation of the biomechanical overload risk in the primary sector activities is more difficult; due to the large variability of the tasks the operators have to perform, depending on crops, operations (seeding, weeding, pruning, harvesting, etc.), the machinery and tools adopted, the, sometimes extreme, weather conditions they have to be carried out with, and the daily and seasonal exposure, as well as the lack of a strict standardization of the work in general [35,36]. Interaction with machines, tools and environments in agriculture, and particularly in forestry, requires therefore a particular attention in the application of risk assessment methods conceived for other contexts, or the formulation of specific methods that consider the distinctive characteristics of these activities [37].Kundu and Gaur [38] compared RULA and REBA in a study addressed at investigating how much these techniques were appropriate for evaluating the postures assumed by agricultural workers. They highlighted some shortfalls in using these techniques to study risk factors associated with agricultural field operations and suggested to add some factors as posture duration, field condition, environmental factors and nutritional status to better assess the occupational risks and possible remedies, especially for agricultural work in the fields.Besides RULA and REBA, other methods have been used to assess the postural risk in manual agricultural activities. For example, Kong et al. [39] compared RULA, REBA, OWAS (Ovako Working Posture Analyzing System [40]) and ALLA (Agricultural Lower Limb Assessment [41]) for various agricultural tasks. They found that ALLA better estimated the risk because it identifies critical issues that the other methods do not analyse and, therefore, it always returns a higher level of risk, compared to the other methods. Ojha and Kwatra [42] combined REBA and VAS (Visual Analog Scale) methods in rice cultivation manual operations. REBA indicated postural load and suggested interesting corrective measures. However, in both studies [39,42], the object of the analysis was a manual activity, while, in developed countries, most of the farming operations require the use of machinery and workers spend many hours in interacting with them, making the investigation of the human–machine interaction more relevant [43].Among the studies specifically dealing with the human–machine interaction in agriculture, RULA is the most frequently adopted method. Vyavahare and Kallurkar [44] used RULA to assess postures assumed during the interaction with agricultural machines such as maize dehusker-sheller. The study analysed and evaluated the risk of various key postures, such as squat, forward/lateral bending, hands flexion/extension, wrist/spine twisting. Putri et al. [45] used RULA to study the use of thresher machines for threshing rice plants. In the first study, RULA helped to optimize a digital human manikin posture, resulting in precise risk assessment and better designed and widely accepted products and workplaces, while, in the second study, RULA provided indications to redesign the machine to reduce injuries and musculoskeletal overload caused by the mismatch between the dimensions of the engine and farmers’ postures and dimensions.In the forestry sector, the studies report postural risk evaluation mainly referred to manual tools. Gallo and Mazzetto [34] applied different methods including REBA and OWAS to evaluate the postural risks from the cutting operations with chainsaw, showing their good applicability for the assessment of WMSD in forestry. The comparison between OWAS and REBA showed that REBA has a higher level of detail of assessment because it considers the angles between body segments and extremities as the wrist, the neck, the elbow and the shoulders (parameters that are not considered by OWAS method) for both sides of the body and, furthermore, it assesses the type of handle coupling and the characteristics of the performed activities. REBA results to be suitable for suggesting interventions to be performed to decrease musculoskeletal overload, even though it has not been developed specifically for forestry operations and it presents some weaknesses during the assessment, as the lack of coded postures as the kneeling posture.Forestry is recognized as a highly hazardous industry [46] and working postures are one of the most investigated risk factors for workers’ safety, even though the postural aspects in the human–machine interaction have been under investigated in this sector [47,48]. With regard to machinery, manually fed wood-chippers are one of the most widely used machines in forestry, agricultural, landscaping and urban tree maintenance to reduce the volume of woods for following disposal or re-use in bio-energy production [47]. The attention to this machine is rising because of the significant increasing interest in biomass production as biofuel [49] and because they are often involved in fatal and non-fatal injuries [50]. Data on the occurrence of accidents provide an objective index of the danger of machinery as well as a valid reason for identifying the most critical features of these machines.In the United States, between 1992 and 2002, Struttman [50] reported 31 fatal and 2042 non-fatal accidents involving wood-chippers. The estimated social cost of these fatal accidents were estimated in 2003 to US$28.5 million; furthermore, an in-depth analysis disclosed that 58% of these fatal accidents involved groundskeepers and machinery operators in the forestry and agricultural sectors [50]. Analyses of the non-fatal accidents showed that 60% of the accidents caused immediate injury or amputation of parts of the upper-body limbs. For 25% of these injuries, the victims were unable to return to work for periods of up to 30 work days [51]. Further studies indicated that more than one-third of these accident victims had less than 11 months of experience in that particular job [50].In the further period between 2008 and 2018, statistics from the United States Occupational Safety and Health Administration of Department of Labor [52] reported 56 incidents during chipping operations: 64% of the accidents dealt with feeding operations and 36% of them were fatal for the operators. From the same statistics, the accidents not related with feeding operations of wood-chippers had a lower rate, 16%, of fatal accidents.Accurate statistics about accidents with wood-chippers are not available from most of the European Union countries [53]. Only the French Ministry of Agriculture, Agrifood, and Forestry reported that at least one severe accident related to the use of wood-chippers occurs every year in France and that, in most of the cases, the operators involved were young apprentices [54], one of the category of users for which warnings to eliminate hazardous behaviours when intrinsically safe machine are used, are less effective [55,56].Based on the literature review, the aim of this study was to identify the more effective and appropriate method between RULA and REBA for the assessment of the risk of biomechanical postural overload, evaluating the postures assumed by an operator during the manual load of a wood-chipper, in a controlled experimental setting. The study further discusses which of the two methods is more precautionary in evaluating postural risk and reports higher risk indices, for the specific activity considered.For the present study, a small manually fed wood-chipper has been selected since: (1) small-size wood-chippers are widespread among forestry, agricultural and urban green maintenance operators; (2) similar types of wood-chippers are used for occasional and accessory operations to the main activity [47]; (3) often experienced, but not specialized, operators are involved in the wood-chipping operations and this category of operators are known to be the most exposed to safety risks [57]. As a consequence, the manual feeding seems to be critical both in terms of safety and health risks for the operators [58,59].The manually fed wood-chippers generally consist of a feeding system, a chipping units based on rotating knives (drum or disc mounted) and on a discharge system. Alternative solutions for the comminution devices are available for production of quality wood chips for fuel [60]. In the case of manually-fed wood-chippers, the operators feed small trees, part of trees and/or branches into the infeed chute by hand with a following risk of biomechanical postural overload, which is typical of manually-loaded wood machines [61].The study was carried out with a manually fed wood-chipper model Tirex, made available by the manufacturer Peruzzo (Peruzzo Srl, Curtarolo, Italy). The machine was connected to the rear three-point hitch of a 55 kW tractor and was powered by the rear Power Take-Off (PTO) (Figure 1). The feeding system consists of a feeding hopper with an infeed chute above 600 mm from the ground level and with a width of 1200 mm at the external edge. The feeding system consists of two horizontal rollers electro-hydraulically controlled by a load-limiting mechanism and located at the end of the infeed chute at a distance of 1200 mm from the external infeed chute limit, in accordance with EN 13,525:2005 + A2:2009 [62]. The machine is equipped with an electro-hydraulic feed control bar fixed on the bottom and along the two sides of the infeed chute. The chipping unit consists of a flywheel (diameter 620 mm, mm thickness 30 mm) rotating at 1500 rpm and with a torque of about 33 Nm. The flywheel is equipped with four knives (thickness 25 mm and length 200 mm) while an anvil is fixed to the frame of the machine. The manufacturer claims the machine is able to chip wood stems with a diameter up to 180–200 mm.The study was conducted at the Azienda agraria sperimentale “L. Toniolo” of the Università degli Studi di Padova (Padova, Italy). Since we were interested in controlling as many variables as possible to better compare RULA and REBA, the evaluation of the differences in risk levels between the two methods was carried out by keeping the machine and participant characteristics constant, while changing the log size. To reduce the inter-individual variability in the adoption of the targeted postures, one single participant was involved in the study. He was a not-specialized operator, with no previous history of either occupational accidents or musculoskeletal disorders. Specific measures were defined for the logs to be used in the tests, with regard to log length and diameter, and an acceptable mass, ranging between 2 and 10 kg, to be manually handled by the operator (Figure 2).Plane-tree wood was used for the tests, preparing both logs of different length and diameters, and branches of more variable dimensions, in order to observe if and how much the size of the wood induced different postures and gestures, which may be associated with different levels of postural risk. The logs were prepared in three different lengths (1 m, 1.5 m, 2 m) and two diameters (65 mm and 135 mm). For each length/diameter combination and for the branches, three elements were prepared to perform three repeated tests. Therefore, six tasks involving logs plus one task regarding branches were performed. The operator performed each task three times to evaluate any intra-individual difference in postures and gestures with respect to the manipulated element. Thus, overall seven tasks were performed for a total of 21 loadings (six tasks for the logs, each repeated three times, and one task for the branch repeated three times). The 21 loadings were randomly performed.The research protocol was approved by the Research Advisory Group of the Institute for Agricultural and Earthmoving Machines (IMAMOTER) of the National Research Council of Italy (CNR) on 28 February 2018.The operator was video recorded and photographed while feeding the machine. One video camera was placed in front of the infeed chute while a second one recorded images from the side of the infeed chute. During the analysis of the recorded videos, every time a log or a branch was loaded, the frame-by-frame vision was carried out until the most critical position was detected. The video analysis was performed with Kinovea software, an open-access video analysis software available online reproducing the video in slow-motion. Kinovea is a valid and reliable method to perform motion and postural analysis [63]. It allows for detecting body and its districts’ angles during the single posture observations: angles were measured according to the reference axes reported in RULA and REBA methods to obtain the corresponding score.The study investigated the human–chipper interaction comparing two observational methods of biomechanical overload postural risk: RULA method [13] and REBA method [14].In RULA method [13], the body is divided into different parts gathered into two groups, A and B: group A includes arm, forearm and wrist of the right and left limb; group B includes the neck, the trunk and the feet. The method consists in assigning a score to each segment depending on the posture taken and it allows to obtain two distinct scores (Scores A and B) through the use of numerical tables or spreadsheets; these scores represent the level of postural load of the musculoskeletal system, determined by the combination of the postures of the whole body. Muscle use and force scores are then added to Scores A and B to obtain two new scores (Scores C and D) that, through a third table or a spreadsheet, allow for obtaining the final score, or Grand Score.Based on the appropriate combination of scores, the final score can range between 1 and 7. The final score is related to four levels of action and four levels of risk. Scores, actions and risk levels for RULA are summarized in Table 1:Action level 1: Low risk level. A score of 1 or 2 indicates that posture is acceptable if it is not maintained or repeated for long periods.Action level 2: Medium risk level. A score of 3 or 4 indicates that further investigation is needed and changes may be required.Action level 3: High risk level. A score of 5 or 6 indicates that investigation and changes are required soon.Action level 4: Very high risk level. A score of 7 indicates that investigation and changes are required immediately.Action level 1: Low risk level. A score of 1 or 2 indicates that posture is acceptable if it is not maintained or repeated for long periods.Action level 2: Medium risk level. A score of 3 or 4 indicates that further investigation is needed and changes may be required.Action level 3: High risk level. A score of 5 or 6 indicates that investigation and changes are required soon.Action level 4: Very high risk level. A score of 7 indicates that investigation and changes are required immediately.In REBA method [14], the body is divided into different segments divided into two groups: the first one includes neck, torso and legs; the second group is composed by arm, forearm and wrist without distinction from the right or the left one. The method consists in assigning a score to each segment of the body according to the posture taken and, using numerical tables or spreadsheet, it allows for obtaining two different scores that represent the level of postural load of the musculoskeletal system determined from the combination of the whole body postures. The two scores should be respectively added to the grip score and to the load and strength score to get two new scores (Scores A and B). From the use of a third table or spreadsheet, it is possible to obtain Score C, which, added to the activity score, allows for obtaining the final score, or Grand Score. The final score can range between 1 and 15 and it is related to five levels of action and five levels of risk. Table 1 reports scores, actions and risk levels for REBA:Action level 0: The risk is negligible so no action is required.Action level 1: Low risk level. The final score between 2 and 3 indicates that changes may be required.Action level 2: Medium risk level. The final score from 4 to 7 indicates the need for measures and further analysis.Action level 3: High risk level. The final score between 8 and 10 indicates the need for an intervention and a change in a short time.Action level 4: Very high risk level. The final score from 11 to 15 indicates that an action is immediately required.Action level 0: The risk is negligible so no action is required.Action level 1: Low risk level. The final score between 2 and 3 indicates that changes may be required.Action level 2: Medium risk level. The final score from 4 to 7 indicates the need for measures and further analysis.Action level 3: High risk level. The final score between 8 and 10 indicates the need for an intervention and a change in a short time.Action level 4: Very high risk level. The final score from 11 to 15 indicates that an action is immediately required.For each task, three risk indices were obtained for each of the two methods, RULA and REBA.The goal was to analyze the highest level of risk calculated, considering the variability of the execution of the gesture, evaluated three times separately. Among the results obtained, the highest value, among the three for each task, and then the highest between the right and left side was selected.The results of the two scores calculated for each task with RULA and REBA were compared to highlight any tendency of any of the two methods to over or underestimate the risk. This was done both on the basis of risks category, considering the action level—color code (Table 1), and the values of normalized indices.In order to compare the differences between REBA and RULA scores, it was necessary to normalize the absolute values as the two methods are based on different scales: RULA has four risk levels that categorize scores from 1 to 7, while REBA is based on five levels of risk that categorize scores from 1 to 15. The use of standardized scores to compare different methods for postural risk assessment has been adopted previously in similar analyses [64,65], thus we normalized the score values ranging from 0 to 1, applying the following formula to each worst value obtained for each task by the two methods:
+        normalized value = [score obs − score value min]/[score value max − score value min].The risk assessment carried out by applying RULA and REBA (Table 2) showed a medium-high level of risk for all the tasks. None of the tasks reported a score referable to a negligible or low risk as well as a very high risk level that would require an immediate intervention.Regarding the differences between the risk indices of each task, it can be seen that, based on the dimensional differences of the handled logs, the risk level is limited in the case of logs with longer length and smaller diameter. This could be due to the posture induced by small length logs which could cause a higher postural overload because of the greater proximity to the machine that could determine worse joint angles. The highest indices were observed in the manipulation of short and larger diameter logs (Task 2), while the lower levels of risk indices occurred in the manipulation of the branches (Task 7). The comparison between the results obtained with RULA and REBA, based on the resulting action level, highlighted a good overlapping between the two methods, with the exception of Task 3, related to wooden elements of an intermediate size (1.5 m) and small diameter (65 mm). In this case, the RULA carried out a more precautionary assessment as it gave an index that fell into a higher action level compared to the corresponding level calculated with REBA. In the comparison of the normalized numerical indices, all of the RULA scores were higher than the corresponding REBA scores for all of the tasks (Table 3). The application of RULA showed to be more precautionary than REBA for all the tasks, as it returned a higher score value in each loading.The present study showed that RULA and REBA are two effective methods for assessing the postural overload determined by incongruous postures adopted in the act of manually loading the wood-chipper with logs of various diameters and lengths. The two methods appeared to be both effective and suitable for the identification and quantification of the level of postural risk using a manually fed wood-chipper, as they highlighted similar levels of urgency of intervention, analysis and modification actions. The high congruence between RULA and REBA evaluations confirmed similar results from other investigations in the industrial context [66]. However, the present study showed that RULA tended to be more precautionary, giving a higher risk index, for all the tasks, consistent with previous studies comparing the two methods in the industrial sector [18]. This underestimation by REBA was also found in KOSHA’s research [67] in different sectors like ship building, automotive, electronics, general manufacturing, and service industries.Comparing the action level in which the calculated indices fall, and, therefore, the corresponding severity and urgency of intervention, RULA and REBA suggested an almost identical level and, in only one case (Task 3, intermediate size and small diameter wooden elements), RULA returned a higher action level than REBA (Table 2). Task 3 is also the one which reported the highest difference between normalized RULA and REBA scores. Comparing normalized scores, RULA always resulted in being more precautionary as it returned higher values for all the tasks. The evaluation of the branches’ manipulation task presented the most contained difference between the two normalized values, but also in this case the risk estimated by RULA, even if slightly, prevailed. RULA appeared to be constantly more precautionary in the risk assessment for all of the tasks considered, both in medium or high risk conditions. RULA also seemed to be more precise as it better highlighted the differences in the level of risk exposure between the manipulation of logs and branches. This could be due to the presence of different postural issues of the upper limbs, such as radial or ulnar deviation of the wrist, rotation of the wrist and movements performed across the body or out to the side, which play an important role in the postural risk assessment of these tasks and are not considered by REBA. About the branches loading, RULA normalized score returned a value of 0.33, which was the lowest of all tasks. REBA did not seem to be influenced by these differences among logs and branches manipulation. The minimum normalized REBA score was 0.29, and it was the same value for both the manipulation of the branches (Task 7) and for the manipulation of logs of Tasks 3 and 5.RULA highlighted how biomechanical workload in the interaction with the wood-chipper is limited to the upper part of the body and it determines an important involvement especially of the wrist and forearm, and on the rachis due to flexion of the chest and postural asymmetries. REBA scores, obtained evaluating the lower limb biomechanical risk in greater detail, had no higher values than RULA in all the tasks, confirming the absence of serious criticalities for the lower limbs in wood-chipper manual feeding.According to the results of the present study—even though they were obtained by observing one single operator and should not therefore be considered as conclusive—RULA is more precautionary, probably because of the more precise workload evaluation of the upper limbs and trunk. RULA considers in a much more limited way the postural and workload conditions of the lower limbs compared to REBA, but, nevertheless, in the present investigation, REBA did not give higher risk estimation values, which means that, during the loading of the wood-chipper, there was not any evidence of risk to the lower limbs (otherwise, REBA would have highlighted it with a higher risk index). Therefore, RULA may be a more adequate method for the assessment of postural risk of interaction with the machinery when exposure to risk of the lower limbs is less relevant. This consideration confirms results from previous studies which reported the effectiveness of RULA in evaluating the interaction with agricultural machines [44].The analysis pointed out different RULA scores due to the different lengths and diameters of logs, with shorter ones determining incongruous postures. This consideration opens new perspectives for the analysis of human–wood-chipper interaction stressing the importance of avoiding short logs during the cutting operations, when they are intended to be chipped. In this way, a postural disadvantage occurs even if the mass is reduced. Future investigation should be addressed at investigating a larger number of operators regarding how much log mass, besides the length, determines incongruous postures (high RULA score) and physical effort (applied force and perceived fatigue). The results of such studies could provide suggestions about an “optimal” cutting length for logs, determining a better postural conditions for operators of manually fed wood-chippers. Indeed, in ergonomics, it is not enough to concentrate on the design of the machine, but it is important to intervene in the different components of human–machine interaction through a holistic approach, as widely documented in the literature and recommended by international standards [68,69,70,71].Besides the attention paid to safety and technical characteristics of the forestry machine, this study highlighted the importance to inform and train the operators to perform gestures, postures and activities in accordance with the ergonomic principles. In this way, in addition to optimize the human–machine interaction, it can be possible to intervene on the organization and management of the whole activity. This integrated approach may contribute to risk reduction and enhance the system productivity and operator’s wellbeing.Agricultural and forestry operators interact with a variety of tasks and machinery, which can require ad hoc methods for the assessment of postural risk. The present investigation showed that the RULA method is suitable for the evaluation of postural overload in the human–machine interaction related to a manually fed wood-chipper, more than the REBA method, because it showed indices that corresponded to a higher level of risk for all the tasks observed, independently of the shape, size, mass of the wooden material and, therefore, it would be a more precautionary method to protect operator’s health. Further considerations about the postural risk of manually-loaded wood-chippers should focus not only on the posture of the operators, but also on the safety standards imposed by the safety regulations (for instance, the former EN 13,525:2005 + A2:2009 [62]). The extended application of RULA in real and differentiated agro-forestry conditions will allow for assessing postural risks and physical effort to improve human–machine interaction and operators’ wellbeing in the wood-chipping activity.M.M.C., A.G., F.C., A.C., E.C. and S.G. conceived and designed the study; A.C. and S.G. collected the data; M.M.C. and A.G. analyzed the data; M.M.C., A.G., F.C., A.C. and S.G. interpreted the data; M.M.C., A.G. and F.C. wrote the paper; A.C., E.C. and S.G. critically reviewed the paper; M.M.C., A.G., F.C., A.C., E.C. and S.G. gave final approval.This study was supported by the “Protection of agricultural machinery operators from crush, entanglement, shearing” (PROMOSIC) project, funded by INAIL, the Italian National Institute for Insurance against Accidents at Work.This research did not receive any specific grant from funding agencies in the public, commercial or not-for-profit sector.The authors declare no conflict of interest.The manually fed wood-chipper used during the test.Example of tasks to be performed during the tests.RULA and REBA scores with the respective action level.RULA: Rapid Upper Limb Assessment; REBA: Rapid Entire Body Assessment.Comparison between RULA and REBA risk evaluation (worst case for each task calculated with each method) for the seven tasks considered with details of the dimensional characteristics of the manipulated wooden elements.1 Note. Tasks 1–6 represent the loading of logs with three different lengths (1 m, 1.5 m, 2 m) and 2 diameters (65 mm and 135 mm); Task 7 represents the loading of branches.Comparison between the normalized scores of RULA and REBA indices for the seven tasks.1 Note. Tasks 1–6 represent the loading of logs with different lengths and diameters; Task 7 represents the loading of branches.

Med-MDPI/ijerph_3/ijerph-16-05-00794.txt

@@ -0,0 +1 @@
+The Chinese Family Planning (FP) programme mainly focuses on married couples, and young unmarried women have limited access. This cross-sectional study aims to identify risk factors related to repeat abortions in Chinese adolescents receiving abortions. Data were collected using a questionnaire for all women seeking abortions within 12 weeks of pregnancy during a period of 2 months in 297 participating hospitals randomly selected across 30 provinces of China in 2013. Only the adolescents (younger than the minimum legal married age of 20 years) were included in this study. Of the 2370 adolescents who were receiving abortions, 927 (39%) were undergoing repeat abortions. The primary reason for the current unintended pregnancies was non-use of contraception (68%). Adolescents receiving abortions who had an increased risk of repeat abortions were those who had children (OR 2.57, 95% CI 1.80–3.67), those who resided in a middle-developed region (OR 1.81, 95% CI 1.30–2.50), those who resided in a relatively poor region (OR 2.40, 95% CI 1.78–3.23), and those who had used contraception during the 6 months preceding the survey (OR 1.38, 95% CI 1.12–1.71 for condom use). The occupation as a student was a protective factor for adolescents (OR 0.64, 95% CI 0.50–0.83). Adolescents should be offered equal access to FP to that of married women in China to reduce unintended pregnancies and repeat abortions. Correct and consistent contraception practice should be promoted.Unintended pregnancy is a global public health problem and is an extremely common occurrence in women’s lives [1]. For example, about half of all pregnancies in the United States were unintended [2]. Induced abortions are commonly used to end unintended pregnancies. A recent study estimated that 35 abortions occurred annually per 1000 women aged 15–44 years worldwide in 2010–2014 [3]. In addition, repeat abortion is another public health problem which cannot be overlooked and presents adverse effects on the reproductive and mental health of women, particularly for the young and unmarried women [4,5]. The identification of risk factors associated with repeat abortions and the development of public health initiatives to reduce the risk of repeat abortions recently witnessed growing international interest [6]. A recent systematic review by McCall et al. identified several determinants of repeat abortions, such as increased age, parity, marital status, contraception use at the time of conception and previous history of abuse or adverse life events [7].In China, approximately six to ten million induced abortions have been reported annually since 2000 [8]. With dramatic social changes associated with rapid economic growth and reform over the past 30 years in China, traditional attitudes towards sex and sexual behavior have changed, and premarital sex has become more acceptable among young unmarried people, including adolescents [9,10,11]. Thus, unintended pregnancies and subsequently induced abortions among them increase. A previous study indicated that about 12% to 32% of unmarried women had become pregnant in China, among whom 86% to 96% had induced abortions [12]. A study conducted in 49 universities across seven cities in China reported that 31.8% of unmarried female students exposed to sexual activities in China had experience of unintended pregnancies, among whom 83.9% chose to terminate their pregnancies through induced abortions [13]. Liu et al. reported that 55.3% of 600 young unmarried women aged less than 24 years who were receiving abortions in Hebei were undergoing a repeat abortion [14]; and another study conducted on similar population in Beijing found that 26.9% of 1478 young unmarried women were undergoing a repeat abortion [15].Numerous reasons result in the unintended pregnancies and induced abortions of young and unmarried women in China. The first reason may be the lack of sufficient programs regarding sex and contraception education in the traditional educational system of China, leading to their limited knowledge or even misconceptions about reproduction and contraception, especially when compared with their western counterparts [16,17]. To significantly reduce the incidence of unintended pregnancies among youth, the latest policy of a ‘Medium- and Long-Term Development Program (2016–2025) for young people (14–35 years)’ was proposed by the Chinese Central Government in 2017 to strengthen the popularization of sexual knowledge and implement sexual health education in conditional schools [18]. The second reason is linked to China’s family planning programme. Although it could provide related technical services (such as contraceptive education, pregnancy check-ups, and abortion services), it mainly focuses on the married couples [9,13], such that young unmarried people have limited access to those services such as contraception counseling and services including long-acting reversible contraceptive (LARC) services. Lack of counseling in family planning services prevents them from seeking contraception services [19]. Although condoms are readily bought and used, unintended pregnancies continue to occur due to their inconsistent or ineffective usage. A strong negative correlation between abortion and contraception use has been reported, such that abortion incidence declines as contraception use increase [20,21]. The provision of contraception by abortion providers, particularly the LARCs, can reduce the repeat abortions [6]. Prior studies in China have demonstrated that unintended pregnancies were directly related to non-use of contraception or ineffective contraception, and most of that had been terminated by induced abortions.Adolescents are a critical target population with regard to global public health outcomes. However, they experience many sexual and reproductive health risks which are often attributable to early and unprotected sexual activities [22]. Compared with adults, they encounter more obstacles towards acquiring contraceptives, resulting in the low rates of their contraception use [23]. And a series of subsequent adverse outcomes are generated, such as unintended pregnancies and induced abortions. Adolescent pregnancies have more complications and are also unwanted in many cases [24]; and pregnancy-related morbidity and mortality are much more prevalent among adolescents than adults, particularly in low-income settings [25]. In 2009, the average abortion rate among all adolescents in the European Union reached 12.2 per 1000 girls aged 15 to 19 years [26]. Furthermore, similar to their adult counterparts, many adolescent girls might face risks of repeat abortions because of ineffective contraception following the prior abortions. Currently, very limited data exist on repeat abortions among adolescents globally. A nationwide retrospective register study in Finland reported that proportion of repeat abortions among adolescents seeking an abortion steadily increased from 8.8% in 1993–1995 to 14.3% in 2007–2009 [27]; the determinants for such increase included age, marital status, type of residence, and second-trimester abortion. In England and Wales, a study based on the data of national statistics on abortions found the percentage of repeat abortions among adolescents undergoing induced abortions increased from 9.1% in 1992 to 13.4% in 2013 [28]. In U.S., results of the Guttmacher’s 2000–2001 Abortion Patient Survey indicated that one in five adolescents obtaining abortions were undergoing repeat abortions [2]. In Kenya where safe abortion was restricted, one in every ten young women (12–24 years) seeking abortion-related care confirmed a previous induced abortion [29]; area of residence (urban versus rural), education, religion, employment, and contraceptive use at the time of current pregnancy were significantly influencing factors. In terms of Asian regions, data extracted from the Georgian Reproductive Health Survey 2010 showed that the percentage of repeat abortions among adolescents who had the experience of abortions was 20% [30], and another study conducted in three provinces in Vietnam found that 7.0% of adolescent abortion seekers were undergoing a repeat abortion [31].In China, adolescents between 10–19 years of age accounted for almost 13.1% of the total population in 2010 [32]. A study in Shanghai verified a repeat abortion percentage of 38.5% among 2343 young unmarried women (≤24 years) undergoing induced abortions, with the determinants identified as age, occupation, education, the age of boyfriend, and cohabitation [33]. Another study conducted in three big Chinese cities reported that 33.0% of 4547 young unmarried abortion seekers (≤22 years) have had previous abortions, and education, migrant status, and contraceptive practice were significantly related to repeat abortions [34]. Nevertheless, to our knowledge, no nationwide study focused on repeat abortions among Chinese adolescents has been performed. In order to fill this literature gap, this study aims to describe the characteristics of Chinese adolescents undergoing abortions in China and to identify potential risk factors related to their repeat abortions.This cross-sectional study was a component of a collaborative research project for ‘Integrating Post-Abortion Family Planning Services in China (INPAC)’, which was funded by the European Commission (EC) under the Seventh Framework Program (FP7). Details of study design of the INPAC project have been published elsewhere [9,35]. The STROBE (Strengthening the reporting of observational studies in epidemiology) guidelines were followed when reporting results in this study [36].The survey was carried out in 30 provinces (autonomous or municipalities) in Mainland China, with the exception of Tibet due to practical constraints on data collection [37]. A stratified cluster sampling design was used. A total of 300 medical institutions that provided abortion services were invited to participate in the survey according to the ‘criterion for medical institution level’ based on size (First-, second-, and third-level) and type (general hospital, maternal and children’s hospital (MCH), and family planning service institution) of institutions. Specifically, each province included ten medical institutions: two general hospitals and one MCH at the third level, two general hospitals and three MCHs at the second level, and one general hospital and one family planning service institution at the first level.A total of 297 medical institutions were finally involved in this survey. Three were excluded because of unfinished data collection. During a period of two months, all the women seeking an abortion within 12 weeks of a pregnancy in the 297 medical institutions were invited to participate in the survey. In addition, only unintentionally pregnant adolescents who were younger than the legal minimum married age (20 years for women in China) and reported valid data regarding first versus repeat abortion were included in this research.A structured questionnaire developed by the INPAC’s consortium was comprised of five sections: (1) General socio-demographic information, (2) Reproductive history, (3) Contraceptive use during the six months preceding survey, (4) Induced abortion history, and (5) Reason for current unintended pregnancy. The questionnaire was completed by abortion service providers in a paper-based or electronic format. Data were collected continuously for two months in each participating institution from March 20 to October 5, 2013. Finally, all data were incorporated and cleaned by the National Research Institute for Family Planning of China.Six variables were used to measure the sociodemographic characteristics. (1) Age was initially a continuous variable and was divided into two categories: <18 and 18–19 years. (2) Education included seven categories: illiterate/semi-illiterate, primary school, junior middle school, senior middle school, senior college, university, and master and above. This study recoded a new variable with three categories: low-level (i.e., illiterate/semi-illiterate, primary school), medium-level (i.e., junior middle school, senior middle school), and high-level (i.e., senior college and above). (3) Occupation included 12 categories: jobless, housework, farmer, worker, business clerk, white-collar worker, civil servant/cadre, teacher/technician, domestic helper, private business owner, student, and other. This study recorded a new variable with five categories: jobless, housework, student, non-professional (i.e., farmer, worker, business clerk, domestic helper, private business owner, and other), and professional (including white-collar worker, civil servant/cadre, and teacher/technician). (4) Hukou status included five categories: local urban, local rural, non-native urban, non-native rural, and other (i.e., foreign and unclear). We recoded the variable of Hukou status into two distinct variables: residence status (i.e., rural, urban, and other) and migration status (i.e., migrant, non-migrant, and other). (5) The 30 provinces were classified into three categories of regions (i.e., relatively poor, middle-developed, and highly developed regions) according to the GDP (Gross Domestic Product) per capita of each province from 1997 to 2012 [38].Reproductive history was measured by two variables: prior pregnancy and parity (number of children). The variable of contraceptive use during the six months preceding the survey included eight categories: non-use, rhythm, withdrawal, emergency contraception, condom, combined oral contraceptive pill (COC), and other. Induced abortion history was displayed by the number of prior induced abortions (including surgical abortion and medical abortion). For this study, we recoded it into a binary variable, i.e., repeat abortion, to measure whether the adolescents had an induced abortion prior to the current one (0 = no, 1 = yes). Reason for current unintended pregnancy included two categories: non-use of contraception and ineffective contraception, i.e., contraceptive failure.Four continuous variables (i.e., age, number of children, number of prior pregnancies, and number of prior induced abortions) were tested for normality first. They all presented an abnormal distribution and were described using the “median” and “interquartile range (IQR)”. Categorical variables were described by “number” and “percentage.”Pearson’s chi-squared tests were performed to assess differences in the proportions of socio-demographic characteristics, reproductive history, contraceptive use during the six months preceding the survey, and reason for current unintended pregnancy between adolescents undergoing repeat abortions and those undergoing the first abortions. We further estimated the crude odds ratio (OR) with 95% confidence interval (CI) for each variable except the self-reported reason of current unintended pregnancy. In addition, a multivariable logistic regression model was set up to explore the risk factors associated with repeat abortions of adolescents who were undergoing abortions. All variables in the model have been checked for interaction. The results were presented as the adjusted OR with 95% CI. All the significance levels were set at p-value < 0.05. The Statistical Package for the Social Sciences 24.0 (SPSS, IBM, Armonk, New York, NY, USA) for Mac was used for data analysis.Ethics approvals were obtained from both ethics committees of the National Research Institution for Family Planning (NRIFP), China, and of the Ghent University, Belgium (B670201421116). All participants signed a written informed consent of which they received a copy. The questionnaire was anonymous and the data were protected according to the European Commission regulations on Data Protection and Privacy guidelines.A total of 2370 adolescents were identified from 80,675 women who participated in the INPAC study. Figure 1 shows the profile of participants. Among the 2370 adolescents receiving abortions, the median number of prior induced abortions was 0 (IQR: 0–1; Range: 0–6). Specifically, 927 (39.1%) were undergoing a repeat abortion, and 206 (9%) for a third time or more. The proportion of adolescents receiving abortions without a prior induced abortion was 60.9%.The median age of these adolescents receiving abortions was 19 years (IQR: 18–19 years). 15.4% were less than 18 years, and the minimum age was 13 years. Among them, 78.9% received a medium-level education. Moreover, 31.7% were engaged in a non-professional occupation, 28.0% were jobless, and 25.5% were students. Over half (55.9%) were from rural areas, and 48.2% were migrants. In addition, 62.2% resided in relatively poor regions. A significant difference was observed between repeat and first abortion seekers with respect to education (p < 0.05), occupation (p < 0.001), and region (p < 0.001). In comparison to adolescents undergoing the first abortion, their counterparts receiving repeat abortions presented a significantly higher proportion in the following characteristics: low-level education, non-students, and residing in relatively poor regions (Table 1). Meanwhile, estimation of crude OR with 95% CI indicated that adolescents who had higher odds of receiving a repeat abortion were those who received a low-level education (OR: 1.99, 95% CI: 1.14–3.47, compared with “high-level education”), those who were engaged in housework (OR: 1.81, 95% CI: 1.18–2.79, compared with “jobless”), those who were non-migrants (OR: 1.19, 95% CI: 1.01–1.41, compared with “migrant”), and those who resided in less developed regions (OR: 2.47, 95% CI: 1.86–3.28 for relatively poor region; OR: 1.99, 95% CI: 1.46–2.72 middle-developed region, compared with “highly developed region”). However, adolescents who were students had higher odds of undergoing a repeat abortion (OR: 0.70, 95% CI: 0.56–0.89, compared with “jobless”).The median number of children that adolescents had was 0 (IQR: 0–0; Range: 0–3). Almost all adolescents receiving abortions (93.0%) had no child, and only 7.0% had one child or more. The median number of prior pregnancies among adolescents was 0 (IQR: 0–1; Range: 0–6). More than half (56.3%) of adolescents receiving abortions did not have a prior pregnancy, 30.8% had one, and 12.9% had two or more. Significant differences were observed between adolescents undergoing a repeat abortion and those receiving a first abortion with respect to parity (p < 0.001) and the number of prior pregnancies (p < 0.001). Adolescents undergoing a repeat abortion presented a significantly higher proportion in the following groups: having one child or more and having one or more prior pregnancies (Table 2). Meanwhile, estimation of crude OR with 95% CI indicated that adolescents who had children (OR: 3.40, 95% CI: 2.43–4.75, compared with “having no child”) and had two or more prior pregnancies (OR: 4.55, 95% CI: 2.34–8.85, compared with “one prior pregnancy”) had higher odds of receiving a repeat abortion.Results indicate that 62.0% of the adolescents receiving abortions did not use any contraceptive measures during the six months preceding the survey. Condom (58.9%) was the most common among adolescents receiving abortions who used contraceptive measures, followed by emergency contraceptive (13.4%), rhythm (10.8%) and other (16.9%, i.e., COC, withdrawal, and other). Among the repeat abortion seekers, 45.1% reported having used contraceptive measures, whereas those who were undergoing the first abortion and had used the measures only reached 33.4% (p < 0.001) (Table 2). Meanwhile, estimation of COR and 95% CI indicated that the adolescents who were receiving abortions and had used contraceptive measures during the six months preceding the survey had higher odds of undergoing a repeat abortion (OR: 2.47, 95% CI 1.63–3.75 for rhythm; OR: 1.52, 95% CI: 1.05–2.21 for emergency; OR: 1.39, 95% CI: 1.14–1.71 for condom; OR: 2.68, 95% CI: 1.75–4.12 for COC, compared with “non-use”).The primary reason for current unintended pregnancies of adolescents receiving abortions was non-use of contraception (67.9%), followed by the ineffective contraception (32.1%). The current unintended pregnancies followed ineffective contraception for 37.9% of repeat abortion seekers and for 28.4% of the first abortion seekers (p < 0.001) (Table 2).All results of estimation of COR and 95% CI for each variable above were displayed in Table 3.In multivariable analysis, considering the high correlativity between the parity and number of prior pregnancies, and that no prior pregnancy would preclude a repeat abortion, we excluded the variable of number of prior pregnancies from the regression model (Table 3). Meanwhile, the variable of self-reported reason for current abortions was not included. After adjusting all remaining variables in the regression model, occupation, region, contraceptive use during the six months preceding the survey, and parity were significantly associated with receiving a repeat versus first abortion among adolescents undergoing abortions. Adolescents who resided in the less developed regions (OR: 1.76, 95% CI: 1.27–2.44 for middle-developed region; OR: 2.33, 95% CI: 1.72–3.15 for relatively poor region, compared with “highly developed region”), had used contraception during the six months preceding the survey (OR: 2.10, 95% CI: 1.32–3.33 for rhythm; OR: 1.56, 95% CI: 1.05–2.33 for emergency; OR: 1.30, 95% CI: 1.02–1.66 for condom; OR: 2.35, 95% CI: 1.47–3.75 for COC, compared with “non-use”), and had children (OR: 2.47, 95% CI: 1.70–3.58, compared with ‘no child’) had higher odds of receiving a repeat abortion. Meanwhile, adolescent abortion seekers who were students (OR: 0.67, 95% CI: 0.52–0.87) had higher odds of receiving a repeat abortion than those non-students. However, the variables of age, education, residence status, and migration status were not significant. In addition, we tested the bivariate interaction effects for all significant variables in univariable analyses, and no significant interaction effects were observed.To our knowledge, this was the first nationwide study that focused on the repeat abortions of Chinese adolescents receiving abortions. Meanwhile, the sample was very large and it was a part of a larger study which covered 30 provinces in China. The overall literature in this area at present was not enough to be done to distinguish. Almost all existing studies in China were conducted in only one or few cities, and most of the participants were women without any restrictions or the young unmarried women.Our study found that 39.1% of the 2370 adolescents receiving abortions were undergoing a repeat abortion, which was similar to existing literature that reported repeat abortion percentage of young unmarried women in China ranging from 26.9% to 55.3% [14,15]. However, it was higher than that of adolescent abortion seekers in Finland, England and Wales, and the United States. 14.3% of adolescents seeking an abortion in Finland during 2007–2009 were receiving repeat abortions [27]; and the percentages of repeat abortions among similar population in the U.S. and England and wales were 13.4% and 20%, respectively [2,28]. No direct evidence was found to explain these differences. However, the gap of sexual education and contraceptive practice among adolescents between China and these developed countries may be relevant [16,17]. In response to that, the Chinese Central Government has proposed a programme called “Medium- and Long-Term Development Program for Young People (14–35 years)” in 2017 to strengthen sexual health education among young people and reduce young women’s risks of unintended pregnancies [18].Results of our study indicated that 16 adolescents receiving abortions reported more than four previous abortions (with one respondent reporting as many as six), which was in line with the maximum number (4–8) reported in previous research conducted among young unmarried women in China [15,39,40]. In addition, the minimum age of adolescents undergoing a repeat abortion was 13 years in our study, which was lower than the finding of 15 years old in prior studies in China [40,41,42]. This outcome implies that very young women were already involved in sexual activities and were at risks of unintended pregnancies and subsequent induced abortions. If women have sex at younger ages, they are less likely to use contraception and at a greater risk of a pregnancy [34]. Therefore, this group of young age should attract more attention from governments and society.In terms of the current unintended pregnancies of adolescents receiving abortions, 67.9% were related to non-use of contraception in our research, which was consistent with the results from prior studies. However, some studies reported ineffective contraception as the primary reason [15,34]. Additionally, more than three-fifths of adolescents receiving abortions in our study did not use any contraceptive measures during the six months preceding the survey. These results suggest extremely inadequate contraceptive practices among Chinese adolescents and further illustrate that adolescents were generally considered to exhibit risk-taking behavior with poor contraceptive practices [43].Our study identified four factors that were significantly associated with repeat abortions among Chinese adolescents receiving abortions. The first factor was occupation. Results indicate that being a student was a protective factor. A decreased risk of repeat abortions was found among adolescents receiving abortions who were students compared with those who were jobless, which was consistent with prior studies in China [15,44]. Normally, adolescents in China aged 19 years or below should still be receiving an education in schools; however, only 25.5% of the adolescents in our study were still students. In comparison with their jobless counterparts with limited education, such adolescents possibly had more contraception knowledge and were more aware of the risks of unprotected sexual behavior.The region was the second factor. One prior study reported that women with low socioeconomic status had an increased risk of repeat abortions compared with those of high socioeconomic strata [5]. Pradhan R. et al. reported that a higher risk of pregnancy among adolescents in Nepal was associated with living in the least resourced region [25]. In our study, although we did not directly measure the personal economic status of adolescents receiving abortions, we explored the effect of regions where they resided, which were classified according to the per capita of GDP and were representative of the general economic situation of adolescents. Findings indicate that residing in less developed regions (middle-developed region or relatively poor region) was a risk factor of repeat abortions of adolescents receiving abortions. Compared with the adolescent abortion seekers residing in highly developed regions, these adolescents who resided in middle-developed and relatively poor regions had 1.81 and 2.40 higher odds of undergoing a repeat versus first abortion, respectively. The less developed the region that the adolescents receiving abortions resided in, the higher the risk of undergoing a repeat abortion of them. This trend may be attributed to limited access to contraceptive measures and sexual health education and knowledge among these poor adolescents [30].The third factor was contraceptive use during the six months preceding the survey. However, the importance of contraception with regard to repeat abortions of women is a controversial issue. In our study, compared with the adolescent abortion seekers who didn’t use contraception during the six months preceding the survey, these adolescents who had used contraceptive measures (rhythm, emergency, condom, and COC) had 1.38–2.58 higher odds of receiving a subsequent versus first abortion. Similarly, a study in Netherlands by Picavet et al. revealed that women who used combined hormonal contraceptive or long-acting method were 1.54 or 1.91 times more likely to have a repeat abortion than those who did not use contraception [45]. In another study conducted in the UK, Fisher et al. also found that women with repeat abortions were twice as likely to take oral contraceptives than women with first abortions [46]. However, the studies in the United States, Hungary, and Georgia documented no significance in terms of contraceptive use relative to repeat abortions [2,30,47]. The effective and correct use of contraception has real potential to decrease risks of unintended pregnancies and repeat abortions [48]. Excepting self-report bias, the results may indicate that although these adolescents were motivated to use contraception because of previous abortions, they were not consistent and correct users [15,34,45]. LARC use during adolescence is safe and most effective [26]; however, due to the restrictions of FP programmes in China, LARCs are mainly delivered through family planning clinics which only target married couples, but the majority of induced abortions are performed in public hospitals where Post-Abortion Family Planning (PAFP) services are often lacking and women who have undergone abortions are usually not referred to family planning clinics for FP counselling and services. The fragmentation of FP services is leaving a high risk to vulnerable groups such as young and unmarried women.Parity was the fourth and a strong risk factor, which has been previously verified in the literature to have a positive effect on increasing the risk of repeat abortions in China and many other countries [2,5,30,45,47,48,49,50,51,52,53,54]. Women who had children were inclined to have repeat abortions compared with those without a child. In our study, 93.0% of the adolescents receiving abortions didn’t have a child; and compared with them, these adolescents who had one or more children had 2.57 higher odds of receiving a subsequent versus first abortion. Possibly, the latter was more casual on contraception and unintended pregnancy, and terminating unintended pregnancies through an induced abortion (even a repeat abortion) was more acceptable for them. Another reason might be that women with higher parity had repeat abortions because they did not want to look after another child [5].Age, education, residence, and migration status were not significantly influencing factors related to repeat abortions of adolescents who were receiving abortions, which contradicted the results of prior studies on young women [15,27,30,34,42,45,47,48,49,50,51,55,56,57].Our study has several potential limitations. First, as the sample in this study was adolescents who were receiving abortions, the results could not be generalized to all the adolescents in China. Second, as the survey is conducted by the way of self-reporting of the adolescents, this would bring a bias of social desirability, for example, adolescents undergoing a repeat abortion might especially feel like they should say they were using contraceptive methods (vs. no method use). Third, the study identified risk factors of repeat abortions among adolescents receiving abortions only using limited variables set in the questionnaire. Other factors that we did not collect and review (such as cohabiting, sexual abuse, age and contraceptive use at the first intercourse, characteristics of sexual partner, the education and occupation of parents, and personal income) are also important. Fourth, our data on contraceptive use were imprecise. Only a single choice was provided for participants when asking about which contraceptive measure they used before. This study overlooked the possibility of the joint use of multiple contraceptive measures. Furthermore, the kinds of contraceptive measure the participants used which led to the ineffective contraception (and resulted in their current unintended pregnancies and induced abortions) remain unidentified. At last, as the data were collected in 2013, there was a delay in dissemination of these results for 6 years. In future study, we aspire to conduct survey on samples of general adolescents and collect additional information to ascertain more risk factors of repeat abortions and determine the importance of contraceptive measures. Meanwhile, qualitative interview and research can be introduced to identify other possible predictors that are difficult to measure.A large percentage of adolescents receiving abortions in China are receiving a repeat abortion, and the low self-reported contraceptive use of adolescents seeking abortions suggests a need for more contraceptive services for this age group. Furthermore, the abortion seekers of adolescents who were more likely to be undergoing repeat abortions were those who were jobless, those who had children, those who resided in middle-developed or relatively poor regions, and those who used contraceptive measures during the six months preceding the survey. Our study further emphasizes that promoting contraception and reducing unintended pregnancies among adolescents remain critical issues. Adolescents should be offered equal access to FP in China as that of married women to reduce unintended pregnancies and repeat abortions, for example, providing LARC services for adolescents. Correct and consistent contraception practice should be promoted.J.L. and W.-H.Z. designed the study. S.W., J.X. and W.-H.Z. designed the questionnaire and collected data. J.L. prepared the first draft. Other authors commented the manuscript. J.L. and W.-H.Z. finalized the manuscript. All authors received the final manuscript.This research was funded by the European Commission (EC)’s Seventh Framework Programme (FP7), grant number 282490.We appreciate all the INPAC group members including Marleen Temmerman (Consortium coordinator), Wei-Hong Zhang (Project Leader) (International Centre for Reproductive health, Ghent University, Belgium); Jian Li, Cheng-Liang Xiong, Jun-Li Liu, Qing-Long Meng, Yan Che, Wei-Li Zhao, and Hui-Ping Zhang (the Chinese Society of Family Planning of the Chinese Medical Association, China); Xu Qian, Hong Jiang, Ji Liang (Fudan University, Shanghai, China); Shangchun Wu, Yan Zhou, Qing Liu (National Research Institute for Family Planning, Beijing, China); Lina Hu, Xiao-Jing Dong, Yi Jiang; Shuai-Bei Liu, Xiao-Ling Gan (Chongqing Medical University/Sichuan University, Chongqing/Chengdu, China); Jørn Olsen, Jiong Li (Department of Clinical Epidemiology, Aarhus University, Aarhus, Denmark); and Rachel Tolhurst, Dusabe-Richards Esther (Liverpool School of Tropical Medicine, Liverpool, UK). Meanwhile, we are grateful to the women who participated in the INPAC and to the 297 hospitals in China for their enthusiasm for collecting high-quality data.The authors declare no conflict of interest.Study profile.Sociodemographic characteristics of adolescents receiving abortions by history of abortion.* Chi-squared test; # Education: low-level (illiterate/semi-illiterate, primary school), medium-level (junior middle school, senior middle school), and high-level (senior college and above).Reproductive history, contraceptive use in six months before survey, and reason for current unintended pregnancy of adolescents receiving abortions by history of abortion.* Chi-squared test. † COC: Combined oral contraceptive pill.Binary logistic regressions on risk factors related to repeat abortions of adolescents receiving abortions.a COR: crude odds ratio. b AOR: adjusted odds ratio. c Model fit information: p-value of Omnibus tests of model coefficients = 0.000, −2LL = 3007.702, Cox & Snell R2 = 0.065, Nagelkerke R2 = 0.088, p-value of Hosmer-Lemeshow goodness-of-fit test = 0.921. d Education: low-level (illiterate/semi-illiterate, primary school), medium-level (junior middle school, senior middle school), and high-level (senior college and above). e COC: Combined oral contraceptive pill.

Med-MDPI/ijerph_3/ijerph-16-05-00795.txt

@@ -0,0 +1 @@
+The Danish Heart Foundation and the non-governmental organization Neighborhood Mothers have co-developed a culturally adapted intervention seeking to promote healthy dietary behaviour among ethnic minority women. This feasibility study explores the potential of the intervention to reach ethnic minority women using health promotion initiatives. Participants attended instructor courses or cooking events, where culturally adapted, healthy recipes were introduced and meals prepared. Feasibility was explored using a mixed-method approach. Surveys were completed by 59 volunteers and 150 participants at five instructor courses and 21 cooking events. Individual interviews and focus group discussions were conducted with volunteers and participants after completion of the intervention. After the intervention, 61% of the 150 participants had high levels of knowledge about dietary recommendations, 96% intended to cook healthy dishes in the future and 84% intended to incorporate measuring equipment into their daily cooking routine. Participants with a high level of knowledge reported intention to change dietary behaviour more often than participants with lower levels of knowledge. Interviews confirmed that the participants cooked healthy dishes after participating, and incorporated knowledge about healthy food practices into their daily cooking. Few participants used measuring equipment. The intervention proved to be feasible as a health promotion initiative targeting a hard-to-reach population.Cardiovascular disease (CVD) is accountable for approximately 17.3 million annual deaths across the globe [1,2]. Social and ethnic inequalities exist in exposure to risk factors and in the burden of CVD. Studies find that an ethnic background is strongly associated with CVD incidence [3,4,5,6] with the risk being particularly high among non-Western migrants [3,4]. In Denmark, some groups of non-Western migrants have twice as high CVD incidence as ethnic Danes [3]. Disease prevention interventions play an important role in CVD prevention. Overcoming ethnic and socioeconomic inequalities in CVD requires disease prevention interventions that accommodate specific needs in terms of language, cultural and psychosocial factors in different population groups [7,8,9,10]. Diversity with regards to socioeconomic position, nationality, culture, language proficiency and, for migrants, time spent in the country of destination are important factors to consider in interventions [7,11]. This diversity, together with general barriers to disease prevention programs within ethnic minority groups, imposes a high demand for tailored disease prevention interventions. This in turn emphasizes the need to identify new and innovative ways to conduct disease prevention in people from varied backgrounds. Researchers within participatory research argue that a collaboration with the target population improves the long-term effects of interventions, as participants become an influential part of the decision-making and design process [12]. This influence and control increases the feeling of ownership among the participants, which has a positive effect on their engagement. Disease prevention interventions targeting people with an ethnic minority background should explore potential benefits of participatory approaches.The Danish Heart Foundation, and Neighborhood Mothers, a Danish non-governmental organization, have collaboratively developed a culturally adapted intervention, seeking to promote healthy dietary behaviour among middle-aged and older ethnic minority women with the aim of preventing CVD. The core concept of the intervention is participatory approaches with a co-designed cookbook based on recipes chosen by the women and adjusted in collaboration with experts to ensure alignment with recommendations for healthy diets. The intervention is embedded within the social networks of female volunteers at Neighborhood Mothers with an ethnic minority background in order to reach ethnic minority women in their own communities. This study explores the feasibility of the intervention with regard to its potential as a preventive intervention targeting ethnic minority women. Specifically, we examine (1) its potential to motivate healthy dietary behaviour and (2) the cultural acceptability of the intervention. The primary aim of the intervention was to increase awareness of healthy cooking practices and improve dietary behaviour among ethnic minority groups at high risk of CVD, through culturally acceptable education. The target group was ethnic minority women from non-Western countries living in communities characterised by low income. Women were targeted in this intervention for two reasons. Firstly, in these communities, families tend to perform traditional gender roles, where women primarily are in charge of choosing and cooking food for the family. Secondly, all volunteers were women, which made women from their local community an accessible target group.The intervention consisted of two phases. During the first phase, volunteers at the Neighbourhood Mothers and dieticians from the Danish Heart Foundation (DHF) co-developed a cookbook with a variety of traditional recipes chosen by ethnic minority volunteers and later adapted to comply with the formal dietary recommendations advised by The Danish Veterinary and Food Administration [13]. Selected recipes from the cookbook can be viewed online [14]. This phase was based on participatory approaches, as ethnic minority women were highly involved in producing and selecting the recipes. In the second phase, DHF held five instructor courses in four areas with similar housing and socio-economic status in Denmark (two in Copenhagen and one in Helsingoer, Aarhus and Herning), inviting members of Neighbourhood Mothers with ethnic minority background to participate [termed ‘volunteers’]. Instructor courses focused on providing volunteers with insight and skills in terms of dietary recommendations and healthy cooking practices using the adapted versions of the traditional dishes from the cookbook. A core element in the education was the measurement of ingredients when cooking, as non-Western migrants tend to use more oil, butter and animal fat than recommended [15,16]. Implementing measuring equipment into their daily cooking was therefore perceived to be important in order to follow the recipes developed for healthy food practices. Furthermore, they were trained in communication skills and skills related to organising cooking events. After participating in the instructor course, the volunteers were encouraged to organise one cooking event each, inviting six to eight ethnic minority women from their community [termed ‘participants’] to join. In line with participatory approaches, the volunteers were responsible for all aspects of planning the cooking event and could organize it in any way they felt was best for the participants. A total of 47 events took place in different communities across the four cities. For most of the cooking events, the volunteers and participants shared the same ethnic background; however, some cooking events were held for participants with different ethnicities. During these events, the volunteers taught participants about dietary recommendations and healthy cooking practices using recipes from the cookbook. At these cooking events, participants received a set of recipes and measuring equipment to take home.This feasibility study was conducted using a mixed-method approach. After the training was completed, surveys were conducted. The survey for the volunteers consisted of 26 items and that for the participants had 30 items. Both surveys included items on demographic and socioeconomic information, level of knowledge regarding dietary recommendations, intention to change dietary behaviour, and satisfaction with the intervention. Items on demographic and socioeconomic information were developed according to recommendations in existing literature and similar national surveys in Denmark [17,18,19,20]. The item measuring level of knowledge asked the respondent to: “Tick all ten healthy food principles.” and were given 15 possible answers. Items regarding intention to change dietary behaviour and satisfaction with the intervention were inspired by items used in prior research, however, developed to fit the specific intervention [21,22]. The items were formulated as follows: “Do you wish to cook the dishes again?” and “Do you think that you will use measuring jugs and spoons in your daily cooking in the future?” The surveys were completed by volunteers at all five instructor courses and by participants at 21 out of 47 cooking events. In order to overcome language and cultural differences, the surveys were translated into four different languages (Turkish, Somali, Urdu and Arabic), in addition to the Danish version. The surveys were pilot tested prior to and after translation, among six persons with different ethnic minority backgrounds. Additionally, semi-structured individual interviews and focus group discussions were conducted with volunteers (n = 7) and participants (n = 8) one to two months after participation in a cooking event. The semi-structured interview guides were developed according to survey responses, of which some needed more exploration including experiences related to participating in the intervention. Furthermore, the interview guides comprised of questions regarding the impact of the intervention, including potential dietary changes the intervention might have led to. Recruitment for interviews was done through snowballing in which volunteers where asked after their cooking event whether they were willing to participate in an interview. The participants were asked in the survey whether they wished to be interviewed. Also, volunteers reached out to participants at the cooking events to help recruit candidates for interviews. We selected volunteers and participants for interviews using a maximum variation strategy to ensure diversity in ethnicity and place of residence.The study objectives were assessed through statistical analyses of the survey data and thematic analysis of the fully transcribed interviews. Descriptive statistics were used to describe the study population. We calculated chi2-tests to examine the crude associations between level of knowledge and intention to change dietary behaviour. All statistical analyses were undertaken using SPSS Statistics version 25 (SPSS Inc., Chicago, IL, USA). In the following paragraphs, we will explain how we conducted analyses specific for each objective. Individual interviews and focus group discussions were fully transcribed and analysed through thematic network analysis. Firstly, we identified basic themes in the interview transcriptions. These basic themes were then grouped into organising themes related to the study objectives. The organising themes were all connected to a global theme equivalent to the primary aim of the intervention.In the survey, intention to change dietary behaviour was measured using two items: whether respondents intended to (1) cook healthy versions of traditional dishes from the cookbook and, (2) use measuring equipment in their daily cooking routines. Responses to these items were explored using descriptive statistics. Furthermore, we analysed whether level of knowledge was associated with intention to change dietary behaviour. We examined this association as this intervention sought to motivate healthy dietary behaviour through improved knowledge on healthy food choices and cooking practices. Level of knowledge was determined taking information from one item in the survey that asked respondents to identify 10 dietary recommendations out of 15 possibilities. We chose to dichotomise the variable into a binary variable according to the distribution of responses. Thereby, high level of dietary knowledge was characterised by identifying 8–10 dietary recommendations, while low level of knowledge was characterised by identifying 0–7 recommendations. When examining the association between level of knowledge and intention to change dietary behaviour, we worked with the hypothesis that participants who had a high level of knowledge were more likely to express intention to improve dietary behaviour than participants with a low level of dietary knowledge. This hypothesis was formulated in line with existing literature, which indicates that knowledge increases the likelihood of changing dietary behaviour [23,24,25,26]. Whether the intervention motivated healthy dietary behaviour was moreover examined in semi-structured interviews in order to gain insights into whether the participants did in fact change their dietary behaviour after participating in the interventionWe examined cultural adaption in semi-structured interviews to investigate the importance of this aspect of the intervention, including the acceptance of the recipes in different cultural contexts and appropriateness of healthy cooking practices.This study was approved by the Danish Data Protection Agency (file number: SUND-2017-14). Furthermore, written informed consent was obtained from each interviewee and survey respondent.As presented in Figure 1, a total of five instructor courses were held with 59 out of 61 volunteers responding to the survey (97% response rate). Moreover, 47 cooking events were held with a total of 422 women participating. Data were collected at 21 cooking events with 150 out of 174 participants responding to the survey (82% response rate). We conducted individual interviews with four volunteers and five participants while focus group discussions were conducted with three volunteers and three participants. The interviews ranged from 11 min to 85 min (average: 56 min). An overview of the study population, showing the baseline characteristics including socioeconomic and demographic factors, is shown in Table 1.As presented in Table 2, the vast majority of the participants (96%) reported an intention to cook the healthy versions of traditional dishes after participating in the intervention. Additionally, 84% of the participants intended to use measuring equipment, such as measuring jugs and spoons, in their daily cooking routines. In terms of level of knowledge, nearly 5% of the participants were not able to identify any of the dietary recommendations, while more than 25% were able to identify all. In more detail, 61% were able to identify 8–10 dietary recommendations—indicating a high level of knowledge; whereas 39% of the participants identified 0–7 recommendations—indicating limited knowledge about healthy food.Table 3 presents the results from chi2-tests examining the crude associations between level of knowledge and the intention to change dietary behaviour. Of the participants who identified 8–10 healthy dietary recommendations, 87% reported the intention to use measuring equipment in their daily cooking routine, and 80% of the participants who identified 0–7 dietary recommendation reported the same intention. With regards to cooking the healthy versions of the traditional dishes, 98% of the participants who identified 8–10 healthy dietary recommendations reported the intention of doing so, while the same was the case for 93% of the participants who identified 0–7 dietary recommendations. The difference is small, and thus, the results only provide indications that participants who were able to identify 8–10 dietary recommendations were more likely to intend both to use measuring equipment and to follow the healthy recipes for the traditional ethnic dishes, compared to the participants who identified fewer dietary recommendations. Likewise, the association was statistically non-significant.The interviews indicated that the intervention continued to affect the knowledge and food practices among volunteers and participants after completion of the intervention. Several of the volunteers and participants interviewed perceived that they were more conscious of their dietary behaviour and had begun to incorporate the dietary recommendations into their daily cooking routines. Measuring equipment, however, was not perceived to be relevant in the daily cooking routine, partly due to lack of time but also since these measuring practices were considered too onerous as it was much more natural for the participants to estimate ingredients by eye.In the interviews, the volunteers had a positive attitude towards the intervention’s foundation in using healthy versions of traditional dishes as reflected in the cookbook. Several volunteers described how they used the book actively when facilitating cooking events. For example, when choosing recipes according to the participants: if the participants were mainly from Pakistan, they would select two to three Pakistani recipes. Thus, the cooking events were culturally relevant for the participants. However, some volunteers experienced the participants having a negative attitude towards the recipes, as they contained less oil and more vegetables than the traditional versions. Nevertheless, in the end, the participants seemed to like the dishes: “I don’t know how many times she said that she was unhappy with it all [the food]... And the funniest thing was, when we were done, she thought it tasted good.”It seemed to be challenging to implement the practice of measuring ingredients, especially of oil, at some of the cooking events. Family traditions and food culture influenced the attitudes of some of the participants towards the measuring equipment. A volunteer described one particular incident:“There were some things that the participants just continued to be resistant towards… Then one of them said: ‘She put too little oil in [the dish]’ and they only needed two tablespoons of oil according to the recipe. Then the other one said: ‘If your mother-in-law sees that, she will be really mad.’”Participants generally expressed satisfaction with the healthy versions of the traditional dishes, and the majority reported that they had introduced the dishes to their families and friends as described in the following quote: “We were supposed to make samosas together, me and my cousin, and then I told her: ‘I don’t think I want to make them with you.’ Because she wanted to deep-fry them and I didn’t want to do that. I said: ‘You should come home to me before Ramadan and taste it [an oven-baked samosa]. It tastes exactly the same. It tastes better in fact. It is... it is a bit more crispy.”In terms of access and acceptability of the intervention, participants expressed that language barriers were overcome as the cooking events were carried out in the primary language of the participants. Furthermore, they perceived the cooking events to be accessible as the community-based approach made it easier to attend events that took place in familiar circumstances and networks.The findings of this study indicate that a culturally adapted CVD prevention intervention is feasible for reaching and engaging ethnic minority women of lower socioeconomic backgrounds. Following the intervention, the majority of the participants had a high level of knowledge about dietary recommendations, intended to cook healthy dishes in the future, and intended to incorporate measuring equipment into their daily cooking routine. Participants with high level of knowledge tended to report intention to change dietary behaviour to a higher extent than participants with a lower level of knowledge. The interviews confirmed that participants incorporated the knowledge about healthy food practices into their daily cooking, although measuring equipment was not perceived to be useful in their daily cooking.Ethnic minorities, in particular non-Western groups and those with lower socioeconomic position are often harder to reach in disease prevention interventions due to barriers related to culture, social factors, language proficiency, and differences in health and risk perceptions [7,11]. This intervention managed to recruit these traditionally hard-to-reach participants from various ethnic backgrounds and low socioeconomic position (detailed information on the ethnic background of the participants is provided in Table S1 in the Supplementary Material). Managing to reach this population is primarily a result of the intervention design building on co-design of both content and delivery with volunteers with ethnic minority backgrounds. This played an important role in both the recruitment process and execution of the intervention. Furthermore, the community-based design, with cooking events being delivered in the neighbourhood of the volunteers, was important for access to and acceptability of the intervention. Using volunteers who share characteristics with the target population as a recruitment strategy to reach ethnic minorities is relatively unexplored. The high participation rates in this study indicate that this recruitment strategy could be feasible for future disease prevention interventions targeting ethnic minorities.Dietary behaviour is negatively impacted by lack of knowledge regarding healthy food, and absence of skills on how to implement dietary recommendations in daily life [23,24,25,26]. This intervention sought to motivate behaviour change by filling potential gaps in knowledge regarding both dietary recommendations and the skills with which to implement them in daily life. Although, we recognise that the overwhelming majority of participants reported an intention to change dietary behaviour, this could partly be a result of social desirability bias, and could also be explained by the so-called ‘intention-behaviour gap’, whereby an intention may not necessarily lead to actual behavioural change. However, interviews with participants confirmed that the intervention had led to more healthy food choices among the interviewees two months after the intervention. In order to determine whether the intention to change dietary behaviour resulted in actual long-term behavioural change, a long-term follow-up is necessary. The ‘intention-behaviour gap’ has been identified in previous research and is described as a black-box of underlying psychological processes through which intention is translated into actual behavioural change [27]. The progression from intention to action is influenced by both internal and external factors including self-efficacy, risk awareness, knowledge, and psychosocial and structural factors [24,26]. Furthermore, these factors might differ according to individual characteristics such as gender, socioeconomic status and ethnicity [26]. According to interviewed participants, barriers to changing their dietary behaviour included food preferences of their husbands and children as well as the stressful and timely structured everyday life. These factors create a gap between intention and behaviour, which form a challenge in disease prevention, when seeking to make sustainable changes in the target population.This intervention targets ethnic minority women who migrated to Denmark. This population has been exposed to a complex combination of risk factors affecting their health-related behaviour: in their country of origin, during the migration process, and in Denmark [28]. A large body of literature has examined the acculturation process, which migrants experience upon arrival in their host country. This process is identified as a multidimensional and dynamic process by which the attitudes, values, beliefs and behaviours of one culture are changed over time [29,30]. The acculturation process is affected by a wide range of factors, including social and contextual factors present in the particular local community in which people live [30]. In Denmark, the majority of migrants live in social housing in areas, which have a high concentration of residents with low income and low educational level [23,31]. Allen et al. (2014) found that greater acculturation among migrants living in low-income housing was associated with poorer diets. These findings indicate that migrants did possess knowledge of healthy dietary practices, and due to accessibility and cultural factors in their country of origin, they had healthier eating habits prior to their arrival to the host country, which is in line with findings from other studies [32,33]. Health promotion interventions should therefore help migrants maintain their healthy dietary behavior throughout the acculturation process in order to reduce their high risk of CVDs, as found in existing literature [3,4]. Drawing on prior knowledge and dietary practices of migrants provides a potential to develop an intervention with a particular high impact. It is therefore valuable to include the knowledge of the target population when designing an intervention and will increase cultural relevance, acceptability, and engagement in the target population. In addition, the accessibility of healthy and affordable food in the neighbourhood influences dietary behaviour [23,30]. Moreover, time is generally a crucial structural barrier for behavioural change, regardless of ethnic background. These social and contextual factors influence the long-term feasibility of interventions aiming to motivate changes in dietary behaviour, as participants might return to their usual habits despite their intention to change. In fact, interviews with volunteers and participants confirm this gap in regards to the use of measuring equipment. The survey responses given right after participation shows an intention to use measuring equipment among the vast majority; however, several interviewees indicated not having used them since due to structural factors and usual cooking routines. This intervention did therefore not manage to close the gap entirely. However, the results do indicate that the intervention was successful in reaching ethnic minorities—an otherwise difficult task in health promotion—as well as motivating healthier food choices. This is a result of the participatory approaches used in this intervention. Particularly, the role of the volunteers acting as role models for the participants. The volunteers and participants share characteristics and ethnic backgrounds, which creates familiarity and trust that the participants might not experience with health providers in the Danish healthcare system. This bond between volunteer and participants is the main strength of the intervention.Ethnic minority people have lower participation rates in disease prevention interventions, and evidence on how to effectively stimulate health behaviour change in ethnic minority populations is limited [7,8,9,10]. According to Davidson et al. (2013), adapted interventions are more likely to succeed in improving health behaviour, including healthy dietary habits. Specifically, adapted interventions consider the characteristics of their target population and context, in order to increase effectiveness and ensure the sustainability of the intervention [7]. When targeting ethnic minorities, it is crucial to consider their cultural background when designing an intervention. This particular intervention used cultural diversity as a foundation for the entire intervention. First and foremost, this intervention introduced culturally adapted recipes in line with dietary recommendations as an incentive to improve dietary behaviour among participants with ethnic minority backgrounds. Participatory approaches to disease prevention has a long tradition in international research; however, in Denmark this is a relatively new strategy in disease prevention targeting people with ethnic minority background. In Denmark, most disease prevention interventions seeking to improve healthy eating focus on dishes popular among ethnic Danes, thus neglecting the food preferences of ethnic minority groups. The findings of this study show the potential of tailoring successful interventions to the target group. Increased knowledge and skills do not necessarily translate directly into behaviour change; however, they are contributing factors [23,24]. Regardless of the short follow-up period, both volunteers and participants revealed in the interviews that they had continued to use the recommended recipes after participating, as well as using the dietary advice to modify other culturally specific recipes. Hence, this intervention has identified a mode of prevention that was feasible in reaching and engaging ethnic minority women in improving dietary behaviours.The primary strength of this study is the participatory approach in which community members are actively involved in the development and organisation of the activities. Furthermore, the intervention was successful in recruitment of ethnic minority women who are often hard to reach in disease prevention interventions. Additionally, the response rate in this study is high considering that ethnic minorities tend to have lower participation rates [7,10].Limitations of this feasibility study include the lack of a baseline survey, which would have made it possible to examine changes in the outcome measures before and after participating in the instructor courses and cooking events. It is, therefore, not possible to draw firm conclusions as to the full effect of the intervention in this study. Furthermore, the follow-up interviews were conducted one to two months after the cooking events, making the follow-up period relatively short. Therefore, long-term impacts of the intervention on the everyday lives of the volunteers and participants could not be determined. Data were only collected from 21 cooking events out of the 47 that were held. Consequently, the study population is small, which is problematic in the statistical analysis. Additionally, the study population might not be representative of the target population, as it contains relatively few people from each of the ethnic minority groups. The recruitment of participants for interviews might be subject to selection bias, as volunteers at the cooking events helped identify potential interviewees. According to several of the volunteers and participants, the surveys were somewhat long, which could influence their responses. Furthermore, the respondents filled out the survey while at the instructor courses and food events, whereby the context might have affected the responses. Additionally, the respondents might wish to appear as positively as possible, inducing a social desirability bias to the responses. This bias is particularly relevant when looking at the reported intention to change dietary behaviour as participants were asked to evaluate a cooking events held by women from their community. It is possible that the participants felt a need to report high levels of intention to change in order for them not to disappoint the women who held the cooking events. It could be seen as an attempt to help each other appear more successful, which may distort the results. However, the high reported rate of intention cannot be explained entirely by social desirability as the participants might in fact intend to improve their dietary behaviour. Furthermore, responses to the surveys are self-reported, which might lead to information bias. Interviews conducted in Danish might be influenced by the language proficiency of the interviewees, as poor Danish proficiency will limit the interviewees in their responses.In conclusion, the intervention proved to be feasible as a CVD prevention intervention in terms of reaching ethnic minority women and positively motivating healthy dietary behaviour. However, the intervention was not successful in motivating participants to use measuring equipment in their daily cooking. Culturally relevant and acceptable health information was important for the ability to engage the target group. Furthermore, drawing on the commitment and community access of volunteers emerged as important. Future studies should investigate the long-term effects of the intervention in order to determine its effect on changing health behaviours in the longer term. Moreover, interventions could increase long-term engagement of participants by arranging for the participants, their families and members of their communities to meet periodically and cook the recipes. Additionally, health interventions targeting migrants should draw on the health knowledge already possessed by migrants and use this as a foundation for the content of intervention. This will likely increase the feeling of authority and improve the impact of the intervention. The impact of such interventions could additionally increase if ethnic minority men were also targeted as food traditions is formed by the social context and all members of that specific context.The following are available online at https://www.mdpi.com/1660-4601/16/5/795/s1, Table S1: Population characteristic according to country of birth.A.V.J.P., S.B. and M.K. contributed to the design of the study, analysis and interpretation of data. A.V.J.P. drafted the manuscript. S.B. and M.K. critically reviewed the manuscript. All authors approved the final version of the manuscript.The research was funded by the Danish Heart Foundation.The authors wish to thank Nida Alsarras, Signe Petersen, and Anne Cathrine Dragsted for their contribution in collecting data, including gathering survey data, recruiting interviewees and conducting interviews.The authors declare no conflict of interest.Data collection.Population characteristics.Distribution of questionnaire responses with regard to level of knowledge and intention to change dietary behaviour.Crude association between level of knowledge and intention to change dietary behaviour.

