ijerph2

Med-MDPI/ijerph_2/ijerph-14-03-00283.txt

@@ -0,0 +1 @@
+Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).Over one million women are employed in child care and are among the lowest wage workers in the US. The health and working conditions of 674 child care workers (118 administrators and 556 staff) from 74 centers is described using baseline data from a larger intervention trial. Participants were 39.9 (±13.0) years old; 55.4% African American, 37.1% Caucasian, and 5.3% of Hispanic ethnicity. Seventy-six percent reported having an Associate’s degree or less; 42% were classified as at or below poverty (<$20,000); and exhibited many health risks such as excess weight, insufficient activity, poor diet, and inadequate sleep. We investigated potential differences by income and job category. Lower income participants were significantly more likely to be current smokers (19.9% vs. 11.7%), drink more sweetened beverages (1.9 vs. 1.5), and report higher depressive symptoms (15.5 vs. 12.6). Administrators worked more hours weekly compared to staff (46.4 vs. 40.6), are less active (100 vs. 126 min/week), more sedentary (501 vs. 477 min/day), and reported higher job demands (13.3 vs. 12.5). Given the numerous health issues and challenging work conditions, we hope our results serve as a call to action for addressing low wages and the work environment as a means of influencing the health and well-being of child care workers.There are more than 1.3 million child care workers in the U.S. [1] of which 95.5% are women [2]. In fact, child care is one of the 10 most prevalent occupations for women [2]. While they are entrusted with our most valuable asset, our children, child care workers are among the lowest paid occupations, often having earnings below poverty level [3,4]. The 2016 Occupation Outlook Handbook reports that the average pay for child care workers in the U.S. is $9.77 per hour or $20,320 annually [1]. As low wage workers, they likely face higher rates of chronic disease and shorter life expectancies [5,6,7]. Furthermore, low wage workers often experience challenging working conditions as demonstrated in a review by Landsbergis and colleagues who found that low wage workers were more likely to experience higher job insecurity and job strain than higher wage workers [8].Very little information is available about the health status of child care workers. The largest study with more than 2000 female child care workers in Pennsylvania found that 75% had one or more chronic health conditions and 20% had three or more conditions [9]. However, the sample was limited to employees at Head Start, a federally-sponsored program that provides comprehensive early childhood education, health, nutrition, and parent engagement services to low-income children and their families. While Head Start programs have more standardized organizational structure and financial support compared to center-based programs, the study suggests that many of the workers were still struggling with multiple health issues. Among the few other studies with child care workers, results suggest elevated levels of emotional distress [10], multiple health risk behaviors [11], and increased prevalence of overweight and obesity [12].Working conditions such as long hours, high job demands, low wages, lack of health benefits, and high turn-over may be impacting the health of child care workers. A recent review by Cumming on child care workers’ well-being identified 30 studies conducted in the U.S. and abroad that help document these challenging working conditions, several of which also demonstrated how these work conditions are related to adverse effects on workers’ psychological and emotional health [13]. Other studies have shown that the working conditions of the child care job can place workers at risk for infectious diseases, injuries, and other occupational hazards [14,15,16]. Unfortunately, very few studies have examined the relationship between working conditions and workers’ physical and emotional health status and/or health behaviors.The purpose of this study is to describe the health status, health behaviors, and working conditions of child care workers and to explore how income and job position may be associated. We believe considering working conditions will address an important gap in the literature about the health of child care workers. We will also discuss the implications of these results for both practice and research that might improve the health and working conditions for child care workers in the future.This study used baseline data from a group randomized control trial (ClinicalTrials.gov registration number NCT02381938) conducted in North Carolina to evaluate the impact of a child care-based worksite wellness intervention. For the purpose of this paper, we used baseline data only, collected prior to randomization and intervention implementation. All protocols were approved by the Institutional Review Board at the University of North Carolina at Chapel Hill (#13-2438).A convenience sample of child care centers and their workers, both administrators and staff, were recruited to join a group randomized control research trial using a multi-phase recruitment strategy. We targeted counties in North Carolina where we had existing community partners that could help facilitate an introduction to local child care centers. Moreover, we targeted geographical areas that had sufficient numbers of medium to large centers to ensure that we would reach recruitment goals. Across four waves, we recruited seven counties that represented a combination of rural, suburban, and urban area, moderate to low income, and similar racial and ethnic populations compared to the state. Centers in these counties were identified from an online database of licensed child care programs available through the North Carolina Division of Child Development and Early Education [17]. Following introductions from the community partners, the research team initiated direct contact with local child care centers via mail/email letters of invitation and follow up telephone calls. During calls, research assistants confirmed center eligibility, reviewed study details, and assessed center interest. Eligible centers had to be in business for at least two years with no plans to close in the following 18 months and have at least four employees (one administrator and three additional staff) willing to participate. Once center eligibility was confirmed, research assistants met with interested staff in-person to confirm eligibility, review study details, and collect informed consent. Eligible staff had to be at least 18 years old, able to read and speak English, and pass a physical activity readiness screening (PAR-Q+ [18]) (and obtain medical approval if required).Participants completed several measures, including physical assessments, physical activity monitoring, and web-based and paper surveys. Data were collected primarily during an onsite visit by data collectors who were trained and certified on all protocols. Specific measures are described below:Physical Assessments. Height, weight, waist circumference, heart rate, and blood pressure were assessed using standard protocols [19,20]. Height was measured to the nearest 1/8 inch with a Shorr measuring board (Shorr Productions, Olney, MD, USA). Weight was measured to the nearest 0.1 pound with a Seca model 874 portable electronic scale (Seca Corporation, Columbia, MD, USA). Height and weight measures were taken with shoes and heavy clothing removed. Waist circumference was measured with a Gulick II (Patterson Medical, Warrenville, IL, USA) measuring tape to the nearest 0.1 cm at the level of the iliac crest. Heart rate and blood pressure were assessed using a digital blood pressure monitor (Omron, Kyoto, Japan) while participants sat quietly with their legs uncrossed and feet flat on the floor. All measures were taken twice. If the two measurements were within reasonable agreement (≤¼ inch for height, ≤1.0 lb. for weight, ≤3.0 cm for waist circumference, and <5 mmHg for systolic and diastolic blood pressure) the final measure was recorded as the average of the two values. Otherwise, a third measure was taken. Height and weight were used to calculate body mass index (BMI) (weight in kg/height in m2) and weight status (underweight = BMI < 18.5, normal weight = BMI 18.5–24.9, overweight = BMI 25.0–29.9, obese = BMI ≥ 30.0). Systolic and diastolic blood pressure were used to determine mean arterial pressure.Physical Activity Monitoring. Physical activity was assessed using one week of accelerometer monitoring using a GT3X ActiGraph monitor (ActiGraph LLC, Pensacola, FL, USA). Participants were fit with a belt that allowed the monitor to be worn over the right hip and were instructed to wear the monitor for the next seven days, 24 h a day except during water activities. Monitors were programmed to sample at 30 Hz. Data from the monitor were downloaded using ActiLife software (ActiGraph LLC). Adult-specific cut points were applied to determine minutes of sedentary activity (0–100 counts/min, <1.5 METS) and minutes of moderate-to-vigorous physical activity (MVPA) (≥2020 counts/min, 3.0 METS) [21,22] for participants with at least eight hours of wear time on a minimum of four days.Web-Based Surveys. Health behaviors were assessed using a series of self-administered surveys collected via an online system known as the Carolina Health Assessment and Research Tool (CHART) [23]. More specifically, CHART assessed dietary intake, tobacco and e-cigarette use, sleep, and emotional health.Dietary intake was assessed using items from the Dietary Screener Questionnaire [24,25] and the Diet History Questionnaire [26]. Within this survey, five items asked specifically about intake of fruits, fruit juice, potatoes, beans and legumes, and other vegetables; and two items asked about intake of sugar-sweetened and artificially-sweetened beverages. The five items related to fruits and vegetables were summed to get a total daily fruit and vegetable intake; and the two items related to beverages were summed to get a total sweetened beverage intake.Tobacco and e-cigarette use was assessed using two items modified from the 2012 Behavioral Risk Factor Surveillance System Questionnaire (BRFSS) [27] that asks participants to estimate the average use of each during the past 30 days. Items were used to identify participants that are current smokers or had ever used e-cigarettes.Sleep duration was assessed using one item extracted from the Pittsburgh Sleep Quality Index (PSQI) [28], which asked participants to estimate their typical hours of sleep per night during the past 30 days.Emotional health was assessed using a combination of indices related to distress, depressive symptoms and job strain. Distress was measured using one item from the Society for Behavioral Medicine’s Common Data Elements [29] which asked participants to rate their level of distress during the past week on a scale of 0 to 10 (0 = no distress, 10 = extreme distress). Depressive symptoms were measured using the 20-item Center for Epidemiologic Studies Depression Scale (CES-D) [30]. As specified in standard scoring protocols, item responses (scored 0–3) are reverse-coded when appropriate and summed to create a total depressive symptom score (ranging from 0−60) with ≥16 indicating high risk of clinical depression. Job strain indicators were measured with 15 items derived from the Job Content Questionnaire (JCQ) [31], specifically items required to calculate job demand (scores ranging from 5−20), job control (scores ranging from 12−48), and job insecurity (0,1). These items ranged from 1 (strongly disagree) to 4 (strongly agree) and were reverse-coded when necessary. Job demands was composed of 5 items (e.g., “My job requires working very fast”) and a higher score indicates more job demands. Job control is comprised of two subscales, skill discretion (6 items, e.g., “My job requires me to learn new things”) and decision authority (3 items, e.g., “I have a lot to say about what happens at my job”) and a higher score indicates less job control. Job insecure participants answered “Strongly disagree” or “Disagree” to “My job security is good” [32].The web-based surveys also captured demographic variables including age, sex, race, ethnicity, education, marriage status, household size, household income, and hours worked (at the enrolled childcare center plus any from a second job). Household income (assessed categorically) was used to identify participants as below poverty (<$20,000) or above poverty (≥$20,000). The below poverty cut point was based on the median household size in our sample (three) and income thresholds set in the 2016 Federal Poverty Guidelines [33] ($20,160 for a household of three). Seventy-four participants declined to answer this question, and were thus considered missing for income-related analyses.Among these demographic items, there was also a question that asked participants to identify their role at the center (e.g., job position), with response options including owner, director, assistant director, lead teacher, assistant teacher, and other. Administrators (defined as owner, director, or assistant director) were asked additional questions about center demographics, including years of operation, hours of operation, enrollment fees, star rating, number of kids in care, number of employees, center type, current participation in the Child Adult Care Food Program (CACFP), and accreditation by the National Association for the Education of Young Children (NAEYC). CACFP is a federally-funded U.S. program that provides reimbursement for meals and snacks served in child care programs that serve low-income families. The NC’s Quality Rating and Improvement Systems (QRIS) [34] star rating is an indicator of quality based on a five-point scale, where a rating of one star corresponds to meeting minimum licensing standards and five stars represents the highest quality and voluntarily compliance with higher standards related to programming and staff education.First, we examined descriptive summary statistics (Table 1 and Table 2). Drawing from the measures described above, we examined demographic characteristics (e.g., age, sex, race and ethnicity, education), indicators of health status and behavior (e.g., weight, waist circumference, heart rate, blood pressure, diet, physical activity, cigarette and e-cigarette use, sleep, emotional health), and indicators of working conditions (e.g., hours worked, job demand, job control, job security). Next, we explored differences in health status, health behaviors, and working conditions by income status (below vs. above $20,000) and job position (administrator vs. staff). Differences were assessed using chi-square tests (categorical outcomes) or two independent sample t-tests (continuous outcomes) using SAS software 9.4 (SAS Institute Inc., Cary, NC, USA). Due to the high number of hypothesis tests conducted, the raw p-values were adjusted to control for the overall Type I error rate; however, these adjusted p-values did not change any conclusions about statistical significance and thus the raw p-values are presented for simplicity in Table 3 and Table 4. While the full sample included 697 child care workers from 74 centers, we narrowed the sample to include only the 674 female workers (118 administrators and 556 staff) for this study. Demographic characteristics of these participants are presented in Table 1. On average, participants were 39.9 (±13.0) years old. The majority (55.4%) were African American, 37.1% were Caucasian, and 5.3% were of Hispanic ethnicity. The majority (76.0%) of participants attained an Associate’s degree or less. Forty two percent of participants were classified as at or below poverty (based on our cut point of <$20,000).We explored demographic differences by job position (administrators vs. staff). Administrators were slightly older than staff (43.8 years vs. 39.1 years); were more likely to be White (45.7% vs. 35.3%); attained more education (46.6% having completed a bachelor’s degree or higher vs. 19.3% of staff); and had higher household incomes (89.1% reported an income greater than $20,000 vs. 44.7% of staff). Administrators were more likely to be married or living with a partner (63.6% vs. 47.3%). No differences were observed between administrators and staff regarding household size.Seventy-four child care centers were enrolled in this study, demographics for which are described in Table 2. Participating centers had a high quality rating, averaging 4.3 (±0.7) stars. Centers had, on average, 56.3 (±32.5) children enrolled and 12.1 (±8.6) staff employed. Centers reported being open for 13.1 (±3.4) h per day and were predominately privately owned (68.1%), though nearly all accepted subsidies (97.3%) and participated in CACFP (84.7%).Participant health indicators are presented in Table 3. Participants appeared to have many indicators of poor health status and behavior. Participants’ average BMI was 34.5 (±9.0) with 22.2% of participants classified as overweight and 66.3% classified as obese. Average waist circumference was 106.5 cm (±18.2), well above the 88.0 cm cut point for women that is associated with increased risk for disease [35]. Twenty-six percent of participants had high blood pressure (defined as pressures at or above 140/90). Although we are unable to report on the use of blood pressure medication, 33.5% of our sample reported on the PARQ+ (from the eligibility screening) being told by a physician that they have high blood pressure.Regarding health behaviors, participants accumulated an average of 122 (±104) min per week of MVPA, which is below the recommended 150 min per week [36]. Also, participants accumulated 481 (±72.4) min per day of sedentary time, roughly 8 h a day. Participants reported eating fruits and vegetables an average of 2.58 (±1.72) times per day and drank sweetened beverages on average 1.71 (±1.82) times per day. Even if participants consumed a cup each time they ate fruits or vegetables, they still consumed far below the 3.5–4.5 cups per day that is recommended for women [37]. Overall, 15.6% were current smokers, and 9.9% reported having used an e-cigarette. On average, participants reported sleeping 6.37 (±1.35) h per night, lower than the 7–8 h per night that is recommended [38]. Perceived level of distress was reported as 4.02 (±2.78) and CES-D depression score was 13.9 (±9.17) with 36.1% reporting scores at or above 16 (the criteria for clinical depression).Significant differences were noted between participants making below $20,000 compared to those with an income above $20,000. Specifically, lower income participants were significantly more likely to be current smokers (19.9% vs. 11.7%, p = 0.0057), drink more sugar-sweetened beverages on a daily basis (1.9 vs. 1.5, p = 0.0057), and report higher depressive symptoms (15.5 vs. 12.6, p = 0.0002). A higher percentage of lower income participants were classified as at or above the typical cutoff of 16 (41.6% vs. 31.3%, p = 0.0094). The lowest income participants (below $20,000) were also less sedentary (468 min/day vs. 491 min/day, p = 0.0006). There were few differences in health indicators by job position; however, compared with staff, administrators accumulated an average of 26 min less of MVPA per week (126 vs. 100 min/week, p = 0.0048) and 24 more minutes of sedentary time per day (477 vs. 501 min/day, p = 0.0051).Job-related health risk indicators are reported in Table 4. On average, participants reported that they worked 41.6 (±11.8) h per week. On average, they rated their job demand as 12.6 (±2.2) and job control as 24.3 (±5.18).There were statistically significant differences in working conditions by income and job position. Not surprisingly, participants making less than $20,000 worked fewer hours compared to those making above $20,000 (39.8 vs. 43.5 h, p = 0.0001). The lowest income participants had less job control, on average (reverse coding, 25.1 vs. 23.4, p = 0.0001) and lower job demands (12.4 vs. 12.8, p = 0.0355). Administrators worked more hours compared to staff (46.4 vs. 40.6 h, p = 0.0001) and were less likely to report perceived job insecurity (6.03% vs. 12.5%, p = 0.0468). Administrators reported higher job demands (13.3 vs. 12.5, p = 0.0007) and better job control (21.6 vs. 24.9, p = 0.0001).This study describes health indicators on a sample of child care workers and selected contextual factors that provide insight into their work environment. Our results align with national data that child care workers are truly low wage workers. Our data also suggest that these workers exhibit many health risks such as excess weight, insufficient activity, unhealthy diet, inadequate sleep, and depressive symptoms. In addition to the hardship posed by low wages, our results confirm challenges of their working conditions such as long hours and high job demands and low job control. Also, this is the first study to explore differences in health and working conditions by household income (±$20,000) and job position (administrators vs. staff). Below, we emphasize the importance of these results in relation to existing literature, with a goal of improving future research and practice with child care workers.The results of this study offer a valuable contribution to research into the health of child care workers, a population that has been largely ignored. Obesity (not just overweight) was an issue for the majority of child care workers in this study. Estimates from our study as well as the study by Sharma and colleagues of Head Start teachers, indicate a higher obesity prevalence among child care workers compared to the general U.S. population of adult women (66.3% and 54.5% vs. 40.4%, respectively) [12,39]. Obesity, in turn, increases risk for a wide array of chronic diseases, including cancer, heart disease, diabetes, kidney disease, and arthritis [40,41,42]. Interestingly, a greater portion of child care workers in this study appear to be sufficiently active compared to the general US population (27.8% vs. 10.7%, when applying the same cut points) [43]. These child care workers reported dietary intake and sleep patterns that are similar to the general population, but again behaviors fall short of national recommendations for overall health. For example, child care workers in our study reported eating fruits and vegetables an average of 2.6 times per day, which is similar to the average of 2.7 servings consumed by adults nationwide [44]; however, both groups fall short of the 3.5–4.5 cups recommended for women [37]. Similarly, child care workers in this study reported 6.4 h of sleep per night, which is slightly less than 6.9 h of sleep per night that most US adults report; however, both are slightly below national recommendations of 7–8 h per night [38].An alarming 36.1% of participants in this study reported CES-D depression scores at or above the criteria for clinical depression, which far exceeds the national rate of depression for Americans (7.6%) and the rate among women between 40–59 years old (12.3%) [45]. This finding corroborates previous research demonstrating elevated levels of depressive symptoms among child care workers, including one study that found depressive symptoms among a nearly a quarter (24%) of child care staff in Head Start programs in Pennsylvania [46,47]. Depression and obesity often co-occur, so that our results warrant further investigation into the reasons why child care workers have high rates of both conditions, and, to explore effective ways of reducing their incidence and unhealthy impacts.Our results also indicate that the lowest paid childcare workers are more likely to report multiple unhealthy behaviors. For example, they are more likely to be current smokers, drink more sweetened beverages, and report higher depressive symptoms. One potential explanation for this pattern of unhealthy behaviors is that lower wage workers are using food, beverages, or cigarettes as a way to cope with challenging work and/or financial conditions. While we cannot be certain of this explanation, we know that these behavioral choices contribute to poor health [48,49,50], and there is evidence from the literature to suggest that lower income populations are more likely to engage in these behaviors [51,52,53]. Future examinations of child care worker health behaviors would benefit from qualitative research that examines how and why these individuals are more likely to smoke or drink sweetened beverages. Then, the next generation of interventions could be tailored to the expressed needs of these individuals. For example, stress management may be a critical component of dietary and/or smoking interventions. And, interventions may need to address the underlying issues related to financial strain, either by offering assistance with finance management strategies and/or advocating for living wages.In addition to the numerous health issues faced by child care workers, our study also highlights their challenging working conditions, including differences by job position. Consistent with existing literature, our study should serve as a call to action for addressing the child care work environment and its impact on workers��� stress and well-being [13]. Our results indicate that administrators report higher demands and higher job control than staff. We may want to investigate ways to increase the job control of the lowest income childcare workers, which are typically staff. A qualitative study by Faulkner and colleagues with home-based and center-based child care workers found that common stressors were parental interactions, caregiving, and the failure of public perception to see child care as a profession [54]. Child care workers also reported sleep disruptions (e.g., dreaming about children/work), and physical exhaustion. Child care work can be challenging, especially for administrators, who are the key gatekeepers to the child care setting. Although new interventions could be helpful, researchers should consider the readiness and capacity of child care administrators and staff when developing interventions so as not to add unnecessary burden to workers as part of well-intentioned initiatives.Our study offers many lessons to help inform future child care-based interventions. Findings emphasize the importance of child care as a setting through which to target health initiatives, especially for those wanting to intervene with low-income women. Like other low wage earners, child care workers experience many risks to their health and well-being. These risks sometimes affect child care workers differently, based on their income and job position. Thus, future efforts to improve the health of child care workers will benefit from multi-level interventions that not only promote healthy behaviors, but also address underlying and interconnected issues related to living wages, health care benefits, and working conditions. Child care workers would also benefit from a coordinated approach to health that not only addresses physical inactivity and dietary behaviors, but also stress management and healthy coping skills. Additionally, it is not enough to focus only on the child care worker, we must address the directors, supervisors, and conditions under which these individuals operate, e.g., the entire work environment. Since child care centers are considered small businesses, we know that these organizations are less likely to provide worksite wellness and health promotion programs, policies, and environmental supports than larger employers [55,56,57]. With more than one million child care workers nationwide, many of whom are women, we need to build the evidence base for effective interventions that are tailored to this important segment of the workforce.We acknowledge several strengths and limitations of this study. A major strength of this study is its contribution to what is currently a very limited body of literature on the health of child care workers. Our study is unique in that it includes data on the child care worker, her health status and health behaviors, and her working conditions. Another strength is the use of objective measures of several physical health indicators (e.g., physical activity, weight, waist circumference, blood pressure). Study limitations are related to cross-sectional data, possible self-selection bias, self-report bias, and unmeasured factors that may impact our findings. Specifically, because our data are cross-sectional at baseline, we cannot establish temporality of our results. We expect to be able to explore changes over time in the larger study which will have multiple measures over time. Another limitation is that we cannot generalize our findings beyond the sample of child care workers in North Carolina due to potential selection bias both at the center and participant levels. Although we cannot be certain of the impact of the bias, it is plausible that volunteers willing to participate in the worksite wellness intervention trial may be healthier than who do not participate. However, upon comparing characteristics of our final sample to the workforce in North Carolina using data from the 2015 Child Care Workforce Survey [58] we found that they appear to be similar in terms of income, education, and quality rating. While we used mostly self-report data which introduces another source of bias, we used primarily well-established instruments with sound psychometric evidence whenever available. In addition, our results may be influenced by unmeasured factors. For example, we did not measure years of work experience, but we know that in North Carolina, directors are in their positions, on average, for 6.4 years, teachers for 3.6 years, and teachers assistants for 2.5 years [58]. So the results we report based on income and/or job position may be better understood if we knew length of time in child care. It is also likely that the relationships we discovered are due to an overlap between income and job position. We did not collect data on the relationship between administrators and staff which can contribute to high job strain if these relationships are negative or otherwise unsupportive. And, we have no information about the stress that child care workers may be under at home which will also likely influence health [59,60]. These are several examples of unmeasured variables that could provide additional insights about our results and future studies should consider.The next generation of research might benefit from considering integrated interventions that address both health promotion and occupational health and safety. The National Institute of Occupational Safety and Health (NIOSH) is advocating for “Total Worker Health” interventions that may be particularly appropriate for this group of workers and workplaces [61,62]. Moreover, future research should include mixed methods studies that would explore reasons why child care workers practice unhealthy behaviors or rate work experiences as high demand/low control; as well as work with center administrators to determine who can best influence the policies and practices in place at child care centers. Promoting the health of child care workers at the workplace with health programs, policies, and environmental supports, along with higher wages, is critical.Child care centers are located in all states and employ over 1.3 million workers nationally [4]. While our results are specific to those who participated in this study, these findings provide useful insights for the larger population of child care workers nationally. These results represent an important call to action for researchers, policy makers, and community leaders who can advocate for living wages for these important members of the workforce, and plan interventions to improve the health and well-being of child care workers in the context of their everyday work conditions.We would like to acknowledge the study interventionist, Ellie Morris, as well as the research assistants, data collectors, and all participating child care staff in this study. This research was funded by the National Heart, Lung, And Blood Institute of the National Institutes of Health (NIH) under award number R01HL119568. This project was conducted out of the Center for Health Promotion and Disease Prevention, a Prevention Research Center funded through a cooperative agreement with the Center for Disease Control and Prevention (CDC) under award number U48DP001944. The content is solely the responsibility of the authors and does not necessarily represent the official views of the NIH or the CDC.Laura Linnan, Dianne Ward, Amber Vaughn, and Lori A. Bateman conceived and designed the study. Lori A. Bateman and Natalie Smith led data collection and processing. Natalie Smith undertook statistical/data analysis. Laura Linnan, Dianne Ward, and Natalie Smith led interpretation of data. Laura Linnan, Gabriela Arandia, Lori A. Bateman, Amber Vaughn, Natalie Smith, and Dianne Ward drafted the manuscript. Laura Linnan, Dianne Ward, and Amber Vaughn obtained funding to carry out the study. All authors were involved in preparing the outline of the manuscript, making comments on the manuscript, and approved the final version of the article. The authors declare no conflict of interest.Participant demographics.Childcare center descriptors.* Average enrollment fee for a 3–5 year old.Physical and mental health risk indicators.* p < 0.05.Job-related health risk indicators.* p < 0.05.