Med-MDPI/ijerph_3/ijerph-16-05-00796.txt

@@ -0,0 +1 @@
+This study was conducted to determine the physiological and psychological benefits of foliage plants as visual stimuli. Twenty-three elementary students (aged 11 to 13 years old) participated in this study. In a crossover design, electroencephalography (EEG) was used to measure and determine the psycho-physiological effects of four different visual stimuli: an actual plant, artificial plant, photograph of a plant, and no plant. Subjective evaluations of emotions were assessed using the profile of mood state and semantic differential methods immediately after exposure to each visual stimulus. A significant decrease in theta waves of the frontal lobe was associated with presentation of the actual plants. This response indicated that the viewing of living plants prompted improvements in the attention and concentration of the elementary students. Furthermore, the presentation of the living plants was associated with more positive mood states, such as feelings of comfort and naturalness. In conclusion, actual plants may improve attention and prompt psychological relaxation in elementary students relative to artificial plants, photographs of plants, or the absence of plants.Owing to rapid urbanization and industrialization, the number and extent of green spaces have decreased and, as a result, humans have become distanced from Nature. This disconnect can spawn social issues, such as human alienation and emotional deterioration [1]. The savannah theory of Orians and the Wilson’s biophilia hypothesis explain the instinctive preferences of humans for the natural world [2,3]. The desire for closeness to nature has increased, and people have introduced plants into their indoor environments [4,5,6].Research show that most people seek nature for pleasure, social activities, or for physical exercise [7]. Kaplan and Kaplan suggested that exposure to the natural environment can effectively restore attention [8]. Ulrich found that stress from the external environment can be reduced through exposure to nature [9]; this observation is supported by reports of reductions in cortisol levels and blood pressure and positive mood changes in response to the natural environment [10,11,12,13].In South Korea, children aged 10–18 years spend an average of 0.62 h outdoors on weekdays, which is very low compared to time spent outdoors by children in Japan, Germany, and the USA [14]. The less children become exposed to the natural environment, the more actively opportunities for children to experience nature should be provided.Park et al. found that horticultural activities were effective in altering physical, psychological, emotional, behavioral, educational, and cognitive aspects [15]; the effects on emotional intelligence and an eco-friendly attitude were highest in children. In childhood, the body and cognitive functions must develop simultaneously to achieve emotional stabilization [16], and educators have suggested that exposure to nature in childhood is important for emotional development [17,18]. Horticultural activities stimulate a variety of emotions and provide opportunities for communication through group activities, allowing children to express their emotions naturally [19]. Bagot and Corraliza et al. suggested that children feel more positive emotions and have a high resilience to stress in nature-friendly schools [20,21]. Based on the results of various studies demonstrating the benefits of the natural environment on the well-being of humans, interest in the relationship between the natural environment and human health has increased [22].Previous studies have showed that the viewing of foliage plants can lead to physiological and psychological relaxation by stabilizing the autonomic nervous system and activating alpha brainwaves [23,24,25]. Sympathetic nerve activity and oxyhemoglobin concentration in the left frontal cortex reportedly decreases following the performance of tasks involving foliage plants relative to tasks without plants [26]. Furthermore, transplanting activities using real plants induced better emotional states and lower stress levels than transplanting activities using artificial flowers and computer tasks [27,28]. Moreover, placing foliage plants indoors has been shown to attenuate techno-stress [29]: the psychophysiological consequences of technology-induced alterations of living, such as an increased blood pressure and eye fatigue. Various studies have been conducted on the physiological and psychological changes that occur in adults after visual stimulation with green plants; however, such studies on children are severely limited. This study therefore investigated both the physiological and psychological responses of elementary students after viewing different foliage plants through electroencephalography (EEG) and the evaluation of participants’ emotional responses.The present study recruited 23 elementary school students. Researchers distributed flyers advertising the recruitment on the bulletin boards of community centers near Konkuk University, such as churches and apartment buildings. The inclusion criteria were as follows: right-hand dominance and no pre-existing physical and emotional disorders that could affect the physiological data. The first criterion was informed by reports that brain activation differs between right- and left-handed individuals [30]. Participants received the equivalent of $20 (USD) as an incentive to complete the experiment. This study was approved by the Institutional Review Board of Konkuk University (7001355-2017-10-HR-206).The experiment was conducted in a room at Konkuk University. A desk and chair was placed in the center of a room. The height of the chair was adjusted so that both feet of the subject would reach the ground. To minimize external visual stimulation, white hardboard paper was placed before the desk, and ivory-colored curtains were installed on either side of it (Figure 1a); a 1.8-by-1.6-m space was thus enclosed. A potted plant was placed on the desk at a distance of 0.5 m from the subject (Figure 1b). The average conditions of the experimental space were as follows: temperature, 21.8 °C; humidity, 25.2%; and illumination, 1465.8 lux.This study employed four visual stimuli: (1) actual plants, (2) artificial plants, (3) photograph of plants, and (4) no plants. As shown in Figure 2, the stimuli were set up as follows: (1) Actual plants: non-patterned Epipremnum aureum were planted in a white flower pot (width 55 cm, height 15 cm). (2) Artificial plants: artificial foliage, similar in appearance to E. aureum, were arranged in a manner similar to the actual plants and in the same kind of white flower pot. (3) Photograph: researchers printed a life-sized color photograph that was taken of the living plant used as the actual plant stimulus in condition (1). (4) Absence of plants: the same white flower pot used in conditions (2) and (3) was partially filled with plain soil.Prior to the experiments, the height, weight, and body composition of the participants were measured using an anthropometer (Ok7979; Samhwa, Seoul, Korea) and a body-fat analyzer (ioi 353; Jawon Medical. Gyeongsan, Korea). Demographic information, such as age and sex, was collected through a survey.A wireless EEG (Quick-20, Cognionics, Inc., San Diego, CA, USA) was attached to the head of the subjects who were then seated in the experimental space for 3 min in order to habituate them to the novel environment. Prior to each trial, the researcher drew lots to determine the order in which to present the four visual stimuli (actual plant, artificial plant, photograph of plant, and no plant). Physiological data were collected for 3 min for each treatment. Before each stimulus was presented, the subjects were asked to look at the white wall in front of them for 1 min to encourage relaxation. After the presentation of each stimulus, the subjects were given a survey regarding their psychological response to the stimulations. The duration of the entire experiment was ~30 min per subject (Figure 3).Brain waves from the cerebral cortex are recorded as electrical signals classified as following types based on their frequencies: gamma (30–50 Hz), beta (14–30 Hz), alpha (8–13 Hz), delta (4–8 Hz), and theta (4–8 Hz) [31]. Each type of wave is association with a different situation: gamma, anxiety and excitement; beta, tension; alpha, relaxation; delta, deep sleep; and theta, shallow sleep [32]. In this study, alpha and theta were analyzed to inform assessments of physiological stability and improvement of attention under the four visual conditions.According to the 10–20 international system of electrode placement, the electrode was attached to the left ear lobe (A1), and the EEG was performed using a total of eight channels: Fp1 (left prefrontal cortex), Fp2 (right prefrontal cortex), F3 (left frontal lobe), F4 (right frontal lobe), P3 (left parietal lobe), P4 (right parietal lobe), O1 (left occipital lobe), and O2 (right occipital lobe) [33]. Only two channels, F3 (left frontal lobe) and F4 (right frontal lobe), were analyzed in this study because the frontal lobe is reportedly associated with rational decision making and judgment (Figure 4).The profile of mood state (POMS) and semantic differential (SD) methods were used to investigate the psychological reaction of each participant to each stimulus.The POMS method was developed by McNair et al. as a way of assessing temporary mood or emotional states that vary according to the surrounding conditions of the subjects [34]. The questionnaire used for this method consists of 30 questions that assess tension-anxiety, depression, anger, fatigue, confusion, and vigor [35]; it was translated into Korean by Yeun and Shin-Park. Total mood disorder (TMD) is assessed through the questionnaire. A value corresponds to each response, and lower total scores indicate better emotional states.The SD method was developed by Osgood as a method of evaluating emotions with adjectives [36]. The questionnaire consists of three categories: comfortable to uncomfortable, natural to artificial, and relaxed to awakening. The participant’s degrees of emotion were scored on a scale of 13 points, and higher scores indicated better emotional states.SPSS (Version 22 for Windows; IBM, Armonk, NY, USA) was used to conduct one-way analysis of variance and Kruskal-Wallis tests. A p-value of <0.05 indicated statistical significance. Demographic data were analyzed using Microsoft Excel (Office 2016; Microsoft Crop., Redmond, WA, USA) to generate descriptive statistics of the means, standard deviations, and percentages.The participants (boys: n = 9, 39.1%; girls: n = 14, 60.9%) in this study were (mean ± SD) 12.2 ± 1.0 years old, 149.6 ± 11.3 cm in height, weighed 44.3 ± 11.8 kg, and their body mass index was 19.4 ± 3.1 kg m−2; according to the World Health Organization, these values fall within the normal range for the age-group (Table 1) [37].The theta waves, which reportedly evince drowsiness and are used as a measure of low attention or concentration [38], were found to be significantly reduced when participants looked at the actual plants relative to the other stimuli (Table 2). This finding suggests that the improved attention of the participants was attributable to the visual stimulus of the actual plants rather than general attentional stability or arousal. Furthermore, the actual plant did not prompt any changes in alpha waves in the children.There were no significant differences in the POMS induced by the stimuli; however, a lower TMD was associated with the presentation of the actual plant (Figure 5). This result revealed the potential of actual plants to prompt positive mood states. The SD findings indicated that the subjects felt more comfortable (p = 0.049) and natural (p = 0.021) when they stared at the living plants (Figure 6).While other studies have only investigated the advantages of green plants, the present study compared the effects of green plants with other visual plant-related stimuli whose presentations evoked the way in which they are actually perceived in daily life. As display technologies advance, images are projected with increasingly greater clarity. However, this study showed that only actual plants can significantly affect theta waves. This may indicate that all five senses, and not just sight, contributed to the alteration of neural activity in a way that could not be simulated by the other stimuli [38].The present study found looking at the actual plants significantly reduced theta waves, suggesting that live plants can improve attention. Park reported that the inhibition of theta waves in 11–15-year-old adolescents contributes heavily to their academic performance [39].Son reported that the placement of green plants in a room changed the brain waves of the people therein, improved their concentration, and relieved their visual fatigue [40]. Kuo reported that living in a natural environment is closely related to the enhancement of attention [41], and contact with nature has been shown to improve the attention of both adults and children [42,43].As aforementioned, alpha waves indicate a state of rest or relaxation [32,44]. Prior studies conducted with adult subjects have shown that viewing of green plants increases cerebral activity and has a positive effect on their psycho-physiology [45,46]. Moreover, a different study demonstrated that the latter effect was more potent when viewing actual foliage plants than when viewing images of plants [47]; this finding may be attributable to the activation of parasympathetic nerves and the suppression of the sympathetic nervous system [23,48,49,50]. In addition, when working with plants, activity of parasympathetic nerves increases and the left frontal cortical concentration of oxidized hemoglobin decreases among adults working with plants relative to those that are not [26]. However, in contrast to these findings, the present study found not changes in the alpha waves induced by the actual plants in children.The SD performed by the present study revealed that the actual plant stimulus induces more positive emotions, such as comfort and feelings of naturalness, than did other stimuli. These effects resemble subjective emotional stabilization, e.g., physiological stability and mood improvement, with which a decrease in right prefrontal cortical oxidative hemoglobin concentration was associated in adults that viewed plants [24]. Park et al. also showed that working with plants increased subjective emotional stability relative to plant-free work [26].Kaplan and Kaplan suggested that the natural environment is effective in restoring attention in humans [8], and Bringslimark et al. reported that indoor plants are effective in reducing stress and for promoting pain relief [2]. In addition, transplanting activities reportedly reduce more stress and induce more positive mood states than does transferring artificial flowers or performing computer-based tasks [27,28]. The application of horticultural therapy was shown to significantly enhance the positive mood of hospitalized patients [51]. According to Van den Berg et al., negative emotions, such as anger, tension, and depression, induced by a horror movie were decreased by 38% after viewing an image of a nature scene [52].This study provided scientific support for the physiological and psychological effects of viewing green foliage in children. Visual stimulation with real plants reduced theta waves, indicative of a lack of concentration. Though the actual plants did not enhance alpha waves, which suggest a relaxed state, the children reported their moods to be more comfortable after viewing the living plants. Thus, the visual stimulation of green foliage plants tended to improve attention and feelings of comfort in elementary school students. This study was subject to two limitations. First, our sample size was small. Future studies should recruit more subjects to analyze more specific physiological mechanisms underlying the visual stimulation of green plants. Moreover, a larger study population would allow for the identification of sex-related differences. Second, we only measured momentary mood changes when the subjects look passively at different plant stimuli.Y.-A.O. and S.-A.P. contributed to the experimental design, data acquisition, statistical analysis, interpretation of results, and manuscript preparation. S.-O.K. contributed to data acquisition and data analysis.The study was conducted with the support of the “Joint Research Project (Project title: Determining the Effects of Urban Agriculture Program for Health of Elementary School Students, Project No. PJ012808)”, Rural Development Administration, Republic of Korea.This paper was supported by the KU Research Professor Program of Konkuk University.The authors declare no conflict of interest.Experimental conditions: (a) Photograph of the experimental room; (b) Arrangement during the experiment.Experimental materials and set-ups used as the visual stimuli: (a) Actual plants, (b) Artificial plants, (c) Photograph of plants, and (d) Absence of plants.Experimental protocol used to assess the effects of exposure to four stimuli. The responses of the participants were assessed via electroencephalography (EEG) measurements and subjective tests.Measuring electrode. Fp: prefrontal cortex; F: frontal lobe; Temporal; C: central; P: parietal; O: occipital; Z: zero, refers to an electrode placed on the midline sagittal plane of the skull; A: the prominent bone process usually found just behind the outer ear.Comparisons of the profile of mood state (POMS) of elementary students in response to four different visual stimuli. Data are presented as means ± standard deviations. NS, non-significant.Comparisons of a semantic differential method (SD) of four different visual stimulations. Data are presented as means ± standard deviations. (a): natural; (b): comfortable; (c): relaxation. * p < 0.05; NS, non-significant.Descriptive characteristics of the elementary students that participated in the experiment.1 Body mass index = weight (kg)/height (m2).Electroencephalography (EEG) results (alpha and theta waves) following visual stimulation.NS: non-significant; * p < 0.05.

Med-MDPI/ijerph_3/ijerph-16-05-00797.txt

@@ -0,0 +1 @@
+Isotemporal substitution model was first developed in 2009 [1] by our research team, whose members also once developed the isocaloric substitution model, widely-regarded as the gold standard method for modeling calorie-containing foods in nutritional epidemiology [2,3,4]. The ISM was developed because epidemiology research in daily activity time, up until recently, did not incorporate that daily time of 24 h a day is finite.Thus, as the original creators of the isotemporal substitution model for physical activity time allocation analysis [1], we read with interest the paper by Biddle and colleagues [5] which described the association between behavior reallocations and risk of metabolic health in a UK cohort. The authors reported the results using the compositional data analysis developed by Chastin et al. in 2015 [6], highlighting the differences between this model and what they called the “traditional” isotemporal substitution model (ISM) developed by Mekary et al. in 2009 [1]. While we commend the authors for their attempt to compare the two methods, we disagree with the explanations given to justify the superiority of the compositional data analysis over the ISM. In fact, compositional analysis is simply a derivative form of ISM that has been already shown in nutritional epidemiology to be a similar model mathematically.Frist, Biddle and colleagues [5] mentioned that while the ISM used absolute values in physical activity, the compositional data analysis used relative values. Interestingly, we have previously discussed the feasibility of using relative values (as in percentages) rather than absolute values in the ISM—we called it the “density model”, in reference to the “multivariate nutrient density model” developed by Willett and Stampfer (1986) [2,3,4] that is being used in nutrition epidemiology. We recommended against the use of the density model in physical activity epidemiology because unlike nutrition intake, which could be given in relative amounts or percentages, physical activity guidelines are given in absolute amounts [7] (e.g., 30 min/day of weight training) instead of relative amounts (e.g., 5% of your waking time spent in weight training). Similarly, a certain percentage of total discretionary activity time could be very heterogeneous among individuals, which makes it hard to interpret and establish physical activity guidelines because different individuals often have widely different total discretionary time available for physical activities. Thus, the same 30 min of running could represent a very different percentage of total activity time. Furthermore, unlike nutrition, almost all physical activity guidelines are given in terms of absolute values and not relative values. These values could vary by age group [7] or by desired outcome (weight loss maintenance [8], weight gain prevention [9], waist circumference change [10], cardiovascular health [11], among others); yet, they do not vary for every individual. In our opinion, this makes the density model, also called compositional data analysis by Chastin et al. (2015) [6], inappropriate to use in physical activity epidemiology. It also lends itself to potentially very inaccurate interpretations by different individuals of varying discretionary activity time.Second, Biddle and colleagues [5] compared their findings using both models and noticed that while the ISM led to symmetric results for the reallocation of behaviors, compositional data analysis led to asymmetric results, which in their opinion was a more accurate estimation of the effect size. For instance, they argued that substituting 30 min/day of moderate to vigorous physical activity for an equal time of sedentary behavior might not necessarily lead to the inverse effect estimate when this substitution was reversed. In our opinion, this depends on the model used. In fact, if data were normally distributed—according to the central limit theorem—and a linear regression model was used to model these associations, symmetry in the results would be expected if substitutions in activities were reversed as associations were estimated in an additive way. However, if the data were not normal or if there were more outliers than expected, the natural log of the outcome would then be modeled and associations would be estimated in a multiplicative way. The symmetry would then be lost in this model. Of note, we showed similar asymmetry in Mekary et al. (2013) [12] when a binary outcome of depression was modeled as an outcome via the Cox proportional hazard model to assess the relationship between activity and clinical depression risk. Similarly, in the paper by Biddle and colleagues [5], the authors used isometric log-ratios as predictor variables. In the original paper by Chastin et al. (2015) [6], the authors log-transformed the non-normally distributed outcomes. This natural log transformation took away the symmetry, as when the data were transformed back to the original scale, the reversal of substitutions of behaviors would no longer lead to symmetrical results. Hence, we disagree with the claim the authors that the ISM led to inaccurate results, which were symmetrical in this case. The ISM model could lead to symmetrical or asymmetrical results based on the scale used, as previously shown [1,12]; thus, symmetry is not a metric of any given model’s superiority per se.Third, Biddle and colleagues [5] compared their findings obtained using the compositional data analysis with the ISM model and noted that results were not materially different. Interestingly, the same conclusions were drawn using either model. Notably, the results from ISM were more appropriate, intuitive, and easy to understand, while the results from the compositional data analysis were hard to articulate. The authors, nevertheless, interpreted their findings using the same language used to interpret the ISM. This simply confirms what is already known from nutritional epidemiology’s decades of research of isocaloric substitution that have compared different substitution approaches [13], where substitution models with absolute macronutrient intakes have shown similar if not the same results as energy density (i.e., calorie percentage) models. Thus, their study simply reconfirms what is already known mathematically, all while confusing the literature with a seemingly different analysis that is more difficult to interpret for physical activity.Taken together, we stand by our original ISM and the accuracy and superior interpretation of the emanating results for physical activity. Other authors have even called the original ISM model a ‘seminal work’ in a recent systematic review of 56 ISM papers worldwide [14]. Moreover, the statistical properties of the ISM encompass those of derivative models, such as compositional analysis, which is merely ISM in another equivalent form. Altogether, we believe that the ISM in absolute units is the appropriate model to use in the arena of physical activity epidemiology, given that physical activity guidelines are provided and conveyed in absolute values rather than relative values.The authors declare no conflict of interest.

Med-MDPI/ijerph_3/ijerph-16-05-00798.txt

@@ -0,0 +1,30 @@
+Thousands of lower Manhattan residents sustained damage to their homes following the collapse of the Twin Towers on 11 September 2001. Respiratory outcomes have been reported in this population. We sought to describe patterns of home damage and cleaning practices in lower Manhattan and their impacts on respiratory outcomes among World Trade Center Health Registry (WTCHR) respondents. Data were derived from WTCHR Wave 1 (W1) (9/2003–11/2004) and Wave 2 (W2) (11/2006–12/2007) surveys. Outcomes of interest were respiratory symptoms (shortness of breath (SoB), wheezing, persistent chronic cough, upper respiratory symptoms (URS)) first occurring or worsening after 9/11 W1 and still present at W2 and respiratory diseases (asthma and chronic obstructive pulmonary disease (COPD)) first diagnosed after 9/11 W1 and present at W2. We performed descriptive statistics, multivariate logistic regression and geospatial analyses, controlling for demographics and other exposure variables. A total of 6447 residents were included. Mean age on 9/11 was 45.1 years (±15.1 years), 42% were male, 45% had ever smoked cigarettes, and 44% reported some or intense dust cloud exposure on 9/11. The presence of debris was associated with chronic cough (adjusted OR (aOR) = 1.56, CI: 1.12–2.17), and upper respiratory symptoms (aOR = 1.56, CI: 1.24–1.95). A heavy coating of dust was associated with increased shortness of breath (aOR = 1.65, CI: 1.24–2.18), wheezing (aOR = 1.43, CI: 1.03–1.97), and chronic cough (aOR = 1.59, CI: 1.09–2.28). Dusting or sweeping without water was the cleaning behavior associated with the largest number of respiratory outcomes, such as shortness of breath, wheezing, and URS. Lower Manhattan residents who suffered home damage following the 9/11 attacks were more likely to report respiratory symptoms and diseases compared to those who did not report home damage.The terrorist attacks in New York City on 11 September 2001 (9/11) led to the destruction of the World Trade Center (WTC) Twin Towers and six other adjacent buildings. The collapse of the two towers released a massive cloud of dust and debris that damaged surrounding buildings within and around the WTC complex, a 16-acre area that was subsequently called Ground Zero [1]. The plume of dust reached an altitude of 1500 meters and, depending on meteorological conditions, extended to lower Manhattan and the New York and New Jersey metropolitan areas. More than 100,000 µm per cubic meter of particles were estimated to be present in the air during the first few minutes after the collapse of each building [2].The contents of settled WTC dust have been extensively analyzed and were characterized as a mixture of cement dust, glass fibers, asbestos, lead, polycyclic aromatic hydrocarbons, polychlorinated biphenyls, organochlorine pesticides, and polychlorinated furans and dioxins [1,3,4,5]. Indoor dust was primarily comprised of inhalable particles (<53 µm). In contrast, most outdoor dust was primarily comprised of larger particles [6].Approximately 25,000 lower Manhattan residents were physically and emotionally affected by this disaster [7]. Two months after the attacks, air quality and surface dust were among the main concerns of lower Manhattan residents [7]. Most WTC-related residential exposures were caused by re-suspension of settled indoor dust [2] during cleaning efforts or in poorly cleaned environments. This is a large motivator for our analyses of cleaning practices.A number of studies demonstrated the deleterious health effects of household exposures in the context of the WTC attacks. Most of them were relatively small case-control studies [8,9,10,11]. Increased asthma prevalence was previously reported among WTC Health Registry (WTCHR) enrollees who experienced a heavy layer of dust in their homes [12]. Acute and chronic exposures were also associated with lower respiratory symptoms among WTCHR participants [13]. However, prior studies, specifically Lin et al. [8] and Maslow [13], had limited samples of residents (n = 1317; n = 479, respectively). Maslow [13] did not evaluate associations between exposures and specific respiratory symptoms. Although Lin et al. [8] did assess lower and upper respiratory symptoms, they did not include diagnosed conditions such as asthma or chronic obstructive pulmonary disease.With a large sample of residents in lower Manhattan and a comprehensive set of outcomes, the present study describes patterns of home damage and cleaning practices, as a surrogate for household exposures since measures of exposure in homes is limited. We initially looked at geospatial relationships between Ground Zero location and exposure and health outcomes. We then conducted multivariate logistic regression analyses to find potential associations between those variables among a large sample of WTCHR respondents.The WTCHR monitors the health of people exposed to the 9/11 WTC disaster through periodic health surveys and is housed within the New York City Department of Health and Mental Hygiene (NYCDHMH). The WTCHR is the largest effort in the United States to monitor health after a disaster of this kind and includes data on 68,444 adults, including lower Manhattan residents, rescue and recovery workers, and building occupants and passers-by. More details on the WTCHR can be found elsewhere [14] and online at https://www1.nyc.gov/site/911health/about/wtc-health-registry.page.For this report, data were derived from WTCHR Wave 1 (W1) survey (baseline), collected between 9/2003 and 11/2004, and Wave 2 (W2) survey (first follow-up), collected between 11/2006 and 12/2007. We included W1 and W2 participants 18 years and older, whose primary residence on September 11th was located in lower Manhattan, south of Canal Street. Residents who performed rescue and recovery work were excluded from this analysis.The Centers for Disease Control and Prevention (CDC) and NYCDHMH institutional review boards approved the Registry protocols. A Federal Certificate of Confidentiality was obtained, as was verbal informed consent.In the W1 Survey, respondents were asked if they were outdoors after the towers’ collapse and to report their location when they first encountered the dust cloud (closest cross street intersection, closest landmark, or closest subway stop) [14]. The W2 survey information was used to categorize dust exposures into “intense”, “some”, and “none”. The “intense” dust exposure was defined as having been in the dust cloud on 9/11 and reporting at least one of five experiences: being unable to see more than a few feet; having difficulty walking or finding one’s way; trouble finding shelter; being covered with dust; or not being able to hear. The “some” category consisted of those who had reported being in the dust and debris cloud at Wave 1 but who did not experience intense exposure and “none” were those who reported no dust cloud exposure at all. These categories of exposure to the dust cloud (1 = none, 2 = some, or 3 = intense) were combined with reported location at the time of first encounter with the dust cloud (reported at W1) to produce Figure 1 for the geospatial analysis.Wave 2 also included detailed questions about conditions inside the homes after 9/11 as well as cleaning practices and replacing of household items. WTCHR participants were asked about the presence of “a fine coating of dust on surfaces”, “a heavy coating of dust on surfaces (so thick one couldn’t see what was underneath)”, “broken window(s)”, “damage to home or furnishings”, and “debris from the disaster”. In addition, enrollees answered whether or not they had personally “cleaned ventilation ducts”, “cleaned with a damp cloth or wet mop or wet sponge”, “used a vacuum (with or without a high-efficiency particulate air or HEPA filter)”, and “dusted or swept without water.” Moreover, enrollees were inquired whether or not they replaced “carpet or rugs”, “furniture (replaced or re-upholstered)”, “drapes, blinds or curtains”, and “air conditioners” as a result of the 9/11 events.Outcomes of interest were self-reported respiratory symptoms and self-reported physician diagnoses of respiratory diseases, asked at both W1 and W2 surveys. Symptoms included shortness of breath, wheezing, persistent chronic cough, and upper respiratory symptoms, first occurring or worsening after 9/11 and present at W2. Respiratory diseases comprised asthma or reactive airways dysfunction syndrome (RADS) and chronic obstructive pulmonary disease (COPD), first diagnosed after 9/11 and present at W2.Initially, to explore potential geospatial relationships between Ground Zero location and exposures and health outcomes, we mapped exposures and health outcomes of WTCHR resident enrollees in lower Manhattan using ArcGIS version 10.2 (ESRI, Redlands, CA, USA). These variables were mapped at the census tract level of geography using 2000 Census tract boundaries. For this part of the analysis, “yes” responses for any of the health outcomes variables were summed per census tracts (Boolean OR, not arithmetic sum). Similarly, exposure variables (fine coating of dust on surfaces, heavy coating of dust on surfaces, broken windows, damage to home or furnishings, and debris from the disaster was present) were also grouped and categorized as “yes” or “no” and mapped by census tract totals. The number of enrollees within each census tract was used as the denominator of the proportions represented on the maps. In addition, point values of dust cloud exposure were modeled using inverse distance weighted (IDW) interpolation to create a surface representation of the dust levels throughout lower Manhattan. IDW uses measured values (point locations) to predict values for unmeasured surrounding areas. We examined demographic characteristics, household exposures resulting from the 9/11 attacks, and cleaning practices among the study population using descriptive statistics (means, standard deviations, and proportions). We also calculated the prevalence (%) of self-reported respiratory symptoms and respiratory diseases. The relationship between cleaning practices and conditions inside the homes after 9/11 was assessed using multivariate logistic regression analyses adjusted for age, race, education, income, sex. Analyses were also controlled for “priority group”, a stratification of registrants into higher and lower priority groups for the purposes of targeting and outreach, which to a degree reflects a resident’s distance from the WTC site. Priority group 1 included residents as of 9/11 at addresses located south of Chamber Street; Group 2 included residents located on or north of Chamber but south of Canal Street; a third resident group, Group 0, was defined upon recognition of the inclusion of respondents living on or north of Canal in zip codes overlapping the Canal boundary or in other NYC boroughs. Also, the analyses were controlled for any exposure outside, to the dust cloud.Each of the cleaning practices (dusted or swept without water; cleaned with damp cloth, sponge, or mop; used a vacuum to clean; and cleaned ventilation ducts) was modeled using logistic regression. Based on medical considerations and plausibility of real relationships being present, the list of all variables was narrowed to 17 potential explanatory variables for cleaning practices using ‘purposive’ or a priori selection. The predictors or covariates included five home exposures (fine coating of dust, heavy coating of dust, broken windows, damage in the home or furnishings, presence of debris in the home). The covariates also included four replacement behaviors (replaced AC, carpets, drapes, furniture). Lastly, each model contained terms for eight confounders, as listed in Section 3.2 (Table 3).The cleaning practices themselves, other than the cleaning behavior specified outcome, are also modeled as predictors. These other cleaning practices were conditioned on the presence of dust in the home. For example, “cleaned ventilation ducts” was modelled including the other three cleaning practices as predictors (i.e., dusted or swept without water; used a vacuum to clean; and cleaned with damp cloth, sponge, or mop). The covariate cleaning practices were considered only if reported in conjunction with the presence of either fine or heavy dust (e.g., vacuum and fine or heavy dust had to be “yes” for the derived covariate “vacuum” to be “yes”). We fitted the four logistic models with all 17 variables included. Adjusted odds ratios (aOR), and their 95% confidence intervals (95% CI), for each cleaning method are provided in Section 3.2 (Table 3). Each health outcome variable was modelled in a similar way. Based on medical considerations and likelihood of real relationships being present, the list of all covariates was narrowed to 21 explanatory variables for health outcomes (again, ‘purposive’ or a priori selection). Each of the six health outcomes was modeled using logistic regression with covariates as follows: 
+          
+            (a)
+            Five demographic variables (age at 9/11, race, sex, education, and income);
+          
+          
+            (b)
+            Resident priority group, indicating degree of registrant recruitment by area (partially a surrogate for location relative to the WTC site);
+          
+          
+            (c)
+            Dust cloud exposure category (intense, some, none);
+          
+          
+            (d)
+            Ever having smoked (at least 100 cigarettes in a lifetime);
+          
+          
+            (e)
+            Five variables of exposures in the home (broken windows, debris in the home, fine dust, heavy dust, and damage in the home);
+          
+          
+            (f)
+            Four variables containing data on behavior regarding replacement of various major items in the home (having replaced air-conditioning, carpeting, drapes, and furniture); and
+          
+          
+            (g)
+            Four variables for the presence of cleaning practices, which are conditioned on the presence of dust in the home.
+          
+        Five demographic variables (age at 9/11, race, sex, education, and income);Resident priority group, indicating degree of registrant recruitment by area (partially a surrogate for location relative to the WTC site);Dust cloud exposure category (intense, some, none);Ever having smoked (at least 100 cigarettes in a lifetime);Five variables of exposures in the home (broken windows, debris in the home, fine dust, heavy dust, and damage in the home);Four variables containing data on behavior regarding replacement of various major items in the home (having replaced air-conditioning, carpeting, drapes, and furniture); andFour variables for the presence of cleaning practices, which are conditioned on the presence of dust in the home.Again, the covariate cleaning practices were considered only if reported in conjunction with the presence of either fine or heavy dust (e.g., vacuum and fine or heavy dust had to be “yes” for the derived covariate “vacuum” to be “yes”). Smoking status was queried as “Have (you) smoked at least 100 cigarettes in (your) entire life?” Adjusted odds ratios (aOR), and their 95% confidence intervals (95% CI), for each health outcome were computed for all variables in the model including demographic, exposure, and cleaning practices. All analyses were performed using SAS version 9.3 (SAS Institute Inc, Cary, NC, USA). A total of 6447 lower Manhattan residents were included in this study. Table 1 shows their demographic characteristics. Mean age on 9/11 was 45.1 years (±15.1 years). Around 42% were male and 67% had completed college or post-graduate work. Forty-five percent had ever smoked cigarettes and 44% reported some or intense dust cloud exposure on 9/11. Figure 1, Figure 2 and Figure 3, respectively, display the geospatial distributions of self-reported outdoor location at the time of first encountering the dust cloud by intensity of exposure, indoor dust and damage to residences, and any reported respiratory symptoms up to 6 years after the 9/11 attacks. Figure 1 and Figure 2 show more frequent reports of higher indoor and outdoor exposures on edges of lower Manhattan, especially the western side along the Hudson River. Over 50% of WTCHR enrollees residing in census tracts in both the lower western and eastern sides of Manhattan reported having dust or debris in the home. The proportion reporting indoor home exposure declined rapidly further north and east up to the Registry Canal Street boundary for residents. The distribution of respiratory symptoms depicted in the Figure 3 map overlaps in large part with the distribution of reported residential exposures. The percentage of enrollees by census tract who reported any respiratory symptom ranged from less than 1% furthest North and East approaching Canal Street to 50% or more of enrollee respondents who lived in census tracts along the Hudson River adjacent to the WTC site.Table 2 shows the types of home damage and household cleaning practices of WTCHR lower Manhattan residents. The majority of residents reported a fine coating of dust on interior surfaces, while 17% had a heavy coating of dust, enough to make it impossible to see what was underneath. Only 6% had broken windows and less than 15% reported damage to home or furnishings. Presence of debris was reported by 12% of the respondents. Cleaning practices varied greatly: more than half of participants reported cleaning with a damp cloth, sponge, or mop, whereas 23% reported dusting or sweeping without water. Around 20% of participants replaced carpets, furniture, or drapes, blinds, and curtains. Almost one third replaced air conditioners (individually, or the building). Figure 2 shows that the majority of respondents reporting any type of home damage were located near the WTC site.Table 3 shows that those who dusted or swept without water were more likely to report the presence of debris (aOR = 1.31; 95% Confidence Interval (CI) 1.06–1.61) as to other reported exposures.Using a damp cloth, sponge, or mop was also associated with reporting a fine coating of dust (aOR = 1.85; 95% CI 1.59–2.14). In contrast, those who used water for cleaning were less likely to report broken windows. Lower Manhattan residents who reported using a vacuum to clean, as well as those who reported cleaning ventilation ducts, were less likely to report a fine or a heavy coating of dust, but more likely to report damage to home or furnishings, Table 3). Cleaning practices were also associated with each other, with cleaning with damp cloth, sponge, or mop having the strongest association with using a vacuum to clean (aOR = 4.71; 95% CI 4.04–5.50) and cleaning ventilation ducts (aOR = 5.28; 95% CI 3.72–7.50). These results were expected, yet including these terms made for more thorough modeling. Prevalence of respiratory outcomes was as follows: shortness of breath (16.1 %), wheezing (10.7%), chronic cough (6.8%), upper respiratory symptoms (60.8%), asthma or reactive airways dysfunction syndrome (RADS) (7.9%), and chronic obstructive pulmonary disease (COPD) (5.4%) (Table 4). A total of 68% of lower Manhattan residents reporting any respiratory outcome were located within 0.5 miles from Ground Zero (Figure 3). Table 5 shows odds ratios and 95% confidence intervals for respiratory outcomes in relation to several characteristics of home damage and cleaning practices. (See explanatory comments given previously, as Table 3 and Table 5 have the same layout). Those who reported a heavy coating of dust had 65% (95% CI 1.24–2.18), 43% (95% CI 1.03–1.97) and 59% (95% CI 1.09–2.28) higher odds of reporting shortness of breath, wheezing, or chronic cough, respectively. Residents who reported damage to home or furnishings had a 33% (95% CI 1.01–1.75) increased odds of reporting shortness of breath and 36% (95% CI 1.06–1.74) increased odds of reporting upper respiratory symptoms. Presence of debris was associated with chronic cough (aOR = 1.56; CI 1.12–2.17) and upper respiratory symptoms (aOR = 1.56; CI 1.24–1.95). Lower Manhattan residents who reported replacing air conditioners had higher odds of reporting upper respiratory symptoms (aOR = 1.32; CI 1.14–1.54). Replacing carpets was associated with increased COPD (aOR = 1.49; CI 1.03–2.16), while replacing drapes was associates with increased shortness of breath (aOR = 1.31; CI 1.03–1.65). Dusting or sweeping without water was the cleaning behavior associated with the largest number (three) of respiratory outcomes. Residents who did that had higher odds of reporting shortness of breath (aOR = 1.37; 95% CI 1.11–1.69), wheezing (aOR = 1.49; CI 1.17–1.90), and upper respiratory symptoms (aOR = 1.28; CI 1.08–1.53). Cleaning with a damp cloth, sponge, or mop and cleaning ventilation ducts were also associated with some respiratory outcomes including URS for damp cloth (aOR = 1.40; 95% CI 1.18–1.64) and wheezing and cough with ventilation duct cleaning (aOR = 1.48; 95% CI 1.08–2.01 and aOR = 1.94; 95% CI 1.38–2.73, respectively).This is the largest study to evaluate respiratory outcomes among lower Manhattan residents who reported household exposures as a result of the 9/11 terrorist attacks. Our findings are consistent with increased respiratory symptoms and diseases, which are associated with several levels of home damage and different cleaning practices, and corroborate other studies in similar populations [8,9,10,11,13] or subset of the WTCHR resident population [13].The explosion and collapse of buildings and subsequent fires at Ground Zero produced a large plume of dust and smoke that released particles and gases into the air. Characterization of both outdoor and indoor dust samples identified asbestos, glass fibers, lead, and polycyclic aromatic hydrocarbons (PAHs), among numerous other contaminants [1,4,5]. In fact, indoor samples contained more inhalable dust particles than did outdoor dust [6]. Our data show that the majority of lower Manhattan residents in the WTCHR may have been exposed to airborne contaminants because of some type of home damage following the 9/11 attacks. The presence of dust was the most common problem, but more severe damage, such as broken windows and furniture, were also reported. A case-control study comparing lower and upper Manhattan residents found that 30.7% of residents in the affected region reported some physical damage to their homes and 86.4% reported dust [8]. As expected, our geospatial analysis demonstrates that a higher proportion of WTCHR participants reporting any home damage were located in the vicinity of Ground Zero.As a result of the presence of dust, debris, and other damage, lower Manhattan residents personally cleaned their homes using a variety of methods. Lin et al. reported that 74.3% of residents in the affected area personally cleaned their homes [8]. The majority of residents participating in the present study reported using wet cleaning practices, such as using a damp cloth, sponge, or mop, which are likely to minimize suspension of dust and have been recommended by the Environmental Protection Agency (EPA) in that context [3,15]. However, many residents reported dusting or sweeping without water and even cleaning ventilation ducts by themselves. These practices were associated with the presence of debris and may have resulted in excessive dust exposure. In addition, numerous residents reported that household items, such as carpets, furniture, curtains, and air conditioners were replaced, which may have been associated with more severe damage to their homes. Among other measures, the EPA designed and implemented a Residential Dust Cleanup Program in 2002, to ensure that lower Manhattan residents were protected from potential WTC-related exposures [16].Upper respiratory symptoms were the most common health outcomes reported by lower Manhattan residents in the WTCHR. This finding is consistent with other WTC studies among residents [10] and rescue and recovery workers [17,18,19]. According to Lin et al., residents living within 1.5 km of the WTC site experienced a 200 percent increase of at least one persistent upper respiratory symptom compared to controls in the Upper West Side of Manhattan [10]. Our study also demonstrates high rates of persistent shortness of breath, wheezing, and chronic cough among lower Manhattan residents. These findings are in line with those of Lin et al., who reported that increased rates of lower respiratory symptoms persisted among lower Manhattan residents two and four years after the 9/11 attacks [9]. Moreover, most residents reporting any persistent respiratory symptoms in our study were located within a 0.5-mile radius of Ground Zero. Lin et al. also described an apparent trend when comparing residents below and not below Canal St.: the former had stronger associations between location and both any new-onset and persistent new-onset respiratory symptoms [9]. Reibman et al. also reported that four times as many lower Manhattan residents presented wheezing compared to a control population. In addition, three times as many residents reported cough and shortness of breath [11].Our results demonstrate an association between respiratory health outcomes and numerous types of home damage and cleaning practices. In particular, the presence of a heavy coating of dust was associated with shortness of breath, wheezing and cough and the presence of debris with chronic cough and upper respiratory symptoms. Lin et al. reported that upper respiratory symptoms were associated with both dust and physical damage to home, with statistically significant cumulative incidence ratios (CIRs) ranging from 1.27 to 1.71. In addition, the association between lower respiratory symptoms and exposures yielded CIRs between 1.31 and 1.44 [8]. In our study, the use of dry cleaning practices, such as dusting or sweeping without water, was associated with shortness of breath, wheezing, and upper respiratory symptoms. In contrast, only one health outcome was associated with cleaning with a damp cloth, sponge, or mop. It is not clear if the cleaning practice with water provided some protective effects by limiting re-suspension of dust, since in our analysis the cleaning practices with and without water were not mutually exclusive. Both practices were also associated with other methods of cleaning. Furthermore, cleaning practice with water was significantly associated with fine coating of dust (OR = 1.85; 95% CI, 1.59–2.14) versus cleaning practice without water was not (OR = 0.99; 95% CI, 0.82–1.20). Nevertheless, Lin et al. did not find statistically significant associations between self-cleaning of homes and lower respiratory symptoms at 2- and 4-year follow ups of lower Manhattan residents (N = 136 and 69 participants, respectively) [9]. A strength of this study is the ability to demonstrate persistent health outcomes in a large sample of lower Manhattan residents exposed in their homes to airborne contaminants resulting from the 9/11 terrorist attacks. Our analyses controlled for potential confounders, such as the intensity of exposure to the cloud of dust and debris, smoking, and priority group or area of recruitment, which may serve as a surrogate for location in lower Manhattan. However, this study is subject to several limitations. Self-selection bias may affect our results. Lower Manhattan residents who developed respiratory symptoms and diseases after 9/11/2001 may have been more likely to enroll in the WTC registry than asymptomatic persons. Medical records were not obtained which might have been used to verify reported conditions. In addition, smoking was defined as ever/never, without taking into account pack-years. Also, there was essentially no ability to validate the reported exposures quantitatively, because environmental sampling was not conducted. Nonetheless, it is unlikely that differences between respondents and non-respondents and misclassification of disease status would have an influence on the intensity of home damage reporting and selection of cleaning practices. Recall bias is also a limitation since questionnaires for W2 were applied over 5 years after the event of interest. In an attempt to account for this aspect, we verified responses on W1 and W2 for consistency. In addition, information on time to return home and use of respiratory protection while cleaning was not available. It is possible that severely damaged home inhabitants took longer to return to their homes or did so only after renovations, which potentially reduced dust exposures. Similarly, residents may have used different degrees of respiratory protection while cleaning surfaces. These factors would have reduced the apparent effect of the reported home damage on respiratory health outcomes.This analysis demonstrates that lower Manhattan residents who suffered home damage following the 9/11 attacks are more likely to report respiratory symptoms and diseases in the WTCHR. It also highlights the specific kinds of damage or other specific exposure events, such as cleaning practices, which are statistically related to such increased symptoms and diseases. These health outcomes persisted for at least 5–6 years after the event, which may have translated into lower quality of life. Conceptualization, V.C.A., R.M.B., M.R.F., L.L.P., S.L.G., J.E.C.; Formal analysis, L.L.P., B.L., H.E.A.; Methodology, V.C.A., L.L.P., S.L.G., H.E.A., Y.K.S., S.B., J.H.S.; Project Administration, V.C.A., M.R.F., R.M.B., J.E.C.; Writing, V.C.A., J.H.S., L.L.P., Y.K.S., R.M.B.; Reviewing and editing, V.C.A., L.L.P., J.H.S., Y.K.S., J.E.C., M.R.F., H.E.A., R.M.B.This study was supported by Cooperative Agreement U50/ATU272750 from the Agency for Toxic Substances and Disease Registry (ATSDR) and the Centers for Disease Control and Prevention (CDC), which included support from the National Center for Environmental Health and by Cooperative Agreement U50/OH009739 from the National Institute for Occupational Safety and Health (NIOSH) and the New York City Department of Health and Mental Hygiene. The findings and conclusions in this report are those of the authors and do not necessarily represent the views of the ATSDR or CDC/NIOSH.The authors declare they have no competing financial interests.Inverse distance weighted (IDW) interpolation of reported levels of dust cloud intensity (1 = none, 2 = some or 3 = intense) among World Trade Center (WTC) Health Registry respondents (n = 23,466) (Wave 1 survey, 9/2003–11/2004).Proportion of World Trade Center (WTC) Health Registry participants reporting any fine coating of dust on surfaces, heavy coating of dust on surfaces, broken window (s), damage to home or furnishings, or debris in their Lower Manhattan residences (n = 6348) (Wave 2 survey, 11/2006–12/2007).Proportion of World Trade Center (WTC) Health Registry participants reporting any respiratory outcome (shortness of breath, wheezing, chronic cough, upper respiratory symptoms, asthma/reactive airways dysfunction syndrome, chronic obstructive pulmonary disease) (n = 6348) (Wave 2 survey, 11/2006–12/2007).Demographic characteristics of 6447 Lower Manhattan residents enrolled in the World Trade Center (WTC) Health Registry.* Percentages do not reflect total N due to missing values.Self-reported household exposures resulting from the 9/11 attacks, cleaning practices, and other characteristics among 6447 Lower Manhattan residents enrolled in the World Trade Center (WTC) Health Registry.* Changes in denominators are due to non-response.Associations between cleaning practices and home exposures among 6,447 Lower Manhattan residents enrolled in the World Trade Center (WTC) Health Registry.* Each model has terms for eight confounders: models are adjusted for any effect due to age at 9/11, education, income level, race, sex, smoking status, and dust cloud exposure and resident priority group. - = not in the particular logistic model. Each column gives a distinct logistic model fit, with the column heading being the dependent (response) cleaning practice variable. (Statistically significant adjusted odds ratios in bold). Prevalence of self-reported post-9/11, respiratory symptoms and respiratory diseases present at Wave 2 (W2) (11/2006–12/2007) surveys among 6447 Lower Manhattan residents enrolled in the World Trade Center (WTC) Health Registry.* Changes in denominators are due to non-response. † RADS = Reactive airways dysfunction syndrome.Post-9/11 respiratory symptoms or diseases present at W2 in relation to home exposures and cleaning practices among 6447 Lower Manhattan residents enrolled in the World Trade Center (WTC) Health Registry.* Each model has terms for eight confounders: models are adjusted for any effect due to age at 9/11, education, income level, race, sex, smoking status, and dust cloud exposure and resident priority group. ** URS = Upper respiratory symptoms. † RADS = Reactive airways dysfunction syndrome. ‡ COPD = Chronic obstructive pulmonary disease. Each column gives a separate logistic model fit, with the column heading being the dependent (response) health outcome variable. (Statistically significant adjusted odds ratios in bold).

Med-MDPI/ijerph_3/ijerph-16-05-00799.txt

@@ -0,0 +1 @@
+The degree of emotional labor in nurses has been associated with negative physical and psychological health indices. The purpose of this study was to examine the relationship between emotional labor and depressive and anxiety symptoms in nurses. Specifically, the study addressed the question of whether anger suppression and anger rumination sequentially mediated the relationship. A total of 99 nurses was recruited from a university hospital in Korea. The questionnaires included instruments assessing emotional labor, anger suppression, anger rumination, as well as depressive and anxiety symptoms. Consistent with our hypothesis, there was a significant indirect effect of anger suppression and anger rumination on the relationship between emotional labor and depressive symptoms, as well as on the relationship between emotional labor and anxiety symptoms. The nurses’ degree of emotional labor, anger suppression, and anger rumination was associated with psychological adjustment. Thus, the impact of nurses’ negative affect needs to be adequately addressed, as inadequate resolution of anger may increase their vulnerability to experience depressive and anxiety symptoms. These findings may contribute to developing a strategy for enhancing nurses’ psychological health.In recent years, the psychological adjustment of the emotional laborers has received a great deal of attention. The need for research on emotional labor experienced by nurses has become inevitable. Emotional labor demands the induction or suppression of feelings to maintain a particular outward appearance [1]. Also, the service provision and emotional expression of nurses needs to appear genuine for patients under circumstances where extreme or negative feelings have been triggered. Therefore, nursing involves emotional caring that requires a balance of emotional engagement and detachment, as hospitals are increasingly demanding emotional labor [2]. Nurses are likely to encounter difficulties in regulating their emotions due to emotional dissonance—a discrepancy between emotions they are actually experiencing and emotions they are required to express.Korean nurses may struggle with difficulties in emotional labor due to the suppression of emotions in order to conform to patients’ and organizational demands. Research suggests that Korean nurses report a greater degree of anger and stress than other professions. Among nurses who reported significant stress, 44.4% chose anger as the most frequently experienced emotion, and they suppressed anger more than they expressed anger [3]. It was concluded that nurses tend to experience negative emotions, in particular anger, and that the emotions are usually suppressed due to job demands.In particular, anger suppression significantly affects psychological adjustment. Previous studies suggested that individuals who engage in greater anger suppression usually display a higher degree of depression and anxiety [4]. Cox, Van Velsor, and Hulgus also reported that women who cannot adequately express anger and opt to divert anger via avoidance of anger awareness or by using indirect means to cope with anger are more vulnerable to depression and anxiety than women who express anger [5]. From these findings, it can be assumed that nurses, who tend to experience frequent anger but also exhibit a tendency to suppress anger, may be particularly vulnerable to depressive and anxiety symptoms.The suppression of negative emotions is associated with rumination, which may eventually cause, maintain, or increase depressive and anxiety symptoms [6]. The potential role of rumination in the context of emotional suppression is partially supported by research examining the relationship between rumination and thought suppression. These studies showed that consciously suppressing unwanted thoughts can ironically increase the availability of the suppressed thoughts [7]. Thus, it is possible that suppression may fuel rumination, which may subsequently increase susceptibility to experiencing depressive and anxiety symptoms.Rumination is not only involved with a low activity mood, such as sadness, but it may also be involved with hyperactive mood states, such as anger. Rumination in this context has been termed “anger rumination”. Anger rumination can be described as the tendency to persistently deliberate on past events that had triggered anger [8]. Although similar in concept, anger rumination has been suggested as an independent cognitive process which follows anger provocation, thereby maintaining or enhancing anger mood and anger-related behavior problems. In addition, higher use of rumination as a coping strategy in an anger-provoking situation contributes to depression and anxiety [8].In summary, extant research indicates that in the process of emotional labor exerting influence on the degree of depressive and anxiety symptoms, the anger-related mechanism may play a significant role. Based on the assumption that there is an integrative causal association between emotional labor, anger suppression and rumination, as well as depressive and anxiety symptoms, the present study sought to empirically examine this possibility. In other words, nurses are required to suppress their anger evoked from emotional labor, and ruminating about anger-provoking incidents and anger mood may likely intensify the problem and possibly contribute to depressive and anxiety symptoms. The present study specifically addressed the question of whether anger suppression and anger rumination sequentially mediate the relationship in order to examine its causal relationship.The current study adopted a cross-sectional self-report survey design in order to examine the sequential mediating effect of anger suppression and anger rumination in the relationship between emotional labor and depressive and anxiety symptoms of nurses. The study protocol was approved by the Yonsei University Institutional Review Board (Approval No. 7001988-201612-SB-102-02).The participants in the study were 99 nurses who worked for at least six months at a university hospital in the Republic of Korea. They were recruited during January of 2017. Data were collected via the distribution of self-report questionnaires given to all nurses in the hospital who expressed interest in participating in the study. Of the 110 surveys distributed, 99 surveys were returned. All 99 surveys were usable for further data analysis without problems concerning missing data or multiple responses. Of the 99 respondents, the mean age was 30 years (standard deviation (SD) = 7.17) and 98.0% were women. The majority of participants were staff nurses (n = 90, 90.9%) who had been working in their current position between 12 and 60 months (n = 51, 51.5%). Their work pattern consisted mostly of triple shifts (n = 89, 89.9%), whereas nine participants (9.1%) reported a non-shifting work schedule (Table 1). The work shifts consisted of a day shift (7:30–15:00), an evening shift (15:00–22:00), and a night shift (22:00–7:30).Emotional Labor—The Korean-Emotional Labor Scale (K-ELS), developed and validated by Jang et al., and based on the characteristics of emotional labor in Korea, was used to measure emotional labor [9]. This tool was composed of 26 items with five subscales, and the total score was used in the present study. Sample items of the K-ELS are “I make efforts not to express negative feelings towards customers” and “I get hurt in the process of facing customers.” The items were rated on a four-point Likert scale ranging from 1 (strongly disagree) to 4 (strongly agree). In the present study, the Cronbach’s alpha coefficient for the K-ELS scale was 0.87.Anger Suppression—The Korean State-Trait Anger Expression Inventory (STAXI-K) [10], which is a modified and validated version of the State-Trait Anger Expression Inventory (STAXI) [11], was used to measure anger suppression. This tool is composed of 10 items for each trait anger and state anger, and eight items for each anger-in, anger-out, and anger-control. The items are rated on a four-point Likert scale ranging from 1 (strongly disagree) to 4 (strongly agree). For this study’s purpose, only the anger-in subscale from the state anger scale was used. Sample state anger scale items of the STAXI-K are “I am angry” and “I yell.” A higher anger-in total score indicates more frequent anger suppression. In the present study, the Cronbach’s alpha coefficient for the anger-in subscale was 0.79.Anger Rumination—The Korean Anger Rumination Scale (K-ARS) [12], which is a modified and validated version of the Anger Rumination Scale (ARS) [13], was used to measure anger rumination. The items are rated on a four-point Likert scale ranging from 1 (almost never) to 4 (almost always in terms of how well the items resemble their beliefs about themselves; higher scores correspond to a greater degree of anger rumination. In the present study, the Cronbach’s alpha coefficient for the K-ARS scale was 0.95.Depressive Symptoms—Depressive symptoms were assessed using the Korean version of the Center for Epidemiologic Studies Depression Scale (CES-D) [14], originally developed by the National Institute of Mental Health [15]. This is a 20-item self-report questionnaire that measures the frequency and severity of depressive symptoms experienced within the last week. The items are rated on a four-point Likert scale ranging from 0 (very rare, i.e., less than one day during one week) to 3 (most of the time, i.e., more than five days during one week). In the present study, the Cronbach’s alpha coefficient was 0.93.Anxiety Symptoms—The State-Trait Anxiety Inventory (STAI), originally developed by Spielberger, Gorsuch, Lushene, Vagg, and Jacobs [16] and adapted for use in Korea [17], was used to measure anxiety symptoms. It measures state anxiety (20 items) and trait anxiety (20 items), where state anxiety represents anxiety that is temporary and changing, and trait anxiety represents anxiety that is permanent and stable. For the purpose of this study, only 20 items measuring trait anxiety were used. The items are rated on a four-point Likert scale ranging from 1 (not at all) to 4 (very much). In the present study, the Cronbach’s alpha coefficient for trait anxiety was 0.91.The sequential multiple mediation analysis suggested by Hayes was conducted using PROCESS Macro for SPSS (IBM Corp., Armonk, NY, USA) [18]. The significance of all the possible indirect effects from the study’s model were examined using the bootstrapping method of 5000 resamples. The 95% confidence intervals that did not include zero indicated a significant indirect effect. Previous research suggested that work shift pattern, length of work experience, and job position may influence nurses’ degree of depressive symptoms [19,20]. Therefore, these demographic variables were controlled as covariates in the statistical analyses.Table 2 presents descriptive statistics and intercorrelations between the study variables. Emotional labor was positively correlated with both anger suppression (r = 0.43, p < 0.001) and anger rumination (r = 0.30, p = 0.003). Anger suppression demonstrated significant positive correlations with depressive (r = 0.30, p = 0.003) and anxiety (r = 0.24, p = 0.020) symptoms. Anger rumination was also positively associated with depressive (r = 0.42, p < 0.001) and anxiety (r = 0.35, p = 0.001) symptoms. However, the degree of emotional labor did not show significant correlations with either depressive (r = 0.13, p = 0.215) or anxiety (r = 0.11, p = 0.289) symptoms.Sequential mediation analyses were conducted after controlling for the work shift pattern, the length of work experience, and the job position as aforementioned (Table 3).Depression—There was a significant total indirect effect, which was the sum of all possible indirect effects in the model between emotional labor and depressive symptoms (effect = 0.16; confidence interval (CI): from 0.05 to 0.33), but a non-significant direct effect (effect = −0.07, t = −0.63, p = 0.531). Regarding each indirect effect, there was a significant indirect effect indicating that the path from anger suppression to anger rumination sequentially mediated the relationship between emotional labor and depressive symptoms (X→M1→M2→Y: effect = 0.12; CI: from 0.04 to 0.26). The indirect effects from emotional labor to anger suppression to depressive symptoms (X→M1→Y: effect = 0.01; CI: from −0.12 to 0.14) and from emotional labor to anger rumination to depressive symptoms (X→M2→Y: effect = 0.04; CI: from −0.03 to0.18) were not significant. The coefficients and standard error (SE) for each path are presented in Figure 1.Anxiety—The total indirect effect was significant (effect = 0.12; CI: from 0.02 to 0.27), but there was a non-significant direct effect (b = −0.03, t = −0.27, p = 0.790). Regarding each indirect effect, there was a significant indirect effect of emotional labor on anxiety through anger suppression and anger rumination, again indicating sequential mediation (X→M1→M2→Y: effect = 0.09; CI: from 0.02 to 0.22). The indirect effects from emotional labor to anger suppression to anxiety (X→M1→Y: effect = −0.02; CI: from −0.14 to 0.12) and the indirect effect from emotional labor to anger rumination to anxiety (X→M2→Y: effect = 0.04; CI: from −0.02 to 0.16) were not significant. The coefficients and SE for each path are presented in Figure 2.Consistent with our hypothesis, the results indicated that emotional labor had an indirect sequential effect on depressive and anxiety symptoms through anger suppression and anger rumination, and the direct effect of emotional labor on depression and anxiety was not significant after controlling for anger suppression and anger rumination. Hence, the nurses’ emotional labor itself may not serve as a risk factor per se for psychological maladjustment. In addition, anger suppression or rumination may not lead to emotional problems. The results of the present study indicate that nurses who experience a high degree of emotional labor, who tended to suppress anger and who also ruminated about the causes and effects of anger-provoking situations, may maintain or intensify anger mood, which ultimately increased their vulnerability to experience depressive and anxiety symptoms.Interestingly, the indirect effect of anger suppression alone was not significant. Specifically, emotional labor was associated with anger suppression, but the pathway from anger suppression to depressive and anxiety symptoms was not significant. It is still an ongoing discussion as to whether emotional suppression always demonstrates a negative effect on psychological adjustment. There are results from meta-analytic studies which suggest that emotional suppression exhibits a negative impact on an individual’s psychological health, but at the same time, there have also been studies that have reported on adaptive aspects of emotional suppression [21,22]. Thus, it cannot be assumed that simply suppressing negative emotions always leads to psychological maladjustment. Moreover, the indirect effect of anger rumination on the path from emotional labor to depressive and anxiety symptoms was not significant. Specifically, anger rumination had a significant effect on depressive and anxiety symptoms, but the path from emotional labor to anger rumination was not significant. Based on our results, the cognitive process of repetitively recalling anger-provoking situations (rumination) may have a significant impact in aggravating depressive and anxiety symptoms, especially when anger is suppressed.In spite of the results from previous studies which reported that emotional labor has a negative effect on psychological adjustment [23], emotional labor and depressive and anxiety symptoms were not significantly correlated in the current study. This result may be related to the characteristic of emotional labor. The two main emotion regulation strategies used by emotional laborers are surface acting, (i.e., displaying nongenuine desired emotions) and deep acting, (i.e., modifying their genuine emotions to actually feel the desired and situationally required emotions) [1]. According to Drach-Zahavy, Yagil, and Cohen, surface acting requires constant effort to suppress negative emotion, which leads to the depletion of psychological resources and as a result impedes one’s well-being [24]. On the other hand, deep acting may work as a protective factor in psychological adjustment, since it also involves efforts to maintain a positive view in negative or stressful situations. Empirical research has also demonstrated that this kind of cognitive effort may have a positive effect. In addition, in social interactions, deep acting may assist in the formation of relatively positive relationships and increase the opportunity to receive positive feedback from others by presenting authentic emotions during interactions [25]. If this effect is applied to nurses, deep acting may contribute to the formation of relatively positive interactions with patients, and nurses may also experience job satisfaction through potential positive feedback. These opposite possibilities may simultaneously affect depressive and anxiety symptoms.The present study may provide valuable information for constructing a specific intervention strategy to protect nurses’ psychological adjustment. More specifically, training in the use of different kinds of adaptive emotional regulation strategies other than anger suppression and anger rumination may be needed. For example, cognitive reappraisal was found to be more effective at reducing and handling anger than efforts to suppress anger [26]. Furthermore, positive reappraisal (i.e., searching for positive aspects or meaning instead of ruminating about anger-provoking events) and positive refocusing (i.e., focusing on pleasant and positive experiences unrelated to the experienced negative events) may be effectively utilized [27]. Supplementing clinically validated effective interventions for depressive and anxiety disorders, such as cognitive behavior therapy (CBT), with such emotion regulation training components may further effectively reduce nurses’ psychological discomfort such as anger and depressive and anxiety symptoms.As reported in the meta-analysis results of Hülsheger and Schewe, there may be a significant moderating effect of cultural differences in the relationship between emotional labor and psychological distress [28]. Eastern cultures are characterized as an interdependent culture that values conformity, and the suppression of emotional expression for the sake of group harmony is frequently demanded [29]. Thus, there may exist a mediating effect of cultural differences on the sequential mediating effect of anger suppression and anger rumination found in the present study. Hence, the results of the current study need to be replicated through cross-cultural research. If differences are found, research identifying the cultural factors that contribute to the differences also needs to be conducted.There is a possibility that the university hospital worked as a protective factor compared to other work environments. According to previous studies, work organization, structure, and policy have a significant impact on employees’ anger [30]. University hospitals have relatively well-established organizational support and a protective system that may work in some degree as a protective factor for the nurses’ anger [10]. Thus, examining the organizational characteristics of nurses working at private hospitals and other emotional laborers that may influence the result is suggested for further research.Also, since the current study was conducted using cross-sectional data, we cannot confirm long-term variation trends in anger suppression and rumination and subsequent depression and anxiety. It is uncertain whether it is a temporary or a long-term effect that anger suppression and rumination have on depressive and anxiety symptoms. A study by Takebe et al. longitudinally examined the relationship between anger suppression and anger rumination [9]. Thus, future studies that examine the long-term effects on depressive and anxiety symptoms are suggested, based on our finding that these variables are significantly associated.The present study suggested the existence of a potential mechanism by which anger that is evoked from emotional labor may be related to depressive and anxiety symptoms. Thus, hospitals should acknowledge that it is critical to manage nurses’ anger through the construction of a specific intervention strategy to protect nurses’ psychological adjustment.The conception and design of the study, or acquisition of data: J.E.K., J.H.P., and S.H.P. The drafting the article or revising it critically for important intellectual content: J.E.K., J.H.P., and S.H.P. The final approval of the version to be submitted: J.E.K., J.H.P., and S.H.P.This work was supported by the National Research Foundation of Korea (NRF) (grant number NRF-2016-11-1385) and the Basic Science Research Program of the National Research Foundation of Korea funded by the Ministry of Science (grant number NRF-2015R1A2A2A04006136).All the procedures performed in the studies involving human participants were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards.On behalf of all authors, the corresponding author states that there is no conflict of interest.Sequential mediation model for emotional labor and depression via anger suppression and anger rumination. Unstandardized path coefficients and SE indicated above. The coefficient appearing above the line connecting emotional labor and depression represents the total effect and the coefficient below the line represents the direct effect. *** p < 0.001.Sequential mediation model for emotional labor and anxiety via anger suppression and anger rumination. The unstandardized path coefficients and SE are indicated above. The coefficient appearing above the line connecting emotional labor and anxiety represents the total effect and the coefficient below the line represents the direct effect. ** p < 0.01, *** p < 0.001.Summary of demographic characteristics (N = 99).Summary of intercorrelations, means, and standard deviations for study variables.Note: Anger suppression (AngerSup); Korean version of the Center for Epidemiologic Studies Depression Scale (CES-D); Korean Anger Rumination Scale (KARS); Korean-Emotional Labor Scale (K-ELS); Korean State-Trait Anxiety Inventory (STAI).Sequential mediation effect of anger suppression and anger rumination in the relationship between emotional labor and depression/anxiety.Note: Emotional labor (X); anger suppression (M1); anger rumination (M2); lower levels for 95% confidence interval (LLCI); upper levels for 95% confidence interval (ULCI).

Med-MDPI/ijerph_3/ijerph-16-05-00800.txt

@@ -0,0 +1,4 @@
+Cross-contamination between occupants in an indoor space may occur due to transfer of infectious aerosols. Computational fluid dynamics (CFD) provides detailed insight into particle transport in indoor spaces. However, such simulations are site-specific. This study couples CFD with statistical moments and establishes a framework that transitions site-specific results to generating guidelines for designing “healthy” indoor spaces. Eighteen cases were simulated, and three parameters were assessed: inlet/outlet location, air changes per hour, and the presence/absence of desks. Aerosol release due to a simulated “sneeze” in a two-dimensional ventilated space was applied as a test case. Mean, standard deviation, and skewness of the velocity profiles and particle locations gave an overall picture of the spread and movement of the air flow in the domain. A parameter or configuration did not dominate the values, confirming the significance of considering the combined influence of multiple parameters for determining localized air-flow characteristics. Particle clustering occurred more when the inlet was positioned above the outlet. The particle dispersion pattern could be classified into two time zones: “near time”, <60 s, and “far time”, >120 s. Based on dosage, the 18 cases were classified into three groups ranging from worst case scenario to best case scenario.Computational fluid dynamics (CFD) has been used as a powerful simulation platform for air flow, thermal distribution, and contaminant and particle transport in the indoor environment for the past 20 years and more. Different CFD models for a range of geometries and ventilation patterns were generated and validated [1,2,3,4,5,6]. The simulations gave detailed insight into the influence of building parameters and indoor air quality (IAQ). However, the knowledge gained from CFD simulations is yet to benefit or influence decision-making or guideline development regarding exposure and infections for occupants of an interior space. The literature published on CFD and indoor air, from 1994 to 2018, dropped from ~1400 to less than 150 and 20 when refined by the words “exposure” and “infection”. Spenglar and Chen [7] in 2000 in the Annual Review of Energy and Environment showcased the potential of CFD to become the design tool for the future, especially to meet the requirement of healthy indoor environments. IAQ must ensure the “health” and “well-being” of the residents as declared by World Health Organization (WHO) in 2000 [8]. ASHRAE standards (62.1 and 62.2) evolved over the years to provide occupants with health, comfort, and productivity [9]. In recent years, indoor environments and, hence, IAQ took on new dimensions with factors such as designing for energy efficiency and sustainability. Guidelines focused on thermal, acoustic, and visual comforts, as well as sick building syndrome [10]. The impact of IAQ on the occupants is multi-fold and is a result of the interactions between the parameters influencing IAQ and the factors impacting health and well-being of residents [11]. There is strong and sufficient evidence to establish relationships between ventilation, air-flow pattern, and the spread of infectious diseases in buildings [12]. One route of transmission is via aerosols, defined as “person-to-person transmission of pathogens through the air by means of inhalation of infectious particles”. Studies show that ventilation rates of 25 L/s per person have the potential to reduce sick building syndrome and sick leave, while low ventilation rates at schools can have a negative impact on school absence and respiratory illness [3]. The efficiency of a ventilation system is tied to the task it is performing [4,5]. Office design has some impact on performance, e.g., women in open-floor plans reported higher long-term sick leave spells [6]. However, increasing ventilation rates can also lead to a negative impact with increased outdoor pollutants coming indoors. Personalized designs, such as introducing chair fans in conjunction with displacement ventilation were also studied to reduce exposure of occupants to pollutants, i.e., particles in the breathing zone [7]. Further investigations into assessing parameters influencing exposure risk between occupants showed that a person-to-person distance of less than 1.1 m in an office increases chances of infection [11]. We spend 90% of our time indoors, and buildings are responsible for the consumption of 20% to 40% of energy in the developed world [13]. To ensure acceptable IAQ levels, without increasing energy costs, identifying and assessing the influence of indoor space parameters such as location of desks and windows relative to engineering controls such as vent locations is vital. Different ventilation patterns were tested to identify the ideal location for a printer so as to reduce particle concentration in the breathing zone [14], and ventilation strategies, particle concentration, and removal efficiency were shown to be dependent on source location [15]. CFD simulations can extract detailed information on the influence of the micro space parameters and IAQ, but the information provided is based on a specific location and its application for general guidelines is limited. In this investigation, a series of simulations are conducted varying different parameters, and the worst and best exposure scenarios are determined through dosage estimation. Idealized two-dimensional large eddy simulations (2D LES) are conducted to understand and extract the “interferences” of the various parameters. The 2D LES approach was applied to understand downstream effects for high-rise buildings [16] and gave accurate information regarding the mechanisms governing vortex shedding around bluff bodies [17]. The study explores coupling CFD simulations with statistical approaches. It proposes a framework that can lead to applying three-dimensional (3D) CFD results in generating guidelines and standards for “healthy” indoor spaces, reducing the transfer of infectious aerosols between occupants. The study assesses the potential of statistical moments to translate CFD data on velocity profiles and particle transport into information relating to indoor conditions and possibility of exposure. Computational domain: The geometry of the computational domain for a representative case is shown in Figure 1a,b. The source of particles in the space is due to respiratory release represented as a “sneeze”. The dimensions of the 2D space are 4.88 × 3.05 m2, representing an office room shared by two occupants. The partition wall height is 1.22 m and is located 2 m from the inlet side. The inlet and outlet positions and the presence or absence of desks are dependent on the specific case as listed in Table 1. A total of 18 cases are simulated. The parameters varied to change the scenario for the cases are hI and hO, which are the height of the inlet and outlet from the floor of the domain, respectively; Dwp, which is the distance from the wall to the partition in the exhale zone; and Dwo, which is the distance from the wall or the side where the inlet is located to the first obstruction encountered from that wall. The “breathing zone” is in the range of 1 to 2 m from the floor. This zone is further separated into the exhalation and inhalation zones. In the “exhalation zone”, the residing occupant sneezes at a height of 1.07 m from the floor, representing a person sitting at a desk, and the “inhalation zone” is where the second office occupant may inhale the exhaled aerosols. It is assumed that the presence of any aerosol in the breathing zone has the potential of causing infection. The sneeze particles are released in an area measuring 0.7 m in length and 0.5 m in height. The area is divided into six segments where 150,000 aerosols are injected randomly with a velocity range of 6–22 m/s with a particle size of 7 μm, assuming a sneeze volume of 1 L [11,18]. To simulate the high momentum at the point of release and the subsequent loss of energy, it is assumed that the particles released close to the source have the maximum velocity, and, as the distance increases, the particle velocity decreases. Mathematical formulation: The Eulerian–Lagrangian framework was applied to simulate the transport of particles in the indoor space. Details of the development were given in Reference [19], and the numerical approach was validated against different geometries [20]. The flow dynamics of the room were captured via large eddy simulation (LES) [21,22]. LES provides an instantaneous velocity field required to calculate the particle trajectories [22]. Equations (1) and (2) solve for mass and momentum conservation, respectively, where ui is the velocity component in x, y, z directions, and τij is the sub-grid scale stress term which was resolved by applying the Smagorinsky model, τij−τkkδij/3=2νtS¯ij, where S¯ij=(∂u¯i/∂xj+∂u¯j/∂xj)/2; finally, the eddy viscosity νt is obtained from νt=CsΔ2|S¯|S¯ij, where ∆ is the grid size, Cs is the Smagorinsky constant, and |S¯|=(2S¯ijS¯ij)1/2 [23]. 
+        (1)∂u¯i/∂xi=0;
+        (2)∂u¯i∂t+∂∂xj(u¯iu¯j)=−1ρ∂p¯∂xi−∂τij∂xj+νt∂2u¯i∂xj∂xj.Particle trajectories were determined applying the force balance equation ∑Fi=m(duP/dt)=FD+FG, where up is the particle velocity of mass m, at location xi, FG is the net gravitational force, FG=(1/6)πdp3(ρp−ρ)g, and FD is the drag force, FD=(1/8C)πdp2ρCD|VR|VR. C is the Cunningham correction factor, CD is the drag coefficient, VR is the relative velocity, dp is the particle diameter, ρp is particle density, and ρ is fluid density. Boundary conditions: The boundary condition of the walls of the room represents no-slip conditions. Uniform velocity profiles are specified at the inlets. The outlets are pressure boundaries with Dirichlet conditions and Neumann conditions applied for all other dependent variables. Particle trajectory calculations terminated when they exited the room. The simulation assumed that the aerosols were spherical and neutral, and no energy loss occurred during its interaction with surfaces. This was assumed to simulate the worst-case scenario, i.e., all expirated aerosols remain in the ambient atmosphere for the longest duration. No evaporation occurs, i.e., the particle size remains constant. One-way coupling was applied, and resuspension was not considered. Solver settings: The equations were discretized based on finite volume techniques having second-order accuracy for time and spatial derivatives. The numerical simulations were conducted using CFD-ACE+ [24]. A blended second-order upwind scheme was applied to resolve the convective diffusive terms, and the temporal terms were solved via the implicit Euler scheme. The convergence criterion was set at 10−7. For each case, the flow field was first generated for 240 s, based on air changes per hour and the dimensions of the geometry; this ensured the domain was filled with air before particle release via a “sneeze”. The simulation was terminated 300 s after the sneeze. Grid independence: A structured computational grid was applied with finer grid resolution at the boundaries, the furniture, and partitions. The grid independence was tested using three resolutions for the case in Figure 1. In LES, to obtain grid independence, often the grid has to be refined until it nearly reaches a DNS (direct numerical simulation)-level resolution and, as a result, loses the LES fundamentals. Hence, the grid convergence index (GCI) approach [25,26] was applied to assess the uncertainty associated with each step of grid resolution for predicting the number of particles remaining in the room, applying GCI=Fs[(fcoarse−ffine)/(1−rP)], where f=(N/No), Fs = 3 (safety factor), r = grid ratio, and p = 2 (formal order of accuracy). The GCI index was 8% when the grids were refined from 25,000 to 50,000, and <2% when the grids went from 50,000 to 80,000, indicating a grid resolution of 50,000 results in an error margin of <2% compared to 80,000. In the current study, the grid resolution of 50,000 was chosen for all the simulations.Design plan: To assess the impact of (1) relative inlet/outlet location, hI/ho, (2) air changes per hour (ACH), and (3) presence and absence of desks, Dwo/Dwp, the design plan was executed as follows:(1) when the inlet and outlet of the domain were located at the top, the ratio hI/ho = 1; when the inlet and outlet were located at opposite walls with the inlet positioned higher than the outlet, hI/ho > 1; and when the inlet was located on the opposite wall but positioned lower than the outlet, then hI/ho < 1, (2)the air changes per hour were set at three levels: 3, 5, and 7 and (3) for the scenarios with desks, the ratio Dwo/Dwp < 1, and in the cases without desks, Dwo/Dwp < 1. Table 1 summarizes the dimensional configuration of the 18 cases with the corresponding schematic. Analysis approach: Spatial and temporal data on velocity and particles were extracted from all cases. The effects of the parameters, hI/ho, ACH, and Dwo/Dwp, on the air flow pattern and on the temporal and spatial particle distribution in the total domain and in the breathing zone were assessed. Statistical moments were applied to describe and distinguish these scenarios and the dosage was calculated. Parameter configurations significantly influencing air flow and particle distribution were determined. The analysis was done with the MATLAB’s statistical package [27].The velocity contour plots in Figure 2 show the resulting air-flow pattern for cases 2, 5, 8, 11, 16, and 18. Cases 2, 5, 8, and 11 were at ACH = 5. In Figure 2a, where both inlet and outlet were in the ceiling (case 2), the circulations with higher velocities were in the upper regions of the room and at the corners. For case 5, Figure 2b, with all other configurations the same as case 2, the presence of desks did not appear to change the air-flow pattern significantly. In Figure 2c,d, cases 8 and 11 appear similar, even though desks were present for the latter and absent for the former. Figure 2e,f show the velocity contour plots for ACH = 3 and 7 when hI/ho < 1 in the presence of desks. The presence of the desks in front of the inlet resulted in a sharp upturn of the air flow at the inlet into the domain, resulting in a different air-flow pattern in comparison to when hI/ho = 1 or hI/ho > 1. The plots confirm the combined influence that indoor parameters have on the air flow in the domain, but it is difficult to distinguish which parameter or configuration is better or worse for the occupants’ well-being.The contour plots do, however, confirm that increasing ACH resulted in fewer locations in the domain where the velocity magnitudes were <0.01 m/s. The regions where such low velocities occurred remained the same for all configurations, as seen in Figure 2. To assess the impact of the parameters on the regions with low velocities, the number of nodes with velocities less than 0.005 m/s was counted and normalized against the total number of nodes in the breathing and non-breathing zones. These were designated as dead zones. The bar charts in Figure 3 compare the percentage of dead zones for the 18 cases. The breathing zones are in blue shades and non-breathing zones are in red shades. The figure shows that increasing ACH, i.e., cases 1, 2, and 3 or cases 4, 5, and 6, resulted in a decreasing percentage of dead zones in the whole domain and in the breathing zone. The percentage of dead zones in the domain nearly halved when air changes were increased from 3 to 5, though the decrease when ACH was increased from 5 to 7 was not as consistent.Comparing cases without desks (Dwo/Dwp = 1) and with desks (Dwo/Dwp < 1), the percentage of dead zones in both regions increased when the inlet and outlet were located at the top, for ACH = 3 (cases 1 and 4). When hI/ho > 1 (cases 7 and 10) and hI/ho < 1 (cases 13 and 16), the dead zone percentage decreased with the inclusion of desks for the whole domain. At ACH = 5, dead zone percentage increased for all regions when hI/ho = 1 (cases 2 and 5) and hI/ho > 1 (cases 8 and 11), and it increased for the breathing zone only when hI/ho < 1 (cases 14 and 17). The dead zone percentage decreased for the non-breathing zones for all cases, with cases 14 and 17 as the only exceptions, when desks were included. At ACH = 7, the exception also occurred when hI/ho < 1 (cases 15 and 18) for the non-breathing zone as well, with a slight increase in the percentage of dead zones when desks were included. Overall, it appears that, when the inlet and outlet location satisfies hI/ho < 1, the trend differs from the other configurations. This can be due to the location of the desks in the domain relative to the inlet position.Figure 4 shows line plots of the normalized velocity for cases 2, 3; 5, 6; 8, 9; 11, 12; 14, 15; and 17, 18 at x = 1.1 m and 3.05 m in the exhale and inhale zones (locations shown in Figure 2d using line probes). Figure 4a compares the velocity profiles for cases 2 and 3 where ACH increases from 5 to 7. The air-flow inlet and outlet were located at the top, i.e., hI/ho = 1, and no desks were present, Dwo/Dwp = 1. Increasing the ACH caused a vortex or recirculation zone to form near the domain for ACH = 7; otherwise, the line plots were similar for both locations. Figure 4b shows cases 5 and 6 with desks, which resulted in velocities near zero at desk heights. Figure 4c,d show line plots for cases 8, 9; and 11, 12, with the inlet located above the outlet, hI/ho > 1. Maximum velocity values occurred at the inlet height for the exhale side, with the lines nearly overlaying. For the inhale side, the plots appeared to “smooth” out as inlet effects diminished (hI/ho < 1 for Figure 4e,f). The peaks were reversed for the exhale and inhale zones in Figure 4e. The presence of desks resulted in sharp peaks at the lower end of the domain (Figure 4f). Overall, the position of the inlet/outlet on the flow pattern appeared to have a lesser effect when hI/ho > 1, and desks appeared to have a lesser effect when hI/ho = 1. Figure 4 shows the influence of the different configurations on the air-flow pattern, and it highlights the difficulty in comparing the effects of the multiple configurations. To better interpret the results across all 18 cases, statistical moments were applied next. The mean, standard deviation, and the skewness were calculated for the average velocities across the room height at x = 1.1 m and 3.05 m. The means increased from ~0.004 m/s to ~0.007 m/s to ~0.010 m/s as the ACH increased from 3 to 7, and they were very nearly the same value for both sides of the partition. Table 2 lists the values for standard deviation and skewness for both exhale and inhale zones, and the change going from one zone to the other. The standard deviation increased with increasing ACH for each zone. Comparing exhale and inhale sides, an impact of the configuration can be seen. The standard deviation decreased from the exhale to the inhale region when hI/ho = 1 and Dwo/Dwp = 1. There was a slight increase when desks were present and ACH = 5 and 7. This indicates that the air flow does not gain momentum for this configuration, suggesting the possibility that contaminant dispersion is influenced mainly by the conditions at the inlet side of the domain. With desks present, the air flow was interrupted and, at higher ACH, some momentum was carried forward, resulting in a rise in the standard deviation. A decreasing trend from the exhalation zone to the inhalation zone can be seen for hI/ho < 1 in the absence of any desks. In the presence of desks, however, the standard deviation increased for all cases moving from the exhale to the inhale side for hI/ho > 1.Skewness gives the direction of the total mass of air. A negative skew indicates that the mass of the air flow is toward x = 0, and a positive skew shows that the mass of air is flowing toward the room end, i.e., where the outlet is located. In the exhale zone, when the inlet and outlet were located at the top, the mass of air flow was toward the inlet, i.e., skewness was negative for cases 1 to 4. In all other cases, for the exhale zone, the air flow was directed toward the outlet. Cases 1, 2, and 3 had no desks. With no desks to break the incoming air stream from the inlet located in the ceiling, the air mass had more space to move in either or both directions. This can also be seen for case 4, where, at ACH = 3, even though desks were present, the air dispersed before the air stream hit the desks. For the inhale zone, cases 2, 9, 12, and 16 had negative skew values. All these cases had unique configurations, and skewness gives insight into the impact of the different configurations. Case 2 is the only case for the group with hI/ho = 1 where the flow of the air mass was toward the exhale zone from the inhale zone (ACH = 5 and no desks were present). For case 9, the group of cases 7, 8, and 9 had the same parameter values (hI/ho > 1, Dwo/Dwp = 1), except for increasing ACH from 3 to 5 to 7, respectively. Cases 10, 11, and 12 were also the same (hI/ho > 1, Dwo/Dwp < 1) except for ACH. One group was without desks and the other was with desks. At ACH = 7, the flow direction was opposite for cases 9 and 12. Case 16 with ACH = 3, on the other hand, also had a negative skew value. For case 16, hI/ho < 1 and desks were present. Skewness transitioned from negative to positive and vice versa for the cases 1, 3, 4, 9, 12, and 16. For cases 1, 3, and 4, where the common factor was hI/ho = 1, the transition was negative to positive. For cases 9, 12 (hI/ho > 1), and case 16 (hI/ho < 1), the transition was positive to negative, i.e., the mass of air moved toward the exhale zone, away from the inlet end, and then reversed direction. To obtain an overall picture of the spatial distribution of the particles, the average and standard moments of the location of the particles for x and y coordinates at the end of the simulation were determined. Figure 5 shows the average and standard deviation (x and y coordinates) for the 18 cases. The values differed for all cases, which confirms that every configuration of ventilation pattern, ACH, and desks resulted in a unique spatial distribution of the particles. A higher value of the mean or average x and y indicated that particles were located toward the inhalation zone or that more particles were located near the ceiling, respectively. Standard deviation quantifies the clustering of the particles in the domain around the average. For example, particles in case 2 (ACH = 5, hI/ho = 1, Dwo/Dwp = 1) were, on average, within 2 m of the entrance, occupying the region right below the breathing zone, but dispersed more in the horizontal direction, staying within a meter of the domain’s floor. For case 17 (ACH = 5, hI/ho < 1, Dwo/Dwp < 1), on the other hand, particles moved to the inhalation zone and were in the breathing region, but clustered in that location.Increasing ACH from 3 to 5 generally pushed the particles toward the inhale zone, for example, in cases 1, 2; and cases 7, 8; except for cases 13, 14 where hI/ho < 1 and Dwo/Dwp = 1, and for cases 17, 18 where hI/ho < 1 and Dwo/Dwp < 1. There was no clear trend when ACH increased from 5 to 7. Most particles remained in the exhale zone for case 7. For case 17, most particles appeared to be in the inhale zone. For cases 13 to 16, the particles were above the breathing zone, whereas, for all the remaining scenarios, the particles were within the height of the breathing zone. Cases 13 to 16 had the common configuration of hI/ho < 1. Once ACH increased from 3 to 5 and 7 for cases 17 and 18, the particles were at the height within the breathing zone.Scanning through the standard deviation values (Figure 6b), it could be concluded that no specific parameter appears to dominate the particle dispersion for both x and y directions. Increasing ACH from 3 to 5 resulted in an increase in the standard deviation value for x when the inlet and outlet were located at the top, i.e., hI/ho = 1 (cases 1, 2; cases 4, 5). For hI/ho > 1 (cases 7, 8; cases 10, 11) and hI/ho < 1 (cases 13, 14; cases 16, 17), the increase was not consistent. When ACH = 7, higher spread occurred for case 12 (ACH = 7, hI/ho > 1, Dwo/Dwp < 1) and case 18 (ACH = 7, hI/ho < 1, Dwo/Dwp < 1). However, particles in case 16 (ACH = 3, hI/ho < 1, Dwo/Dwp < 1) also had a standard deviation value nearly the same as cases 12 and 18, even though the ACH was 3. The least particle dispersion occurred for the cases 7, 10, 11, and 14 for the x value. The values for y location followed the same trend as x and were always less than x, except for case 14 (ACH = 5, hI/ho < 1, Dwo/Dwp = 1), where dispersion of the particles was slightly more in the vertical direction than in the horizontal direction.The standard deviation values for air flow in Table 2 were compared to the trends in Figure 5b. Focusing on cases with relatively more clustering for both x and y values (i.e., cases 7, 10, 11, 14, and 17), it can be seen that smaller magnitudes of standard deviation for air flow were seen for cases 7 (~0.019 both sides) and 10 (~0.013 both sides) when ACH = 3 and hI/ho > 1 in the absence and presence of desks, respectively, compared to the other cases. However, for cases 11, 14, and 17, the standard deviation for air flow was within the magnitude of the other cases (~0.021 to ~0.029) even though the standard deviation values for particles indicated clustering. In these three cases, the inlet and outlet were at the opposite end (case 11, hI/ho > 1; case 14, hI/ho < 1; and case 17, hI/ho < 1) and ACH was either 5 (case 11, 17) or 7 (case 14). Desks were present for cases 11 and 17 but absent for case 14. The temporal evolution of the particle number in the breathing zone and in the whole domain for all cases is shown in Figure 6. The plots show the unique impact of each configuration of the 18 simulations. The effect of increasing ACH from 3 to 7 is shown in every plot. The first column in the figure is for the cases with the presence of the partition only, Dwo/Dwp = 1, and the second column is for the cases with both partition and desks, Dwo/Dwp < 1. The first row is for hI/ho = 1, the second row is for hI/ho > 1, and the third row is for hI/ho < 1. Lines in the plot represent particle number evolution in the whole domain, and lines with markers represent evolution in the breathing zone only. Assessing the influence of increasing air changes, at ACH = 7, a larger number of particles left the domain compared to at ACH 3 and 5. However, the least removal also occurred when hI/ho < 1 and Dwo/Dwp = 1 for ACH = 7 (case 15). There is no clear interpretation as to the impact of the presence or absence of desks or the inlet/outlet locations. When the inlet and outlet were located at the top, in the presence of desks, more particles exited the room for ACH = 5 and 7. There was little or no change in the trend for ACH = 3. When hI/ho < 1 for ACH = 5 and 7, the presence of desks was observed to have the same effect; however, overall, a smaller number of particles were removed when compared to hI/ho = 1. For hI/ho > 1, the presence of desks resulted in the entrapment of higher particles in the whole domain, and the number leaving the domain was small. Particles cycled in and out of the breathing zone with the air flow. The peak of the cycles was dependent on the total remaining particles in the domain. Hence, more particles returned to the breathing zone in the following cycle if particles remained in the room. Cases 13, 14, and 15 (Figure 6e) illustrate this clearly. For ACH = 3, it appears that particles in the breathing zone left the room; however, for ACH = 5 and 7, the pattern indicates a return earlier for ACH = 7 than that for ACH = 5. Hence, for ACH = 3, the particles return to the breathing zone after a longer time period. The peak of the cyclic behavior coincides with the trend line for the change in the total number of particles in the domain.To assess the temporal trend for all cases, the average and standard deviation change over time for the x and y locations of the particles in the whole domain and in the breathing zone were plotted (Figure 7). The dispersion trend of the particles could be classified under “near time”, <60 s after release, and “far time”, >120 s after release, along the horizontal direction of the room (Figure 7a,c). For the first 60 s, the average location of the particles was on the inlet side, around the exhale zone, before dispersing afterward. After that, the effects of the room configuration appeared to take over with the average value of x location, i.e., the spatial distance from the inlet side or the exhale region, increasing. The value of x was mainly higher for higher ACH and for the configuration hI/ho > 1, while it was lower for ACH = 3. Figure 7c is the corresponding temporal change of the standard deviation of the spatial x locations. The standard deviation is a representation of the particle clustering trend. Within the first minute of release, the standard deviation was ~0.5 or less for all cases. After 120 s, the particles started dispersing for some scenarios, and, for others, the particles remained clustered for the entire particle tracking duration. Figure 7b shows a distinct difference in the average location of the particles along the room height. Particles released in the configuration where the inlet/outlet was located at the top appeared to congregate in the lower portion of the space (below 1.5 m) for all ACH and in the presence and absence of desks. Case 7 was the only exception, where the particles constantly stayed at the same average height. For all cases, the particles congregated in and around the breathing zone. Figure 7d shows the corresponding standard deviation change of the particle positions for y. The standard deviation increased beyond 120 s.The total number of particles inhaled over time was calculated using Equation (3), where Dt is the dosage or the number of particles inhaled over a specific time period t, Npt is the number of particles (maximum, minimum, or mean) in the breathing zone for time t, f is the fraction of particles that will enter the respiratory tracts based on a particle size of 0.05, and fRt accounts for breathing over the time period t, assuming a representative adult population and that the amount of particles that can be breathed in t time is Npt [28]. 
+        (3)Dt=Npt×f×fRt.Figure 8a shows the results for the maximum and minimum dosage for the 18 cases, and Figure 8b shows the change in average dosage over 300 s. The maximum inhaled dosage increased exponentially and overlapped for all cases, with the maximum value at 4 × 104. For the minimum value, it varied with the highest being near 1 × 104 for cases in which hI/ho > 1 and desks were present with the exception of when ACH = 7 (case 12). The minimum hovered near zero and decreased for the other configurations of hI/ho. There appeared to be a change in the behavior after 120 s and then after 180 s. Figure 8b captures the average number of particles exposed to over the time period. The plot is broken down into three groups. Group 1 refers to the cases where there was an exponential rise. The average dosage in group 1 was also closer to the maximum line in Figure 8a. Group 2 refers to cases which had an exponential rise until 120 s or 180 s and then leveled off. The dosage amounts also fell between the maximum and minimum boundaries. Group 3 covers the remaining cases, which fell near the minimum band of Figure 8a. Some cases of this group also started with an exponential rise, but this rise was less steep than that in group 1. Others transitioned into an exponential rise after 120 s or 180 s. Table 3 classifies the groups and the related cases with their specific configurations. In group 1, hI/ho > 1 for all cases and ACH = 7 was absent. Hence, the “worst-case scenario” occurred when the inlet was located above the outlet for lower air changes. At higher air changes, the effects of the location were minimized. The six cases in group 2 were equally divided with two cases each for the ratio hI/ho, i.e., when ACH = 7, hI/ho > 1, and when ACH = 3, hI/ho < 1, further illustrating the inter-dependent relationship between the relative location of inlet/outlet and air changes per hour. In group 3, which can be classified as the “best-case scenario”, there were four cases for the inlet and outlet located at the top, and the remaining four cases for the inlet located below the outlet. The influence of the presence or absence of desks was neutral, comparing the number of cases in group 1. In group 2, there were more cases with desks than without desks. Transitions at different times occurred in the cases in group 2, indicating that desks influenced the outcomes. Group 3 justified having no desks, as cases where no desks were present dominated the group, indicating the inhalation dosage of particles increased for the occupants when desks were present. This study established a framework of extracting information from CFD simulations and coupling it with statistical moments for identifying guidelines for minimizing exposure to occupants. Eighteen cases were simulated applying a 2D LES Eulerian–Lagrangian framework. The impact of the inlet/outlet location (hI/ho), air changes per hour (ACH), and presence/absence of desks (Dwo/Dwp) was assessed. If inlet/outlet were both located on the ceiling, then hI/ho = 1; if the inlet was located above the outlet, then hI/ho > 1; and if the inlet was located below the outlet, then it was assigned as hI/ho < 1. ACH was assigned three values (3, 5, or 7), and desk presence was denoted by Dwo/Dwp < 1 and Dwo/Dwp = 1. The effects on the air-flow pattern and particle trajectories were unique for the different configurations. To interpret the information and translate it to guidelines, the average velocity and particle number in the room was assessed using the first three moments: mean, standard deviation, and skewness. Mean, standard deviation, and skewness of the velocity profiles gave an overall picture of the spread and movement of the air flow in the domain. Inlet/outlet locations influenced the change in the standard deviation of the velocities at the exhale and inhale zones, with the air flow “spreading” into the domain for hI/ho < 1. Skewness captured the transition of mass of air-flow direction going from negative (toward exhale zone) to positive (toward inhale zone). A single parameter or configuration did not dominate, confirming the significance of considering the combined influence of multiple parameters for determining localized air-flow characteristics. The spatial distribution of the particles and the temporal variation of their number was also tracked. The average and standard deviation of the x and y locations of the particles at the end of the simulation were calculated. For hI/ho < 1, more particles stayed above the breathing zone compared to other scenarios. There was a clear trend of particles being located near the outlet for ACH increasing from 3 to 5, while, from 5 to 7, the same conclusion could not be drawn. However, more particles left the room for ACH = 7 from the whole domain compared to other values of ACH. More particle dispersion occurred in the x direction when the inlet and outlet were located at the top, with more clustering occurring for hI/ho > 1, with the influence of ACH dominating when ACH = 7. The temporal change of the average and standard deviation values of x and y showed that the particle dispersion pattern could be classified into two time zones: “near time”, <60 s, and “far time”, >120 s. While the maximum dosage inhaled was the same for all scenarios, the minimum exposure trend changed after 120 s, with the exposure either decreasing or continuing to rise, with another change occurring in the trend at 180 s for a few cases. Based on the average dosage intake, the 18 cases were separated into three groups. The relative dosage that the occupants were exposed to was dependent on multiple variables. The combination of the different variables resulted in “groups” of configurations ranging from worst-case to best-case scenario. The cases in group 3 (best-case scenario) all had decreasing standard-deviation trends for velocity profiles (except for case 1 which had a slight increase). All air changes were present in this group. The position of the inlet above the outlet was absent in this group. Group 2 had cases where skewness transitioned, and the dosage trend flattened after an initial exponential rise. The dead zone percentages in the breathing regions for the cases in groups 1 to 3 were compared. No specific trend was noted. For the worst-case scenario, the inlet was always above the outlet, and ACH = 7 was absent. These results can vary based on the physics introduced and by changing the geometry to a 3D domain. In a 3D domain, mixing will be enhanced and, hence, the spatial and temporal distribution, and the associated dosage will differ. However, by identifying the minimum, average, and maximum dosage trends, the best- and worst-case scenarios can be categorized. The framework can be applied to identify design approaches that may limit exposure. Additionally, the results can guide the selection of cases that might have to be simulated in a 3D domain to extract additional information and insight. The 2D LES approach gave a clear picture of the influence of the multiple variables and different configurations on the air-flow pattern and particle transport. The simulation results coupled with the inferences from statistical moments gave a map toward utilizing the information from more detailed simulations. However, the flow obtained was not a perfect representation of a real-life flow field in a 3D room, and the effects resulting from changing the room configuration in each case were, therefore, enhanced to a certain level. To capture the 3D physics governing the flow and particle transport, the framework established in the study can be extended to a series of 3D simulations. The steps outlined in the current study will have to be expanded to account for the influence of the third dimension. The results obtained will change with the introduction of physico-chemical mechanisms governing particle fate and transport as required, such as the evaporation of the sneeze droplets in a space maintained at a specific temperature and humidity. Conceptualization, S.H.; formal analysis, S.H. and F.B.O.; investigation, S.H.; methodology, F.B.O.; supervision, S.H.; writing—original draft, S.H.This research received no external funding.The authors declare no conflicts of interest.Problem scenario: (a) partition only; (b) partition and desks. Dimensions are in meters.Velocity contour plots at air changes per hour (ACH) = 5 for (a) case 2, (b) case 5, (c) case 8, and (d) case 11; velocity contour plot at ACH = 3 for (e) case 16, and at ACH = 7 for (f) case 18.Percentage of dead zones in breathing and non-breathing parts of the domain.Average velocity line plots; solid and dotted lines are the normalized velocities across the height of the room at exhale (x = 1.1 m) and inhale (x = 3.05 m) locations, respectively; lines without markers are for ACH = 5 and those with markers are for ACH = 7; (a) cases 2 and 3, (b) cases 5 and 6 (c) cases 8 and 9, (d) cases 11 and 12, (e) cases 14 and 15, and (f) cases 17 and 18.(a) Average (m) of x and y locations for the 18 cases; (b) standard deviation (m) for x and y locations for the 18 cases at the end of simulation.Temporal change of particle number in the domain and in the breathing zone (solid lines—whole domain; lines with markers—breathing zone; blue, red, and black represent ACH = 3, 5, and 7, respectively): (a) cases 1, 2, and 3; (b) cases 4, 5, and 6; (c) cases 7, 8, and 9; (d) cases 10, 11, and 12; (e) cases 13, 14, and 15; (f) cases 16, 17, and 18.(a) Average x; (b) average y; (c) standard devaiton of x; and (d) standard deviation of y for particle locations. Black = Dwo/Dwp = 1, red = Dwo/Dwp < 1; dotted lines: ACH = 3, dashed lines: ACH = 5, and solid lines: ACH = 7; no marker: hI/ho = 1, “o”: hI/ho >1, and “*”: hI/ho < 1.(a) Maximum and minimum, and (b) average number of particles that an adult is exposed to after a “sneeze” over 300 s or a five-minute period.Dimensional configuration of the 18 cases simulated. ACH—air changes per hour.Standard deviation and skewness to assess the changes from exhale to inhale zones.I = increase; D = decrease; SI = slight increase.Details of the configurations of the cases in each group.

Med-MDPI/ijerph_3/ijerph-16-05-00801.txt

@@ -0,0 +1 @@
+Objectives: Several studies have shown mortality and suicide risk in substance use disorders, and autopsy findings with respect to the used substances. However, there seems to be a gap in the knowledge about substances misused in life and at death at the within-person level. Methods: All consecutive, autopsied patients during 1993 to 1997, who had been in contact with the Addiction Centre in Malmö from 1968, were investigated (365 subjects). Drug misuse in the long-term course noted in case records was related to autopsy findings. Self-inflicted death (suicide/undetermined suicide/accidental overdose) was compared with natural death. Results: Benzodiazepine misuse was associated with a high risk of autopsy findings of the substance in suicide and death of undetermined intent. It was also associated among non-misusers, but less so. An alcohol level above 1‰ was found more often in self-inflicted death. Prescription opioids at autopsy were mainly found in self-inflicted death among non-misusers. Heroin misuse was related to overdose. Central nervous system stimulants (CNS-S) and cannabis were rarely found in self-inflicted death among previous misusers. The overlap between depression in life and antidepressants at death was low. Conclusions: Benzodiazepines and alcohol seem to disinhibit suicidal tendencies. Suicide risk among users of cannabis and CNS-S may be related to other risk factors than acute use. Implications for suicide prevention are discussed.Substance use disorders are associated with increased mortality, including self-inflicted death such as suicide, death of undetermined intent, and accidental overdoses by heroin [1,2,3,4]. Substances used and misused have also frequently been found in medico-legal samples.Among dependent or regular users of illicit opiates, overdose has been found to be the most common cause of increased mortality, and suicide is another major cause of death [5,6]. Premature death, including suicide, has been found among amphetamine users [7,8]. Findings about cannabis use and mortality have been contradictory [9,10]. An increased risk of suicide has been found among cannabis users, but this was explained as markers of psychological and behavioural problems rather than overdose [11]. Increased risk of accidental overdose among users of heroin and cannabis in a criminal justice sample has also been found [12]. In medico-legal studies. heroin has been found to be a main intoxicant among illicit drug misusers in Sweden and Norway, and use has increased [13,14]. In addition to heroin, other intoxicants include amphetamine, tetrahydrocannabinol (THC), benzodiazepines, ethanol [15] and cocaine [16].According to one study, misuse of benzodiazepines among psychiatric patients could not be related to increased mortality compared to controls with similar diagnoses but no misuse [17]. However, Tiihonen et al. found that use of benzodiazepines was associated with an increased risk of suicide, which did not apply to antidepressants and antipsychotics [18,19]. Benzodiazepines have increasingly been found in autopsy samples [15,20,21] during the past decade, with increasing poly-drug use [22], although these are sometimes seen as incidental findings [23] and only constitute a minor contribution to suicide [24,25]. To our knowledge, the risk of intoxication among users of prescription opioids has only been studied indirectly by comparing prescription rates with fatalities, and some investigators have found corresponding decreased rates with less prescription [26,27,28]. Recent studies have shown increased rates of death with increased prescription [29,30]. One autopsy study reported that fatal intoxications with prescription opioids (buprenorphine, codeine, dextropropoxyphene, fentanyl, methadone, oxycodone, tramadol) were found at increasing rates in Finland [31] and increasing rates of heroin and prescription opioids (including oxycodone) in the US [16].Alcohol use disorders are associated with an increased risk of completed suicide [3]. From a medico-legal point of view, ethanol appears to predominate among mono-intoxications [24,32] and is also commonly found in addition to other drugs [15]. In fatal intoxications, ethanol appears to predominate among mono-intoxications [24,32].Substance use disorders are, after depression, the most commonly found diagnoses in suicide victims [33,34]. Although suicide and death of undetermined intent are multifactorial, where diagnoses, psychosocial stressors, somatic disease, etc. are important risk factors, the substance itself contributes to the risk by its physical and psychological effects. This is why misused substances are commonly found at autopsy, and several studies have shown increased risk of mortality among those who misuse. However, there is little knowledge about the link between misuse in life and autopsy findings at the within-person level. In one study, it was found that substances found at autopsy had also been used at index admission [35]. Another study showed that the overlap between histories of help-seeking and substance misuse is not simple: some people with a history of misuse show no evidence of use at the time of death and in autopsies, and some with no record of seeking help show evidence of use at death [36]. Consequently, there is a gap in the knowledge concerning the extent to which people who misuse certain substances in life also use them at self-inflicted death on the within-person level. To address this gap, we have revisited a sample of substance users in Malmö, examining clinical data and autopsy findings.This study is based on a sample of patients treated at the Addiction Centre in Malmö from 1968 onwards, who died in the period 1993 to 1997, and who were autopsied at the Department of Forensics. A pseudo-prospective design was employed, enabling collection of independent information on these cases, including previous addiction, diagnoses, and causes of death. As this design was time-consuming, extending over several years, we chose to revisit the sample rather than start a new sample. The sample has been presented in a previous study on unnatural death and drugs used in life, concluding that number of substances used in life is related to an increased risk of death of undetermined intent and heroin overdoses, but not suicide [37]. A more recent study on the same sample showed that suicidal communication was more often considered non-serious before death of undetermined intent than before suicide, and that the undetermined group also showed higher levels of alcohol in blood at autopsy [38]. The sample does not reflect the substance use pattern of today, but the physical and psychological effects we intended to study remain the same.The aim of this study was to compare substances misused in life with autopsy findings on a within-person level, to explore the link between substance use disorder in life and autopsy findings. The study design enables exploration of this link, overlooked by previous studies on substance use disorders in life or autopsy findings only. Unnatural self-inflicted death (suicide, death of undetermined intent, and accidental overdoses) was compared with natural death. Differences in rates may be an indication that the substance had contributed to death. The difference between self-inflicted and natural death for “non-misusers” was also calculated. Risk of suicide, death of undetermined intent, and heroin overdoses during the follow-up by substance used in life were estimated. Violent death by external causes was also described. Background factors, such as number of substances used, psychiatric diagnoses (including depression), and non-fatal suicidal behaviour were considered. Finally, poisoning and violent self-inflicted death were compared, and drugs used at death were described.All persons autopsied during 1993–1997 at the Forensic Department were matched with the patient register for Addiction Centre in Malmö (originally called Clinical Alcohol Department), which included inpatients and outpatients at the clinic from 1968 and onwards, a follow-up of almost 30 years. Cases with toxicological data were included. The study procedure is presented in Figure 1.A forensic examination sampling procedure was used. This was carried out on all consecutive autopsies of patients who had been in contact with the Addiction Centre at Malmö University Hospital. In Sweden, forensic examination applies to most subjects who have died outside hospitals by suspected natural causes (disease) but with no medical history that can explain the death, or by non-natural manners (trauma including homicide, suicide, death of undetermined intent, and accidental fatal intoxications). Fatal intoxications could be intentional, as in suicide, of unknown intent, as in the case of death of undetermined intent, or probably unintentional, as is usually the case when the opioid drug previously used is involved [39]. The only accidental intoxications in the present sample were heroin overdoses.Suicide was defined as follows: “Different manners of non-natural death have different numbers of undetermined cases in terms of intent; for example, in a hanging or a shooting it is usually easy to differentiate between a suicide or a trauma (or a crime), while for drowning, traffic accidents or intoxication it is more difficult. Circumstantial findings, such as suicide notes, expressed intent or other findings such as self-inflicted cutting of the wrist followed by drowning, are suggestive of the intent” (E950-953, X60-84-ICD—International Classification of Disease—9 and 10) [40,41]. Death of undetermined intent was defined as follows: “When crime can be ruled out and it cannot be established whether the manner of death is a suicide or an accident, the manner of death is recorded as death of undetermined intent” (E980, Y10-34). Heroin overdose (E850, X40-49) was another cause of self-inflicted death, and it is mostly considered unintentional [6]. The coroner also assessed the contribution of any individual substances to death.Poisoning and violent (such as hanging, drowning, jumping, burning) methods of self-inflicted death, suicide, and undetermined intent, were noted. Substances found at autopsy, such as benzodiazepines, prescription opioids, and antidepressants, were recorded by manner of death.Death by trauma, such as fall from height, car accident, and homicide, was considered as violent death (when it was not a suicide). All other cases were considered as natural death, i.e., when the death was caused by disease alone. In some cases, toxicological data were not necessary, for instance some cases of trauma with no suspicion of intoxication.Mean age at death was 58 years for natural death, 52 for violent, 52 for undetermined intent, 49 for suicide, and 38 for heroin overdoses.In the period 1993 to 1997 inclusive, 394 consecutive forensic autopsies were performed on previous patients at the Department of Forensic Medicine in Lund. In one case of violent death, it could not be determined whether death was self-inflicted or caused by another person, and case records could not be found for five patients. In 23 cases there was no blood sampling for any substance, so the final sample comprised 365 patients: 319 men and 46 women.A pseudo-prospective design was used, in which investigation was carried out within a few days of death. One member of the research team (AF) performed the interviews with the staff at the Addiction Centre. As the interviews were performed shortly after death, neither the interviewer nor the interviewees knew the manner of death, whether it was suicide or not. This enabled a blind approach on a reasonable sample size and a reasonable length of follow-up time using the case records. The sampling was carried out in the period 1993–1997, but there have been no significant changes in methodology since then, and the physical and psychological effects of the substances remain the same. The interviewer (AF) blindly evaluated the records of those who had been in- or outpatients at the Addiction Centre in Malmö University Hospital. The items were recorded on a questionnaire, with reports on type and characteristics of the addiction of legal and illegal drugs, physical disease, hallucinations, etc. In this study, information on type of addiction, suicidal behaviour, and diagnoses according to ICD was recorded.Substance use was diagnosed according to ICD 8 and 9 [40,41] for all inpatients, who constituted 76% of the sample. The remaining 24% had been admitted as outpatients and had applied because they subjectively considered themselves to have a substance use problem and need treatment. It is safe to conclude that almost all outpatients fulfilled the criteria for alcohol dependence and/or had a drug problem. Up to 1994, all the patients treated at the Department of Clinical Alcohol Research were admitted for alcohol problems; after that date, some patients may have been treated for narcotic misuse only, i.e., not for alcohol problems, but in only four cases could alcohol use disorder not be confirmed, though it may still be suspected (1% of the total sample and 4% of the illegal drug users).Other psychiatric diagnoses according to ICD 8 and 9 [40,41] were also recorded, including misuse of prescribed substances, psychoses and complications to alcohol use disorders. Depression is an important risk factor for suicide, and notes regarding depression and antidepressant medication in the case records were recorded as depression.Misuse involved both prescribed and illegal drugs. The former was divided into benzodiazepines (and a few cases of “z drugs”) and prescription opioids (dextropropoxyphene and codeine; oxycodone was not misused in Sweden at the time); illegal drugs were divided into opioids, cannabis, and central nervous system stimulants (CNS-S), mainly amphetamine and a few cases of cocaine. All drug misuse was recorded regardless of whether it was the main drug or not. Drugs that are not misused, such as antipsychotics, were not recorded. However, use of antidepressants as a marker of depression was noted.We present the analysis by drug used in life and found at autopsy. An individual could use more than one drug in life, so more drugs could be found at autopsy. We compared number of drugs used in life and at autopsy for each drug to find out whether poly-drug use was more common for any particular drug as a possible confounder.Fisher’s exact test was used to compare substances used in life and found at autopsy. As there were multiple comparisons, a Bonferroni correction was applied and the significance was set at p < 0.0025. A multiple logistic regression was used for comparison between suicidal ideation, suicide attempt, diagnoses, and depression and the different substances used.Ethical approval was not required for deceased persons in Sweden at that time, and the case records all concerned deceased persons, i.e., none of the sample was still alive. However, the National Board of Forensic Medicine approved the study.Psychiatric diagnoses and previous non-fatal suicidal behaviour noted in the case records by substance use in life are presented in Table 1. In addition, there were 18 cases of diagnoses of substance use disorder by legal drugs, four of which were not contemporary to another psychiatric diagnosis. Other psychiatric diagnoses tended to be less often reported among cannabis users. Depression was also reported but rarely, in only 20 cases (5–8 per cent for each substance), with no significant difference between the substances used.Mean number of substances used in life by substances used are presented in Table 2. There were on average around two extra substances, slightly higher for heroin users, and somewhat lower for benzodiazepines.Manner of death related to substances used in life is presented in Table 3. Self-inflicted death was more common among patients who had used other substances in addition to alcohol (63–70% versus 31%), mainly because there were no heroin overdoses in the alcohol-only group. The suicide rates were similar for the different substances (around 10%). Death by undetermined intent was common, but somewhat less so among those who had used heroin or alcohol only.Among those who died by suicide or death by undetermined intent, there were 26 cases of violent method.Mean number of substances used in life by specific substances used are presented in Table 4. There were similar rates for the different substances used, but the rate was somewhat lower for benzodiazepines. Substances used in life related to autopsy findings by manner of death (unnatural or natural death) are presented in Figure 2.Heroin at autopsy among heroin users was significantly more often found in self-inflicted death compared with natural death (Fisher’s exact test: 23/30 vs. 0/13 p < 0.0001). Prescription opioids were also marginally significantly more often found in self-inflicted death (Fisher’s exact test: 18/24 vs. 3/9 p < 0.0441). The same association was found for benzodiazepines (Fisher’s exact test: 40/57 vs. 4/29 p < 0.0001). When self-inflicted death by heroin overdoses was excluded, use of benzodiazepines was the only substance that remained associated (Fisher’s exact test: 26/40 vs. 4/29 p < 0.0001).In contrast, CNS-S and cannabis were rarely found at autopsy among those who had used the substances in life, regardless of manner of death. No correlations could be shown.Among those who had used alcohol only, no significantly higher rates of alcohol use were found at autopsy in self-inflicted than natural death (63/148 (43%) vs. 52/156 (33%)). However, there were higher rates of alcohol level above 1‰ among those who died of self-inflicted death (Fisher’s 11/49 (22%) vs. 34/58 (58%) p < 0.0002).A comparison between substances found at autopsy by manner of death (self-inflicted or natural death) among non-misusers in life is shown in Figure 3.Illegal substances were rarely found, regardless of manner of death.Prescription opioids were more often found at autopsy in self-inflicted deaths compared with natural death among those who had not used the substance in life (Fisher’s exact test; 38/129 vs. 16/201, p < 0.0001). If heroin overdoses were excluded, the correlation remained significant (Fisher’s exact test: 16/201 vs. 27/118, p < 0.0003).Benzodiazepines were significantly more often found at autopsy in self-inflicted deaths compared with natural death even among those who did not have a known use/misuse during life (Fisher’s exact test: 32/96 vs. 13/165, p < 0.0001). If heroin overdoses were excluded, the result remained significant (Fisher’s exact test: 30/91 vs. 13/165 p < 0.0001).However, the rates were significantly higher for users than for non-misusers (Fisher’s exact test: 26/40 vs. 30/91 p < 0.0010).Alcohol intoxication was found in similar rates in violent and poisoning death (suicide and death of undetermined intent) (Fisher’s exact test 15/26 vs. 69/105, p < 0.4966.) Similar rates of alcohol levels above 1‰ were found in both types of death (Fisher’s exact test: 6/15 vs. 15/69; p < 0.1875).Prescription opioids at autopsy were related to poisoning in self-inflicted death (Fisher’s exact test: 33/105 vs. 1/26 p < 0.0025).Eighteen patients died by violent manner of death. Twelve were positive for high levels of alcohol; mean 3.08‰ (range 2.3–4.6‰). There was one case of benzodiazepine, one of cannabis, and four of CNS-S, but none of opioids. Only two were not positive for any drug. There were two cases of homicide, one positive for alcohol (3.4‰). Intoxication was therefore very common, mainly with respect to alcohol.The contribution of the substances found at autopsy were investigated. There were 56 cases of benzodiazepine found at autopsy in suicide and undetermined intent (30 among previous users), of which there was one case (2%) of major cause of death, 29 cases (52%) of contribution, and 26 cases (46%) of occasional findings, including z-drugs.For 34 cases of intoxication of prescription opioids in suicide and undetermined intent, there were 15 (44%) main causes of death, 12 (35%) contributions and 7 (21%) occasional. Twenty-four had used dextropropoxyphene; others had used mainly codeine.Among those who had used cannabis in life, eight had died by suicide by intoxication, but none had been using the drug at death. Three out of 14 who died by undetermined intent had used cannabis, and three out of 16 who died by heroin overdoses. Other intoxicants used were prescription opioids, ethanol, benzodiazepines, venlafaxin, paracetamol, and carbon monoxide.Among those who had used CNS-S, seven had died by suicide, one of whom had used amphetamine. Four out of 18 who died by undetermined intent had used CNS-S. Other stimulants used were ethanol, opioids, ketamine, paracetamol, karbamazepin, karisopropodol, teophyllin, and orphenadrin.Among those who had used heroin in life, four had died by suicide, three of whom had been using the drug at death. Seven had died by undetermined intent, one of whom had used heroin. One of those had possibly used heroin before death. They had also used ethanol, prescription opioids, benzodiazepines, karisopropodol, venlafaxin, orfenadrin, and karbamazepin.There was only one case of overlap between depression detected in life and antidepressants found at autopsy. Of the 20 patients with signs of depression according to case records, seven died by their own hand (35%) versus 146/181 (45%) not diagnosed with depression, thus no significantly higher suicide rate for depressives. Twenty-two positive cases of antidepressants were found at autopsy. Of these 18 died by their own hand vs. 134/324 among those negative (Fisher’s exact test; p < 0.0002). In four cases, antidepressants were considered the main cause of death, seven contributing together with other substances, and seven occasional, so antidepressants were considered as cause of death in more than half the cases (61%). Lithium was found in one case of death of undetermined intent.This study examines the link between substances used in life and autopsy findings on the within-person level.Benzodiazepine misuse was associated with self-inflicted death. It was also associated with self-inflicted death among non-misusers, but less so. For prescription opioids, there was a significant association with self-inflicted death among non-misusers, regardless of the inclusion of heroin overdoses. Heroin use was, as expected, associated with a risk of overdose. Cannabis and CNS-S were rarely found in self-inflicted death. Alcohol levels above 1‰ were associated with self-inflicted death, but there was no significant difference between poisoning or violent method. Prescription opioids were negatively associated with violent method. High levels of alcohol were often found in violent manner of death.According to a recent review of 17 studies, a positive correlation between prescription benzodiazepines and non-fatal and fatal suicidal behaviour was reported in most studies, with few exceptions [42]. This has since been questioned [43]. Epidemiological, clinical, laboratory-based, and neuro-biological studies were included. Possible mechanisms of pro-suicidal effects could be an increase in impulsivity or aggression. The reviewed articles included the two above-mentioned longitudinal studies [28,29], where there was an increased risk of suicide in people with schizophrenia, who were prescribed benzodiazepines, while antipsychotics reduced the risk in both studies and antidepressants in the latter. Other studies in the review reported prescription of benzodiazepines shortly before suicide. The present study gives further support for this view, by adding autopsy findings of benzodiazepines among previous users, thereby indicating that the substance was also used at the time of death. There was no indication of specific comorbidity to explain the increased risk, but there were only two cases of anxiety diagnoses (one with misuse of benzodiazepine), which may be an underestimate. We cannot definitely rule out self-medication at the time of suicide, nor that the findings represent withdrawal symptoms.Alcohol at levels above 1‰ was also associated with an increased risk of self-inflicted death. This level of alcohol has also been found in suicide victims with a brittle/sensitive personality [44] and high levels of alcohol were also related to suicide among patients with suicidal ideation in a previous study on the present sample [38], both indicating disinhibition of suicidal impulses. Furthermore, there were equal rates of positive blood sampling in poisoning and violent suicides and deaths of undetermined intent and also similar rates of detection of high alcohol levels (above 1‰), regardless of violent method. This agrees with a study by Jones et al. [45] showing no significant difference in detection of alcohol in poisoning and hanging suicides. A possible interpretation is that alcohol is compatible with, or even induces, fatal suicidal acts, regardless of the toxic effect.The association between heroin use and fatal heroin overdoses was expected and has repeatedly been shown in previous literature [5,6]. Prescription opioids were sometimes used in addition to heroin in heroin overdoses. They were also found at autopsy at higher rates in comparison with natural death among non-misusers. When poisoning was compared with violent suicide/death of undetermined intent, prescription opioids were associated with poisoning rather than violent death. These findings may indicate that the toxicity is the main reason of the fatal outcome rather than psychological effects, such as aggression and depression, in agreement with the known effects of opioids. Regular use was not shown to be an increased risk. Dextropropoxiphene was available at the time, which probably increased the risk of self-inflicted death. Reduced suicide rates were shown after withdrawal of this drug from the market [21].Cannabis and CNS-S were rarely found at autopsy, even in previous users, so acute use could not be associated with self-inflicted death among persons who had previously used these substances. In a recent review [46], it was concluded that there is lack of evidence that acute use of cannabis increases the imminent risk of suicidality, while evidence tends to support that chronic acute use can predict suicide. One study [11] concluded that suicide rates have been found to be increased among cannabis users, but they were explained as markers of psychological and behavioural problems rather than overdose. The present study supports this view. One study showed that there was a smaller number of deaths with amphetamines and cocaine compared to opioids, in agreement with the present findings [4]. However, that study dealt with autopsy findings and did not consider risk among previous users.The overlap between depression detected in life and antidepressants found at autopsy was remarkably low, only one case. The medication was also considered to contribute to death in more than half of the cases. Depression in alcohol use disorder is probably undertreated and difficult to treat. Men with depression and alcohol use disorder have been shown to have a very high risk of suicide, 16.2% [47]. Furthermore, in comorbid cases of depression and alcohol use disorders, alcohol turned out to be more common as first diagnosis in the same sample [48]. This means that detection and treatment of depression secondary to alcohol are important in the efforts to prevent suicide.In summary, there was a high risk of self-inflicted death for all the substances, regardless of the acute effect. There was no obvious difference in psychiatric comorbidity or poly-drug use between the substances, apart from fewer diagnoses among cannabis users. However, the latter did not show lower rates of self-inflicted death. The chronic effect on the risk of suicide and death of undetermined intent needs to be considered and investigated. There was a modest overlap between substances used in life and autopsy findings, in agreement with the study on detection of alcohol at autopsy and alcohol use reported in primary care [36]. Another study reported that the same substances used at index admission could be found at autopsy. but did not consider to what extent the substances were found [35]. Substance use as such is a risk factor for suicide, where the use of the substance itself contributes, but the contribution may vary for different substances, as has been clarified in the present study.Gamma-amino butyric acid (GABA-ergic) substances, such as benzodiazepines and alcohol, seemed to facilitate suicidal behaviour on a psychological level, although it may be difficult to distinguish between physical and psychological effects. The fact that alcohol was also common in violent suicide or undetermined death is compatible with a psychological effect as well, as the physical effect did not contribute to death. Opioids, legal or illegal, appear to mainly have a physical effect, as prescription opioids were associated with suicide in casual use and negatively associated with violent suicide or death of undetermined intent.Cannabis may have a very limited toxic effect on suicidal behaviour, and social factors may contribute to the suicide risk. CNS-S do not often facilitate suicidal feelings, but aggression may lead to violent behaviour [49]. The number was too small to show an association with violent suicide in the present sample.The almost non-existent overlap between depression in life and antidepressants at autopsy may be an indication of low detection and undertreatment of depression.To prevent suicide in substance use disorder, treatment of the misuse itself, and poly-drug use, is of major importance. Benzodiazepines should be prescribed with caution to people with substance misuse. Investigation of suicidal ideation during drug use would be of major importance, as the substance may facilitate suicidal behaviour. Reduced availability of more toxic opioids has already been found to reduce suicide rates [21]. Psychosocial factors should not be overlooked. Comorbidity with depression needs investigation, and treatment of depression in alcohol use disorders, although presenting a challenge, is very important. The association between cannabis and schizophrenia [50] is another possible link to suicidal behaviour.The major strength of this study was the combination of clinical data with a long-term follow-up and autopsy findings on a within-person level, an approach that, to the best of our knowledge, has very rarely been used. The design enabled an exploration of the link between the use of substances in life and autopsy findings. The research assistant who evaluated the case records did not know the cause of death, so the assessments were unbiased in terms of knowledge of the manner of death, a problem usually inherent in a retrospective design. The case records have been shown to be of good standard by validation against self-reports [51].In this study, natural death was restricted to those who were autopsied. Some people treated at the Addiction Centre had died at the hospital. Those patients were either not autopsied at the Forensic Department or not at all, so the risk of self-inflicted death is overestimated. However, those who had died at the hospital could not be assumed to be more often positive for the different substances compared to those who had been autopsied, so the higher doses found in self-inflicted death is probably reliable.The substances used are not representative of substances used today. However, the physical and psychological effects are the same, and these effects were the aim of this investigation. Furthermore, the increasing poly-drug use today makes it even more difficult to study the effect of individual substances. We could not control for the effect of use of more than one drug, but the number of drugs was similar. This sample is not representative of users of benzodiazepines and prescription opioids without using other substances, mainly alcohol, as well.Additional psychiatric diagnoses may be underestimated, but this may be equally applicable to the variables compared.Blood-sampling of drugs and not brain autopsy was used, so we have only information on the recent intake of substances before death.This study supports the view that a psychological effect facilitating suicidal behaviour, in addition to a medium toxic effect, contributes to the high risk of self-inflicted death patients who use the GABA-ergic substances benzodiazepine and alcohol. The latter substance was also commonly found in natural death, but in less toxic doses. Legal and illegal opioids, not unexpectedly, appear mainly to have a toxic effect, as accidental overdoses were common and violent death very rare. Use of illegal drugs, such as cannabis and CNS-S, shows a high risk of self-inflicted death from accidental overdoses to suicide, but the chronic risk appears more important than the acute use of the drugs. More knowledge about comorbidity with other diagnoses appears to be an important topic for future research. Detection and treatment of depression in substance use disorders appears to be a challenge.Caution should be exercised in prescribing benzodiazepines and opioids to people with an alcohol problem. Inquiry about suicidal tendencies and also a history of suicide attempts in patients with substance use disorders, also during intoxicated states, may be important in the efforts to prevent suicide. Finally, people with substance use disorders who are not registered at the hospital should be a topic for future research.Conceptualisation: L.B., P.L. and M.B.; Investigation: A.F.; Writing—original draft: L.B.Governmental funding of clinical research within the Swedish NHS (National Health Service) and Ellen and Henrik Sjöbring’s Memorial Foundation supported the study.The study was sponsored by Skåne University Hospital and the Faculty of Medicine, Lund University and Ellen and Henrik Sjöbrings Memorial Foundation. Leslie Walke revised the manuscript.The authors report no conflicting interests.Flow diagram showing the sampling procedure.Autopsy findings by manner of death related to substances misused in life and previous depression; violent death (vd) heroin overdose (ho), undetermined intent (u) suicide (s), no drug. Unnatural and natural death. y-axis, number of cases.Autopsy findings by manner of death among non-misusers and no previous depression; violent death (vd) heroin overdose (ho), undetermined intent (u) suicide (s), no drug. Unnatural and natural death. y-axis: number of cases.Type of misuse related to suicidal behaviour and diagnoses in case records.* Diagnoses cannabis versus other drugs: p < 0.008 (multiple regression).Mean number of drugs misused in life by each drug used in life. Alcohol excluded.Manner of death by substance misused in life in addition to alcohol. (There is an overlap of substances, so the sum does not correspond to individuals.). CNS-S: central nervous system stimulants.Mean number of drugs found at autopsy by drug used in life.

Med-MDPI/ijerph_3/ijerph-16-05-00802.txt

@@ -0,0 +1 @@
+Wetland plants that cover the wetlands play an important role in reducing pollutants. The aim of this study was to investigate the effect of two plant species on microbial communities and nitrogen-removal genes and to evaluate the contributions of absorbing pollutants by Canna indica (CI) and Cyperus alternifolius (CA) to the removal performance in both a vertical subsurface flow constructed wetland and a horizontal subsurface flow constructed wetland, which were part of a full-scale hybrid constructed wetland system. The microbial assemblages were determined using 16S rRNA high-throughput sequencing. Results showed that the presence of CI and CA positively affected microbial abundance and community in general and which was positive for the total bacteria and ammonia nitrogen removal in the CWs. The higher abundance of Nitrospirae appeared in the non-rhizosphere sediment (NRS) than that in the rhizosphere sediment (RS). More denitrification genes were found in NRS than in RS. The copy numbers of narG, nirS and nosZ genes for CA were higher than those for CI. Wetland plant species can significantly (P < 0.05) affect the distribution of microbial communities in RS. Plant selection is important to promote the development of microbial communities with a more active and diverse catabolic capability and the contribution of plant absorption to the overall removal rate of wetland system can be neglected.Constructed wetlands (CWs) have be reported as energy saving, reliable operation and convenient for maintenance in many studies [1,2]. The practical application of the system has been developed from single types to hybrid systems recently. A hybrid constructed wetland (HCW) is composed of two or more types of CWs to take the advantages of individual CWs and complement each other [1].Wetland plants, as an important component of constructed wetlands, have attracted worldwide attention [2]. The root system of wetland plants can promote contaminant binding, more carbon sources and more active redox environment [3]. In turn, it directly or indirectly stimulates soil microbial activity. The roles of wetland plants that covered CWs were thought to be well established [4]. Nevertheless, many wetland plant species usually exhibit different characteristics, such as biomass, root densities and root exudates [3,5]. Previous studies demonstrated that plant offered more suitable conditions via root and root exudates (e.g., the radial oxygen leakage (ROL) and excretion of carbon) for bacterial growth and thereby increased the bacterial population [3,6]. In contrast, plants root and exudates could also play a role in reducing nitrification by an inhibitory root exuded compound [7] or reducing denitrification by increasing oxygen which was less favorable for denitrifiers in wetlands. Therefore, the practical plantation is still premature as we lack the appropriate knowledge of microbial community shift with RS and NRS and species [4,8].At present, many studies on wetlands are based on natural wetlands and laboratory scale [2,4,9]. Few studies had been carried out on plants in practical full-scale engineering systems. Most of previous studies investigated the single removal of nutrients or heavy metals by plants [2,10]. The relations between plant biomass and pollutants concentration in the tissues and their impact factors had not showed consistent results [2,11,12]. Moreover, the allocation for the metals removal in different plant tissues had not been clearly identified.The objectives of the present study were: (i) to investigate effect of two plant species on microbial community, nitrifying, denitrifying and anaerobic ammonia oxidizing (anammox) bacteria in both a vertical subsurface flow constructed wetland (VSFCW) and a horizontal subsurface flow constructed wetland (HSFCW), which were part of a full-scale hybrid constructed wetland (HCW) system; (ii) to evaluate the contributions of absorbing nutrients and heavy metals by Canna indica (CI) and Cyperus alternifolius (CA) to the removal performance in both a VSFCW and a HSFCW.The HCW system was constructed in the northwest of Foshan City, Guangdong Province in Southern China (112°58′16.74″ E, 23°12′37.70″ N) and had operated for six years and two months before sampling. Briefly, the HCW system was composed of a VSFCW, a free water surface constructed wetland (FWSCW) and a HSFCW in series (Figure 1). A detail of the HCW design can be seen in Supplementary Material (SM).The vegetation samples were collected from the VSFCW and HSFCW in June 6, which approached the annual harvesting date (Figure 1). Four CI and four CA were collected which heights represented the average growth of whole plants in the zones. CI samples were numbered V-CI1, V-CI2, H-CI3 and H-CI4, CA samples were numbered V-CA1, V-CA2, H-CA3 and H-CA4 (Figure 1). The water samples were sampled at inlet and outlet of CWs. Both rhizosphere sediment (RS) samples and non-rhizosphere sediment (NRS) samples were collected at depth of 20 cm, about 0.5m horizontal distance between them. A detail of the sampling procedure can be seen in Supplementary Material (SM).The extraction, detection and determination procedures of Genomic DNAs of the sediment samples can be seen in Supplementary Information.The copy numbers of nitrogen transformation genes were determined by the quantitative real-time polymerase chain reaction (PCR) method. For quantification of 16S rRNAs and functional genes, which include total bacteria 16S rRNAs, hydrazine synthase (hzsA), AOB ammonia monooxygenase (amoA), membrane-bound nitrate reductase (narG), dissimilatory cd1-containing nitrite reductase (nirS) and nitrous oxide reductase (nosZ) genes, MyiQ5 (BIO-RAD, Hercules, CA, USA) based on SYBR Green II method was used. The primer pairs and amplification programs are summarized in the Table S1 (in Supplementary Information). The details of PCR mixture and PCR procedure can be seen in Supplementary Information.Primers for sequencing were 515F (5′-GTG CCA GCM GCC GCG GTA A-3′) and 806R (5′-GGA CTA CHV GGG TWT CTA AT-3′), with different barcodes for the V4 region of bacterial and archaeal 16S rRNA gene [13]. The detail of PCR mixture and PCR procedure can be seen in Supplementary Information. A mixture of the amplicons was then used for sequencing on the Illumina MiSeq platform (paired-end 250-bp mode) at the Guangzhou Magigen Biotechnology Co. Ltd. The sequences have been deposited in the NCBI Sequence Read Archive under the accession number SRP186692.Reads with a quality less than 30 at the 3’ end were trimmed. The quality sequences were clustered into operational taxonomic units (OTUs) at the 97% similarity level. Taxonomic assignment was determined at the 80% threshold. The detail of high-throughput sequencing data analysis can be seen in Supplementary Information.Relative abundance (%) of individual taxa within each community was estimated by comparing the number of sequences assigned to a specific taxon versus the number of total sequences obtained for a sample. Calculations of alpha-diversity (including Faith’s phylogenetic diversity (PD), Chao1, observed species, Shannon and Simpson) and beta-diversity (Bray−Curtis distance) metrics were based on a subset of 10,335 randomly selected sequences from each sample. Bray−Curtis-based principal coordinate analysis (PCoA) was used to show the differences among the sediment samples of the two types of CWs. Beta diversity was also evaluated with the bacterial community data to examine the differences of community patterns related to CW structures and sampling locations via the Venn diagram.All the vegetation samples were transported to the laboratory in cold, washed and separated to roots, rhizomes, stem, leaves and flower. To measure dry weight, vegetation samples were dried at 75 °C for 48 h to a constant weight [14,15]. The dry weights were determined, and the dried samples were powered and stored for further analyses.The water samples were used to measure pH, dissolved oxygen (DO), redox potential (Eh), chemical oxygen demand (COD), ammonium nitrogen (NH4-N), nitrate nitrogen (NO3-N), total nitrogen (TN), total phosphorus (TP), Cd, Cu, Ni and Zn. The sediment samples were used to measure pH, Eh, total organic carbon (TOC), NH4-N, NO3-N, TN, TP, Cd, Cu, Ni and Zn. The plant samples were used to measure TN, TP, Cd, Cu, Ni and Zn. The measurements were repeated for three times for each sample. The methods used to measure the physicochemical indices of the samples were described in Supplementary Information.All statistical analyses were implemented using SPSS 18.0 software (SPSS, Inc., Chicago, IL, USA). Kolmogorov-Smirnov test was used to determine whether the data distribution is normal. An independent sample t-test was used to examine the difference between the biomass of same plant species in different types of CWs. Pearson correlation coefficients were calculated to evaluate the relationships between biomass and physicochemical parameters. One-way analysis of variance (ANOVA) was performed to compare differences in physicochemical properties of water among the sampling sites, as well as physicochemical properties of sediment, nutrients concentration, heavy metal concentration and functional genes numbers. Post-hoc tests with Duncan’s statistics at p = 0.05 was performed to analyze the differences between groups of data.High-quality sequences of 397,181 were acquired from the 24 sediment samples, with a range of 10,335 to 24,652 sequences per community (Table S2 in Supplementary Information). In total, 154,009 OTUs were identified, of which 12.5% were detected in only a single sample. Rarefaction analysis indicated that the full extent of microbial diversity was generally high in the sediment samples of the CWs (Figure S1 in Supplementary Information).In sediment communities, Bacteria was dominated (>94.3%) with a low proportion of Archaea (<5.7%) in the VSFCW and HSFCW. The Relative abundance of Archaea (4.7%) in CA groups was higher than that (4.7%) in CI groups (Figure 2). Among the 22 bacterial phyla identified, Proteobacteria were predominant (24.4–60.6%) across all the sediments, with other abundant phyla, such as Chloroflexi (8.2–23.3%), Acidobacteria (3.1–10.6%), Bacteroidetes (2.0–14.3%) (Figure 2). Remarkably, the majority of archaeal sequences were affiliated with Crenarchaeota, which abundance reached the maximum in the NRS of CA in VSFCW. The abundance of Nitrospirae obviously fluctuated among all samples. The abundance of Nitrospirae in the NRS was higher than that in the RS and its abundance in CA was higher than that in CI. The abundance of Cyanobacteria showed higher in the HSFCW than the VSFCW.At the genus level, members from Nitrospira, Leptolyngbya, Thiobacillus and Geobacter were most frequently detected (Figure 3). Leptolyngbya showed its highest abundance (17 ± 3.0%) in NRS of CI in VSFCW and the abundance in NRS was higher than that in RS of two plant species. Thiobacillus showed the highest abundance (19.8 ± 1.7%) in CA RS in VSFCW and the abundance in RS was higher than that in NRS for two plant species. The relative abundance of Nitrospira in NRS was higher than that in RS. Geobacter showed the opposite trend with Nitrospira in NRS and RS in two plants, the abundance in RS is higher than that in NRS in two plants. Dechloromonas showed higher abundance in V.CI.NRS samples.The richness and diversity indexes No. of OTUs, Faith’s PD and chao 1 ranged from 4109 to 5145, from 269 to 351 and from 11,533 to 16,279, respectively (Table S2 in Supplementary Information). For CI in the HSFCW and CA in VSFCW, the microbial diversity was significantly higher in the RS than that in the NRS (p < 0.005), for CA in the HSFCW and CI in VSFCW, the microbial diversity showed no significant difference in the RS and NRS (Table S2 in Supplementary Information).Results of the Unifrac-based PcoA showed that the samples distributed in different parts of the data space (Figure S2 in Supplementary Information), indicating significant differences in sediment community compositions among the sampling sites. For VSFCW, the bacterial communities in the CI.NRS and the CA.NRS shared 1154 OTUs (20.2%) and 1402 OTUs (24.8%) were shared by two communities in the CI.RS and the CA.RS, respectively (Figure 4). For HSFCW, the bacterial communities in the CI.NRS and CA.NRS shared 1162 OTUs (18.8%) and 1457 OTUs (23.9%) were shared by two communities in the CI.RS and CA.RS, respectively. For two CWs, the OTUs was significantly higher in CA group than that in CI group. The results showed that the microbial diversity of CA group was higher than that of CI group.Results of total bacteria 16S rRNAs, hzsA, AOB amoA, narG, nirS and nosZ genes of 16 sediment samples in the two wetlands are shown in Figure 5. For each plant, the abundance of these genes varied significantly among the eight sites (p < 0.05), except for nirS in the samples of CI. For CI, the copy numbers of total bacterial 16S rRNAs varied among 1.76 × 109~1.70 × 1010 copies g−1 and showed significantly higher number in the RS than the NRS in VSFCW (Figure 5a). Similar differences were also presented for hzsA and AOB amoA. The copy numbers of narG and nosZ showed lower in the RS than the NRS although no significant differences were found in VSFCW and HSFCW. The copy numbers of nirS showed no significant differences between the RS and the NRS in VSFCW, showed irregular in HSFCW. For CA, the copy numbers of the total bacterial 16S rRNAs varied among 4.68 × 109~2.51 × 1010, showed higher numbers in the RS than the NRS (Figure 5b). Almost similar differences were also presented for hzsA and AOB amoA. The copy numbers of narG as well as nirS were lower in the RS than the NRS, also significant differences only within HSFCW. The copy numbers of nosZ showed slightly lower in the RS than the NRS. The copy numbers of narG, nirS and nosZ genes for CA were higher than those for CI.Physicochemical properties of wastewater and sediment are shown in Tables S3 and S4 (in Supplementary Information). The concentrations of all pollutant indicators decreased from the inlet to the outlet in each CW except for NO3-N. DO, Eh, COD, NH4-N, TN and SO42− at the inlets of VSFCW were significantly different from those of HSFCW (p < 0.05), except pH, NO3-N and TP (p > 0.05). At the outlets, DO, Eh, COD, NH4-N, NO3-N and TN were significantly different among the two types of CWs (p < 0.05), while pH, TP and SO42− were not significantly different (p > 0.05). At an average sewage flow of 1120 m3 during sampling period, the removal loads of COD, NH4-N, TN and TP in VSFCW were 37.6 g m−2 d−1, 4.22 g m−2 d−1, 3.59 g m−2 d−1 and 0.37 g m−2 d−1, respectively. While those in HSFCW were 3.69 g m−2 d−1, 0.78 g m−2 d−1, 3.28 g m−2 d−1, 0.12 g m−2 d−1, respectively.The biomass (g DW m−2) of two wetland plants grown at the end of 150 days in the VSFCW and HSFCW of the HCW system was shown in Table 1. The biomass of roots, stems and leaves of CI was significantly higher than that of rhizomes and flowers (p < 0.05). The biomass of CI was higher than that of the CA in the VSFCW but did not reach a significant level (p > 0.05). The biomass of CI was significantly higher than that of CA in HSFCW (p < 0.05). The aboveground biomass of CI was significantly higher than that of the underground (p < 0.05) and the aboveground biomass of CA was lower than the underground biomass but was not significantly different (p > 0.05).The concentrations of nutrients as well as heavy metals in CI roots were significantly different among four sites (p < 0.05 or p < 0.001) (Figure 6). Contrary to TP, TN concentrations in the roots of CI were lower than those in the roots of CA. The metals concentrations in the roots of CI were higher than those in roots of CA, except that Cd was almost similar between them. The concentrations of nutrients and heavy metals in rhizomes, leaves and flowers were shown in Figure S3 and Table S5. The concentrations of TN in other plant tissues of two plant species showed similar differences like those in roots. In brief, the TN concentrations in CI were lower than those in CA, whereas the concentrations of heavy metals in CI were higher than those in CA. The concentrations of nutrients were not significantly different between the underground and aboveground tissues of CI, as well as CA (p > 0.05), while four kinds of heavy metals in the underground tissues were significantly higher than the aboveground tissues.Proteobacteria as frequently the dominant phylum in all kinds of wetland, including the genera Zoogloea, Comamonas, Thiobacillus, Nitrosospira, Denitratisoma, Azonexus and Azospira, is usually considered played a vital role in the removal of organic matter and nitrogen [1]. More DO concentration (Table S3) in the influent leaded to more microbial abundances of Proteobacteria and Cleroflexi in VSFCW than in HSFCW (Figure 2), which contributed to the removal of a large number of organic and nitrogen pollutants in VSFCW. The copy numbers of total bacterial 16S rRNAs, hzsA and AOB amoA also showed similar patterns (Figure 5). Rich diversity of the microbial community and the high proportion of nitrogen-cycle bacteria were essential for nitrogen removal in this treatment system [16]. The higher relative abundance of Chloroflexi in all sites indicated the rapid growth of filamentous bacteria and their potential role in carbon cycling during litter decomposition and provided available carbon source to the heterotrophic bacteria [17]. The higher abundance of Nitrospirae which performed nitrite oxidation in the second step of nitrification in the NRS than that in the RS may be due to the inhibition of nitrifying bacteria by root exudates [7,8,18]. In addition, as plants need less energy to absorb ammonia nitrogen than nitrate nitrogen [19], relatively low concentration of ammonia nitrogen and relatively high concentration of nitrate nitrogen in rhizosphere may inhibit the metabolism of nitrifying bacteria. The higher abundance of Nitrospirae in CA than that in CI showed that CA has better nitrification ability. The microbial genera, such as Geobacter, Methanosarcina, Nitrospira and Thauer showed obviously different relative abundances between VSFCW and HSFCW in this study (Figure 3). Due to the different flow direction and oxygen carrying capacity of the influent for two CWs, the phenotype of microbial community was affected by different redox potentials in substrates, more aerobic bacteria, such as Nitrospira were found in VSFCW, while more anaerobic bacteria, such as Methanosarcina were found in HSFCW. More information about how wetland structure affected microbial phenotypes in the HCW systems studied can be seen in our previous studies [1].Wetland plants could provide a large surface area for attached microbial growth and supply reduced carbon and oxygen in the rhizosphere [1]. Most previous studies reported plant presence positively affected microbial abundance and community [20,21]. Plant productivity and below ground biomass impacted anaerobic microbial metabolism in a potted plant study [9]. Nevertheless, the root exudates may have a direct impact on nitrogen cycling, as they may inhibit nitrification process by soil nitrifying microorganisms (Figure 2 and Figure 3) [8]. The root exudates of CA group had more strong inhibition compare to CI group. The denitrifying bacteria, Dechloromonas [22], showed higher abundance in V.CI.NRS samples (Figure 3), which indicates that some denitrifying bacteria may be inhibited by the ROL of root system. A few studies have found no significant difference in microbial communities between root and non-root systems [23,24].Wetland plant species may significantly affect the distribution of microbial communities in RS. The CW systems planted with Vetiveria zizanioides or Juncus effusus L. showed much higher bacterial abundance but lower archaeal abundance [25]. There were significantly more bacteria on P. australis roots when compared to the roots of Phalaries arundinacea [26]. Denitrification strongly depends on both the presence of emergent plants and the denitrifier communities selected by different plant species [27]. Nevertheless, other studies showed some plants appear to exert little effect on the structure of microbial communities in constructed wetlands [28,29,30]. Those results indicates that plant selection is important to promote the development of microbial communities with a more active and diverse catabolic capability [31]. Higher microbial richness and diversity, as well as a higher abundance of bacteria, archaea, anaerobic ammonium oxidation (Anammox) bacteria and key genes (amoA, nosZ, nirS and narG) involved in nitrogen metabolism could affect the degradation of organic compounds and conversion of nitrogen in CWs [32,33].For CI in the HSFCW and CA in VSFCW in our study, the microbial diversity was significantly higher in the RS than that in the NRS (p < 0.005). This finding was similar to previous reports in which the abundances of bacteria and total cells showed significantly higher values in the rhizosphere [17,34]. Our results showed that the microbial diversity of RS was similar that of NRS in two plants. Another study [35] also found significant differences in the microbial community but not in the microbial abundance among different plant species in constructed wetlands in summer. The low percentages of shared OTUs (Figure 4) indicated core microbes were obviously affected by two different plant roots, resulting in significant differences. Our results showed that the microbial diversity of CA was greater than that in CI, this is related to the larger underground biomass and denser roots of CA than those of CI.The copy numbers of the total bacterial 16S rRNAs and functional genes were reported to vary considerably with surroundings, sampling location and depth [36]. In the present study, their copy numbers were among the range of values that had been reported [36,37]. The functional genes of hzsA and AOB amoA in the sediment were higher in the RS compared with NRS in general, indicating that vegetation was positive for the ammonia nitrogen removal in the CWs. The primary influence of plant presence was believed to be related to ROL and its effect on rhizosphere redox [18] and the release of carbon compounds by plants [4]. Even if facultative anaerobic microbes (e.g., anammox bacteria harboring hzsA gene) prefer to gather in root surface for micro-aerobic environment, Eh value of 0 was recommended for the single-stage partial nitritation/anammox process in the previous study [38]. More denitrifiers (harboring narG, nirS and nosZ genes) found in NRS than in RS might be due to the inhibition by ROL due to the susceptibility of denitrifying bacterial genes to oxygen in the denitrification pathway [39]. The copy numbers of narG, nirS and nosZ genes for CA were higher than those for CI, indicated CA had better denitrification ability and CI was more affected by ROL than CA. The total bacterial 16S rRNAs and AOB amoA had higher copy numbers of in the RS of CA than that of CI, which suggested more organic matter decomposers and ammonia oxidizers in the RS of CA. This may attribute to the significantly higher underground biomass and denser roots of CA than those of CI, which promoted the aerobes reproduction.Our results showed that two species of plants had implications in multiple steps of the nitrogen cycle and could significantly (p < 0.05) alter the nitrogen removal microbes in CWs. Although CI and CA presented some differences in the effect on the functional genes, they had similar and positive effects on the degradation of organic compounds and ammonia oxidation process in general. However, the root exudates or certain plant species may have a direct impact on carbon and nitrogen cycling and may inhibit nitrification process by nitrifying microorganisms [8]. Some studies found wetland plants had negative or little influence on bacterial population or functional potential [7,18], in contrast to other studies [6,40]. Therefore, the biological inhibition for nitrogen removal could depend on different plant species and root exuded compound.CI and CA are known for phytoremediation, their biomass affects the pollutant uptake capacity and the total removal capacity of CWs. The total root biomass can significantly influence the removal of ammonia and total dissolved phosphorus via processes such as plant uptake and nitrification [2]. The differences of nutrients in water and sediment were probably the main reason for the difference in the biomass of the same plant in different types of constructed wetlands (Tables S3 and S4). The uptake of NH4-N by plants required less energy than NO3-N and most of the large aquatic plants showed adaptation to use NH4-N as an inorganic nitrogen source [19]. This also can be proved by that NH4-N concentrations showed better correlations with the biomass than NO3-N (Table S6). In this study, the mixed supply of NO3-N and NH4-N could promote more growth of plants and produce higher absorb contributions more than a single supply of NH4-N or NO3-N [41,42]. No obvious harm on the growth of the two plant species was found as their biomass were comparable to those reported [2]. The allocation of nutrient and heavy metal in plant tissues in this study were similar to the previous reports [10,12]. In this study, we found that both of plants had stronger ability to accumulate Cu and Zn than other metals, which may be due to the uptake preferences to heavy metal [43]. CA had a good cumulative capacity for Cu and Zn and reached 75% and 6.72% of those in substrate [44], respectively. Our results confirmed that CI and CA tend to accumulate more heavy metals in roots than other tissues.The contribution of plant absorption to pollutant removal can be estimated by comparing the pollutant removal load of the two treatment systems and the amount of pollutant absorbed by plants. The amount of pollutants absorbed by plants is proportional to the biomass and the concentration of pollutants in tissues. The difference of TN in different plant species was mainly related to plant species. In addition, it was positively related to the concentration of nutrient elements (e.g., TN) and biological abundance around root system. Microbes, such as ammonia oxidizing bacteria (AOB) and archaea (AOA) abundance were regulated by the levels of nitrogen, phosphorus and organic carbon in CWs [45]. While the results (Figure 5) also confirmed the copy numbers of the total bacterial 16S rRNAs, narG, nirS and nosZ genes for CA were higher than those for CI. As the front part of the whole HCW system, VSFCW removed more pollutants and the removal rates of organic pollutants and ammonia nitrogen were close to 60%. While HSFCW as the back part of the whole system, the removal rates of organic pollutants and ammonia nitrogen were 35% and 52%, respectively. The removal rate of nitrate nitrogen in HSFCW was higher than that in VSFCW. The pollutant removal loads of COD, ammonia nitrogen and TN in the whole wetland system were comparable with those of reported system [46], which has a hydraulic residence time of 10 days. Good combination of different wetland structures promoted the treatment performance. Obviously, VSFCW located in the front of the system undertook more pollutant removal than HSFCW. Based on the 150-day growth results in chapters 3.4 and 3.5, it can be seen that the contribution of plant absorption to the overall removal rate of wetland system can be neglected.Wetland plants as the important component of constructed wetlands play a role in pollutants removal by affecting microbial communities rather than by absorbing pollutants. Results showed the presence of CI and CA positively affected microbial abundance and community in general and which was positive for the total bacteria and ammonia nitrogen removal in the CWs. Their root exudates may have a direct impact on nitrogen cycling, as they may inhibit nitrification process by soil nitrifying microorganisms. The root exudates of CA group had more strong inhibition compare to CI group. Some denitrifying bacteria may be inhibited by the ROL of root system. CA had better denitrification ability and CI was more affected by ROL than CA. Wetland plant species can significantly affect the distribution of microbial communities in RS. Plant selection is important to promote the development of microbial communities with a more active and diverse catabolic capability and the contribution of plant absorption to the overall removal rate of wetland system can be neglected. In the application of CWs, CI and CA should be planted at a more suitable density, which should be intensified when planted in VSFCW with main aim to remove organic matter and ammonia nitrogen. While in HSFCW with denitrification as its main function, planting should be reduced to meet only the necessary available carbon sources, especially for CI.The following are available online at https://www.mdpi.com/1660-4601/16/5/802/s1, Figure S1. α-Diversity comparison; Figure S2. The analysis chart of PCoA (Weighted Unifrac) for samples; Figure S3. Nutrients concentration, heavy metal concentration in rhizomes of the two plants at the eight sites in the HCW system. Table S1. Quantitative PCR primers and thermal cycling programs in this study; Table S2. Summary of 16S RNA Miseq sequences, operation taxonomic units (OTUs), and microbial diversity of sediment samples; Table S3. Averaged water quality monitoring data of a VSFCW and a HSFCW belonging to the HCW system; Table S4. Physicochemical properties of the rhizosphere sediment (RS) and non-rhizosphere sediment (NRS); Table S5. Nutrients concentration and heavy metal concentration in the leave and flower of two plants at the eight sites; Table S6. Pearson correlation coefficients between the plants biomass and physicochemical parameters of RS.Methodology, T.H.; Validation, Y.W., J.Y.; Formal Analysis, Y.W.; Investigation, C.C., X.F. and D.W.; Data Curation, Y.W.; Writing–Original Draft Preparation, Y.W.; Writing–Review & Editing, Y.W.; Supervision, R.Z. and R.H.This study was Funded by the Research Fund Program of Guangdong Provincial Key Laboratory of Environmental Pollution Control and Remediation Technology (2018K09), the Science and Technology Planning Project of Guangdong Province, China (2017A020216004, 2017A020216003) and the National Natural Science Foundation of China (No. 31702391).The authors thank the reviewers for their useful comments.The authors declare no conflict of interest.Geographic location, layout, photos and sampling locations of the study hybrid constructed wetland (HCW) system. V = VSFCW = vertical subsurface flow constructed wetland; FWSCW = free water surface constructed wetland; H = HSFCW = horizontal subsurface flow constructed wetland. Arrows show water flow direction in the system. Each circle represents the sampling locations of Canna indica (CI). Each star represents the sampling locations of Cyperus alternifolius (CA).Relative abundance (%) of dominant microbial taxa across all analyzed sediments revealed by 16S rRNA MiSeq sequencing, at phylum level, mean relative abundance > 1%. In the sample number: V = VSFCW; H = HSFCW; CI = Canna indica; CA = Cyperus alternifolius; RS = rhizosphere; NRS = non-rhizosphere; three duplicates in each zone were numbered with 1, 2 and 3.Relative abundance (%) of top 30 microbial genera across the two wetlands revealed by 16S rRNA MiSeq sequencing. The samples numbers were shown in Figure 2.Venn diagram showing the unique and shared OTUs (0.03 phylogenetic distance). Venn diagram for four sampling zones in VSFCW (a) and four sampling zones HSFCW (b). The calculation of the Venn diagram was based on results of three duplicates in each zone. The samples numbers were shown in Figure 2.Quantitative analysis of microbial 16S rRNA and functional genes in the RS and NRS in VSFCW and HSFCW. (a) Canna indica (CI); (b) Cyperus alternifolius (CA). Error bars represent standard deviation calculated from three independent experiments. The samples numbers were shown in Figure 2.Nutrients concentration (mg g−1 DW), heavy metal concentration (μg g−1 DW) in roots of the two plants at the eight sites in the HCW system. (a) total nitrogen (TN); (b) total phosphorus (TP); (c) Cd; (d) Cu; (e) Ni; (f) Zn. The samples numbers were shown in Figure 1.Biomass (g DW m−2) of two wetland plants grown at the end of 150 days in the VSFCW and HSFCW belong to the HCW system.Notes: DW represents dry weight; the samples numbers were shown in Figure 1.