Med-MDPI/ijerph_2/ijerph-14-03-00284.txt

@@ -0,0 +1 @@
+Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).Caring for HIV/AIDS patients and/or drug users requires health care workers (HCWs) to have good knowledge of the issues. Cultural differences in HCWs, combined with professional ethics and personal beliefs, could also result in conflicting attitudes, leading to difficulties related to looking after people affected by HIV/AIDS or drug users. A cross-sectional study was carried out to assess the attitude towards HIV/AIDS patients and/or drug users in a sample of workers operating in a large university hospital in southern Italy. A total of 736 workers were surveyed from May to November 2016. During the periodic occupational health surveillance, a questionnaire was administered about attitudes of discrimination, acceptance and fear towards these patients. Respondents showed average levels of acceptance to HIV/AIDS and drug user patients. As years of experience and professional training increased, scores for discrimination, acceptance of HIV/AIDS, acceptance of drug users and fear decreased. Factors positively influencing levels of attitudes were being female and younger. Supplementary education is needed to strengthen the awareness of HCWs.People living with HIV/AIDS require ongoing health care services as they are potentially at increased risk of developing disorders including cardiovascular and liver disease, accelerated bone loss, metabolic disorders, etc. [1,2,3]. Those able to access medical care are living longer and have improved their health thanks to antiretroviral medications. These patients are experiencing acute HIV-related chronic episodes and other types of illness that can require hospitalization and/or supportive care arrangements [4,5]. Several studies have investigated the attitudes, knowledge and practices of health care workers (HCWs) towards patients with HIV/AIDS and underlined that HCWs still fear the disease and behave prejudicially toward HIV/AIDS patients [6,7,8,9,10,11].Factors which influence these attitudes include fear of contagion associated with the uncertainty of care and the awareness of feeling useless in providing care for patients with a potentially fatal disease [12].Drug users are hospitalized more frequently than the general population [13,14,15,16]. Within hospitals, people who use substances encounter significant barriers to access care [4,17]. They are often labeled as being “challenging, manipulative, drug-seeking and demanding” by HCWs [18,19]; they also meet difficulties, receive substandard care and frequently leave hospitals against medical advice [20,21,22].Taking care of HIV/AIDS patients and/or drug users requires HCWs to have good knowledge of their unique issues. Cultural differences in HCWs, combined with professional ethics and personal beliefs, could also result in conflicting attitudes, which may lead to difficulties related to caring for HIV/AIDS patients and/or drug users [4,12].The purpose of this study was to assess the attitude towards patients affected by HIV/AIDS and/or drug users in a sample of HCWs operating in a large university hospital in southern Italy.A cross-sectional study was conducted from May to November 2016 at the University Hospital of Catania (Italy) where the health care workforce consists of ≅2800 HCWs.The study was performed as part of the periodic occupational health surveillance and it required no formal approval by the local ethics committee, which was nevertheless consulted and granted its informal authorization. Participants were informed about the study aims and procedures and gave their written informed consent to participate.This study was conducted using a self-administered multiple-choice questionnaire developed by See et al. [23]. This tool allows us to evaluate four aspects: discrimination, acceptance of HIV/AIDS patients, acceptance of drug users and fear. For each of the four aspects there were four multiple choice questions with answers ranging from: “strongly disagree” = 0, “disagree” = 1, “agree” = 2, “strongly agree” = 3. Negative questions were reverse-scored to ensure that the direction was consistent with all items and higher scores represented a more positive professional attitude [23]. The original version of the questionnaire was translated into Italian by an expert mother tongue translator.A further section was added in order to collect information about age, gender, schooling, professional training and occupational history.After emphasizing the importance of the topic, the questionnaire was explained and distributed by two trained occupational physicians, then the self-administered questionnaires were completed by the HCWs in an anonymous and voluntary manner.In order to proceed with a statistical analysis, HCWs were grouped according to their tasks: physicians, graduate sanitary (biologists, physicists, chemists, psychologists); nurses and midwives; healthcare assistance staff (physiotherapists, orthoptists, logotherapists); healthcare diagnostic staff (lab technicians, radio technicians, audiometric technicians, neurophysiopathology technicians). Then, employees were divided by work environment: surgery, medicine and services. The latter include support departments like anesthesiology, radiology, laboratories.Data were analyzed with the software SPSS 22.0 (SPSS Inc., Chicago, IL, USA) for Windows. Descriptive analyses were performed using frequencies’ percentages. Scores were reported as mean ± standard deviation (SD).For the bivariate analysis, t-tests were performed to evaluate differences in quantitative variables. The one-way variance analysis (ANOVA) was used to determine any statistically significant differences between the groups’ means.A logistic regression model was used to detect possible factors associated with attitude status and age, gender, professional training and service years. Moreover, gender and age, as possible confounders, were included in the regression model when analyzing professional training and service years. Results are expressed as odds ratio (OR) with 95% CI. The level of significance was set at p ≤ 0.05.In this study, 736 HCWs were examined by university occupational doctors; all participants were asked to complete the questionnaire, which was administered to 713 workers (97% response rate), whereas 3% (n = 23) of the HCWs refused to participate in the survey, owing to lack of time needed to reply to questions.A total of 353 (48%) were male, with a mean age of 41.2 ± 16.7 years and a duration of employment of 15.4 ± 12.3. The characteristics of the sample are reported in Table 1. Over 80% of the sample was made up of physicians (40%) and nurses and midwives (42%). The remaining part (18%) was health care assistance staff, health care diagnostic staff and graduate sanitary. However, the sample was equally distributed in a cross three work environments: surgery, medicine and services.Table 2 shows the scores of the questionnaires. Female HCWs showed more statistically significant acceptance towards drug users than male HCWs, and expressed no negative emotions such as discrimination and fear.As age increased, a progressive reduction of the scores was observed in all four aspects: discrimination, acceptance of HIV/AIDS patients, acceptance of drug users and fear.The ANOVA test showed statistically significant reductions in age bands >30 years as to discrimination and acceptance of HIV/AIDS, and >40 years as to the acceptance of drug users and ≥50 as to fear.There is a statistically significant gap between physicians/nurses/midwives and the remaining investigated staff in all four aspects. The former show significantly greater discrimination and fear, with significantly less acceptance and fear, and so do those working in the surgery department and services areas.The health care assistance staff only showed a significant score in relation to fear.The results of the logistic regression (see Table 3) pinpoint how male subjects show greater discrimination and fear than female workers. Being over 40 is considered a risk factor related to a discriminatory attitude, acceptance of drug users and fear in general. After 50 years old, the workers’ level of acceptance of HIV/AIDS patients decreases. Having more than three years of professional training in biological risks and drug user management turns out to be a risk factor, given all the attitudes analyzed. The same happens for those with more than 11 years of service.The questionnaire used in this survey was developed by See et al. [23] and modified for the purpose of this study. The response rate was 97%, and thus it well represents the sample analyzed.As observed by See et al. [23], the questionnaire was reliable and valid for assessing the professional attitude of HCWs towards serving HIV/AIDS patients and/or drug users.The results allow us to observe that discrimination and fear scores significantly decreased among HCWs by age, suggesting poorer professional attitudes within these domains, while the acceptance of HIV/AIDS and/or drug users decreased if the sample age increased.From the data of our study it is observed that being older than 40 seems to be a significant risk factor in terms of discriminatory attitudes, low tolerance towards drug users and the generation of fear.Furthermore, starting from 50 years of age, employees manifest poor tolerance (acceptance) of HIV/AIDS patients.Similar results have been reported by other surveys [6,7,8,9,10,11] that dealt with HCWs’ attitudes towards HIV patients.Logistic regression highlighted how male workers show discrimination and fear attitudes to a greater extent than female ones. These data were observed for the first time after this survey.The analysis of the scores to the 16 questions, asked in relation to the role/task of each HCW, allows to detect that physicians, nurses and midwives show similar values. This, as already reported in a previous survey conducted by See et al. [23], showed that HCWs, who are more exposed to biological risks, accept HIV/AIDS patients and drug users in a similar manner.Although HCWs had negative feelings about HIV/AIDS patients [24], their integrity allowed them to overcome their fear of HIV infection [25,26]. Thus, HCWs accept both HIV/AIDS patients and drug users in a similar way [23].In the same fashion, the years of service seem to be an unfavorable factor, as HCWs with more than 11 years of service showed greater discriminatory attitudes, poor tolerance towards HIV/AIDS patients and drug users and more fear. These data are in line with the observation according to which older workers develop significantly negative feelings towards these kinds of patients compared to younger HCWs.It may be that such results may be accounted for by the fact that the cultural environment they grew up in allowed younger HCWs to get to know these pathologies more deeply, as well as the social levels where it is usually possible to detect them [7].Indeed, many people believe that HIV/AIDS is contracted due to immoral behavior, such as risky sexual intercourse and illegal use of drugs [20,23].Predictors for a positive attitude were previous experience in caring for HIV/AIDS patients, age and having no children [10].Besides, as far as the information/education of HCWs about biological risks and drug user management is concerned, they seem to have an essential role, since there are better results in those employees who received specific training less than three years before. Additionally, the scores regarding discrimination, tolerance and fear were significantly lower.In a recent survey on nurses conducted by Marranzano et al. [8], it was observed that training, competence in the care of patients with HIV/AIDS and the prevention of HIV-related occupational risks are of paramount importance.In the same way, several studies have shown that the competence of physicians and nurses in the care and prevention of alcohol and drug addiction is strategically important [27,28,29].Unlike many other diseases, HIV/AIDS cannot be cured. Its symptoms can become severe, and it is associated with stigma. Hence, many HCWs are afraid of contact with HIV/AIDS patients [9,10]. In the first 10 years after HIV/AIDS appeared, many studies discussed the reluctance of HCWs to have contact with HIV/AIDS patients due to fear of contracting the virus [10,11,12].As far as drug and alcohol addicts are concerned, many physicians and nurses took on a reluctant behavior in approaching these patients, fearing they could lie about their use of substances [27,30,31]. The score of fear in this study indicates that HCWs are still scared of HIV patients’ and drug users’ behavior.As observed in other studies, the positive attitudes of HCWs have been correlated with high levels of HIV/AIDS and alcohol consumption knowledge [5,10,11,12,13,19], and also through on-going training [8].Moreover, as described by others, it is important to examine not only the effectiveness of the program in changing the awareness, attitudes and behavior of HCWs, but the durability of its impact, even through follow-up program carried out every six to 12 months [8,10,25].Drawbacks of the present study relate to getting results from a single hospital; therefore, they might not reflect HCWs’ attitudes and concern in other hospitals in Italy, even though the results are comparable to those observed in other surveys [6,7,8,9,10,11,23].Furthermore, despite the questionnaire being self-administered, the intervention of two occupational physicians may have partially influenced the participants’ responses.In conclusion, in order to improve HCWs’ attitude towards patients, there are several possibilities that involve various professional figures, i.e., toxicologists, infectivologists, psychologists, etc. As far as the occupational medicine competence area is concerned, it seems to be of great importance to set up intense education programs, to spread information about contagion pathways, and to use general and personal protection devices by adopting safety procedures in accordance with internationally recognized organizations’ guidelines, preventing accidents and/or inconveniences to patients.In order for these education programs to effectively reach the HCWs, they should be put among the hospital’s strategic objectives by the hospital management. Such programs should be mandatory every three years and their attendance should confer credits to the participants’ skills and acquired job-content knowledge and implementation; besides, follow-ups should be carried out every six months.The authors have no support or funding to report. The authors are grateful to Patrizia Mangano for the collaboration in the research. The participation of the Health Care Workers is this study is gratefully acknowledged. Caterina Ledda and Venerando Rapisarda conceived and designed the experiments; Francesca Cicciù and Beatrice Puglisi performed the experiments; Caterina Ledda analyzed the data; Tiziana Ramaci contributed analysis tools; Giuseppe Nunnari critically reviewed the manuscript and contributed analysis tools; Caterina Ledda and Venerando Rapisarda wrote the paper.The authors declare no conflict of interest.Sample characteristics.Scores obtained from HCWs.* Statistically significant difference.Logistic regression.* Statistically significant difference.