Med-MDPI/ijerph_3/ijerph-16-05-00803.txt

@@ -0,0 +1,44 @@
+In this work, a two-strain dengue model with vertical transmission in the mosquito population is considered. Although vertical transmission is often ignored in models of dengue fever, we show that effective control of an outbreak of dengue can depend on whether or not the vertical transmission is a significant mode of disease transmission. We model the effect of a control strategy aimed at reducing human-mosquito transmissions in an optimal control framework. As the likelihood of vertical transmission increases, outbreaks become more difficult and expensive to control. However, even for low levels of vertical transmission, the additional, uncontrolled, transmission from infected mosquito to eggs may undercut the effectiveness of any control function. This is of particular importance in regions where existing control policies may be effective and the endemic strain does not exhibit vertical transmission. If a novel strain that does exhibit vertical transmission invades, then existing, formerly effective, control policies may no longer be sufficient. Therefore, public health officials should pay more attention to the role of vertical transmission for more effective interventions and policy.Dengue fever is one of the most important re-emerging vector-borne diseases. The primary vector, Aedes aegypti has endured several attempted eradication campaigns, but both the vector and the virus have revealed themselves to be extremely resilient to control measures. Due to rapid urbanization, global travel, and environmental change, public health officials in the world face enormous future challenges from emerging or re-emerging infectious diseases [1]. Over the next 20 years with the largest share of the international growth coming from the Asia and Latin America, regions where dengue is endemic, mass transportation is indeed an important factor in the long-range dispersal of dengue [2,3,4,5]. Dengue puts 40% of the global population at risk with 50 to 100 million infections per year [1]. Despite intensive vector control programs, many countries have experienced dengue re-emergence over the last few decades [1,6].There are only two diseases that have been successfully eradicated: smallpox in 1979 and just recently rinderpest has been declared eradicated by the UN, due in large part to an effective vaccine and aggressive vaccination program [7]. Although vaccines exist for many other diseases, cost and even public perception can limit vaccine coverage, hamper the establishment of herd immunity, and preclude disease eradication [8,9,10]. In 2015, the first dengue vaccine was used in Mexico, however, the effectiveness of the dengue vaccine is still under investigation [11], thus the mitigation and prevention policies have focused on breeding site reduction (elimination of mosquito breeding sites) and spraying programs; that is, they have focused on controlling the vector [6]. The primary drivers of species extinction are habitat disturbance and direct elimination (harvesting, hunting, etc.) [12,13]. Botanical extracts have been pursued as an alternative means of vector control, but their effectiveness (as part of a dengue control strategy) has yet to be ascertained [14,15]. More recent work explored the impact of modern countermeasures such as the Sterile Insect Technique (SIT), the Release of Insects carrying Dominant Lethal genes (RIDL) and the release of Wolbachia-infected mosquitoes [16]. However, Ae. aegypti has demonstrated an affinity to the urban landscape and ability to thrive even in countries with strict control programs [3,17,18].There are five distinct serotypes in dengue virus: DENV-1, DENV-2, DENV-3, DENV-4, and DENV-5 [1]. The disease symptoms range from asymptomatic, mild dengue fever (DF), to severe stages such as dengue hemorrhagic fever (DHF), and dengue shock syndrome (DSS) [19,20]. Dengue virus is mainly transmitted to humans through the bite of infected female mosquitoes of the Aedes species; this process is called horizontal transmission. Also, the mosquito becomes infected when it bites an infected human. However, there exists the possibility of vertical transmission of DENV from the infected female to her eggs (note that vertical transmission occurs in the mosquito population only not in the human population). Recent studies have shown clear evidence of vertical transmission of dengue in the mosquito population for Ae. aegypti and Ae. albopictus [21,22,23]. Moreover, other findings have explored that vertical transmission involving Ae. aegypti and Ae. albopictus species is feasible in captivity and in nature [24,25,26,27,28,29]. Vertical transmission provides a possible mechanism supporting virus dengue persistence in the absence of a recognized host and/or under unfavorable conditions for mosquito activity [30]. A literature review is performed on the presence of natural vertical transmission of DENV in Ae. aegypti and Ae. albopictus [31].Mathematical modeling of vector-borne diseases has evolved from simpler models [32] to more complex models that include climate changes, socio-economic changes and urbanization [33]. Geographic heterogeneity and climate change are some of the key factors for recurrent vector-borne diseases in many tropical/subtropical countries [34]. Systematic reviews on mathematical and statistical models have been performed for the transmission dynamics of dengue [35,36]. Particularly, mathematical analyses of the role of vertical transmission were carried out in [22,37,38]. Their results confirmed the idea that vertical transmission can be an essential mechanism that favored the maintenance of the virus even with low human densities [37]. On the other hand, a substantial proportion of vertical transmission (when vertical transmission is over 20%) could enhance the persistence of the dynamics of dengue disease and otherwise, the role of vertical transmission was negligible [22].Our work is motivated by the 2000–2001 dengue outbreak in Peru, where two strains of DENV-2 are co-circulating (American and Asian of the DENV-2 serotype), and particularly with vertical transmission in the Asian strain of DENV-2 [38]. An invading strain of dengue virus (DENV-2) from Asia rapidly circulated into Peru eventually displacing DENV-2 American. Some fields studies have demonstrated the percentage of natural vertical transmission of DENV from the female to her eggs by analyzing the presence of DENV in terms of the minimum infection rates (MIR) [39], suggesting that Aedes species display different susceptibilities to dengue virus infections. Laboratory experiments also have supported the hypothesis that higher infection rates exist when Ae. aegypti is exposed to the DENV-2 Asian strain in comparison to DENV-2 American [40]. The biological mechanisms behind the displacement of DENV-2 American by DENV-2 Asian at the population level was carried out in the previous work ([38] and references therein). Their results highlighted the importance of vertical transmission, observing that lower transmission rates of DENV-2 Asian are sufficient for displacing DENV-2 American in the presence of vertical transmission.We suggest that vertical transmission, an often overlooked transmission pathway for dengue fever, may contribute to the difficulty of controlling the disease. While we have mentioned some of the numerous political and ecological reasons for the failure of previous eradication campaigns [9], this paper aims to elucidate some implications of vertical transmission on an attempt to control an outbreak of dengue fever. In the present work, we formulate an optimal control problem to identify optimal control strategies for a two-strain dengue model with vertical transmission in the mosquito population. Because vertical transmission is often considered to not be a major factor in dengue transmission, we model the effect of a control measure that does not directly impact vertical transmission. We then compare situations where vertical transmission is and is not a significant mode of dengue transmission. In the next section, we will develop the system with control and develop conditions for the existence of optimal control. Then we present some numerical results and finally we discuss the implications of optimal control in the two-strain model with vertical transmission.A two-strain dengue model was developed to assess the dynamics of two-strain competition, motivated by the context of the 2000–2001 dengue outbreak in Peru [38]. Previous to 2000 only DENV-1 and DENV-2 American genotypes had co-circulated in Peru with neither DHF nor DSS cases reported [41,42]. The absence of DHF and DSS in Peru prior to 2000, in the presence of co-circulating DENV-1 and DENV-2 American, had been explained, using the data of experiments carried out in laboratories. These studies identified partial cross-immunity conferred by DENV-1 against DENV-2 American but not conferred against the 2000–2001 invading DENV-2 Asian strain [41]. Currently, at least four dengue serotypes are found in Peru: DENV-1, DENV-2 (American and Asian), DENV-3, and DENV-4 serotypes. The displacement of DENV-2 American by the DENV-2 Asian has also been associated with the appearance of DHF in the Americas [43]. This shows biological evidence supporting the greater virulent strength of DENV-2 Asian.Although vertical transmission has been mostly understudied in models of dengue, recent results [38] have demonstrated that vertical (transovarial) transmission has both primary and secondary effects in facilitating the invasion and persistence of novel strains of dengue. Dengue management policies exist virtually everywhere dengue fever is a major health concern, yet the fact that dengue outbreaks are increasing in severity and frequency suggests we need to better understand control strategies and how to evaluate them [44]. Among these features is vertical transmission which will be explored by considering a population that is impacted by two variants of the same serotype of dengue simultaneously: one that exhibits vertical transmission as a significant mode of disease transmission (DENV-2 Asian) and one that does not (DENV-2 American). In the present work, we extend the previous model [38] by incorporating a time-dependent control function.We use a compartmental modeling framework where each compartment, shown in Figure 1 and Figure 2 by a letter within a box, denotes a class of individuals. Then the arrows represent the flows of individuals between different states [45]. Let S represent the number of susceptible hosts (humans). These individuals are antigenically naive to the particular strain of dengue being modeled but may have had previous exposures to other strains. DAm and DAs are individuals infected with genotypes of dengue 2, DENV-2 Asian and DENV-2 American, respectively. H represents individuals who have developed DHF, R and is recovered individuals. N is the total human population size that is assumed constant since the change in population size is insignificant for a short time period. V is the class of susceptible vectors (female mosquitoes). WAm and WAs are mosquitoes that carry strain DENV-2 American and DENV-2 Asian, respectively. M is the total vector population size, is assumed constant which is biologically reasonable within a short time scale as well. Then we can write the system of equations representing our model as:
+	  (1)S˙=μN−βAm(1−u(t))SWAmM−βAs(1−u(t))SWAsM−μSD˙Am=βAm(1−u(t))SWAmM−(δ+μ)DAmD˙As=βAs(1−u(t))SWAsM−(δ+α+μ)DAsH˙=αDAs−(δ+μ)HR˙=δDAm+δDAs+δH−μRV˙=μmM−pμmWAs−θAm(1−u(t))VDAmN−θAs(1−u(t))VDAsN−μmVW˙Am=θAm(1−u(t))VDAmN−μmWAmW˙As=θAs(1−u(t))VDAsN+pμmWAs−μmWAs
+	  Note that N˙=0 and M˙=0 when we add all the equations in the above system (1). In this work, a control function (u(t)) is modeled by preventive control efforts: preventive control efforts may involve the application of a pesticide (sprays), reduction of vector breeding grounds, mosquito repellents, or the results of education campaigns, which increase personal protection. It is assumed that these preventive interventions do not reduce the total vector population significantly, and the effect of these interventions implicitly translates in reductions of transmission between vectors and hosts per unit time.Therefore, u(t) is the percentage reduction in infection due to the effect of control measures. Then βi(1−u) is the effective transmission force for strain i. Note we assume no a priori knowledge of what strain a particular individual has, thus the control measure is independent of the strain. Furthermore, since we are primarily interested in modeling the effect of a control measure, we assume that the reduction in effective contacts impacts mosquitoes equally well as humans. Thus, their effective force of infection is also reduced by u(t). We also assume that one strain, DENV-2 Asian, is more virulent, leading to cases of DHF and also exhibiting vertical transmission with some probability p times the basic fecundity function while the other strain, DENV-2 American, does not.When the control function u(t)≡0, system (2) is said to be autonomous. The basic reproductive number of an epidemiological model generally determines whether or not the disease will die out or persist [45]. For the autonomous system, if we consider each strain independently, then the reproductive number for DENV-2 American is R0Am=βAm(δ+μ)θAmμm, and the reproductive number for DENV-2 Asian is R0As=p2+p22+βAs(δ+α+μ)θAsμm. Then the basic reproductive number is R0=max[R0Am,R0As] (more detailed computations are found in [38]). The basic reproductive number is a central component of the model that can distinguish between different qualitative behavior in the autonomous system.A central component of the control problem is the optimization, in this case, minimization, of an objective function. We are interested in controlling an outbreak of dengue, thus we want to minimize the number of infected humans and the cost of implementing control efforts as well. However, we are also interested in preventing future outbreaks, thus we want to minimize the number of infected mosquitoes and individuals both during the course of our control measure and when our control policy has ended at time t=T. Then the corresponding objective function is:
+	  (2)J(u(t))=∫0Tw1(DAm(t)+DAs(t))+w2(WAm(t)+WAs(t))+12w3u2(t)dt+w4(DAm(T)+DAs(T))+w5(WAm(T)+WAs(T))
+	  where w1 is the weight constant for host infections, w2 is the weight constant for vector infections. 12w3u2(t) is the cost of control with the weight constant w3 and included as a quadratic term for the existence of optimal control due to the convexity of a control function in the objective function. Lastly, w4 and w5 are the weight constant for the payoff term (at the final time, t=T). If we let X be the vector of our state variables which is restricted to the positive orthant, X∈R+8, then X* is the optimal solution that corresponds to the optimal control function u* such that
+      J(u*)=min{J(u)|u∈Ω},
+      where Ω={(u(t)∈L1|0≤u(t)≤1,t∈[0,T]}. Then the Hamiltonian of our system is
+      (3)H^(X,u)=w1(DAm(t)+DAs(t))+w2(WAm(t)+WAs(t))+12w3u2(t)+λ1μN−βAm(1−u(t))SWAmM−βAs(1−u(t))SWAsM−μS+λ2βAm(1−u(t))SWAmM−(δ+μ)DAm+λ3βAs(1−u(t))SWAsM−(δ+α+μ)DAs+λ4αDAs−(δ+μ)H+λ5δDAm+δDAs+δH−μR+λ6μmM−pμmWAs−θAm(1−u(t))VDAmN−θAs(1−u(t))VDAsN−μmV+λ7θAm(1−u(t))VDAmN−μmWAm
+      (4)+λ8θAs(1−u(t))VDAsN+pμmWAs−μmWAs,
+      where λi are the co-state or adjoint variables [46]. Then, by Pontryagin’s Maximum Principle [47], our optimal solution can be found by simultaneously solving the adjoint system:
+	  dλ1(t)dt=−∂H^∂S=(λ1−λ2)βAm(1−u)WAmM+(λ1−λ3)βAs(1−u)WAsM+λ1μdλ2(t)dt=−∂H^∂DAm=(λ6−λ7)θAm(1−u)VN+(λ2−λ5)δ+λ2μ−w1dλ3(t)dt=−∂H^∂DAs=(λ6−λ8)θAs(1−u)VN+(λ3−λ5)δ+(λ3−λ4)α+λ3μ−w1dλ4(t)dt=−∂H^∂H=(λ4−λ5)δ+λ4μdλ5(t)dt=−∂H^∂R=λ5μdλ6(t)dt=−∂H^∂V=(λ6−λ7)θAm(1−u)DAmN+(λ6−λ8)θAs(1−u)DAsN+λ6μmdλ7(t)dt=−∂H^∂WAm=(λ1−λ2)βAm(1−u)SN+λ7μm−w2dλ8(t)dt=−∂H^∂WAs=(λ1−λ3)βAs(1−u)SN+(λ6−λ8)pμm+λ8μm−w2,
+	  with the transversality conditions at t=T
+      λ1=λ4=λ5=λ6=0λ2=λ3=w1λ7=λ8=w2,
+      and the optimality condition
+      (5)∂H^∂u=w3u+(λ1−λ2)βAmSWAmM+(λ1−λ3)βAsSWAsM+(λ6−λ7)θAmVDAmN+(λ6−λ8)θAsVDAsN,
+      where ∂H^∂u=0 at u=u*. We can solve this for the optimal control function u* with the constraint that u must be between 0 and 1 to get
+      u*=minmax0,(λ2−λ1)βAmSWAmw3M+(λ3−λ1)βAsSWAsw3M+(λ7−λ6)θAmVDAmw3N+(λ8−λ6)θAsVDAsw3N,1.
+	  This type of optimal control formulation has several applications in mathematical biology [46,48,49,50,51]. Although proof of the existence of optimal control is left to the Appendix A, the solution to our control problem will be a piecewise smooth control function. For the purposes of this article, what is important is the qualitative shape of this control function. Because it is unclear what the costs of these control policies are relative to the effective reduction in transmission, more insight may be gleaned by examining the qualitative features of the control function as the relative costs are changed.Each numerical solution is performed over a period of three years to give account for transient dynamics. In reality, a control policy would also be evaluated over short, medium and long term time periods, and three years seemed sufficient for our numerical results. The default parameters for all simulations are listed in Table 1 unless otherwise indicated.First, we investigate the impact of different values of the relative cost of control (w3) on the controlled dengue dynamics. The weight constant can be considered as the relative cost of control implementation, and a larger value represents a relatively higher cost. Figure 3 and Figure 4 illustrate the impact of control weight constants under several values of w3=1,0.5,0.05. As mentioned in the previous section, there are two cases of the basic reproduction number: either one of the two strains is dominant (R0=max[R0Am,R0As]). Overall, the impact is straightforward; for higher costs, the control decreases, which leads to larger outbreaks.If the cost of the control function is comparable, on the same order of magnitude (w3=1), to the costs incurred from the disease, then there is no incentive to invest heavily on control. We see this in Figure 3 where not much effort is spent on the control function. However, if the control becomes less expensive, or analogously the costs from disease become more expensive, then it is worthwhile to invest in eliminating the disease and preventing an outbreak. Note that with sufficient effort the control function can mitigate the current outbreak and prevent future ones (the damped oscillations predicted in the autonomous model) as seen in Figure 3. This is the case where the basic reproductive number, R0, is greater than one (1.4), otherwise there would be no outbreak and control would be moot.However, the left panels in Figure 3 and Figure 4 are when R0As<R0Am, i.e., the strain without vertical transmission is the dominant strain during an outbreak. If we keep the same basic reproductive number but instead chose the outbreak to be dominated by the strain with vertical transmission, R0As>R0Am, then we get the scenarios depicted in the right panels of Figure 3 and Figure 4. Here we see that when the cost of control is comparable to the cost of the disease, we get the same results as before. When the cost of control is too high, we cannot completely control the outbreak and we must respond to rises in prevalence, Figure 4. However, at the same level of relative costs where the outbreak was controlled before, here we are unable to fully control the outbreak. The total number of cases is larger and there is a small secondary outbreak. In order to fully control the outbreak, we have to reduce the relative costs even further than in the previous case. Vertical transmission (p the proportion of eggs hatched infected with dengue) made the outbreak more difficult to control because the control function did not prevent the development of newly infected mosquitoes from infected eggs.In the previous results (Figure 3 and Figure 4), the level of vertical transmission was relatively low (p = 0.0103). Now, we investigate the impact of vertical transmission on the controlled dengue dynamics (four different values of p=0.01,0.1,0.5,0.7 are used). Figure 5 and Figure 6 display the total proportion of infected and optimal controls under four different values of p using the same value of w3=1 and the same value of R0=1.4. Note that in order to keep the same value of R0, β1 and β2 are varied as well. Again, we present two cases of the basic reproduction number: either one of the two strains is dominant (R0=max[R0Am,R0As]), which are displayed in Figure 5 and Figure 6.In Figure 5 (when the American strain is dominant), the impact is straightforward; for a higher vertical transmission rate, the control increases (b), which leads to smaller outbreaks (a). Interestingly, Figure 6 (when the Asian strain is dominant) shows a counterintuitive effect of p; for a higher vertical transmission rate, even though the control increases (b), it becomes harder to control the outbreak (a). This confirms that, in particular, when the dominant strain is Asian, a higher vertical transmission rate increases the difficulty of controlling the outbreak under the same level of R0=1.4.To further see the impact of vertical transmission, we measured the total value of the objective function and the cumulative incidence as functions of p and β2. Figure 7 and Figure 8 illustrate the results under two weight constants (low cost using w3=0.01 and high cost using w3=2). If the relative cost of control is higher, then the total costs are proportionally higher as well. Regardless of the costs of control, having a large force of vertical transmission makes an outbreak extremely expensive to control, Figure 7b. This is due to the fact that the control policy cannot directly stop the generation of infected mosquitoes via vertical transmission, and thus are penalized by the number of new infections those mosquitoes cause, Figure 8b. As seen in both Figure 7 and Figure 8, under the parameter values used here, the impact of p on the objective function value and the cumulative incidence is more significant than the impact of β2. In the low-cost case, the outbreak is manageable except p is very high (higher than 80%). On the other hand, the high-cost case, the outbreak is manageable only when p∈[0,30]% and β2∈[0,0.5].For a fixed force of horizontal transmission, β2, we can see how the total costs of control and the severity of an outbreak vary directly with changes in the force of vertical transmission, p, and relative cost of control, w3. Figure 9 displays the results under two horizontal transmission rates (low using β2=0.05 and high using β2=0.21667). As the horizontal transmission increases (β2), we notice that larger outbreaks occur for smaller values of vertical transmission, Figure 9d. Large values of vertical transmission can cause larger outbreaks with associated larger costs Figure 9b. For the lower horizontal transmission case, the outbreak is manageable except p is very high (over 70%) while the outbreak is manageable only when p∈[0,10]% for the higher horizontal transmission case.We developed an optimal control framework to identify optimal control strategies for a two-strain dengue model with vertical transmission in the mosquito population. Our model is motivated by the 2000–2001 dengue outbreak in Peru, where two strains of DENV-2 are co-circulating, and particularly with vertical transmission in the Asian strain of DENV-2. We evaluate the role of vertical transmission in the controlled dengue dynamics. Our results indicate that the controlled dengue dynamics are strongly dependent on the following three key factors: p, β2, and w3. Overall, controlling the outbreaks is more difficult as vertical transmission (p), β2, and the relative cost increase.Under the moderate level of R0=1.4, the outbreak can be well controlled (when vertical transmission also is moderate). Especially, for the case of unlimited resources available (the relative cost is inexpensive w3=0.01), controlling the outbreak is sufficiently effective even when p is high (p∈[0,70]%). As the relative cost becomes more expensive, controlling the outbreak is effective only in the smaller range of β2 and p. If the cost is high (w3=2), then the outbreak will be extremely expensive and impossible to control (β2>0.05 and p>10%). Regardless of the relative cost of the control, outbreaks are extremely hard to manage when both horizontal and vertical transmission becomes higher. Finally, it is impossible at all as all of them increase (w3>1, β2>0.2 and p>10%). Although this situation is unrealistic, it highlights the importance of a control strategy, whether highly cost efficient or otherwise to take into consideration all possible transmission pathways.Moreover, our findings highlight the importance of vertical transmission in the two-strain dengue dynamics. The two-strain model considers the competing dynamics of these two DENV-2 strains (the resident or the American type and the invasive more virulent Asian strain). The three critical factors mentioned above (p, β2, and w3) play a more significant role when the outbreak is dominated by the invasive DENV-2 Asian strain (i.e., R0As>R0Am). Since data from the 2000–2001 outbreak in Peru showed that DENV-2 Asian had displaced DENV-2 American [42], more careful prevention plans should be implemented when DENV-2 Asian strain is dominant.As the relative cost of the control function is reduced, the proportion of infected people decreases. However, when the outbreak is dominated by the strain without vertical transmission, then the outbreak can be controlled more easily than when the outbreak is dominated by the strain with vertical transmission. In this case, the role of vertical transmission rate becomes negligible. Therefore, the effectiveness of control is strongly sensitive to various factors including the relative cost, dominant strains, the level of horizontal transmission, and vertical transmission. There are also similar results, observing that the role of vertical transmission is sensitive to other various factors [22].Diseases have been and continue to be a major public health challenge, with outbreaks of infectious diseases capable of causing tremendous loss of life in relatively short periods. There are various strategies to controlling an epidemic (including vaccination, isolation, and social distancing) that have been used to study disease prevention/mitigation in various contexts (see [8,9,45,46,49] and references therein). Note that the current model could incorporate different control measures such as vaccination, treatments, chemical insecticide for adult mosquitoes or destruction of breeding sites (i.e., killing immature and aquatic stages). Some of these countermeasures have been implemented and compared in an optimal control framework [54,55,56]. Instead, our control is modeled as the effect of such preventive countermeasures mentioned above. We assumed that these preventive interventions do not reduce the total vector population significantly, and the effect of these interventions implicitly translates in reductions of transmission between vectors and hosts.This implicit approach of control is employed, so the role of vertical transmission is made transparent as possible. However, the current study with such a simple assumption has limitations. For instance, the application of a pesticide (sprays), or reduction of vector breeding grounds will change the vector population size and the life-span of vector as well. Similarly, other control methods (mosquito repellents, or the results of education campaigns) will have different impacts on dengue transmission dynamics. It requires to develop relevant mathematical models, then to drive resulting optimality systems. Therefore, further extensive simulations and analyses should be carried out in future work. Furthermore, as the epidemiological and morbidity burden associated with dengue increase substantially, it becomes more critical to measure estimates of health and economic costs of the disease [5,57,58]. Extensive cost-effectiveness analyses based on real dengue burden should be carried out in future research.We have modified the previous model proposed in [38] by incorporating a time-dependent control function. Our model is motivated by the 2000–2001 dengue outbreak in Peru, where an invading strain of dengue virus (DENV-2) from Asia rapidly circulated into Peru eventually displacing DENV-2 American. As the likelihood of vertical transmission increases, outbreaks become more difficult and expensive to control. This is of particular importance in regions where existing control policies may be effective, and the endemic strain does not exhibit vertical transmission.This paper illuminates some of the implications of a control strategy that ignores the role of vertical transmission. If horizontal transmission is the dominant mode of transmission, and the moderate level of the basic reproductive number combined with the inexpensive cost of control, then the impact of vertical transmission may be negligible, and hence, the dengue outbreak is manageable. However, if any of those conditions are not met, vertical transmission may render a perfectly adequate control policy useless.There is some evidence that genetic changes in either the vector or the virus may facilitate vertical transmission [24,26,28,59]. The unbeknownst proliferation of these genetic mutants can establish an alternative pathway of dengue transmission leading to unexpected outbreaks and perplexing regulators using policies that should be effective. Since the force of vertical transmission can increase both the costs associated with controlling the vector and the burden of dengue cases, public health officials should pay more attention to the role of vertical transmission for more effective interventions and policy.For research articles Conceptualization, D.M., A.M., and S.L.; Methodology, D.M. and S.L.; Visualization, D.M. and S.L.; Writing—original draft, D.M., A.M. and S.L.; Writing—review and editing, A.M. and S.L.This research was funded by the National Research Foundation of Korea (NRF) grant funded by the Korean government (MSIP) (NRF-2018R1A2B6007668).We would like to thank two anonymous reviewers for their careful reading and valuable comments that helped us to improve the paper.The authors declare no conflicts of interest.The optimal control exists under very general constraints. First note that solutions are bounded in the positive orthant: the derivative is non-negative on the zero boundary and dNdt=0, dMdt=0 where N is the sum of the host variables and M is the sum of the vector variables. The optimal control exists if the following conditions are satisfied [60]:
+		The set of controls and corresponding state variables is non-empty.The control set, Ω, is convex and closed.The right hand side of system (2) is bounded by a linear function in the state and control.The integrand of the objective functional is convex and bounded below by c1(|u1|2+|u2|2)β2−c2 and the Lipschitz condition is satisfied.The payoff function is continuous.
+		The set of controls and corresponding state variables is non-empty.The control set, Ω, is convex and closed.The right hand side of system (2) is bounded by a linear function in the state and control.The integrand of the objective functional is convex and bounded below by c1(|u1|2+|u2|2)β2−c2 and the Lipschitz condition is satisfied.The payoff function is continuous.
+          Proof
+          
+            
+              If we consider the vector of state variables x=[S,DAm,DAs,H,R,V,WAm,WAs]T, then we can write our system of equations as
+            x˙=f(x,u).
+			
+              Since we know our state variables are bounded in the positive orthant, the particular form of our system of equations dictates that f(x,u) is bounded. Thus there exists a unique solution to our system given suitable initial conditions.
+            
+            
+              By the construction of Ω, this condition is clearly met.
+            
+            
+              The total population for both the host and vector systems is constant, thus all solutions are bounded. The control function is also bounded, thus the right-hand side can be bounded by a linear function in the state and control.
+            
+            
+              The integrand is linear in the state variable and quadratic in the control function, and thus clearly convex. Furthermore, the Lipschitz is condition is clearly satisfied as the integrand is bounded below since both the state and control are non-negative.
+            
+            
+              The payoff function is clearly continuous by construction.
+            
+          
+        If we consider the vector of state variables x=[S,DAm,DAs,H,R,V,WAm,WAs]T, then we can write our system of equations as
+            x˙=f(x,u).
+			Since we know our state variables are bounded in the positive orthant, the particular form of our system of equations dictates that f(x,u) is bounded. Thus there exists a unique solution to our system given suitable initial conditions.By the construction of Ω, this condition is clearly met.The total population for both the host and vector systems is constant, thus all solutions are bounded. The control function is also bounded, thus the right-hand side can be bounded by a linear function in the state and control.The integrand is linear in the state variable and quadratic in the control function, and thus clearly convex. Furthermore, the Lipschitz is condition is clearly satisfied as the integrand is bounded below since both the state and control are non-negative.The payoff function is clearly continuous by construction.Host model flow diagram: S is the class of susceptible individuals who can become infectious with either DENV-2 American genotype, DAm, or DENV-2 Asian genotype DAs via infectious female mosquitoes W carrying the corresponding strain. In this model, only individuals infected with the Asian genotype can progress to DHF, H, and all infected individuals can recover, R. Note that the control function (1−u(t)) is modeled as the reduction efforts in the transmission rate from S either to DAm, or DAs.Vector model flow diagram: V is the class of susceptible female mosquitoes that can become infected with either DENV-2 American genotype WAm or DENV-2 Asian genotype WAs via contact with an infectious human, D carrying the corresponding genotype. Vertical transmission only occurs in mosquitoes infected with genotype Asian. In this model, there is a constant birth rate, but a proportion, p, of those births by mosquitoes carrying genotype Asian, WAs, enter directly into the infectious class. Note that the control function (1−u(t)) is modeled as the reduction efforts in the transmission rate from V either to WAm, or WAs.As the relative cost of the control function, w3, is reduced, the proportion of infected people decreases. However when the outbreak is dominated by the strain without vertical transmission, (a) then the outbreak can be controlled more easily than when the outbreak is dominated by the strain with vertical transmission, (b) In the latter case, the cost of control must be reduced even further to effectively control the outbreak.As the relative cost of the control function decreases, it is used more frequently and is able to control the outbreak. If the relative cost is expensive, then it is used sparingly and in response to outbreaks. Notice the peaks occur right after an increase in the prevalence of dengue in the corresponding panel of Figure 3.When the dominant strain is DENV-2 American, as the level of vertical transmission increases, the level of optimal control increases (b). Therefore, it is easier to control the outbreak (a).When the dominant strain is DENV-2 Asian, as the level of vertical transmission increases, the level of optimal control increases (b). However, it is harder to control the outbreak (a).Even with an “effective” control program, a high vertical transmission rate can render the health policy moot regardless of the cost of additional control is low (a), or high (b).The outbreak can be well controlled except when vertical transmission is extremely high. Although this situation is unrealistic, it highlights the importance of a control strategy, whether highly cost-efficient, left, or otherwise to take into consideration all possible transmission pathways.Regardless of whether horizontal transmission is low, left panels, or moderate, right panels, a high level of vertical transmission can create extremely large, and costly outbreaks, top panels. If the relative costs of controlling the outbreak are low, w3, then the epidemic can still be controlled, bottom panels. However, if the cost is high, then the outbreak will be extremely expensive and impossible to control.Default Parameter Values: Biological parameters may vary across geographic and temporal scales, however, most of the values are taken from related literature or estimated to achieve the desired reproductive number.

Med-MDPI/ijerph_3/ijerph-16-05-00804.txt

@@ -0,0 +1 @@
+Bedouin society has undergone rapid changes over the past decade. The younger generation of Bedouin women is better educated, which has enabled them to enter different professions, increased their incomes and elevated their social status. We examined the sense of coherence (SOC) and its components of meaningfulness, manageability and comprehensibility as well as the use of coping strategies among Bedouin women from three age groups. We also investigated the coping resources and strategies before determining the relationships between these variables in the three groups. One hundred ninety-six women participated in the study. Differences were found mostly between the oldest age group (61 years and older) and the two younger groups (21–40 and 41–60 years old). The oldest women reported less meaningfulness and used less positive reframing, planning, humor and acceptance. In terms of coping strategies, venting was used more by the youngest group whereas behavioral disengagement was used more by the oldest group. In the younger groups, SOC and its components were positively correlated with the use of coping strategies that are considered to be adaptive and with emotional support. However, the correlations between these factors were negative among the oldest group, which points to non-adaptive coping strategies used by these women. These results are discussed in light of the salutogenic, stress-appraisal and coping theories.The Bedouin of the Negev are a minority group in Israel. They are Muslim Arabs who have inhabited the Negev desert since the fifth century Common Era (CE). Traditionally, they lived in nomadic or semi-nomadic tribes. However, over the past half century, they have experienced a rapid and dramatic transition. In the past decade alone, Bedouin society has undergone tremendous change, including the increased exposure to higher education and more interaction with Israeli Western society. These changes may have affected the coping resources of Bedouin women and their effectiveness in reducing stress [1]. Despite these changes, this society has preserved traditional customs and values, which emphasize the collective over the individual and collectivism remains a central component of Bedouin culture. In Bedouin culture, there is an emphasis on the mutual responsibility of the group for the individual that is accompanied by the cultural norms of respect for people in authority, privacy of feelings, politeness and belief in fate [2,3]. However, in recent years, Bedouin society has experienced significant changes [1]. The modernization and urbanization processes that the Bedouin community has undergone in recent years have disrupted traditional frameworks and led to the fragmentation of the cultural hierarchy. These changes have also highlighted the conflict between traditional Bedouin values and Western Israeli values, which are in direct competition with each other [4]. As a result, some of the society’s collectivist nature has been lost and this is reflected in the changes in traditional frameworks, the loss of traditional authority and the striving of young people for positions of leadership and individuality [5]. These changes have also influenced the women of this society. Bedouin women are now entering the higher-education system at dramatically higher rates. In 2018, the number of Bedouin women enrolled in the institutions of higher education was 50% higher than it was a decade earlier. However, this increased participation in higher education has not yet been completely translated into the workforce and most of these women still work in the education system [6].The current study compared three generations of Bedouin women. Our goal was to compare these women in terms of the coping strategies that they use when faced with various stressful situations and their coping resources. Specifically, we focus of their sense of coherence (SOC) and its components of meaningfulness, manageability and comprehensibility.The theoretical foundation of the present study is the transactional theory of stress and coping developed by Lazarus and Folkman [7]. They defined coping as the actual effort that is made in the attempt to make a perceived stressor more tolerable and minimize the distress induced by the situation. Most models of coping assume that the individuals who cope more effectively with stressful life events show lower levels of psychological problems and subsequently experience a better quality of life [7]. The importance of coping has been stressed in research conducted in recent decades. The studies of coping with stressful situations have shown that the emotionally focused strategies tend to be associated with more psychological problems whereas the use of problem-solving strategies or active coping tends to be linked to better wellbeing [8,9]. In addition to the increased stress that has accompanied the process of change in this society, the stress of aging should also be acknowledged. Getting older is stressful in itself. As individuals age, they experience social loss and general physical weakening, which can threaten their quality of life [10]. Since there are only limited studies focusing on the responses of Bedouin women to different types of stress, it is important to study coping strategies among Bedouin women as those strategies could have important implications for their wellbeing.Several studies conducted in the last decade have pointed to the crosscultural importance of coping strategies. However, it seems that overall, among Arabs from various areas in Israel, similar to those living in other Western cultures, the implementation of problem-solving strategies acts as a protective factor whereas the implementation of emotional coping strategies seems to be maladaptive and results in more psychological problems and poorer wellbeing [11,12,13,14]. More specifically, among Bedouin women, seeking social support and regulating feelings and actions were found to be related to lower levels of wellbeing [15]. An exception for the well-known relationships between emotional coping and mental health was observed for the specific coping strategy of accepting responsibility. Although this strategy is considered to be a form of emotional coping, it was found to be negatively correlated with wellbeing among a group of Bedouin women [16].Antonovsky [17] suggested the conceptualization of salutogenesis (the “origin of health”) in stress research and claimed that this is a universal concept that applies to all cultures. This is a continuum model that suggests that rather than classifying health/illness dichotomously, it is more appropriate to view each individual at any given moment as being somewhere on the ease/dis-ease continuum [18]. The most important concept in this model is SOC, which represents the motivation and the internal and external resources that one can use to cope with stressors. Thus, this SOC plays an important role in the way that one perceives challenges. Sense of coherence is a global orientation of seeing the world as more or less comprehensible (the internal and the external world are perceived as rational, understandable, consistent and expected), manageable (the individual believes that s/he has available resources to deal with situations) and meaningful (the motivation to cope and the commitment to emotionally invest in the coping process) [18].The research conducted around the world, which has mostly been undertaken in Western societies, has found strong relationships between SOC and various health factors, including physical, mental and social functioning. Thus, an individual with a stronger SOC is expected to report better health [19]. In contrast to the strong conclusions regarding SOC and health in Western societies, some recent studies focusing on the responses to stressful events within Bedouin society have yielded different results. Although, SOC has been found to serve as a protective factor and to be negatively correlated with a range of psychological problems in some cases [20], SOC was not found to be related to mental health problems or was found to be positively related to such problems in most cases [21,22]. This complicated picture was presented in a recent study, which demonstrated that certain subgroups of Bedouin society benefit from SOC as a protective factor whereas SOC seems to be a maladaptive resource among other subgroups [23]. These results called for further investigation and the present research will try to fill part of this gap as it explores the relationships of SOC and its components with various coping strategies.Several studies have investigated the relationships between SOC and coping strategies. One study of women with breast cancer in Sweden found that women who reported a strong SOC also reported using a variety of coping strategies. These women also reported using the coping strategies of distraction, situation redefinition, direct action, relaxation and to a lesser extent, religion. Using a variety of strategies enabled these women to choose the proper coping strategy and therefore, resulted in them coping better and reporting better psychological health [24]. A different study found that during a political violent event, both problem-solving and emotional coping strategies mediated the relationship between SOC and psychological problems. Although that study demonstrated a positive relationship between SOC and problem-solving, which was negatively correlated to psychological problems in turn, the relationships with emotional coping were reversed. SOC was negatively correlated with emotional coping, which was positively correlated with psychological problems in turn [25].Following this literature review, we aim to explore several questions relating to SOC and coping strategies among Bedouin women. First, we would like to compare women from three age groups (young, middle-aged and older women) in terms of their SOC and its components of meaningfulness, manageability and comprehensibility, as well as their use of various coping strategies (i.e., self-distraction, active coping, emotional social support, instrumental social support, behavioral disengagement, venting emotions, positive reframing, planning, humor, acceptance and self-blame). Second, we would like to examine the relationships between SOC and the various coping strategies in the three age groups. Third, we aim to determine whether there are any differences in the direction or strength of these relationships among the different age groups.One hundred ninety-six women who were aged 21–83 years old participated in this study. The mean age was 52.80 (SD = 16.82.) As for marital status, 22 women (11.9%) reported that they were single, 115 (62.2%) were a first wife, 40 (21.6%) were a second wife and 8 (4.3%) reported their family status as “other”. Furthermore, 11 women did not report their marital status. The years of formal education varied greatly among the women from 0–18 years, with a mean of 5.76 (SD = 5.77). As for current work, most of the women (62.9%) reported that they were not working at all at the time of the study.The women were divided into three age groups. There were 49 women in the youngest group (21–40 years old), 73 women in the middle-aged group (41–60 years old) and 74 women in the oldest group (61–83 years old). The women in the youngest and middle-aged groups were more educated than the older women (M = 9.57 years, SD = 4.76; M = 8.18 years, SD = 5.56; M = 0.82 years, SD = 1.90, respectively; F = 76.14, p < 0.001). The middle-aged group included the most workers as 52.2% of the youngest group, 58.3% of the middle group and only 5.5% of the oldest group reported that they were currently working. However, both the youngest and the middle-aged groups differed significantly from the oldest group in terms of work (X2 = 49.90, p < 0.001).After receiving ethics approval from the university department’s IRB committee, the participants filled out anonymous self-report questionnaires between February and May 2018. The questionnaires were administered in Arabic, the native tongue of the participants. We used snowball, convenience sampling and adhered to all ethical guidelines. Participation was voluntary and the participants were informed that the researcher was interested in their experiences. Participants were free to withdraw their participation for any reason at any time during the questionnaire procedure.The collected demographic data included age, marital status, years of formal education and working status.This was a 28-item tool that measures coping strategies, using a four point Likert scale ranging from 1 (usually don’t do it at all) to 4 (usually do it a lot). The questionnaire is designed to fit different situations. The Brief COPE [26] items are divided into 14 subscales, which each contain two items. The means of each pair of items were used to create the subscales. Correlations were computed for each two items in order to determine reliability. Only the scales with reliability scores of r > 0.20 were included in the study. Thus, the following scales were used: self-distraction (r = 0.22), active coping (r = 0.22), emotional social support (r = 0.49), instrumental social support (r = 0.52), behavioral disengagement (r = 0.51), venting emotions (r = 0.26), positive reframing (r = 0.51), planning (r = 0.22), humor (r = 0.33), acceptance (r = 0.29) and self-blame (r = 0.31).Sense of coherence [18] was measured using a series of semantic differential items that were rated on a 7-point Likert-type scale with anchoring phrases at each end. High scores indicated a strong SOC. An account of the development of the SOC scale and its psychometric properties appears in Antonovsky’s writings, which demonstrated that it was reliable and reasonably valid [17,18]. In this study, SOC was measured using the long-form scale consisting of 29 items. This scale includes items, such as: “Doing the things you do every day is” with answers ranging from 1 (a source of pain and boredom) to 7 (a source of deep pleasure and satisfaction). In the present study, we used the three subscales of SOC, which were namely meaningfulness, comprehensibility and manageability, and global SOC. The Cronbach’s alpha coefficient for the entire SOC scale was 0.82. The Cronbach’s alpha coefficients for the individual subscales were α = 0.81 for meaningfulness, α = 0.77 for comprehensibility and α = 0.79 for manageability.First, the frequencies, means and standard deviations of the demographic characteristics of the participants and the study variables were calculated. Second, in order to answer the first question, one-way ANOVA with LSD post hoc comparisons was conducted to understand the differences in the study variables among the three age groups. Third, we used Pearson correlations to explore the relationships among the study variables in each of the age groups. Finally, we computed Fisher z-scores to examine significant differences among these correlations in the three age groups.The preliminary results showed that for the entire sample, the SOC scores were quite low and there was little variance around the mean (M = 3.57). As for the coping strategies, it seems that this sample of women used self-distraction (M = 3.02) and self-blame (M = 3.13) as their main means of coping. The least commonly used coping strategies were emotional support (M = 2.06) and humor (M = 1.77).To answer the first question, one-way ANOVA was conducted to compare the means of the different variables among the three age groups. The results of this analysis are presented in Table 1.The results show that differences were most frequently seen between the youngest and the oldest groups, which were mainly in the coping-strategies variables. To a lesser extent, differences were observed between the middle-aged group and the oldest group. The differences between the youngest group and the middle-aged group were noted for only two scales of positive reframing and humor.Pearson correlations were used to answer the second research question. For each age group, a correlation was run to explore the relationships between SOC and the coping strategies. The results of this analysis are presented in Table 2.Our results revealed significant correlations between SOC and most of the coping strategies. As shown in Table 2, when we controlled for education, the direction, strength and significance of the correlations remained approximately the same in most cases. It should be noted that in the older age group, within which there was almost no variation in education, the correlations remained exactly the same. However, differences in the strength and the direction of the correlations were observed among the three age groups. Therefore, in order to understand whether there were significant differences between the groups in terms of the strength of these correlations, several z-tests were run to examine differences among the groups in the correlations between SOC and the different coping strategies. The results of this analysis are presented in Table 3.In most cases (for all of the strategies, except for humor and acceptance), significant differences were observed. The most prominent differences were observed between the two younger groups, which were namely the youngest and the middle-aged group, and the oldest group. For emotional support, instrumental support, venting, positive reframing and planning, there were differences in the strength of the relationships and the direction of the relationships. In contrast, for active coping, behavioral disengagement and self-blame, the only differences were in the strength of the relationships. The only difference observed between the youngest group and the middle-aged group was in the strength of the relationship between the use of the coping strategy of planning and SOC, with a stronger relationship observed in the younger group.To summarize, our results point to the fact that differences appeared mainly between the two younger groups and the oldest group. The youngest group and the middle-aged group tended to use the various coping strategies to the same extent and in most cases, they tended to make more use of coping strategies compared to the oldest women. Additionally, the relationships between SOC and the different coping strategies also differed mainly between the younger two groups and the oldest group. In contrast, among the oldest group, SOC was usually negatively correlated with the use of the different strategies. Among the youngest group and the middle-aged group, SOC was usually positively correlated with the use of adaptive strategies and negatively correlated with the use of emotional and non-adaptive strategies.This study examined Bedouin women from three age groups who represent different generations in a society that has undergone tremendous changes in recent years. Our aim was to explore similarities and differences in these women’s methods of coping and their coping resources. We also wanted to find out how the reported coping strategies and coping resources are related to each other in each of the examined groups. First, we examined some demographic background variables and as expected, we found that the older women were less educated than the younger women. Additionally, the older women were much less likely to work.A preliminary examination revealed that all three groups reported very low SOC levels compared to those usually seen in Western societies and compared to the objective scale, with scores close to the mean of the scale. These results resemble those of other studies that have compared the Bedouin population with Jewish Israeli population (e.g., [27]). As Antonovsky [18] claimed, it could be that one has to belong to a society whose values and traditions are stable in order to develop a strong SOC. However, the transition that Bedouin society is going though [1] might result in and be reflected by the weak SOC exhibited by the participants in our study. The fact that our results show a trend of stronger SOC among the younger groups might indicate a minor shift in this society.Regarding our first question, significant differences were found between the oldest women and the two younger groups. The meaningfulness component of SOC was strongest among the youngest women as it seems to be more accessible for them This indicates that for the younger generation, having meaning in one’s self and taking care of one’s self are seen to be more legitimate. On the other hand, the values on which the oldest women were raised and which were important for the former social structure of the Bedouin society are less prevalent among the younger generations. Thus, most of the values that the oldest women live for (e.g., strong relationships within the extended family, more power and respect for older people, very traditional and conservative sex roles) have been weakened and some of them no longer exist among the younger members of their community. The values that these women have lived for their entire lives are fading, leading to a decrease in the amount of respect granted to those who were formerly viewed as wise elders and a decrease in the amount of power held by those elders. Thus, this creates a situation in which these older women experience less meaningfulness in their lives. The same generational gap was observed in the behavioral components of the coping strategies. Indeed, planning, which has not traditionally been highly valued in the Bedouin context, was observed more among the members of the younger generations than among the members of the oldest generation. It seems that the younger generation is more aware of the importance of planning for the future, such as building a career and integrating that career with family life. In contrast, the oldest women believe that faith and religious values are the most important priorities. They believe that everything that happens is God’s will and therefore, there is no need to plan.Differences in coping strategies were also observed for behavioral disengagement and venting. Behavioral disengagement was the only strategy that the oldest women used more than by their younger counterparts. It could be that the older women who are less educated and work less tend to have weak social status and therefore, they disengage more. To the extent that these older women believe that they have no control over outcomes and are at the mercy of luck or fate, they will most likely feel helpless and rely upon passive and avoidant coping strategies, such as behavioral disengagement. These women do not have a wide repertoire of coping skills to use so they lean on disengagement, which is more available to them. The other coping strategy for which we observed differences, venting, is used more by the youngest women. Traditionally, a ‘good’ woman does not complain or whine. It could be that women use this strategy more as they become more Westernized.Our main aim in this study related to the relationships among the core concept of salutogenesis, sense of coherence (SOC)—a global orientation to see the world as more or less comprehensible, manageable and meaningful—and various coping strategies rooted in the stress, appraisal and coping theory; the problem-solving or emotional-coping orientation; and differences in these relationships among the different age groups. The oldest women seem to be embedded in the traditional collectivistic culture. Therefore, the role of SOC among these eldest women is different from the role of SOC among the youngest and the middle-aged groups. Our results indicate that among the oldest women, SOC is a protective factor as far as behaviors that are not adaptive among the oldest generation of the Bedouin population. Coping strategies, such as emotional and instrumental social support, do not help individuals from traditional collectivistic culture to overcome stressful events [16]. Thus, we can conclude that SOC first protects by preventing the use of non-adaptive behaviors. However, the role of SOC among the younger women who are becoming more Westernized appears to be more similar to the role of SOC in Western societies, in which SOC promotes good health [19]. Furthermore, SOC also seems to promote and advance adaptive behaviors among the younger women.Several limitations should be acknowledged. First, all of the data were collected via self-report questionnaires. Therefore, all data are subjective. Second, a potential degree of sample bias cannot be ruled out as we investigated a relatively small sample (especially of the youngest age group), which was not a representative sample of Bedouin woman.Based on the salutogenic theory and stress, appraisal and coping theories, we aimed to explore the role of coping resources and strategies when facing various stressful events. We explored this among three generations of Bedouin women who belongs to a society that has undergone rapid and dramatic changes during the past decade. We found some significant differences in the use of the various coping strategies and in the meaningfulness component of SOC among the three age groups. Moreover, we also found that SOC only prevents maladaptive behaviors for the oldest women while it also promotes the use of more adaptive coping strategies among the younger groups. It should be noted that the oldest women are an ‘at risk’ population as they do not make optimal use of coping strategies and SOC is a limited resource for them. These results give us, as researchers, an opportunity to demonstrate Antonovsky’s claim regarding the stability of SOC in a society that is undergoing tremendous changes. Further research should be conducted to expand our understanding of this matter. In the future, researchers might consider conducting a longitudinal study of the stability of SOC that would also consider variables, such as cultural change, aging, sociopolitical change and public-education achievements.All authors have designed the study. O.B.-L. has designed and written the manuscript. S.A.-K. has designed and helped with the statistical analyses. K.A.-S. has collected the data. E.H author has reviewed the manuscript.This research was funded by the Israel Science Foundation (ISF), grant number 723/15.The authors declare no conflict of interest.Means and standard deviations of the study variables among the three age groups.ᶺ p = 0.09; ᶺᶺ p = 0.07; * p < 0.05; *** p < 0.001. Note: Venting (ac, p = 0.02); Acceptance (ac, p = 0.03).Correlations between SOC and use of different coping strategies among the three age groups when we controlled for education.* p ≤ 0.05; ** p ≤ 0.01; *** p ≤ 0.001.Differences in the strengths of the correlations in the different age groups, as represented by z-scores.* p ≤ 0.05; ** p < 0.01; *** p < 0.001.

Med-MDPI/ijerph_3/ijerph-16-05-00805.txt

@@ -0,0 +1 @@
+Parent-child conversations contribute to understanding and regulating children’s emotions. Similarities and differences in discussed topics, quality of interaction and coherence/elaboration in mother-child conversations about emotional experiences of the child were studied in dyads who had been exposed to interpersonal trauma (N = 213) and non-trauma-exposed dyads (N = 86). Results showed that in conversations about negative emotions, trauma-exposed children more often discussed trauma topics and focused less on relationship topics than non-trauma-exposed children. Trauma-exposed dyads found it more difficult to come up with a story. The most common topics chosen by dyads to discuss for each emotion were mostly similar between trauma-exposed dyads and non-trauma-exposed dyads. Dyads exposed to interpersonal traumatic events showed lower quality of interaction and less coherence/elaboration than dyads who had not experienced traumatic events. Discussion of traumatic topics was associated with lower quality of mother-child interaction and less coherent dialogues. In conclusion, the effect of the trauma is seen at several levels in mother-child interaction: topics, behavior and coherence. A focus on support in developing a secure relationship after trauma may be important for intervention.Parents have a crucial role in helping children understand their inner world and supporting them in emotion regulation [1,2]. During daily conversations between parents and children, children learn which feelings are accepted and which are not. They learn if there are emotions that should be dampened and others that should be intensified. In addition, children discover whether there are emotional subjects to which their parents are attentive compared to others to which they are not. During these conversations children also learn whether they can rely on their parent to assist them through the reconstruction of a sequence of events in a meaningful and coherent manner. They gain an understanding of whether their parent is able to help them cope with negative feelings and regain a sense of confidence and security, particularly when recalling past aversive or traumatic events [3,4]. Parents who are capable of providing an enabling, child-focused, and organized emotional climate serve as a psychological secure base from which children can freely explore their emotional inner world. Inattentive and non-accepting parents on the other hand may negatively affect their child’s ability to create coherent and meaningful narratives about their emotional experiences [1,5].Parental traumatic experiences may impact parents’ ability to support their children [6] and influence children’s emotion conversations with their parents. Emotion conversations between parents and children help children to understand their emotions, both regarding everyday events, as well as regarding highly significant and emotionally laden events. Meaning making regarding these events is important for a healthy emotional development [7]. Meaning making may be of particular importance for very stressful and emotion-laden traumatic events, such as child exposure to domestic violence or sexual abuse [8].Co-constructed parent-child emotion conversations can be analyzed on three levels: (1) behavior of both partners, (2) coherence/elaboration of the conversation, and (3) content of the conversation [9,10]. The behavior of both partners describes the sensitivity of the caregiver and the cooperation of the child. Coherence refers to the narratives’ organization and describes the degree to which a story makes sense [8]. Coherent stories are internally consistent, fluent, detailed and focus on the emotion they ought to describe. As coherent stories are usually more detailed, they are likely to be more elaborate and consist of more words. The content of the conversation refers to what parent and child decide to discuss together. Most studies so far have focused on the way parent and child interact with each other and the coherence/elaboration of the narrative (e.g., [11,12]). The content of the conversation, particularly the topic parent and child decide to discuss together, has so far been largely neglected. The current study aims to fill this gap by focusing on all three aspects—the quality of mother-child interaction, coherence/elaboration of the narratives and the content of the stories—in mother-child conversations about different emotional experiences of dyads with and without trauma exposure. Exposure to traumatic events of either parent, child or both may hinder attunement between parent and child. This is unfortunate, because in highly stressful emotion-laden conversations maternal guidance is of great value for the child [13]. In families exposed to interpersonal traumatic events the parent-child interaction and coherence of the conversation are often of lower quality than in non-risk families. In addition, children with a less enabling parent likely feel more constrained to extensively explore and share their emotions and have less elaborative conversations compared to children with an enabling parent. Koren-Karie and colleagues [12] studied how sexually abused mothers talk with their non-abused children about emotional events. They found that the degree of maternal resolution of the trauma moderated the relationship between the abuse and the quality of mother-child interaction. More resolved mothers have been able to integrate the trauma in their view of the world and come to terms with the abuse, enabling them to focus more on the present and the current signals of their child. These mothers showed more sensitive guidance, their children were more cooperative, and their conversations were more coherent. Not only parental interpersonal trauma affects the quality of parent-child interaction. Additionally, when the child has been exposed to traumatic events, this influences mutual communication. Van Delft and colleagues [14] compared the quality of emotion conversations of sexually abused children with non-abused children and found a lower quality of interaction, less maternal sensitive guidance and less cooperation of the child. In a sample of mother-child dyads who both had been exposed to interpersonal trauma, in the case of exposure to marital violence, violence-exposed dyads showed less maternal sensitive guidance, less child cooperation and a lack of coherence/elaboration compared to a matched normative sample [15]. In addition, children showed striking behavior in interaction with their mothers after trauma exposure, such as overprotectiveness towards the parent and extreme aggressiveness [16].Parental exposure to interpersonal trauma impacts well-being which in turn impacts parenting and mutual communication (e.g., [17,18]). Exposure to interpersonal trauma, such as marital violence or maternal childhood sexual abuse, has been shown to be associated with more aggressive parenting, less warmth, less consistency in parenting and less attention for the emotional needs of children [16,17]. Parents and children contribute differently to the emotion conversation and also influence each other both positively and negatively. Maternal acceptance and encouragement have been associated with children’s emotional openness [19] and cooperation [20]. By contrast, parents who were unsupportive or rejecting were less likely to have children willing to be emotionally open with their parents [19]. As parent-child conversations are based on reciprocity, quality of the conversation may be negatively influenced both when parents or children are exposed to traumatic events. Indirect evidence for this hypothesis comes from the findings of Van Delft and colleagues [14] and Koren-Karie and colleagues [12] on the effect of interpersonal traumatic experiences on emotion conversations. While in the study by Van Delft and colleagues only the children had been exposed to trauma and not the parents, in the study by Koren-Karie and colleagues only the mothers had been exposed to trauma and not the children. In both studies the quality of interaction and coherence of conversation was of lower quality in families which were highly impacted by the interpersonal traumatic events. Little research attention has been paid towards the content in parent-child emotion conversations. Grych and colleagues [21] found that children exposed to interparental violence had fewer positive representations of parent figures in a narrative task than control children. Shields and colleagues [22] found that maltreated children’s parental representations in a narrative task were less positive and more negative than non-maltreated children’s representations. Both studies used the narrative of the children and did not make use of a co-constructed parent-child narrative.Two studies looked into references to emotions during jointly parent-child conversations [23,24], both comparing securely attached children with insecurely attached children in a normative sample. However, in both studies the focus was on the frequency of discussion of positive and negative events and the chosen topics were not part of the studies. In sum, only the rather global distinction between positive and negative topics has been examined. We could find no studies examining content of conversations of dyads with and without trauma exposure. The discussed content of the conversation may be related to the quality of the mother-child interaction and coherence/elaboration of their conversation. In order to discuss negative emotional experiences or traumatic events in an open, regulated, and coherent manner, children need a sensitive parent who provides them with an accepting and containing atmosphere and creates a secure base from which they can safely explore and share their thoughts, feelings, and (when occurring) painful memories [10]. However, establishing such an open and enabling atmosphere may be difficult for both partners, especially after experiencing interpersonal traumatic events for either parent, child or both [12,14,15].Coherence and quality of parent-child interaction have been associated with emotion understanding; elaborative coherent stories within a secure relationship contributed to more emotion understanding [2]. Children whose parents are better able to provide a secure base will have access to a wider range of topics compared to children whose parents have more difficulties with providing sensitive guidance of emotion conversations. These children will be able to describe more elaborative well-organized stories in collaboration with their parents. Finally, Laible and Thompson [24] found that insecure dyads have more difficulty discussing emotions and made less references to feelings than secure dyads. Attachment has been associated with quality of interaction [25], which has been associated with trauma exposure (e.g., [15]). This suggests that trauma-exposed dyads will have more difficulty to come up with a topic for an emotion conversation. Exposure to interpersonal trauma has been found to impact parenting and parent-child communication [17,18], resulting in lower quality of interaction and less coherent and elaborative dialogues in trauma-exposed dyads (e.g., [15]). So far, little research attention has been paid towards the content in parent-child emotion conversations and its relations with the quality of interaction and coherence/elaboration. The objectives of this study were to examine similarities and differences in content, quality of interaction and coherence/elaboration in mother-child conversations about emotional experiences of children in dyads who have been exposed to interpersonal trauma compared with non-trauma-exposed dyads.1. Dyads exposed to traumatic experiences will refer to these traumatic experiences when discussing both positive and negative emotions. In addition, trauma-exposed dyads will describe fewer positive stories and more negative stories about parental figures. Differences in other topics are explored as well. 2. Trauma-exposed dyads are expected to discuss a smaller range of topics than non-trauma-exposed dyads, and will more often not be able to decide on a topic together.3. In trauma-exposed dyads the mother-child interaction is expected to be of lower quality than in non-trauma exposed dyads. Dyads exposed to traumatic experiences will provide less coherent and less elaborative emotion stories than non-trauma exposed dyads.4. Discussion of traumatic topics will be associated with lower quality of mother-child interaction and less coherent dialogues. No specific hypotheses were made regarding discussion of other topics and quality of interaction or coherence.A total of 299 mother-child dyads from different samples recruited in the Netherlands and Israel were included. The samples were combined into a group of participants exposed to interpersonal trauma and a control group (see Table 1 for descriptive statistics). There were no differences in children’s age or gender between the two groups, but in the trauma group the years of maternal education was lower and participants from the Netherlands were overrepresented.All participants participated, as part of other studies [12,14,15,26,27], in the Autobiographical Emotional Events Dialogue (AEED; [28]). For the current study, all transcripts were recoded for content analysis, and the originally coded assessments of quality of mother-child interaction and coherence were used. Coders were blind to group allocation (trauma/non-trauma) and different coders coded the content of conversations and quality of mother-child interaction/coherence. Content of mother-child emotion conversation was measured with a newly developed coding system [29] for content analysis of the Autobiographical Emotional Events Dialogue (AEED; [28]). In this task mother and child were asked to co-construct four separate stories about events in which the child felt happy, scared, angry, or sad. The dialogues were transcribed verbatim. The discussed event for each emotion was classified into mutually exclusive topics based on the essence of the story to avoid multicollinearity (e.g., trauma-related, parent positive, parent negative, peer-related) and dichotomously coded (no = 0, yes = 1). First, the first two authors read together and discussed 94 transcripts to define common topics. A codebook was composed based on these topics generalized over emotions. In the ensuing coding process, topics were added when the existing themes did not cover a discussed topic. A total of 20% of transcripts were double coded for reliability. Differences between coders were resolved through discussion. In the next stage, themes that described similar topics (e.g., stories about different family members besides the core family were combined into the category ‘other family members’) or were difficult to differentiate (e.g., some stories were about ‘parents’ instead of ‘mother’ or ‘father’ resulting in the combined category ‘parent-related’) were combined. Finally, dialogues on very rare topics were recoded—if possible—into a more common topic, resulting in the final set of topics for analyses presented in Table 2.Quality of mother-child interaction was also assessed in the AEED [28]. Transcripts were coded with the AEED coding system and quality of mother-child interaction was scored on seven scales for the mother and seven parallel scales for the child. Rating of the transcripts was done by marking indicators for the various scales as they appear throughout the transcript and then assigning a score on each of the scales both on the frequency and strength of these indicators. The scales are: Focus on the task (Mother/child is focused on the child’s emotions with no shifts to irrelevant details or to mother’s own feelings); Clear boundaries (Mother/child keeps their appropriate roles: mother does not force her own ideas/emotions on the child or becomes overwhelmed by the child’s themes, and the child does not assume a parental role such as promising to protect the mother); Acceptance and tolerance (Mother/child enables the other to express a wide range of emotional themes without being defensive or judgmental); Hostility (Mother/child shows thematic hostility, anger or derogation); Involvement and reciprocity (Mother/child is positively engaged in the task and shows genuine interest in the stories); Containment of negative feelings (Stories with negative themes are ended with positive resolutions and an emphasis on the child’s coping abilities and strength); Structuring and Elaboration (Mother facilitates the child in narrating rich and coherent stories and child tells rich and detailed stories). Each scale was scored between 1 and 9, and a higher score represented more of the coded behavior (in all scales except for Hostility high scores reflect positive behaviors. Scores for Hostility were inverted for the sum scales). For example, a mother who mostly accepts her child’s ideas but there are one or two indices in which she responds with slight impatience would get a high score on the Acceptance scale, whereas repeated rejection of the child’s ideas will lead to a low score on this scale. In addition, a very strong marker of specific behavior can lead to a low score, for instance when a mother derogates her child (e.g., “What a stupid story! Is that your best example? Couldn’t you come up with a better example? Now sit still and behave like a normal boy!”). All maternal scales were summed in a measure for ‘maternal sensitive guidance’ and all child scales combined represented ‘child cooperation and exploration’. Coders were blind for group allocation (trauma/non-trauma). Coders were trained by the developer of the coding system (N. Koren-Karie) and established adequate reliability. Adequate inter-rater reliabilities were reached for the composite scores (ICCs ranging from 0.76 to 0.95 for maternal sensitive guidance, 0.65 to 0.95 for child cooperation). Maternal sensitive guidance ranged from 21 to 60 (M = 43.31, SD = 7.37, possible range 9–63) and child cooperation and exploration ranged from 23.5 to 62 (M = 45.14, SD = 6.63, possible range 9–63). In addition to the scales, striking features in mother-child communication, such as extreme aggressiveness, child parentification or maternal frightening behavior, were coded dichotomously (present = 1, absent = 0) per emotion.Coherence of the conversation was assessed with two overarching scales of the AEED coding system: Adequacy of the story (Mother and child describe separate stories for all four emotions and the stories match the emotions or themes they ought to describe); and Coherence (Mother and child construct stories that are fluent and clear). These scales are also scored between 1 and 9. High scores on these scales are given when mother and child construct stories that are internally consistent, fluent, detailed and match the emotions they ought to describe. The stories have a beginning, middle and end. The coder understands what the described event was and what the child was thinking, feeling and doing during that event. In coherent narratives there are no shifts to irrelevant topics or details. Negative emotions are not left open, but rather the stories end happily when possible or with an emphasis on the child’s strength and ability to competently deal with negative emotions. Inter-rater reliability for this composite score ranged from 0.73 to 0.93. Quality of the dialogue ranged from 2 to 17 (M = 8.69, SD = 3.65, possible range 2–18). For all analyses SPSS (version 25) was used. Winsorizing was used to replace one outlier for maternal sensitive guidance. A total of 14 missings on maternal years of education were imputed with the total mean for analyses, but not for descriptive purposes. Differences in content of emotion dialogues between families with and without trauma are described with frequency analyses and tested for significance with chi-square tests. Differences in quality of mother-child interaction were tested with t-tests. The association between the content of the emotion conversations and quality of mother-child interaction and coherence are described using regression analyses, while controlling for children’s age and maternal years of education. Topic choices of dyads who had been exposed to interpersonal trauma and dyads who had not been exposed to trauma were compared. As expected, in frequency analyses differences were found between discussed topics in dyads. For all negative emotions, trauma (e.g., incidents of parents fighting) was discussed more often in trauma-exposed dyads. Contrary to expectation, in conversations about the emotion happy, trauma-exposed dyads were more likely to discuss positive stories about parental figures than dyads without trauma. Overall, however, trauma-exposed dyads focused less on relationships (see Table 2 for an overview of topics per emotion between groups). Only for the emotion sad were trauma-exposed dyads less likely to come up with a story than non-trauma exposed dyads.Interestingly, the most common topics chosen by dyads to discuss for each emotion were similar between families with and without exposure to interpersonal trauma, except for the discussion of trauma for scared in trauma-exposed dyads (Table 3).A description of the quality of mother-child interaction in samples with and without trauma exposure is provided in Table 4.Mothers of trauma-exposed dyads were less sensitive in interaction with their children (t(296) = 5.18, p < 0.001) and children were less cooperative (t(296) = 4.39, p < 0.001). Exploratory analyses did not find significant differences within the trauma group between dyads where only mothers, only children, or both partners had been exposed (maternal sensitivity: mother/child trauma vs. both trauma, resp. t(358) = −0.28, p = 0.780, t(343)= −0.07, p = 0.947; child cooperation: resp. t(358) = −0.05, p = 0.963, t(343) = −0.56, p = 0.576).The additional questions regarding striking features in mother-child interaction showed the mothers in the trauma group to be more self-involved: they focused more on their own emotions (happy: t(280.2) = −3.65, p < 0.001, scared: t(235.6)= −2.61, p = 0.010), more often rejected the topic of the child (scared: t(212) = −2.88, p = 0.004), and more often told the story of the child by themselves (happy: t(291.1) = −2.28, p = 0.024). In trauma-exposed families the dialogues were less coherent (t(137.2) = 3.83, p < 0.001). Coherence was not worse in samples in which both partners had been exposed to trauma compared to samples in which only one partner had been exposed to trauma (mother/child trauma vs. both trauma, resp. t(358)= 0.43, p = 0.670, t(343)= −0.33, p = 0.739). See Table 4 for a description of coherence of emotion conversation in samples with and without trauma exposure.Mothers of trauma-exposed dyads elaborated less while describing an emotional event than mothers of dyads in which no partner had been exposed to trauma (happy: F(1, 297) = 9.03, p = 0.003, scared: F(1, 297) = 24.97, p < 0.001, angry: F(1, 297) = 12.25, p = 0.001, sad: F(1, 297) = 20.38, p < 0.001). Similar results were found for children of trauma-exposed dyads (happy: F(1, 297) = 5.87, p = 0.016, scared: F(1, 297) = 17.76, p < 0.001, angry: F(1, 297) = 7.78, p = 0.006, sad: F(1, 297) = 5.36, p = 0.021) (see Table 5).Controlled for children’s age and maternal years of education, topics of emotion conversation were regressed on quality of mother-child interaction and coherence. To limit the number of analyses, composite scores for relationship-related topics (stories on peers and family combined, for happy positive themes and for scared, angry and sad negative themes) and object-related topics were calculated for all four emotions, and regressed together with the topic ‘trauma’ and ‘no story’ (coded when the dyad was not able to come up with a story to discuss). Discussion of the topic ‘trauma’ was negatively associated with maternal sensitive behavior (scared: β = −0.19, p = 0.001; angry: β= −0.25, p < 0.001), child cooperation (scared: β = −0.13, p = 0.027, angry: β = −0.12, p = 0.038) and coherence (scared: β = −0.16, p = 0.008). ‘No story’ was negatively associated with quality of mother-child interaction and coherence for almost all emotions (maternal sensitive behavior: happy: β = −0.12, p = 0.047, scared: β = −0.14, p = 0.014, angry: β = −0.12, p = 0.043; child cooperation: happy: β = −0.23, p < 0.001, scared: β = −0.13, p = 0.026, sad: β = −0.17, p = 0.003; coherence: happy: β = −0.18, p = 0.002, scared: β = −0.21, p < 0.001, angry: β = −0.13, p = 0.028, sad: β = −0.19, p = 0.001). Discussion of negative relationship-related topics for the emotion angry was negatively associated with maternal sensitive behavior (β = −0.13, p = 0.038). Additional analyses showed that particularly discussion of negative stories about parents (β = −0.17, p = 0.007), and not about siblings (β = −0.05, p = 0.405) or peers (β = −0.02, p = 0.718), was associated with lower maternal sensitive guidance. No other associations between the composite scores for relationship-related topics and object-related topics for the other emotions and maternal sensitive behavior, child cooperation or coherence were found. Trauma-related topics were assumed to be more difficult to discuss than daily problems such as a fight with a classmate or a bad grade, placing dyads who have been exposed to trauma at higher risk for negative mother-child interaction. However, additional analyses within the sample of trauma-exposed dyads showed there were no differences in maternal sensitive guidance (t(210) = 1.22, p = 0.226), child cooperation (t(210) = 1.34, p = 0.183) or coherence (t(210) = 0.85, p = 0.397) between the dyads who discussed a topic coded under ‘trauma’ and dyads who did not discuss a traumatic topic. Hierarchical regression analyses with the control variables added in the first step, sample allocation in the second step, and the topics (trauma, no story, relationship-related, object-related) in the third step showed similar results. Exposure to trauma significantly predicted maternal sensitive behavior, child cooperation and coherence. Of the coded content only ‘no story’ was significantly associated with lower maternal sensitive behavior, child cooperation and coherence. As hypothesized, trauma-exposed dyads more often discussed trauma-related topics, although the number of positive or negative stories about parental figures was not significantly different. The fact that children of trauma-exposed dyads chose the trauma as topic for discussion more often than non-trauma exposed children is not really surprising since this is a salient topic for both conversation partners. Contrary to expectation, trauma-exposed dyads did not describe fewer positive and more negative stories about parental figures. This finding has been found in two studies with abused children participating in a narrative task [21,22]. In the current study, the mutually exclusive coding of the topic of a story might have hindered coding more stories as negative about parental figures in trauma-exposed dyads. On the other hand, compared to earlier studies, mother and child participated together in a discussion task, which could make it more difficult for the child to discuss negative stories with the parent in which the parent is the focus of the negative emotions, especially when the mother-child interaction is of lower quality; and could be a sign of the child’s tendency to please the mother. Support for this interpretation can be found in the fact that for the emotion happy, trauma-exposed dyads were even more likely to discuss positive stories about parental figures. Overall, trauma-exposed dyads focused less on relationships in their conversations than non-trauma-exposed dyads. In light of previous research with traumatized individuals this can be explained by the detrimental effects of trauma exposure and posttraumatic stress symptoms, such as emotional numbing, on attachment behavior and interpersonal functioning. Traumatic memories can lead to distancing and avoidance of interpersonal triggers that create traumatic re-enactments [30]. To protect oneself from traumatic memories individuals rather focus on topics which are not relationship-related and therefore ‘safe’.Contrary to our hypothesis, dyads exposed to interpersonal trauma did not discuss a smaller range of topics than non-trauma-exposed dyads. However, they did have more trouble deciding on a topic to discuss together and were less likely to come up with a story than non-trauma exposed dyads. The similar range of topics discussed by trauma-exposed and non-trauma-exposed dyads shows that, despite trauma, on a daily basis children are still mainly consumed by worries about friends, sports or common scary things such as thunderstorms. As expected, dyads exposed to traumatic events showed lower quality of interaction than dyads who had not experienced traumatic events, and their conversations were less coherent and elaborate. These findings have been found in previous studies (e.g., [15]). In addition, mothers in the trauma samples were more self-focused and more often rejected the child’s story, which may be related to their higher probability to not be able to decide on a topic together (Hypothesis 2). These findings are in line with previous research showing that traumatized mothers have greater difficulty with parenting [31] and children need their parents as a psychological secure base from which to explore their emotional inner world [1,5]. Additional explorative analyses showed that the quality of mother-child interaction was similar in samples in which either one (child/mother) or both partners had been exposed to traumatic experiences. Although this result needs to be interpreted with caution, because of small sample sizes of the samples in which only one partner had been exposed to trauma, this is an interesting finding. One would expect that mothers, who are the mature and experienced partner of the dyad, will have more impact on the conversation than the child, which has been found before [32]. Trauma-exposure however may have such a great influence that the dyad as a whole is negatively influenced, regardless of which partner was exposed to the traumatic events. Studies on the effect of sexual abuse of one’s child on parental functioning show similar results [33]. This stresses the vulnerability of the dyad, since their ability to create a meaningful, focused and coherent narrative together is negatively affected when one part of the dyad has suffered trauma. As hypothesized, discussion of traumatic topics was associated with lower quality of mother-child interaction and less coherent dialogues. This could not be explained by the fact that trauma-related topics may be more difficult to discuss than more daily problems, suggesting that it is particularly the exposure to trauma and not the discussion of the traumatic events which is associated with the lower quality of mother-child interaction. Discussion of relationship-related topics was not related to the quality of mother-child interaction, except for an association between more negative discussion about parents and lower quality of maternal behavior. This seems contrary to our other findings that trauma-exposed dyads have more difficulties discussing negative events about parents with parents. On the other hand, it may be simply the case that children of less sensitive mothers have more negative incidents between themselves and their mother to discuss.To our knowledge this is the first study which looks into the association between the content of emotion dialogues and quality of mother-child interaction and coherence in dyads exposed to interpersonal trauma compared to dyads not exposed to trauma. Strengths of the study are the use of an observation measure for coding the quality of mother-child interaction, and the combination of different samples from previous studies enabling us to compare relatively large samples of families with and without exposure to interpersonal trauma. The samples with and without trauma exposure were comparable on all background variables except for maternal years of education and country of sampling. A limitation of the study is that both the content of the conversations and the quality of the interaction are measured with the same instrument (AEED). However, for both constructs different coding systems were used and were coded by different coders who were blind to group allocation. In the current study, only mothers and their children participated. Fathers interact differently with their children than mothers [34]. Most studies on the effect of trauma on parenting have been done with mothers, and little is known on how trauma-exposure affects parenting in fathers [35]. In future studies it would be interesting to look into how trauma affects emotion conversations between fathers and their children as well as emotion conversations within family systems of parents and siblings. Children may derive different forms of support from different persons, and may therefore benefit differently from conversations with their fathers and their mothers. More insight into these support networks of children may be helpful in the provision of treatment after trauma exposure. In addition, the analyses in the current study on (lack of) differences in quality of interaction between samples in which one partner or both partners had been exposed to trauma gave interesting food for thought, but were underpowered. A larger study with equal cells for trauma exposure of one (separately for mother and child) or both partners would help to clarify this point. Exposure to interpersonal trauma affects not only individual family members but also the parent-child dyad as a whole and its lasting effects are still noticeable many years later. The effect of the trauma is seen at several levels: sensitive guidance of the mothers, cooperation of the child, the ability of the dyad to form a coherent and elaborate story and the topic the dyad choses to talk about. In this study a link has been found between exposure to interpersonal traumatic events and emotion dialogues. Higher quality of parent-child interaction, including narratives, has been associated with better child functioning (e.g., [10]). Clinicians working with families exposed to interpersonal trauma may pursue problematic conversations about emotions as a possible explanation for children’s problems and might consider interventions that focus on this aspect of parent-child interaction. Conceptualization, M.M.O. and N.K.-K.; Formal analysis, M.M.O.; Funding acquisition, M.M.O.; Investigation, M.M.O., N.K.-K. and A.E.B.-H.; Writing—original draft, M.M.O.; Writing—review & editing, M.M.O., N.K.-K., J.C.d.S., P.D.D.G. and C.S.This study was funded by the Koninklijke Nederlandse Akademie van Wetenschappen (Royal Dutch Academy of Sciences; KNAW), project number UPS/PM/3971.The authors would like to thank Ivanka van Delft, Margreet Visser and Rachel Getzler-Yosef for the use of the data from their PhD-projects, and Naomi Koning for her help in coding the Dutch transcripts for content analyses.The authors declare that they have no competing interests. The funders had no role in the design, execution, interpretation or writing of the study.Descriptive statistics.a) Percentages based on the trauma group only.Frequency of topics between dyads with and without interpersonal trauma exposure.* p ≤ 0.05; ** p ≤ 0.01; *** p ≤ 0.001.Most common topics per emotion between dyads with and without trauma exposure.Quality of mother-child interaction in samples with and without trauma exposure.Number of words of mothers and children used for describing an emotional event.