Med-MDPI/ijerph_2/ijerph-14-03-00285.txt

@@ -0,0 +1 @@
+Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).(1) The study aim was to assess disparities in non-retention in HIV care and non-viral suppression among non-Hispanic Black Caribbean immigrants living with HIV in Florida. (2) We analyzed cases involving individuals, aged ≥13, who met CDC HIV case definition during 2000–2014. Chi square test was used to evaluate differences in non-retention and non-viral suppression by country of origin/race/ethnicity. Multilevel logistic regressions with three referent groups [US-born Blacks, Hispanics, and non-Hispanic Whites (NHWs)] were used to estimate adjusted odds ratios (aOR). (3) Caribbean-born Blacks were less likely to be retained in care or be virally suppressed than US-born Blacks, Hispanics, and NHWs. Bahamians, Haitians, and Trinidadians and Tobagonians had increased odds of non-retention (aOR 3.13, 95% confidence interval [CI] 2.40 –4.10; aOR 1.52, 95% CI 1.40–1.66; aOR 2.30, 95% CI 1.38–3.83), and non-viral suppression (aOR 3.23, 95% CI 2.48–4.21; aOR 1.82, 95% CI 1.68–1.98; aOR 1.76, 95% CI 1.06–2.90) compared with NHWs. (4) Caribbean-born Blacks living with HIV infection are less likely than other racial/ethnic groups to be retained in care and/or achieve viral suppression. Further research is urgently needed to determine social, cultural, and biological factors that contribute to this disparity.The rate of HIV diagnoses in Florida is almost three times the national average [1]. Reduction in HIV incidence is possible if a larger proportion of persons living with HIV are consistently engaged and retained in care [2]. Retention in care is a critical factor for persons living with HIV to achieve viral suppression. Persons living with HIV who are in routine care and achieve undetectable viral loads minimize the chances of transmitting the virus to HIV-uninfected individuals [3].Just over one percent of the US population is composed of Caribbean immigrants [4]. Florida is home to approximately 40.0% of the total Caribbean-born population in the United States [5]. From 2001 to 2007, Caribbean-born immigrants, primarily from Haiti, Jamaica, Trinidad and Tobago, and the Bahamas, comprised approximately 54.1% of the cases of HIV foreign-born Blacks in the United States [6]. Haitian-born immigrants continue to be at highest risk for HIV acquisition, accounting for 16% of all HIV cases in South Florida (Miami-Dade, Broward, and Palm Beach Counties) while representing only 2% of the total Florida population [7].Compared to other populations in Florida, Caribbean immigrants are more likely to be screened later for HIV, resulting in delayed diagnosis [8]. Delayed diagnosis in Caribbean populations suggest other gaps in the HIV care continuum, including insufficient access to and/or engagement in care once diagnosed, leading to impaired disease management and lack of viral suppression, which can also impact primary prevention of HIV in uninfected populations [9].Factors impacting the HIV continuum of care of Black Caribbean immigrants are under-researched. The main aim of the present study is to partially address this gap by assessing the disparity of non-retention in HIV care and non-viral suppression for non-Hispanic Black Caribbean-born immigrants.We used de-identified records, obtained from the Florida Department of Health (DOH) Enhanced HIV/AIDS Reporting System (eHARS), of Florida residents aged 13 years or older who were diagnosed with HIV infection from 2000 to 2014 and met the Centers for Disease Control and Prevention (CDC) case definition for HIV [10]. Data in eHARS are sourced primarily from health care provider reports, laboratory reports, and data extracted from medical records by county health department staff.Retention in care during 2015 was defined as engagement in care two or more times at least three months apart during 2015. We used an expanded definition for retention in care that took advantage of ADAP and Ryan White data to ensure that all medical encounters were captured. The standard definition of retention in care uses only evidence of a laboratory test [11]; however, given the additional information we have linked to our Florida surveillance data, we chose to use a more conservative measure. A person was engaged in care if there was evidence of at least one documented laboratory test, a prescription filled through the AIDS Drug Assistance Program (ADAP) (for those in ADAP), or a physician visit documented in one of the Ryan White databases (for those receiving services through the Ryan White HIV/AIDS program). HIV viral load suppression during 2015 was defined as having a viral load of <200 copies/mL in the last laboratory test performed during 2015. The last viral load test during the measurement year is used by the US Department of Health and Human Services HIV/AIDS Bureau to measure program performance [12].Non-Hispanic Whites (NHW), US-born Blacks, Hispanics, and all foreign-born Blacks from non-Spanish speaking Caribbean countries were included in the analysis to assess disparities affecting Caribbean-born Blacks. Non-Hispanic White individuals living with HIV were used as a referent group because we hypothesized that this group had the greatest contrast with Caribbean–born persons living with HIV, and would therefore highlight disparities affecting Caribbean-born Blacks. Hispanic immigrants, who may have also been from the Caribbean region [13], were used to explore ethnic differences and or/similarities. US-born Blacks were selected as the final referent group to eliminate race as a confounder between Caribbean- and US-born Blacks, allowing for other contextual variables, such as country of origin or culture, to be considered as explanatory variables for the two primary outcomes.Caribbean countries with fewer than 70 HIV cases were grouped together for analysis [Guyana (n = 26), Turks and Caicos (n = 20), Barbados (n = 18), St. Lucia (n = 14), Dominica (n = 10), Antigua and Barbuda (n = 6), Grenada (n = 8), St. Vincent and the Grenadines (n = 7), Bermuda (n = 2), British Virgin Islands (n = 5), St. Kitts and Nevis (n = 4), and Cayman Islands (n = 2)]. Countries that were Departments or Constituents of other countries (Dutch Antilles and French Guiana), were excluded. Records were excluded from the analysis if: HIV diagnosis occurred under 13 years of age (n = 270), missing month and year of HIV diagnosis (n = 79), missing country of birth (n = 907), missing or invalid residential ZIP code (n = 1306), or HIV diagnosis occurred at a correctional facility (n = 3195).Individual-level variables available in the eHARS dataset for cases included current residential ZIP code; month and year of HIV diagnosis and AIDS diagnosis (if applicable); country of birth; age at HIV diagnosis; sex; race/ethnicity; HIV transmission mode; and whether the diagnosis occurred at a correctional facility. AIDS case definition was met if the person’s medical record indicated the development of an AIDS-defining illness, a CD4 lymphocyte count <200 cells/μL, or CD4% of total lymphocytes <14 [14]. People were classified as being born in the United States if they were born in mainland United States, Hawaii, Alaska, Guam, or the US Virgin Islands. Individuals from the US Virgin Islands were included with mainland US instead the Caribbean because of the similarity in terms of access to care. Puerto Ricans were included among the Hispanic group. Individuals with a reported mode of transmission of men who have sex with men (MSM) combined with injection drug use (IDU) were grouped with those who reported mode of transmission as IDU only, because, of the two risk categories, IDU is generally higher risk than MSM.Following a previously used procedure to create a socio-economic status (SES) index (described elsewhere with 2002–2008 American Community Survey estimates [15]), we used 2009–2013 American Community Survey estimates in the present study to create the SES index for our analysis. To categorize current ZIP code tabulation areas (ZCTA) into rural or urban, we used Categorization C of Version 2.0 (University of Washington, Washington, WA, USA) of the Rural-Urban Commuting Area (RUCA) codes, developed by the University of Washington WWAMI Rural Research Center [16].Individual- and neighborhood-level data were merged by matching the current ZIP code with the ZIP code’s corresponding ZCTA. First, we compared individual- and neighborhood-level characteristics by country of birth and ethnicity. We used the Cochran-Mantel-Haenszel general association statistic for individual-level variables controlling for ZCTA, and the chi-square test for neighborhood-level variables. Multi-level (Level 1: individual; Level 2: neighborhood) logistic regression modeling was used to account for correlation among cases living in the same neighborhood. To explore varying disparity, three referent groups were used: US Blacks, Hispanics, and non-Hispanic Whites. Crude and adjusted odds ratios and 95% confidence intervals were calculated comparing cases by country of birth and ethnicity. To identify unique predictors of viral suppression and retention for Caribbean-born Blacks, separate models were estimated stratifying by country of birth. For the stratification, results are only presented for Bahamas, Haiti, and Jamaica—there were no statistically significant results for other Caribbean countries, and the model did not converge for Trinidad and Tobago due to a smaller sample size. Odds ratios were adjusted for year of HIV diagnosis, sex at birth, age, mode of HIV transmission, AIDS diagnosis by 2015, neighborhood SES, and rural/urban status. SAS software, version 9.4 (SAS Institute, Cary, NC, USA) was used to conduct analyses [17]. The study was reviewed and approved by a local Institutional Review Board, and the Florida Department of Health designated this study as non-human subjects research.Demographic characteristics of the study population are presented in Table 1. Haitians and Bahamians had a more even gender distribution among Caribbean countries, while Trinidadians and Tobagonians, Jamaicans, and US-born Blacks had a higher percentage of males living with HIV; Hispanics and NHWs also had the highest percentages of male cases. US-born Blacks, Bahamians, and Hispanics had the most individuals diagnosed in the 13–24 year-old age group (23.2%, 15.9%, and 12.8%), while individuals from Haiti, Jamaica, and Trinidad and Tobago had the most cases 50 years of age and older (24.7%, 21.8%, and 20.0%). Compared to Hispanics and NHWs, the Caribbean groups and US-born Blacks had a lower proportion of people reporting MSM as mode of HIV transmission. Compared to other racial/ethnic groups, NHW cases (25.4%), cases from Trinidad and Tobago (38.6%), and Hispanic cases (42.7%) had a lower proportion of people who had an AIDS defining illness by 2015. More Jamaicans and ‘Other’ Caribbean immigrants reported living in a ZCTA in the highest SES quartile compared to other Caribbean countries (10.4%, 9.4%); and NHWs and Hispanics had a higher percentage of individuals in the two highest SES ZCTA quartiles compared to US-born Blacks (21.6%, 12.9%).The analysis demonstrated that none of the groups is close to attaining the 2020 United Nations goal of 10% or less of HIV infected individuals being non-retained in care or achieving viral suppression [18], and non-Hispanic Caribbean immigrants are far from this goal (Figure 1). Overall, 33.2% of the population was not retained in care, and, 39.3% did not achieve viral suppression (data not presented). Of all the groups in the present analysis, persons living with HIV from Bahamas, Trinidad and Tobago, and Haiti had the highest percentages of non-retention and non-viral suppression (non-retention: 56.6%, 54.3%, 39.2%; non-viral suppression: 63.0%, 51.4%, 47.9%).HIV cases from Jamaica did not differ statistically from any of the reference groups, and Other Caribbean individuals were not different compared to US-born Blacks in non-retention in care (Table 2). For the remaining Caribbean countries, disparities were greater when compared to NHWs, followed by Hispanics, and US-born Blacks. In the fully adjusted models, the rates of non-retention were highest for Bahamas, Trinidad and Tobago, and Haiti, and the disparity was greatest compared to non-Hispanic Whites (retention adjusted odds ratio [aOR] 3.13, 95% confidence interval [CI] 2.40–4.10; viral suppression aOR 2.30, 95% CI 1.38–3.83; 1.52, 95% CI 1.40–1.66).The post-hoc stratification revealed some insight into contributing variables to the results (Table 3). Other or unknown mode of HIV transmission was associated with increased odds of non-retention for all Caribbean countries (Bahamas aOR 4.62, 95% CI 1.78–12.18; Haiti aOR 3.46, 95% CI 2.82–3.72; Jamaica OR 2.18, 95% CI 1.28–3.71.); however, not having AIDS by 2015 was a risk factor for non-retention (aOR 2.80, 95% CI 1.49–5.25; Haiti aOR 3.24, 95% CI 2.82–3.72; Jamaica aOR 2.84, 95% CI 1.95–4.12). For Bahamians, individuals who were diagnosed between 2000–2004 were more likely not to be retained in care (aOR 3.02, 95% CI 1.42–6.44). For Haitian and Jamaican HIV-infected individuals, there were greater odds of non-retention if they were diagnosed in earlier years (Haiti: 2000–2004 aOR 1.90, 95% CI 1.60–2.25; 2005–2009 aOR 1.44, 95% CI 1.20–1.72. Jamaica: 2000–2004 aOR 1.67, 95% CI 1.05–2.65; 2005–2009 aOR 1.60, 95% CI 1.04–2.46), and if they were male (Haiti: 1.46, 95% CI 1.27–1.68. Jamaica: aOR 2.28, 95% CI 1.50–3.48). Additional protective factors for Haitians and Jamaicans were being diagnosed at an older age, and for Haitians only, living in rural areas (aOR 0.21, 95% CI 0.05–0.85).For this clinical outcome, individuals from Bahamas and Haiti consistently presented as being less likely to achieve viral suppression compared to all three referent groups (Table 4). Similar to the non-retention results, the greatest disparity occurred in the comparison to NHWs (aOR 3.23, 95% CI 2.48–4.21; aOR 1.82, 95% CI 1.68–1.98). Jamaicans and Trinidadians and Tobagonians only had statistical differences compared to NHWs, but not to any other racial/ethnic group (aOR 1.29, 95% CI 1.09–1.52; aOR 1.76, 95% CI 1.06–2.90). The one exception was that Jamaicans were more likely to achieve viral suppression compared to US-born Blacks (aOR 0.83, 95% CI 0.70–0.97).The post-hoc stratification for non-viral suppression (Table 5) revealed higher odds for Bahamians, Haitians and Jamaicans diagnosed in earlier years (Bahamas: 2000–2004 aOR 2.26, 95% CI 1.07–4.75; Haiti: 2000–2004 aOR 1.65, 95% CI 1.40–1.93; 2005–2009 aOR 1.42, 95% CI 1.20–1.67. Jamaica: 2000–2004 aOR 1.89, 95% CI 1.26–2.84), or if the individual was a Haitian or Jamaican male (aOR 1.33, 95% CI 1.16–1.52, aOR 2.37, 95% CI 1.60–3.52). Other or unknown mode of HIV transmission was associated with increased non-viral suppression for individuals from Bahamas and Haiti (aOR 7.42, 95% CI 2.25–22.44; aOR 2.51 95% CI 2.07–3.06). Being diagnosed at an older age was protective for Jamaicans (25–49 years of age aOR 0.46, 95% CI 0.27–0.78; ≥ 50 years of age aOR 0.29, 95% CI 0.16–0.56); and it should be noted that Haitians approached significance for the same trend in age at diagnosis. Not having an AIDS diagnosis by 2015 was associated with increased odds of non-viral suppression for all three countries (Bahamas aOR 2.59, 95% CI 1.37–4.89; Haiti aOR 2.17, 95% CI 1.91–2.48; Jamaica aOR 1.96, 95% CI 1.39–2.77). For Haitians only, being MSM or residing in a rural area was protective and lowered the likelihood of non-suppression (aOR 0.76, 95% CI 0.61–0.96; aOR 0.21 95% CI 0.06–0.78).The proportion retained in care or achieving viral suppression was lowest for those in the Bahamas, Haiti, and Trinidad and Tobago. These countries also account for the highest HIV rates in the Caribbean region [18,19]. Bahamians’ odds were least favorable for both outcomes, and Trinidad and Tobago and Haiti were second for non-retention and non-viral suppression, respectively compared to NHWs.As hypothesized, the disparity was greatest between the Caribbean groups and non-Hispanic Whites. Caribbean individuals also did worse than Hispanics for both outcomes. Different immigration policies between the two dominant Caribbean and Hispanic countries in the population may be part of the explanation for the differences between Hispanics and non-Hispanic Caribbean-born Blacks. Compared to US-born Blacks, persons living with HIV from Caribbean countries were also less likely to be retained in care or achieve undetectable viral loads, implying that the disparity may not fully be accounted for by race, but that other cultural and socio-economic considerations should be considered.Hispanics living with HIV in South Florida are predominantly from the Caribbean region, specifically Cuba [20,21]. Among Black persons living with HIV in Florida, the majority are from Haiti [4,22]. During the time of the study, U.S. immigration policies for Cuba and Haiti differed in that Cuban immigrants were provided a level of state protection that facilitated easier access to medical care and other social services [21]. Therefore, despite both immigrant groups having strong ethnic ties with substantial receiving communities in South Florida [23,24], the differing socio-political context may increase HIV risk for Haitians over Cubans, and decrease Haitian individuals living with HIV subsequent engagement and retention in care, and disease management. Although immigration policy may not have a direct impact on individuals in our study who may or may not be immigrants, the existing immigration policies can indirectly influence the differing health culture of Hispanics and Caribbean-born Blacks where Caribbean-born Blacks may be less prone to access services because of perceived obstacles associated with the perception of an unfavorable socio-political climate.Our finding that Bahamians had the worst odds for both outcomes against all referent groups was somewhat unanticipated, as they are considered to be less vulnerable compared to Haitians and other Caribbean groups residing in South Florida [19]. However, this may be attributed to a possible misclassification between Bahamians and Haitians. There is a relationship between Bahamas and Haiti and a history of Haitian immigration to the Bahamian islands [24,25,26]. Based on this historical precedent, it is possible that “Bahamian” cases may have been conflated with Haitian cases, or at minimum, a portion of the Bahamian cases may have Haitian heritage. Given the limitations of the dataset, this possibility could not be explored, but should be studied further as there may be differences in culture between the two countries that can influence health-seeking behavior and decision making.Haitians living in rural areas were more likely to be retained in care and achieve viral suppression. There is evidence of rural areas having more social support [27] and less substance use [28], which are both factors that can improve persons living with HIV’s adherence to HIV medication and treatment [29]. Outside of Haitian immigrants, which were the only Caribbean immigrants with a population large enough to estimate rural/urban effects, it is difficult to postulate the effect of rural/urban status on non-retention and non-viral suppression among other Caribbean countries, as the percentages of non-Haitian Caribbean immigrants residing in rural areas in this population were low (average of 5% or less).Trinidad and Tobago emerged as one of the countries with people living with HIV who were more likely not to be retained in care or not achieve viral suppression. Of note, the proportion of Trinidadians and Tobagonians with MSM as a mode of HIV transmission was higher than for any other Caribbean group (Table 1). The sample size of Trinidadian and Tobagonian immigrants living with HIV (n = 70) did not allow for within group analysis to assess the association between MSM mode of transmission/sexual behavior with the outcomes of interest, indicating another area that warrants further investigation.Finally, although Jamaica has among the highest HIV rates in the Caribbean region [19], in this study, Jamaican immigrants living with HIV had the most favorable outcomes. It was beyond the scope of our analysis to determine what factors may have contributed to these results, but additional investigation into which factors assist Jamaican immigrants living with HIV in Florida to be retained in care and achieve viral suppression may help inform the development of efficacious interventions for immigrants living with HIV from other Caribbean countries.Consistent with other HIV continuum care research, men did worse than women, and being diagnosed at an older age was a protective factor [30,31,32]. “Other” mode of HIV transmission and having an AIDS defining illness had significant associations with the two primary outcomes for all three countries, but the direction of association was different for these two variables. “Other” mode of transmission increased odds of non-retention and non-suppression, and having an AIDS defining illness was protective in this study. Previous research has found that individuals diagnosed with AIDS are more likely to be in care and be retained in care [33], but additional research is needed to determine the mechanism of association for AIDS diagnosis reducing risk of non-viral suppression. More information is also needed about “Other” mode of transmission (i.e., hemophilia, blood transfusion, and perinatal exposure), which may indicate other unknown underlying issues other than what is reported, including possible existing issues with the data capture instrument, or issues related to social stigma when people living with HIV are self-reporting this type of information [33,34]. Over 1300 cases were excluded from the analysis because of missing zip code which may suggest housing instability/insecurity that can also impact retention and viral suppression [35].Although heterosexual mode of transmission accounted for most of the Caribbean cases, homophobia for MSM cases may impact engagement in HIV care. Regardless of sexual orientation, the need for privacy and anonymity because of fear of disclosure status in their communities, and associated stigma and discrimination can also deter routine engagement in care [36].Factors driving non-retention and non-viral suppression are intersectional and complex for persons living with HIV who are also immigrants [37]. Primary data collection and qualitative methods on additional demographic factors and social determinants are needed to examine additional reasons for less retention and viral suppression in non-Hispanic Black Caribbean immigrants. Our analysis could not account for the role of acculturation, age of immigration, and migratory patterns of immigrants who may travel frequently to their countries of birth; nor could it account for second-generation born immigrants, or US-born Blacks who were ethnically non-Latino Caribbean immigrants. Additionally, our data may have had some differential selection for the retention analysis as cases who accessed services from ADAP and Ryan White programs may have had a higher probability of being selected for analysis. White Caribbean-born cases may have identified as non-Hispanic White; however, preliminary analysis indicated that over 95% of HIV cases from the Caribbean in this dataset were Black.Florida continues to be one of the states most affected by HIV, having among the highest rates of new infections and people living with HIV, and Blacks having the highest HIV prevalence in the country [1]. This paper extends our previous findings about delayed diagnosis for Caribbean individuals living with HIV by exploring later stages of the HIV continuum of care and assessing non-retention and non-viral suppression disparity. The expanded analysis indicated that poorer health outcomes and lack of engagement in care for non-Hispanic Black Caribbean persons living with HIV may be related to factors other than race and be more related to cultural differences and socio-political context. HIV interventions that are developed to target this group should be comprehensive and take into consideration other social determinants that are exacerbating the disparity.Research reported in this publication was supported by the National Institute on Minority Health & Health Disparities (NIMHD) under award R01MD004002 and P20MD002288 of the National Institutes of Health. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.All authors were involved in the design of the study and interpretation of findings. Lorene Maddox provided and managed the dataset. Diana M. Sheehan analyzed the secondary data and drafted the manuscript. All authors critically reviewed and revised the final document. All authors read and approved the final manuscript for submission.The authors declare no conflict of interest.Non-retention and non-viral suppression by country of birth for persons living with HIV, 2015.Demographic characteristics of persons living with HIV by race, ethnicity, and country of birth, retained in care, age 13 and older, Florida, 2000–2014, (n = 56,119).Note: ZCTA, ZIP code tabulation area; IDU, injection drug use; MSM, male to male sexual contact; RUCA, rural urban commuting area; SES, socioeconomic status. Percentage may not add up to 100 due to rounding; a Includes Guyana (n = 26), Turks & Caicos (n = 20), Barbados (n = 18), St. Lucia (n = 14), Dominica (n = 10), Antigua and Barbuda (n = 6), Grenada (n = 8), St. Vincent and the Grenadines (n = 7), Bermuda (n = 2), British Virgin Islands (n = 5), St. Kitts and Nevis (n = 4), Cayman Islands (n = 2); b Excludes cases diagnosed under 13 years of age (n = 270), missing month and year of HIV diagnosis (79), missing country of birth (n = 907), missing or invalid residential ZIP code (n = 1306), diagnosed in a correctional facility (n = 3195). c Includes cases reported as both IDU and MSM/IDU.Non-retention in care for Black Caribbean-born immigrants compared to US-born non-Hispanic Blacks, Hispanics, and non-Hispanic Whites in FL in 2015.Note: OR, odds ratio; CI, confidence interval; Adjusted Odds ratios: Controlling for individual-level variables (year of HIV diagnosis, sex at birth, age, race, mode of HIV transmission) and neighborhood-level variables (SES index and rural/urban status); ** p ≤ 0.001.Adjusted odds ratios and 95% confidence intervals for non-retention by selected characteristics, stratified by Caribbean country of birth, 2015.Note: IDU, injection drug use; MSM, male to male sexual contact; SES, socioeconomic status; aOR, adjusted odds ratio; CI, confidence interval; Odds ratios adjusted for individual-level variables (year of HIV diagnosis, sex at birth, age, race, and mode of HIV transmission), and neighborhood-level variables (SES index and rural/urban status); * p ≤ 0.05; ** p ≤ 0.001; a Includes cases reported as both IDU and MSM/IDU.Non- viral suppression for Black Caribbean-born immigrants compared to US-born non-Hispanic Blacks, Hispanics and non-Hispanic Whites in FL in 2015.Note: OR, odds ratio; CI, confidence interval; Adjusted Odds ratios: Controlling for individual-level variables (year of HIV diagnosis, sex at birth, age, race, mode of HIV transmission) and neighborhood-level variables (SES index and rural/urban status); ** p ≤ 0.001.Adjusted odds ratios and 95% confidence intervals for non-suppression by selected characteristics, stratified by Caribbean country of birth, 2015.Note: IDU, injection drug use; MSM, male to male sexual contact; SES, socioeconomic status; aOR, adjusted odds ratio; CI, confidence interval; Odds ratios adjusted for individual-level variables (year of HIV diagnosis, sex at birth, age, race, and mode of HIV transmission), and neighborhood-level variables (SES index and rural/urban status); * p ≤ 0.05; ** p ≤ 0.001.

Med-MDPI/ijerph_2/ijerph-14-03-00286.txt

@@ -0,0 +1,33 @@
+Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).The sudden death of a friend or relative, particularly by suicide, is a risk factor for suicide. People who experience sudden bereavement report feeling highly stigmatised by the loss, potentially influencing access to support. We assessed whether perceived stigma following sudden bereavement is associated with suicidal thoughts and suicide attempt. We analysed cross-sectional survey data on 3387 young adults bereaved by the sudden death of a close contact. We tested the association of high versus low perceived stigma (on the stigma sub-scale of the Grief Experience Questionnaire) with post-bereavement suicidal ideation and suicide attempt, using random effects logistic regression, adjusting for socio-demographic factors, pre-bereavement psychopathology, and mode of sudden bereavement (natural causes/unnatural causes/suicide). Subjects with high perceived stigma scores were significantly more likely to report post-bereavement suicidal thoughts (adjusted odds ratio (AOR) = 2.74; 95% confidence interval (CI) = 1.93–3.89) and suicide attempt (AOR = 2.73; 95% CI = 2.33–3.18) than those with low stigma scores. People who feel highly stigmatised by a sudden bereavement are at increased risk of suicidal thoughts and suicide attempt, even taking into account prior suicidal behaviour. General practitioners, bereavement counsellors, and others who support people bereaved suddenly, should consider inquiring about perceived stigma, mental wellbeing, and suicidal thoughts, and directing them to appropriate sources of support.The search for modifiable risk factors for suicide underpins the suicide prevention research agenda. Sudden bereavement, particularly by suicide [1], is now recognised as a robust risk factor for suicide [2], but explanations for this are unclear. Studies controlling for mental illness indicate that neither heritability [2,3] nor assortative mating [4] completely account for the observed association. An alternative explanation is perceived stigma; the subjective awareness of others’ negative attitudes [5]. This is a common feature of sudden or violent bereavements and may influence access to support [1]. Stigma is also potentially modifiable [6]. Studies comparing grief reactions after different causes of death reveal that experiences of stigma, shame, and concealing the cause are reported after all modes of bereavement, but particularly after violent deaths [1,7] and specifically suicide [8]. Accounting for high levels of perceived stigma has been found to attenuate the association of suicide bereavement with suicide attempt [9], suggesting its role as a mediator of suicide risk. The implication is that anti-stigma interventions might reduce the risk of suicide attempt in people who experience sudden bereavement, perhaps by reducing distress and/or optimising support. The means by which stigma creates barriers to help-seeking have been well-described in relation to mental illness [10], but less well in relation to sudden bereavement [7,11,12,13]. In people with mental illness, stigma is hypothesised to contribute to suicidality through factors such as social isolation, hopelessness, and a perception of being a burden [14]. The same might be theorised after sudden bereavement, when avoidance might arise due to embarrassment, or fear of appearing socially incompetent [15]. Feeling stigmatised by a death contributes to a sense of thwarted belongingness and poor social support; both of which may engender suicidal thoughts [16]. Our objective was to investigate whether high levels of perceived stigma after sudden bereavement are associated with suicidal behaviour. To do this, we analysed British cross-sectional survey data on adults who had experienced sudden bereavement. Our hypothesis was that high stigma scores are associated with post-bereavement suicidal behaviour and depression. To build our understanding of mechanisms, we also hypothesised that high stigma scores would be negatively associated with social support, and receipt of formal or informal support. To understand what differentiates those who attempt suicide from those who consider suicide after sudden bereavement, we hypothesised that high stigma scores are associated with suicide attempt in the sub-group of those with suicidal thoughts following bereavement [17]. Finally, we hypothesised that the effect of high stigma scores on primary outcomes would be modified by gender and by mode of bereavement, such that it would be more pronounced in women and in people bereaved by non-suicide causes.We analysed data from the UCL Bereavement Study [8,9]. This was a UK-wide cross-sectional survey of young adults aged 18–40 working and/or studying at UK higher education institutions (HEIs) who had experienced the sudden bereavement of a close friend or relative. This study had focused on young adults due to concerns about their risk of suicide [18] and the difficulties of engaging young suicidal men with services [19]. Full details of sampling for this closed online survey have been described elsewhere, including the survey instrument (see Supplementary Materials) [8,9]. Sampling via institution-wide email lists (to all staff and students) avoided the biases associated with recruiting a help-seeking sample, and was felt to be the most efficient, comprehensive and pragmatic means of recruiting a hard-to-reach population of young adults [20]. Of 5085 respondents to the survey, we included those who consented to participate, completed a stigma score, and specified their mode of bereavement (n = 3387). The study was approved by the UCL Research Ethics Committee in 2010 (ref: 1975/002). All participants provided online informed consent. Our exposure measure was high perceived stigma of the bereavement, defined using the 10-item stigma subscale of the Grief Experience Questionnaire (GEQ) [21]. The GEQ is a standardised, self-administered instrument for the assessment of the phenomenology of grief. It was originally developed in the U.S. using qualitative data from individuals bereaved by natural causes, accidental death, and suicide [22], and subsequently validated [21]. The stigma sub-scale includes items describing perceptions of others’ avoidance and lack of concern (see Box 1), capturing perceived rather than personal stigma. Responses to items in each subscale are rated using a 5-point Likert-style frequency scale, generating subscale scores of 5 to 25 (at 0.5 intervals). The majority of studies measuring GEQ scores use GEQ subscales rather than overall GEQ scores, allowing them to delineate specific components of grief [8,23,24,25]. Based on precedent [23] and the normal distribution of stigma scores in this sample, we used the mean to dichotomise stigma scores, classifying them as low (5 to 12) or high (12.5 to 25) to aid clinical interpretation.Stem: Since the death how often did you….
+		 
+		 
+        feel like a social outcast?
+		
+		
+        feel like no-one cared to listen to you?
+		
+		
+        feel that neighbours and friends did not offer enough concern?
+		
+		
+        feel avoided by friends?
+		
+		
+        think people were gossiping about you or the person?
+		
+		
+        think that others didn’t want you to talk about the death?
+		
+		
+        feel somehow stigmatised by the death?
+		
+		
+        feel like people were probably wondering about what kind of personal problems you and the person had experienced?
+		
+		
+        think that people were uncomfortable offering their condolences to you?
+		
+		
+        feel like the death somehow reflected negatively on you or your family?
+		
+		feel like a social outcast?feel like no-one cared to listen to you?feel that neighbours and friends did not offer enough concern?feel avoided by friends?think people were gossiping about you or the person?think that others didn’t want you to talk about the death?feel somehow stigmatised by the death?feel like people were probably wondering about what kind of personal problems you and the person had experienced?think that people were uncomfortable offering their condolences to you?feel like the death somehow reflected negatively on you or your family?Our primary outcomes were self-reported suicidal ideation (“Have you ever thought of taking your life, even though you would not actually do it?”) [26] and self-reported suicide attempt (“Have you ever made an attempt to take your life, by taking an overdose of tablets or in some other way?”) [27] post-bereavement. These standardised, validated measures were derived from the Adult Psychiatric Morbidity Survey (APMS) [28], a regular national population survey in England, qualified by whether each was before or after the sudden bereavement, or both, to derive an incident measure. Our three secondary mental health outcomes were post-bereavement non-suicidal self-harm (self-poisoning and self-injury without suicidal intent) using the standardised, validated APMS measure [27] (adapted as above); post-bereavement suicidal and non-suicidal self-harm (aggregating the suicide attempt and non-suicidal self-harm measures, to correspond to that used in a major longitudinal study of self-harm in England [29]); and post-bereavement depression, using the Composite International Diagnostic Interview (CIDI) screen for lifetime depression [30], also validated for use in an online questionnaire [31] (adapted for incident cases as above).Our three self-reported support measures were level of current social support (using a standardised ordinal measure from the APMS [28]); receipt of any formal bereavement support (using a binary measure developed for this study); and receipt of any informal bereavement support (using a binary measure developed for this study). Classification of formal and informal bereavement support was derived from similar British [32] and international [33] studies of service use . Self-help was excluded due to problematic categorisation in relation to formal versus informal bereavement support [34]. Thus, formal support was defined as that received from healthcare or social services staff; psychological therapists or counsellors; voluntary sector helplines or counsellors; police officers; funeral directors; coroners’ officers; teaching staff; school or HEI counselling services; line managers, or employer counselling services. Informal support was defined as that received from friends; family; spiritual/religious advisors, or complementary and alternative medicine practitioners.We selected nine confounding variables on the basis of existing literature and clinical judgement: age; gender; socio-economic status (using the UK Office for National Statistics Standard Occupational Classification [35]); mode of sudden bereavement; kinship to the deceased; family history of suicide (excluding an index bereavement by suicide); pre-loss depression; pre-loss suicidal and non-suicidal self-harm; and years since sudden bereavement. Mode of bereavement was classified via self-report as bereavement by suicide, bereavement by sudden natural causes (e.g., cardiac arrest), and bereavement by sudden unnatural causes (e.g., accidental death). In the case of exposures to more than one mode of sudden bereavement, all those bereaved by suicide were classified as such, regardless of other exposures. Those bereaved by non-suicide death were asked to relate their responses to whichever person they had felt closest to, with exposure status classified accordingly. Missing data for model covariates and outcomes were less than 7%.We investigated simple associations between the outcome variables and exposure using χ2 tests or one-way analysis of variance, as appropriate.We investigated the relationship between outcomes and high stigma scores using multilevel regression models with HEI as random effect, to take into account the clustering effect at the HEI level. We used ordinal logistic regression to investigate the relationship between social support and high levels of perceived stigma scores. All multivariable models included the nine pre-specified confounding variables described above. Models were fitted using complete case analysis. We used the Bonferroni correction to set a significance threshold of p = 0.006 for multiple testing.To test whether the effect of high stigma scores on primary outcomes varied by gender and by mode of bereavement, we added interaction terms to these models, using a less stringent p-value threshold (p = 0.1) to reflect the limited statistical power of interaction tests.To test an additional research question about whether high perceived stigma helps differentiate those who attempt suicide after bereavement from those with suicidal ideation after bereavement, we ran our multivariable model for suicide attempt in the sub-sample of those who reported suicidal thoughts or attempts post-bereavement (n = 1510). We ran a series of a priori defined sensitivity analyses to assess the robustness of our main findings when taking into account biases introduced by <7% missing data and by our sampling strategy. In the first and second analyses, we used best-case and worst-case scenarios to impute missing values by recoding all missing values on outcomes/covariates as positive (e.g., no suicidal ideation/attempt) or as negative (e.g., suicidal ideation/attempt) respectively [36]. In the third and fourth, we used more stringent inclusion criteria: dropping the 10 HEIs that modified the stipulated recruitment method, and the 18 HEIs with participant numbers below the median cluster size. Finally, we conducted linear regression to test whether there was a linear association between stigma scores and outcomes.All analyses were conducted using Stata version 12 (Stata Corp. 2011. Stata Statistical Software: Release 12. College Station, TX, USA).The majority of the sample were female (81%), of white ethnicity (90%), bereaved by sudden natural causes (61%), and reported the death of a relative (71%). The mean time elapsed since bereavement was 5 years (standard deviation (SD) = 5.3 years; range = 1 day to 30 years), with no group differences (Table 1). The age of the deceased varied from 0 (for miscarriage or stillbirth) to 100 years, and median age was significantly younger for those reporting high (median age = 45; inter-quartile range (IQR) = 22–58) versus low stigma scores (median = 50; IQR = 23–70). The group reporting high stigma scores were more likely to be women, students, those in higher social classes, and those educated to a higher level than the group with low stigma scores. They were also significantly more likely to have been bereaved by suicide, and to have had a history of suicidal or non-suicidal self-harm and of depression prior to the loss. Amongst subjects who had made a suicide attempt since the bereavement, those who reported high stigma scores were significantly less likely to have sought help for it than those reporting low stigma scores.Overall, 32% of the sample had received no informal support after the bereavement (Table 2).In an adjusted analysis (Table 3), high stigma scores were associated with a significantly higher probability of post-bereavement suicidal thoughts (AOR = 2.74; 95% CI = 1.93–3.89), suicide attempt (AOR = 2.73; 95% CI = 2.33–3.18), non-suicidal self-harm (2.16; 95% CI = 1.76–2.64), any self-harm (AOR = 2.25; 95% CI = 1.85–2.74), depression (AOR = 3.84; 95% CI = 3.21–4.59).High stigma scores were positively associated with poor social support (AOR = 2.86; 95% CI = 2.44–3.34), and use of formal bereavement support (AOR = 1.87; 95% CI = 1.60–2.19), but negatively associated with use of informal bereavement support (AOR = 0.48; 95% CI = 0.41–0.57).In the sub-sample of n = 1510 individuals who reported suicidal thoughts or attempts post-bereavement, we found a significant association between high stigma scores and post-bereavement suicide attempt (AOR = 1.90; 95% CI = 1.32–2.72; p = 0.001).Gender did not modify the associations between stigma and primary outcomes, but there was an interaction with mode of bereavement, such that the magnitude of the association between high stigma and suicidal ideation was higher for those bereaved by sudden natural death (AOR = 3.08; 95% CI = 2.52–3.76) or sudden unnatural death (AOR = 3.02; 95% CI = 2.13–4.28) than for those bereaved by suicide (AOR = 1.61; 95% CI = 1.09–2.39).The magnitude and direction of adjusted odds ratios for primary outcomes were unchanged in four sensitivity analyses simulating potential biases introduced by missing data and by our sampling strategy. Conducting the analysis using linear regression showed that stigma scores were significantly associated with suicidal ideation (adjusted coefficient = 0.033; 95% CI = 0.029–0.037; p ≤ 0.001) and suicide attempt (adjusted coefficient = 0.009; 95% CI = 0.006–0.107; p ≤ 0.001). On all other measures, it also showed directions of associations consistent with those in our main analysis (p ≤ 0.001 in all cases).The findings of this analysis of British cross-sectional data support our hypothesis that people who feel highly stigmatised by the sudden death of a friend or relative are at increased risk of suicidal thoughts, suicide attempt, non-suicidal self-harm, and depression. The cross-sectional nature of the data limits interpretation of the chronology of the pathways between high stigma scores and outcomes. However, associations with support measures suggested a buffering effect of social support on the negative effects of perceived stigma: those with low perceived stigma scores were more likely to report use of informal support, whereas those with high stigma scores perceived poor social support. Contrary to our hypothesis, use of formal support was more likely in people who felt highly stigmatised, perhaps because they could not rely on friends and family. It is possible that informal support plays an important role in preventing and/or redressing perceived stigma and in mitigating the effects of stigma on suicidality and depression. The results of our interaction tests are interpretable in the context of previous findings from this dataset, namely the increased risk of suicide attempt in people bereaved by suicide [9]. Amongst those bereaved by non-suicide causes, it was those perceiving high levels of stigma who were much more likely to report suicidal thoughts. Previous studies measuring the stigma of sudden bereavement have compared groups defined by cause of death [1,9], but none have explored whether stigma scores per se are associated with adverse outcomes. Instead, qualitative approaches have been used to describe the nature of the stigma experienced by people bereaved traumatically, and how the “death taboo” influences their and others’ avoidance of the topic [11]. Qualitative studies of the stigma perceived by people with mental illness identify the anticipation of negative consequences as a key theme in relation to help-seeking behaviour [10]. The same might apply after bereavement, particularly where the bereaved anticipate social awkwardness [15]. There is some evidence that the stigma attached to help-seeking for mental health problems [10] also applies to bereavement support groups [38], even despite social expectations for the bereaved to engage with support [39] so they can quickly “move on” with their grief [40,41,42].We analysed data from a large, UK-wide sample of 3387 bereaved adults using a validated stigma measure. We tested specific hypotheses formulated on the basis of theory and clinical experience, and our models were adjusted for pre-selected potential confounders. Results were robust to sensitivity analysis simulating potential biases. We acknowledge the potential for male non-response bias, and selection bias of highly educated adults from HEIs. This, and the restricted 18–40 age range, suggest that this study’s findings may only be generalizable to highly-educated young women in the UK. All measures were potentially subject to recall bias. Although validated, our measure of lifetime depression was derived from a brief screening tool [30], and may have over- or under-estimated past depression where used as a potential confounder in multivariable models. Our measures of formal and informal support use were subjective, and represent both preferences and availability. Although the GEQ stigma subscale captures stigmatising aspects of the death specifically, perceived stigma may have been compounded by stigmatising depression or suicidal behaviour. As this was a cross-sectional study, it was not possible to ascertain the temporal sequence of outcomes, including whether suicidal behaviour following bereavement had preceded the awareness of stigma, or of lack of support. However, quantitative [43] and qualitative [44] studies identify perceptions of a lack of support immediately after a sudden death. This study has identified perceived stigma after sudden bereavement as a potentially useful marker for suicidality, depression, and for poor social support. Indeed, among those who had felt suicidal following sudden bereavement, perceptions of high stigma helped differentiate attempters from ideators. Identification of people who feel highly stigmatised after bereavement creates an opportunity to intervene and prevent suicide attempt. General practitioners and bereavement counsellors who encounter bereaved people might consider inquiring about perceived stigma as a way of building a rapport before probing, where appropriate, for low mood and thoughts of self-harm and suicide. Anyone in contact with someone bereaved traumatically has a role in providing information on sources of voluntary sector bereavement support [45,46,47]. Given our findings, this is particularly important for those who feel most stigmatised. Specific resources are available for people bereaved by suicide [46,48], recognising their elevated risk of suicide [2,4] and efforts to target this group in suicide prevention strategies [49,50]. The role of stigma as a putative mediator in the association between sudden bereavement and suicide-related outcomes, and the role of informal support as a moderator of this effect, would need formal testing using longitudinal approaches. If stigma is confirmed as mediating risk of suicidality, the next step would be to develop and trial individual-level or community-level anti-stigma interventions. These might address the barriers to seeking or receiving support, and potentially reduce suicide rates. Cultural dimensions of grief [51] suggest that the development of anti-stigma interventions will need to be based on the findings of qualitative studies. These should explore why bereaved people in different communities feel stigmatised, and how this influences help-seeking behaviour and mental health, but also to understand how informal networks perceive their role and what prevents them from offering adequate support.The results of this study suggest that people who feel the most stigmatised by a sudden bereavement are at greater risk of suicidal behaviour and depression, and are more likely to feel inadequately socially supported by friends and family. High perceived stigma helps differentiate those who attempt suicide after sudden bereavement from those who consider it. Clinicians who inquire about perceived stigma are in a position to identify suicidal distress and address support needs. All those who have contact with people bereaved traumatically should direct them to appropriate sources of support, to overcome any barriers to help-seeking and the effects of perceived stigma. The following are available online at www.mdpi.com/1660-4601/14/3/286/s1.This work was supported by a Medical Research Council Population Health Scientist Fellowship to AP (G0802441). The Medical Research Council, as with other UK Research Councils, provides UCL with funds to cover open access for research papers it has funded. We would like to thank all the HEIs from England, Wales, Northern Ireland, and Scotland that consented to participate in the UCL Bereavement Study, listed below, and all the bereaved individuals who took time to respond to the online survey. Participating HEIs: Bishop Grosseteste University College Lincoln; Bournemouth University; Central School of Speech and Drama; City University; Cranfield University; Courtauld Institute; De Montfort University; University of Greenwich; King’s College London; Liverpool Institute for Performing Arts; Liverpool John Moores University; London Metropolitan University; Norwich University College of the Arts; Royal Veterinary College; School of Oriental and African Studies; St George’s, University of London; Staffordshire University; Trinity Laban Conservatoire of Music and Dance; UCL; University Campus Suffolk; University of Bedfordshire; University of Chester; University of Cumbria; University of Leeds; University of Liverpool; University of Oxford; University of Southampton; University of Worcester; University of Westminster; Queen Margaret University; Heriot-Watt University; Scottish Agricultural College; University of Dundee; Cardiff University; Cardiff Metropolitan University (formerly University of Wales Institute Cardiff); Queen’s University Belfast; University of Ulster.Alexandra Pitman conceived the study with Michael King and David Osborn, and designed the analysis plan in collaboration with Khadija Rantell, Louise Marston, Michael King and David Osborn. Alexandra Pitman conducted the data analysis. All authors interpreted data. Alexandra Pitman wrote the paper, with input from Khadija Rantell, Louise Marston, Michael King and David Osborn. Alexandra Pitman had full access to all the data in the study, takes responsibility for the integrity of the data and the accuracy of the data analysis, and is the guarantor.The authors declare no conflict of interest. The Medical Research Council had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, and in the decision to publish the results.Characteristics of participants by high versus low perceived stigma scores.SD = standard deviation; IQR = inter-quartile range; a using Grief Experience Questionnaire (GEQ) stigma sub-scale score dichotomised at mean into low (5 to 12) and high (12.5 to 25). † pre-specified covariate entered into adjusted models. b significance threshold of p = 0.05; not adjusted for multiple testing. c socio-economic status using the five categories from UK Office for National Statistics. d SAPAS-SR screen for personality disorder [37]. Summary of outcomes by low versus high perceived stigma scores.GEQ = Grief Experience Questionnaire, a using GEQ stigma sub-scale score. a using Grief Experience Questionnaire stigma sub-scale score dichotomised at mean into low (5 to 12) and high (12.5 to 25). b significance threshold of p = 0.05; not adjusted for multiple testing. c measure of social support from Adult Psychiatric Morbidity Survey [28].Estimates of the association between high stigma scores and outcomes.GEQ = Grief Experience Questionnaire. a using Grief Experience Questionnaire stigma sub-scale score dichotomised at mean into low (5 to 12) and high (12.5 to 25). b using corrected significance threshold of p = 0.006. c adjusted for nine pre-specified confounding variables: age; gender; socio-economic status; mode of sudden bereavement; kinship to the deceased; family history of suicide (excluding index bereavement); pre-loss depression; pre-loss suicidal and non-suicidal self-harm; and years since index bereavement. 