Med-MDPI/ijerph_3/ijerph-16-05-00806.txt

@@ -0,0 +1 @@
+This paper explores the validity (sensitivity and specificity) of different cut-off levels of the UNICEF/Washington Group Child Functioning Module (CFM) and the inter-rater reliability between teachers and parents as proxy respondents, for disaggregating Fiji’s education management information system (EMIS) by disability. The method used was a cross-sectional diagnostic accuracy study comparing CFM items to standard clinical assessments for 472 primary school aged students in Fiji. Whilst previous domain-specific results showed “good” to “excellent” accuracy of the CFM domains seeing, hearing, walking and speaking, newer analysis shows only “fair” to “poor” accuracy of the cognitive domains (learning, remembering and focusing attention) and “fair” of the overall CFM (area under the Receiver Operating Characteristic curve: 0.763 parent responses, 0.786 teacher responses). Severe impairments are reported relatively evenly across CFM response categories “some difficulty”, “a lot of difficulty” and “cannot do at all”. Most moderate impairments are reported as “some difficulty”. The CFM provides a core component of data required for disaggregating Fiji’s EMIS by disability. However, choice of cut-off level and mixture of impairment severity reported across response categories are challenges. The CFM alone is not accurate enough to determine funding eligibility. For identifying children with disabilities, the CFM should be part of a broader data collection including learning and support needs data and undertaking eligibility verification visits.It is critical that education data systems are disaggregated by disability to measure progress in achieving access to quality education for children with disabilities, and efforts to enable this are moving forward globally. Disability-disaggregated education data are required to track progress towards various frameworks including the Convention on the Rights of Persons with Disabilities (CRPD) [1], the Sustainable Development Goals (SDG) [2] and the Incheon Strategy to “Make the Right Real” for Persons with Disabilities in Asia and the Pacific [3]. There is widespread consensus on the urgency to support Ministries of Education (MoEs) to disaggregate their Education Management Information Systems (EMISs) by disability, and the importance of doing so using tools which are valid and internationally comparable [2,4,5]. Given the complexity of disability measurement, efforts to develop and agree upon tools for disability measurement that are valid, feasible and comparable have taken statisticians and researchers decades. Whilst debate remains lively, the urgency to gather baseline data for the SDGs has required consensus. In a statement titled Disability data disaggregation joint statement by the disability sector [6], peak disability agencies such as the International Disability Alliance, the World Health Organization, UNICEF, United Nations Development Programme, and the UN Partnership to Promote the Rights of Persons with Disabilities, amongst others, agreed that the Washington Group on Disability Statistics (WG) modules should be used to disaggregate data sets to measure SDG indicators; the WG Short Set of questions for adults and the UNICEF/WG Child Functioning Module (CFM) for children.The CFM has been developed for measuring child functioning in surveys with parents/caregivers as proxy respondents for the child’s functioning. It has been validated in different settings [7,8,9,10,11,12,13,14,15,16] and was finalised in 2016. The CFM is designed for children between two and 17 years and covers a range of areas for measuring functioning difficulties, including: seeing, hearing, walking, self-care, speaking, learning, remembering, anxiety/worry, depression/sadness, controlling behaviour, attention/concentrating, accepting changes in routine and making friends. Response categories for most questions are: “no difficulty”, “some difficulty”, “a lot of difficulty” and “cannot do it at all”.Recent advice from the United States Agency for International Development (USAID), a key donor, is that the WG Short Set and/or the CFM should be used wherever possible in USAID-funded education programs to disaggregate data sets [17]. This is a positive indication of donor commitment to measuring outcomes towards fundamental human rights. If the CFM is to be used to disaggregate EMISs, it is critical that its properties are understood when proxy respondents are teachers and to test its measurement accuracy when used for education systems.There are various purposes for which disability identification is needed in data aimed at ensuring inclusion of children with disabilities. From determining funding eligibility at an organizational or an individual level, determining learning and support needs of a student, to comparing equalization of access to socioeconomic rights through disability-disaggregated census or household survey data. The purpose has implications for the approach to disability identification, and for the degree of accuracy required in the instrument that determines the classification of disability. Madden highlighted the importance of designing valid tools which take into account the evidence for and consequences of score interpretation and use, and establishing meaningful thresholds on the spectrum of disability experience [18].Disability can be seen as a continuum ranging from minimal difficulties to fundamental impacts on a person’s life. On an instrument designed to measure this functioning to disability continuum, the point along the span which is used to define someone as having a disability is referred to in this paper as a cut-off. It is critical that the rationale and implications of the cut-off are clearly understood. If for example the cut-off is relatively low on the continuum and includes mild disabilities (such as difficulty seeing which can be entirely overcome with glasses), the number of children counted as having disability will be high. Whereas if severe disability is the cut-off (having a great deal of difficulty with basic functions), the number of children counted as having disability will be comparatively low. The cut-off level must be appropriate to, and will alter depending on, the purpose for identifying disability. Education systems may consider it important to identify children with mild and moderate disability to enable early intervention and educational accommodations, whereas a scheme establishing eligibility for monetary benefits may target a higher level of disability [19].The recommended criterion for identifying disability using CFM is having difficulty functioning at the level of at least “a lot of difficulty” [20], or “daily” for anxiety and depression questions. The USAID guidance document [17] states that “for a more nuanced analysis of disability, the answers can be used as a regular scale, with “cannot do it at all” denoting severe disability while “some difficulty” denoting minor disability in each functional domain. Answers across all domains can also be combined into a larger scale.” (p. 4). However recent studies in Cameroon, India and Fiji [12,21,22,23] indicate that there is a significant variation in how parents choose response categories to report functioning difficulties and that the cut-off “a lot of difficulty” misses significant numbers of children with moderate to severe impairments. That is, this cut-off had low sensitivity in identifying disability. When used in large household surveys or censuses the importance of these differences may be considered within acceptable margins of error. However, within an education system the tool is used for different purposes and a response cut-off with a high sensitivity is needed. Sensitivity and specificity are a trade-off and selecting a lower severity response category, for example “some difficulty”, may result in lower specificity. That is, the chance increases of falsely identifying some children as disabled who do not have a disability.In a rapidly modernising information technology age, EMISs are increasingly based on individual electronic data files [24]. Data from these systems are not only used to monitor and evaluate progress towards inclusive education at a large area level but are capable of and being used to determine individual student eligibility for funding related to disability status. A tool appropriate for national surveys may not also be reliable or valid in identifying individual students’ levels of functioning. It is critical that people making decisions about incorporating disability within EMISs understand that tools they are being advised to use for national or large area monitoring may have limitations for individual level assessments.This study was undertaken in the context of an Australian aid funded education sector project in Fiji. The required purposes for disability data in Fiji’s EMIS included identification of children with disabilities, by disability type and severity, to enable resource allocation based on individual level data, and to enable monitoring, planning and reporting against policy and other commitments. The key question for the Fiji MoE was the extent to which the CFM is effective when used by teachers to identify the presence and severity of disability amongst children in Fiji. Validity and reliability of specific domains (seeing, hearing, walking, speech and cognition) were reported elsewhere [21,22,23]. This paper focuses on the performance of the CFM as a whole. With the overarching aim of identifying a valid, reliable and feasible method for Fiji to identify children with disabilities in schools to enable monitoring, planning and reporting against policy commitments, the objectives of this paper are to:(1)Determine the validity (sensitivity and specificity) of different cut-off levels of the CFM for predicting the presence of disabilities in primary school aged Fijian children compared to standard clinical assessments of impairment.(2)Determine the inter-rater reliability between teacher and parent CFM responses.Determine the validity (sensitivity and specificity) of different cut-off levels of the CFM for predicting the presence of disabilities in primary school aged Fijian children compared to standard clinical assessments of impairment.Determine the inter-rater reliability between teacher and parent CFM responses.A cross-sectional diagnostic accuracy study, two-gate design with representative sampling [25] was undertaken from March-July 2015 in Fiji. In diagnostic accuracy studies, the index test whose accuracy is being investigated (CFM) is compared to reference standard (clinical) tests, sometimes termed “gold standards” [26,27]. The purpose of a diagnostic accuracy study is to evaluate the ability of the index test to correctly classify study participants into two categories, those with and without the ‘target condition’. Diagnostic accuracy is based on measuring sensitivity and specificity values at each cut-off level. For the purpose of assessing the sensitivity and specificity of the CFM against the reference tests, we have essentially defined disability as clinically assessed impairment of a moderate or more severe level. There are inherent limitations in assuming that medical impairment assessments are “gold standards” for disability. However, this approach enabled a validated, consistent and objective means of measuring an aspect of disability, i.e., impairment, against which the self-report-based CFM could be compared.Ethics approvals were obtained from the University of Melbourne’s Human Research Ethics Committee (#1543942, 17/03/15) and the Fiji MoE’s ethics committee (RA09/15, 5/03/15). All subjects had written consent and children’s assent was obtained prior to each clinical assessment. Sampling was purposive regarding school selection and student participation. Participants for the study were 5–15 year old students recruited from ten special schools and five inclusive education (mainstream) schools from the four administrative divisions in Fiji. Children invited to participate included: all children in the special schools, and all children in the mainstream schools previously identified by the school to have disabilities, and selected controls matched by age, sex, ethnicity and location (Table 1). The flowchart of participation is shown in Figure S1 (Supplementary Material). Invitations to parents were included in the information and consent process for participation of the children. Teachers in all study schools were informed of the research and given information and consent forms. After the children had been assessed and parents interviewed, respective teachers of the children were provided questionnaires to complete. Representative sampling focused on including cases with mild/moderate through to profound impairment to minimise “spectrum effect”, whereby a sampling bias towards including only cases with more significant impairment can lead to higher estimates of sensitivity and specificity [25]. This was operationalized in two ways: (i) by keeping tallies on impairment levels of children throughout recruitment and working closely with schools to achieve a mixture of impairment severity levels; and (ii) by assessing large numbers of children who were not initially identified by schools as having disability, which resulted in a sample with a full spectrum of function/impairment, including those around the lower or borderline end, which was necessary to minimise “spectrum effect”. Sample size was estimated based on minimum number to achieve a sensitivity or specificity of 0.85 (prevalence 0.10, alpha 5%, 1-beta 80%; CI 95%, lower confidence limit 0.65) [28]. A target of 52 cases and 52 controls were sought under each of five impairment domains (vision, hearing, musculoskeletal, speech and cognition).This study used a draft of the CFM (5–17 year age group) current at February 2015, with permission from UNICEF and the Washington Group. Appendix A lists the differences between the version used in the study and the final version of the CFM, which is available from www.washingtongroup-disability.com. Translation and pretesting processes are described in [21]. For the diagnostic accuracy analysis in this paper, only seven CFM domains are included (seeing, hearing, walking, speaking, and three cognitive domains—learning, remembering and focusing attention) as these relate directly to constructs measured in the clinical assessments.For clarity, the term “CFM-7” is used throughout this paper when referring to this group of domains. For other analysis in the paper the remaining domains (self-care, anxiety/worry, sadness/depression, controlling behaviour, accepting changes to routine and making friends) are included and the term “CFM-13” is used to refer to the entire module. Table 2 provides the wording of the CFM questions and response categories and illustrates the domains referred to by the terms CFM-7 and CFM-13.Clinical tests were undertaken for vision, hearing, musculoskeletal impairment, speech and cognition using reference standard (clinical) tests considered the best available tests regarding the conditions of interest [26,27]. The clinical tests for this study were selected based on international standards for vision and hearing and well validated tools for speech, musculoskeletal impairment and cognitive impairment. Detailed descriptions of these assessments and how they were implemented in this study are available elsewhere [22,23,24] and summarised in Appendix B.Case definitions. Vision impairment: presenting visual acuity in the better eye <6/18 and ≥6/60 (moderate), <6/60 and ≥3/60 (severe) and <3/60 (blind) [29]. Hearing loss: 41–60 dBA (moderate), 61–80 dBA (severe) and ≥81 dBA (profound). Children identified on the Rapid Assessment of Musculoskeletal Impairment with structure impairment including “severe”, “moderate” and “mild” effect on the musculoskeletal system’s ability to function as a whole were identified as cases with mobility impairment [30]. Children identified to have impairment only affecting the upper limb were excluded to enable comparison with the CFM question on walking. Speech impairment: Intelligibility in Context Scale [23,31] scores: 1.8 to <2.5 (moderate) and 1.0 to <1.8 (severe). Cognitive impairment: assessed using the Cambridge Neuropsychological Test Automated Battery (CANTAB) [32], subjects with Overall Impairment Scores of 3 (moderate) and 4–5 (severe) [22].Assessment camps were run over two to five days at each school in rooms set up with multiple assessment stations. Parents/caregivers attended the screening camp where an interviewer administered the CFM in a location separate from the reference standard assessments, using either the Fijian, Fijian-Hindi or English version depending on parent preference. Interviewers had received a half-day training in administration of the questionnaire. In-situ training also occurred during the early stages of data collection, with the lead researcher providing clarification about administration as questions arose. It was self-completed by teachers either during the camp or within the following week; teachers received no training other than instructions to carefully follow the skip-prompts in the questionnaire. The clinical team were blinded to the CFM results and teachers and parents were blinded to each other’s CFM responses and to clinical results.Statistical analysis was undertaken using SPSS Version 24 (IBM, Armonk, NY, USA) and MedCalc v.17.6 (MedCalc Software, Ostend, Belgium). Descriptive statistics were calculated for participant demographics and CFM-7 results were cross-tabulated by clinical results. To analyse diagnostic accuracy of the CFM-7, the case definition was: child has impairment in at least one of the five clinical assessments (see “Case definitions” above). The definition to determine CFM-7 response was the highest level of difficulty reported against any of the seven domains. For example, for a child assessed as having “a lot of difficulty” seeing and “some difficulty” speaking, the overall CFM response would be recorded as “a lot of difficulty”.Sensitivity (Sn), specificity (Sp) and likelihood ratios (LR) were calculated for each respondent type (parent or teacher) for each cut-off level. True positives are children with impairments (assessed using the reference standard (clinical) assessments, defined by the case definitions in Section 2.2.2), who are correctly identified by the CFM as having difficulty in the respective functioning domain. True negatives are children without impairments who are correctly identified by the CFM as not having difficulty in the respective domain. False positives are children without impairments who are incorrectly identified by the CFM as having difficulty. False negatives are children with impairments who are incorrectly identified by the CFM as not having difficulty. Positive (and negative) LRs indicate how many times more likely a positive (or negative) test result is obtained when the target condition is present than when it is absent:Sn = true positives/total casesSp = true negatives/total controlsPositive LR = Sn/(false positives/total controls)Negative LR = (false negatives/total cases)/SpSn = true positives/total casesSp = true negatives/total controlsPositive LR = Sn/(false positives/total controls)Negative LR = (false negatives/total cases)/SpReceiver operating characteristic (ROC) curves were constructed separately for parent and teacher CFM-7 responses to determine the Area Under the ROC Curve (AUC). ROC curves are constructed by plotting the false-positive rate (1—specificity) against the true-positive rate (sensitivity) at each cut-off value defined by the CFM and then drawing a line from x = 0, y = 0 through the values at each cut-off point; the AUC is an overall figure of diagnostic accuracy with a perfect test having a value of 1.0 and a value of 0.5 suggesting that the test result is no better than chance [33,34]. AUC interpretations were classified as excellent (0.96–1.0), very good (0.9 to <0.96), good (0.8 to <0.9), fair (0.7 to <0.8), poor (0.6 to <0.7), and useless (0.5 to <0.6) [33]. ROC curves used dichotomous clinical variables, differentiating cases and controls based on definitions outlined earlier.The Youden Index (YI) was calculated for each ROC curve to determine the statistically “optimal” cut-off level for each disability domain (seeing, hearing, walking, speaking, learning, remembering and focusing attention) and respondent type. The YI is the maximum vertical distance between the ROC curve and the line of random chance ([x = 0, y = 0] to [x = 1, y = 1]) and is calculated as maximum (Sn + Sp − 1). That is, the cut-off point at which (sensitivity + specificity − 1) is maximal, is taken to be the “optimal” cut-off point [35]. Importantly, the YI gives equal weight to false positive and false negative values, which means that it does not vary based on the context or aim of the test. For the purpose of this study, it is a useful index to provide consistency in our comparisons between disability domains, the CFM as a whole, and respondent types. For determining the best choice, or contextually “optimal”, cut-off level for Fiji’s MoE, the advantages and disadvantages of valuing sensitivity or specificity more highly are considered in depth in the Discussion.Throughout the paper, results related to parents as proxy respondents are denoted by a subscript P and those by teachers by a subscript T.For the domains without clinical reference standards in this study (self-care, anxiety, depression, controlling behavior, accepting changes to routine, and making friends), proportions of the sample reported as ≥ “some difficulty” and ≥ “a lot of difficulty” were compared. These two cut-off values were compared because the recommendation from the WG is to use “a lot of difficulty” [20,36] but previous results have raised concerns about the low sensitivity of this cut-off [13,21,23]. Also, a comparison of the clinical impairments of children identified at both cut-offs was undertaken, comparing “some difficulty” to ≥ “a lot of difficulty” on the CFM-13.Inter-rater reliability between parents and teachers was tested using a two-way random, absolute, average-measures intra-class correlation (ICC) [37]). Using Cicchetti’s classification [38], IRR interpretations were classified as: poor (<0.40), fair (0.40–0.59), good (0.60–0.74), and excellent (0.75–1.00). Cicchetti is slightly more generous than other classifications [39,40].Spearman’s rho correlation coefficient was used to test correlations between age, gender, and school type and CFM items, using the criteria: very high (0.90–1.00), high (0.70 < 0.90), moderate (0.50 < 0.70), low (0.30 < 0.50) and negligible (0.00 < 0.30) [41].Unless otherwise noted the two CFM questions on difficulty being understood when speaking by people: (1) inside the household, and (2) outside the household, have been combined as per the WG recommendation - to use the most severe response reported for either question [20].The sample included 472 children with mean ± SD age of 10.2 ± 2.6 years (range: 5 to 15 years) in Classes 1 to 8, including approximately half from special and half from mainstream schools (Table 1). There were 231 cases in the study and 241 controls, determined by clinical assessments. Cases included 35 children with vision impairment ranging from moderate vision impairment to total blindness, 60 children with hearing impairment from moderate hearing loss to profound deafness, 42 children with mild to severe mobility impairments, 71 children with moderate to severe speech impairment, and 125 children with moderate to severe cognitive impairment (Table S1). The mean age of cases was 10.15 years and controls was 9.71 years. Females made up 37.2% of cases and 51.0% of controls. Ninety-eight teachers participated, of whom 69% were female. Of the parents/guardians of the cases: 56% were mothers, 19% fathers, and 25% other (grandparent, aunty, uncle, guardian); the highest level of education was primary for 25%, secondary for 56% and higher education for 19%. Of the parents/guardians of the controls: 60% were mothers, 25% fathers, and 15% other; the highest level of education was primary for 22%, secondary for 63% and higher education for 15%.Table S2 presents values of area under the curve (AUC), sensitivity, specificity, the Youden Index for the optimal cut-off points and likelihood ratios from the construction of ROC curves. Table 3 provides a summary of key data from Table S2.Domain-specific results shown in Table 3 (eg., seeing, hearing) are discussed elsewhere and provided here to enable comparison with the overall CFM-7 result (see Table 2 for definition of CFM-7). In summary, the accuracy (AUC) of the CFM items on seeing, hearing, walking and speaking were higher than the items on learning, remembering and focusing attention. The lower levels for learning, remembering and focusing attention led to the CFM-7 as a whole having an AUC that was only “fair” (0.763P/ 0.786T); with slightly better overall accuracy by teachers. As shown in Table S2, levels of sensitivity were very consistent between parents and teachers across the cut-off levels, with “some difficulty” being 0.98P/0.96T, “a lot of difficulty” being 0.55P/0.57T, and “cannot do at all” being 0.23P/0.22T. Whilst teachers had higher specificity than parents at the cut-off “some difficulty” (0.33P/0.42T), results were more consistent at the higher levels; “a lot of difficulty” being 0.80P/0.82T, and “cannot do at all” being 0.99P/0.99T.Table 4 presents the spread of CFM-7 responses across impairment levels - none, mild, moderate and severe. Table S3 provides an extended presentation of Table 4 showing cross-tabulation of the highest level of severity of the child on any of the five reference standard results (vision, hearing, musculoskeletal, speech, cognition) with the highest level of difficulty reported for that child on any CFM-7 response.There was strong consistency between parent and teacher results in the overall proportions of children categorised as having “a lot of difficulty” (25.8%P/26.5%T) and “cannot do at all” (11.4%P/12.5%T). Parents reported slightly more children as having “some difficulty” (44.9P/39.3%T) and slightly fewer children as having “no difficulty” (17.6%P/21.7%T). Most moderate impairments are reported by parents and teachers as “some difficulty”. Severe impairments are reported approximately evenly across three CFM response categories: “some difficulty”, “a lot of difficulty” and “cannot do at all”. Most children with no impairments are mainly reported as having “no difficulty” (33.9%P/43.8%T), or “some difficulty” (47.4%P/39.1%T). However, a notable proportion (17.8%P/16.0%T) are reported as having “a lot of difficulty”, which is predominantly related to items on learning, remembering and focusing attention (as shown in Table S3). Children with mild impairments are mainly reported as having “some difficulty” (42.1%P/58.8%T) and “a lot of difficulty” (47.4%P/29.4%T).Problematically, the response category “some difficulty” includes children with a wide range of functioning. Of children with moderate clinical impairments, 52.4%P/47.3%T are reported as just having “some difficulty”, and of the children with severe impairments, 38.8%P/34.4%T are recorded as just “some difficulty”.Table 3 (and Table S2) show the YI for parent and teacher responses at the cut-off levels “some difficulty” and “a lot of difficulty” for each domain-specific question and for the CFM-7. For all seven domain-specific questions, the YI for the cut-off “some difficulty”, for both parent and teacher responses, is clearly higher than the YI for the cut-off “a lot of difficulty”. However, when considering the accuracy results for the CFM-7 (that is, the combined results), this is reversed and the cut-off “a lot of difficulty” is the highest.The positive likelihood ratio at the level of “some difficulty” is 1.46P/1.66T, compared to 2.78P/ 3.21T at the level “a lot of difficulty”. This means that the cut-off “some difficulty” provides a ‘minimal increase’ in the probability of the CFM-7 identifying disability in a child with disability compared to a child without. This is improved upon only somewhat by the cut-off “a lot of difficulty” which provides a ‘small increase’. The negative likelihood ratios for the overall CFM-7 at the cut-off “some difficulty” indicate a ‘large and often conclusive’ decrease in the likelihood that a negative result comes from a child with disability than from a child without disability. Whereas at the cut-off “a lot of difficulty” there is only a ‘small’ to ‘minimal decrease’ in this likelihood. These results should be interpreted cautiously though because the confidence intervals for the higher cut-offs were very wide due to small sample sizes.Table 5 summarizes the analysis of CFM domains that did not have clinical reference standard tests—self-care, anxiety, depression, controlling behaviour, accepting changes, and making friends. It highlights the proportion of responses for each domain at the level of at least “some difficulty” compared to at least “a lot of difficulty”. Parents and teachers reported a similar proportion having at least “some difficulty” with self-care (20.1%P/21.6%T), with “good” correlation between respondents (0.72). However, teachers reported a higher proportion having at least “a lot of difficulty” with self-care (2.3%P/6.2%T). Parents and teachers reported a similar proportion of the sample as feeling anxious or depressed “weekly”, but correlation was “negligible” (≤ to 0.26). Teachers reported a higher proportion of the sample as feeling anxious or depressed “daily”. Whilst data are not shown here, teacher responses showed a high correlation between learning and remembering (0.758), and depression and anxiety (0.729), and a moderate correlation between accepting changes to routine and focusing attention (0.546), self-care and walking (0.520), learning and being understood outside (0.511), focusing attention and learning (0.502), and accepting changes to routine and learning (0.502). Parent correlations for the same domains were far lower, ranging from 0.152–0.527.Overall, the proportions of children reported as “some difficulty” in the domains in Table 5 seem very high, but without a reference standard it is not possible to know whether this is reflective of disability.To further explore the rate of clinical impairments amongst children identified at the two cut-off levels (“some difficulty” and “a lot of difficulty”), Table 6 shows the frequencies of any impairment occurring amongst children reported as having “some difficulty” compared to ≥ “a lot of difficulty” on any question on the CFM-13. Table 7 is similar, but shows the frequencies of the individual impairments. As expected, with the larger number of questions on the CFM-13, slightly fewer children are missed compared to the CFM-7.Table 7 shows that children with moderate impairments that would be missed if the cut-off were “a lot of difficulty” are spread across all types of impairments, however it is the cognitive impairments that are missed more than other impairment domains.Using the “a lot of difficulty” cut-off, 39.7%P/33.3%T of the children with moderate impairments and 27.5%P/20.5%T of the children with severe impairments would be missed. Of all the types of impairment, those with moderate or severe cognitive impairment form the greatest proportion of children who would be missed if the cut-off were “a lot of difficulty”. These results do not indicate how many children with other impairments such as psychosocial or behavioural (which require other clinical assessments) may be missed.Inter-rater reliability between parents and teachers, assessed using ICC, varied considerably across disability domains as shown in Figure 1.For the overall CFM-13 it was 0.68 (95% CI 0.60–.73). The range of ICC was 0.22–0.82 across the individual domains. Domains with better ICC (0.61–0.82 were hearing, walking, speaking, self-care, seeing and learning. Domains with lower ICC (0.22–0.33) were anxiety, sadness, controlling behaviour, focusing attention and accepting changes to routine. Table 6 shows better correlations for overall categorisation of children with no impairment (0.61) and mild impairment (0.85) across the categories “some difficulty” and “a lot of difficulty”. However, correlations are worse for children with moderate impairment (0.06, not significant) and severe impairment (0.55). On the whole, correlations between teachers and parents were variable.This study identified that the CFM is a useful core aspect of data required for disability disaggregation of Fiji’s EMIS and that teachers are adequately accurate proxy respondents to the CFM. However, the mixture of severity of impairments reported across CFM response categories and ambiguity in the choice of cut-off level, in both parent and teacher results, are limitations of the CFM and indicate that the CFM may not be accurate enough to be used as the sole method for identifying children with disabilities.The first objective of this study was to determine the validity (sensitivity and specificity) of the CFM, which is operationally defined as the extent to which an overall score on the CFM at a given cut-off level identifies children who have an impairment as assessed using reference standard, or “gold standard”, clinical measures. For assessing sensitivity and specificity of the CFM, this paper effectively defines disability as clinically assessed impairment of a moderate or more severe level. There is debate about this medical perspective but for our purposes, it provides an objective assessment (in the sense of being made independently of those who stand to gain or lose from the assessment, or might perceive that they do), and so we have accepted it as the best available reference standard.Overall diagnostic accuracy (a combined value of sensitivity and specificity) of the CFM was found to be just “fair” based on combined results from seeing, hearing, walking, speaking, learning, remembering and focusing attention, i.e., CFM-7. This is substantially lower than the previously reported accuracy of individual domain-specific questions on speaking, walking, seeing and hearing [21,23], which are perhaps more observable functions. The cognitive domains had “fair” to “poor” accuracy (22). Given the variation in accuracy across the different domains in the module ranging from excellent to poor, it is not surprising that overall accuracy is only “fair”. This finding indicates that CFM-7 may not be accurate enough to be used as the sole method for identifying children with disabilities.Whilst diagnostic accuracy of parent observations related to seeing, walking and speaking is stronger than that of teachers, teacher accuracy is acceptable, ranging from “good” to “very good” (between 0.823–0.909). Conversely, for the domains learning, remembering and focusing attention, teacher results are stronger than parent results. For hearing, the accuracy is high and very similar between respondent types.To disaggregate Fiji’s EMIS by disability, it is important to identify the appropriate cut-off level of the CFM. The field testing of CFM as part of population-based surveys in Samoa, Mexico and Serbia showed that the “some difficulty” cut-off estimates a very high prevalence compared to the “a lot of difficulty” cut-off [15]. The cut-off recommended by UNICEF/ Washington Group is “a lot of difficulty” [20]. However, in our study a significant proportion of children with moderate or higher clinical impairment were reported as having only “some difficulty” on CFM-7, comprising seeing, hearing, walking, speaking, learning, remembering and focusing attention domains (Table 3). These children would therefore miss out on services if the cut-off were “a lot of difficulty”. Based just on these domains, approximately half of children with moderate clinical impairments (52.4%P/47.3%T) and a third of children with severe impairments (38.8%P/34.4%T) would miss out on services if the cut-off level were “a lot of difficulty”. However, when CFM-13 was considered (which includes the additional 6 questions), not surprisingly the chance of missing children is reduced, and the proportions were reduced to some extent. Despite this, 39.7%P/33.3%T of children with moderate clinical impairments and 27.5%P/20.5%T of children with severe impairments would be missed. When domain-specific findings are considered, it is the children with moderate-severe cognitive impairments who miss out in greatest numbers [21,22,23]. The decision to select a cut-off must also consider the fact that 47.8%P/39.1%T of children with no clinical impairment are reported as having “some difficulty”. Our findings indicate that children reported as having “some difficulty” can neither be ignored nor be assumed to have disability.The cross-tabulation also highlights the fact that the three CFM response categories—“some difficulty”, “a lot of difficulty” and “cannot do at all”—do not relate to the same levels of severity across different functioning domains. This is in contrast with the recommendations on the interpretation of these categories by UNICEF/Washington Group [20] and USAID [17]. Whilst most moderate impairments are reported as “some difficulty”, children with severe impairments are showing up relatively evenly across the three response categories, and the response categories do not have the same meaning across different domains. For example, the category “cannot do at all” picks up a large proportion of children with severe musculoskeletal impairment yet it picks up only approximately 2% of children with severe cognitive impairment. This extreme response category is used to a small extent for questions on hearing, walking, speaking and seeing, but almost never used for questions on learning, remembering and focusing attention.The CFM is described as being able “to determine the proportion of those who have mild difficulties (at least some difficulty on one or more domains of functioning), or moderate levels of difficulty (those who respond at least a lot of difficulty) or those with severe difficulties (those who respond cannot do at all)” [36] (p. 487). However, our findings suggest that this interpretation of the CFM response categories across disability domains would not work in Fiji. Mitra emphasised the value of using a “trichotomy” (severe, moderate and no difficulty), in which classification of people with moderate functional difficulty was based on “some difficulty” in at least one domain with no higher levels of difficulty recorded [43]. This is consistent with our finding that the cut-off “some difficulty” included most of our children with moderate impairments, however the challenge remains that many children without impairments were also recorded as having “some difficulty”.The ROC curve results from earlier reports were complicated and varied across domains and methods, including sensitivity, specificity, the Youden Index and likelihood ratios. For the domains seeing, hearing, walking and speaking, “some difficulty” was a far more accurate cut-off than other levels [21,23]. The cognitive domains learning, remembering and focusing attention also indicate the cut-off “some difficulty” as the best, with teacher results being superior to parents at identifying children with cognitive impairments [22].However, contrary to the individual domain-specific results, the diagnostic accuracy results for the CFM-7 showed “a lot of difficulty” as the best cut-off, albeit only marginally better. This is because at “some difficulty” sensitivity is excellent (0.98P/0.96T) but specificity is very poor (0.33P/0.42T). At the cut-off “a lot of difficulty” specificity was much better (0.80P/0.82T) but sensitivity dropped significantly (0.55P/0.57T). Notably, the Youden Index for the overall CFM was quite low at either cut-off (0.31P/0.40T for “some difficulty” and 0.36P/0.39T “a lot of difficulty”). This was not surprising given the disappointing diagnostic accuracy of the CFM-7 as only “fair”. These results further highlight an important shortcoming in diagnostic accuracy of the CFM-7: there is no clear and strong cut-off response category for the overall CFM and the cut-off which performs best for individual functional domains is different from that for the overall module.The high proportion of children reported as having “some difficulty” on the six domains without a clinical reference standard highlights the need for further research to understand the impact of the cut-off level on identifying children with difficulties in these domains.The second objective was to determine the inter-rater reliability between teacher and parent CFM responses. Our study showed that IRR of the CFM-13 is “good” (0.68), which in theory contributes to the case that the CFM can be used with teachers as respondents. However, there is great variation in IRR across domains [21,22,23]. The potentially more observable domains (hearing, walking and speaking) have “excellent” IRR followed by “good” IRR for self-care, seeing and learning.However, IRR needs to be considered in relation to accuracy. For example, if both respondents are equally “wrong”, the IRR may be high but this does not mean the tool is useful. Or, if parent responses are “wrong”, a low IRR could be positively interpreted in terms of teacher use of the tool. Considering accuracy together with IRR between parents and teachers, the most accurate and reliable CFM questions relate to the domains of seeing, hearing, walking and speaking. Of the CFM questions for which this study does not have clinical reference standards (and therefore no diagnostic accuracy analysis)—self-care, anxiety, sadness, controlling behaviour, accepting changes and making friends—it is harder to interpret the largely poor IRR results. This may reflect poorly on the questions or may imply varying perspectives and accuracy between parents and teachers; teachers may be in a better position to make a relative judgment for some of these items. The higher correlations between teacher results for domains which might be expected (anxiety and depression; learning and remembering; changes to routine and focusing attention) provide some indication that teachers are observing these functional domains more consistently than parents and that teacher results may be more accurate in these domains. In relation to anxiety and depression, the results highlight a potentially important role for teachers in Fiji in identifying children at risk of psychosocial distress. These issues both point to important areas for future research. Research is required to investigate parent and teacher response accuracy for these domains.Fiji’s MoE has committed to provide inclusive education in a way which leaves no one behind [44] and following this study commenced disability inclusion grants to schools, calculated by number of children with disabilities. Messick [45] and Shepard [46] championed the importance of undertaking “consequential validity”, or investigation and prediction of positive and negative social consequences of a test. The implication of Fiji’s policy, in relation to this study, is that if a cut-off level has a low sensitivity it misses out eligible children, which would be the case if “a lot of difficulty” were used. Hence to ensure children are not missed the cut-off “some difficulty” must be used. However, given the significant proportion of children classified as “some difficulty” who do not have disability, follow-up assessments are required to verify presence of disability (and to identify children for whom referral services are required).Conversely the low specificity of the “some difficulty” cut-off has cost implications regarding verification visits. Travelling to remote areas to assess children simply based on a self-reported “some difficulty” response would be cost-prohibitive and an inefficient use of already stretched MoE staff time. A solution to this challenge may be found in another series of results from the study, to be discussed in a subsequent paper, showing that the combination of CFM data and learning and support needs data enables a much more accurate estimation of disability. This would reduce false positives on the list of children who need verification visits.An essential feature of the CFM to highlight, in relation to assessing disability for funding eligibility, is the self-report nature of the tool. Whether the respondent is a parent/caregiver or a teacher, the results can be biased if there is perceived financial advantage in reporting higher levels of difficulty. The disability verification visit is necessary to pre-empt over-reporting. These visits involve qualified MoE district officers visiting the schools to discuss the results with teachers and undertake basic tests with the identified children, such as visual acuity tests (Snellen chart), observations of gross and fine motor function, classroom observation, review of student records, etc. The visit offers the chance for monitoring and mentoring of efforts towards disability-inclusive education.An important limitation common to all diagnostic accuracy studies is the assumption that the clinical assessment standards are 100% sensitive and specific themselves. That is, that the tests for vision, hearing, musculoskeletal impairment, speech and cognition are indeed “gold standards” against which the CFM can be measured. The justification for selection of the five clinical assessments along with measures to ensure accuracy of the tests and to reduce classification bias [47] have been presented in detail elsewhere [21,22,23] and is summarised in Appendix B.The five clinical assessments did not cover all the functioning constructs that are covered by the whole CFM (the CFM-13), specifically self-care, anxiety/worry, depression/sadness, behaviour and socialisation. We attempted to overcome this limitation by making interpretations based on IRR and simple proportions reported in different severity levels of the CFM-13. However, an outstanding recommendation for further research is for a diagnostic accuracy study which adequately covers these constructs.A relatively high proportion of cases were from special schools (76.2%) due to the limited numbers of children with disabilities in mainstream schools. To achieve the required sample size across all five impairment groups, recruitment had to allow for this imbalance. Despite this, the target sample of 52 in each clinical impairment category was not reached for children with vision impairments (n = 35) and musculoskeletal impairments (n = 42). Future research should aim to rectify this sampling disparity and shortfall.An important limitation relates to generalizing the findings to other populations. Of the parents/caregivers of the cases, 19% had attained a tertiary education, which is higher than the national average [48]. The level amongst controls was 15%, which is closer to average. This highlights potential differences related to parents of children in special schools, but importantly raises the question of difference between parents of children with disabilities in school compared to those who are out of school. Future research should include out-of-school children with disabilities, whose parents may respond differently to the CFM questions.Another limitation is that 62.8% of cases were male compared to 49.0% of controls and the mean age of cases was 10.15 years compared to 9.71 years amongst controls. However, correlations between age, sex and the CFM questions were explored, and the impact of these variations appears to be negligible. Age had significant but negligible correlation with the domains learning (0.164), remembering (0.118) and depression (0.097). Sex had significant but negligible correlation with the domains speaking (0.092), learning (0.144), controlling behaviour (0.156), focusing attention (0.096) and making friends (0.097).Finally, the authors acknowledge the limitations of categorizing IRR values into the classifications “excellent/good/fair/poor” because it is dependent on the purpose for which the test is to be used. For the purpose of this study however, the categories provide a convenient means of comparing individual domains and the overall CFM-13.The UNICEF/WG Child Functioning Module is an important new instrument for disability disaggregation of datasets particularly considering the urgency to collect baseline information for the SDGs. When evaluated as a whole it achieved only a “fair” level of accuracy to identify children with disabilities in Fiji. This contrasts with earlier domain-specific findings which showed “good” to “excellent” accuracy for seeing, hearing, walking and speaking.The choice of cut-off level and the mixture of severity of impairments reported across response categories are particular challenges for the CFM. Specifically, the response category “some difficulty” includes children with severe impairments as well as children with no impairments, with uneven results across disability domains. In the context of Fiji’s education system, children reported as having “some difficulty” can neither be ignored nor be assumed to have disability. There is no clear and strong cut-off response category for the overall CFM and the cut-off which performs best for individual functional domains is different from that for the overall module. While the CFM provides useful data for Fiji’s EMIS, the CFM is not accurate enough on its own for identifying children with disability for the purpose of determining funding eligibility.We recommend that children with disabilities are identified using CFM plus additional data on learning and support needs and that verification visits are undertaken to confirm funding eligibility.The following are available online at https://www.mdpi.com/1660-4601/16/5/806/s1, Figure S1: Flowchart of participation, Table S1: Clinical characteristics of the study sample, Table S2: Extended data for Table 3—Diagnostic accuracy of the CFM-7 compared to five reference standard assessments, parent versus teacher responses, at different cut-off levels, Table S3: Extended data for Table 4—Cross-tabulation: Child Functioning Module results (CFM-7) by the results of the reference standard tests for vision, hearing, musculoskeletal, speech and cognition.Conceptualization, B.S.; methodology, B.S. and M.M.; software, B.S.; validation, B.S. and M.M.; formal analysis, B.S.; investigation, B.S.; resources, B.S.; data curation, B.S.; writing—original draft preparation, B.S.; writing—review and editing, B.S., M.M. and B.M.; visualization, B.S.; supervision, M.M. and B.M.; project administration, B.S.; funding acquisition, B.S.This research was part-funded by Department of Foreign Affairs and Trade, Australian Government, grant number 66440.The authors wish to sincerely thank the staff, families and students of the many schools involved in the study, partner agencies Pacific Disability Forum and the Pacific Islands Forum Secretariat, and the Access to Quality Education Program.The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results. The views expressed herein are those of the authors and not necessarily those of the Commonwealth of Australia. The Commonwealth of Australia accepts no responsibility for any loss, damage or injury resulting from reliance on any of the information or views contained in this publication.(1) The draft version did not include reference to contact lenses, which is now in the final version, as shown in italics in the question: When wearing his/her glasses or contact lenses, does (name) have difficulty seeing?(2) The word “focusing” has been replaced by “concentrating” in the final version.“Does (name) have difficulty concentrating on an activity that he/she enjoys doing?”(3) The question on difficulty controlling behaviour has been changed from the draft version used in this study:“Compared with children of the same age, how much difficulty does (name) have controlling his/her behaviour?” with response options: No difficulty/The same or less/More/A lot moreTo the final version: “Compared with children of the same age, does (name) have difficulty controlling his/her behaviour?” with response options: No difficulty/Some difficulty/A lot of difficulty/Cannot do at all.(4) The word “very” has been inserted in the questions on anxiety and depression in the final version.“How often does (name) seem very anxious, nervous or worried?”“How often does (name) seem very sad or depressed?”(5) The sequence of the last 8 questions has been changed in the final version.In the draft version the CFM questions were in this sequence: learning, remembering, anxiety, depression, controlling behaviour, focusing attention, accepting changes in routine, and making friends.In the final version the CFM questions are in this sequence: learning, remembering, concentrating (formerly focusing attention), accepting changes in routine, controlling behaviour, making friends, anxiety and depression.The reference standard (clinical) tests for this study were selected based on international standards for vision and hearing and well validated tools for speech, musculoskeletal impairment and cognitive impairment. The clinical team consisted of trained vision and hearing technicians and physiotherapists.Vision assessment was performed with torchlight examination, visual acuity with Snellen chart, pinhole testing and refraction using a Topcon autorefractor. The following levels of vision impairment were included as cases: presenting visual acuity in the better eye <6/18 and ≥6/60 (moderate), <6/60 and ≥3/60 (severe) and <3/60 (blind) [21].Hearing assessment was performed by observation with otoscope and air conduction audiometer. The pure-tone audiometry values for four frequencies in each ear, including 0.5, 1, 2 and 4 kHz, were averaged and the threshold level of the better ear was used to determine cut-off for cases. The following levels of hearing loss were included as cases: 41–60 dBA (moderate), 61–80 dBA (severe) and ≥81 dBA (profound). Greater than 30 or 31 dBA is commonly used as a criterion for hearing impairment in children [49,50], however >40 dBA was used in this study to identify children with clinically relevant hearing impairment due to the ambient noise levels in the assessment rooms in the schools. This is consistent with the extensive prior experience of the hearing assessors in Fiji and with other studies in developing countries [51,52]. Children found to have impacted wax or foreign bodies in the ear had this removed and were tested for hearing after removal.Musculoskeletal assessment was undertaken using the Rapid Assessment of Musculoskeletal Impairment (RAMI) [30]. Through consultations with the Ministry of Health senior physiotherapist, it was established that there is no standard assessment used or validated for children of this age group for Fiji. Based on a literature review of assessment tools, the RAMI was deemed to be the best available method for establishing presence or absence of mobility impairments in this study setting [53]. The RAMI includes an initial set of five questions, such as, “Do you have any difficulty using your legs?”, with corresponding questions about duration indicating that it has lasted more than one month or is permanent. This is followed by observation of a series of gross and fine motor activities. In children where one or more of the five questions was answered positively, and one or more of the duration questions was “Yes”, and one or more of the observations indicated difficulty with the activities, children were assessed further for the extent of the effect on the musculoskeletal system. The RAMI does not consider functioning with equipment. Children identified on the RAMI to have impairment only affecting the upper limb were excluded for this analysis on walking difficulty. Children identified on the RAMI with structure impairment including “severe”, “moderate” and “mild” effect on the musculoskeletal system’s ability to function as a whole were identified as cases with mobility impairment [30].Speech was assessed by administering the Intelligibility in Context Scale (ICS) [31] to parents. The ICS was selected as the tool to identify children with speech difficulties for several reasons: at time of data collection, there were no speech-language pathology services in Fiji and no speech assessment tools developed or validated in Fiji [54]. It can be administered by non-specialists. It can be used irrespective of language or number of languages spoken by the child [55,56], which is important in Fiji where many people are multilingual [57]. It assesses intelligibility and comprehensibility, which are comparable constructs to CFM questions on difficulty being understood when speaking. The ICS had already been rigorously translated into Fijian and Fiji-Hindi and has been widely used both with children with speech sound disorders [31,58] and with typically developing speech [32,58,59]. For our study, case definition for speech difficulties were ICS scores: 1.8 to <2.5 (moderate) and 1.0 to <1.8 (severe).Cognitive impairment was assessed using the Cambridge Neuropsychological Test Automated Battery (CANTAB) [32] and cases included subjects with CANTAB Overall Impairment Scores of 3 (moderate) and 4–5 (severe). CANTAB, designed to be non-linguistic and culturally independent, has been validated with children to assess a range of cognitive functions [32,60,61,62] and has been used with children in a range of settings globally including where English is not the first language [63,64]. Five sub-tests, recommended by Cambridge Cognition to provide an overall assessment of cognitive function, were implemented in this order: Motor screening (MOT), Paired Associates Learning (PAL), Spatial Working Memory (SWM), Stockings of Cambridge (SOC) and Reaction Time (RTI).Inter-rater reliability between parents and teachers of the overall CFM (CFM-13) and of individual domains.Demographic characteristics of the study sample.* Other: grandparent, aunty, uncle, guardian.CFM domains, question wording and response categories; coded to indicate which group of domains was used in the various analyses in this paper.No difficultySome difficultyA lot of difficultyCannot do at allPeople inside this householdPeople outside this householdDailyWeeklyMonthlyA few times a yearNeverNo difficultyThe same or lessMoreA lot more** The CFM includes questions to establish whether the child wears glasses, uses a hearing aid, or uses any equipment or receives assistance for walking. If the child does use the assistive device, the question for seeing is “When wearing his/her glasses, does (name) have difficulty seeing?” Similar questions are asked for hearing and walking. The CFM has separate questions for difficulty walking with and without equipment for children who need equipment. Analysis for this paper includes: difficulty walking for children who do not need equipment, plus those who require equipment but have difficulty walking without their equipment (this allows comparison with the Rapid Assessment of Musculoskeletal Impairment which tests function without equipment).Diagnostic accuracy of the Child Functioning Module (CFM-7); parent versus teacher responses, comparing two cut-off levels: “some difficulty” to “a lot of difficulty”.Cross-tabulation: Child Functioning Module results (CFM-7) by level of impairment.* Child is recorded in the highest level of difficulty from any of the CFM-7 questions on seeing, hearing, walking, being understood when speaking, learning, remembering and focusing attention, and in the highest level of severity from any of the five reference standard assessments for vision, hearing, musculoskeletal, speech and cognitive impairment.Proportion endorsing each domain at the cut-off level “some difficulty” compared to “a lot of difficulty”, and inter-rater reliability between parents versus teachers.ICC = Intraclass correlation; Ω = more difficulty and a lot more difficulty.Frequencies of any impairment occurring amongst children reported as having a highest level of difficulty of at least “some difficulty” compared to at least “a lot of difficulty” on any question on the CFM (CFM-13), comparing parent and teacher responses.Frequencies of five types of impairment occurring amongst children reported as having a highest level of difficulty of at least “some difficulty” compared to at least “a lot of difficulty” on any question on the CFM (CFM-13), comparing parent and teacher responses.VI = Vision impairment; HI=Hearing impairment; MSI = Musculoskeletal impairment (mobility only); ¥ Visual Acuity of better eye; NPL—no perception of light; CF2m—counting fingers at 2metres. ^ Severity for the Rapid Assessment of Musculoskeletal Impairment was determined using the parameters for the percentage of function outlined in the International Classification of Functioning (ICF) [42]. Percentage loss of the musculoskeletal systems ability to function as a whole. ₱ Intelligibility in Context Scale—scores between 1.0–2.43 (detailed in [23]. For this paper, severe vision impairment and blindness are combined in one category and severe and profound hearing impairment are combined in one category. Results with these severities separately reported is available in [21].