Med-MDPI/ijerph_2/ijerph-14-03-00287.txt

@@ -0,0 +1 @@
+Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).Whether parental supply of alcohol affects the likelihood of later adolescent risky drinking remains unclear. We conducted a systematic review and meta-analysis to synthesize findings from longitudinal studies investigating this association. We searched eight electronic databases up to 10 September 2016 for relevant terms and included only original English language peer-reviewed journal articles with a prospective design. Two reviewers independently screened articles, extracted data and assessed risk of bias. Seven articles met inclusion criteria, six of which used analytic methods allowing for meta-analysis. In all seven studies, the follow-up period was ≥12 months and attrition ranged from 3% to 15%. Parental supply of alcohol was associated with subsequent risky drinking (odds ratio = 2.00, 95% confidence interval = 1.72, 2.32); however, there was substantial risk of confounding bias and publication bias. In all studies, measurement of exposure was problematic given the lack of distinction between parental supply of sips of alcohol versus whole drinks. In conclusion, parental supply of alcohol in childhood is associated with an increased likelihood of risky drinking later in adolescence. However, methodological limitations preclude a causal inference. More robust longitudinal studies are needed, with particular attention to distinguishing sips from whole drinks, measurement of likely confounders, and multivariable adjustment.Risky consumption of alcohol is a leading threat to adolescent health globally because of its role in the aetiology of intentional and unintentional injury, mental disorders, and sexually transmitted infection [1,2]. Risky drinking is defined as consumption of ≥5 drinks in a single episode at least monthly. The European School Survey Project on Alcohol and Other Drugs (ESPAD) reported that one in twelve adolescents at the age of 13 or below drank alcohol riskily in 2015 [3]. In the USA, 14% of 12–20 year-olds reported drinking ≥5 drinks on one or more occasions in the previous month, and this age group made 188,706 emergency room visits due to injury and other alcohol-related conditions in 2011 [4]. In addition to illegal sales, sources of alcohol for adolescent (i.e., under the legal age of purchase) drinking include parents, other relatives, and peers [5]. Parents may directly influence their children’s drinking by offering sips of alcoholic drinks at dinner or on special occasions, by supplying alcohol at supervised parties, or by permitting them to take alcohol to drink in unsupervised settings [6]. In Australia and the UK, where drinking per se is not illegal but where purchase is illegal under the age of 18 years, more than a third of adolescents report receiving alcohol from their parents [7,8]. Some research suggests that parents give their children alcohol to teach them how to drink responsibly and to prevent risky drinking with peers [9,10,11].Research regarding the impact of parental supply of alcohol on adolescent risky drinking has produced conflicting results. A 2014 narrative review of the literature found that parental supply of alcohol was associated with heavy episodic drinking and higher risk of alcohol-related harm in 10 studies; but seven studies found it to be protective against such harm [6]. Some studies showed that parental supply was more prevalent in supervised than in unsupervised settings [12,13], with the latter being associated with a higher incidence of risky drinking among 13–17 year-olds [14]. In other studies parental supply was found to be associated with lower risk of hazardous drinking and related problems [15,16]. It is important to note that many of the studies included in the review were cross-sectional, such that the temporal relation between the hypothesized exposure and outcome could not be established. In addition, several studies did not adjust estimates for likely confounding variables (e.g., parent drinking [17,18,19]) so that estimates of association may be biased. Thus, the potential impact of parental supply of alcohol on adolescent risky drinking remains unclear.There have been no reviews synthesizing longitudinal studies to examine associations between prospectively measured parental supply of alcohol and later adolescent risky drinking. We sought to critically examine longitudinal studies with prospective measurement of exposures, and to conduct a meta-analysis to determine whether parental supply of alcohol is associated with later risky drinking.We used the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) [20] guidelines to formulate the basis of pre-specified eligibility criteria using the PICO (P—Populations/People/Patient/Problem, I—Intervention(s), C—Comparison, O—Outcome) worksheet and search strategy (Table 1) [21].Only prospective longitudinal studies (prospective cohort studies and randomized or non-randomized intervention trials) were eligible for inclusion; cross-sectional and retrospective studies being excluded. We included estimates based on assessment of outcome 12 months, or as close to 12 months as possible, after assessment of exposure. Articles analysing parental supply based on adolescent-, parent-, or both adolescent- and parent-report were eligible for inclusion. Only peer-reviewed journal articles published in English were included and there were no exclusion criteria regarding year of publication.Eight electronic databases were searched (Medline, MEDLINE In-Process and Other Non-Indexed Citations, EMBASE, PsycINFO, CINAHL, Scopus, Dissertations and Theses, and Cochrane Library) with the last search carried out on 10 September 2016. We searched for the following terms: parental provision, social hosting, parental source of alcohol, youth, student, teenage, underage, minor, risky drinking, excessive drinking, and binge drinking. We modified and used appropriate mesh terms in the databases with the assistance of the health librarian at the University of Newcastle. Table A1 presents an example of a search strategy performed in EMBASE where 107 articles were found. We screened titles and abstracts using the eligibility and exclusion criteria. Potential eligible articles for data extraction were identified after full-text review. Two reviewers (S.S. and M.K.) independently performed these two stages of screening. Disagreements were resolved by consensus or after consultation with a third reviewer (K.K.). Forward (Google Scholar) and backward searches (bibliographies of included articles) were conducted to find articles that might have been missed during initial database searches. A third reviewer (K.K.) independently reviewed the final included articles to confirm they met the inclusion criteria. The review was registered in PROSPERO [22] on 21 January 2016, prior to the analysis being undertaken (registration number CRD42016032409).Two authors (S.S. and M.K.) extracted information (population, intervention, outcome, study design, statistical methods, and results) using the Cochrane Public Health Group Data Extraction and Assessment Template [23] to tabulate findings of included articles. Finally, they independently assessed risk of bias using the Newcastle–Ottawa Scale for evaluating the quality of nonrandomized studies in meta-analyses [24]. Three factors were considered to score the quality of included studies: (1) selection, including representativeness of the exposed cohort, selection of the non-exposed cohort, ascertainment of exposure, and demonstration that at the start of the study the outcome of interest was not present; (2) comparability, assessed on the basis of study design and analysis, and whether any confounding variables were adjusted for; and (3) outcome, based on the follow-up period and cohort retention, and ascertained by independent blind assessment, record linkage, or self-report. We rated the quality of the studies (good, fair and poor) by awarding stars in each domain following the guidelines of the Newcastle–Ottawa Scale. A “good” quality score required 3 or 4 stars in selection, 1 or 2 stars in comparability, and 2 or 3 stars in outcomes. A “fair” quality score required 2 stars in selection, 1 or 2 stars in comparability, and 2 or 3 stars in outcomes. A “poor” quality score reflected 0 or 1 star(s) in selection, or 0 stars in comparability, or 0 or 1 star(s) in outcomes (Table 1).For articles that reported suitable statistics, a meta-analysis with a random effects model was conducted [25], using the metan command, specifying random, in Stata 13 [26]. There was methodological heterogeneity, studies having applied different measures of exposure and outcome. One study [27] reported results as correlation coefficients. For meta-analysis, we transformed the correlation coefficients into standardized mean differences and then converted them into log odds ratios (logORs) and standard errors (SElogORs). For binary outcomes, ORs and SEs were transformed into logORs and SElogORs. Finally, we pooled logORs and SElogORs of each study to produce summary effect sizes in a forest plot as an OR with 95% confidence intervals (CI). Heterogeneity of findings was assessed using χ2 and I2 statistics [28]. Analyses with p < 0.05 were interpreted as significant. We conducted a sensitivity analysis by examining change in the overall estimate after removing each study in turn, excluding the weaker studies, and excluding those studies that assessed both parent and child report. We assessed publication bias using funnel plots, contour-enhanced funnel plots, and both Begg’s [29] and Egger’s [30] tests.Figure 1 summarises the selection of articles for review. Initial database searches identified 460 records and these were imported into Endnote X7 [31]. From backward and forward searches, three additional articles were identified for further screening. After removing 168 duplicate articles, 284 remained for title and abstract screening. Articles that did not meet inclusion criteria were not carried forward for full-text review, i.e., review articles, conference abstracts, cross-sectional or retrospective studies, studies in which the exposure was not parental supply of alcohol, or where the outcome was not risky drinking. Twenty full-text articles were assessed closely for eligibility, resulting in seven eligible articles from which data were extracted, and results summarised. Of these seven studies, suitable summary statistics were available from six studies for meta-analysis. The remaining one article [32] used analytic methods that do not produce effect estimates that can be converted to ORs.Two studies were conducted in Sweden, two in the USA, one in The Netherlands, one in Australia and one each in the USA and Australia (Table 2). The follow-up period for all studies was ≥12 months, and samples ranged in age from 12 to 15 years at baseline. The age at last follow-up ranged from 14 to 31 years. Sample sizes ranged from 428 to 1945 participants. Parental supply of alcohol was reported by an adolescent, or by both a parent and an adolescent. Most studies were conducted in school settings, and all were published during 2003–2015.Parental supply of alcohol was defined in different ways across studies, including alcohol being supplied at home, direct offers of alcohol by parents to their children in different drinking contexts (home alone, in a party, pub or club, in a park or car), and alcohol consumption at home on weekdays versus weekends. Outcomes included a range of definitions such as heavy episodic drinking (≥5 drinks on a single occasion) [33], problem drinking (as per the Rutgers Alcohol Problem Index [34], or lifetime DSM-IV alcohol abuse and dependence, based on 16- and 14-item scales [34], respectively), drunkenness (fell down or became sick due to alcohol use) [35], risky drinking (≥5 drinks on a single occasion) [36], and alcohol-related harm [27,37].The study results are summarised in Table 2. In all of the studies, parental supply of alcohol was associated with increased risky drinking in mid- or late adolescence. In one study, the association was not significant for boys; however, the point estimate was in the hypothesised direction [33]. Table 3 provides quality scores for the studies, assessing risk of bias. Three studies were of good quality [27,33,36], one was of fair quality [38], and three were of poor quality [32,35,37]. A causal inference is constrained by risk of bias in some studies, the main concerns being measurement of the exposure (a lack of distinction between sips and whole drinks) [27,32,33,36,37], the lack of adjustment for key potential confounders (e.g., parent drinking, and parent rules about alcohol) [27,36,37,38], or a lack of clarity as to whether key confounders had been adjusted for [32,33,35]. Two studies focused explicitly on drinking at home as an exposure. In a study of Australian adolescents (wave 1, mean age 15 years), Degenhardt and colleagues found that drinking at home with family in mid-adolescence was associated with a higher risk of drinking in a range of unsupervised settings, and of becoming a risky drinker in late adolescence [36]. In a Dutch study, Van der Vorst et al. found that drinking at home in early adolescence was associated with problem drinking later in adolescence, the association being similar irrespective of whether the drinking occurred with parents or peers [32]. In a USA cohort, Warner and White [37] found that an onset of drinking before age 11 years in family gatherings was associated with increased risk of “problem drinking” between 3 and 19 years later (OR = 2.86, 95% CI = 1.36, 6.00). Early onset of drinking outside family gatherings was associated with substantially higher risk of later problem drinking (OR = 8.32, 95% CI = 2.28, 30.4) [37].In a comparison of cohorts in the USA state of Washington, and the Australian state of Victoria, alcohol use among 14 year-olds under adult supervision either “at parties” or “at dinner or a special occasion” was found to be associated with higher levels of alcohol-related harm a year later (correlation coefficient = 0.22, p < 0.05) [27]. In a study of USA children, parental supply of alcohol at age 12 years was associated with an increasing trajectory of drunkenness and risky drinking [38]. In a Swedish cohort, parental supply of alcohol at home was associated with increased lifetime prevalence of drunkenness in boys (OR = 1.95, 95% CI = 1.18, 3.20) and girls (OR = 2.76, 95% CI = 1.54, 4.95) compared with adolescents who were not supplied with alcohol [35]. In another Swedish cohort, parental offers of alcohol to 7th graders (aged 13 years) were associated with increased risky drinking in 9th grade (aged 15 years). In adjusted models the association was significant for girls (OR = 1.80, 95% CI = 1.18, 2.75) but not for boys (OR = 1.25, 95% CI = 0.83, 1.89) [33]. Our search revealed a small number of prospective cohort studies from four high income countries with traditionally restrictive approaches to alcohol [39]. It is plausible that the association between parental supply and adolescent risky drinking is different in countries in which drinking small amounts more frequently is the prevailing consumption pattern, e.g., those in southern Europe [40]. Non-exposed groups were selected from the same source population as the exposed group in all studies.The exposures of interest were ascertained from child report in four studies [27,33,36,38] and from reports of both a parent and the child in the other three studies [32,35,37]. If participants generally under-reported parental supply (non-differential misclassification), ORs would be attenuated [41], i.e., the true increase in risk of adolescent risky drinking associated with parental supply would be larger than the estimates suggest. There is evidence to suggest that parents are not a reliable source of information about whether they supply their children with alcohol. In a study involving an anonymous survey of New Zealand school children aged 13–17 years, and a telephone (confidential but not anonymous) survey of their parents, 36% of children reported that their parents had given them alcohol to drink in unsupervised settings in the preceding month, while only 2% of parents reported that they had supplied alcohol to their children for unsupervised drinking in the same period [42]. It is unknown whether such misreporting would be differential or non-differential with respect to the outcome of adolescent risky drinking, such that the likely direction of bias in the estimate of association is indeterminable. This uncertainty about the effects of misclassification of exposure also applies to the problem of counting sips as drinks.The effects on estimates of systematic misreporting of the outcome are also difficult to assess and depend on whether misreporting varies as a function of exposure status [41]. Methodological research suggests that reporting of alcohol consumption is fairly robust in conditions in which respondents have no reason to expect judgement (negative or positive) from researchers, parents, or peers, on the basis of their responses, e.g., where questionnaires are completed anonymously [43].Studies involved parental consent [27,32,33,35,36,37,38,44] and student assent [27,37,38] prior to data collection. In three studies the paper specifically indicated that participants were assured of confidentiality [33,35,38], and in two it was noted that participants were given the opportunity to refuse to participate or answer questions [35,38]. It is unclear what conditions prevailed in the other studies, though it should be noted that in all of the papers it was stated that ethical approval had been received from an institutional review committee.Several studies [27,32,36,37,38] either did not use multivariable analyses to model outcomes [27,32,36] or did not clearly specify what potential confounders were adjusted for [32,35]. Likely confounders include parental drinking, peer and sibling drinking, family income, ethnicity, and religiosity, all of which have been found in prospective cohort studies to be associated with the outcome, namely adolescent risky drinking (e.g., [45]), and are plausibly associated with the exposure (parental supply) [46]. Accordingly, it is likely that effect estimates have been inflated by confounding.Rates of loss-to-follow-up ranged from 3% to 15%, suggesting a low overall potential for attrition bias. The median duration of follow-up was ≥12 months, a period probably long enough for outcomes to occur if parental supply were a causal factor.Of the six studies with data suitable for meta-analysis, two estimated ORs stratified by sex, while the remaining four reported combined ORs, producing a total of eight estimates. Figure 2 presents a forest plot with effect sizes and 95% CIs. All of the ORs were >1, indicating that parental supply of alcohol was associated with twice the odds of later adolescent risky drinking (OR = 2.00, 95% CI = 1.72, 2.32; I2 = 26.4%; p = 0.218). The I2 statistic indicates that the estimates are consistent across the studies. We found the effect estimates from sensitivity analyses were consistent with the effect estimate from the primary analysis (Table A2).One study [32] used analytic methods (path analysis) producing estimates of association that we could not include in the meta-analysis. It found positive associations between parental supply of alcohol and adolescent risky drinking, making it at least broadly consistent with the meta-analytic results.The funnel plot (Figure 3) is asymmetrical, suggesting the possibility of publication bias. A contour-enhanced funnel plot (Figure 4) also suggests the possibility of publication bias as missing studies are mostly in the non-significance area. Begg’s (p = 0.336) and Egger’s test results (p = 0.689) do not confirm this observation; however, they are limited by the small number of studies.The results of this systematic review and meta-analysis suggest the possibility that parental supply of alcohol in childhood increases the odds of later adolescent risky drinking; however, a causal inference is limited by a high likelihood that estimates are inflated by a lack of control for confounders, and a risk of publication bias. If the findings do reflect a true effect, the following aspects of parental supply are implicated: direct supply of alcohol by parents, offers of alcohol by parents, adolescent drinking under parental supervision, adolescent drinking at home, and adolescent drinking in family gatherings.We included only prospective cohort studies providing the basis for establishing that parental supply preceded the outcome of adolescent risky drinking, excluding simple reverse causality as an explanation for the association. However, we cannot exclude more complex competing explanations for the findings. For example, parental supply may initially facilitate moderate drinking in early or mid-adolescence which in turn potentiates heavier drinking and thereby demands on parents or peers to supply alcohol in the larger amounts necessary for risky drinking. Testing such explanations requires assessment at multiple time points and analytic approaches (e.g., marginal structural models [47]) that can model iterative (i.e., time-dependent) processes.Strengths of the review include the comprehensive search strategy, independent screening, study identification and coding, and the risk of bias assessment. The use of meta-analysis increased the precision of the key point estimate, and formal assessment of publication bias has helped to qualify that estimate. Some studies were judged to be high in risk of bias, particular concerns being unreliable measurement of exposure, and lack of adjustment for confounding variables. It is possible that the literature is biased by non-publication of small studies with null findings or findings suggesting that parental supply is protective against adolescent risky drinking.We standardized effect estimates for the purpose of comparison. The transformations we performed (e.g., correlation coefficient to Cohen’s d to lnOR) may have introduced error producing wider confidence intervals for estimates. The variety in definition and measurement of exposures was sufficient to compromise the comparability of studies and it highlights the importance of context in the construct of parental supply. For instance, Warner and White [37] did not define what drinking in a family gathering meant in practice. We assumed it included parents supplying alcohol to their children to drink at family gatherings. Conversely, we assumed drinking outside family gatherings, e.g., with peers, did not involve parental supply, yet qualitative research suggests it is likely that in some situations more complex combinations of parent and peer supply occur [48]. In the study by Danielsson and colleagues [33], we deemed “parental offers of alcohol” as equivalent to “parental supply of alcohol”; however, the paper does not indicate whether adolescents accepted the offers. Contact with the authors confirmed that the questions asked did not permit a judgement to be made about whether the offers resulted in supply or consumption. We reasoned that, in any case, an offer alone may plausibly confer risk by communicating a permissive attitude toward adolescent drinking, as some survey data suggest [49]. Similarly, the exposure “adult supervised drinking” (used in [27]) does not define the relationship of the adult supervisor to the adolescent drinker, such that some instances of what were treated as parental supply may in fact have been supply by other adults.Whether children were allowed to drink whole beverages or merely sip their parents’ alcoholic beverages under supervision was not distinguished in most studies [27,32,33,36,37,44]. In the wider literature on drinking initiation, sipping is often categorized as drinking, yet there is evidence from one prospective cohort study that, in contrast to consuming whole drinks, sipping is not associated with later risky drinking [50]. In a recently published prospective cohort study we found that parental supply of alcohol (of whole drinks, not merely sips) measured when children were around 13 years-old, was not associated with risky drinking (>4 drinks in a single episode in the preceding year) up to three years later, after adjustment for parental drinking, access to alcohol without parents knowing, alcohol-specific rules, monitoring, family factors, family alcohol problems, child factors, and peer factors [51]. Importantly, unadjusted analyses showed a positive association between parental supply of alcohol and risky drinking that disappeared in multivariable models. The study had high retention (>85% three years after baseline), and the cohort is broadly representative of the Australian population of the same age [52], however, it remains possible that evidence of risk associated with parental supply will emerge as members of this cohort enter their late teens, when the prevalence of risky drinking typically increases sharply in Australia.Prospective cohort studies suggest that parental supply of alcohol in childhood increases the likelihood of risky drinking later in adolescence but the potential for bias in this literature is judged to be high. Further longitudinal studies are needed, with particular attention to distinguishing parental supply of sips versus whole drinks, the meaning of supervised drinking, measurement of likely confounders, and adjustment for them in multivariable models. Studies are needed in cultures with traditionally low restrictions on youth drinking (so-called “wet” societies, e.g., in southern Europe [40]), and in low and middle income countries where alcohol consumption is increasing as economies grow rapidly [53].The following are available online at www.mdpi.com/1660-4601/14/3/287/s1, Table S1: PRISMA Checklist.The research was funded by an Australian Research Council Discovery Project (DP106668), an Australian Rotary Health Research Grant, and the Foundation for Alcohol Research & Education. Sonia Sharmin received a Ph.D. scholarship from University of Newcastle, Australia and Australian Rechabite Foundation. Kypros Kypri’s contribution was funded by a National Health & Research Council Senior Fellowship (APP1041867).Sonia Sharmin and Kypros Kypri conceptualized the study and contributed to the interpretation of results and risk of bias assessment. Sonia Sharmin completed initial database searches and with Mausma Khanam independently screened articles, extracted data and assessed risk of bias. Kypros Kypri assessed included articles whether these met inclusion criteria. Sonia Sharmin conducted data analysis and prepared the manuscript. Monika Wadolwski, Mausma Khanam, Kypros Kypri, Richard P. Mattick and Raimondo Bruno critically reviewed the manuscript for important intellectual content and approved the final version as submitted. All authors have agreed to be accountable for all aspects of the work.The authors declare no conflict of interest.Search strategy used for EMBASE.Sensitivity analysis.No.: Number.PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) study flow diagram.Meta-analysis forest plot.Funnel plot of the eight estimates available for meta-analysis. SE: Standard error.Contour-enhanced funnel plot of the eight estimates available for meta-analysis.PICO Worksheet (parental supply of alcohol and adolescent risky drinking).PICO: Population, Intervention, Comparison, Outcome.Study characteristics and results.1 Frequency of drinking six cans of medium-strength beer or four cans of normal beer or four large bottles of strong cider, or a bottle of wine, or half a bottle of spirits on an occasion; 2 Drinking ≥5 drinks in a row; 3 Families (father, mother, and two siblings). CI: Confidence interval; HED: Heavy episodic drinking; OR: Odds ratio; RR: Relative risk.Risk of bias assessment (Newcastle–Ottawa Quality Assessment Scale criteria).Good quality: 3 or 4 stars (★) in selection domain AND 1 or 2 stars in comparability domain AND 2 or 3 stars in outcome domain; Fair quality: 2 stars in selection domain AND 1 or 2 stars in comparability domain AND 2 or 3 stars in outcome/exposure domain; Poor quality: 0 or 1 star in selection domain OR 0 stars in comparability domain OR 0 or 1 stars in outcome/exposure domain.