Med-MDPI/ijerph_3/ijerph-16-05-00807.txt

@@ -0,0 +1 @@
+Diffusion of cholera and other diarrheal diseases in an informal settlement is a product of multiple behavioral, environmental and spatial risk factors. One of the most important components is the spatial interconnections among water points, drainage ditches, toilets and the intervening environment. This risk is also longitudinal and variable as water points fluctuate in relation to bacterial contamination. In this paper we consider part of this micro space complexity for three informal settlements in Port au Prince, Haiti. We expand on more typical epidemiological analysis of fecal coliforms at water points, drainage ditches and ocean sites by considering the importance of single point location fluctuation coupled with recording micro-space environmental conditions around each sample site. Results show that spatial variation in enteric disease risk occurs within neighborhoods, and that while certain trends are evident, the degree of individual site fluctuation should question the utility of both cross-sectional and more aggregate analysis. Various factors increase the counts of fecal coliform present, including the type of water point, how water was stored at that water point, and the proximity of the water point to local drainage. Some locations fluctuated considerably between being safe and unsafe on a monthly basis. Next steps to form a more comprehensive contextualized understanding of enteric disease risk in these environments should include the addition of behavioral factors and local insight.Monitoring the water supply for diarrheal pathogens in informal settlements (IS) is imperative to improve the health of residents while also being imperative in achieving the targets set out in the United Nations’ Millennium Development Goals [1,2,3,4]. From the research perspective, the need is to understand how, why and when water contamination occurs [5]. While traditional epidemiological data analysis will always be important, this is a complex water-risk landscape comprised of spatial, temporal and behavioral factors that might not so easily be reduced for analysis. Which water points become contaminated and when does this occur, and then how does the environmental configuration, meteorological changes and human action lead to this fluctuation are all important questions. A more nuanced contextualized reading of this complex landscape could lead to both improved daily outcomes, and the ability to respond in the face of a major outbreak. Basic questions that need to be asked include which water points are the most vulnerable to contamination, and does this risk change over time? Even supposedly official “clean” city water piped underground can become contaminated through human activities, such as illegal tapping [6], while reservoirs, irrespective of the cleanliness of the water supply, can become contaminated by having dirty containers lowered into them. Emerging cases of cholera and diarrheal diseases are not only point-in-space outcomes, but also inputs into a complex interconnected system of water access, sanitation and drainage. While conceptually we understand these relationships, we are impeded by a lack of data that reflect the temporally dynamic and spatially granular complexity of the Water, Sanitation, and Hygiene (WASH) [7,8,9] landscape within an informal settlement. Simply put, case locations, water point and drainage testing, fine scale environmental layers, and localized metrological data are all important, but all pose considerable logistical challenges in collection at the scale of intervention, meaning at a granular scale that can support effective community or household intervention. Other contributing factors such as human intra settlement mobility, activity space and context are similarly absent. The goal then is to develop an understanding that includes dynamic epidemiological data collection nuanced by behavioral and/or environmental influences. Ideally, regular water point testing should be supplemented with surveys that capture localized environmental change. Data to be collected would be the setting of each water point with regards to local drainage, standing water and mud, and the condition of the ground at the actual water access point. In this paper, we advance this perspective by focusing on the longitudinal testing of water points for fecal coliforms and V. cholerae, with a simultaneous micro-environmental survey. Our aim is to not just analyze spatial and temporal patterns in fecal coliforms but to also develop one of the first contextualized readings of this complex epidemiological, environmental and behavioral system. Our study locations are a coastal informal settlement (IS) in Port au Prince, Haiti. The geographic pattern of enteric diseases in an IS is related to the fecal-oral means of transmission [10]. Simply put, household water supplies become contaminated with a diarrheal pathogen. The way this happens is complex and multi-scalar. At the coarsest scale of impact climate change leading to sea level rise, heat/increased water temperature, and rainfall can lead to outbreaks of diarrheal diseases, including cholera [11,12]. However, these broader externalities work through localized environmental configurations. For example, flooding or overland flow from drainage ditches containing waste water might contaminate clean water reservoirs during rainy seasons or after intense rain events (such as a hurricane) [13,14,15], but how that water is stored and accessed either protects against or facilitates that contamination. In both Peru and Bangladesh, during strong El Nino years, cholera rates rose in sync with local sea surface temperature [16,17]. In Ecuador, diarrhea rates were found to be positively correlated with seasonal extreme rainfall events [18]. More specifically for this paper, hospital admissions for diarrhea-related illnesses peak in association with seasonal increases in precipitation for the informal settlements of Port au Prince [19]. Yet an IS is not a homogenous space; even with such externalities there is localized variation in disease risk. Where is the water point located (does the surrounding relief contribute to area flooding), how is the water dispensed (is the source a container sunk into the ground with no seal or other barrier), how close are drainage channels or toilets, and is there standing water or mud around the water point where people rest their water containers? Adding further complexity is that these risks are also temporal, as water and mud from cumulative rain events can create a breeding ground suitable for propagating pathogenic bacteria [20]. It is therefore important to develop a more nuanced appreciation of this space in addition to more traditional hotspot and pattern analyses. In this paper we explore the interlinked pattern of water point fecal coliform contamination and micro environments for IS neighborhoods in Port-au-Prince, Haiti, that were originally refugee camps after the earthquake of 2010 [21]. In times of such upheaval enteric disease often occurs, and while the current cholera situation in Haiti (as of 2018) is not as severe as that in Yemen where deteriorating hygiene, sanitation systems and water shortages are contributing to a massive human toll [22,23,24], parallels can be drawn due to the human displacement and disruption to clean water and effective sanitation practices that occurred after the earthquake [25,26]. Cholera is now endemic in the region, and knowing how and where disease risk is greatest, especially within the Port au Prince IS, can help spatially prioritize intervention during future outbreaks [27].In response to the post-earthquake cholera epidemic, in 2012 Haitian Group for the Study of Kaposi’s Sarcoma and Opportunistic Infections (GHESKIO) organized a cholera vaccination campaign that covered 70,000 people in six IS of Port au Prince [28] (The work described in this paper is part of a partnership with the University of Florida Emerging Pathogens Institute in Gainesville, Florida and GHESKIO which is one of the largest non-profit healthcare providers in the Caribbean. GHESKIO was founded in 1982 in response to the AIDS epidemic, and is now the largest provider of HIV/AIDS and tuberculosis (TB) services in the Caribbean. GHESKIO set up a “tent city” to house over 7000 refugees, many from the three neighborhoods (for ease of description referred to as EA, EC and EB), after the earthquake in 2010, providing shelter, security, food, clean water, health services and educational opportunities. GHESKIO has continued to provide health and social services to the study neighborhoods, and has also trained over 800 community members as “Health Agents” to improve sanitation and access to clean water, and to survey for infectious diseases, malnutrition, and to refer patients to the GHESKIO clinic. During the cholera epidemic the group established an emergency cholera treatment center, and developed comprehensive disease reduction strategies including providing chlorinated water, building of latrines, and establishing rehydration posts, a 250-bed tent hospital, and 10 community health posts). Two of these communities, along with a third unvaccinated neighborhood, are the setting of the study described here. Each of the three study areas are coastal, peri-urban, and to some degree have been reclaimed from the sea through the process of trash dumping and compaction. Each neighborhood also has uneven access to clean water and sanitation systems [19]. The environmental and coliform bacterial data collection described in this paper is part of an initiative to compare and contrast the long-term effectiveness of the original vaccination campaign, while at the same time providing a more detailed understanding of ongoing diarrheal disease risk in the neighborhoods [19]. The three neighborhoods have “typical” informal settlements characteristics for the region with structures ranging from concrete block buildings to homes made of corrugated metal and other scavenged materials (Figure 1). Transportation paths are mostly dirt streets or narrow winding alleys. There is virtually no legal built environment infrastructure (such as sewage disposal, electricity, piped water, trash collection), while water is purchased from vendors at a variety of different access points.Figure 1 contains three inset images from the study neighborhoods named as EA, EC, EB; 1A shows the coastal area EC with the field team walking through accumulated trash piles which often contain fecal matter to take an ocean sample. 1B is water testing in the main drainage channel that separates the neighborhoods EA and EB. The buildings in the background of 1B (located in EC) are a combination of brick/concrete and metal sheet buildings while 1C displays a typical residence found in neighborhood EA. It should be noted that the purpose of this paper is not to focus on the neighborhoods as different units, indeed EA and EB are only separated by the large Bois de Chêne canal. The naming of EA and EB mirrors local operational logistics and is used here for convenience. Instead we treat each testing location by type and as input into multi-scalar complex landscape, one in which variations can occur over relatively small distances. Drainage in the study region is comprised of open channels of varying sizes often full of trash, refuse, human waste and scavenging animals. As can be seen in Figure 2A, which shows one such channel, a risk scenario often described in the literature are children playing in mud or water contaminated with fecal matter [29]. These channels either flow into progressively larger drains, such as the one seen in Figure 1B, or directly into the sea. It should be noted, however, that many of these drains do not have a planned structure to them, but instead are where water flows finding the least resistance. At the sea edge, most notably in neighborhood EC, the land has been pushed out into the water, with the constant dumping of trash forming an unstable surface on which concrete buildings have been built. This can be seen in the background of 1A with new homes being constructed on the trash. The location of this photo is actually several meters into the open ocean on the map of Figure 1 due to the overhead imagery being from 2010.Figure 2B shows a water point operator filling a customer’s bucket. In the background, one of the project team is testing a water sample taken from the cistern. Water point type varies in the neighborhoods. In Figure 2B an operator distributes from an above-ground reservoir, while elsewhere a concrete cistern or plastic drum reservoir might be set into the ground. For water points without a tap, the customer drops his/her bucket into the storage container, or uses a plastic vessel tethered to the water point. Most of these reservoir-type water points receive water from city trucks. In EC, water points are more frequently stand pipes that have been illegally tapped into the city’s water supply. To explore longitudinal patterns of contamination of these water points, we pose the following questions: Is there a space-time variation in household water contamination risk, as assessed by fecal coliform (FC) counts, between testing locations?Is there a water point type and location that pose the greatest risk to public safety?How can micro environmental contextual factors explain the results to the above questions?Is there a space-time variation in household water contamination risk, as assessed by fecal coliform (FC) counts, between testing locations?Is there a water point type and location that pose the greatest risk to public safety?How can micro environmental contextual factors explain the results to the above questions?Beginning in October 2016, water samples were collected monthly from thirty-eight locations in three IS neighborhoods of Port au Prince. Of eighteen sample sites selected in the EA neighborhoods, ten water points were used for drinking, and/or washing/bathing, one was used for washing/bathing only, four were open drainage channels, and three were ocean collection sites. EB had six sample sites, all of which are a source for drinking and washing water. The EA and EB neighborhoods were only separated by a single large drainage channel. Approximately three kilometers west, the unvaccinated EC neighborhood had fourteen sample locations including six stand pipes illegally connected into the city’s underground water supply, all of which are used for drinking and washing. Five sample locations are drainage channels and the remaining three are from the ocean boundary. A team of researchers visited each site to draw water samples which were returned to GHESKIO. Each collected water sample was tested in the field for temperature, turbidity, dissolved solids, pH, and dissolved oxygen, before the sample underwent laboratory testing for FC counts using the membrane fecal coliform (mFC) agar method [15]. Briefly, for the FC assay, we collected 500 mL water in a sterile Nalgene (MilliporeSigma, Burlington, MA, USA) bottle from each sampling site. One hundred mL (100 mL) of directly collected water, and/or, as needed, ten-fold dilutions of the water in phosphate buffered saline (PBS, pH 7.2) [vol/vol] to a final volume of 100 mL were passed through a 0.22 µM filter (MilliporeSigma, Burlington, MA, USA) using vacuum-driven force. The filter was then aseptically transferred onto an mFC agar (MilliporeSigma, Burlington, MA, USA) and the culture plate was incubated overnight at 44.5 ± 0.2 °C. The number of blue bacterial colonies grown on mFC agar were counted and the results were based on the average of triplicate mFC culture plates for each independent water sample tested. mFC culture plates exhibiting no coliform colonies were considered to have <1 cfu/100 mL sample water. As a negative control, an aliquot of 100 mL PBS was processed for FC using mFC agar during each independent experiment.Starting in February 2017, in addition to the water samples, micro environmental surveys were also collected by the field team around each sample site using Contour +2 and Patrol Eye Body cameras. These spatial video (SV) cameras have an internal global positioning system (GPS) receiver, which means the videoed environment around each testing site can be subsequently viewed for risks using CameraPlayer, software the team has developed to simultaneously locate each image frame on a map [30]. After each monthly collection, video and GPS paths were downloaded and meta data sheets completed, which recorded the technical performance of each camera. A combination of the video and overhead aerial imagery was used to verify the accuracy of the GPS path and correct any errors using bespoke software designed by the team for these types of environments [30] (Figure 3). Then, each located video frame became a digitizing source, allowing the team to map “risk” features into Google Earth or as a Geographic Information System (GIS) layer. These risk factors include standing water, mud, trash, and human activity [31,32,33,34]. After the initial SV visit, a map was created for each water point depicting its immediate environmental setting including its proximity to other features such as drainage channels. Once an initial environmental assessment had been made of each test location, each subsequent visit was compared for change. This was achieved using an environmental assessment index created with scores ranging from one (least severe risk) to ten (most severe risk) for standing water, mud, trash, and the amount of human activity within a ten meter buffer of the test site. These environmental assessments follow similar methods applied in Haiti, Nicaragua, and Tanzania [30,31,32,33]. This type of mapping of micro-environmental characteristics has been successfully used to track and predict the distribution of both waterborne and water-vectored diseases [35]. While there is always a degree of subjectivity in assigning an environmental score (for example a trash value of 10 for Figure 1A, and 8 for Figure 1B), the same two-person team involving the main coder, and then a project manager overseeing the process with random checks, was used for all sites. This, as much as possible, helped ensure standardization in coding.The FC counts for each test site, and the associated environmental index were then mapped using ArcGIS 10.4 (ESRI 2011, ArcGIS Desktop, Redlands, CA, USA) and tabulated for exploratory data analysis. Water sampling at thirty-eight test sites began in October 2016 and SV environmental assessments commenced in February 2017. In order to tease out the spatial and temporal variation between sites and the implications these results have for neighborhood residents, the results of the monthly data collection trips are presented in two ways: firstly as a FC comparative table for just household water points and then as a combined FC—environmental assessment. Table 1 displays the results of a ranked FC summary for 22 water points, and two environmental sites (EC 11 and EA 16) which are included for comparative purposes; EC 11 is the juncture of a drainage channel entering the ocean (Figure 4A), and EA 16 is a concrete holding point sourced with ocean water which is also used by the community for household purposes (Figure 4B). All things being equal, it is expected that EC 11 and EA 16 should score the highest FC counts and so they are included as a benchmark of risk by which the other household water points can be judged. Table 1 also includes the water point type and source. Table 1 ranks each water point according to its FC count as compared to the other 24 tested locations for that month. Therefore, it expected that EC 11 and EA 16 should score 24 and 23 (the highest possible ranks) for every month. These rankings are also summed for the entire testing period to show the relative stability in risk for each test site. An alternative way to visualize the risks can be seen in Figure 5 which graphs the FC counts (log transformed) for each of the test sites in the EA neighborhood. No obvious neighborhood-wide pattern is evident, apart from the elevated risk for April and May, which corresponds to the rainy season. However, the same risk is not apparent for the second rainy season (September and October). Indeed, this graph shows that there is considerable fluctuation across all sites, which means that more aggregate “findings” could lose the importance of more localized factors. The FC counts were then compared against the environmental contamination scores for each month. Each waterpoint was graphed to include the scores for each contamination category, and the FC count (logged to allow for more easy comparison). Figure 6 provides an example for EB 4.Using all these outputs, the following profiles were developed for overall IS area. While the EC neighborhood was generally considered safe, the one exception was EC 13 which rose from a low monthly FC count to over 35 million cfu/100 mL (colony forming units per 100 milliliters of collected water) (Table 1). The area around the water point had high levels of standing water and moderately high levels of mud, though not dissimilar from conditions at the other water points in the neighborhood. Indeed, even when FC counts at EC 13 returned to zero the environmental contamination risks persisted for another two months. The FC counts in the EB neighborhood were higher than EC, especially EB 1 through 5. Most of these water points are located next to a busy road. Four of the six sites are cistern-style vending points that involve dipping a bucket into the reservoir. Interestingly, over the project period EB 4 changed from a ground-level bucket-retrieval style cistern, to a piped tap by June that connects to the underground city water system. Figure 6 displays the impact of this change with FC counts dropping dramatically by July. The environment around the water point had displayed localized flooding, which would have provided contamination when the cistern was close to ground level, but would have been less problematic when the tap was introduced. FC levels of at least 3000 cfu/100 mL occurred in March and April, before dropping off to 100 in May and then climbing to 690,000 in June when the new tap was introduced. In the months after, until November, levels fell to zero. The spike in June might be explained by the disruption caused by the changeover. EB 2 was another water point that experienced a physical change during the study period, as the once jagged and pitted concrete and gravel plinth around the cistern was concreted over. While this permanently reduced the amount of standing water immediately next to the cistern, the potential for contamination remained close by on the sidewalk and roadside which continued to have some of the worst environmental assessment scores in the neighborhood. Interestingly this change had no impact on FC count trends, with a decline from 1350 to 1 from July to August, then rising again to over 2000 in September.Three of the six collection sites in the EB neighborhood yielded their highest contamination levels in June with EB 1 recording over 1 million cfu/100 mL, EB 4 over 690,000, and EB 6 over 370,000. However, this was not a neighborhood-wide pattern as EB 2, 3, and 5 all reported levels of zero for the same month. Counter to this, each of the EB sites, apart from EB 2, had their lowest environmental assessment scores for June. Indeed, June is tied with May and October as the months of least concern (on average across all six test sites) and is just behind August. Interestingly, October was the only month to have a zero FC count at all the EB household water access points. The most consistently problematic water point was EB 1, a cistern situated at the corner of an extremely busy road and a paved side street, while also being located within 20 m of the Bois de Chêne canal, one of Port-au-Prince’s main open-air sewage canals. EB 1 consistently recorded positive FC results with levels only falling below 50 for four of the months. EA generally had the worst water quality. Two of the safer water points, EA 1 and EA 2 were proximate, had a similar water delivery style and cistern structure, and had comparable environmental risks. Both sites saw an FC spike from April to May before becoming relatively safe for the remainder of the study period despite the micro environment around each being consistently wet and muddy. This same FC peak in April and May also occurred in eight of the other eleven water points, with seven of these recording their highest levels during this period (see Figure 6). A secondary FC peak occurred in July which affected six test sites, two of which recording their highest counts. While EA 4 and EA 18 were the most consistently contaminated cisterns (each had only one month with zero FC), their environmental assessments were different; EA 4 generally had proximate water and mud during peak FC months, but EA 18 only had moderate levels of environmental risk coinciding with high FC counts. Possibly a more concerning environmental risk factor for EA 18 was that it had one of the highest human activity measures and is located next to the previously mentioned Bois de Chêne canal. EA 4 and EA 5 were the only water points that displayed a connection between FC counts and their immediate environmental risks. In addition to FC, we isolated two V. cholerae nontoxigenic O1 strains from EA 10 and EA 14 in December 2017. Eight drainage channels and six coastline/ocean sites were tested in the EC and EA neighborhoods. As one would expect, these sample sites had the highest concentrations of FC. In general, for the EC neighborhood while there was a consistently high FC count across the months; August saw levels rise for all sites. Even though this was followed by a slight fall, counts remained in the tens of millions through the rest of the study period. There was a consistent level of standing water and mud around each of the EC drainage sites, though the amount of trash varied markedly, being especially high around EC 11, 12, and 14. Some of the sites had visible feces, large amounts of trash and high levels of mud and standing water around the channel. Each drainage channel in the EC neighborhood recorded high FC levels throughout the study period. The source of water into these channels was mostly household runoff (probably also containing some human waste) and rainfall flow. EC 9 was a smaller channel that flowed across a minor road in the southern portion of EC. This flow (at least visibly) appeared to be superficial and serving a local drainage purpose rather than being a collection point of sewage from the broader neighborhood. The environmental scores were lower compared to the other EC drainage sites, though people were often sitting immediately next to the water which flowed under/next to a residence. As an indication of the heterogeneity of risk in the neighborhood, while EC 9 had the lowest FC counts, EC 8 just 15 m away, had the highest FC levels. Indeed, when compared to all neighborhood drainage channels in the EA and EC neighborhood, EC 8 ranked as the fourth most contaminated while EC 9 was the least. Four EA drainage sites displayed similar patterns to the EC locations but were notably more severe. EA 12, 13, and 14 were the first, second, and third most contaminated drainage channels in all the study. Interestingly, EA 15, is the previously mentioned Bois de Chêne canal, which flows for several kilometers through Port-au-Prince serving as a catch-all drainage channel, yet it only ranked sixth in terms of FC counts. The canal, which appears to have numerous environmental risks, being full of trash, visible sewage, and feeding pigs, is approximately 15 m wide where the EA 15 samples were taken. Yet while environmental conditions at EA 15 were among the visibly worst at any site, and though FC counts were still high, the less visibly concerning and more proximate to residences EA 12, 13, and 14 drainage samples consistently scored higher.The three EC ocean sites had lower FC counts than those recorded for the drainage channels (presumably because of sea water dilution) though they still remained consistently high across all months. Just as with the drainage channels, FC peaked during August, though differing in hen returning to a pre-peak level. The only exception was EC 2 which had a second peak in December, similar to those also found in the drainage channels. The closest tested drainage (EC 11) was nearly 200 m away, so it is unclear why this secondary spike occurred here but not at the other EC ocean sites. Environmental assessments for the EC ocean sites were somewhat meaningless except for the extraordinary amounts of trash and feces found at the ocean edge.While the EA drainage sites had higher FC counts than their EC counterparts, the reverse was true for the ocean samples drawn in both neighborhoods. Just as with the EC sites, the land edge to these sample locations had excessive amounts of trash. However, while EA 6 spiked in August, similar to the EC ocean sites, EA 7 had several minor peaks and EA 8 had no single elevated month, though again, as expected, all three maintained generally high FC levels throughout the study period. It is widely acknowledged that residents of informal settlements suffer elevated exposure to disease, and that much of this risk is a result of clean water access and sanitation systems. Conceptually, we also understand that this risk is not homogenous but rather a localized function of water point type, drainage, nearby environmental risks, and behavior. In this paper we have attempted to take the first steps in documenting this geographic complexity using monthly FC counts and environmental assessments. Interpretation of these results led to the conclusion that there is a broad difference between the neighborhoods with EC being far safer than EA or EB. Interestingly EC is also the neighborhood that did not receive the cholera vaccine though, obviously the FC counts reported here are not indicative of a specific disease, rather a general level of risk. One explanatory factor for better results in EC was the method of water delivery, which was illegally sourced taps. For the other neighborhoods, where the water was often supplied from a reservoir at or close to ground level, contamination could occur through overland flooding or generally wetter conditions around the water point. This was especially true if the customer used either a tethered container or their own bucket to dip into the reservoir. In these cases, contamination could occur through human contact or by setting the vessel down in pathogen rich water or mud. While these were the general patterns, nuance and variation resulted in a heterogenous surface of risk within each neighborhood. For example, two water points built around a plastic in-ground drum below a city water pipe (EA 1 & 2) were relatively FC free for most of the study period, while the next closest site to these, at 70 m away (EA 18), recorded constant contamination. However, the proximate environmental conditions around EA 18 were visibly better than for EA 1 and 2. In another example, while the results in EC suggest a tap is the safest method of access, and with the one EB site improving its water quality after switching to this method further supporting this, there were still taps in the EA neighborhood that recorded months with a high FC. This suggests that other factors, especially the method of water source and storage, still play an important role in contamination. Similarly, proximate environmental risks alone do not predict contamination. Just as with the EA 1 and 2 comparisons with EA 18, the EC neighborhood generally had poor environmental conditions around each water point. This suggests that even if the conditions for potential contamination exist, protective factors (delivery type), and enhancing factors (nearby drainage) will create a heterogeneous intra neighborhood risk map. There are still more general observations that hold relevance for all study locations. For example, Figure 7 maps all water points in the highest risk category of Table 1 as yellow circles. Two potential spatial patterns emerge, the proximity to the main drainage channel (EA 11, 16, 17, 18 and EB 1), and being at the edge of the neighborhood drainage pattern where water channels reach the sea. This geographic pattern is further supported by EA 4 and 5 which occupy the next two places in the ranking behind the top risk grouping. This map suggests a problematic link when a water point is close to a drainage channel; indeed, each of the top four ranked sites are within 40 m of one another. While proximity to a drainage channel should be considered a risk, this relationship does not correlate with the size or appearance of the channel. While there was a distance decay effect of contamination increasing the closer a water point is to the largest and most visibly problematic Bois de Chêne canal, especially in EB, there were smaller drainage channels that ran through the built-up areas that had even higher FC counts. One possible reason might be that the increased flow in the larger channel leads to FC dilution. Irrespective, the fact that interior drainage channels proximate to both homes and water access points had such high FC counts is alarming, especially for those nearby water points sitting in standing water and mud. Future research should certainly explore this water point—drainage system nexus further.Another location of immediate concern is EA 16, which is sourced by ground-filtered ocean water. The high FC counts recorded here might not raise alarm because water from this location is only used for cleaning and bathing; however when using this as a benchmark, other nearby water points serving household needs had higher FC counts in March, May, June, and August. It is worth re-emphasizing this point—water points where community members would go to purchase household water actually had higher FC counts than a location sourced by the ocean. Adding further temporal complexity, the EA 16 highest FC count occurred between September and December which was when the majority of the other sites experienced their lowest levels.This last example also illustrates the danger in relying on a cross sectional FC and environmental assessment. While certain water points were continuously problematic, others fluctuated in risk level by the month of study (for example as seen in Figure 5). Part of the explanation for this temporal variation was meteorological impacts, as generally there are patterns of increased FC counts during the rainy season. The mechanism of contamination here is that overland flow from the drainage ditches either directly comes into contact with the water reservoir or the water/mud around the water point. This pattern is supported in the EC neighborhood which showed no such monthly pattern because there is less likely to be contamination from drainage to a tapped water supply. However, the generally wet and muddy conditions in this neighborhood mean that the same contamination pathways exist and might lead to either subsequent risks with transportation or storage in the home. It should also be noted that while meteorological conditions are not likely to vary within the neighborhoods, or even significantly across the three IS, the localized environmental conditions will lead to different factors increasing or ameliorating that risk. This helps explain the variability seen in Figure 5, though other human-caused changes, such as seen in Figure 6, can also change these temporal patterns. It is for reasons such as these that we need to supplement more traditional epidemiological analysis with a more contextualized understanding of the landscape. There are several limitations to the current study that could be addressed in future work. Understanding these micro environmental interactions would obviously benefit from fine scale meteorological data. These types of data (for example weekly rainfall) have proved valuable in recent predictive cholera models for Yemen [36]. Unfortunately, the authors do not know of any suitable fine scale meteorological data for Port au Prince. The team has recently purchased a single weather station to gain some basic rainfall data. Failing this, rain effects can be assessed using proxy measures such as standing water or mud. This leads to the question of how frequently should the environmental surveys be conducted? While we have collected by month, should these be collected every week, or after a single rain event? One challenge to exploring some of these questions in a more controlled manner is the challenging nature of these environments. For example, all data collection in the EA neighborhood ceased since April 2018 because of the ongoing gang violence in the area.While we have used the SV data to map basic environmental risks, there is always the risk of subjectivity. For this project, and to counter inter rater reliability issues, a single coder with a supervisor overseeing and sample checking all digitizing was used. This has proven successful in other environments. Could the SV data be mined further, such as how high is the water access point from the ground, and therefore could flood waters directly enter the storage container? Are there other concrete surfaces close by that could be used as a staging point to rest buckets? While we have considered spatial and temporal variation in FC, supplemented with micro environmental assessments, we have not, as of yet, included human insight into these processes. We do not know the choices involved in selecting a water point, or how families interact with the neighborhood waste water system. We do not know how conditions change at night, or even how a typical day brings the members of a family into contact with other enteric disease-causing environments. For example, as seen in Figure 2, we have visual evidence that children play in the drainage channels that probably contribute to the contamination of the neighborhood water points. Therefore, while spatio temporal variation of water point contamination is vital and should continue, future work should contextualize these findings with behavioral insights. One method which has proven successful in providing such contextualization is spatial video geonarratives (SVG), where local commentaries are added to the SV and then mapped [35,37]. SVG could reveal insights into water point choice, community knowledge on water storage and transport risks, perceived risks when rains occur, and more general daily activity patterns, including where children go to play. Seasonal factors, micro environments around each water point, and proximity to broader drainage patterns should all be considered when determining water point contamination risk. This study has found that the two strongest predictors of water safety are (positive) using a tap to limit environmental contamination, and (negative) being proximate to neighborhood drainage patterns. Even for daily prevention strategies the results presented here suggest where water chlorination is immediately needed, and which water points should be more closely monitored. This insight will not only aid practitioners and epidemiologists working to reduce contamination levels but may also influence local prioritization of vending sites. The same information could be used to help inform community groups as to where their water sources are cleanest, and in so doing place more pressure on local vendors to improve their water points.While there were no cholera cases in these neighborhoods during the project period, one would expect future outbreaks to contain aspects of the results described here. Of the three neighborhoods, EC appears to be safest even though the generally wetter conditions mean that there is a contamination risk during water transport. This is encouraging as this neighborhood was not part of the initial cholera vaccine strategy. Neighborhood EB would have a varying risk, with one water point (C1) needing immediate attention. The cholera risk would vary within neighborhood EA, with those living closer to the “edge”, meaning being proximate to the main drainage channel or the sea, being most at risk due to overland flow from local drainage channels emerging from the densely settled area where presumably the cases would occur. As a result, several water points would need to be closely monitored (V 9 through V 18). These insights might be extremely useful during an outbreak when resources are limited and time of intervention vital. However, for all neighborhoods, the potential for contamination exists, and while we might advocate for spatial prioritization, these should be in addition to more global risk reduction strategies. Finally, SV data also allow for virtual return trips to investigate other disease risks, in effect breaking down research silos. The SV revealed environmental risks connected to other health hazards—drainage channels that flow close to (and in some cases under) houses with people often sitting nearby. Not only is this a concern for water borne disease risk but also as a mosquito breeding ground. Another hazard, trash, which can be used in addition to sewage and standing water as a proxy for leptospirosis [38], varied dramatically across the neighborhood with amounts dramatically rising towards the ocean. These trash piles not only attract a wide range of animals (including stray dogs and presumably rats), but also contain a high amount of human feces. Similarly, the SV data could be repurposed for non-health usage. A large part of the EC neighborhood is built upon semi-compacted trash and rubble that has amassed over the past 20 years. Unregulated disposal of refuse and subsequent building has allowed this area to expand into the sea. These reclaimed areas are vulnerable to future liquefaction associated with seismic activity, as well as sea water inundation through sea level rise and storms/hurricanes. Evacuation and other hazard mitigation strategies, along with security, infrastructure, amenity and even education planning could benefit from these spatially linked visuals, especially if collected over time, to make more informed intervention strategies. Conceptualization, A.C., V.R. and J.G.M.J.; methodology, A.C.; software, J.A.; validation, M.T.A., A.A. and V.R.; formal analysis, A.C., R.S., M.T.A., A.A., M.M.A., M.H.R.; resources, J.W.P.; data curation, S.B.; writing—original draft preparation, A.C. and R.S.; writing—review and editing, all authors; visualization, A.C., R.S. and J.A.; supervision, A.C., V.R., J.W.P. and J.G.M.J.; project administration, J.G.M.J.; funding acquisition, J.G.M.J.This research was funded by the National Institute of Allergy & Infectious Diseases RO1 Al126357 Cholera Transmission and Evolution in Port-au-Prince, Haiti 2016-2021.We would all like to thank the GHESKIO and Gressier teams that conduct the environmental sampling activities: Serge Emmanuel, Ricles Saint-Louis, Ricardo Paul, Sylvain Dumerlin, Makenson. Andrew Curtis would like to thank all the students who have worked in the GIS Health & Hazards Lab. He would also like to thank Martin Kennedy, Mick Moss and Dee Palmer for providing such wonderful distractions—the writing of this manuscript has benefited greatly from their presence in and around the office. And last, but not least, Kenneth Wilson, who provides constant support and inspiration. I can’t express how important our meetings are to me. The authors declare no conflict of interest.Examples from the three study neighborhoods. (A) study neighborhood EA; (B) study neighborhood EB; (C) study neighborhood EC.Two example locations of concern; (A) is a drainage channel full of trash and waste in which children are playing, and (B) is a public water point.Maps displaying GPS (global positioning system) path corrections in the EC study area. In each image, the blue line shows the walking path connecting test locations. The path on the left is the GPS extracted from the camera. The path on the right has been corrected.Two environmental testing locations used as a comparison with the household water point samples. (A) is an ocean site (EC 11) while (B) is a tidal water filled reservoir (EA 16).Fecal Coliform counts per month for the water point testing sites in the EA neighborhood.An example of the Fecal Coliform counts and environmental scores for one example water point sample site.Generalized locations of high risk sample sites in the EA and EB neighborhoods. Arrows indicate the increases in contamination the closer the site is to either the Bois de Chêne canal (A and B) or the coastal edge (A).Monthly Fecal Coliform counts for each of the tested water points.Test sites with a 0* indicate no value for that month, either because the water point was not available (for example being locked), it was dry having <1 cfu/100 mL sample water, or there was a problem with the sample. Months exceeding 100 cfu/100 mL. (colony forming units per 100 milliliters of collected water) are colored light grey, those exceeding 1000 are colored dark grey, and those exceeding 100,000 are in light red. EA, EC and EB are the three study neighborhoods.

Med-MDPI/ijerph_3/ijerph-16-05-00808.txt

@@ -0,0 +1,10 @@
+Singapore experienced its first Zika virus (ZIKV) cluster in August 2016. To understand the implication of human movement on disease spread, a retrospective study was conducted using aggregated and anonymized mobile phone data to examine movement from the cluster to identify areas of possible transmission. An origin–destination model was developed based on the movement of three groups of individuals: (i) construction workers, (ii) residents and (iii) visitors out of the cluster locality to other parts of the island. The odds ratio of ZIKV cases in a hexagon visited by an individual from the cluster, independent of the group of individuals, is 3.20 (95% CI: 2.65–3.87, p-value < 0.05), reflecting a higher count of ZIKV cases when there is a movement into a hexagon from the cluster locality. A comparison of independent ROC curves tested the statistical significance of the difference between the areas under the curves of the three groups of individuals. Visitors (difference in AUC = 0.119) and residents (difference in AUC = 0.124) have a significantly larger difference in area under the curve compared to the construction workers (p-value < 0.05). This study supports the proof of concept of using mobile phone data to approximate population movement, thus identifying areas at risk of disease transmission.The emergence of dengue (DENV), chikungunya (CHIKV) and zika (ZIKV) [1,2,3,4,5] viruses into new susceptible human populations, driven in part by human travel on both local and global scales, represents significant public health concerns, particularly in Singapore with suitable climatic conditions and constant urbanization. Unlike CHIKV and DENV, ZIKV’s association with Guillain–Barré syndrome and microcephaly cases was a major cause of public alarm during the epidemics [6]. In addition, a study has reported that the social and economic impacts of the recent Zika outbreak in Latin America and the Caribbean alone could cost countries in the region of an estimated 7–18 billion dollars from 2015 to 2017 respectively [7]. Reports in 2016 confirmed the introduction and spread of ZIKV in Southeast Asia and that the likelihood of under-diagnosis is high [8]. It is imperative to have timely health care response and outbreak control measures, which are focused in areas predicted to be at the highest risk of transmission, to effectively control outbreaks of new infectious diseases [9].Human movement is known to play an important role in the transmission dynamics of diseases [10,11,12]. With the recent emergence of mobile phone calling data and associated location information, experts have suggested using mobile phone data to understand the spatial and temporal spread of disease incidences [13]. One such study was recently conducted on the 2010 Cholera epidemic in Haiti, whereby anonymized mobile data enabled the prediction of epidemic spread [14]. In addition, another study on the same epidemic also showed that it was feasible to monitor real-time population movements using mobile phone data during infectious disease outbreaks, thus allowing for better preparedness in controlling outbreaks [15]. Studies conducted by Amy Wesolowski et al. have shown the possibility of using mobile phone–based mobility estimates to predict the geographic spread of vector-borne diseases, such as dengue epidemics as well as malaria prevalence [16,17].In August 2016, Singapore experienced its first local zika outbreak [5,18]. When the first locally acquired ZIKV case in Singapore was identified, the Ministry of Health (MOH) conducted active case findings for the purpose of outbreak investigations and public health measures [19,20]. The National Environment Agency (NEA) intensified its battle against the Aedes mosquitoes to interrupt transmission. In addition to vector control measures, public outreach was stepped up to urge residents to clear potential breeding habitats to prevent mosquito breeding.As ZIKV was newly introduced in Singapore, this outbreak provided an opportunity to understand the implication of human movement on disease spread. A total of 460 ZIKV cases were reported in Singapore in 2016, among which 64.8% (n = 298) belonged to the first ZIKV cluster, known as the Aljunied Crescent cluster. We postulated that human movement from the Aljunied Crescent cluster locality led to the spread of ZIKV and that different segments of population contributed differently to the spread. Therefore, a retrospective study was conducted to determine how the movement of different groups of individuals (construction workers, residents, visitors) from the Aljunied Crescent cluster locality to other localities influenced ZIKV transmission. We used reported cases of ZIKV infection to validate the accuracy of the identified areas. The results of this study would enable the identification of risk areas of disease transmission areas for targeted intervention. The key aim of this study is to see if only human movement data is sufficient to accurately predict the spread of the disease. Such a predictive tool could help public health officers prioritise areas for investigation, inspections and vector control, to preemptively mitigate the risk of transmission.ZIKV case data. In Singapore, the MOH is responsible for the epidemiological and clinical management of ZIKV, while NEA is responsible for the vector surveillance and control. A ZIKV case is defined by a person who is tested positive for confirmed ZIKV in serum or urine samples by the reverse transcription polymerase chain reaction method. All cases are notified to MOH and epidemiological investigations are then conducted via interviews to obtain epidemiological data. Epidemiological information such as residential addresses and dates of onset of illness and diagnosis of patients were collected for this study. The permission to use these data was approved by MOH, in Singapore on 5 July 2018. All laboratory-confirmed and locally-acquired ZIKV cases reported to MOH in 2016 were anonymized and aggregated to hexagons prior to analysis.ZIKV cluster. In Singapore’s Aedes borne disease control programme, a ZIKV cluster of two cases located within 150 m from each other and with onset dates within 14 days [5], suggests a localized transmission. Subsequent cases that fulfill the same criteria are also tagged to the same cluster. Using these criteria, the Aljunied Crescent cluster was considered active from 29 August 2016. As all locally acquired ZIKV cases were used in the analyses, the assumption we made was that all the cases were epidemiologically linked to one another. Hence, we assumed that the spread of the ZIKA transmission around the island started off from cases at the Aljunied Crescent cluster locality.Study area. Singapore is divided into regular hexagons each with a circum-radius of 165 m and an average area of 0.07193 km2. There is a total of 11,401 hexagons, out of which 5,176 (45.4%) are residential hexagons. The Aljunied Crescent ZIKV cluster locality, which is defined as the area where the ZIKV clustered cases are reported, consists of 24 hexagons, with a total area of 1.72 km2 (Figure 1). The Aljunied Crescent ZIKV cluster locality comprises an approximate human population size of 34,000. The area is a mixture of residential buildings; commercial buildings such as shop-houses; and other amenities like schools, swimming complex, hospitals as well as a train station. During the period of this analysis, there was an ongoing residential construction site located within the cluster locality. The Aljunied-Kallang neighbourhood, where the cluster occurred, has a historical presence of Ae. aegypti and was categorised as a high-risk area for dengue transmission [5].Mobile phone data. Call data records (CDRs) provide a detailed temporal and spatial movement of subscribers of mobile telephone companies (Telcos). The present study was based on all CDRs from three one-week periods from 13–19 June 2016, 25–31 July 2016 and 8–14 August 2016. This period was selected because the transmission period of the ZIKV cluster was from 31 July 2016 to 25 September 2016. Hence, the period selected for this analysis starts from 2 months prior to the onset of the cluster, as well as during peak transmission periods of the cluster. The data contained a total of 70,147 subscribers (for the three weeks) of StarHub limited (Ltd), one of the three major Telcos in Singapore, which constituted approximately 30% of the coverage of total mobile subscribers in Singapore. Aggregated Telco insights are availed through data products developed on anonymized data, respecting consumer privacy and strict adherence to personal data protection legislation. The network is represented by nodes and entities, where nodes are the entities and links are the relationships between the entities. In this study, the nodes refer to the regular hexagons of size 165 m, and the links refer to the connectivity between the pairwise hexagons (i.e., number of subscribers travelling from one hexagon to another). GRID360, a geo-location product developed by StarHub Ltd. to anonymize and aggregate island-wide geolocation-related data onto regular hexagons of size 165 m, was deployed by StarHub Ltd. to detect dwelling points of individual subscribers and subsequently their origin and destination locations of travel by analyzing their cell phones’ network activities. The GRID360 origin–destination location engine detects a dwelling point (origin or destination) when the subscriber has stayed in a hexagon for 30 min or longer. Each unique subscriber in our data corresponds to a single geographical location and is generated when the following occur:Whenever there are activities on the phone which engage a service of the Telco, such as usage of mobile data, including applications running in the background, calls or short message services.Whenever the phone moves into another location area code.Periodically, there is a location update (roughly once every 2 to 3 h) if there is no activity or movement to another location area code.Whenever there are activities on the phone which engage a service of the Telco, such as usage of mobile data, including applications running in the background, calls or short message services.Whenever the phone moves into another location area code.Periodically, there is a location update (roughly once every 2 to 3 h) if there is no activity or movement to another location area code.Firstly, human movement from Aljunied Crescent ZIKV cluster locality was developed. The Aljunied Crescent ZIKV cluster locality was the origin. All other hexagons with movements into them were destinations. Human movement data was categorized into three groups of individuals: (1) construction workers, (2) residents and (3) visitors.Construction workers: Subscribers who dwelled in the construction site hexagon within the Aljunied Crescent ZIKV cluster locality at least five times per week for more than 4 h on each occurrence, were at least 18 years old and were from the following countries (Malaysia, People’s Republic of China, India, Sri Lanka, Thailand, Bangladesh, Myanmar, Phillipines, Hongkong, Macau, South Korea and Taiwan) and categorized as construction workers. The countries were chosen as per the Ministry of Manpower’s work permit requirement criteria for construction workers [21].Residents: Subscribers who dwelled in the Aljunied Crescent ZIKV cluster locality at least five times per week for more than 4 h on each occurrence and resided in the locality (based on their home addresses) were categorized as residents. There will be no overlap with the category of construction workers, as construction workers do not have residential addresses within the area.Visitors: subscribers who dwelled in the Aljunied Crescent ZIKV cluster locality at least five times per week for more than 4 h on each occurrence and did not reside in the locality (based on their home addresses) were categorized as visitors. In this category also, subscribers who overlapped with the category of construction workers were excluded from this group.Analyses of the retrieved Telco data using a chi-square test showed that the data was comparable to Singapore’s population data for each of the variables: age group, gender and ethnicity. Hence, we assume that the data is representative of the whole population.The odds ratio (OR) was calculated to find out the odds of ZIKV cases reported in a hexagon that had human movement from the cluster locality into a hexagon, compared to the odds of ZIKV cases reported in a hexagon in the absence of human movement from the cluster locality into a hexagon. OR was calculated for total subscribers included in the study, as well as individually for different group types. A chi-square test for independence was used to test for significance between the variables. OR was calculated using the Formula below:OR = (a/c)/(b/d)
+		where:“a”denotes the number of hexagons reported with at least one ZIKV case and at least one person moving into a hexagon.“b”denotes the number of hexagons with no ZIKV case and at least one person moving into a hexagon.“c”denotes the number of hexagons with at least one ZIKV case and nobody moving into a hexagon.“d”denotes the number of hexagons with no ZIKV case and nobody moving into a hexagon.denotes the number of hexagons reported with at least one ZIKV case and at least one person moving into a hexagon.denotes the number of hexagons with no ZIKV case and at least one person moving into a hexagon.denotes the number of hexagons with at least one ZIKV case and nobody moving into a hexagon.denotes the number of hexagons with no ZIKV case and nobody moving into a hexagon.The Receiver Operating Characteristic (ROC) curve in this analysis depicts the sensitivity (true positive rate) and specificity (true negative rate) of using origin–destination location insights to predict the ZIKV occurrences in hexagons. Sensitivity refers to the proportion of hexagons that had movement and ZIKV cases (true positive rate); and specificity refers to the proportion of hexagons that had no movement and no ZIKV cases (true negative rate). False positive rate refers to the hexagons that were identified as having ZIKV cases although they did not. False negative rate refers to the hexagons that were identified as not having ZIKV cases although they had. ROC curves were generated for each of the three groups of individuals. The AUC of the ROC curves provides a measure of how well the human movement of the particular group is able to predict the occurrence of the ZIKV cases in the hexagons. The ROC curves of the three groups were independent of each other. Hence, a comparison of independent ROC curves was used to test the statistical significance of the difference between the areas under the ROC curves of the three groups of individuals (comparing ROC curve of residents with ROC curve of visitors, comparing the ROC curve of residents with the ROC curve of construction workers and comparing the ROC curve of visitors with the ROC curve of construction workers) with the non-parametric method described by Delong et al. [22].The value of AUC using the empirical method is calculated by summing the area of the trapezoids that are formed below the connected points making up the ROC curve. From DeLong et al. [23], we define the T1 component of the ith subject, V(T1i) as
+        V(T1i)= 1n0∑j=1n0¥(T1i,T0j)And define the T0 component of the jth subject, V(T0j) as
+        V(T0j)= 1n1∑i=1n1¥(T1i,T0j)
+        where
+        ¥ (X,Y)= 0 if Y>X,¥ (X,Y)= 12 if Y=X,¥ (X,Y)= 1 if Y<XThe empirical AUC is estimated as
+        AEmp=∑i=1n1V(T1i)/n1=∑j=1n0V(T0j)/n0The variance of the estimated AUC is estimated as
+        V(AEmp)=1n1ST12+1n0ST02
+        where ST12 and ST02 are the variances
+        ST12=1nj=1∑i=1ni[V(T1i)−AEmp]2,i=0,1Spatial maps were created in ArcGIS version 10.5 (Environmental Systems Research Institute, Inc. (ESRI, Redlands, CA, USA). R software version 3.0.2 (R: A language for environment and statistical computing, Vienna Austria) was used for statistical tests. Statistical significance was assessed at the 5% level.A total of 460 ZIKV cases were reported, among which 64.8% (n = 298) belonged to the Aljunied Crescent cluster. The majority of the initial cases (80% of the 1st three weeks of onset of the illness, epidemiological week 31–33) reported in the cluster were workers from the construction site (Figure 2). The ZIKV density map showed that though cases were concentrated in the eastern side of Singapore, we also see sporadic cases spread around the island (Figure 3).There were a total of 5176 residential hexagons in Singapore, and the 460 ZIKV cases were reported in 173 hexagons. Out of these, 24 hexagons belonged to the Aljunied Crescent Locality. The odds ratio of ZIKV cases in a hexagon visited by a subscriber from the Aljunied Crescent cluster locality, independent of the group of individuals, is 3.20, reflecting a higher count of ZIKV cases when there is a movement into a hexagon from the cluster locality. It is statistically significant, as evidenced by the 95% CI (2.65–3.87) which does not include the value 1 (the null hypothesis is understood to be odds ratio = 1) and a Chi-square test of independence yields a p-value < 0.05. Moreover, when we take into consideration the different groups of individuals, the effect of this confounding variable give us similar odds ratios. Table 1 presents the total number of subscribers moving into a hexagon from the cluster locality and the odds ratio of reported ZIKV cases among different groups of individuals. The odds ratio of having ZIKV cases reported in a hexagon visited by a subscriber from the cluster locality for construction workers is 1.15 and 0.3 lower than residents and visitors respectively, as we had anticipated. The reason is because construction workers had limited movement, with at least 20% fewer hexagons visited compared to residents and visitors. In contrast, the odds ratio of ZIKV cases from the cluster locality moving into a hexagon for residents is higher at 4.24. Furthermore, for all groups of individuals, the overall case burden of hexagons visited by at least one person from the Aljunied Crescent cluster locality is significantly higher than that of remaining hexagons, as evidenced by the 95% CI. Figure 4 shows that residents and visitors had a broader range of movement, with a higher density of subscribers found in the southern and south-eastern part of Singapore. The spatial distribution of the subscribers is comparable to the spatial distribution of the population living in Singapore, as shown by the human population density segregated by subzones [23]. In contrast, construction workers had a more limited range of movement.Comparison of the independent ROC curves showed that visitors and residents have a significantly larger difference in area under the curve (AUC) compared to the construction workers using the non-parametric method of Delong et al. (p-value < 0.05) (Figure 5, Table 2). There is no significant difference in AUC between residents and visitors (p-value > 0.05)This study shows that there are higher odds of ZIKV cases being reported in areas that were visited by people from areas with active ZIKV transmission compared to areas that were not (Table 1). However, the AUC (visitors = 0.686, residents = 0.691, construction workers = 0.567) was moderate for visitors and residents and low for construction workers suggesting that using only human movement is not good enough to predict risk areas with high accuracy. Nevertheless, the findings are particularly important for improving early containment efforts of vector-borne diseases. This finding is supported by several studies that have demonstrated the role of human movement in the spatial spread of diseases [10,11,12,13]. Singapore is highly susceptible to infectious disease outbreaks due to multiple factors: (i) Singapore being a travel hub, (ii) having one of the highest population densities in the world (7796 people per square kilometer) [24] and (iii) improved local transportation that facilitates virus dissemination. Hence, human movement data can be used as a risk assessment tool to identify risk areas to ensure that resources are deployed in a strategic and sustainable manner.Human movement is not the only risk factor for ZIKV transmission, as it is a vector-borne disease. The data shows that human movement during the daytime is heavily aggregated in south and southeast Singapore. However, we did not have any clusters in many of these locations. For active transmission to be established in the areas visited by the individuals from the cluster locality, other critical factors such as land use, demography, presence of vectors, presence of naïve humans, human density and conducive weather parameters are required. The accuracy of the potential risk areas identified can be improved by overlaying the spatial distribution of Aedes population. The technique was explored in a recent study in Namibia, where human movement data together with case data was used to prioritize areas for surveillance and control of Malaria [25,26]. In Singapore, a similar technique is currently used for dengue control. A dengue risk map was developed by modeling multiple layers of data such as past dengue cases, presence of mosquitoes, human population density and environmental factors to guide vector control operations [27].In Singapore, the range of movement of different segments of population can differ based on their lifestyles. Therefore, we categorized human movement data into different groups of individuals (residents, visitors and construction workers) to find out the impact of each group on the spread of the disease. A past study conducted in the Environmental Health Institute has shown that a construction site is a potential environmental driver for increased dengue transmission, and that the odds of dengue clusters, associated with construction sites expanding into large clusters, are 9 to 17 times higher than clusters not associated with construction sites [28]. This is because large-scale building construction activities often create environmental conditions conducive for vector-borne disease transmission due to the increased likelihood of Aedes breeding habitats. Also, construction workers, who are generally from dengue non-endemic countries, are prone to dengue infections. As a result, we postulated that they might be a key factor in facilitating dengue transmission in residential areas that are located in close proximity to construction sites. ROC analysis showed that residents and visitors of the cluster were more likely to cause the spread of ZIKV around the island as compared to the construction workers. This is likely due to the smaller range of movement of construction workers, who tend to stay within their worksites and dormitories. However, it is possible that workers who stayed within their worksites might have been responsible for diffusing ZIKV in the residential areas close to worksites especially at the beginning of the outbreak, which is then spread to other parts of the island. We also noticed that the dormitories were located in areas with low Ae.aegypti population, which could be a possible reason why an outbreak was averted [29]. By categorizing human movement based on the population of interest, this study has shown us more insights into the spatial distribution in relation to disease spread which can be important to better understand localized spread of diseases, such as within a community, for targeted control efforts.One key limitation of this study is in the quantification of human movement based on mobile data. The inference of the different categories in the data is unlikely to be without error. For example, the method to infer the categories was a trade-off in tuning the appropriate hyper parameters to maximize the chances of identifying data that exhibited a category. For instance, construction workers’ behavior is sufficient to believe that they were likely to be workers associated with the construction site in the cluster. Additionally, the data, being dependent on the frequency of mobile phone usage, can vary for different groups. For example, no data will be generated for mobile phones that have been switched off. Another concern is that the data collected by the Telco used in this study, represented approximately one third of the total mobile phone users in the market. Nonetheless, this data is expected to have little impact on the spatial patterns observed in this study.Another key limitation is that data was collected from three one-weeks periods in June, July and August and not throughout the epidemic transmission. An analysis of CDRs using Kendall’s tau-b correlation of the movement of the general human population has shown that there is no significant difference in the movement data on a month-to-month basis. Therefore, the data is expected to be a sufficient representation of movement data throughout the epidemic transmission.Another significant limitation is that the association between human movement and ZIKV cases was done as a binary analysis. As the mobile phone data was aggregated to a hexagon level, analyses did not take into account the association between the distance of movement between zones and intensity of ZIKV transmission. More in-depth data should be collected for such analyses. For successful future use of this method in outbreak control of infectious diseases, having access to aggregate data on mobility between areas on a regular basis is necessary to identify risk areas. Hence, it will be ideal to have infrastructure and systems in place for such aggregated connectivity data to be collected from all the Telcos present in Singapore to assist decision making processes for future outbreaks.In summary, this study supports the proof of concept of using mobile phone data to approximate population movement, which can be used to identify areas at risk of disease transmission. This method is especially useful for newly introduced human-to-human diseases. By quickly identifying possible areas of transmission, operations can be carried out in a timely manner to help mitigate the spread. Further studies are warranted to include other risk factors such as vector population and herd immunity to improve the accuracy of the risk areas. Despite having a low mosquito population, Singapore continues to be under threat of vector-borne diseases transmission due to several factors such as current dynamics of climate change, globalisation, travel, trade, socioeconomics as well as viral evolution [30]. Hence, the current vector control regime has to be constantly enhanced by incorporating data analytics to be more effective.Conceptualization, L.-C.N.; Data Curation, J.R., S.-H.L. and Y.-H.T.; Formal Analysis, J.R., J.O., S.-H.L., Y.-H.T. and W.B.; Investigation, J.R., J.O.; Methodology, L.-C.N. and J.R.; Project Administration, L.-C.N.; Resource, L.-C.N.; Supervision: G.Y. and C.-S.C.; Validation, G.Y. and C.-S.C.; Visualization, J.R.; Writing—Original Draft Preparation, J.R.; Writing—Review & Editing, C.-S.C. and G.Y.The study was funded by NEA, Singapore. We thank colleagues of Communicable Diseases Division, MOH, Singapore who contributed to the collection of the epidemiological data of ZIKV cases and colleagues in NEA for analytical guidance.The authors declare that they have no competing interests.Study area. The shaded green area is the Aljunied crescent ZIKV cluster boundary. The hexagon highlighted in red contains the implicated construction site. The figure was created using ArcGIS version 10.5.Weekly number of reported ZIKV cases in 2016. Cases in Aljunied cluster from August–October 2016 highlighted in red and orange; cases working in the adjacent construction site are red and the rest of the cases in the cluster are orange.Spatial distribution of Zika cases in 2016. The spatial distribution of Zika cases in Singapore was generated using the kernel density tool in the spatial analyst toolbox of ArcGIS 10.5 ArcMap software (ESRI, CA, USA) based on a search radius of 400 m. Case density values were classified into four classes of <25th (2 cases/km2), 25th–50th percentile (3–8 cases/km2), 51st–75th percentile (9–29 cases/km2) and more than 75th percentile (>29 cases/km2) quantiles, using the quantile classification method and were displayed in tones of yellow, orange and red respectively, as shown in the legend.Maps (A–C) show the movement of construction workers, residents and visitors from the Aljunied Crescent ZIKV cluster, respectively. The number of subscribers that move into the residential hexagons from the Aljunied Crescent ZIKV cluster were classified into four classes of <25th (1 subscriber) in yellow, 25th–50th (2–3 subscribers) in green, 51st–75th (4–8 subscribers) in blue and more than 75th (9 and above subscribers) in red using the quartile classification method. Five-thousand-one-hundred-and-seventy-six residential hexagons are overlaid on the map of Singapore. A similar movement pattern was observed for the three periods that were selected. The Aljunied Crescent ZIKV cluster locality is overlaid on the map. The figure was created using ArcGIS version 10.5.ROC curves (sensitivity and specificity) for predicting the ZIKV occurrences based on the human movement of the three groups of people: blue line denotes visitors, red line denotes residents and green dotted line denotes construction workers. The area under curve (AUC) for visitors, residents and construction workers are 0.686 (0.640–0.732 95% CI), 0.691 (0.647–0.734 95% CI) and 0.567 (0.533–0.600 95% CI), respectively.Comparison of the number of subscribers moving into a hexagon from the Aljunied Crescent cluster and the OR of ZIKV cases reported in a hexagon form the Aljunied Crescent cluster for the three groups of individuals.Comparisons of independent ROC curves with the non-parametric method of Delong et al. [22] of the three groups of individuals: construction workers, visitors and residents.

Med-MDPI/ijerph_3/ijerph-16-05-00809.txt

@@ -0,0 +1 @@
+These authors contributed equally to this work.The current study was designed to investigate the short-term effects of policosanol consumption on blood pressure (BP) and the lipid parameters in healthy Korean participants with prehypertension. A total of 84 healthy participants were randomly allocated to three groups receiving placebo, 10 mg of policosanol, or 20 mg of policosanol for 12 weeks. Based on an average of three measurements of peripheral BP, the policosanol 20 mg group exhibited the most significant reduction, that is, up to 7.7% reduction of average systolic BP (SBP) from 136.3 ± 6.1 mmHg (week 0) to 125.9 ± 8.6 mmHg (week 12, p < 0.001). Between group comparisons using repeated measures ANOVA showed that the policosanol 20 mg group had a significant reduction of SBP at 12 weeks (p = 0.020) and a reduction of diastolic BP (DBP) at 8 weeks (p = 0.041) and 12 weeks (p = 0.035). The policosanol 10 mg and 20 mg groups showed significant reductions in aortic SBP of 7.4% and 8.3%, respectively. The policosanol groups showed significant reductions of total cholesterol (TC) of 9.6% and 8.6% and low-density lipoproteins (LDL-C) of 21% and 18% for 10 mg and 20 mg of policosanol, respectively. Between group comparisons using repeated measures ANOVA showed that the policosanol (10 mg and 20 mg) groups at 12 weeks had a significant reduction of TC (p = 0.0004 and p = 0.001) and LDL-C (p = 0.00005 and p = 0.0001) and elevation of %HDL-C (p = 0.048 and p = 0.014). In conclusion, 12-week consumption of policosanol resulted in significant reductions of peripheral SBP and DBP, aortic SBP and DBP, mean arterial pressure (MAP), and serum TC and LDL-C with elevation of % HDL-C.Hypertension is a key risk factor for the incidence of stroke and cardiovascular disease (CVD) and is often accompanied by dyslipidemia and diabetes [1,2]. It has been well established that the serum level of high-density lipoprotein-cholesterol (HDL-C) is inversely correlated with the incidence of CVD, diabetes, and Alzheimer’s disease [3,4,5]. However, the percentage of HDL-C in total cholesterol (TC), rather than amount of HDL-C (mg/dL), is considered a more important factor to predict the risk of incident hypertension [6]. Korean subjects with prehypertension have higher serum TC, triglycerides (TG), and serum TG/HDL-C [7]. In the same context, a study of Middle Eastern women showed that TG/HDL-C are strong predictors of incident hypertension [8]. A link between dyslipidemia, obesity and hypertension via stimulation of aldosterone synthesis has been proposed [9]. Recently, higher VLDL was shown to increase aldosterone production in mitochondria via binding of scavenger receptor-class B type I (SR-BI) and stimulation of several signaling pathways in acute aldosterone secretion and sustained aldosterone production [10]. Modified LDL, such as oxidized LDL and glycated LDL, also stimulate aldosterone release via Jak-2 activation for adrenocortical steroidogenesis [11].Policosanol is mixture of eight aliphatic primary alcohols purified from sugar cane wax (Saccharum officinarum L.). These primary alcohols range from 24–34 carbon atoms, with octacosanol, triacontanol, dotriacontanol, hexacosanol and tetratriacotanol as main constituents. Consumption of policosanol can reportedly reduce TC and LDL-C with less oxidation of LDL [12,13]. In our previous studies, incorporation of policosanol into the core of HDL enhanced HDL functions via enhancement of anti-glycation, anti-apoptosis, and CETP inhibition [14]. Policosanol supplementation for 9 weeks in zebrafish had serum lipid-lowering and HDL-C-elevating effects via CETP inhibition; policosanol also ameliorated fatty liver changes [15]. Policosanol supplementation in Korean participants raised serum HDL-C and enhanced HDL functionality to inhibit oxidation and glycation of LDL and HDL [16]. Policosanol therapy for 8 weeks by healthy female subjects who had prehypertension resulted in lower blood pressure and CETP activity by elevating HDL/apoA-I contents and enhancing HDL functionalities, including cholesterol efflux and insulin secretion [17]. Eight weeks of policosanol supplementation in spontaneously hypertensive rats (SHR) resulted in remarkable decreases of blood pressure in a dose-dependent manner [18]. In addition to increasing the HDL-C level, long-term (24 weeks) consumption of policosanol lowered BP while enhancing the athero-protective functions of HDL as well as its antioxidant, anti-glycation, and anti-atherosclerotic activities [19].Based on the previous studies, in the current study we aimed to analyze the effects of short-term (12 weeks) policosanol consumption on peripheral BP and central aortic BP along with any changes in the serum lipid profile in healthy Korean subjects who had prehypertension.The policosanol raw material (also known as sugar cane wax alcohol) was obtained from Rainbow & Nature Pty, Ltd (Thornleigh, NSW, Australia). The policosanol consisted of several alcohol chains of various lengths. More than 90% of policosanol contents were higher aliphatic wax alcohols. Individual alcohols present in policosanol are identical with our previous report [19] and other reports for genuine policosanol [20]: namely, 1-tetracosanol (C24H49OH, 0.1–20 mg/g);1-hexacosanol (C26H53OH, 30.0–100.0 mg/g); 1-heptacosanol (C27H55OH, 1.0–30.0 mg/g); 1-octacosanol (C28H57OH, 600.0–700.0 mg/g); 1-nonacosanol (C29H59OH, 1.0–20.0 mg/g); 1-triacontanol (C30H61OH, 100.0–150.0); 1-dotriacontanol (C32H65OH, 50.0–100.0 mg/g); 1-tetratriacontanol (C34H69OH, 1.0–50.0 mg/g).The placebo tablet had the same taste and smell with an identical colour. It contained the same basic pigment (Gardenia Blue color) and ingredients such as lactose, cellulose, glycerin fatty acid ester, magnesium stearate, etc, except for the policosanol.Healthy male and female volunteers who had prehypertension (systolic 130–139 mmHg, diastolic 80–89 mmHg) were recruited by newspaper advertisement. All participants were pre-screened for suitability and the inclusion criteria were as follows: age 19–65 years old who had pre-hypertension without any history of endocrinological disorder. Heavy alcohol drinkers (>30 g ethanol (EtOH)/day) and those who had consumed prescribed drugs associated with hyperlipidemia, diabetes mellitus, or hypertension were excluded. All participants had unremarkable medical records without illicit drug use or past history of chronic diseases. On the first visit, all participants (n = 84) were grouped randomly by dice casting as group 1 (placebo), group 2 (policosanol 10 mg), or group 3 (policosanol 20 mg). All participants in each group consumed policosanol for 12 weeks, according to the study design (Figure 1). The Institutional Review Board at Yeungnam University (Gyeongsan, South Korea) approved the study and endorsed the protocol (IRB #7002016-A-2016-021), and the participants signed an informed consent form prior to the study beginning.This study was a double-blinded, randomized, and placebo-controlled trial with a 12-week treatment period. Participants were informed to consume one tablet per day containing policosanol (10 mg or 20 mg) or placebo, and film coated tablets were manufactured by CosmaxBio, Inc. (Jecheon, Korea) for this study. The ingredients, manufacturing process, and facility were approved by the Korean FDA. All participants were advised to avoid aggressive changes in dietary habits and excessive alcohol consumption (less than 30 g of ethanol per day). They were also instructed to avoid intense exercise (less than 30 min per day at 60–80% maximum capacity). If subjects had a sedentary lifestyle before the study, they were encouraged to balance their lifestyle during the policosanol consumption period to avoid bias due to exercise or other lifestyle habits.On the visit day, anthropometrical parameters such as height, body weight, body mass index (BMI), subcutaneous fat (kg), and visceral fat mass (kg) were individually measured for each participant. The parameters were measured at the same time of day at 4-week intervals using an X-scan plus II body composition analyzer (Jawon Medical, Gyeongsan, Korea).Measurement of BP was carried out between 8 a.m.–12 p.m. on visit day because the participants had to do overnight fasting for blood collection. BP was measured using three measuring instruments at each visit and the average was recorded at 4-week intervals. First, we used a mercury sphygmomanometer for manual measurements by a licensed technician (S.-J.K.). Second, digital BP device (Omron HBP-9020, Kyoto, Japan) was employed. Third, SphygmoCor system (AtCor Medical, Sydney, Australia) was employed to measure peripheral and central aortic BP [21]. Central (aortic) BP provides more useful prognostic information than peripheral BP due to the proximity to important organs such as the heart, brain and kidneys. A meta-analysis of several longitudinal studies revealed that quantification of central aortic BP had good clinical significance and was a better predictor of cardiovascular events [22]. Using the SphygmoCor system, mean arterial pressure (MAP) was also measured, which defines the pressure during a single cardiac cycle, and was estimated using the formula (aortic SBP + aortic DBP x 2)/3.After overnight fasting, blood was collected from each participant on the visit day. For plasma collection, blood was collected in a vacutainer (BD Biosciences, Franklin Lakes, NJ, USA) containing EDTA (final concentration of 1 mM) at weeks 0 and 12 by low-speed centrifugation (3000g) and stored at −80 °C until analysis. Total cholesterol (TC), triglyceride (TG), HDL-C, and glucose levels were measured in plasma using commercially available kits (Cleantech TS-S; Wako Pure Chemical, Osaka, Japan). Plasma aldosterone was measured by radioimmunoassay (RIA) using an instrument (1470-Gamma Counter, PerkinElmer (Waltham, MA, USA) at the Seegene Medical Foundation (Seoul, Korea). All values are expressed as the mean ± SD (Standard deviation) in the tables. SBP and DBP and blood profile were distributed normally as assessed by a Shapiro-Wilk test in Table 1 and Table 2. The differences in the placebo or policosanol 10 mg and policosanol 20 mg among the groups and over the follow up time were compared using repeated measures ANOVA with peripheral SBP, DBP and lipid profile, as listed in Table 3. When the ANOVA test for repeated measures was significant, the least significant difference (LSD) test was applied for post hoc pairwise multiple comparisons within the four paired means (0, 4, 8, and 12 weeks) and among the three groups (placebo, policosanol 10 mg, and policosanol 20 mg). The Student’s t-test for paired samples was used to compare the mean body composition, peripheral BP, aortic BP, and blood profile, as listed inTable 1, Table 2 and Table 4. The Student’s t-test (Table 1, Table 2 and Table 4) was used to compare the beginning of the study (at 0 week), which was used for homogeneity, with the middle of the study (4, 8, and and 12 weeks) to support the results of repeated measures ANOVA (Table 3 and Table 5).In Figure 2, the data reveals a change in difference in all groups from a 100% initialized BP as mean ± (standard error of mean) SEM. A correlation study was performed using a Pearson’s test. Differences with a p value of <0.05 were considered significant and the statistical software SPSS (version 23.0; SPSS, Inc., Chicago, IL, USA) was used for statistical analysis. As shown in Table 1 and Table 2, there were no significant changes in BMI and body fat contents after 8- or 12-week consumption in all groups; values remained steady with BMI 22–24 kg/m2 and 1.6–2.1 kg of visceral fat. Based on the average of three methods of measurement of peripheral BP at week 8 (Table 1), the policosanol 20 mg group showed the greatest reduction of 6.1% and 6.0% for SBP and DBP, respectively, compared with week 0. SBP was reduced from 136.3 ± 6.1 mmHg (week 0) to 128.4 ± 9.1 mmHg (week 8, p < 0.001) and DBP was reduced from 84.2 ± 7.3 mmHg (week 0) to 79.5 ± 7.8 mmHg (week 8, p < 0.001). The policosanol 10 mg group showed 4.5 % and 3.6% reduction of SBP and DBP, respectively, at week 8 compared with week 0. The placebo group showed no change in brachial SBP and DBP after 8 weeks. We also quantified the confidence interval (CI, usually 95%) of blood pressure values in the different groups to measure the range that provides the variability of the observed population statistic (precision) and to report the probable relationship in the population from which the data were extracted (accuracy). For the placebo group, the measured CI at week 0 and week 8 for the peripheral and aortic pressure were SBP 95% (−0.24, 6.42), (1.68, 10.2) and DBP 95% (−1.78, 3.87), (−0.08, 5.12). For the policosanol 10 mg group, the measured CI at week 0 and week 8 for the peripheral and aortic pressure were SBP 95% (1.81, 10.2), (4.26, 12.4) and DBP 95% CI (0.12, 5.96), (0.40, 6.76). For the policosanol 20 mg group, the measured CI at week 0 and week 8 for the peripheral and aortic pressure were SBP 95% (5.53, 10.26), (3.54, 9.76) and DBP 95% (2.85, 6.52), (1.28, 7.06).As shown in Table 2, the policosanol 20 mg group exhibited the most remarkable decrease in average SBP from 136.3 ± 6.1 mmHg at week 0 to 125.8 ± 8.7 mmHg at week 12 (7.7% reduction; p < 0.001). The policosanol 10 mg group showed a 6.1% reduction in SBP from 135.8 ± 12.3 mmHg at week 0 to 127.7 ± 9.6 mmHg at week 12 (p < 0.001), while the placebo group showed no change in BP between week 0 (134.4 ± 8.8 mmHg) and week 12 (132.1 ± 10.2 mmHg). For the placebo group, the measured CI at week 0 and week 12 for the peripheral and aortic pressure were SBP 95% (−1.52, 6.13), (−0.077, 8.83) and DBP 95% (−0.65, 5.28), (−0.38, 6.86). For the policosanol 10 mg group, the measured CI at week 0 and week 12 for the peripheral and aortic pressure were SBP 95% (4.25, 11.91), (5.46, 12.61) and DBP 95% (0.44, 6.96), (−0.817, 6.81). For the policosanol 20 mg group, the measured CI at week 0 and week 12 for the peripheral and aortic pressure were SBP 95% (8.06, 12.96), (7.31, 12.6) and DBP 95% (3.72, 8.55), (1.20, 7.83).As shown in Table 3, from repeated measures ANOVA with SBP data, the policosanol groups (10 mg, 20 mg) showed significant differences from the placebo in point of time (p < 0.001) and the time and group interaction (p = 0.012). Further analysis with repeated measures ANOVA showed that both policosanol 10mg (p < 0.001) and 20 mg (p < 0.001) groups showed significant differences within the group, whereas the placebo group showed no significance within the group (p = 0.286). One-way ANOVA between the groups showed no significant difference among the three groups at 0 week (p = 0.753), 4 weeks (p = 0.770), and 8 weeks (p = 0.631), except for 12 weeks (p = 0.020). Using the least significant difference (LSD) method as a post-hoc test with the 12-week data, the policosanol 20 mg group showed a significant difference compared to the placebo group, whereas the 10 mg group showed no difference. Within each policosanol group, multiple comparisons using the LSD method revealed the policosanol 10 mg group to show a significant reduction of SBP at 4 weeks (p = 0.0004), 8 weeks (p = 0.007), and 12 weeks (p = 0.0002) compared to that at 0 week. The policosanol 10 mg group also showed a significant decrease in SBP at 12 weeks (p = 0.025) compared to that at 4 weeks. The policosanol 20 mg group showed a more significant reduction of BP at 4 weeks (p < 0.001), 8 weeks (p < 0.001), and 12 weeks (p < 0.001) compared to that at 0 week. The policosanol 20 mg group also showed a significant reduction of BP at 8 weeks (p = 0.021) and 12 weeks (p < 0.001) compared to that at 4 weeks. In addition, the policosanol 20 mg group showed a significant reduction of SBP at 12 weeks (p = 0.007) compared to that at 8 weeks.From repeated measures ANOVA with DBP data (Table 3), the policosanol groups (10 mg and 20 mg) showed a significant difference from the placebo in the point of time (p < 0.001) and time and group interaction (p = 0.023). Further repeated measures ANOVA revealed significant differences between the policosanol 10mg (p = 0.028) and 20mg (p < 0.001) groups, while placebo group showed no significance within group (p = 0.270). One-way ANOVA analysis between the group showed that the three groups were similar at 0 week (p = 0.831) and 4 weeks (p = 0.532). On the other hand, the same analysis revealed a significant difference at 8 weeks (p = 0.041) and 12 weeks (p = 0.035). Using LSD method as a post-hoc test with the 8-week and 12-week data, the policosanol 20 mg group showed significant differences compared to the placebo group. Within the each policosanol group, multiple comparisons using the LSD method showed that the policosanol 10 mg group had a significant reduction of DBP at 4 weeks (p = 0.021), 8 weeks (p = 0.042), and 12 weeks (p = 0.028) compared to that at 0 week. The policosanol 20 mg group showed a more significant reduction of DBP at 4 weeks (p = 0.002), 8 weeks (p < 0.001), and 12 weeks (p < 0.001) compared to that at 0 week. The policosanol 20 mg group also showed a significant reduction of DBP at 8 weeks (p = 0.002) and 12 weeks (p < 0.001) compared to that at 4 weeks. These results suggest that 20 mg of policosanol is significantly more effective in lowering the SBP and DBP than 10 mg of policosanol with dose responsiveness according to repeated measures ANOVA. Figure 2 depicts the changes in the peripheral blood pressures (SBP and DBP 100 % initialized) during 12 weeks consumption of policosanol 10 mg, 20 mg, and placebo group. Between group comparisons using the LSD method, as a post-hoc test, showed that the policosanol 20 mg group had a significant reduction of SBP at 12 weeks up to −7.7% (p = 0.020) and a reduction of DBP at 8 weeks (−5.5%, p = 0.041) and 12 weeks (−7.1%, p = 0.035) compared with placebo. In contrast, the policosanol 10 mg group showed no significant difference compared with placebo at each time point, although the 10 mg group showed significant reduction of SBP and DBP from baseline of 0 week.Based on the SphygmoCor measurements, the placebo group did not show significant changes in aortic SBP and DBP during the 12-week consumption. However, between 0 and 8 weeks, the placebo, policosanol 10 mg, and 20 mg groups showed aortic SBP reductions of 4.8%, 6.9%, and 5.6%, respectively. Aortic DBP was reduced significantly in policosanol 10 and 20 mg groups by 4.0% and 4.7% respectively, at week 8. Mean arterial pressure (MAP) was reduced in the policosanol groups in a dose-dependent fashion. In the policosanol 10 mg and 20 mg groups, MAP was reduced by 4.1% and 6.0%, respectively, while the placebo group showed no change (Table 1).In week 12, as shown in Table 2, the placebo group had a similar aortic SBP: 122.1 ± 11.8 mmHg (week 0) and 117.8 ± 9.0 mmHg (week 12). In the policosanol 20 mg group, aortic SBP was reduced by 8.3% from 120.7 ± 8.7 mmHg (week 0) to 110.7 ± 9.0 mmHg (week 12, p < 0.001). The policosanol 10 mg group showed also significant reductions by 7.4% in aortic SBP from 123.0 ± 12.5 mmHg (week 0) to 113.9 ± 8.6 mmHg (week 12, p < 0.001). Although aortic DBP was significantly reduced only in the policosanol 20 mg group by 5.1% reduction, however, MAP was significantly reduced by 5.2% and 7.5% in the policosanol 10 mg and 20 mg groups, respectively. Interestingly, the peripheral SBP and DBP, aortic SBP and DBP, and MAP were lowered in a policosanol dose-dependent manner. After 12 weeks, as shown in Table 4, TC in the policosanol groups was significantly reduced by 9.6% and 8.6% by 10 mg and 20 mg, respectively, compared with the TC level in week 0 (Table 4). The placebo group showed an 8.3% increase in TC at week 12. Serum TG and glucose level were not changed in any groups after 12 weeks. However, serum HDL-C was significantly increased by 16% and 20% in the 10 and 20 mg policosanol groups, respectively, while the placebo group was not increased significantly. The percentage of HDL-C in TC (%HDL-C/TC) was also remarkably elevated in the policosanol groups in a dose-dependent manner. In the policosanol 10 mg group, %HDL-C/TC was increased to 25.9 ± 6.5% in week 12 compared with 20.2 ± 4.7% in week 0. In the policosanol 20 mg group, %HDL-C/TC was increased to 26.6 ± 6.0% at week 12 from 21.3 ± 8.4% at week 0 (Table 4). TG/HDL-C level was significantly reduced in the policosanol groups in a dose-dependent manner, while the placebo group displayed no change.  In the policosanol 10 mg group, TG/HDL-C decreased to 3.0 ± 2.5 in week 12 from 3.2 ± 2.7 in week 0, while policosanol 20 mg decreased TG/HDL-C to 2.4 ± 1.6 in week 12 from 3.4 ± 3.6 in week 0. LDL-C level increased by 10% in the placebo group after 12 weeks, while the policosanol groups showed remarkable decreases of 20% and 18% in LDL-C for the 10 and 20 mg groups, respectively. As shown in top panel of Table 5, from repeated measures ANOVA with TC data, the policosanol groups (10 mg, 20 mg) showed only significant differences from the placebo in point of time × group interaction (p = 0.009). Although there were no difference in group and time, however, the time × group interaction was significantly different. Further analysis with repeated measures ANOVA showed that both policosanol 10 mg (p = 0.010) and 20 mg (p = 0.024) groups showed significant differences within the group, whereas the placebo group showed no significance within the group (p = 0.203). One-way ANOVA between the groups showed no significant difference among the three groups at 0 week (p = 0.915), 4 weeks (p = 0.169), and 8 weeks (p = 0.167), except for 12 weeks (p = 0.0005). Using the least significant difference (LSD) method as a post-hoc test with the 12-week data, the policosanol 10 mg group (p = 0.004) and 20 mg group (p = 0.001) showed a significant difference compared to the placebo group. Within each policosanol group, multiple comparisons using the LSD method revealed the policosanol 10 mg group to show a significant reduction of TC only at 12 weeks (p = 0.001) compared to that at week 0. The policosanol 10 mg group also showed a significant decrease in TC at 12 weeks (p = 0.010) compared to that at 4 weeks. The policosanol 20 mg group showed a more significant reduction of TC at 4 weeks (p = 0.005), 8 weeks (p = 0.015), and 12 weeks (p = 0.029) compared to that at week 0. These results suggest that policosanol groups showed significant reduction of TC with time dependent manner, while placebo group was not. As shown in middle panel of Table 5, from repeated measures ANOVA with LDL-C data, the policosanol groups (10 mg, 20 mg) showed significant differences from the placebo in point of group (p = 0.035), time (p < 0.001), and time × group interaction (p < 0.001), suggesting that LDL-C was significantly and distinctly reduced in policosanol groups. Further analysis with repeated measures ANOVA showed that both policosanol 10mg (p = 0.001) and 20 mg (p < 0.001) groups showed significant differences within the group, whereas the placebo group showed no significance within the group (p = 0.103). One-way ANOVA between the groups showed no significant difference among the three groups at 0 week (p=.916), 4 weeks (p = 0.226), and 8 weeks (p = 0.148), except for 12 weeks (p < 0.001). Using the least significant difference (LSD) method as a post-hoc test with the 12-week data, the policosanol 10 mg group (p = 0.00005) and 20 mg group (p = 0.0001) showed a significant difference compared to the placebo group. Within each policosanol group, multiple comparisons using the LSD method revealed the policosanol 10 mg group to show a significant reduction of LDL-C at 12 weeks (p < 0.001) compared to that at week 0. The policosanol 10 mg group also showed a significant decrease in LDL-C at 12 weeks (p = 0.001) compared to that at 4 weeks. The policosanol 20 mg group showed a more significant reduction of LDL-C at 4 weeks (p = 0.008), 8 weeks (p = 0.0002), and 12 weeks (p = 0.0001) compared to that at week 0. These results suggest that policosanol groups showed significant reduction of LDL-C with time dependent manner, while placebo group was not.As shown in bottom panel of Table 5, from repeated measures ANOVA with %HDL-C data, the policosanol groups (10 mg, 20 mg) showed significant differences from the placebo in point of time (p < 0.001), and time × group interaction (p = 0.003). Further analysis with repeated measures ANOVA showed that both policosanol 10 mg (p < 0.001) and 20 mg (p < 0.001) groups showed significant differences within the group, whereas the placebo group showed no significance within the group (p = 0.311). One-way ANOVA between the groups showed no significant difference among the three groups at 0 week (p=.684), 4 weeks (p = 0.918), and 8 weeks (p = 0.769), except for 12 weeks (p = 0.035). Using the least significant difference (LSD) method as a post-hoc test with the 12-week data, the policosanol 10 mg group (p = 0.048) and 20 mg group (p = 0.014) showed a significant difference compared to the placebo group. Within each policosanol group, multiple comparisons using the LSD method revealed the policosanol 10 mg group to show a significant elevation of %HDL-C at 8 weeks (p = 0.002) and 12 weeks (p < 0.001) compared to that at week 0. The policosanol 10 mg group also showed a significant elevation in %HDL-C at 8 weeks (p = 0.026) and 12 weeks (p = 0.00007) compared to that at 4 weeks. The policosanol 20 mg group showed a more significant elevation of %HDL-C at 8 weeks (p = 0.00001), and 12 weeks (p = 0.00004) compared to that at week 0. The policosanol 20 mg group also showed a more significant elevation of %HDL-C at 8 weeks (p = 0.0002), and 12 weeks (p < 0.001) compared to that at week 4. These results strongly suggest that policosanol groups showed significant elevation of %HDL-C with time dependent manner, while placebo group was not. Serum aldosterone was remarkably decreased in the policosanol groups by 35% and 24% for the 10 and 20 mg groups, respectively, while the placebo group showed no change after 12-week consumption.Pearson’s correlation analysis of peripheral and aortic blood pressure with lipid profile after 12 weeks of placebo and policosanol treatment is shown in the supplementary Tables (Tables S1–S3). After 12 weeks of placebo group, peripheral blood pressure and aortic pressure were significantly correlated with MAP. However, in week 0, the peripheral DBP was negatively associated with % HDL-C. (Table S1). After 12 weeks of therapy with 10 mg of policosanol, the correlation among peripheral, aortic pressure and the lipid profile was improved. The peripheral SBP was positively correlated with TG, TG/HDL, and MAP. Moreover, the peripheral DBP was negatively correlated with % HDL (Table S2). In the same group, the aortic SBP was also positively correlated with TG/HDL and MAP (Table S2). The correlation was improved for participants in the 20 mg policosanol group. A similar trend was seen for the correlation between different blood pressure measurements and lipid parameters in participants who consumed 20 mg of policosanol for 12 weeks, especially the peripheral blood pressure (Table S3). The aortic blood pressure was correlated with MAP. In week 0, the peripheral DBP and aortic DBP were significant with TG, TG/HDL, and MAP (Table S3).Hypertension is closely linked to the incidence of metabolic syndrome [23], which involves abdominal obesity, high serum TG level, low HDL-C level, and insulin resistance. Treatment of prehypertension is very important to reduce the risk of stroke, coronary artery disease, impairment of cognitive function, and chronic kidney disease [24]. It is also well known that hypertensive people have higher serum levels of TC and LDL-C and lower serum levels of HDL-C than normotensive subjects [7,25].In the current study, at week 0, all participants in this study who completed 12 weeks consumption (n = 76) had similar BP (around 135.6 ± 9.1 mmHg of SBP and 84.6 ± 7.5 mmHg of DBP), serum TC (188 ± 20 mg/dL) and LDL-C (130 ± 32 mg/dL). However, policosanol consumption for 12 weeks lowered the peripheral BP, aortic BP, MAP, and serum levels of TC and LDL-C in a dose-dependent manner. In particular, the policosanol 20 mg group showed a remarkable reduction in peripheral SBP and DBP in a time-dependent manner between each group of different doses compared with the placebo group after 12 weeks from repeated measurement ANOVA (Table 3). The policosanol 20 mg group maintained the significantly lower DBP compared with placebo after 8 weeks. These outcomes are in good agreement with our recent reports, which showed that consumption of policosanol for 8 or 24 weeks lowered BP and serum TC and LDL-C via CETP inhibition and enhancement of HDL functionality [17,19]. The improvement in the serum lipid profile in the current study correlates well with enhancement of HDL functionality in our previous reports [17,19]. A contemporary meta-analysis of 22 studies also described that policosanol could be used to lower lipid levels and to safely elevate HDL-C levels [26]. The LDL-C lowering by policosanol can contribute to the decrease in aldosterone production via inhibition of scavenger receptor-mediated signaling as suggested previously [9,10]After 8 weeks, the policosanol 10 mg group showed 3.4% reduction in peripheral DBP; however, the policosanol 20 mg group exhibited a 5.5% reduction in peripheral DBP (Figure 2). Interestingly, there was similar efficacy between the policosanol 10 mg and 20 mg groups to lower serum TC and LDL-C and raise HDL-C. Indeed, at 12 weeks, the policosanol 20 mg group (7.5% reduction from 0 week) showed greater reduction in MAP than the 10 mg group (5.2% reduction from 0 week). This result shows that policosanol improved not only peripheral BP, but also helped to alleviate aortic BP and MAP, especially 20 mg of policosanol. A higher MAP is correlated with higher incidence of hypertension [27,28] and metabolic syndrome [29].In conclusion, 12-week consumption of policosanol resulted in significant reductions of peripheral SBP and DBP, aortic SBP and DBP, mean arterial pressure (MAP), and serum TC and LDL-C with elevation of %HDL-C. The BP-lowering efficacy of policosanol is also well correlated with reductions in serum levels of TG and TG/HDL-C.The following are available online at https://www.mdpi.com/1660-4601/16/5/809/s1, Table S1: Pearson’s correlation analysis between at the baseline and after 12 weeks in the placebo group, Table S2: Pearson’s correlation analysis at the baseline and after 12 weeks of consuming policosanol 10 mg, Table S3: Pearson’s correlation analysis at the baseline and after 12 weeks of consuming policosanol 20 mg. Data curation, H.-J.P., D.-J.J. and S.-J.K.; Funding acquisition, S.-J.K., M.-A.B. and J.-R.K.; Supervision, K.-H.C.; Writing—original draft, K.-H.C.; Writing—review & editing, D.Y. and K.-H.CThis work was supported by a grant from the Ministry of Trade, Industry and Energy, Korea (grant no. 2016-10063396) and the Medical Research Center Program (2015R1A5A2009124) through the National Research Foundation (NRF), funded by the Ministry of Science, ICT and Future Planning of Korea.The authors declare no conflict of interest.Design of study and participants. Inclusion criteria were normolipidemic, normoglycemic, and healthy subjects who had prehypertension (systolic 130–139 mmHg, diastolic 80–89 mmHg). M: Male; F: Female.Change of peripheral blood pressures (SBP and DBP) during 12-week consumption of policosanol groups and placebo group from repeated measurement ANOVA. Data are expressed as mean ± SEM. SBP: systolic BP; DBP: diastolic BP; SEM: standard error of mean.Change of blood pressures after 8-week policosanol consumption.Data are expressed as mean ± SD. * p < 0.05 vs. week 0; ** p < 0.01 vs. week 0; *** p < 0.001 vs. week 0 in each group. M: male; F: Female; BP: blood pressure; SD: Standard deviation.Change of blood pressures after 12-week policosanol consumption.Data are expressed as mean ± SD. * p < 0.05 vs. week 0; ** p < 0.01 vs. week 0; *** p < 0.001 vs. week 0 in each group.Repeated measures ANOVA of peripheral systolic BP and diastolic BP between the three groups.a data of 0 week; b data of 4 week; c data of 8 week; d data of 12 week.Change of serum lipid profile after 12-week policosanol consumption.Data are expressed as mean ± SD, * p < 0.05 vs. week 0; ** p < 0.001 vs. week 0 in each group. TC, total cholesterol; TG, triglyceride; HDL-C, High-density lipoprotein-cholesterol; LDL-C, low-density lipoprotein-cholesterol.Repeated measures ANOVA of serum TC, LDL-C, and %HDL-C profile between the three groups.Data are expressed as mean ± SD. TC, total cholesterol; HDL-C, High-density lipoprotein-cholesterol; LDL-C, low-density lipoprotein-cholesterol. a data of 0 week; b data of 4 week; c data of 8 week; d data of 12 week.