Med-MDPI/ijerph_2/ijerph-14-03-00288.txt

@@ -0,0 +1,208 @@
+Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).Particulate matter has recently received more attention than other pollutants. PM10 and PM2.5 have been primarily monitored, whereas scientists are focusing their studies on finer granulometric sizes due both to their high number concentration and their high penetration efficiency into the respiratory system. The purpose of this study is to investigate the population exposure to UltraFine Particles (UFP, submicrons in general) in outdoor environments. The particle number doses deposited into the respiratory system have been compared between healthy individuals and persons affected by Chronic Obstructive Pulmonary Disease (COPD). Measurements were performed by means of Dust Track and Nanoscan analyzers. Forty minute walking trails through areas with different traffic densities in downtown Rome have been considered. Furthermore, particle respiratory doses have been estimated for persons waiting at a bus stop, near a traffic light, or along a high-traffic road, as currently occurs in a big city. Large differences have been observed between workdays and weekdays: on workdays, UFP number concentrations are much higher due to the strong contribution of vehicular exhausts. COPD-affected individuals receive greater doses than healthy individuals due to their higher respiratory rate.Outdoor air pollution is an important health threat for humans in both developed and developing countries. The World Health Organization (WHO) estimated that in 2012 ambient air pollution caused three million premature deaths worldwide. These deaths have been mainly ascribed to exposure to particulate matter (PM) of 10 microns or less in diameter (PM10) [1]. The latest scientific research on the association between PM exposure and negative human health outcomes has been focused on fine (PM2.5) and UltraFine Particles (UFPs < 100 nm). UFPs play a significant role in PM-induced adverse effects, both for the pulmonary system and other sites such as the cardiovascular and nervous systems [2,3,4,5]. Furthermore, these findings are supported by the inverse relationship occurring between particle size and their toxicity, i.e., particles with smaller granulometric sizes are the most dangerous for human health [6]. In addition, UFPs exhibit some adverse properties such as high surface area to mass ratio, ultrahigh reactivity, and smaller size than the dimensions of cellular structures. These characteristics allow UFPs to easily adsorb organic molecules and reach cellular targets of the pulmonary system and other systems [7,8,9]. In addition, particles with a diameter greater than 2.5 µm are efficiently removed from the atmosphere though dry and wet deposition processes, whereas particles with a diameter less than 1 µm (PM1) persist for a longer period, contaminating outdoor and indoor air [10], and can be transported long-range.Despite all this evidence, UFPs have not yet received the proper attention from environmental policy makers or consideration in air quality regulations. Thus, an important issue in this field is a better understanding of UFP sources, composition, and size distribution, as well as their routes of potential exposure and their impact on adverse outcomes [5]. Regarding the UFP sources, it has been demonstrated that the major contributors to UFP air pollution are anthropogenic emissions, such as combustion engines and power plants [5]. In particular, scientific literature evidences that, in urban areas, UFPs mostly derive from diesel and automobile exhaust [11,12,13]. Thus, it is fundamental to monitor UFP traffic-emissions in order to assess human exposure and to evaluate the related health risks. Difficulties related to this kind of monitoring are the very quick evolution of UFP concentrations and size distribution when they are released into the atmosphere. These phenomena can be influenced by several parameters, such as the meteorological conditions (e.g., wind speed, wind direction, and height of the boundary layer), the levels of some precursor gases (i.e., SO2 and NOx) and other pre-existing particles [14]. Furthermore, exposure levels to anthropogenic emissions can greatly fluctuate, even at short distances, and are strongly related to personal activities. Consequently, UFP exposure assessment should overcome the limitations of the traditional analyzers installed in fixed monitoring stations [11].Within this context, the aim of the present study is to estimate the potential UFP exposure of people living in typical urban environments. We performed an intensive monitoring campaign in the urban areas of a big city (Rome, Italy), characterized by different levels of traffic density during the winter season in severe atmospheric conditions (atmospheric stability conditions, high pressure for long periods, pollutant accumulation, traffic restrictions, etc.). Along with PM measurements, we investigated UFP number concentrations and size distributions and evaluated individual exposure. These data were used to assess the doses of particles deposited into the respiratory system for each investigated urban scenario.This paper represents the first attempt to describe the doses deposited in the respiratory tract of people commuting in a large city. Moreno et al. [15] compared atmospheric contaminants inhaled by travelers during bus, subway train, tram, and walking journeys through the city of Barcelona; this paper was focused on the chemical aerosol speciation, carried out in the 10–300 nm particle range. Many other studies were based on the chemical speciation and the relevant influences on the distribution of transport-related air pollutants in urban air [16,17,18,19,20,21,22,23,24,25,26,27,28], whereas no papers deal with the dose deposited in the human respiratory tract.Aerosol measurements have been carried out for some typical scenarios currently encountered in urban environments. Measurements have been performed along an urban path in Rome, including areas with different vehicular traffic densities. Furthermore, the ultrafine particle exposure of pedestrians waiting for a bus or standing close to a traffic light along a high traffic density road, waiting to cross the road, was evaluated.A 40-min walking path was considered, as, on average, this would be the time required to reach a workplace in Rome. The selected path (Figure 1) proceeds from Piazza dei Navigatori to Ostiense train station across a distance of 2.5 km and an estimated time travel of 40–50 min.The entire sampling campaign was carried out in January–February 2016 for 30 workdays and 6 holidays. During this period, the considered path was followed both on workdays and holidays. The data reported in the present study are representative of a typical daily trend in such a period.Three different areas can be identified on the selected path (Figure 1): the first one, labeled “1”, is along “Via Cristoforo Colombo” and is characterized by high vehicular traffic density (cars, buses, trucks, motorcycles); the second one, labeled “2”, runs along “Circonvallazione Ostiense Street” and is at medium traffic density; and the last one, labeled “3”, “Piazza 12 Ottobre 1492”, is a low vehicular traffic density area.By using a camera, the number of vehicles circulating per unit of time was evaluated; on average, about 200 vehicles (e.g., cars, buses, and motorcycles) per minute passed in zone 1, an average of 44 vehicles per minute passed in zone 2, and 2 vehicles per minute passed in zone 3 on workdays and 10 cars on holidays.It is important to point out that, during the sampling period, the weather in Rome was unfavorable to the pollutant dispersion. High barometric pressure characterized the entire campaign period and induced the Rome City Mayor to limit the car circulation on workdays and to prohibit it on holidays. The data reported as holidays refer to Sundays, when car traffic was completely halted, except emergency vehicles, buses, and taxis.Aerosol measurements were performed to evaluate the exposure of pedestrians waiting at a bus stop and at a traffic light close to a high traffic density road. In the first case, an average waiting time of about 20 min was considered. In the second case, 50 s were considered on workdays and 40 s on holidays (the duration of the red traffic light). Both operations were carried out on workdays and on holidays.Throughout the campaigns, a baseline measurement was performed in order to evaluate indirect vehicular traffic influence; the measurements were carried out in Villa Ada Park, a park considered the background area of Rome in terms of air pollution and which is quite far from direct vehicular traffic emissions.Different PM size fractions have been investigated along with the temporal trend of particle number concentration in the 11–365 nm size range. A backpack was equipped with a portable DustTrack system and a Nanoscan instrument. The DustTrakTM II Aerosol Monitor 8532 (TSI, Shoreview, MN, USA), a handheld battery-operated, data-logging, light-scattering laser photometer, simultaneously measures both mass and size fractions (PM1, PM2.5, PM4, or respirable fraction, PM10, and Total PM size fractions). It uses a sheath air system that isolates the aerosol in the optics chamber to keep the optics clean for improved reliability and low maintenance. A NanoScan SMPS 3910 (TSI), which adopts a scanning mobility particle sizing technology, was also used to measure particle number concentrations in the range 10–365 nm [29]. Aerosol concentrations were measured with 60 s time resolutions in thirteen size channels (i.e., 11.5 nm, 15.4 nm, 20.5 nm, 27.4 nm, 36.5 nm, 48.7 nm, 64.9 nm, 86.6 nm, 115.5 nm, 154.0 nm, 205.4 nm, 273.8 nm, and 365.2 nm). The attention was focused on ultrafine particles, i.e., on size channels from 11.5 nm to 115.5 nm (and the first three modes, which are important in the accumulation mode).Two probes (anti-electrostatic tubes) were placed on the jacket lapel to get detailed information on personal exposures, on opposite sides in order to avoid interferences and to prevent air vortices that could affect the measurement.The methodology of particle dose calculation has been thoroughly described in previous papers [30,31,32,33,34,35,36], where the particle deposition in the human respiratory system was evaluated using the Multiple-Path Particle Dosimetry model (MPPD v2.1, ARA 2009, ARA, Arlington, VA, USA). This model calculates, in the respiratory tract, the deposition and clearance of mono- and poly-disperse aerosols in the range from ultrafine to coarse particles in the respiratory tract [37,38]. A stochastic lung model was considered because it provides more realistic lung geometry than the symmetric one considered in the International Commission on Radiological Protection (ICRP) model [39]. In the MPPD model, the ten stochastic lungs are ordered according to size (total number of airways) from the smallest to the largest, and the approximate size percentile of each lung is provided [40]; the 60th percentile human stochastic lung has been considered in this study. The following settings were considered in the MPPD model: (i) an uniform expansion of the lung; (ii) an upright body orientation; and (iii) nasal breathing with an inspiratory fraction and a no pause fraction. Moreover, the following parameters were used for a Caucasian adult male under the sitting-awake condition based on the ICRP report [39]: (i) a functional residual capacity (FRC) of 3300 mL; (ii) an upper respiratory tract (URT) volume equal to 50 mL; (iii) a 10.7-min−1 and 13.1-min−1 breathing frequency for healthy and COPD (Chronic Obstructive Pulmonary Disease) patients, respectively [41]; and (iv) an air volume inhaled during a single breath (tidal volume, Vt) of 1.25 L [42].The total doses of particles deposited in the human respiratory tract vs. time has been estimated for healthy individuals (DHealthy) and individuals affected by Chronic Obstructive Pulmonary Disease (COPD) [43] according to the following equations:
+        
+          
+            
+              
+                
+                  
+                    D
+                    
+                      H
+                      e
+                      a
+                      l
+                      t
+                      h
+                      y
+                    
+                  
+                  
+                    (
+                    t
+                    )
+                  
+                  =
+                  
+                      ∑
+                    
+                      i
+                      =
+                      1
+                    
+                    
+                      13
+                    
+                  
+                  
+                    D
+                    i
+                    
+                      H
+                      e
+                      a
+                      l
+                      t
+                      h
+                      y
+                    
+                  
+                  
+                    (
+                    t
+                    )
+                  
+                
+				 
+				with
+				 
+                
+                  
+                    D
+                    i
+                    
+                      H
+                      e
+                      a
+                      l
+                      t
+                      h
+                      y
+                    
+                  
+                  
+                    (
+                    t
+                    )
+                  
+                  =
+                  
+                    F
+                    i
+                    
+                      H
+                      e
+                      a
+                      l
+                      t
+                      h
+                      y
+                    
+                  
+                  ×
+                  
+                    C
+                    i
+                  
+                  
+                    (
+                    t
+                    )
+                  
+                  ×
+                  
+                    V
+                    t
+                  
+                
+              
+            
+          
+        
+        
+          
+            
+			 
+              
+                
+                  D
+                  
+                    C
+                    O
+                    P
+                    D
+                  
+                
+                
+                  (
+                  t
+                  )
+                
+                =
+                
+                    ∑
+                  
+                    i
+                    =
+                    1
+                  
+                  
+                    13
+                  
+                
+                
+                  D
+                  i
+                  
+                    C
+                    O
+                    P
+                    D
+                  
+                
+                
+                  (
+                  t
+                  )
+                
+              
+			   
+			  with
+			   
+              
+                
+                  D
+                  i
+                  
+                    C
+                    O
+                    P
+                    D
+                  
+                
+                
+                  (
+                  t
+                  )
+                
+                =
+                
+                  F
+                  i
+                  
+                    C
+                    O
+                    P
+                    D
+                  
+                
+                ×
+                
+                  C
+                  i
+                
+                
+                  (
+                  t
+                  )
+                
+                ×
+                
+                  V
+                  t
+                
+              
+			  
+            
+          
+        
+       where FiHealthy and FiCOPD are the fractions of the particle deposited doses in the respiratory tract of healthy people and individuals affected by COPD, respectively; Ci(t) is the particle number concentration measured by NanoScan; and Vt is the tidal volume, i.e., the volume of air involved in each respiratory act. The FiHealthy and FiCOPD deposition fractions were derived from Löndahl et al. [41]; the authors measured experimentally the deposition fractions of aerosol emitted from diesel engines on a group of eight healthy volunteers and 10 volunteers suffering from COPD. A Vt of 0.86 L has been assumed for both cases, whereas a respiratory frequency of 10.7 min−1 and 13.1 min−1 has been assumed for healthy people and COPD-affected people, respectively.The methodology reported in the Experimental section was applied to the overall campaign. First, the different fractions of aerosol were estimated; Table 1 shows the data on PM concentrations measured during the workdays.The comparison between the fraction levels evidenced that PM2.5 represented 95% of PM10 and that 94% of PM10 was made of PM1.A similar situation occurred for the measurements performed during holidays (Table 2), wherein PM1 (average 32 µg·m−3) represented 91% of the PM10 (average 35 µg·m−3).Table 3 shows the Pearson’s correlations among particulate matter at different granulometric sizes. Correlation coefficients were higher for fractions from PM1 to PM10; they ranged from 0.999 for PM2.5–PM1 to above 0.9 for other fractions below PM10. These PM fractions were so highly correlated because, in the investigated periods, PM pollution was mainly due to fine aerosol deriving from anthropogenic sources. Correlation coefficients between total PM and the other PM fractions were lower (0.66–0.74), very likely because of the resuspension of road dust due to the nearby vehicular traffic. The PM composition of Rome’s urban air, extensively studied by Perrino et al. [44] and Avino et al. [45], was characterized by a crustal contribution that ranged between 7%–30%, whereas the remaining amount was due to anthropogenic sources. This last fraction was formed by 30%–40% of carbonaceous particulate matter [46] and 20%–35% of ammonium sulphate and nitrate. Minor percentage components, but ones that were really important for their public health relevance, included Polyciclic Aromatic Hydrocarbons (PAHs), nitro-PAHs, oxo-PAHs [47], and heavy metals [48]. Specifically, PM1 was the most significant component of PM10 and, therefore, it is important to analyze its size distribution. To this aim, a NanoScan instrument was used. Figure 2 shows the total concentration values in two typical days, a workday and a holiday, along the considered urban path; the profiles represent the average of all the measurements performed during the entire campaign.The first relevant finding is that during workdays the submicron particle concentration reached a value of about 120,000 # cm−3 and it never dropped below 25,000 # cm−3. However, when the vehicular traffic was completely halted (no vehicles passing-by due to red traffic light), data were very different; the submicron particle concentrations were considerably lower (about 23,000 # cm−3), and the minimum result was approximately 5000 # cm−3. As reported in Figure 2, the highest concentrations were reached along Via Cristoforo Colombo (10 car lanes, track width 25 m), a high traffic density road during workdays.The area tagged “Park area” (Figure 2) is located close to a high traffic density road; the peak value detected was due to truck emissions because of work in progress. Although truck emissions inside a park are not a currently encountered situation, it is not so rare when the park is close to an avenue with 10 lanes and frequent maintenance works. Therefore, this data set can be considered as representative of a real exposure scenario in a densely populated area of Rome. In the area tagged “High traffic area” (along Via Cristoforo Colombo), the highest values were due to the continuous transit of many vehicles during the “green” traffic light, while the lowest level occurred when the traffic light was red. Moving on the path, the particle concentration levels drop, due to the presence of trees, which act as a natural barrier to the particles and allow a reduction of the exposure level. The last peak in the “High traffic area” was due to vehicles waiting to cross the road at the traffic light. In the area tagged as “Medium traffic area”, particle concentrations were almost constant and were lower than in the previous zone. The same occurred in the “Low traffic area” zone, where the only (very low) peak was recorded near a market with several vehicles nearby. During holidays, the submicron particle concentration values were very low (below 20,000 # cm−3) and almost constant over the time.Table 4 summarizes the aerosol size distribution data measured in both workdays and holidays. The Coefficient of Variation (CV %) gives an index of the variability of the measurements.During workdays, the particle concentrations for size channels from 11.5 nm to 48.7 nm showed high CVs (up to 100% for the first three channels) due to the fast temporal evolution of such particles, whereas the CVs were below 50% for size fractions above 48.7 nm. During holidays, the trend was quite similar, although with lower values. All these data confirm that the different particle behavior strictly depends on the granulometric size fractions; nucleation particles are quickly generated by autovehicular exhausts, and their concentration rapidly drops due to diffusion, coagulation, and deposition mechanisms.UFP percentage contribution was about 75% and confirms previous measurements, performed in a street canyon in downtown Rome, in which their percentage ranged between 70%–95% [49,50]; studies performed in other urban environments, in cities other than Rome, reported a percentage contribution of particles below 100 nm of about 80% of the total particle number concentration [51,52,53,54,55].The Pearson’s correlation coefficients among different size fractions (Table 5) evidence the common origin of ultrafine particles. For instance, in Table 5, coefficient correlations above 0.7 up to 0.98, calculated for particles ranging from 36 nm to 86 nm, denote the contribution of diesel engine emissions. This occurrence is confirmed by data shown in Figure 3, in which the size distribution for a diesel engine bus (“waiting for” scenario) is reported.The temporal concentration levels of each NanoScan size channel are reported in Figure 4. Peak concentrations are due to nucleation particles (NanoScan size channel from 11.5 nm to 48.7 nm) that were continuously released by the intense autovehicular emissions. In contrast, the greater contribution in the low traffic zone was due to larger particles because nucleation particles were rapidly dispersed and removed through coagulation and deposition mechanisms. These features have an important influence on the doses deposited into the human respiratory tract. This issue will be discussed in the Section 3.2, both for healthy people and patients affected by COPD.People waiting at bus stops is a common scenario in urban areas. In such a simulated scenario, PM10 (overall mean value 154 µg·m−3) was composed of 94% PM1 (overall mean value 144 µg·m−3) (data not shown), and the exposure to submicron particles is very high on workdays due to long bus-waiting times and high particle concentration. Total particle concentrations between 50,000 and 100,000 # cm−3 and peak concentrations as high 415,000 # cm−3 were measured on bus arrivals.Figure 3 shows the aerosol number size distributions representative of the “waiting” and “bus coming” scenarios. In the first case, a trimodal size distribution was measured (with modes at about 15.4 nm, 27.4 nm, and 115.5 nm). In the second case, a single mode of 36.5 nm was measured. In particular, the “bus coming” scenario is characterized by a monomodal distribution, centered at about 36.5 nm, coherent with the measurements performed by Kittelson for diesel engine exhaust size distribution [56]. The differences between the two size distributions reported in Figure 3 are due to the different bus engine loads [57] occurring in the “waiting” and “bus coming” scenarios.Similar trimodal and monomodal distributions were measured upon waiting at traffic lights. Bimodal size distributions (modes at 15.4 nm and 115 nm) were also measured, depending on the engine emissions and loads. Total particle number concentrations as high as 100,000 # cm−3 were measured for traffic lights in red status [58] due to the simultaneous presence of a high number of vehicles waiting in the queue, whereas, during the green traffic light, particle levels below 40,000 # cm−3 were measured.Table 6 summarizes the main findings in terms of particle doses for healthy and COPD-affected people, respectively, along with the dose increment % estimated for COPD-affected individuals.Figure 5 shows the cumulative doses of particles deposited in the human respiratory system during the city path (workdays) by comparing healthy persons and patients affected by COPD, whereas Figure 6 shows the particle doses deposited in the respiratory system of healthy and COPD-affected persons in 1 min time intervals (the Nanoscan scan time) as functions of time and of particle diameter. People with COPD disease received higher particle doses than healthy persons, due to their increased respiratory rate, i.e., the difference is about 7.3% (city path). Similar trends were found for holiday dosimetry estimates; in this case, the difference between the two profiles was 7.7%, whereas the particle number deposited was about five orders of magnitude less that calculated for the previous simulation (Table 6).For the “bus waiting” scenario, the cumulative dose for COPD-affected individuals was about 5.2% and 8.4% greater than for healthy individuals during workdays and holidays, respectively. Figure 7 shows the particle doses deposited into the respiratory system of healthy and COPD-affected persons in 1 min time intervals as functions of time and of particle diameter for the “bus waiting” scenario. In comparison with the city walk path, peak doses were about 3-fold higher and in the same range from 20.5 nm to 36.5 nm as for the city walk path. However, the peak doses for the “bus waiting” scenario were about threefold higher for healthy individuals than for COPD-affected individuals.Finally, simulations of individuals waiting at a traffic light were also performed. Figure 8 shows the cumulative particle doses deposited into the human respiratory system of healthy and of COPD-affected individuals waiting at a traffic light on a typical workday. As observed in the previous scenario, cumulative particle doses were 7.71% higher for persons suffering from COPD than for healthy persons during workdays and >8.54% during holidays.It should be considered that waiting for the “green” traffic light for ten minutes in most cases is an unlikely situation; nonetheless, in some other cases, such as a police officer standing outside near a traffic light for much more than ten minutes, the simulation is certainly much more significant.This paper deals with an important potential threat to human health, ultrafine particle exposure and the relevant dose deposited into the respiratory tract of pedestrians in big urban areas such as Rome, considering both healthy people and more susceptible individuals, such as those affected by COPD. UFPs are becoming an issue investigated by several authors, and data sets are being collected in different urban areas. However, health effects have been mainly associated with particulate matter (i.e., PM10 and PM2.5) [59], whereas studies that correlate UFPs and human health in outdoor environments are scarce. Moreover, health effects are not directly linked to particle concentration, but they are associated with the particle doses deposited into the respiratory system. To date, dosimetry data are not so widely reported in literature for urban environments. To the authors’ knowledge, this is the first paper that investigates the deposition dose for healthy and COPD-affected individuals in real urban scenarios. This kind of evaluation is highly relevant because, in 2013, COPD affected 329 million people or nearly five percent of the global population (males and females with the same incidence) [60,61] with consequently very high social costs (i.e., $2.1 trillion in 2010 [62]). The present study shows that individuals with COPD receive higher particle doses (+7%–9%) compared to healthy people. The measurements performed in downtown Rome display great differences between workdays and holidays, with the former reaching submicron particle values definitely higher than the latter and thus highlighting the strong contribution of autovehicular traffic to ultrafine particle generation.This research was supported by INAIL grants from 2016–2018.P.A. and M.M. are responsible for the research design. C.N. and P.A. conducted the field work. The paper was written by P.A. with significant contributions by M.M., C.P., and M.V. All the authors approved the paper.The authors declare no conflict of interest.The path from Piazza dei Navigatori to Ostiense train station. Zone 1 = high vehicular traffic density area; zone 2 = medium vehicular traffic density area; zone 3 = low vehicular traffic density area; and zone 4 = Park.Typical trend of the submicron particle concentration values (# cm−3) sampled along an urban path during workdays (blue line) and holidays (red line).Comparison of the aerosol size distribution profiles in the “waiting for” (blue line) and “bus coming” (red line) scenarios.Typical trends of different size fractions during workdays (a) and holidays (b).Typical profile of the cumulative doses of particles deposited in the human respiratory system of healthy persons and patients affected by Chronic Obstructive Pulmonary Disease (COPD) along a city path on workdays.Particle number doses deposited in 1 min time intervals in the human respiratory system of healthy (a) and COPD-affected (b) persons as functions of particle sizes and time along a city path in workdays.Particle number doses deposited in 1 min time intervals in the human respiratory system of healthy (a) and COPD-affected (b) persons as functions of particle sizes and time for the bus stop waiting scenario on workdays.Cumulative particle doses deposited in the human respiratory system of healthy people and persons affected by COPDs; the simulation was performed while the pedestrian waited for green light at a traffic light on holidays.Particulate matter (PM) data (as µg·m−3) measured using DustTrack during the simulation of the city path during workdays. Variability is calculated as Coefficient of Variation (CV %).PM data (as µg·m−3) measured using DustTrack during the simulation of the city path during holidays (variability calculated as CV %).Pearson’s correlation between different PM sizes during workdays (a) and holidays (b).Particle measurements (# cm−3) performed by Nanoscan during the city path on workdays (a) and on holidays (b) (variability calculated as CV %).Pearson’s correlation coefficient among the submicron particles in different size fractions on workdays (a) and on holidays (b). Values > 0.7 are reported in italic.Total doses of aerosol deposited in the human respiratory system, relevant UltraFine Particle (UFP) % contribution for the healthy and COPD-affected individuals, and % dose increments estimated for COPD-affected individuals in comparison with healthy individuals (Δ %), both on workdays and holidays.

Med-MDPI/ijerph_2/ijerph-14-03-00289.txt

@@ -0,0 +1 @@
+Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).Particulate matter (PM) such as ultrafine particulate matter (UFP) and the organic compound pollutants such as polycyclic aromatic hydrocarbon (PAH) are widespread in the environment. UFP and PAH are present in the air, and their presence may enhance their individual adverse effects on human health. However, the mechanism and effect of their combined interactions on human cells are not well understood. We investigated the combined toxicity of silica nanoparticles (SiNPs) (UFP) and Benzo[a]pyrene (B[a]P) (PAH) on human endothelial cells. Human umbilical vascular endothelial cells (HUVECs) were exposed to SiNPs or B[a]P, or a combination of SiNPs and B[a]P. The toxicity was investigated by assessing cellular oxidative stress, DNA damage, cell cycle arrest, and apoptosis. Our results show that SiNPs were able to induce reactive oxygen species generation (ROS). B[a]P, when acting alone, had no toxicity effect. However, a co-exposure of SiNPs and B[a]P synergistically induced DNA damage, oxidative stress, cell cycle arrest at the G2/M check point, and apoptosis. The co-exposure induced G2/M arrest through the upregulation of Chk1 and downregulation of Cdc25C, cyclin B1. The co-exposure also upregulated bax, caspase-3, and caspase-9, the proapoptic proteins, while down-regulating bcl-2, which is an antiapoptotic protein. These results show that interactions between SiNPs and B[a]P synergistically potentiated toxicological effects on HUVECs. This information should help further our understanding of the combined toxicity of PAH and UFP.Air pollution has been linked to increasing cardiovascular and pulmonary diseases disease morbidity and mortality [1,2,3,4]. It has been shown to escalate the progression of atherosclerosis, and induce plaque stability, thrombogenesis, vascular dysfunction, and autonomic imbalance [5]. These are linked to the development of cardiovascular disease. Air pollutants consist of particulate matter (PM), organic compounds, gases, and toxic metals [6,7]. The PM is categorized into PM10 (diameter less than 10 μm), PM2.5 (diameter less than 2.5 μm), and ultrafine particles (UFPs) (diameter less than 0.1 μm) [8]. The UFPs are the most harzardous pollutants of the PMs owing to their size, which allows deeper inhalation and penetration into the body’s organs (citation). Organic compound pollutants are numerous, but polycyclic aromatic hydrocarbons (PAH), which are produced during the incomplete combustion of fossil fuels, are the most ubiquitous in ambient air [9]. They occur as gases or are bound to PM. Ultrafine particles (UFP) such as SiNPs are widely spread in the atmosphere during in-dust periods [10] and construction-related industries [11]. In addition, their increased production and application in the industrial, commercial, and biomedical fields has also increased their environmental presence [12,13]. Moreover, PAH such as B[a]P easily binds to UFPs [9]. Currently, B[a]P-bounded PM is receiving special attention [9], because of the possibility of bounded UFPs, such as SiNPs, penetrating the lung and reaching other organs via the circulatory system [14]. It has been demonstrated that SiNPs and B[a]P are capable of entering the human body via inhalation, ingestion, and dermal contact, and can distribute themselves in nearly all of the organs through the blood stream [9,15,16]. Upon entering the bloodstream, SiNPs and B[a]P have direct contact with blood vessels and the heart endothelial cells lining, where they may induce toxicity. Endothelial cells are the biological barriers which mediate the clearance of nanoparticles, and maintain vascular function and homeostasis [17].SiNPs and B[a]P have been shown to individually induce endothelial cell toxicity, resulting in endothelial dysfunction [18,19,20,21]. Endothelial dysfunction has been associated with multiple cardiovascular events that cause vascular wall damage, atherosclerotic plaque, and consequently, it promotes vascular injury [22]. However, these studies only investigated single chemical exposure toxicity, while in the environment, these chemicals exist as a mixture and their synergistic effects may be underestimated. This is because these chemicals, when in a mixture, can interact additively or synergistically to exert a larger effect than that which has been predicted [23]. Chemicals may act in the same way, resulting in a dose (additive interaction) or effect addition, or they may enhance the toxicity of one another (synergy interation) [23]. Therefore, it is imperative to assess the effects and mechanism of UFPs and PAH co-exposure on endothelial cells. In order to determine these effects and mechanism, we assessed B[a]P and SiNP co-exposure toxicity to human endothelial cells. A human umbilical vein endothelial cells (HUVECs) line is often used for in vitro studies of endothelial cell function [24]. B[a]P and SiNPs interactions were investigated by assessing cellular oxidative stress, DNA damage, cell cycle arrest, and apoptosis. The Stöber technique, as illustrated by [25], was used to make SiNPs. Approximately 4 mL of Ammonia, 2 mL of water, 50 mL of ethanol, and 2.5 mL of tetraethylorthosilicate (TEOS) were mixed. The mixture was constantly stirred at 150 rpm for 12 h at 40 °C and centrifuged for 15 min at 12,000 rpm. Thereafter, it was washed three times with deionized water, and then dispersed in 50 mL of deionized water. A stock solution of 50 mM BaP concentration was prepared by dissolving 0.05111 g B[a]P in DMSO, resulting in a total solution of 4.05 mL.The size and distribution of the SiNPs were determined by a Transmission Electron Microscope (TEM) (JEOL, Tokyo, Japan) and imageJ software. Dynamic light scattering (DLS) was applied for measuring the hydrodynamic size and zeta potential of SiNPs in different solution media, using a zeta electric potential granulometer (Malvern, Worcestershire, UK). SiNPs were first sonicated for 5 min before measurements were taken.HUVECs were obtained from Shanghai Institutes for Biological Sciences, Shanghai, China. Cells were cultured in DMEM (HyClone, Pittsburgh, PA, USA), mixed with 10% fetal bovine serum (Gibco, Pittsburgh, PA, USA), and incubated at 37 °C and 5% CO2. After 24 h, the cells were then exposed to SiNPs and/or B[a]P for another 24 h. To minimize the aggregation of SiNPs, they were sonicated at 160 W, 20 kHz, for 5 min, before being added to the culture medium. The treatment groups where designed as SiNPs (10 μg/mL), B[a]P (1 mM), and SiNPs + B[a]P (10 μg/mL + 1 mM), with 1% DMSO culture medium and pure culture medium as control groups.The viability of the HUVECs was detected by a Cell Counting Kit-(CCK-) 8 (KeyGEN, Nanjing, China). Briefly, cells with 1 × 104 cells per well were adhered to the bottom of 96-well plates for 24 h, followed by SiNPs and/or B[a]P. After 24 h incubation, the equivalent amount of CCK-8 reagent was put in each well and measured by a microplate reader at 492 nm (Thermo Multiskan MK3, Pittsburgh, PA, USA). To evaluate the cytotoxicity of the SiNPs and B[a]P mixture, we first tested the cell viability at 2.5, 5, 10, and 20 μg/mL concentration of SiNPs, and 0.25, 0.5, 1, and 2 μM concentration of B[a]P, and then selected a No Observed Effect Concentration (NOEC) for the assessment of the mixture’s toxicity.For the screening of the intracellular ROS level, flow cytometry was used with an oxidation-sensitive probe, the 2′, 7′-dichlorofluorescein diacetate (DCFH-DA) (JianCheng, Nanjing, China). First, the cells were treated with SiNPs and/or B[a]P for 24 h, and were then double washed with PBS and incubated for a further 30 min at 37 °C in the dark with serum-free DMEM medium containing 10 μM DCFH-DA. Consequently, we collected the cells and washed them using PBS, before determining the fluorescent intensities and percentage of positive cells at 488 nm excitation, 525 nm emission, using a flow cytometer (Becton Dickison, Franklin Lakes, NJ, USA). Oxidative damage is determined by the malondialdehyde (MDA) content, and the superoxide dismutase (SOD) and glutathione peroxidase (GSH-px) activities. After 24 h exposure to SiNPs and/or B[a]P, the cells were washed with ice-cold PBS, and lysed in ice-cold RIPA (DingGuo, Beijing, China) lysis buffer for 30 min. The lysates were centrifuged at 12,000 rpm for 10 min, and supernatants were collected to measure the MDA content, and SOD and GSH-px activities. Commercially available kits (Jiancheng Bioeng Inst., Nanjing, China) were used for the measurements. The protein concentration was determined using a bicinchoninic acid (BCA) protein assay (Pierce, Rockford, IL, USA).To quantify the DNA damage, a single cell gel electrophoresis (SCGE) kit (Research Biolab, Beijing, China) was used. After 24 h of cell treatment with SiNPs and/or B[a]P, 1 × 106 cells were collected and washed with PBS, and about 10 μL of cell suspension was combined with 90 μL agarose and moved to agarose-coated slides, concealed with cover slips, and cooled at 4 °C for 4 min. The slides were placed in fresh lysis solution for 2 h at 4 °C in the dark, and were then electrophoresed at 25 V, 300 mA, 1 V/cm for 30 min, for DNA unwinding. Propidium iodide (PI) was used to stain the slides, which were then detected by a fluorescence microscope (Olympus, Tokyo, Japan). The CASP software was used to compute the DNA damage rate, tail DNA percentage, tail length, and Olive Tail Moment (OTM).The cell cycle distribution was determined by a cell cycle detection kit (KeyGen, Nanjing, China). HUVECs (1.0 × 106/well) were plated and treated in 6-well plates (three wells per group). Cells were treated by SiNPs and/or B[a]P for 24 h, and were then fixed in ice-cold 70% ethanol at 4 °C overnight. After that, the cells were incubated at 37 °C for 30 min with 100 μL Rnase A and 400 μL PI, respectively. Finally, the samples were examined by a flow cytometer (FC500, Beckman Coulter, Brea, CA, USA).Apoptosis in HUVECs was detected by an annexin V and propidium iodide (PI) assay kit (KeyGen, Nanjing, China). After being treated with SiNPs and/or B[a]P for 24 h, the HUVECs were immersed in 500 μL binding buffer and stained with 5 μL Annexin V-FITC for 15 min, before being treated with 5 μL PI at room temperature. The cells were later loaded on a flow cytometer (Millipore, Billerica, MA, USA), and data from 10,000 cells/sample were analyzed at 488 nm. To analyze whether combined SiNPs and B[a]P exposure influences the expression of the G2/M DNA damage checkpoint and apoptosis regulators, we measured the protein levels of Chk1, Cdc25C, cyclin B1/Cdc2, Bcl-2, Bax, Caspase 9, and Caspase 3 in HUVECs, using western blot analysis. The cells were lysed through the RIPA buffer on ice. Equal amounts of protein lysates were loaded onto SDS-polyacrylamide gel electrophoresis, to be separated, and were then transferred to polyvinylidene fluoride membranes (PVDF) (Millipore, Billerica, MA, USA). 5% skim milk mixed with Tris-buffered saline (TBS) was used to block the PVDF membranes for 1 h. Then, the PVDF membranes were cultured with the appropriate primary antibodies (Cell Signaling Technology, Beverly, MA, USA) at 4 °C for the whole night. The PVDF membrane was washed with TBST three times and was incubated with fluorescent secondary antibodies (Cell Signaling Technology, Beverly, MA, USA) for 1 h in the dark. After having been rinsed with TBST, blotted proteins were detected and imaged through the Odyssey Infrared Imaging System (LI-COR Biosciences, Lincon, NE, USA). Data were analyzed using the Image J software (National Institutes of Health, Bethesda, MD, USA).One-way analysis of variance (ANOVA) was used to determine the differences between the treatment groups. While two factorial analysis of variance (ANOVA) was used to determine SiNPs and B[a]P interactions, an F value greater than a F(4,10) = 3.4780 critical value indicates that there is a deference between the treatment groups. Their marginal means were compared using profile plots (interaction plots), as described by Ennos [26] and Yu, et al. [27]. The interaction plots of estimated marginal means for visual exploration of interactions involving combinations of between-subjects and/or within-subjects factors, was constructed using the General Linear Model (GLM) command in SPSS. The results were displayed graphically using the plots where parallel lines indicated an additive effect, while nonparallel lines indicated a synergy interaction effect. The statistical significance was considered at a p value < 0.05. All of the experiments were performed in triplicate and expressed as a mean ± SD. SPSS. (Version 16.0, SPSS Inc., Chicago, IL, USA) software was used to perform statistical analyses. SiNPs were near-spherical and well isolated, as shown in Figure 1A, while their sizes were normally distributed, with an average diameter of 62.88 ± 10.16 nm (Figure 1B). The hydrodynamic sizes of SiNPs were measured in distilled water, DMEM medium, 1% DMSO DMEM, and 10% serum DMEM exposure media. There was an inverse relationship between the hydrodynamic sizes and zeta potential of SiNPs, as shown in Table 1. The data shows an increase in the hydrodynamic size with a decrease in the zeta potential in different media.The cell viability was reduced with an increasing dose of both SiNPs and B[a]P, as shown in Figure 2A,B. A NOEC for both SiNPs and B[a]P was chosen to investigate their combined exposure effect and interaction. A total of 10 μg/mL of SiNPs and 1 μM of B[a]P were chosen. The cell viability of SiNPs, B[a]P, and their mixture exposure groups, was 95.4%, 86.7%, and 73.5%, respectively (Figure 2C). The data demonstrated a synergy interaction between SiNPs and B[a]P (F = 6.476, p = 0.021), which was further proved by the profile plots (Figure 2D). The data on morphological changes observed in treated cells are presented in Figure 3A–E. The morphological changes showed that SiNPs + B[a]P treated cells had a reduced cell density and irregular cell shapes, as shown in Figure 3E. There was a significant increase in the intracellular ROS levels of the SiNPs and B[a]P + SiNPs treated groups, as shown in Figure 4A. The data indicate that the co-exposure of SiO2NPs and B[a]P could generate more intracellular ROS than individual exposure. The factorial analysis provides evidence of a synergy interaction between SiNPs and B[a]P (F = 7.301, p = 0.027, Figure 4B).The data on MDA content, and SOD and GSH-px activities, and the level of cells exposed to SiNPs and/or B[a]P, are presented in Figure 5A,C,E, respectively. The results show an increased MDA content and decreased SOD and GSH-px activities in cells exposed to SiNPs and/or B[a]P. The co-exposure of SiNPs and B[a]P co- significantly increased the MDA content, while reducing SOD and GSH-px activities, when compared to the control or other treated groups. This suggests that co-exposure enhances oxidative damage to a greater extent than individual chemical exposure. Factorial analysis further demonstrates the synergistic interaction in the increase of MDA content (F = 5.084, p = 0.026, Figure 5B) and decrease of GSH-px activity (F = 11.174, p = 0.006, Figure 5F), with an additive effect of decreasing SOD activity (F = 3.506, p = 0.143, Figure 5D).The DNA damage results of different treated groups are presented in Figure 6A–E. A significant difference is exhibited by the co-exposed group, when compared to the other groups. Factorial analysis confirms that the co-exposure of SiNPs and B[a]P induces DNA damage in a synergistic manner (F = 28.392, p < 0.001, Figure 6G). The majority of cells in the SiNPs and B[a]P co-exposed group were arrested in the G2/M checkpoint of the cell cycle (Figure 7). This suggests that the co-exposure of SiNPs and B[a]P induced cell cycle arrest at the G2/M checkpoint. A summary of the percentage of cells in the G0/G1 phase, S phase, and G2/M phase, is represented in Figure 7F. Factorial analysis shows that there was a synergistic interaction between SiNPs and B[a]P cell cycle arrest during induction (F = 27.637, p = 0.001, Figure 7G). Western blot results show a significant increase in Chk1 protein expression and decrease in Cdc25C, cyclin B1, and Cdc2 proteins expression, in the B[a]P + SiNPs group, when compared to the other groups (Figure 7H,I).Apoptosis was significantly induced in cells treated with SiNPs. Moreover, it was significantly higher in the combined exposure group than the single exposure group (Figure 8F). This demonstrates that co-exposure enhanced the apoptotic rate in HUVECs, when compared to individual chemical compounds. Factorial analysis shows a synergistic interaction between SiNPs and B[a]P in inducing the apoptosis of HUVECs (F = 23.838, p = 0.001, Figure 8G).Several studies have linked ambient PM with the increased morbidity and mortality of cardiovascular and pulmonary diseases [2,3,4]. Of special interest are the UFPs, which are more hazardous and can lead to worse health effects [28,29]. This is due to the complexity of their biological effects, which not only depend on their individual response, but also on the interaction with other pollutants. UFP, along with PAH, can enhance the resulting effects. Therefore, it is important to identify the combined biological effects of UFP and the attached toxic PAH, since they are widely spread in the atmosphere during in-dust periods [10] and constructions [11]. In addition, their increased application in the industrial, commercial, and biomedical fields, is also increasing their environmental presence [12,13]. On the other hand, PAH produced during the incomplete combustion of fossil fuels is the most ubiquitous in ambient air [9].The SiNPs used in this study were small in size and spherical in shape (Figure 1A,B). This may have facilitated their cellular uptake [30]. Moreover, their smaller sizes produced a larger surface area for B[a]P adsorption, thus increasing the B[a]P cytoplasm bioavailability. The present study suggests that the NOEC of SiNPs and B[a]P, when co-exposed to human endothelial cells, induces oxidative damage, resulting in DNA damage, cell cycle arrest at the G2/M checkpoint, and the apoptosis of HUVECs. A synergy interaction between SiNPs and B[a]P was involved in enhancing their toxicity. To gain a greater understanding of the mechanism involved in SiNPs and B[a]P co-exposure-induced biological effects, such as those affecting cellular morphology, cell viability was measured as cytotoxicity indicators in HUVECs. We assessed the morphology of HUVECs exposed to SiNPs and/or B[a]P for 24 h by optical microscopy (Figure 3). A cell density reduction, irregular shape, and cellular shrinkage were observed. Cellular morphology changes have always been chosen as a pointer in determining the cytotoxicity [31]. To confirm and analyze this observation, the cell viability was measured (Figure 2C). Our data revealed that the co-exposure of SiNPs and B[a]P induced cytotoxicity in a synergistic manner. To confirm oxidative DNA damage, we carried out intercellular ROS generation and oxidative damage analyses. Both SiNPs and B[a]P have been demonstrated to induce DNA damage through ROS generation [21,32,33,34]. The results show that co-exposure to SiNPs and B[a]P synergistically induced intracellular ROS production, lipid peroxidation, and a decrease in GSH-px activities in cells. Moreover, the decrease in SOD activities was due to the additive effect of B[a]P and SiNPs. SOD specifically catalyzes the dismutation of the superoxide anion to O2 and H2O2, and the disturbed SOD activity may increase the level of ROS, thus leading to redox imbalance, which is an initial result of ROS-induced DNA damage [35]. Intracellular GSH has been known to act as a major nonenzymatic antioxidant or free radical scavenger that protects cells against oxidative stress [35]. The inverse relationship between the ROS level and the GSH activities observed in this study indicates that the free radical species generated by the synergy interaction between SiNPs and B[a]P, reduces the intracellular antioxidant level. Moreover, free radicals also result in the production of malondialdehyde, an indication of lipid peroxidation. Our data demonstrate that the generation of intracellular ROS caused oxidative damage, followed by the production of lipid peroxidation and the inhibition of antioxidant activities. In the present study, our results showed that the degree of DNA damage was significantly high during the co-exposure of B[a]P and SiNPs (Figure 6). Furthermore, the results show that the co-exposure of B[a]P and SiNPs inhibited HUVECs proliferation by inducing G2/M cell cycle arrest (Figure 7). The cell response to DNA damage always involves multiple repair mechanisms and checkpoint responses that can interrupt cell cycle progression or alter DNA replication [36]. Cells usually initiate cell cycle checkpoints in order to detect and repair damaged DNA, for the purpose of maintaining genome stability [37]. In the case of incomplete damaged DNA repair, the checkpoints will arrest cell cycle at either the G0/G1, S, or G2/M phase. Cells with DNA damage are prevented from entering mitosis (M phase) at the G2/M checkpoint [38]. This delay provides extra time for DNA damage repair [39]. However, when the DNA damage is so severe that it exceeds the cellular repair capacity, apoptosis occurs. In the current study, we confirmed that the co-exposure of SiNPs and B[a]P elicited cell cycle arrest at the G2/M checkpoint. Our data also show that Cdc25C, Cdc2, and cyclin B1 were notably suppressed in HUVECs, after exposure to B[a]P + SiNPs for 24 h, while Chk1 was significantly increased. Checkpoint kinase 1 (Chk1), a vital kinase for conserving genome stability, is always triggered in response to DNA damage. It plays an important role in the cell cycle checkpoint control, DNA damage repair, and DNA damage-induced apoptosis [40,41]. Chk1 is primarily involved in the G2/M checkpoint signal transduction pathway [42,43]. Chk1 is triggered to inhibit the activation of Cdc25C, consequently resulting in the downregulation of cyclinB1 [44]. Cdc2 and cyclin B1 are necessary for the cells to enter into the mitotic phase. Cdc2 always attaches to cyclin B1 during the the G2/M transition. The blockage of the cyclin B1/Cdc2 complex results in G2/M cell cycle arrest [45]. This study demonstrate that the co-exposure of B[a]P and SiNPs induced a toxic effect of endothelial cells through triggering the Chk1-dependent G2/M DNA damage checkpoint signaling pathway.Our data show that apoptosis was significantly induced in cells treated with combined SiNPs and B[a]P (Figure 8). The up-regulation of pro-apoptotic proteins (bax, Caspase-9m and caspase-3) and downregulation of anti-apoptotic protein bcl-2, confirms that the co-exposure of SiNPs and B[a]P induced apoptosis in HUVECs. p53 has been postulated to up-regulate Bax [46]. Besides, the entry of bax into the mitochondrial membrane may instigate p53-mediated apoptosis [46]. Caspases are triggered through proteolytic cascades, which can amplify an initially small amount of caspase activity to levels which are sufficient for the initiation of apoptosis [47]. Bax, Caspase-9, and Caspase-3 are involved in the intrinsic apoptosis pathway [47]. Hence, this study demonstrates that the intrinsic apoptosis pathway is involved in SiNPs and B[a]P co-exposure-induced apoptosis.In summary, the co-exposure of UFPs and PAH may induce some unanticipated toxicity, even beyond the well known toxicities of the individual compounds. Our results reveal that the co-exposure of SiNPs and B[a]P caused excessive oxidative stress, leading to DNA damage, cell cycle arrest, and apoptosis. The Chk1-dependent G2/M DNA damage checkpoint signaling pathway and intrinsic apoptosis pathways were involved in inducing cell cycle arrest and apoptosis, respectively. The toxicity was potentiated by synergistic interactions between SiNPs and B[a]P. This study provides evidence of the interactions between UFPs and PAH, causing cardiovascular toxicity. This work was supported by the National Natural Science Foundation of China (no. 81230065, 81502830, 81571130090), Beijing Natural Science Foundation Program and Scientific Research Key Program of Beijing Municipal Commission of Education (KZ201410025022), and Scientific Research Common Program of Beijing Municipal Commission of Education (KM201610025006).C.O.A. developed the concept, carried out the research, and wrote the article. J.W., H.H., L.F., and X.Y. contributed to the revision of the article. J.D. and Z.S. contributed to the concept development, quality assurance, and revision of the article. The authors declare no conflict of interest.Characterization of SNPs. (A) TEM images show spherical SNPs with good monodispersity in distilled water; (B) Size distribution of SNPs showing a normal distribution curve, mean = 62.720 ± 10.917.Effects of SiNPs and/or B[a]P on HUVECs’ viability. (A) Cell viability of various concentrations of SiNPs; (B) Cell viability of various concentrations of B[a]P; (C) Cell viability of HUVECs treated with DMSO (1%), SiNPs (10 μg/mL), B[a]P (1 μM), and their mixture (10 μg/mL + 1 μM); (D) Profile plot shows a synergy interaction between SiNPs and B[a]P (F = 6.476, p = 0.021). * p < 0.05, ** p < 0.01 for treated group compared to control, while # p < 0.05 for combined groups compared to single treated groups.Morphological changes in HUVECs observed under an electron microscope after 24 h of exposure to B[a]P and/ SiNPs. (A) Control group; (B) HUVECs exposed to DMSO (0.1%); (C) HUVECs exposed to B[a]P (1 μM); (D) HUVECs exposed to SiNP (10 μg/mL); (E) HUVECs exposed to B[a]P + SiNPs (10 μg/mL + 1 μM).Intracellular ROS generated by treated HUVECs. (A) ROS level; (B) Interaction plots showing a synergy interaction between SiNPs and B[a]P (F = 7.301, p = 0.027). * p < 0.05, ** p < 0.01 for the treated group compared to the control, while # p < 0.05 for combined groups compared to single treated groups.HUVEC oxidative stress caused by SiNPs and B[a]P co-exposure. (A) Malondialdehyde content increased; (B) Profile plots shows that SiNPs and B[a]P synergistically increased the malondialdehyde content (F = 5.084, p = 0.026); (C) Decreased superoxide dismutase activity; (D) Profile plots shows that the superoxide dismutase activity decrease was additive (F = 3.506, p = 0.143); (E) Decrease in glutathione peroxidase activity; (F) Profile plot shows a synergy interaction in the decrease of glutathione peroxidase activity (F = 11.174, p = 0.006). ** p < 0.01 for the treated group compared to the control, while # p < 0.05 for the combined group compared to single treated groups.DNA damage in HUVECs induced by SiNPs and/or B[a]P exposure. (A–E) show the representative fluorescence images of PI-stained nuclei of the control and cells treated with DMSO, B[a]P, SiNPs, and B[a]P + SiNPs, respectively; (F) Shows the DNA damage rate, tail DNA percentage (%), tail length, and OTM; (G) Interaction plot illustrates the synergistic effect of SiNPs and B[a]P on the DNA damage of HUVECs. * p < 0.05, ** p < 0.01 for the treated group compared to the control, while # p < 0.05 for the combined group compared to single treated groups.Cell cycle phase distribution in various treatment groups. (A) Control; (B) DMSO (0.1%); (C) B[a]P (1 μM); (D) SiNPs (10 μg/mL); (E) B[a]P + SiNPs (1 μM + 10 μg/mL) treatment; (F) Percentages are mean ± SD of each cell cycle phase for triplicate experiments; (G) Factorial analysis plots show a synergy interaction between SiNPs and B[a]P (F = 27.637, p = 0.001). ** p < 0.01 for the treated group compared to the control, while # p < 0.05 for the combined group compared to single treated groups; (H) Western blot results show a significant increase in Chk1 expression and decrease in Cdc25C, Cyclin B1, and Cdc2 expression; (I) the graph shows a significant change in the protein expression of the co-exposed group compared to the other groups.Apoptosis of HUVECs induced by SiNPs and/or B[a]P. (A) Control; (B) DMSO; (C) B[a]P; (D) SiNPs; (E) B[a]P + SiNPs; (F) The percentage of apoptotic cells; (G) Synergistic interaction between SiNPs and B[a]P illustrated by interaction plots (F = 23.838, p = 0.001. * p < 0.05, ** p < 0.01 for the treated group compared to the control, while # p < 0.05 for the combined group compared to single treated groups; (H) Western blot results show a decrease in Bcl-2 expression and increase in Bax, Caspase 9, and Caspase 3 expression in the co-exposed group; (I) the graph shows a significant change in the protein expression of the co-exposed group compared to the other groups.Hydrodynamic size and Zeta potential of silica nanoparticles in dispersion media.