Med-MDPI/ijerph_3/ijerph-16-05-00810.txt

@@ -0,0 +1 @@
+Different studies around the world indicate that the percentages of overweight and obesity in childhood and adolescence are high. In this context, it would be useful to have a common, valid, and reliable instrument to assess health behaviors of families that allows comparisons of data from different countries. The objective is the adaptation of a Spanish version of the Family Health Behavior Scale (FHBS). The questionnaire originally developed by Moreno group was translated and adapted following the International Test Commission protocol. Its psychometric properties were evaluated through analysis of internal consistency, factor analysis and other evidences of validity. The Spanish version of the FHBS demonstrated adequate reliability coefficients, and its factor structure sufficiently replicated that obtained by the original measurement. The results suggested that the adapted version of the questionnaire was an adequate and valid measure for the evaluation of family health behaviors related to the prevention of overweight and obesity.Overweight and obesity rates in children and adolescents are currently a concern both in developed countries and in those with emerging economies [1]; due to the complications and associated risks they may pose, as well as the impact on physical [1,2] and psychological [3,4,5] quality of life of these future adults. Similar warnings have been raised in Spain, where different studies indicated that the percentages of overweight and obesity in child and adolescent populations have stabilized at high levels [6,7] and are increasing at lower ages and at lower economic levels [8].Assuming overweight and obesity are largely preventable [9], the final report of the Commission on Ending Childhood Obesity stated, “Academic institutions can contribute to addressing childhood obesity through studies on biological, behavioural, and environmental risk factors and determinants, and the effectiveness of interventions in each of these.” ([10], p. 37). To prevent instead of acting once overweight and obesity have appeared, families’ healthy habits have an important role to play. In this sense, a significant relationship has been found between the intergenerational transmission of obesity and family lifestyles that cannot be explained exclusively by genetic components [11,12]. Therefore, early detection and intervention on unhealthy family behaviors would help to reduce the prevalence of weight problems.Health professionals have been concerned with the assessment and measurement of behavioral factors and habits related to overweight in childhood and adolescence. Between the several instruments of the parents’ mealtime routines that Vaughn, Tabak, Bryant, and Ward [13] have revised, the Family Health Behavior Scale (FHBS) developed by Moreno et al. [14] allows an efficient assessment of family behaviors related to obesity in childhood during primary and secondary education years. It contains 27 items and provides information on four areas: Parent Behaviors, Physical Activity, Child Behaviors, and Mealtime Routines. The questionnaire evidenced adequate validity and reliability data for USA population.From an international perspective, the problem with the evaluation and measurement of behavioral and social factors is that they are strongly conditioned by cultural and linguistic determinants, unlike physical and biological measures. This makes the comparative use between countries of behavioral scales difficult. The existence of a common, brief, and valid instrument to assess family health behaviors would be useful to understand the possible role of countries socio-cultural factors on children and adolescents’ overweight and obesity. Instruments as the FHBS could play that role, but there is no similar questionnaire in Spanish that briefly and effectively measures family behavioral habits related to unhealthy children’s weights. In order to perform this assessment of family health behaviors in Spain with a reliable and valid instrument, we consider the adaptation and validation of the FHBS to the Spanish language and culture as the objective of this study. The sample recruitment was tailored to the characteristics of the original sample used by Moreno et al. [14] in the development and validation of the original FHBS scale in the USA. Our sample was composed of 360 caregivers of children of school age ranging from 5 to 12 years (M = 8.4, SD = 2.5), of which 44.4% were girls.Caregivers provided the age, sex, height, and weight of the children. The children’s body mass indexes (BMI) adjusted for the sex and age were calculated using the application developed by Romero [15]. The median of the BMI percentile of the minors was 59.1 (SD = 31.6). According to these percentiles, 64.3% of the children showed a healthy weight, 13.0% were overweight, 13.8% were obese, and 8.9% were underweight. No statistical differences were found between the distributions of BMI groups in both samples (K-S Test D = 0.5 p = 0.699). Table 1 compares the data of the sample of children participating in this study with those of the original sample [14].Parents or guardians communicated data on their children and adolescents’ age, sex, height, and weight. The family health habits were measured with the FHBS questionnaire developed by Moreno et al. [14] that evaluates the healthy behaviors of the family related to food and physical exercise. The scale has 27 items divided into four subscales, one referring to Parent Behaviors (10 items), and the remaining three referred to Physical Activity (6 items), Mealtime Routines (5 items), and Children Behaviors (6 items). Each item is answered from a five-point Likert scale that goes from 0 (almost never) to 4 (nearly always). The total scores per subscale are obtained by adding the item scores, after reverse scoring the negative items, so that a higher score means a higher frequency of healthy behaviors.For the process of adaptation and validation of the FHBS questionnaire to Spanish, the criteria proposed by the International Test Commission [16] and Hambleton [17] were followed. Linguistic equivalence was sought through a process of translation and retro-translation. Cultural equivalence was assessed through interviews and consultations with experts. Finally, the statistical equivalence was verified analysing the psychometric properties of the questionnaire administered to a Spanish sample with similar characteristics to those of the original study in the USA.Once the agreement of the original authors was obtained, a professional translated the items of the FHBS from English to Spanish. After comparing the Spanish translation with the original version of the questionnaire, a pilot version was obtained. This pilot version of the questionnaire was administered in individual interviews with parents of children enrolled in primary education. Parents were asked to review the items, identify any words or concepts in the questionnaire that they did not understand. After these contributions, slight modifications were made. In parallel, experts in pediatrics and nutrition from a third level hospital were consulted to evaluate the content of the items of the translated questionnaire. The most relevant change that occurred with respect to the original questionnaire was that the item “My child eats three meals a day” was modified to “My child eats four or five meals a day”, adjusting it to Spanish cultural habits. Once the final version of the questionnaire was prepared, it was sent to another translator to carry out a retro-translation from Spanish to English. With the exception of the aforementioned item, it was found that both the literal meaning and the sense of the items of the Spanish version were coincident with the original. In this way, the linguistic and semantic equivalence of the Spanish version of the FHBS questionnaire with respect to the original was assumed, and it was considered ready to be administered.Caregivers of children enrolled in primary schools of the province of Seville, Southern Spain, were recruited. All the schools were from urban, middle-class populations located in the metropolitan area of Seville (688.171 inhabitants). Most of them contacted through a face-to-face procedure, in which undergraduate students of the University of Seville went to the participating schools. Only a small sample (n = 27) were contacted through the Internet once the school principals authorized the survey, without apparent impact on the resulting data. In all cases, together with the distribution of the FHBS questionnaires, the parents or guardians of the minors provided informed consent to participate. The Ethics Committee of the institution where researchers collaborate authorized the study (Virgen del Rocio Hospital, Seville). The analysis techniques used were: (a) descriptive analysis of the items; (b) analysis of the internal consistency of the subscales and the total scale of the FHBS using Cronbach’s alpha coefficients; (c) factor analysis through the FACTOR program (Ver. 10.3.01 XP [18,19]); (d) calculation of factor scores, reliability estimates and their correlations and with the BMI measures; and (e) binomial logistic regression analysis to predict from the factor scores the dichotomized classification between the healthy weight group against overweight and obese groups.Most of the items presented mean values above 2 in the items’ scale from 0 to 4, except in the case of item number 5 (“My child eats frequently during the day”, from the Children Behaviors scale) whose 95% confidence interval around their mean did not exceed the threshold. This is consistent with the analysis of corrected item-total correlations, which only in item 5 showed a clearly inadequate level (r = 0.053). In all the items the maximum range of responses was found, with 95.8% of the sample (345 subjects) answering all the items and only 24 blank responses appeared, distributed in 16 of the items. As for the skewness and kurtosis of the responses to the items, it was found that in 17 of the items the absolute value of 1.00 was exceeded in the asymmetry indexes, in the kurtosis indices or in both, indicating that their distributions do not correspond to those of a normal curve. The multivariate analysis of Mardia [20], corroborated statistically the absence of normality in the case of kurtosis (B = 34.21, p < 0.0001). After inverting the scores of the negative items, new scores were obtained for each subscale, following the original item distribution of Moreno et al. [14], as well as for the total of the FHBS. To facilitate the interpretation of these scores, they were linearly transformed to scales from 0 to 100. The descriptive results of these scores show averages between 59.08 and 70.65, except for the Subscale of Mealtime Routines that rises to 91.98. This subscale is also the one with the lowest standard deviation (SD = 10.14). Regarding the distributions of these scores, in all cases, the skewness and kurtosis indexes are lower than an absolute value of 1.00, except for the Mealtime Routines subscale, which has a clearly skewed distribution towards the higher values.Cronbach’s alpha coefficients showed sufficient internal consistency for both the FHBS Total scale (α = 0.746) and the Parent Behaviors subscale (α = 0.761). In the rest of the subscales, the value of 0.70 was not reached, being especially low the value of the subscale of Mealtime Routines (Table 2).Given the behavior in terms of skewness and kurtosis of the items, a factor analysis was made from the polychoric correlation matrix, as recommended by Lorenzo-Seva and Ferrando [18,19]. This analysis was limited to four factors to assess the validity of the model obtained by Moreno et al. [1], using the procedure of robust weighted squares (Unweighted Least Squares, ULS) and a method of oblique rotation Promin [21] to achieve greater factorial simplicity. As a final criterion, the solution was limited to a maximum of 100 iterations or to a convergence value of 0.00001.The analysis of the adequacy of the correlation matrix was considered satisfactory, according to the Bartlett sphericity test (M = 0.00134, χ2 = 2208.5, df = 351; p = 0.00001), and the Kaiser–Meyer–Olkin index (KMO = 0.70537). This result justified the limitation imposed on the four-factor solution, which explained up to 46.3% of the variance found in the 27-item questionnaire. The adjustment of the four-factor model was satisfactory (GFI = 0.96; RMSR = 0.06), showing an adequate Bentler’s simplicity index (S = 0.98).As can be seen in the rotated factor matrix of factorial loadings (Table 3), the content of the four factors extracted coincided with those of the original work: Children Behaviors, Physical Activity, Mealtime Routines, and Parent Behaviors. Considering only the items with absolute loading greater than 0.40, following the same criteria as Moreno et al. [14], the classification of the items coincided in 81.5% of cases with the original (22 of 27 items). Of the five items whose results do not coincide with the original solution, in four of them it is due to their factor loadings being lower than 0.40 (items 3, 5, 24, and 27), while one item loaded on the Mealtime Routines factor rather than the Parent Behaviors factor (item 4: “My child has help choosing healthy foods”).Unlike the factorial solution found by Moreno et al. [14] the Child Behaviors factor explained the highest proportion of variance (24.0%), while the Parent Behaviors factor explained the least amount of variance (8.2%). Reliability estimators based on factor analysis (MacDonald’s Omega, Ω) show all satisfactory values for the four factors, and higher than 0.79 in all of the factor scores.The analysis of the correlations (Table 4) between the scores of the different factors showed that the Child Behaviors factor was not related with the other three factors (Parent Behaviors, Mealtime Routines, and Physical Activity), which were related with each other. The correlations of the scores on the four factors of FHBS with the BMI were negative and significant for the subscales of Parent Behaviors, Physical Activity, and Children Behaviors, which shows that lower scores on healthy family habits are associated with a slight increase in BMI.The analysis of the means and 95% confidence intervals of the factor scores in the four factors according to the different groups of classification of the BMI (underweight, healthy weight, overweight, and obese), is shown in Figure 1. These scores were typified with an average around 0.00 and a standard deviation of 1.00. The scores of the groups of children with overweight and obesity were below the average (negative), while the groups of healthy and underweight were above the average (positive), being in all cases the lowest scores for the obese group, indicating more obesogenic family health habits. Figure 1 shows a curved descending pattern from the healthy weight group to the obese group in the family health habits. The underweight group scores similar to the healthy weight group. In any case, the distribution of the means of the scores of Child Behaviors and Mealtime Routines was overlapping between the different groups taking into account the 95% confidence intervals around the means showing no discriminative power. However, in the case of the scores of the Physical Activity and Parent Behaviors, the healthy weight group showed an average outside the confidence interval of the obese group.With the aim of replicating the analysis of the possible validity of measures of family health habits to predict overweight and obesity, a binomial logistic regression model was carried out following the procedures of Moreno et al. [14]. As in the original study, the children classified with underweight were excluded, resulting in a sample size of 316 for this analysis. As a criterion variable, we differentiated two groups: those classified as overweight and obese on one side, from those classified as healthy weight. Using this classification and the total scores in the FHBS scale as a predictor variable, the results indicated that each point of increase in the FHBS scale implied a decrease of 4.3% in the probability of being in the overweight or obese group (OR = 0.957, 95% CI 0.934–0.980; p < 0.001). In comparison, the original study demonstrated a 3.9% decrease in the likelihood of being overweight or obese for every point increase. This model correctly predicted the BMI classification of 71.5% of the participants, which was higher than the 62.8% found in the original study.Bivariate correlation analyses were performed to determine if the zBMI was significantly related to the subscales and the total score of the FHBS. For these analyses, those classified as underweight or healthy weight were excluded, in order to replicate the original study [14]. This resulted in a subsample of 93 children with a BMI equal to or greater than the 85th percentile. The results indicated that in this group zBMI was only significantly and inversely related to the Physical Activity scale (r = −0.279; p = 0.008). The scales of Parent Behaviors, Children Behaviors, and Mealtime Routines, as well as the total scores in the FHBS, did not correlate significantly with zBMI (with values between −0.163 and 0.051). The analysis of the bivariate correlations between zBMI and the scores obtained from the factor analysis offered similar results.The Spanish adaptation of the Family Health Behavior Scale [14] proved to be adequate and useful as an instrument to detect behaviors related with the risk of overweight or obesity. It is a short scale to be used by any professional who works with children and wanted to assess their family health habits. The comparisons with the results of the original study of Moreno et al. [14] showed similarities and differences. The classification of the items obtained from the factor analysis reproduces sufficiently the content of the items of the original study, with four clearly differentiated areas: Child Behaviors, Parent Behaviors, Physical Activity, and Mealtime Routines. The results of the factor analysis show good indexes of adjustments and the factor scores offer adequate estimates of reliability, showing as valid alternatives to the direct scores in the FHBS. Regarding the values of internal consistency, these were lower than those of the original study, both in the total scores of the FHBS and in the scores of the different subscales. However, the greatest consistency was found in the measure of Parent Behaviors and Total Scores, as in the original study. Conversely, Mealtime Routines was the least consistent score in the Spanish sample evaluated in this study, showing that perhaps the items of this subscale are not discriminative for this population. In relation to the variance explained by each of the factors, the Child Behaviors measure were the most determinant in the Spanish population, compared to the Parent Behaviors that were the most important in the USA population. It stands out that the subscale of Mealtime Routines did not provide much information in this Spanish sample, since all the subjects demonstrated high scores, failing to discriminate between participants with more or less healthy habits. It may be due to differences associated to Mediterranean culture on meals schedule in Spain, determined by work and school timeframe that facilitate well-established family customs. Especially in southern Spain, it is usual to have shared mealtime routines: e.g., most of the families eat at the table and children do not choose their meals. On the other hand, the correlations analysis revealed significant relationships between the BMI measures and all factors except for Mealtime Routines. Taking into account the four groups considered according to their BMI (underweight, healthy weight, overweight, and obese), the results showed a consistent trend with the hypothesis that worse family health habits were more common among the children with overweight and obesity. In addition, the binomial logistic regression model reproduced similar results to those of the original study. Thus, the total scores in the FHBS significantly predicted the weight classification of the healthy weight group and the overweight and obese groups. When the analysis focused only on the group of children with overweight and obesity, as in the original study, Physical Activity was significantly and inversely related to measures of BMI, showing convergent validity. In contrast, the correlations between the zBMI scores with the total scores of the FHBS and those of the subscale of Parent Behaviors were not significant, unlike the original study. One limitation of this study would be that parents provided the anthropometric measures (height and weight), and because of that, they may be inaccurate. However, this scale can be considered a primary prevention tool because it can detect unhealthy behaviors before the increase in BMI. A follow-up assessment protocol included at different periods could detect the risk of overweight and obesity. An educational program that promotes healthy behaviors such as those included in this scale could keep a healthy weight. Obviously, an evaluation made exclusively with one scale of family behavioral habits is incomplete, so sociodemographic and personal variables must be taken into account. For future studies, it would be desirable to relate the FHBS scores with more precise anthropometric and sociodemographic measurements.We present a valid and reliable instrument applied to a large sample of children with a wide range of age. It is a user-friendly scale to be answered by parents and guardians. It measures four categories of healthy behaviors with only 27 items in Spanish language. As in the original Moreno’s paper, the FHBS helps to discriminate between overweight and healthy weight children and correlates significantly with BMI.The resulting scale should allow for international comparisons with minor language and cultural adaptations regarding family health habits related to obesity and overweight. This scale would be useful to detect and prevent obesity and overweight in any context, such as educational, social, or clinical.All authors read and approved the final manuscript. Conceptualization M.-D.L.-F., J.-F.L.-O., and I.A.-C.; Investigation M.-D.L.-F., J.-F.L.-O., M.G.-d.-T.-G., I.A.-C., and R.J.M.-C.; Methodology and Formal Analysis M.G.-d.-T.-G. and R.J.M.-C.; Writing—Original and Draft Preparation M.-D.L.-F. and R.J.M.-C.; Writing—Review and Editing M.-D.L.-F., J.-F.L.-O., M.G.-d.-T.-G., I.A.-C., R.J.M.-C., and J.P.M.This research was partially supported by grant CTS-152 Comprehensive Paediatrics and Paediatric Psychology, Andalusian Research Plan (Spain).The authors would like to thank the principals and teachers of the primary schools that collaborate with this study, as well as the parents and guardians that provided their children’s data.The authors declare no conflicts of interest.Means and 95% Confidence Intervals of the factor scores for the BMI classification groups.Comparison of the present study and Moreno et al. [14] sample characteristics.Comparison of internal consistency (α) in the present study and Moreno et al. [14].Matrix of rotated factorial loadings, factors’ eigenvalues, explained variances, and reliability estimates.a Items with reversed scores. b Item of the original questionnaire with a different classification.Matrix of correlations between the scores obtained from the factor analysis and the BMI scores.* p < 0.05; ** p < 0.01; N = 360.

Med-MDPI/ijerph_3/ijerph-16-05-00811.txt

@@ -0,0 +1,25 @@
+The Syrian crisis began on 15 March 2011. It is one of the bloodiest and complicated conflicts in the world today. Although almost eight years have passed over this tragedy, civilians continue to suffer from conflicts and destructions in the area. As a result, this situation disregards human life and the number of people in need increases day by day. Particularly, people who have to live in the conflict area encounter troubles with regard to health, shelter, food and other needs. Thus, we have focused on identifying the Primary Health Care Center (PHCC) locations within Idleb Governorate in Syria. Data is extracted from a sample containing 23 sub-districts in the governorate and a total of 338 communities. We have formulated a mixed integer-weighted goal programming model and combined it with a Geographic Information System-GIS (ArcMap). The model is solved via an optimization package and moreover, sensitivity analyses are conducted to achieve a more in-depth study. Our aim was to have 60 PHCCs out of 77 available candidate PHCCs and the model located 42 PHCCs in total, by allocating 379,080 people, with a total cost of USD 1,000,353 and a cash for work amounting to USD 163,549. Accordingly, the model’s outputs and sensitivity analyses are expected to help decision-makers in case of such disasters.Disasters may occur anywhere in the world and may be categorized as natural or man-made disasters. While a natural disaster emerges from natural hazards on earth (such as floods, earthquakes, volcanic eruptions, hurricanes, tsunamis, etc.); man-made disasters are caused by human behaviors. Some of these include, among others, industrial, transport or public health accidents, terrorism, crimes against humanity and warfare. It is crucial to state that warfare may be one of the most devastating types of man-made disasters as it has both severe and long-term implications. Even though the warfare or conflict ends, serious problems such as its damages to the country, environment, infrastructure, healthcare services; the lack of food, water and other resources, displaced people and disease outbreaks may continue to be prevalent [1]. Warfare and other conflicts have materialized many times across the globe. Wars have further damages in addition to the killing of people. Among these damages we can count forced migration and displacement as important ones. Nowadays, The Office of the United Nations High Commissioner for Refugees (UNHCR) estimates that there are approximately 68.5 million people who are subjected to forced displacement at global level worldwide and have been forced to leave their homes because of battles and armed conflicts [2]. Of these 68.5 million people, about 40 million are internally displaced people (IDPs); approximately 25.4 million are refugees and nearly 3 million are asylum seekers [2]. While a refugee is defined as someone who has been forced to run away from his/her own country due to persecution, war or violence, IDPs are described as the people or group of people who have left behind their home, habitual residence and livelihood, but have not crossed any international border [2]. Even though their own country is the source of their displacement, IDPs proceed to wait for protection by their government. The harsh fact about refugees is that the citizens of only three countries constitute more than half of 24.5 million refugees worldwide: South Sudan (about 2.4 million), Afghanistan (nearly 2.6 million) and the Syrian Arab Republic (approximately 6.3 million). This demonstrates that the Syrian Arab Republic has the highest share among these three countries and it has led us to the Syrian Crisis which began in March 2011. It became an international crisis from that day on and accounted for the largest migration of people since World War II. It materialized as a crucial humanitarian emergency through the establishment of refugee camps, seeking for shelter and the presence of thousands of asylum seekers in different countries. These humanitarian emergencies devastate the local government’s capacity to handle and provide affected populations with basic necessities, such as food, clothing, shelter, water, non-food items (blankets, sheets, cooking items, soaps, etc.), security and healthcare facilities. Thus, refugees’ and IDPs’ safety and healthcare needs have evolved into a more and more significant topic all over the world. The Syrian Crisis is a typical example of health challenges encountered by refugees, IDPs and host countries [3]. Due to troublesome conditions under which they live, they are defenseless and various factors including fatigue, the lack of food and clean drinking water as well as poor hygiene influence their health [4]. Hence, the planning of healthcare response to natural and man-made disasters has attracted more attention in the last decades [5,6,7,8,9]. In this respect, it is noteworthy to point out primary healthcare centers (PHCCs) at local level as they are very vital to offer health facilities and services. During a disaster, irrespective of being natural or man-made, PHCCs play a critical role in saving as many lives as possible. They constitute the key structural and operational part of public health services in developing countries, as in Syria. It is difficult to estimate requirements and figure out the allocation of PHCCs and appropriate resources, particularly in the case of any humanitarian emergency like battles and forced migration such as the Syrian case and accordingly, this situation poses an enormous danger. Although forced migration in the area is a long-lasting process and has apparent threats to the health of the population, it is not exactly categorized as a global health issue, yet [10]. On the contrary, responses to the health needs stemming from the excess number of forcibly-displaced people have proven to be insufficient to a large extent.To gain a better understanding of the subject, we have reviewed the related literature under two titles: “primary health care centers in disasters” and “location-allocation problems in disasters.”Before researching literature for primary health care centers in disasters, we reviewed health issues in disasters. For the issue of health within natural disaster settings, studies were rare and we realized that a great majority of studies were based on earthquakes [11,12,13,14,15]. For the issue of health during man-made conflict settings, the studies can be collected under five titles: disaster response [16], health system performance [17,18], health activities and capabilities [19,20,21,22]; psychosocial issues [23,24,25,26] and health requirements [27,28]. In respect to PHCCs within natural disaster settings, some studies focused on earthquakes [29], hurricanes [30] and floods [31]. Other studies included: a national study of connections among health centers and the emergency preparation and response plan attempted in their own societies [32], analyzing the level of disaster preparedness in public hospitals and applying the hazard vulnerability analysis (HVA) tool [33], developing and accessing health care in disrupted societies and studying frameworks in primary care [34], and evaluating health-care providers’ insights of their knowledge, abilities and preparedness for disaster management [35]. In relation to PHCCs in armed conflict settings, some studies focused on: access to water and sanitation [36], non-governmental organizations’ (NGOs) role in delivering therapeutic health services in conflict settings [37], the integration of staff’s well-being into the primary health care (PHC) system [38], non-communicable diseases (NCDs) among refugees [39] and the PHC’s role to promote an incorporated distribution of care to refugees [40]. Furthermore, some studies focused on the scope of mental, neurological and substance use (MNS) services in refugee camps [41]; specifying the problems and summarizing the precedence’s and obstacles in negatively influenced health care systems by armed conflict [42]; medical condition, unfulfilled requirements and the provision of health services among refugees [43] and health workers’ multifactorial behavioral interference to the adults in the primary care settings regarding conflicts [44]. In conclusion, many researchers have studied primary health care subjects in natural and man-made disasters, but there is no literature related to the location-allocation problem regarding primary health care centers.With regard to location-allocation problems in medical settings during or in preparation for disasters, which is also the focus of this study, some studies adopted multi objective models for the purposes of covering as many patients as possible [45] and of minimizing: the travel time of patients [45,46], queue duration [45], the total mortality risk value of patients [46], total response time by taking costs into consideration [47,48], total network access time and the operating costs of a shelter and the cost of failure to accommodate all evacuees [49], the total demand-weighted transportation time between facilities and the cost of lost demand at the points of demand and hospitals [50], relief distribution network [51], expected costs over all scenarios [52] and the total sum of distance plus expected value [53]. Since the Syrian war is currently the most complicated and bloodiest conflicts in the world, the humanitarian community is expected to provide emergency and life-saving PHCCs to the Syrian region apart from all aforementioned reasons. Despite the significance of the issue and the existing studies mentioned above, there is no study handling the PHCC location-allocation problem (identifying the right places to locate PHCCs so as to cover a maximum number of people suffering from health problems in the region with multi objectives) within the Syrian context. This implies that academic and scientific research interest in this field is still insufficient. These factors make this study a valuable contribution to literature. Therefore, due to the importance of the Syrian Crisis and the significance of its indirect global implications in recent years, this paper aims to bridge this gap and to be a valuable resource in such conflict areas by formulating a multi-objective model and integrating it into a real case study in the area. In this multi-objective model, as in [45], we have aimed to maximize the number of people served from the whole demand and minimize overall costs (as in [49]) at the same time. While accomplishing these objectives, we have focused on maximizing the number of located PHCCs with health availability factors (solar power, basement, internet service, laboratory service, blood grouping service and vaccination) and the cash for work amount of the located PHCCs simultaneously, differently from what was previously studied in the literature. These were accomplished through a mixed integer-weighted goal programming model developed for a real case study. The proposed model is solved by an optimization software. Another differentiating feature of this study is the use of real data directly collected from beneficiaries and consultants in the area.As for the real case study, we selected Idleb Governorate, one of 14 governorates in Syria. There is an ongoing inflow of displaced people to the region and there are overcrowded camps and temporary shelters in the area. Livelihood and income are the most pressing needs in the governorate and these lead to difficulties in buying food and other basic items. Furthermore; schools, hospitals and other civilian infrastructure have been destroyed or damaged and access to basic services is deteriorating. Most of the PHCCs in Idleb are operated and funded by a local non-governmental organization or international non-governmental organization in coordination with the health directorate in the area. Although humanitarian non-governmental organizations are trying to relieve the suffering of people, they cannot fully respond to a long-term humanitarian crisis of such magnitude. These issues encouraged us to choose this governorate as the real case study for the purposes of identifying the locations of PHCCs and allocating people in the area to the located PHCCs so as to alleviate the impacts of this man-made disaster. Consequently, the remaining sections of this paper are organized as follows. In Section 2, the materials and methods employed in the study are described in detail. This section covers data collection and model formulation. The developed model is applied to solving the location-allocation PHCC problem in the area. Section 3 contains the results out of the relevant model and sensitivity analyses. Discussions and some suggestions for future studies are specified in Section 4. In the end, final remarks related to our work are indicated in Section 5. Methodology adopted in this study for identifying PHCCs and allocating selected people to these PHCCs is demonstrated in Figure 1. Since direct beneficiaries are the most important stakeholders in this conflict area (and also in this study) and look forward to overcoming the detrimental impacts of the conflict, our approach is mainly based on direct beneficiaries. The methodology used contains four fundamental stages: the first stage of the methodology operated a needs assessment with the help of focus group discussions (FGDs), questionnaires, etc., in order to evaluate the needs of and the most significant issues for all relevant beneficiaries as the authorities required for publishing this data are not present in the conflict area. The data was collected between 5 March 2018 and 30 May 2018. Data regarding Idleb Governorate was extracted from a sample including 23 sub-districts within the governorate and a total of 338 communities by contacting direct beneficiaries and performing key source interviews with people who are aware of the issues in the society.The second stage of the methodology was identifying the objectives and constraints of the model. This stage also involved the identification of priorities and penalties of objectives with regards to the beneficiaries and experts by weighting each objective by means of a multi-criteria decision making technique, Analytic Hierarchy Process (AHP). The third stage of the proposed methodology comprised of building the roads network dataset in the projected field utilizing the Geographic Information System (GIS) and updating it consistently by also covering risky roads with the aim of specifying origin-destination (OD) cost matrix by the distance between nodes (communities) and PHCCs. The fourth and last stage of the model was constructing the mathematical model with the introduction of a mixed integer model and solving it with the optimization package software so as to identify PHCCs and allocate people in the area to these centers. Moreover, sensitivity analyses were performed to discuss the results.The problem addressed in this study was a part of the facility location-allocation problem, a branch of operations research associated with locating or positioning at least a new facility in between a number of existing or candidate facilities. The goal of this operations research specialty was optimizing (maximization/minimization) one or more objective functions such as profit, revenue, cost, travel distance, coverage, etc. Numerous fields of the application comprising private and public facilities, business areas, national and international fields are analyzed in the related literature [54]. Furthermore, facility location is studied by several researchers in terms of humanitarian relief [55,56,57,58,59]. It is possible to encounter multi-objective problems or multi-objective decision making (MODM) problems in the real world. By considering a myriad of interactions within the model, MODM methods endeavor to identify the best alternative which ideally satisfies the decision-maker (DM) by means of accomplishing satisfactory results for a set of objectives [54]. Yet, as in MODM, many real world decision making problems have contradictory objectives and this issue should be analyzed to obtain accurate results. In addition to this, an ideal solution in MODM is described as the one ending with an objective function’s optimum value at the same time within an effectual solution while none of the objective functions can be upgraded without damaging other objectives [60]. Thus, we took into account all these aspects and integrated facility location problems within MODM environment. Such problems might contain several objective functions to be achieved such as minimizing the total cost; maximizing the coverage; minimizing the longest distance from the existing facilities and maximizing the service, etc. In the light of this information, we introduced, through this paper, a mixed integer model incorporating capacitated maximal covering, fixed charge cost and some specific features of health centers in the Syrian context. These features are summarized as follows: Availability factors: Laboratory service,Blood grouping service,Vaccination,Solar power,Basement,Internet service.Laboratory service,Blood grouping service,Vaccination,Solar power,Basement,Internet service.Economic factors such as cash for work: Cash for work corresponds to the wages paid for workers to repair the PHCC and to make it habitable. Assumptions utilized in the model are stated below:Each demand node can be served as an entire unit from a PHCC or not served at all (0 or 1 without any fraction).Amount of each demand and its location are fixed.Paths throughout the updated built road networks are accessible and there is no broken or closed street.Variable costs for each allocated group of people at each location are related to the amount of people allocated in the PHCC irrespective of the PHCC’s location. It means that the cost of allocating a person to a certain PHCC is the same cost of allocating him/her to another PHCC in another location.Each demand node can be served as an entire unit from a PHCC or not served at all (0 or 1 without any fraction).Amount of each demand and its location are fixed.Paths throughout the updated built road networks are accessible and there is no broken or closed street.Variable costs for each allocated group of people at each location are related to the amount of people allocated in the PHCC irrespective of the PHCC’s location. It means that the cost of allocating a person to a certain PHCC is the same cost of allocating him/her to another PHCC in another location.Consistent with the requirements for building the model, this paper uses the following parameters and variables for the multi-objective mixed integer model as shown in Table 1.In the following parts, we demonstrate the multi-objective mixed integer model with objective functions and constraints.Objective functions:(1)Maximize ∑i∈I∑j∈Iaij Zijhi
+        (2)Minimize  ∑j∈JfjXj+R∑i∈I∑j∈J Zijhi+TC∑i∈I∑j∈J Zijhidisij
+        (3)Maximize  ∑j∈JSEjXj
+        (4)Maximize  ∑j∈JBjXj
+        (5)Maximize  ∑j∈JISjXj  
+        (6)Maximize  ∑j∈JLabjXj  
+        (7)Maximize  ∑j∈JBGj Xj
+        (8)Maximize  ∑j∈JVacjXj
+        (9)Maximize  ∑j∈JCWj XjSubject to
+        (10)∑j∈Jxj ≤ P
+        (11)∑j∈Jzij ≤ 1   ∀ i ∈ I
+        (12)∑j∈Jaijzijhi ≤ Kj    ∀ j ∈ J
+        (13)Zij ≤ aij Xj     ∀ i ∈ I , ∀ j ∈ J
+        (14)Xj , Zij∈ [0, 1].Equations (1) to (9) correspond to the objective functions of the proposed multi-objective model. Equation (1) maximizes the number of people served out of the whole demand while Equation (2) minimizes the total cost covering three factors: a fixed cost which is the cost of setting up and opening a PHCC; a variable cost to run a PHCC within a year and a transportation cost to arrive at the related PHCC. Equations (3) to (8) maximize the number of located PHCCs with availability factors: Equation (3) maximizes the number of located PHCCs with the availability of solar power at location j; Equation (4) maximizes the number of located PHCCs with the availability of basement at location j; Equation (5) maximizes the number of located PHCCs with the availability of internet service at location j; Equation (6) maximizes the number of located PHCCs with the availability of laboratory service at location j; Equation (7) maximizes the number of located PHCCs with the availability of blood grouping service at location j; and Equation (8) maximizes the number of located PHCCs with the availability of vaccination at location j, respectively. Equation (9) maximizes the cash for work amount for located PHCCs. Constraints of the proposed model are presented in Equations (10) to (14). While Equation (10) limits the number of located PHCCs to be less than or equal to a specific value (P), Equation (11) guarantees that each demand can be covered mostly once. Equation (12) limits each PHCC to cover demand nodes with less capacity or a capacity equal to its own capacity. Equation (13) expresses that demand at node i∈I cannot be covered unless at least one of the PHCC sites covering node i is located. Binary variables for located PHCCs and covered nodes are presented in Equation (14).Through an extensive literature review, we noticed that numerous methods have been built to deal with multi-objective problems [60,61,62,63,64,65,66,67,68]. Thus, the problem addressed in this paper is figured out via weighted goal programming. In the following section, we provide a brief introduction to weighted goal programming. Goal programming is one of numerous techniques for dealing with the modeling, solution, and analysis of multiple and conflicting objective problems. A traditional goal programming model contains constraints and a set of goals, all of which are taken into account simultaneously [69]. However, the final goal is to handle various objects which might be conflicting in the real world. This turns researcher’s attention to weighted goal programming (WGP), which is a type of goal programming and enables to optimize several objectives at once. This is achieved by converting crucial objectives (particularly those in contradiction) into goals and considering the remainder of objectives as constraints. Since trade-offs occur between objectives through deviation variables, negative and positive deviation variables are identified one by one for each goal corresponding to the over- and under-achievement of related goals. Hence, a single objective (achievement) function in the WGP minimizes the sum of undesirable deviations from the target goal values and results in a compromised solution between contradictory goals [70]. Any deviation is undesired, and the relative importance of each deviation variable is expressed by the relevant weights. They can be set either by expert estimation or a technique serving to that purpose (multi-criteria decision making techniques) [71]. Goals (namely a set of objectives) in WGP are commonly measured in different measurement units and they cannot be summed up as this would lead to incommensurability [72]. Deviations are scaled by utilizing the normalization technique to get rid of different units for various goals. Out of several normalization techniques (percentage normalization, Euclidean normalization, etc.); we employed percentage normalization in this paper. Thus, in relation to our objective functions with regard to cost (USD), allocated people (persons) and others, each deviation is converted into a percentage value apart from its target level. This enables to measure all deviations in the same units as a percentage. Our work contains multiple attributes and utilizes Multi Criteria Decision Making (MCDM) to address the relevant issues. MCDM techniques deal with decision making problems encompassing contradictory and miscellaneous criteria and objectives. At this point, it is noteworthy to point out that the Analytic Hierarchy Process (AHP) approach is very appropriate for group decision making as it contributes to numerous group preference collection methods [73,74]. AHP (developed by Saaty [75]) relies on expert judgments to obtain priority scales using pairwise comparisons. Throughout this process, comparisons are based on a scale of judgments (Table 2) that demonstrates to which extent one element dominates over another for a given attribute. Furthermore, it has theories to predict decision makers’ consistency of priorities [76]. Weights derived from the pairwise comparisons of AHP can be directly integrated into a WGP model [71]. Numerous studies reported the advantage of AHP for criteria weights [70,77,78,79,80,81,82,83]. Taking into account these features, it is employed in this study to weight objectives in WGP.Table 2 demonstrates the rating scale utilized for pairwise comparisons in AHP. For detailed information about this method, readers should refer to [75].In this paper, the pairwise comparison matrix of objectives was acquired by three expert decisions and their weights are presented in Table 3. The following equations (Equations (15) and (16)) were applied to check the consistency of responses acquired from decision experts. Equation (15) expresses the consistency index (CI) for a pairwise comparison matrix where λmax is the largest eigenvalue of the comparison matrix and n is the dimension of the matrix or the number of decision criteria. Equation (16) expresses the consistency ratio (CR) where RI(n) is a random index varying depending upon the size of matrix [76]. Random index values of random matrices are presented in Table 4. 
+        (15)CI= λmax−nn−1
+        (16)CR= CIRI(n).If the CR is equal to or less than 0.1, it is consistent and acceptable, however if it exceeds 0.1, the judgment sets may be too inconsistent to be reliable and the decision makers are asked to repeat pairwise comparisons to accomplish consistency in their responses. In our AHP, the CR was 0.005 meaning that it is consistent. The weighted goal programming formulation of the problem is given below and then, modified equations are listed subsequently.
+        (17)Minimize ∑n(pn− dn−+pn+dn+)1RHSn (%).Equation (17) minimizes the total deviations related to objective functions bearing in mind the penalty of each objective and percentage normalization according to RHSn “right hand sides” of the goal targeted for constraints from (18) to (26) as stated below. Thus, Equations (18) to (26) demonstrate the soft constraints taken into account in the weighted goal programming. Subject to:(18)∑i∈I∑j∈Iaij Zijhi+d1−− d1+  = RHS1
+        (19)∑j∈JfjXj+R∑i∈I∑j∈J Zijhi+TC∑i∈I∑j∈J Zijhi disij +d2−− d2+  = RHS2
+        (20)∑j∈JSEjXj+d3−− d3+  = RHS3
+        (21)∑j∈JBjXj+d4−− d4+  = RHS4
+        (22)∑j∈JISjXj +d5−− d5+ = RHS5
+        (23)∑j∈JLabjXj+d6−− d6+ = RHS6
+        (24)∑j∈JBGjXj +d7−− d7+ = RHS7
+        (25)∑j∈JVacjXj+d8−− d8+ = RHS8
+        (26)∑j∈JCWjXj+d9−− d9+ = RHS9.In the weighted goal programming model, Equations (10) to (14) are utilized in the same way as described before.By way of utilizing the methodology detailed in Section 2, we have performed a case study in Idleb Governorate of Syria. Idleb Governorate is situated in the northwest of Syria, has a border with Turkey and it has been a conflict area since the Syrian crisis started in 2011. It has an approximate area of 6097 km2 and the population estimate of the Governorate for 2010 (prior to the war) was about 1,464,000. Due to the crisis, no updated population estimate is available for the area. Furthermore, because of the crisis in the country, there are fluctuations in population and many people have immigrated from Aleppo, Eastern Ghouta, Homs or Daraa to Idleb. It might have evolved in this way because Idleb seems to be safer when compared to other Governorates in Syria and this makes it a good place for the settlement of internally displaced people (IDPs). Hence, our data collection in the area resulted in a population of 1,852,440 people. Figure 2 demonstrates the distribution of nodes and candidate PHCC locations for the location-allocation problem addressed in the study. Data regarding the following parameters of the proposed model was collected by operating an assessment via surveys and FGDs: Demands at nodes;Availability factors of candidate PHCCs (solar power, basement, internet service, laboratory service, blood grouping service and vaccination);Coverage distance;Fixed cost of locating a PHCC at candidate locations;Capacity of each candidate location;Cash for work amount at each candidate location.Demands at nodes;Availability factors of candidate PHCCs (solar power, basement, internet service, laboratory service, blood grouping service and vaccination);Coverage distance;Fixed cost of locating a PHCC at candidate locations;Capacity of each candidate location;Cash for work amount at each candidate location.More information about the parameters described above in the form of charts and figures is available in this interactive link (https://goo.gl/x2GjVv). The study, as displayed in Figure 2, covers 338 nodes and identifies 77 candidate PHCCs with the aforementioned fixed costs, variable costs and availability factors. Of the 77 candidate PHC centers, the maximum number of located/selected PHCCs are designated as 60 (p = 60), due to budgetary and management considerations, since the maximum number of possible PHCCs is 77 and if beneficiaries and experts are determined to identify a larger number, it means that they need more procedures and regulations to control it. On the contrary, if they consider a smaller number to be selected, it means that they will not be able to allocate a lot of people in this vulnerable area. As a result, they have found a compromise throughout FGDs/surveys (please see Figure 3 for detailed information about the flowchart of this process). Distances between nodes and candidate locations are generated utilizing GIS (ArcMap 10.4.1, Esri, Redlands, CA, USA). In this process, a roads network dataset is built by constructing the Origin-Destination (OD) matrix. The desired RHSn, pn− and pn+ values which will be utilized in the weighted goal programming model are acquired according to the process in Figure 3 and AHP are as follows:Regarding the RHS of our first objective, the target value of allocated people, we have set our target value as 1,852,440 following data collection since we aim to allocate all people in the case study area.Among candidate PHCCs; 31 have laboratory service, 33 have blood grouping service, 60 have vaccination, 18 have solar power, 36 have basement and 65 have internet service. Through the process in Figure 3, we determined these objective’s target values as: 30, 30, 30, 18, 30 and 30, respectively.According to the results obtained by AHP and depicted with Table 3; for every objective function(n); pn− values are set as: p1− 45, p2− 0, p3 − 5, p4− 5, p5− 5, p6− 5, p7− 5, p8− 9 and p9− 9. pn+ values are set as: p1+ 0, p2+ 14, p3+ 0, p4+ 0, p5+ 0, p6+ 0, p7+ 0, p8+ 0 and p9+  0. Here, goals though 1 to 9 correspond to: allocated people objective, total cost objective, cash for work, solar power, basement, internet service, laboratory service, blood grouping service and vaccination, respectively.Total cost budget and cash for work target values are determined as USD 1,000,000 and USD 100,000 via the process in Figure 3. Regarding the RHS of our first objective, the target value of allocated people, we have set our target value as 1,852,440 following data collection since we aim to allocate all people in the case study area.Among candidate PHCCs; 31 have laboratory service, 33 have blood grouping service, 60 have vaccination, 18 have solar power, 36 have basement and 65 have internet service. Through the process in Figure 3, we determined these objective’s target values as: 30, 30, 30, 18, 30 and 30, respectively.According to the results obtained by AHP and depicted with Table 3; for every objective function(n); pn− values are set as: p1− 45, p2− 0, p3 − 5, p4− 5, p5− 5, p6− 5, p7− 5, p8− 9 and p9− 9. pn+ values are set as: p1+ 0, p2+ 14, p3+ 0, p4+ 0, p5+ 0, p6+ 0, p7+ 0, p8+ 0 and p9+  0. Here, goals though 1 to 9 correspond to: allocated people objective, total cost objective, cash for work, solar power, basement, internet service, laboratory service, blood grouping service and vaccination, respectively.Total cost budget and cash for work target values are determined as USD 1,000,000 and USD 100,000 via the process in Figure 3. Then, demands, candidate locations and constraints are taken into account and the problem is solved via an optimization package software. The process of identifying the aforementioned values via FGDs with key stakeholders is composed of five stages and its flowchart is demonstrated in Figure 3 below. In the first stage, possible targeted values of RHSs to be considered depending on similar projects and situations are identified. The second stage defines main stakeholders that are connected with the specific value (financial aspect, humanitarian context, administration as well as donors, partners, beneficiaries, representatives, etc.) and will be involved in the next stages and achieves the related value. In the third stage, stakeholders are surveyed to find out their opinions and recommendations supported by reasons and clarifications. The fourth stage involves conducting FGDs to examine the results acquired in the third stage and discussing them. In the fifth and last stage of this process, estimated values are compromised by establishing the highest consensus value for most of the stakeholders by considering each value’s range of changes to be handled in sensitivity analyses until achieving a specific point “allocating as many people as possible” or “the full capacity of PHCCs”. We solved the model and attained the results demonstrated in Figure 4. It depicts achievement ratios (%) compared to the objectives targeted in the study. Since we aim to achieve every objective by 100% (horizontal orange column in Figure 4), it can be observed that four objectives acquire this (blue columns in Figure 4). For instance, we fulfilled an achievement ratio of 100% in terms of total cost budget. We fulfilled an achievement ratio of 120% and 103% for the objectives of PHCCs with internet service and PHCCs with vaccination, respectively. These mean that the number of located PHCCs with these availability factors is higher than the targeted values. Due to the conditions in Syria and Idleb, this is a significant and positive situation. Hence, the higher the number of PHCCs fulfilling these availability factors is, the higher the number of people benefitting from these centers is. We observe that achievement ratios for the objectives “PHCCs with laboratory service” and “PHCCs with blood grouping” are below the achievement ratio of 100% (77% and 80%, respectively). Achievement ratios for PHCCs with solar power and PHCCs with basement are 50% and 67%, respectively. A good compromise was achieved regarding these four results. The value of the objective “cash for work” is momentous for humanitarian contexts such as Syria-Idleb because the cash for work provided to vulnerable families as wages in return for working can alleviate the suffering of such persons in conflict areas. It is seen that this objective is achieved by 164% in this study, which is a crucial positive feature of the results achieved out of the model. Most importantly, we realized that our goal of “allocated people” did not achieve the targeted level (20%) because the total capacity of candidate PHCCs is 856,000. Since it is a conflict area and people are seeking health care, our model aimed to allocate all people (1,852,440 people) in the relevant area of Idleb. We aimed to include a maximum of 60 PHCCS out of 77 available candidates and the model located 42 PHCCs in total. Moreover, the number of allocated people (and all other objectives) depends mostly on the cost budget and it includes multiple factors: fixed cost, running cost and transportation cost. Within a cost budget of USD 1,000,000, the model allocates 379,080 people, which corresponds to an achievement ratio of 20% approximately. Since all our objectives except the number of allocated people, PHCCs with solar power, PHCCs with basement, PHCCs with laboratory service and PHCCs with blood grouping service achieved the targeted level, we focused on improving these objective values after general analyses. Figure 5 highlights the results of the PHCC location-allocation problem addressed in the study by demonstrating the nodes covered, the PHCCs located and the covered nodes allocated to the located PHCCs all together on a map. It also contains the summary of the addressed problem. The model selects 94 nodes out of 338 nodes by allocating them to the located 42 PHCCs. The total allocated population is 379,080 people and this corresponds to 44% of the PHCC’s total capacity, which is 856,000.Figure 6 and Figure 7 demonstrate the availability of internet service, solar power and basement of the located PHCCs and the availability of health factors (laboratory service, blood grouping service and vaccination) regarding the results of the PHCC location-allocation problem. In Figure 6, PHCCs with the availability of internet service are symbolized by blue bars; with the availability of solar power by yellow bars and with the availability of basement by orange bars. In Figure 7, blue bars display the located PHCCs with laboratory service while yellow bars and orange bars display the ones with blood grouping service and vaccination, respectively.In this part, we dealt with sensitivity analyses from three perspectives: (i) according to the model results, some objectives are under the targeted achievement ratio. However, it is a conflict area and our primary target is allocating as many people as possible, thus our main goal is to improve the objective of “coverage.” Since coverage, the number of allocated people, depends mainly on the cost budget within the study, a sensitivity analysis was implemented by increasing the cost budget within a specific range while keeping other inputs constant to find out how sensitive the objective regarding the number of allocated people is to these changes. This analysis enables us to observe other objectives’ changes as the cost budget is increased. (ii) We conducted sensitivity analyses through which the target of allocated people is decreased within a specific range until the total capacity of PHCCs is utilized in order to observe the achievement ratio of all objectives. (iii) As for the availability factors, the achievement ratio of which is under the targeted ratio (solar power, basement, laboratory service and blood grouping service), we decreased the RHS values of these parameters within a specific range so as to detect changes regarding other objectives. The following Figure 8, Figure 9 and Figure 10 display the results of the aforementioned sensitivity analysis. In these figures, OBJ1 to OBJ9 correspond to the following descriptions, respectively:OBJ1    : Value achieved in the Coverage; OBJ2    : Value achieved in the Total Cost;OBJ3    : Value achieved in the located PHCCs with Solar Power;       OBJ4    : Value achieved in the located PHCCs with Basement;OBJ5    : Value achieved in the located PHCCs with Internet Service;OBJ6    : Value achieved in the located PHCCs with Laboratory Service;OBJ7    : Value achieved in the located PHCCs with Blood Grouping Service;OBJ8    : Value achieved in the located PHCCs with Vaccination;OBJ9    : Value achieved in the Cash for work,Occupancy : Occupancy ratio of PHCCs (obtained by dividing the total number of allocated people to the total capacity of PHCCs).Figure 8 demonstrates the graph of achievement ratios for objectives acquired as a result of conducting sensitivity analysis by changing the cost budget. In this analysis, we increased the total cost budget by USD 200,000 at every model run, and observed the results regarding objectives 1 to 9. Although there is no significant difference on the achievements rates from USD 2,600,000 to USD 3,950,000, we achieved the maximum number of allocated people at USD 3,950,000. Through this cost budget, the model allocated 803,180 people to the 60 located PHCCs of which 18 have solar power, 30 have basement, 53 have internet service, 30 have laboratory service, 32 have blood grouping service and 44 have vaccination. Within this cost budget, the value of cash for work is acquired as USD 184,852. While the model firstly allocated 379,080 people, which corresponds to 20% of total PHCCs capacity occupancy, the model allocated 803,180 people by obtaining a PHCC occupancy rate of 94% after sensitivity analysis. All related achievement ratios are available in Figure 8. There are two reasons for the failure to acquire an achievement ratio of 100% regarding the occupancy parameter: firstly, since we located PHCCs and there is a coverage distance defined as the maximum distance people can travel to reach to the PHCC, some people could not be allocated to any located PHCC even if all PHCCs are located. Secondly, even if a PHCC is located, in the event that the number of people in the demand node nearby is higher than that PHCC’s remaining capacity, this node cannot be covered by that PHCC and the remaining capacity of PHCC will be the same without serving to any more people. The population of the case study is 1,852,440 but in addition to this, the total capacity of candidate PHCCs is 856,000. By considering that it is a conflict area and all people in the field will require health care, we targeted to allocate all of the population in the first place. However, since PHCCs currently have a specific capacity, we decreased the targeted number of allocated people within a specific range to observe results and achievement ratios in this part of the sensitivity analysis. Figure 9 demonstrates the achievement ratios of objectives as a result of conducting a sensitivity analysis within the target of allocated people. If we aim to allocate 856,000 people by keeping all other parameters in the model constant, the model allocates 697,500 people to the located 52 PHCCs. Even though we aim to utilize the PHCCs in full capacity and occupancy, the model can at most allocate 81% of the targeted allocated people to 81% of PHCCs’ total capacity mostly due to a limited cost budget of USD 1,000,000 in these analyses. Achievements ratios for other objectives can be obtained from Figure 9. With regard to the last sensitivity analysis, we focused on four availability factors: solar power, basement, blood grouping service and laboratory service. We decreased the target values (RHS values) of these availability factors and the results of achievement ratios for all objectives are depicted in Figure 10. This figure indicates that changing the RHS values of these availability factors almost does not affect the achievement ratios of other objectives. These changes only impact the achievement ratios of these objectives. In Figure 10, these changes can be tracked by following the green line with square markers for solar power (OBJ3); the purple line for basement (OBJ4); the orange line for laboratory service (OBJ6) and the light blue line for blood grouping service (OBJ7).In this study, we employed a multi-objective decision-making methodology to identify the optimum PHCC locations for the people in the north of Syria-Idleb, and to allocate the selected people nodes to the located PHCCs. Since it is necessary to consult with direct beneficiaries in humanitarian context as they are suffering from a myriad of miseries and we are trying to relieve their miseries, we mostly counted on direct beneficiaries because of the absence of authorities that can directly provide such data. In the first stage of our four-stage method, we operated a needs assessment with the help of focus group discussions, questionnaires, etc., so as to assess the needs of people. In the second stage, we set the objectives and constraints of the proposed model throughout AHP and FGDs. We constructed the roads network dataset utilizing the Geographic Information System and updated it in association with risky roads in order to build the origin-destination cost matrix in the third stage. In the last stage of the model, we built the mixed integer model, adapted it into a weighted goal programming model and solved it via an optimization package software and conducted a set of sensitivity analyses. After the execution of the model, it allocated 379,080 people with a cost of USD 1,000,353 and a cash for work of USD 163,549. A total of 42 PHCCs were located by considering the constraints and objectives identified in advance based on the region’s humanitarian context, the needs of people in the area and the indicators of stakeholders. Of these 42 PHCCs, nine PHCCs have solar power, 20 PHCCs have basement, 36 PHCCs have internet service, 23 PHCCs have laboratory service, 24 PHCCs have blood grouping service and 31 PHCCs have vaccination.To improve the results achieved in the analysis, we conducted sensitivity analyses and encountered three cases: (i) since the monetary budget is limited, the model managed to allocate 20% of people. Thus, the monetary budget should be increased to achieve better results in locating more PHCCs and allocating more people to them. (ii) As the main focus of this study is to allocate people to PHCCs and they have capacity constraints, the capacity of PHCCs should be increased in order to allocate and serve more people in this vulnerable area. (iii) Although some objectives are under the targeted level of achievement, changing their RHS values hardly affects the achievement ratios of other objectives. This action only affects their own achievement ratios.This paper focuses only on the northern part of Syria and assesses the relevant PHCCs in terms of nine criteria. Our primary aim was to identify the locations of PHCCs within Idleb Governorate and to relieve the miseries of the people in the area. Thus, labor factors and medical resources are not included in our model. These are the limitations of this paper and researchers can consider the following suggestions within their future studies so as to overcome these deficiencies: PHCCs or health care facilities can be assessed with more criteria such as the availability of running water and availability of electricity in hours.Criteria such as education, access to food and water can also be handled alongside the criteria/objectives addressed in this study.In future studies, labor factors (doctors, nurses, technicians and guards) and medical resources (beds, drugs, etc.) can be included in the model.Sensitivity analyses can be performed by changing multiple parameters simultaneously.Other regions of Syria can also be added into the relevant area, which can make the paper more comprehensive.A web-based tool can be designed incorporating the mathematical model and GIS and adapted to various similar problems.A dynamic model might be proposed to deal with the high degree of uncertainty regarding such problems.The problem can be handled by different techniques such as heuristic, meta-heuristic methods, hybrid models and social simulations.A conflict risk assessment can be applied to investigate the connection between the risk of armed conflict/ongoing crisis and a set of indicators such as education, infrastructure and access to health care facilities and food.PHCCs or health care facilities can be assessed with more criteria such as the availability of running water and availability of electricity in hours.Criteria such as education, access to food and water can also be handled alongside the criteria/objectives addressed in this study.In future studies, labor factors (doctors, nurses, technicians and guards) and medical resources (beds, drugs, etc.) can be included in the model.Sensitivity analyses can be performed by changing multiple parameters simultaneously.Other regions of Syria can also be added into the relevant area, which can make the paper more comprehensive.A web-based tool can be designed incorporating the mathematical model and GIS and adapted to various similar problems.A dynamic model might be proposed to deal with the high degree of uncertainty regarding such problems.The problem can be handled by different techniques such as heuristic, meta-heuristic methods, hybrid models and social simulations.A conflict risk assessment can be applied to investigate the connection between the risk of armed conflict/ongoing crisis and a set of indicators such as education, infrastructure and access to health care facilities and food.A conflict can influence human life in every aspect, particularly if it takes place in a densely-populated area and lasts for almost 8 years, as in Syria and specifically in Idleb. In this humanitarian emergency environment, people live in refugee camps, seek asylum and they even need basic items such as food, water, blanket, etc., apart from health. People need health care in every environment, but this proves to be more significant especially in a conflict area because conditions are more desperate since they live in harsh conditions, they are prone to infectious diseases, and they cannot access health care, which brings about an elevated mortality rate. The literature review revealed that, although there are numerous studies about primary health care subjects in natural and man-made disasters, there are no studies related to the location-allocation problem regarding PHCCs. Accordingly, there is no study handling the PHCC location-allocation problem within the Syria context which means that academic and scientific interest to this field is still unsatisfactory. These were the motives of this paper and therefore, this paper focused on identifying the locations of PHCCs and allocating people in the area to the located PHCCs with the aims of bridging this gap and being a valuable resource in such conflict areas. These contributions are attained by formulating a multi-objective model and integrating it into a real case study in the area so as to alleviate the dire impacts of this man-made disaster. From a humanitarian perspective, the proposed methodology is applicable in conflict areas to achieve the most feasible solution by combining the multi-objective mathematical model with GIS by utilizing real data from the area. These findings are expected to enable benefactors to respond to the needs of people, especially in these humanitarian contexts. Last but not the least, this paper’s main objective is to ensure accessibility to this model by any country, authority or institution. The purpose of this paper is to be useful for all humanitarian contexts and other services. In this paper, we identified the optimum location of PHCCs under specific constraints in a conflict area. Even decision variables, parameters and criteria may differentiate depending on the specific framework of the related problem area (this problem area can be general service centers, education, or other services), but our model can be used for different situations in other countries or conflict areas.  Conceptualization, P.M., M.K. and J.H.; methodology, P.M., M.K. and J.H.; software, P.M. and J.H.; validation, P.M. and J.H.; formal analysis, P.M.; investigation, P.M., M.K. and J.H.; resources, P.M and J.H.; data curation, J.H.; writing—original draft preparation, P.M., M.K. and J.H.; writing—review and editing, P.M., M.K.; visualization, P.M. and J.H.; supervision, M.K.; project administration, M.K.; funding acquisition, M.K.This research was funded by Çukurova University (Adana, Turkey), Coordination Unit for Scientific Research Projects, grant (project) number FDK-2017-8162.The authors declare no conflict of interest.Proposed methodology adopted in the study.Distribution of nodes and candidate Primary Health Care Centers (PHCCs) for the relevant location-allocation problem.Flowchart of focus group discussions (FGDs)/surveys utilized in the study.Outputs achieved in the study as a percentage of the targeted level.Results of the PHCC location-allocation problem addressed in the study.Availability of internet service, solar power and basement for the located PHCCs.Availability of health factors for the located PHCCs.Achievement ratios of objectives acquired as a result of conducting sensitivity analyses regarding the cost budget.Achievement ratios for objectives as a result of conducting sensitivity analyses in the target of allocated people.Achievement ratios for objectives as a result of conducting sensitivity analyses regarding availability factors.Meanings of model parameters and variables.Rating scale utilized in Analytic Hierarchy Process (AHP) (adopted from Saaty [75]).Pairwise comparisons with the final weight of each objective by AHP.Random index values [76].

Med-MDPI/ijerph_3/ijerph-16-05-00812.txt

@@ -0,0 +1 @@
+Background Few reports have evaluated the relationship between changes in postprandial blood pressure and the severity of autonomic dysfunction in patients with type 2 diabetes. This was a cross-sectional study designed to investigate postprandial blood pressure changes in individuals without type 2 diabetes and patients with type 2 diabetes and mild or severe cardiac autonomic dysfunction. Methods Forty patients with type 2 diabetes mellitus and 20 individuals without type 2 diabetes participated in this study. Fifty-two participants underwent a meal tolerance test. Blood pressure (brachial systolic blood pressure (bSBP) and central systolic blood pressure (cSBP)), electrocardiogram recordings, and blood samples were assessed before and after meal ingestion. Patients with diabetes were divided into two groups based on their coefficient of variation of R–R intervals (CVRR): a normal or mildly dysfunctional group (mild group, CVRR ≥ 2%; n = 20) and a severely dysfunctional group (severe group, CVRR < 2%; n = 15). Results In the control group, bSBP and cSBP did not significantly change after meal ingestion, whereas both decreased significantly at 60 min after meal ingestion in the mild and severe groups. While blood pressure recovered at 120 min after meal ingestion in the mild group, a significant decrease in blood pressure persisted at 120 min after meal ingestion in the severe group. Conclusions Based on these results, adequate clinical attention should be paid to the risk of serious events related to postprandial decreases in blood pressure, particularly in patients with diabetes and severe cardiac autonomic dysfunction.Diabetes mellitus (DM) is a chronic disease characterized by hyperglycemia resulting from defects in insulin secretion, action, or both [1]. Long-term complications of DM include retinopathy, nephropathy, and neuropathy [1]. Diabetes mellitus is a growing global concern, with the International Diabetes Foundation predicting that the number of people with diabetes will increase to 629 million worldwide by 2045 [2].Previous studies have shown that the prevalence of hypertension is increased in patients with DM [3]. Tight control of blood pressure (maintenance below 130/80 mmHg) is clinically recommended in patients with type 2 diabetes mellitus (T2DM) and hypertension by the Japanese Society of Hypertension’s 2014 guidelines [4]. Therefore, many patients with DM and hypertension develop hypotension as an undesired effect of blood pressure control measures [5].A critical symptom of hypotension in patients with DM is postprandial hypotension (PPH) [6]. Postprandial hypotension is commonly defined as a decrease of more than 20 mmHg in systolic blood pressure (SBP) within 2 h of the start of a meal, and it is an important clinical problem that disposes patients to syncope, falls, angina pectoris, and cerebrovascular events [7,8]. Tabara et al. [9] previously reported that lacunar infarctions were significantly more common in elderly individuals with a postprandial decrease of more than 10 mmHg in systolic blood pressure (SBP) than was found in controls. Sasaki et al. [6] also suggested that PPH might be an important clinical symptom of cardiovascular disturbances in patients with DM. In general, PPH may be more common and further associated with more severe outcomes in patients with DM and autonomic dysfunction [10]. Although the precise pathophysiological mechanisms underlying PPH remain unknown, Tanakaya et al. [11] have suggested that a lack of compensatory sympathetic activation contributes to PPH in patients with DM. Given this, PPH may be related to autonomic dysfunction. However, only a few reports have evaluated the clinical relationship between changes in postprandial blood pressure and the severity of autonomic dysfunction.Further, for blood pressure measurements, the importance of measuring central systolic blood pressure (cSBP), as performed in this study, is that this measure is more closely related to cardiovascular events compared to peripheral blood pressure and is recognized by studies such as the Anglo-Scandinavian Cardiac Outcomes Trial Conduit Artery Function Endpoint (ASCOT-CAFE) study [12,13]. However, studies of PPH that include cSBP as an outcome have not been performed yet.Given this gap in the existing research, we hypothesized that there might be a difference in the change in postprandial blood pressure in patients with T2DM and mild or severe cardiac autonomic dysfunction when measuring both brachial systolic blood pressure (bSBP) and cSBP and that there would be a positive correlation between the decrease in postprandial blood pressure and the severity of autonomic dysfunction.The purpose of the present study was as follows: (1) to investigate differences in the change in postprandial blood pressure in individuals without T2DM and patients with T2DM and mild or severe cardiac autonomic dysfunction by measuring brachial systolic blood pressure (bSBP) as well as cSBP; and (2) to clarify the relationship between changes in postprandial blood pressure and the severity of autonomic dysfunction.This was a cross-sectional study, which included 40 patients with T2DM and 20 individuals without T2DM. Patients with T2DM were recruited from outpatients and inpatients at Akashi Medical Center (Akashi, Hyogo, Japan) and Tatsuno Central Hospital (Tatsuno, Hyogo, Japan). All patients were diagnosed according to the Japan Diabetes Society’s T2DM criteria [14]. Study inclusion criteria included a patient age of 40–80 years and 6.0% ≤ HbA1c ≤ 12.0%. Study exclusion criteria included a diagnosis of type 1 diabetes, comorbid cardiovascular or other vascular diseases, arrhythmias such as atrial fibrillation, severe liver dysfunction, severe renal disease (estimated glomerular filtration rate < 30 mL/min/1.73 m2), pregnancy, and treatment with a β-blocker. We also recruited 20 individuals without T2DM through advertisements at the Graduate School of Health Sciences, Kobe University (Kobe, Japan). The only control group inclusion criterion was an age of 40–80 years. Control group exclusion criteria included pre-existing diabetes, cardiovascular disease, arrhythmias such as atrial fibrillation, pregnancy, being a night-shift worker, and smoking. The study was approved by the Ethics Committee of the Graduate School of Health Sciences, Kobe University (approval no. 14). The purpose and risk of this study was explained to each participant before their written informed consent was obtained.Figure 1 depicts the experimental protocol and timeline. Participants were asked to refrain from consuming food and drink that contained caffeine or alcohol and were instructed not to perform any exercise after 22:00 the previous night. They were asked to finish breakfast by 7:30 on the day of measurement. The administration of any hypertensive drugs was prohibited at least 24 h before the study. Patients were also advised not to take any antidiabetic drugs (including insulin) from the evening before the study until the study was completed. Before the assessment, height and body weight were measured. We measured the height with the participants’ shoes off, and the weight with light clothing and empty pockets.Assessments began at 11:30. After 10 min of rest in a seated position, blood pressure and radial arterial waveforms were measured. The measurement of cSBP was performed simultaneously. An electrocardiogram (ECG) was then obtained while the patient was in a supine position. After this, intravenous blood samples were collected from the brachial vein. Starting at 12:00, participants were served a test meal (E460F18, Kewpie, Tokyo, Japan; 460 kcal, 56.5 g of carbohydrate, 18.0 g of protein, 18.0 g of fat, and 670 mg of sodium). The participants were those who had completely consumed a meal within 15–20 min. The same set of measurements was then repeated at 60 (13:00) and 120 (14:00) minutes after the participants began their meal. During the measurement period, participants were allowed to drink only water (less than 500 mL) in addition to the test meal.Brachial systolic blood pressure, brachial diastolic blood pressure (bDBP), and cSBP were measured using a digital automated sphygmomanometer (HEM-9000AI, Omron Healthcare, Kyoto, Japan). Brachial systolic blood pressure and bDBP measurements were recorded using a cuff around the right upper arm. Pressure waveforms in the radial artery were recorded non-invasively using a tonometric approach. Early systolic pressure (SBP1), late systolic pressure (SBP2), and augmentation index (AI) were also measured and calculated automatically by the HEM-9000AI using the following equations: AI = (SBP2 − DBP)/(SBP1 − DBP), SBP2 (mmHg) = AI × (SBP − DBP) + DBP. Because Takazawa et al. [15] demonstrated a good correlation between cSBP and SBP2, cSBP calculations were based on the SBP2 values obtained in the present study. Furthermore, bSBP, bDBP, and cSBP were each measured twice and the average values were used for further analyses. Electrocardiogram recordings (Cardio Star FX-7432, Fukuda Denshi, Tokyo, Japan) were obtained across 200 pulses using lead II. R–R intervals of 200 pulses recorded on the ECG, coefficient of variation of R–R intervals (CVRR), and average heart rates (HRs) were calculated as follows: CVRR = (standard deviation/mean value of R–R intervals) × 100 (%). These assessments were performed by a medical technologist and a trained assistant. Venous blood samples (10 mL) were obtained by a medical doctor. Plasma glucose, serum insulin, triglycerides, high-density lipoprotein (HDL) cholesterol, and low-density lipoprotein (LDL) cholesterol levels were measured by standard methods using an autoanalyzer.Data were expressed as the mean value ± standard error. Body mass index (BMI) was calculated using the following equation: BMI = body weight (kg)/height (m)2. We divided the patients based on their CVRR into two groups: a normal or mild cardiac autonomic dysfunction (mild group; CVRR ≥ 2%) and a severe cardiac autonomic dysfunction (severe group; CVRR < 2%). A CVRR value of 2% is considered the critical level below which diabetic autonomic neuropathy occurs [16]. Differences among groups at baseline were evaluated using a one-way analysis of variance (ANOVA) and the Bonferroni post-hoc test. Differences between the mild and severe groups at baseline were evaluated using t-tests and Fisher’s exact test. Changes in measured variables following meal ingestion were examined using two-way ANOVAs for repeated measures and the Bonferroni post-hoc test. Pearson’s correlation analyses were performed to identify correlations between preprandial blood pressure (bSBP and cSBP), and the magnitude of postprandial change in bSBP over 1 or 2 h (∆bSBP1hr, or ∆bSBP2hrs) or cSBP over 1 or 2 h (∆cSBP1hr, ∆cSBP2hrs), calculated by subtracting preprandial blood pressure from postprandial blood pressure. A correlation analysis was also used to assess any correlation between preprandial CVRR and the magnitude of postprandial blood pressure change (∆bSBP1hr, ∆bSBP2hrs, ∆cSBP1hr, and ∆cSBP2hrs).A stepwise multiple linear regression analysis was performed to evaluate the independent variables that were associated with the magnitude of postprandial change in blood pressure for all subjects. The independent variables were ∆bSBP1hr, ∆bSBP2hrs, ∆cSBP1hr, and ∆cSBP2hrs. The following factors were used as dependent variables: age, BMI, CVRR, bSBP, and plasma glucose at baseline. The effectiveness of the models was assessed by p-values, and the completeness of the models were assessed by the coefficients of determination. Statistical analyses were conducted using SPSS version 20 (IBM, Chicago, IL, USA). A p-value of less than 0.05 was considered to indicate statistical significance.Figure 2 shows the participant flow through this study, which included 35 patients and 17 individuals. The mild group (CVRR ≥ 2%) was composed of 20 patients, and the severe group (CVRR < 2%) was composed of 15 patients. Baseline patient characteristics are shown in Table 1 and Table 2. In this study, 21 patients were treated with diet and exercise only, 14 patients were administered oral hypoglycemic agents, and eight patients received insulin therapy. In addition, 14 patients received the following drugs for treatment of hypertension: angiotensin-receptor blockers (n = 12) and calcium channel blockers (n = 8). The mean CVRR in the severe group was 1.45 ± 0.11%, which was significantly lower than that in the mild group (3.23% ± 0.19%, p < 0.01). In addition, the mean CVRR in the mild group was significantly lower than that in the control group (p < 0.05). There were no significant differences in age or measures of blood pressure, such as bSBP and cSBP, among the groups.During meal tolerance testing, plasma glucose significantly increased at 60 min in all groups (Figure 3A). Serum insulin concentrations also significantly increased at 60 min in the control and mild groups, while no significant difference was observed in the severe group (Figure 3B).After meal ingestion in the control group, bSBP and cSBP did not significantly change, while HR significantly increased at 60 and 120 min (Figure 4). In the mild and severe groups, bSBP significantly decreased at 60 min (from 136.5 ± 3.35 to 127.8 ± 2.78 mmHg in the mild group, p < 0.01; from 131.8 ± 3.87 to 122.6 ± 3.22 mmHg in the severe group, p < 0.01). This decrease in bSBP recovered by 120 min in the mild group (134.3 ± 3.04 mmHg), whereas a significant decrease in bSBP persisted at 120 min in the severe group (122.2 ± 3.51 mmHg, p < 0.01, Figure 4A). Similarly, cSBP significantly decreased in the mild and severe groups at 60 min after meal ingestion (from 141.5 ± 3.56 to 130.7 ± 2.99 mmHg in the mild group, p < 0.01; from 135.9 ± 4.11 to 122.8 ± 3.46 mmHg in the severe group, p < 0.01); the decrease in cSBP recovered at 120 min in the mild group (138.1 ± 3.22 mmHg), whereas a significant decrease in cSBP persisted at 120 min in the severe group (124.6 ± 3.72 mmHg, p < 0.01, Figure 4B). HR did not significantly change in either the mild or severe groups (Figure 4C).Pearson’s correlation analysis revealed that ∆bSBP1hr was significantly and negatively correlated with bSBP at baseline (r = −0.519, p < 0.001), whereas ∆bSBP2hrs was not. Similarly, ∆cSBP1hr was significantly and negatively correlated with cSBP at baseline (r = −0.377, p = 0.006), whereas ∆cSBP2hrs was not. ∆bSBP2hrs (r = 0.415, p = 0.003) and ∆cSBP2hrs (r = 0.430, p = 0.002) were significantly and positively correlated with CVRR, whereas ∆bSBP1hr and ∆cSBP1hr did not significantly correlate with CVRR.As shown in Table 3, multiple linear regression analyses revealed that bSBP at baseline was a significant predictor of ∆bSBP1hr and had a greater influence than CVRR, as indicated by a higher standardized coefficient. In addition, ∆bSBP2hrs was independently related to CVRR, whose standardized coefficient also revealed a greater influence than bSBP. Similarly, ∆cSBP1hr was independently related to cSBP at baseline, and ∆cSBP2hrs was independently related to CVRR with a higher standardized coefficient that indicated a greater influence than cSBP.The main findings of this study are as follows: (1) cSBP, as well as bSBP, decreased after meal ingestion in patients with T2DM and both mild and severe cardiac autonomic dysfunction; furthermore, the decline in bSBP and cSBP at 60 min after meal ingestion correlated with preprandial bSBP and cSBP, respectively. (2) Postprandial decreases in both bSBP and cSBP persisted longer in patients with T2DM and severe cardiac autonomic dysfunction, resulting in a magnitude of decline in bSBP and cSBP at 120 min after meal ingestion that was independently associated with CVRR. These results suggest that adequate clinical attention should be paid to possible serious events related to PPH, especially in patients with T2DM and severe cardiac autonomic dysfunction.In this study, bSBP decreased at 60 min after meal ingestion not only in patients with severe autonomic dysfunction but also in those with mild autonomic dysfunction, suggesting that a postprandial decrease in blood pressure might occur in the early stage of T2DM. In the present study, blood pressure was maintained and HR was increased in middle-aged individuals without diabetes after meal ingestion. The difference in postprandial decreases in blood pressure between individuals without T2DM and patients with T2DM could be attributed to differences in their compensatory response to postprandial hemodynamic change [7]. That is, in the individuals without diabetes, HR increased significantly after a meal to compensate for the decreased blood pressure, which resulted to a stable postprandial blood pressure. On the contrary, such an increase in HR was not observed in patients with T2DM. This lack of HR increase might lower the blood pressure.Our study results are in concordance with those of previous studies. For instance, some studies have found that postprandial blood pressure was maintained in healthy participants despite vasodilation and splanchnic blood pooling caused by gastrointestinal vasoactive peptides [17,18,19]. In healthy participants, HR and cardiac output increased significantly after meal ingestion to compensate for the lowering of blood pressure, which resulted in the maintenance of postprandial blood pressure [7,17]. However, in patients with T2DM, these compensatory responses may be diminished by cardiac autonomic dysfunction such that blood pressure cannot be maintained. Indeed, in the present study, CVRR was lower in the mild and severe groups than in the control group, revealing that cardiac autonomic dysfunction may occur in T2DM. In addition, there were no significant changes in HR in patients with T2DM and either mild or severe cardiac autonomic dysfunction. Previous studies have similarly reported that meal ingestion resulted in blood pressure reductions without changes in cardiac output or HR in patients with DM [11,20,21]. Some studies have also suggested that several sympathetic cardiovascular indexes increased in response to meal ingestion in healthy participants, whereas these compensatory responses were diminished in patients with autonomic dysfunction [17,22]. These studies are consistent with our findings, collectively suggesting that postprandial decreases in blood pressure in patients with T2DM may be due to cardiac autonomic dysfunction.In the present study, we found that ∆bSBP1hr significantly correlated with preprandial bSBP. Previous studies have made similar observations among elderly or hypertensive patients. For example, Vaitlevicius et al. [23] reported that maximal postprandial reductions in SBP occurred between 45 and 60 min and were inversely correlated with preprandial SBP values in elderly nursing-home residents. Puisieux et al. [24] also reported this relationship between preprandial SBP and postprandial blood pressure decline and suggested that higher preprandial SBP levels were associated with greater postprandial blood pressure declines and an elevated risk of PPH in elderly individuals. Mitro et al. [18] further observed a high prevalence of PPH at 60 min in hypertensive patients and suggested that the decrease in baroreflex sensitivity associated with blood pressure elevations might lead to PPH. Collectively, the conclusion that PPH is associated with hypertension may support our findings here. Furthermore, the present study suggests that postprandial decreases in SBP at 60 min in patients with T2DM may be related to hypertension rather than the degree of cardiac autonomic dysfunction.Although the precise mechanisms which result in PPH have not been clearly elucidated, earlier studies [19,25,26,27,28,29,30,31] had revealed several hemodynamic changes after meal ingestion. Meal ingestion promotes the secretion of vasodilators like insulin and neurotensin [8,19,25,26,27,28,29,30,31]. These agents cause vasodilation and splanchnic blood pooling, which can lead to a reduction in cardiac output, and thus, may contribute to the development of PPH [7,25]. In the present study, serum insulin increased after meal ingestion in patients with mild cardiac autonomic dysfunction but did not significantly change in patients with severe dysfunction. Despite these differences in insulin levels after meal ingestion, decreases in blood pressure were similar in patients with both mild and severe dysfunction at 60 min. Some previous studies have showed that exogenous insulin administration may lower blood pressure and sometimes result in syncope in patients with T2DM [26,27]. Insulin has further been implicated in the etiology of postprandial hypotension [28]. However, the results reported here suggest that insulin may not play a major role in PPH in patients with T2DM. Alternatively, eating possibly promotes neurotensin secretion from the small intestine in healthy participants and patients with autonomic failure [29,30]. In addition, voglibose, an alpha-glucosidase inhibitor, prevented PPH by reduced splanchnic blood pooling due to an inhibition of neurotensin release in patients with neurologic disorders [31]. Therefore, neurotensin, rather than insulin, might be the primary driver of PPH pathogenesis [19].We also found that CVRR was independently related to ∆bSBP2hr and had a greater influence than bSBP. This suggests that more severe cardiac autonomic dysfunction might delay recovery from decreases in blood pressure. To the best of our knowledge, this study is the first to report that postprandial decrease in blood pressure persists longer in patients with T2DM and severe cardiac autonomic dysfunction. Furthermore, impaired sympathetic activation due to severe cardiac autonomic dysfunction might not increase peripheral resistance, which was diminished by vasodilation after meal ingestion. Masuda et al. [32] suggested that to prevent PPH, sympathetic nervous activity during eating needs to be two to three times higher than that which occurs, on average, during daily activity. Therefore, in patients with T2DM and severe cardiac autonomic dysfunction, sympathetic dysfunction after meal ingestion might persist longer, resulting in PPH. Moreover, vasodilator secretion may persist longer due to delayed gastric emptying in these patients and may thus account for persistent peripheral vasodilation and decreases in blood pressure in patients with severe dysfunction. However, further studies are needed to investigate differences in the compensatory mechanisms of patients with T2DM and mild autonomic complications compared to those with severe autonomic complications.In the present study, as with bSBP, cSBP also decreased in patients with both mild and severe autonomic dysfunction. The decrease in the severe group persisted for up to 120 min after meal ingestion. Therefore, we extended the findings observed in bSBP to cSBP. As mentioned above, in general, cSBP is recognized to be more closely related to cardiovascular events compared to peripheral blood pressure, as demonstrated by studies such as ASCOT-CAFE study [12,13]. cSBP is considered to be the blood pressure at the aortic root, which is directly applied to key organs, such as the heart and brain. Previous reports have described how elevation of the cSBP induces coronary arteriosclerosis [33,34]. As the observation and reduction of cSBP contribute to the prevention of cardiovascular events, measurement of not only brachial blood pressure but also cSBP may be useful in the prevention and treatment of cardiovascular diseases [15]. This finding of the present study is of clinical significance because it suggests that the persistence of postprandial decrements to cSBP may not only result in dizziness and lightheadedness but may also trigger transient ischemic attacks [35]. Our results revealed an obvious persistence in postprandial decreases in cSBP in patients with T2DM and severe autonomic dysfunction, suggesting that the influence of postprandial decrease in blood pressure may be even greater than might be expected. Given these results, adequate attention should be paid to PPH, especially in patients with T2DM and severe autonomic dysfunction.Our study has some limitations that warrant discussion. First, we did not measure gastrointestinal hormones beyond insulin. Our results suggest that insulin might not play a major role in PPH in patients with T2DM. It is necessary to further investigate other hormonal mechanisms and the rate of gastric emptying and mesenteric blood flow to clarify additional potential mechanism of PPH. Second, as blood pressure was only measured for 120 min after meal ingestion, blood pressure recovery from hypotension was not observed in patients with severe autonomic dysfunction. While postprandial decreases in blood pressure recovered at 120 min after meal ingestion in patients with mild autonomic dysfunction, this decrease in blood pressure persisted in patients with severe autonomic dysfunction. Had we taken measurements for a longer period of time after meal ingestion, we might have been able to clarify the point at which postprandial hypotension recovers in patients with severe autonomic dysfunction. Third, the present study’s sample size was relatively small. In addition, we did not perform power calculations to determine the number of subjects. Fourth, in the control group, there was no established definition of CVRR, and an oral glucose tolerance test (OGTT) or HbA1c evaluation were not performed. In the control participants, the plasma glucose before meal ingestion was 99.3 ± 1.81 mg/dL, and none had levels exceeding 110 mg/dL, which is the criterion for impaired fasting glucose (IFG). Because we did not perform an OGTT in the control group, we cannot exclude the possibility that individuals with impaired glucose tolerance (IGT) were included. While this is a limitation, the plasma glucose at 120 min after meal ingestion was 108.5 ± 6.59 mg/dL, and no individuals had levels higher than 140 mg/dL. Therefore, we believe that it is unlikely that individuals with IGT were included in the control group and that the probability of misclassification bias is low. Finally, while numerous mechanisms may account for blood pressure reduction after meal ingestion, these remain poorly understood and worthy of additional investigation.The present study revealed that postprandial decreases in bSBP and also cSBP occurred in patients with T2DM and both mild and severe cardiac autonomic dysfunction. Furthermore, we found that postprandial decreases in blood pressure persisted for longer in patients with T2DM and severe cardiac autonomic dysfunction. In addition, cardiac autonomic dysfunction was associated with the magnitude of blood pressure decline at 120 min after meal ingestion. Given these findings, attention should be paid to the possible serious events related to PPH, especially in patients with T2DM and severe cardiac autonomic dysfunction.Conceptualization, M.H. and H.S.; Data curation, M.H., S.K., K.S., and H.S.; Formal analysis, M.H.; Investigation, M.H.; Methodology, M.H. and H.S.; Project administration, M.H. and H.S.; Software, M.H.; Supervision, K.P.I. and H.S.; Validation, K.P.I.; Visualization, M.H. and K.P.I.; Writing—original draft, M.H.; Writing—review & editing, M.H., K.P.I., and H.S. All authors approved the manuscript for submission.This research received no external funding.The authors would like to thank the participants in this study.The authors declare no conflict of interest.Experimental protocol. Abbreviations: BP, blood pressure; ECG, electrocardiogram.Participant flow. Abbreviations: T2DM, type 2 diabetes mellitus; CVRR, coefficient of variation of R-R intervals.Changes in plasma glucose (A) and serum insulin (B) during the meal tolerance test. The control group is indicated by the black line and circles (●); the mild group by the blue line and squares (■); and the severe group by the red line and triangles (▲). ** p < 0.01, * p < 0.05 vs. 0 min.Changes in brachial systolic blood pressure (SBP) (A), central SBP (B), and heart rate (C) during the meal tolerance test. The control group is indicated by the black line and circles (●); the mild group by the blue line and squares (■); and the severe group by the red line and triangles (▲). ** p < 0.01, * p < 0.05 vs. 0 min.Baseline characteristics.Data given as mean ± SE. ** p < 0.01, * p < 0.05 vs. Control. †† p < 0.01 vs. Mild Group. Abbreviations: bSBP, brachial systolic blood pressure; bDBP, brachial diastolic blood pressure; cSBP, central systolic blood pressure; CVRR, coefficient variation of R–R intervals; HDL, high-density lipoprotein; HR, heart rate; LDL, low-density lipoprotein.Baseline characteristics of patients.Data given as mean ± SE. Abbreviations: DM, diabetes mellitus; HbA1c, hemoglobin A1c.Multiple linear regression analysis for predicting the magnitude of postprandial blood pressure change.Abbreviations: BP, blood pressure; B, unstandardized coefficients; CI, confidence intervals; β, standardized coefficients; bSBP, brachial systolic blood pressure; cSBP, central systolic blood pressure; CVRR, coefficient variation of R–R intervals; SE, standard error; 1 hr, one hour; 2hrs, two hours.

Med-MDPI/ijerph_3/ijerph-16-05-00813.txt

@@ -0,0 +1 @@
+The methodology of intervention studies on physical activity (PA) promotion is of great importance regarding evidence development in complex interventions. The aim of this review was to provide an overview of the methodological quality of those studies which reported statistically significant effects of interventions promoting PA. PUBMED was searched for reviews on PA promotion to identify studies reporting effective interventions with participants of working age (16–67 years). Selected reviews were screened and data from primary studies with effective interventions were extracted to assess methodological quality. Forty-six reviews with 600 primary studies were identified, of which 33 met the inclusion criteria. Twenty-one studies were conducted as randomized controlled trials, 13 included an intervention control group, 25 measured PA by questionnaire, and 13 included objective measurements. Information on used statistics was often scarce, and long-term follow-up measurements were frequently missing. The overall methodological quality was moderate for randomized studies and low for non-randomized studies; information on methods and results was often lacking. To overcome these methodological issues, standardized guidelines for reporting study results should be considered, not only when publishing results but also when designing studies. This review provides a solid foundation for the development of practical advice for planning application-oriented studies in PA promotion.Physical activity (PA) is a widely accepted cornerstone of a healthy lifestyle and a favored target of public health efforts [1]. Nevertheless, promoting physical activity remains a challenge due to the complexity of the interventions and the various influencing factors such as personal, social, and environmental conditions [2,3,4]. From a practice-oriented perspective, the question of the intervention components which are associated with increased effectiveness in promoting PA, e.g., setting, delivery mode, study population or delivery provider, are of high relevance for maximizing the effectiveness of PA promotion programs [5,6]. However, there is considerable heterogeneity in promoting PA regarding settings and content, as well as the delivery mode and means of access [7,8]. In this context, several systematic reviews have already tried to identify components associated with the effectiveness of PA promotion, although most of these reviews have shown inconclusive results [6,8,9,10]. Overall, the reviews could not prove a clear relationship between effectiveness and the intervention components. Neither the intervention setting [9,10] nor the delivery mode [5,8,9,10] showed homogenous results.While the question of the effectiveness of specific intervention components is of great interest for practitioners, the methodology of intervention studies on PA promotion is of great importance regarding evidence development in complex interventions. To provide the best internal validity possible, interventions on promoting PA should refer to the standards of evidence-based medicine [11,12] including the use of the best available methods as well as the precise reporting of methods and results. While randomized controlled trials (RCTs) are considered the gold standard for high internal validity [13,14], the feasibility of RCTs in PA promotion is widely debated [15,16]. Among other things, the practicability of RCTs in the field of multifaceted and complex interventions as well as the accompanying costs are questioned [16]. With regard to reporting, several guidelines have emerged over the years. According to the CONSORT checklist for the transparent reporting of trials [14], the trial should report, inter alia, the description of the trial design, the sample size calculation as well as the completely defined, pre-specified primary and secondary outcome measures, including how they were assessed [14]. The aim of the present review was to assess the methodological quality of studies reporting effective interventions in PA promotion. Due to the high number of studies investigating interventions in PA promotion, the present review focused on studies reporting effective interventions since those are the ones being read and cited most often [17,18]. Moreover, previous research has shown that methodological quality influences conclusions of effectiveness in such a way that poorer study quality increases the likelihood of reporting statistically significant effects [19,20].This review was conducted following the international guidelines established by PRISMA (preferred reporting items for systematic reviews and meta-analyses) [21]. Since this review was based on a review of the literature, no ethical approval was required.This review focused on articles published between January 2007 and August 2016 in English or German. To include studies which had already been cited elsewhere, two researchers (JT, FL) independently searched the database PUBMED for reviews. The following keywords (including medical subject headings) were used and combined by the Boolean operator “AND”: physical activity, promotion, and intervention. The tags “child” and “school” were excluded by the Boolean operator “NOT” to directly reduce the number of results. Truncations (child*, school*) were used. Subsequently, the reviews identified were checked for primary studies reporting effectiveness in PA promotion. In an additional manual search, reviews of reviews and review protocols were checked for further reviews which fit the inclusion criteria and were not found in the PUBMED search. In the first search, the article type filter was set to review. In a two-phase screening process, first, the reviews’ titles and abstracts were independently screened for eligibility by two researchers (JT, FL). In the second phase, the full texts of relevant reviews were acquired and assessed against the inclusion criteria, which were 1) reviews or meta-analyses focusing on 2) the effectiveness of PA interventions among 3) participants of working age (16–67 years). Reviews were excluded if 1) PA was not a primary outcome, and 2) they only focused on sedentary behavior (e.g., sitting time), fitness (e.g., maximal oxygen consumption), or vital parameters (e.g., blood pressure) (see Table 1). If a selected review did not report information about the participants’ age range or the primary outcome, the review was provisionally included, and its primary studies were later checked separately for these inclusion criteria. If none of the primary studies complied with the inclusion criteria, the whole review was excluded subsequently.Based on the results of the review search, the primary studies of the reviews reporting an effective PA intervention were included in the second search. After excluding duplicates, two researchers (JT, FL) screened the titles and abstracts and independently checked the following inclusion criteria: 1) published 2007 or later, 2) primary outcome PA (frequency, duration and/or intensity), and 3) the authors reported a statistically significant effectiveness of the intervention for the PA outcome. Exclusion criteria for the primary studies in the second search were the same as described for the first search. The full texts were assessed by the researchers if the relevant information for inclusion could not be identified by the abstract screening process. Disagreements regarding inclusion were resolved through discussion involving the two researchers checking for eligibility (JT, FL) and an additional researcher (KR). Full (100%) consensus was achieved. All the studies that were excluded during the screening processes were recorded, along with the reasons for exclusion.The data extraction was performed by one researcher (LMR) and cross-checked by a second researcher (KR), with reference to the full text of the article. Following the CONSORT checklist [14], data on ten predefined study components were extracted and summarized in a data extraction template: Study design;Control group (CG) condition (intervention CG (we define an intervention CG as a group that receives any form of intervention, including usual care and placebos, and is called a CG by the authors of the primary study), non-intervention CG (we define a non-intervention CG as a group which is instructed to continue their lifestyle.) or no CG);Sample size (number of participants analyzed in primary outcome results);Operationalization of PA (subjective and/or objective measurements, observed time frame);Reporting of sample size calculation and achievement of desired number of cases (yes/no);Reporting of intention-to-treat (ITT) analyses (we define ITT analyses as analyses of study participants according to the original group allocation, regardless of non-compliance or inconsistency with the study protocol. We did not differentiate between the different methods of handling missing outcome data [22]) (yes/no);Checking for baseline group differences (yes/no);Reporting of drop-out analyses (yes/no);Reporting of standardized effect sizes of the results (e.g., Cohen’s d);Follow-up measurements (after the measurement at the end of the intervention).Study design;Control group (CG) condition (intervention CG (we define an intervention CG as a group that receives any form of intervention, including usual care and placebos, and is called a CG by the authors of the primary study), non-intervention CG (we define a non-intervention CG as a group which is instructed to continue their lifestyle.) or no CG);Sample size (number of participants analyzed in primary outcome results);Operationalization of PA (subjective and/or objective measurements, observed time frame);Reporting of sample size calculation and achievement of desired number of cases (yes/no);Reporting of intention-to-treat (ITT) analyses (we define ITT analyses as analyses of study participants according to the original group allocation, regardless of non-compliance or inconsistency with the study protocol. We did not differentiate between the different methods of handling missing outcome data [22]) (yes/no);Checking for baseline group differences (yes/no);Reporting of drop-out analyses (yes/no);Reporting of standardized effect sizes of the results (e.g., Cohen’s d);Follow-up measurements (after the measurement at the end of the intervention).In addition, the Delphi list [23] was used to assess the methodological quality of the studies included. This list is commonly used in systematic reviews [24,25] and has a comparatively greater validity of evidence than other standardized quality checklists [26]. Two researchers (LMR, KR) independently rated the studies’ quality by assigning a value of 0 or 1 for the 9 items of the Delphi list (1 point = “yes”; 0 point = “no” or “don’t know/not reported”). An interrater reliability analysis for individual Delphi scores was performed using the Kappa statistic.The combination of the keywords resulted in a total of 176 reviews whose titles and abstracts were screened for eligibility. In the next step, the full texts of the remaining 59 reviews were checked. A total of 19 reviews were excluded because of exclusion criteria being found in the full text. Five of these were provisionally excluded because they were reviews of reviews (RoR) or review protocols, which were later used for a manual search. Six reviews were added by manual search. Therefore, 46 reviews were included, resulting in 808 primary studies about PA interventions. After removing duplicates and checking for eligibility, 775 studies were excluded. As a result, a total of 33 studies were included in the present review. The flow chart diagrams (see Figure 1 and Figure 2) give an overview of the literature search process.Table 2 shows the results of the data extraction process. The individual Delphi scores for the studies included a range between 0 and 7 points, with an average of 5.1 out of 9 possible points for randomized studies and 1.3 points for non-randomized studies (see Table 2). The interrater reliability analysis showed a very good [25] interrater agreement with Kappa = 0.93 (p < 0.001) for individual Delphi scores. No study attained the maximum score of nine points. Four studies included the blinding of the outcome accessors, patients, and/or an intervention provider. No other study reported or used any kind of blinding procedure. The biggest similarities between the studies included were in the study design. The majority of the studies (n = 27; 81.8%) were conducted as randomized trials, 21 thereof (63.6%) as RCT. Correspondingly, most studies (n = 26; 78.8%) had some kind of control group condition. While 13 (39.4%) had an intervention control group, 13 (39.4%) used a non-intervention control group and seven (21.2%) had no control group at all. Of those without a control group, four studies had only one intervention group, one had two intervention groups and two had three such groups.Moreover, almost all (n = 26) of the studies with more than one group (n = 29) presented results for baseline group differences. The sample sizes of the studies included ranged from 25 to 1239 research participants, with most studies (n = 19, 57.6%) having more than 100 participants in total.Another similarity between the included studies was the absence of long-term follow-up measurements. Only four (12.1%) studies presented follow-up results based on an additional measurement after the measurement at the end of the intervention period. The biggest differences between the studies included the operationalization of PA. While the time frame predominantly focused on a typical week or the last seven days (n = 23, 69.7%), the measures were varied. More than half of the studies (n = 20; 60.6%) used subjective measures, while about a quarter (n = 8; 24.2%) reported objective measurements and five studies (15.2%) a combination of both. The most commonly used subjective measure was the International Physical Activity Questionnaire [59] (n = 9, 27.3%). Pedometers (n = 7, 21.2%) and accelerometers (n = 6, 18.2%) were used almost equally as objective measures. Of those using objective measures, six studies instructed participants to manually record their daily/weekly data in a step log.Further differences are related to the reporting of the statistics used and results. Less than a third (n = 10; 30.3%) of the studies conducted an a priori sample size calculation, 14 (42.4%) studies reported ITT analyses and less than a third (n = 9; 27.3%) included a standardized effect size in the description of the individual results. Moreover, 11 (33.3%) studies performed a drop-out analysis.The aim of the present review was to assess the methodological quality of intervention studies which effectively promoted PA. In total, we summarized data from 33 effectiveness studies reporting effective interventions in PA promotion, most of which were conducted as RCTs with either intervention or non-intervention control groups, moderate Delphi scores and usually more than 100 participants overall. In terms of data collection, questionnaires were used more frequently than accelerometers and pedometers. Follow-up data after the measurement at the end of the intervention period were usually not collected or reported. The biggest differences between the studies were found in the reporting of the statistics used. While most studies checked the baseline data for statistically significant group differences, ITT and dropout analyses as well as sample and effect size calculations were reported considerably less often.Overall, the methodological quality of the 33 studies which effectively promoted PA was moderate. While most studies applied some of the standard methods of securing high quality, such as the use of randomized controlled designs [13,14], there is still a lot of room for improvement regarding the reporting of methods (e.g., sample size calculation) and results (e.g., standardized effect sizes). Although reporting guidelines such as CONSORT [14] have existed for several years and many journals refer to the CONSORT statement in their “instructions to authors” section [60], the information within the studies published before and after the release of the CONSORT statement remains heterogeneous. For example, only a third of all studies included in our review reported a priori sample size calculations, although the benefits of an optimal sample size is well known: too-small sample sizes are more prone to bias [61,62], whereas a large number of participants consumes more resources and facilitates the detection of statistically significant changes which are not necessarily clinically relevant results [62,63], not to mention the ethical issues arising from exposing large numbers of participants to possibly non-effective interventions [62,64].A pleasing result of the current review was the prevalence of RCTs. The promotion of PA usually takes place in complex settings [65,66], where it is hard to administer RCT implementation. Nevertheless, the results of the present review show that more than 80% of the identified studies managed to include some kind of randomization procedure, showing that an approximation to the standards of evidence-based medicine [11,12] is possible and that statistically significant results can be obtained under these complex conditions. Furthermore, four studies even managed to apply blinding procedures. This is a positive sign for evidence development and should be an incentive for researchers to use RCTs and, thus, secure high quality in future studies [13,14].In line with the prevalent use of RCTs, the majority of the studies included control groups or used various intervention groups to examine the effectiveness of the individual interventions. However, a more frequent use of intervention control groups is desirable, since non-intervention control groups cannot rule out a possible placebo-effect of the interventions [67]. Moreover, intervention control groups would add knowledge through the comparison of different kinds of interventions and, hence, facilitate the search for the most effective interventions [67]. An even further strengthening of the research would be the more frequent reporting of standardized effect sizes, which would allow quantitative comparison between different treatments [63,67,68] and could also be used for sample size calculations in advance of a study [63].A striking feature of our results is the frequent use of questionnaires as the instrument of choice for the operationalization of PA. It stands to reason that subjective measurements increase the probability of a study to be reported as effective, but it is noticeable that more than 75% of the studies included a subjective measurement for the primary outcome, whereas only five studies used an additional objective measure. Obviously, data collection via questionnaires is much cheaper and easier to manage than most objective assessments [69,70], especially in large sample sizes. However, since both subjective and objective measures have individual strengths and weaknesses [71] and the respective data can differ widely [69], a combination of both is advisable to adjust for the individual weaknesses and to obtain a holistic view of individuals’ PA levels [72]. A positive aspect of additionally using accelerometers or pedometers is its objectivity, or rather that it is independent of the participants’ ability to accurately recollect the duration and intensity of their PA, which is often the subject of misperception [69,73,74]. The often-used method of letting participants write down their objectively measured data in a daily log (and then only evaluating this log data), however, reduces the objectivity of the data collection process since it brings back social desirability and the possibility of data bias. A possible solution to this problem could be the use of devices with sufficient memory capacity, so that the need to record data by hand becomes obsolete. The present review shows that the application of standardized reporting guidelines needs to become more established in the field of PA promotion. Moreover, these guidelines should not only be used when publications are being written, but also should be considered when studies are planned, to make sure that all of the necessary information will be available when the results are published. To our knowledge, this review is the first to address the methodological quality of studies effectively promoting PA. Our review does not claim to provide a full overview of the methodology used in the studies. Instead, it is limited to those studies reporting successful interventions, since effective studies are usually those who attract more attention by being cited more often [17,18]. In line with that, we included studies that had already been cited in other reviews. The aim was to consider their specific methodological structure to provide implications for future studies and to sensitize readers to methodological issues when interpreting the respective results. Nonetheless, studies showing non-effective results should have the same methodological quality as studies with positive findings. Methodologically poor quality studies can not only lead to the overrating of false positive results but also to the erroneous rejection of interventions that would have shown effectivity if evaluated better. Moreover, the need for good methodology is not only restricted to effectiveness studies but also applies to every other kind of study and research question, respectively.Due to the large number of studies investigating PA promotion, we restricted our literature search to one database and studies which investigated participants of working age (16–67 years). Moreover, the decision to limit the searches to articles published between 2007 and 2016 was a pragmatic one to reduce the number of included studies to a manageable amount. Wider inclusion criteria may have resulted in a larger number of studies being included and a different occurrence of the selected methodological characteristics. Nevertheless, our study could demonstrate that the methodological quality of studies leaves a lot of room for improvement and that attention should be paid to study quality when summarizing study results.The selection of the ten methodological study components for this study, in addition to the standardized Delphi list, is another strength of this review. We included different components such as the operationalization of PA, which are not covered by the Delphi list, to provide an extensive view of each study’s methodology, with a special focus on PA research. However, it must be mentioned that the Delphi list was originally created for RCTs [23]. Although almost two-thirds of the identified studies were RCTs, the Delphi list may be regarded as a suboptimal choice for the quality assessment. Although other quality assessment methods which are able to deal with non-randomized trials exist (e.g., the EPHPP Quality Assessment tool [75] and ROBINS-I [76]), we chose the Delphi list because of its psychometric values [26] and its frequent use in reviews [24,25]. Moreover, along with other quality assessment tools for RCTs, the Delphi list is often used as the basis for developing further scales [26]. For these reasons and because we did not want to combine two different assessment tools, we decided to also apply the Delphi list to non-RCTs.In addition to that, some of the reviews from which we extracted primary studies used quality assessment tools as well. However, due to the diverse instruments used in those reviews as well as missing quality appraisal for some of the primary studies, we decided to not include the quality assessment of the other reviews in our results.Methodological weaknesses may increase the probability of bias and unknown sources of error affecting study results. From a scientific point of view, the broad implementation of RCTs in the investigated intervention studies is pleasing, but weaknesses in the reporting of methods and results could still be identified. The challenge remains of overcoming the weaknesses identified and increasing the quality and explanatory power of study results; especially, the reporting of statistics, the combination of measurements as well as the use of long-term follow up measurements need to be improved. A more frequent adherence to guidelines for publishing study results is advisable. In addition to the existing guidelines for the publication of study results, guidelines for designing and conducting application-oriented studies on promoting PA are needed. The present review provides a first step for the development of these guidelines. The supplementary materials are available online at https://www.mdpi.com/1660-4601/16/5/813/s1. Conceptualization, K.R. and A.S.; Methodology, K.R., L.A.L.D. and A.S.; Validation, K.R., F.L., L.-M.R. and J.T.; Formal Analysis, K.R. and L.-M.R.; Investigation, K.R., F.L., L.-M.R. and J.T.; Resources, I.F.; Writing—Original Draft Preparation, K.R.; Writing—Review & Editing, K.R., L.A.L.D., I.F., F.L., L.-M.R., A.S. and J.T.; Visualization, K.R. and L.-M.R.; Supervision, A.S. and I.F.; Project Administration, A.S. and I.F.; Funding Acquisition, A.S. and I.F. All authors read and approved the final manuscript.This work was supported by the German Federal Ministry of Education and Research [reference: 01EL1425A].The authors thank the German Federal Ministry of Education and Research for funding the cross-cutting issue “physical activity” in the research association TRISEARCH.The authors declare no conflict of interest.Flow chart of the review selection process (step 1). Since not all included reviews provided suitable primary studies, we provide a Supplementary appendix with all reviews from which we included primary studies.Flow chart of the study selection process (step 2).Eligibility criteria for the literature search. PA: physical activity.Methodological components of effective PA interventions.Sample Size Calculation (SSC)ITT Analysis (ITT)Check for Baseline Differences (CBD)Drop-Out Analysis (DOA)SSC -ITT -CBD ✓DOA -SSC ✓ITT ✓CBD ✓DOA ✓SSC ✓ITT -CBD ✓DOA ✓SSC -ITT -CBD ✓DOA -SSC -ITT-CBD ✓DOA -SSC -ITT -CBD ✓DOA -SSC -ITT -CBD ✓DOA -SSC ✓ITT ✓CBD ✓DOA -SSC -ITT ✓CBD ✓DOA -SSC -ITT -CBD ✓DOA -SSC ✓ITT -CBD -DOA -SSC -ITT -CBD -DOA -SSC -ITT ✓CBD ✓DOA -SCC -ITT -CBD ✓DOA -SSC -ITT -CBD -DOA -SSC -ITT -CBD ✓DOA ✓SSC -ITT ✓CBD ✓DOA ✓SSC -ITT ✓CBD ✓DOA -SSC ✓ITT ✓CBD ✓DOA ✓SSC ✓ITT ✓CBD ✓DOA ✓SSC ✓ITT ✓CBD ✓DOA ✓SSC -ITT ✓CBD ✓DOA ✓SSC -ITT -CBD -DOA -SSC ✓ITT -CBD ✓DOA ✓SSC ✓ITT -CBD ✓DOA -SSC -ITT -CBD ✓DOA -SSC -ITT -CBD -DOA -SSC -ITT ✓CBD ✓DOA ✓SSC -ITT ✓CBD ✓DOA ✓SSC ✓ITT ✓CBD ✓DOA -SSC -ITT ✓CBD ✓DOA -SSC -ITT -CBD -DOA -SSC -ITT -CBD -DOA -✓: used/done; -: not used/not done/not reported; CBD: check for baseline-differences; CG: control group, whose participants received no intervention; DOA: drop-out analysis; ICG: control group that gets any form of intervention, including usual care and placebos; IG: intervention group; ITT: intention-to-treat analyses; PA: physical activity; RCT: randomized controlled trial; SSC: sample size calculation. a Recorded daily steps were reported in a step log. b n for at-risk subsample; total population for analysis IG n = 827, CG n = 890. c Sample size for the analysis not reported in the article; baseline values listed above.

Med-MDPI/ijerph_3/ijerph-16-05-00814.txt

@@ -0,0 +1 @@
+Features of the environment may modify the effect of particulate matter ≤2.5 µm in aerodynamic diameter (PM2.5) on health. Therefore, we investigated how neighborhood sociodemographic and land-use characteristics may modify the association between PM2.5 and cardiovascular mortality. We obtained residence-level geocoded cardiovascular mortality cases from the Massachusetts Department of Public Health (n = 179,986), and PM2.5 predictions from a satellite-based model (2001–2011). We appended census block group-level information on sociodemographic factors and walkability, and calculated neighborhood greenness within a 250 m buffer surrounding each residence. We found a 2.54% (1.34%; 3.74%) increase in cardiovascular mortality associated with a 10 µg/m3 increase in two-day average PM2.5. Walkability or greenness did not modify the association. However, when stratifying by neighborhood sociodemographic characteristics, smaller PM2.5 effects were observed in greener areas only among cases who resided in neighborhoods with a higher population density and lower percentages of white residents or residents with a high school diploma. In conclusion, the PM2.5 effects on cardiovascular mortality were attenuated by higher greenness only in areas with sociodemographic features that are highly correlated with lower socioeconomic status. Previous evidence suggests health benefits linked to neighborhood greenness may be stronger among lower socioeconomic groups. Attenuation of the PM2.5–mortality relationship due to greenness may explain some of this evidence.Cardiovascular disease (CVD) is the most common cause of death in the United States [1]. In the late 20th century there was a declining trend in CVD mortality attributable to health care and public health advancements, but CVD mortality is no longer improving [2].Numerous clinical, behavioral, and personal characteristics, such as age, sex, obesity, and smoking, have been established as risk factors for the development of CVD [3]. Yet, beyond these well-known clinical factors, there are a wide range of environmental and sociodemographic exposures that might influence CVD risk [4] including air pollution [5,6,7,8,9,10], socioeconomic status (SES) [11,12], and neighborhood greenness [13].Air pollution, a well-established risk factor for CVD morbidity and mortality [5,6,7,8,9,10], was identified recently by the Global Burden of Disease study as a highly ranked risk factor for CVD death [14]. A recent meta-analysis concluded that the relative risk for myocardial infarction is 1.02 times higher for every 10 µg/m3 increment increase in short-term exposure to PM2.5 [15]. Another systematic review that pooled studies of short-term exposure to PM2.5 found a 0.80% increase in CVD mortality per 10 µg/m3 increment increase in exposure [16].Both traditional CVD risk factors and air pollution exposures are related to neighborhood sociodemographic and land-use characteristics. There is an increasing prevalence of some traditional CVD risk factors (such as diabetes, obesity, and hypertension) over the last two decades and there is evidence that people of lower socioeconomic status are at higher risk [11,12]. This link between low socioeconomic status and CVD risk may be related to residence in areas with high levels of air pollution combined with little financial and physical access to health care services [17,18,19,20]. Sociodemographic characteristics were found to modify the association between PM2.5 and mortality in several studies, showing larger effects among communities of non-white race [21], lower socioeconomic status [21,22], lower education [23], and higher unemployment rate [24,25].Other neighborhood land-use characteristics such as neighborhood greenness and walkability may affect human health directly or through modification of the PM2.5 effect on health. A number of studies have defined neighborhood greenness as the amount of vegetation in a defined buffer around the residential address [13,26,27]. The common hypothesized mechanisms used to explain the beneficial effect of green space on cardiovascular health involve providing a location for physical activity [28], increasing social engagement, and lowering stress levels [29,30]. Greenness is also thought to reduce exposure to air pollution either by filtering the air or, more likely, by creating a buffer space between sources and individuals who might be exposed [31]. Neighborhood walkability is a concept of how amenable an area is for routine walking. Walkability indices are commonly comprised of land-use components that are linked to walking behavior [32] and measures of the neighborhood’s accessibility, connectivity and density [26,27]. A highly walkable built environment may be correlated with higher air pollution levels, and may lead to higher personal exposure to air pollution as individuals are spending more time outdoors [27]. At the same time, it may be more conducive to physical activity.While studies have assessed the direct effect of these land-use characteristics on human health [13,27,33], evidence on whether they modify PM2.5 effects is scarce. In recent years, the few studies that assessed the modification of the PM2.5 effect by neighborhood greenness have shown contradicting results. For example, one study found higher air quality associated health risks among people who reported not using green areas [34]. Another study found larger mortality effect estimates with increasing neighborhood greenness [35]. One study has shown conflicting effect modification of the association between air pollution and mortality by neighborhood greenness across urban and rural areas [36]. To our knowledge, only one study has examined the modification of the PM2.5 effect on mortality risk by neighborhood walkability, and has found no evidence of an interaction between PM2.5 and walkability [37].In this study, we aim to assess the joint modification of the PM2.5 related cardiovascular risk by land-use characteristics (neighborhood greenness and walkability) and neighborhood sociodemographic characteristics (income, race, population density, and education) while controlling for individual sociodemographic characteristics. The Massachusetts Department of Public Health registers mortality records with information on residential address, place of death, age at death, gender, ethnicity, education, occupation, the exact date of death, and the underlying cause of death. We obtained the records of all decedents who died of a cardiovascular cause (International Classification of Disease, 10th Revision, group I) [38] between the years 2001–2011 and were 40 years or older.This study was approved by the Harvard T.H. Chan School of Public Health Human Subjects Committee and by the Massachusetts Department of Public Health. We estimated average daily PM2.5 from a model generating predictions at a 1 × 1 km spatial resolution. Briefly, mean daily PM2.5 concentrations were estimated daily in each 1 × 1 km grid cell by calibrating Aerosol Optical Depth (AOD) with monitored PM2.5 using mixed effect models with spatial and temporal predictors and a random slope for day and nested regions. Out-of-sample “ten-fold” cross-validation showed excellent model performance (mean out-of-sample R2  =  0.88). For a more in-depth description, please refer to Kloog et al. (2014) [39]. We assigned each decedent the daily values for the grid in which his/her residential address was located.We adjusted our models for temperature and day of the week. We estimated daily average temperature from a model that incorporated moderate resolution imaging spectroradiometer (MODIS) land surface temperatures (LST) data, land-use regression (LUR) variables (percent urban, elevation, normalized difference vegetation index) and a random intercept and slope for surface temperature for each day. Out-of-sample ten-fold cross-validation showed excellent model performance (mean out-of-sample R2 = 0.94). For more in-depth information, refer to Kloog et al. (2014) [40]. As with PM2.5, each decedent was assigned the daily values for the grid in which his/her residential address was located.We considered neighborhood land-use characteristics and sociodemographic characteristics as potential modifiers.Neighborhood greenness was estimated using a 250 × 250 m spatial grid, and each decedent was assigned the value for the grid in which his/her address was located using the Normalized Difference Vegetation Index (NDVI). Seasonal NDVI estimates were obtained from MODIS from NASA’s Terra satellite for all the years included in the study period. NDVI exposure in the season of death was assigned to each subject. During the photosynthesis process, chlorophyll in plants strongly absorb visible light (0.4–0.7 μm) while leaves reflect near-infrared light (0.7–1.1 μm). The satellite image provides information on these two measurements, and the NDVI calculates the ratio of the difference between the near-infrared region and red reflectance to the sum of these two measures. NDVI values range from −1.0 to 1.0, with larger values indicating higher levels of vegetative density. Previous studies of neighborhood greenness have used a 250 m area and a larger buffer size (1000–1250 m radius) as an approximate measure of residential and walking greenness exposure, respectively [41,42,43]. Therefore, as a sensitivity analysis, we used the seasonal mean NDVI data in a 1250 m area surrounding each address [27]. Neighborhood walkability index at the census block group-level was assigned based on the US Environmental Protection Agency (EPA) Smart Location Database. The Smart Location Database provides nationwide geographic data on 90 attributes summarizing characteristics such as housing density, diversity of land-use, neighborhood design, destination accessibility, transit service, employment, and demographics at the census block group-level [44]. We created a nationwide walkability index combining z-scores of three components: population density (people/acre) on unprotected land; street intersection density (weighted, with auto-oriented intersections eliminated); and land-use diversity based on the mix of retail, office, service, industrial, entertainment, education, healthcare, and public administration employment in the census block group. The index ranges between −2.9 and 16, where higher values indicate a more walkable neighborhood [44]. We obtained the following characteristics for each block group from the 2014 Census 5-year estimates from the American Community Survey (ACS) which spans from 1 January 2010 to 31 December 2014: percentage of residents by race (white, black or other), percent of residents with no high school diploma, median household income, percent of residents below the poverty line, and population density. We assessed the association between CVD mortality and PM2.5 using the time-stratified case-crossover approach [45], with the case day defined as the date of death for each person. Control days were defined as every 3rd day before and after the case day, and within the same month and year as the case day. Based on our previous work, we defined the exposure window for the main analysis as the moving average concentration of PM2.5 of the current and previous days. All models were adjusted for day of the week and temperature [40], with the same exposure window as PM2.5. Personal SES is controlled for by design as the same person is compared on case and control days. Season is controlled for by design as cases and controls are in the same month.We assessed the modification of the PM2.5 and CVD mortality association in three phases. First, we assessed modification by neighborhood greenness and walkability independently by separately adding multiplicative interaction terms for each. Then, we assessed joint modification by adding interaction terms between PM2.5; each land-use characteristic, and each of the neighborhood sociodemographic characteristics. Then, for simplicity, we presented the modification of the PM2.5 effects by the land-use characteristics in models stratified by the neighborhood sociodemographic characteristics.To make sure the modifying effect of NDVI is not influenced by seasonality, we added a sensitivity analysis in which we assessed the modification effect of the PM2.5–CVD association by annual averages of NDVI. Results are presented for each 10 µg/m3 increase in PM2.5 at the 25th and 75th percentiles of the land-use modifier within each stratum (below or above the median value) of neighborhood sociodemographic characteristic. Analyses were performed using the ‘survival’ package in R version 3.5.2 [46].We included 179,986 persons who resided in Massachusetts and died of CVD between 2001 and 2011. The mean age of death was 80 years, about 94% of the decedents were white, 45% were males, and the majority had a high school education or less (Table 1). The interquartile range (IQR) of the two-day average concentration of PM2.5 was between 6.8 µg/m3 and 12.3 µg/m3, and the mean value was 10.2 µg/m3. Over the study period, the mean block group PM2.5 values were lower in rural areas and higher in urban areas, as expected, reaching a maximum value of 17.4 µg/m3 (Figure S1). The mean NDVI ranged between 0.00 and 0.87, and the mean walkability index ranged between −2.93 and 16.87. As expected, the walkability index was highest in urban areas and lowest in rural areas (Figure S2). NDVI tended to be higher in rural areas and lower in urban areas, although this was not true in all locations (Figure S3).An increase of 10 µg/m3 in exposure to PM2.5 in lag days 0–1 was statistically significantly associated with a 2.54% (1.34%; 3.74%) increase in CVD mortality. No association was observed with exposure to temperature in lag days 0–1 (−1.38% (−3.49%; 0.77%), per 10 °C).Among all available cases, NDVI and walkability did not modify the association with PM2.5. The PM2.5–mortality association did not change across levels of NDVI. Results were similar when calculating the NDVI value in the 250 m and 1250 m buffer around the decedents’ addresses (Table 2).When we assessed the joint modification effect of NDVI or walkability and the neighborhood sociodemographic characteristics, we observed a statistically significant joint modification of the PM2.5–CVD associations by NDVI and population density (interaction p value = 0.001), NDVI and percentage of white population (interaction p value = 0.019), and NDVI and percentage of persons with no high school diploma (interaction p value = 0.043). We found no joint modification by walkability and the neighborhood sociodemographic characteristics (Table S1). We observed smaller PM2.5 effects with increasing values of NDVI among people who resided in areas of higher population density, areas with a lower percentage of white population, and areas with a higher percentage of residents without a high school diploma. In areas of lower population density, a higher percentage of white residents, or a lower percentage of residents without a high school diploma, we observed larger PM2.5 effects with increasing values of NDVI (Figure 1 and Table S2). The sensitivity analysis, using annual NDVI averages, showed similar results (Figure S4). We also observed smaller PM2.5 effects with increasing values of NDVI in neighborhoods with a higher percentage of poverty and lower median household income, though these findings were not statistically significant (data not shown).In this study, we observed statistically significant increased risk for CVD mortality associated with exposure to PM2.5. Among all cases, the risk was not modified by neighborhood greenness or walkability. However, we found that PM2.5-related cardiovascular mortality was lower with higher NDVI only in neighborhoods with higher population density, a lower percentage of white population, or a higher percentage of residents without a high school diploma. Although the joint interaction was not statistically significant, we observed smaller PM2.5 effects in greener neighborhoods with a higher percentage of poverty and lower median household income as well. No joint effect modification was observed for walkability.Our finding of joint effect modification of the PM2.5–CVD mortality association, by NDVI and sociodemographic factors, is generally consistent with evidence from previous studies suggesting that the health benefits of exposure to green space are stronger among low socioeconomic groups and in urban areas [36,47,48]. Similar to our findings, a study that assessed the association between green space and cardiovascular mortality found reduced mortality risk among those exposed to greener environments only amongst the most socioeconomically deprived groups [48]. A study that included 49 U.S. cities found protective effects of neighborhood greenness on mortality and life expectancy only in areas of lower socioeconomic status [28]. Another study conducted in the U.S. found a positive association between access to quality green space and reduced psychological distress among deprived urban populations [49]. A possible explanation for why the effect of air pollution is weaker in greener neighborhoods with lower socioeconomic status may be related to different usage of green space in these areas. Higher access to green space among deprived communities may contribute to a reduction of health inequalities by increasing frequency or time spent on outdoor activities, which contribute to a better psychological and physical health [50]. In addition, deprived communities may get a larger benefit from green spaces because residents of these communities may be more likely to spend time outside and interact with their immediate neighborhood compared to those in wealthier neighborhoods [51,52,53].To our knowledge, there are no studies that assessed the difference in the joint modification of the PM2.5 effect by greenness and SES levels. However, studies that assessed the modification of the PM2.5 effect by SES alone found that poorer neighborhoods are often characterized by higher pollution levels and larger PM2.5 effects [20]. Another finding of our study is the synergistic effect of PM2.5 and greenness found in neighborhoods with a lower population density, higher percentage of whites, and higher education level. A possible explanation for this finding is that the benefits of neighborhood greenness could be compromised by conditions and lifestyles that are prevalent in greener rural areas (such as car dependency) and distance to health care [28]. We found no modification of the effect of PM2.5 by neighborhood walkability. Like our findings, a study in California that compared the ischemic heart disease mortality risk between neighborhoods according to their walkability score found that the estimated mortality risk contributed by air pollution exposure did not differ by the neighborhood walkability score [37].A major strength of this study is the use of geocoded mortality cases with high spatially and temporally resolved exposure. In addition, using joint interaction analysis, we were able to assess the complex relationship between multiple environmental exposures and mortality.Our study had several limitations. First, similarly to other studies that assess the effect of air pollution on human health, there is a possibility of exposure measurement error due to the lack of information on participants’ activity space. Second, we did not have the ability to evaluate NDVI and walkability at the same exposure window as PM2.5. NDVI was calculated seasonally, and walkability was calculated once during the study period. However, since these land-use characteristics vary slowly over time, we do not expect it to bias our results. Third, the sociodemographic neighborhood-level modifiers included in our analysis may not have adequately captured differences across neighborhoods and possibly underestimated the neighborhood effects. In addition, in this study design, we were not able to account for personal usage of green space.In conclusion, PM2.5-related CVD mortality risk was smaller in highly populated greener neighborhoods with lower SES. Cumulative evidence suggests that health benefits linked to neighborhood greenness may be stronger among the lower socioeconomic groups. This may explain the beneficial effect of greenness found only in highly dense neighborhoods with a lower percentage of white residents and lower education. Our findings confirm the importance of addressing multiple aspects of an individual’s environment when assessing the effect of air pollution on human health.The following are available online at https://www.mdpi.com/1660-4601/16/5/814/s1, Table S1: The joint modification of the PM2.5 effect by NDVI, walkability and neighborhood sociodemographic characteristics, Table S2: Percent increase in mortality for 10 µg/m3 of average PM2.5 in the same and previous day: modification by NDVI within strata of low and high sociodemographic characteristic, Figure S1: Block group average concentrations of PM2.5 in the study area across Massachusetts (2001–2011), Figure S2: Block group average walkability index in the study area across Massachusetts (2001–2011), Figure S3: Block group average NDVI in the study area across Massachusetts (2001–2011), Figure S4: The percent increase and 95% confidence intervals in cardiovascular mortality, associated with PM2.5, in the 25th and 75th percentiles of NDVI, by neighborhood sociodemographic characteristics.Conceptualization, A.Z.; Methodology, M.Y.S., J.E.H., A.Z.; Formal Analysis, M.Y.S.; Investigation, M.Y.S., P.J., I.K., K.J.L., K.F.; Writing—Original Draft Preparation, M.Y.S; Writing—Review & Editing, P.J., I.K., J.E.H., J.D.S., F.L., K.J.L., M.P.F., K.F., A.Z.; Supervision, F.L., M.P.F., A.Z.; Funding Acquisition, A.Z., J.D.S.This study was supported by the National Institute of Minority Health and Health Disparities grant P50MD010428; National Institute of Environmental Health Sciences grants P30 ES000002 and R01 ES024332; National Cancer Institute grant R00 CA201542; and Environmental Protection Agency grants RD83615601 and RD83587201. Its contents are solely the responsibility of the grantee and do not necessarily represent the official views of the USEPA. Further, USEPA does not endorse the purchase of any commercial products or services mentioned in the publication.The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.The percent increase and 95% confidence intervals in cardiovascular mortality, associated with PM2.5, in the 25th and 75th percentiles of NDVI, by neighborhood sociodemographic characteristics. p25 and p75 refer to percentiles of NDVI.Population characteristics: cardiovascular mortality cases in Massachusetts 2001–2011.The percent change in cardiovascular mortality associated with 10 µg/m3 increase in PM2.5 in the 25th and 75th percentiles of NDVI and walkability measures.1 The walkability index was created using the z scores of the following three components of the US EPA Smart Location Database: (1) gross population density (i.e., people/acre on unprotected land); (2) eight-tier employment entropy (i.e., land-use diversity of employment mix of retail, office, service, industrial, entertainment, education, healthcare, and public administration occupations); (3) street intersection density (i.e., summary of the total intersection density, weighted to reflect connectivity for pedestrian and bicycle travel). 2 Normalized Difference Vegetation Index (NDVI) is the ratio of the difference between the near-infrared region and red reflectance and the sum of these two measures. NDVI values range from −1.0 to 1.0, with larger values indicating higher levels of vegetative density.

Med-MDPI/ijerph_3/ijerph-16-05-00815.txt

@@ -0,0 +1 @@
+The authors contributed equally to this work.Sarcoid-like granulomatous diseases (SGD) have been previously identified in cohorts of World Trade Center (WTC) dust-exposed individuals. In the present studies, we analyzed lung and/or lymph node biopsies from patients referred to our clinic with suspected WTC dust-induced lung disease to evaluate potential pathophysiologic mechanisms. Histologic sections of lung and/or lymph node samples were analyzed for markers of injury, oxidative stress, inflammation, fibrosis, and epigenetic modifications. Out of seven patients examined, we diagnosed four with SGD and two with pulmonary fibrosis; one was diagnosed later with SGD at another medical facility. Patients with SGD were predominantly white, obese men, who were less than 50 years old and never smoked. Cytochrome b5, cytokeratin 17, heme oxygenase-1, lipocalin-2, inducible nitric oxide synthase, cyclooxygenase 2, tumor necrosis factor α, ADP-ribosylation factor-like GTPase 11, mannose receptor-1, galectin-3, transforming growth factor β, histone-3 and methylated histone-3 were identified in lung and lymph nodes at varying levels in all samples examined. Three of the biopsy samples with granulomas displayed peri-granulomatous fibrosis. These findings are important and suggest the potential of WTC dust-induced fibrotic sarcoid. It is likely that patient demographics and/or genetic factors influence the response to WTC dust injury and that these contribute to different pathological outcomes.Following the collapse of the World Trade Center (WTC) towers on 11 September, 2001, toxic dust was suspended in the air of lower Manhattan and Brooklyn. Over time, rescue and recovery workers exposed to WTC dust developed various respiratory diseases involving both the airways (asthma-like airway hyper-reactivity, bronchiolitis), and less commonly, the lung parenchyma (eosinophilic pneumonia, fibrosis, sarcoid-like granulomatous disease) [1,2]. Most notable is sarcoid-like granulomatous disease (SGD), which has been diagnosed at relatively high rates in cohorts of firefighters and rescue workers who have been followed since 2001 [3]. This condition is distinct from frank sarcoidosis which is a diagnosis of exclusion, once other known causes of granulomatous disease have been ruled out. It may be that WTC dust-induced SGD is a unique pathology, with potential for specific therapeutic interventions. Although WTC dust has been extensively characterized [4], little is known about the pathophysiologic mechanisms underlying the development of SGD and other lung pathologies. Studies in rodents have reported increased expression of genes associated with inflammation and oxidative stress in the lung following WTC dust exposure [5,6]. However, in these studies, granulomatous inflammation was not observed, making determination of the pathogenesis of WTC dust-induced SGD in humans, problematic.In this report, we describe the pathologic findings in seven patients exposed to WTC dust who presented with ambiguous clinical pulmonary diagnoses. In all of the patients, we identified markers of injury, oxidative stress, inflammation, and epigenetic changes in lung and/or lymph nodes. Based on pathologic findings, five of the patients demonstrated evidence of SGD and three of these had peri-granulomatous fibrosis. These findings are important as they suggest the possibility of development of fibrosis in patients with WTC dust-induced SGD. This may lead to changes in clinical outcomes and treatment strategies.Investigators at the Environmental and Occupational Health Sciences Institute (EOHSI) of Rutgers University are part of a New York/New Jersey consortium that has been following a cohort of rescue and recovery workers exposed to dust and other materials at the WTC site. These individuals included rescue, recovery, debris-cleanup, law enforcement, and related support service workers, and volunteers in lower Manhattan (south of Canal St.), the Staten Island Landfill, and/or the barge loading piers, who worked on-site for at least 4 h/day between 11 and 14 September, 2001, for at least 24 h during September 2001, or for at least 80 h total time between September and December 2001 [7]. From among approximately 2000 patients followed at EOHSI, 7 were referred for biopsy between 2007 and 2011, if a final pulmonary diagnosis despite symptom review and full evaluation including pulmonary function testing (PFT) and computerized tomography (CT) scan interpretation required tissue analysis. Patient demographics were collected, including WTC dust exposure levels as described by Wisnivesky et al. [2]. Lung and/or mediastinal lymph node specimens were obtained for both histopathologic evaluation and immunohistochemical staining. All subjects gave informed consent for inclusion before they participated in the study. The study was conducted in accord with the Declaration of Helsinki, and the protocol was approved by the Ethics Committee of the Rutgers Institutional Review Board.PFTs were performed using NSpire Health pulmonary functioning testing devices (NSpire Health, Longmont, CO, USA), that were calibrated in the Rutgers University Robert Wood Johnson Medical School Pulmonary Function Laboratory. Test results were interpreted according to American Thoracic Society guidelines and results collected for each patient [8]. Tidal breathing CT was performed from lung apex to lung base without the use of intravenous contrast material. All results were interpreted by board-certified radiologists.We collected lung samples by transbronchial biopsy (TBB) from 3 patients, and by video-assisted thoracoscopy (VATS) from 2 patients. Three patients had mediastinal lymph node biopsies obtained via mediastinoscopy (Med). Samples were fixed in 3% paraformaldehyde for 4 h and then transferred to 50% ethanol. Histological sections (4 μm) were prepared and stained with hematoxylin and eosin (H and E) or Masson’s trichrome stain. Tissues were analyzed by a board-certified pulmonary pathologist (M. Deen) for the extent of inflammation, including macrophage and neutrophil localization, alterations in alveolar epithelial barriers, fibrin deposition, edema, granuloma formation, and fibrosis. Representative images were acquired at high resolution (magnification 60×) using an Olympus VS120 Virtual Microscopy System, scanned and viewed using OlyVIA version 2.6 software (Olympus Life Sciences, Center Valley, PA, USA). For immunohistochemistry, tissue sections were deparaffinized with xylene (4 min, ×2) followed by decreasing concentrations of ethanol (100%–50%) and finally, water. After antigen retrieval using citrate buffer (10.2 mM sodium citrate, 0.05% Tween 20, pH 6.0) and quenching of endogenous peroxidase with 3% H2O2 for 10–30 min, sections were incubated for 1–4 h at room temperature with 5–100% goat serum to block nonspecific binding. This was followed by overnight incubation at 4 °C with rabbit IgG or rabbit polyclonal anti-cytochrome b5 (1:250, Abcam, Cambridge, MA, USA), anti-cytokeratin 17 (1:1000, Abcam), anti-heme oxygenase (HO)-1 (1:650, Enzo Life Sciences, Farmingdale, NY, USA), anti-lipocalin (Lcn)-2 (1:250, Abcam), anti-inducible nitric oxide synthase (iNOS) (1:500, Abcam), anti-cyclooxygenase (COX)-2 (1:100, Abcam), anti-tumor necrosis factor (TNF)α (1:50, Abcam), anti-ADP-ribosylation factor-like GTPase (ARL)11 (1:100, Bioss Antibodies, Woburn, MA, USA), anti-mannose receptor (MR)-1 (1:500, Abcam), anti-galectin (Gal)-3 (1:400, R&D Systems, Minneapolis, MN, USA), anti-transforming growth factor (TGF)β (1:50, Abcam), anti-histone H3 (1:50, Cell Signaling, Danvers, MA, USA), or anti-mono-methyl histone H3K4 (1:100, Cell Signaling) antibodies. Sections were then incubated with biotinylated secondary antibody (Vector Labs, Burlingame, CA, USA) for 30 min at room temperature. Binding was visualized using a DAB (3,3’ diaminobenzidine) peroxidase substrate kit (Vector Labs). Sections of lung and/or lymph nodes from all 7 subjects were analyzed for each antibody. Patient demographics are presented in Table 1. The patients were predominantly male (100%), white (non-Hispanic; 100%), and former smokers (57%). Based on pathological findings (see below), patients were divided into two groups; those with SGD (n = 5) and those with other pathology (n = 2). Of the patients with SGD, 100% were white males, overweight or obese (BMI = 25–29 (20%); BMI ≥ 30 (80%)), 80% were <50 years old (mean age 43.5), and 60% never smoked (Table 1). Four of the five patients with SGD had an intermediate WTC dust exposure level, and one had a high level of exposure [2]. The two non-SGD patients had high WTC dust exposure levels (Table 1). The clinical, radiographic and pathologic characteristics of patients undergoing biopsy are presented in Table 2. The reasons for biopsy in patients with SGD (patients #1–5) were mediastinal/hilar adenopathy and pulmonary nodules (Table 2). Other patients displayed a mix of lymphadenopathy, fibrosis, nodules, and emphysema. Of patients with SGD, patients #1 and #2 had normal pulmonary function, while patients #3 and #4 displayed restrictive physiology and patient #5 had mixed obstruction and restriction (Table 3). The two patients without SGD (#6 and #7) had restrictive physiology (Table 3).Biopsy samples from three of the seven patients (patients #1, #3 and #4) showed well-formed non-caseating granulomas in lymph nodes suggesting SGD; patient #2 exhibited scant, poorly formed granuloma in the lung (Table 2 and Figure 1). In two of the patients (#3 and #4), only lymph nodes were biopsied; in one patient (#2), only lung parenchyma was biopsied. In patient #1, both lymph nodes and lung were biopsied, and both tissue types exhibited non-caseating granulomas (Figure 1). Masson’s trichrome staining of tissues revealed the presence of well-formed granulomas and evidence of peri-granulomatous fibrosis in lymph nodes of patients #1, #3 and #4 (Figure 2). Of the other three patients without granulomatous changes (#5–#7), pathological diagnoses in the lung included mild fibrosis, bronchial inflammation and acute and chronic lung injury (Table 2 and Figure 1). Patient #5 was diagnosed later with SGD at another medical facility after obtaining additional tissue via VATS.Analysis of tissue sections by immunohistochemistry revealed the presence of cytochrome b5 and cytokeratin 17, markers of injury in lung (patients #1, # 2, #5, #6 and #7) and lymph nodes (patients #1, #3 and #4) (Figure 3). Whereas cytochrome b5 expression was uniformly distributed throughout the lung, including alveolar macrophages and epithelial cells, cytokeratin 17 was more prominent in alveolar macrophages. However, the intensity of expression of both of these markers varied between patients. Markers of oxidative stress including HO-1 and Lcn-2 were upregulated in alveolar macrophages and lymph node biopsies from all patients (Figure 4). As observed with cytochrome b5 and cytokeratin 17, the intensity of expression varied between patients. We also noted upregulation of the proinflammatory proteins iNOS, COX-2, TNFα, and ARL11 to varying degrees in alveolar macrophages and lymph nodes in all patient samples examined (Figure 5 and Figure 6). Anti-inflammatory/pro-fibrotic markers of macrophage activation, including MR-1, the galactoside-binding lectin, Gal-3, and TGFβ were also upregulated (Figure 7 and Figure 8). Additionally, histone H3 and methylated (K4) histone H3K4, epigenetic markers of inflammation/fibrosis were identified in lung and lymph nodes with varying intensities in all patient samples examined (Figure 9).The present studies report that five of the seven patients referred to our clinic for diagnostic evaluation exhibited SGD, as evidenced by non-caseating granulomas in lung and/or lymph nodes. SGD has previously been described among a subset of WTC rescue and recovery workers [9,10]. Mechanisms underlying the development of this disease remain unknown [11,12,13,14,15]. Non-caseating granuloma formation is a type of foreign body reaction which involves trapping of remnants of foreign materials that cannot be degraded and/or destroyed by macrophages [16]. The appearance of non-caseating granulomas in WTC dust-exposed patients is in line with findings that WTC dust contained organic and inorganic particles which have been implicated in the development of granulomatous pulmonary disease [17].Inflammation and oxidative stress are common responses to inhalation of particles and fibers including silica and asbestos, components of WTC dust, and they are thought to be involved in pulmonary disease pathogenesis [18,19,20]. As a first step in elucidating WTC dust-induced SGD mechanisms, we analyzed the expression of markers of oxidative stress, inflammation, and injury in lung and/or lymph nodes of WTC dust-exposed patients followed in our clinic. Lung and/or lymph node samples from all patients examined stained positively for markers of pro-inflammatory M1 macrophages including iNOS, COX-2, TNFα and ARL11 [21,22,23]. iNOS and COX-2 mediate the production of reactive nitrogen species and proinflammatory eicosanoids, respectively. These mediators are known to contribute to M1 macrophage activation and lung injury induced by diverse toxicants, and they may play a similar role in the pathogenic response to WTC dust [21]. In this regard, COX-2 has been reported to be upregulated by silica, a component of WTC dust, in rodents, cultured fibroblasts, and in sarcoid granulomas [4,18,24,25,26]. ARL11 has recently been identified as a regulator of proinflammatory macrophage activation and TNFα release [22,27]. TNFα has been implicated in lung injury induced by silica [26], suggesting a potential mechanism of disease pathogenesis following WTC dust exposure. Previous studies showed that WTC dust induces oxidative stress [5,6]. Consistent with these reports, tissue samples from all patients examined were found to stain positively for the oxidative stress markers, HO-1 and Lcn-2. In addition to their anti-oxidant activity, these proteins promote anti-inflammatory responses. Macrophage Lcn-2 has previously been shown to reduce granulomatous inflammation in mycobacterial pulmonary infections [28]. Upregulation of HO-1 and Lcn-2 in granulomas of WTC dust-exposed patients may reflect a compensatory attempt to limit inflammation and granuloma progression. Increases in proinflammatory macrophage mediators and oxidative stress were associated with lung damage as evidenced by upregulation of cytochrome b5 and cytokeratin 17 in macrophages and epithelial cells. Similar increases in these proteins have been reported in acute lung injury induced by inhaled ozone and S. aureus enterotoxin [23,29,30]. The activity of proinflammatory/cytotoxic M1 macrophages is balanced by anti-inflammatory/pro-fibrotic M2 macrophages, which downregulate inflammation and initiate wound repair. However, when overactivated, M2 macrophages promote fibrosis [21]. Findings that macrophages in histologic sections of WTC dust-exposed patients stained positively for MR-1 and Gal-3 suggest M2 macrophage activation [31]. This is in accord with findings that Gal-3 promotes M2 polarization of macrophages and contributes to lung fibrosis [32,33]. Coordinate with the presence of M2 macrophages, we found that tissue samples stained positively for the pro-fibrotic protein TGF β. Of note, patients with well-formed granulomas (#1, #3 and #4) also exhibited peri-granulomatous fibrosis, as evidenced by trichrome staining. Fibrotic sarcoidosis is characterized pathologically as fibrotic destruction at sites of prior granulomatous inflammation [34]. The granulomas are thought to function as a nidus for the development of fibrosis that can encompass larger areas of the respiratory tract, resulting in collagen deposition in broncho-vascular tracts and interlobular septae, cystic distortion and honeycombing of the lung [35,36]. It remains to be determined if peri-granulomatous fibrosis in our series is an early indicator of progressive fibrotic disease in patients with WTC dust-induced SGD. The most recent follow up of a large WTC dust-exposed firefighter cohort with SGD did not report clinical evidence of fibrotic disease [13]. Our pathological series suggests that fibrosis may develop in these patients with time. This is important as the development of fibrosis in sarcoidosis represents a significant change in the clinical course that is associated with increased morbidity and mortality [34,35]. It is important to note that expression of inflammatory proteins was observed in all patients exposed to WTC dust, regardless of the presence of granulomas. As we did not have controls to assess for comparison, a causal link has not been established. It is likely that specific patient demographics, such as smoking, obesity, age, and/or genetic/epigenetic factors influence the immune response to WTC dust injury, contributing to different pathological outcomes. In this regard, our patients with SGD were predominantly non-smokers, obese and <50 years of age at diagnosis. This is similar to the demographic profile of larger cohorts of New York City firefighters and first responders exposed to WTC dust [3,13]. Evidence suggests that sarcoidosis is more likely to develop in young, obese, non-smokers [37,38]. These factors are known to affect the adaptive immune response by inducing T-cell differentiation towards a Th1 phenotype [39,40,41,42]. This may contribute to the development of WTC dust-induced SGD. Lung and lymph node biopsies in our cohort, also stained positively for histone H3 and H3K4, suggesting the possibility that epigenetic factors may also contribute to SGD. All five patients with SGD had intermediate/high exposure to WTC dust. Previous studies demonstrated a strong correlation between WTC dust exposure levels and the development of lung disease with very high exposure levels causing the most disease [2]. Our studies suggest that intermediate levels of WTC dust exposure may be a sufficient risk factor for developing SGD with fibrosis. It is also possible that heterogeneity in WTC dust, such as composition (silica, metal) and particle size, affect the pathologic response [36]. These possibilities require further investigation. In this case series studies, we are limited in any inferences we can make about associations, since our sample size was small and non-random, and we did not have case-matched controls.In summary, our studies demonstrate the presence of SGD in five of seven patients exposed to intermediate/high levels of WTC dust. Lung and/or lymph nodes exhibited signs of oxidative stress, inflammation, tissue damage, and fibrosis. This was associated with upregulation of proinflammatory and pro-fibrotic markers and epigenetic alterations. Our observations of peri-granulomatous fibrosis in some of the patients with SGD are novel. Further studies are required to elucidate the precise mechanisms underlying this pathology in subsets of WTC dust-exposed patients.Conceptualization, V.R.S., H.K. and D.L.L.; data curation, S.H.; formal analysis, V.R.S. and J.R.; funding acquisition, H.K., I.U. and D.L.L.; investigation, V.R.S. and D.L.L.; methodology, V.R.S., K.N.V., J.C. and M.D.; project administration, D.L.L.; resources, D.L.L.; supervision, V.R.S. and D.L.L.; writing—original draft, V.R.S. and J.R.; writing—review and editing, S.H., H.K., I.U., R.L., J.S., J.D.L. and D.L.L.This work was supported by the National Institutes of Health [grant numbers: ES005022, ES004738, and AR055073], and the National Institute for Occupational Safety and Health [grant number: U100H008239].The authors declare no conflict of interest.Histology of lung and lymph nodes. Biopsies of lung (top and middle panels) and lymph nodes (bottom panels), collected from World Trade Center (WTC) rescue and recovery workers, were sectioned and stained with Hematoxylin and Eosin (H&E). Inset, patient number. Magnification, 20×; arrows, alveolar macrophages; arrowheads, lymphocytes.Trichrome staining of lung and lymph nodes. Biopsies of lung (top panel) and lymph nodes (bottom panels), collected from WTC rescue and recovery workers were sectioned and stained with Masson’s trichrome. Inset, patient number; NC, tissue not collected. Magnification, 13.2×.Cytochrome b5 and cytokeratin 17 expression in lung and/or lymph nodes collected from WTC rescue and recovery workers. Sections were immunostained with antibody to cytochrome b5 or cytokeratin 17. Binding was visualized using a 3,3’ diaminobenzidine (DAB) peroxidase substrate kit. One representative section from each patient is shown. Inset, patient number. Magnification, 60×; arrows, alveolar macrophages.Hemeoxygenase (HO)-1 and Lipocalin (Lcn)-2 expression in lung and/or lymph nodes collected from WTC rescue and recovery workers. Sections were immunostained with antibody to HO-1 or Lcn-2. Binding was visualized using a DAB peroxidase substrate kit. One representative section from each patient is shown. Inset, patient number. Magnification, 60×; arrows, alveolar macrophages.Inducible nitric oxide synthase (iNOS) and Cyclooxygenase (COX)-2 expression in lung and/or lymph nodes collected from WTC rescue and recovery workers. Sections were immunostained with antibody to iNOS or COX-2. Binding was visualized using a DAB peroxidase substrate kit. One representative section from each patient is shown. Inset, patient number. Magnification, 60×; arrows, alveolar macrophages.Tumor necrosis factor (TNF)α and ADP-ribosylation factor-like GTPase (ARL)11 expression in lung and/or lymph nodes collected from WTC rescue and recovery workers. Sections were immunostained with antibody to TNFα or ARL11. Binding was visualized using a DAB peroxidase substrate kit. One representative section from each patient is shown. Inset, patient number; NA, tissue not available. Magnification, 60×; arrows, alveolar macrophages.Mannose receptor (MR)-1 and Galectin (Gal)-3 expression in lung and/or lymph nodes collected from WTC rescue and recovery workers. Sections were immunostained with antibody to MR-1 or Gal-3. Binding was visualized using a DAB peroxidase substrate kit. One representative section from each patient is shown. Inset, patient number. Magnification, 60×; arrows, alveolar macrophages.Transforming growth factor (TGF)β expression in lung and/or lymph nodes collected from WTC rescue and recovery workers. Sections were immunostained with antibody to TGFβ. Binding was visualized using a DAB peroxidase substrate kit. One representative section from each patient is shown. Inset, patient number; NA, tissue not available. Magnification, 60×; arrows, alveolar macrophages.Histone H3 and methylated (K4) histone H3K4 expression in lung and/or lymph nodes collected from WTC rescue and recovery workers. Sections were immunostained with antibody to H3 or H3K4. Binding was visualized using a DAB peroxidase substrate kit. One representative section from each patient is shown. Inset, patient number; NA, tissue not available. Magnification, 60×; arrows, alveolar macrophages.Demographic characteristics of patients undergoing biopsy.Abbreviations: SGD, sarcoid-like granulomatous disease; BMI, body mass index; WTC, World Trade Center.Clinical, radiographic and pathologic presentation of patients.Abbreviations: PT#, patient number; CT, computerized tomography; LAD, lymphadenopathy; VATS, video-assisted thoracoscopy; Med, mediastinoscopy; TBB, trans-bronchial biopsy; SGD, sarcoid-like granulomatous disease; NC, sample not collected.Pulmonary function test findings in patients undergoing biopsy.Abbreviations: PT#: patient number; FEV1: forced expiratory volume in 1 s; FVC: forced vital capacity; VC: vital capacity; TLC: total lung capacity; DLCO: diffusing capacity for carbon monoxide; WNL: within normal limits; RTLD: restrictive thoracic lung disease; RILD: restrictive interstitial lung disease; OLD: obstructive lung disease. Bolded values were below normal confidence intervals.