Med-MDPI/ijerph_2/ijerph-14-03-00290.txt

@@ -0,0 +1 @@
+Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).Data from the Australian Longitudinal Study on Women’s Health were used to examine how work was associated with time spent sleeping, sitting and in physical activity (PA), in working women. Young (31–36 years; 2009) and mid-aged (59–64 years; 2010) women reported sleep (categorised as shorter ≤6 h/day and longer ≥8 h/day) and sitting time (work, transport, television, non-work computer, and other; summed for total sitting time) on the most recent work and non-work day; and moderate and vigorous PA (categorised as meeting/not meeting guidelines) in the previous week. Participants reported occupation (manager/professional; clerical/sales; trades/transport/labourer), work hours (part-time; full-time) and work pattern (shift/night; not shift/night). The odds of shorter sleep on work days was higher in both cohorts for women who worked shift or night hours. Longer sitting time on work days, made up primarily of sitting for work, was found for managers/professionals, clerical/sales and full-time workers. In the young cohort, clerical/sales workers and in the mid-aged cohort, full-time workers were less likely to meet PA guidelines. These results suggest multiple behaviour interventions tailored to work patterns and occupational category may be useful to improve the sleep, sitting and activity of working women.Given that most working adults spend around a third of their day at work, the occupational environment is likely to have a bearing on time spent sitting, sleeping and being active, and on subsequent health risks. Long periods spent sitting, sleep duration outside the recommended range (e.g., 7–8 h) and a lack of moderate to vigorous physical activity (MVPA) are associated with increased cardio-vascular disease risk, incident type 2 diabetes and premature mortality [1,2,3,4,5]. The duration and timing (night/day) of paid work, as well as the demands of work in terms of seated or active tasks, have been associated with overall daily sitting, activity and sleep, and with transport and leisure-time physical activity. Hence relationships between these three behaviours (sitting, sleep, physical activity) are complex and likely to be intertwined with patterns and duration of work.Several small studies have examined associations between occupation, work hours and sitting time. One Australian study has shown that call centre workers spend significantly more time in sedentary behaviour while at work, than office and customer services employees [6]. Other studies have shown that managers and professionals report more occupational sitting time than technicians and trade workers, but that leisure sitting time is similar in these employee groups [7,8]. Hours of paid work may also be influential, as one study has shown that women who work part-time sit less during the whole day than those who work full-time [9]. There is also some evidence that more time spent sitting at work is associated with more time sitting at home, as shown by a study that found female University employees who sat most at work, also sat most for television viewing, on work and non-workdays [10]. As most studies to date have involved small samples in selected workplaces, larger studies are required to clarify the relationships between work hours and sitting time in different occupational groups.A recent review has examined associations between occupation and physical activity (PA). It found strong associations of leisure time physical activity with occupation type, with higher leisure-time MVPA in professional than in blue collar occupations in most studies, although findings were mixed. In the same review there was a negative association between hours worked and leisure time MVPA [11]. The authors stressed that confounding factors such as hours of work, work demands, and work-related physical activity were often not accounted for [11]. The issue of whether leisure-time PA is influenced or compensated by occupational activity is unclear. In a study of office, call centre and customer service employees in Australia, all employees engaged in more sedentary behaviour and less light intensity activity on work days than non-work days [6] but MVPA was higher on work days than non-work days [6]. Conversely, another study showed that MVPA in office workers was very similar on work and non-work days [12]. The amount of leisure-time MVPA may also be influenced by the physical demands of work; however, findings have been equivocal, with high workplace PA associated with low [7] or high levels of non-occupational PA [11], while other studies showed no difference [13]. These mixed results suggest that the relationship between occupation and non-work activity is currently unclear and should be investigated. There is also evidence to suggest that occupational factors may impact sleep quality and duration. Data from the 2009 U.S. Behavioural Risk Factor Surveillance System showed that adults who worked more than 40 h per week were 65% more likely to report insufficient sleep (<7 h per day) than those who worked less than 40 h per week [14] and that those working in the manufacturing sector were more likely than people in the general working population to sleep less than 6 h per night [14]. Working night shifts has also been associated with short sleep duration, as one study found that 44% of night shift workers report sleeping less than 6 h per day compared with 30% of day shift workers [14]. Thus, sleep duration may, not surprisingly, differ according to level and type of employment.While it is apparent that patterns of sitting, sleeping and being active vary markedly across groups in different work settings, these patterns may be even more complex in women. This is because women are more likely to juggle paid and unpaid work, Australian Bureau of Statistics data show more women work part-time than men, and that women spend more time in household duties than men of the same work status (part-time or full-time) [15]. There is also some evidence that women’s health may be particularly at risk from sedentary behaviour, as shown by associations of television viewing with metabolic syndrome [16] and all-cause mortality [17] in women, but not in men. Finally, the occurrence and nature of sleep disorders appears to be different in women and men [18,19] and gender differences in achieving 150 min/week of MVPA have also been noted [11,20]. Time spent sleeping, active and sedentary are mutually exclusive but as time in a day is finite these are likely to compete with or influence each other, particularly on work days where work tasks involve either active or sedentary behaviours. To date, no research has assessed whether time spent sitting, sleeping, and being physical active, differs among women in different occupational groups. Given that there are now more women in the paid workforce [21], it is important to understand patterns of sitting, sleep and PA in women, so that appropriate workplace health interventions can be developed. Data from the Australian Longitudinal Study on Women’s Health provide a unique opportunity to examine how occupation and work hours are associated with time spent in these three key health behaviours in a large population cohort of young and mid-aged working women. It is hypothesised that women in different occupations and with varied paid work hours may have different patterns of sleep, PA and sitting on work days, but that on non-work days their patterns will be similar.The Australian Longitudinal Study on Women’s Health (ALSWH) is a prospective cohort study of women’s physical and mental health, psychosocial aspects of health (such as socio-demographic and lifestyle factors) and use of health services [22]. The overall aim of the study is to examine the relationships between these factors and to inform governments on implications for health policy and practice. In 1996, three cohorts of women, young (born 1973–1978, aged 18–23 years; n = 14,247), mid-aged (born 1946–1951, aged 45–50 years; n = 13,715) and older (born 1921–1926, aged 70–75 years;  n = 12,432), were recruited. The sample was randomly drawn from the Australian national Medicare health insurance database which includes all Australian citizens and permanent residents [23]. The women completed a mailed survey every three years. The study has ethical approval from the Universities of Queensland and Newcastle Ethics Committees, and informed consent was received from all respondents.Data (cross-sectional) for this paper were taken from the 2009 survey of the young cohort and the 2010 survey of the mid-aged cohort, as these surveys included questions on domain specific sitting time and sleep duration. Participants were included if they reported being employed and answered the questions on domain specific sitting time, sleeping time, PA engagement and demographic variables included in the analyses (see the Supplemental Figure S1 for a flow chart of participant numbers). As the focus is on working women, data from the older cohort were not included.Participants were asked to report the time (hours and minutes) spent sitting on their most recent work and non-work day, separately for work, transport, television viewing, leisure-time computer use and all other purposes. Times spent in all domains were summed separately for work days and non-work days to give overall sitting time per work day and non-work day. These questions were based on the validated questionnaire developed by Marshall et al. [24]. Participants with data for total sitting time of greater than 24 h/day was not included in the analyses. Participants were also asked to report their sleep duration (hours and minutes) for the most recent work and non-work day. Data for participants who reported sleeping for greater than 24 h/day was not included in the analyses. Sleep duration was reported as a continuous variable (h/day) and categorised as shorter (≤6 h/day compared with >6 h/day) or longer sleep (≥8 h/day compared with <8 h per day) duration on both work days and non-work days. Sleep cut points were based on literature suggesting harmful effects of sleeping for 6 h or less and for 8 h or more [1,2,25]. PA was measured using the Active Australia Questionnaire [26], which asks participants to report the frequency and duration (hours) of brisk walking and moderate and vigorous intensity activity for transport or leisure in the last week in bouts of 10 min or more. PA was categorized as meeting guidelines or not, using a PA score of MET (metabolic equivalent) minutes/week, calculated as the sum of the products of total weekly minutes in each of the three categories and their generic MET values (walking minutes × 3.0 METs) + (moderate-intensity PA minutes × 4.0 METs) + (vigorous intensity PA minutes × 7.5 METs). A score of ≥600 was used to denote meeting current PA guidelines (≥150 min of moderate-intensity activity per week) [26].Participants reported their main occupation in the following categories: (a) manager or administrator; (b) professional; (c) associate professional; (d) tradesperson or related worker; (e) advanced clerical or service worker; (f) intermediate clerical, sales or service worker; (g) intermediate production or transport worker; (h) elementary clerical, sales or service worker; or (i) laborer or related worker. These categories were collapsed into manager or professional (a, b and c); clerical or sales (e, f or h) and trades, transport or laborer (d, g or i). Participants reported the average number of hours worked per day; responses were categorized as part-time (<35 h/week) or full-time  (≥35 h/week) under the heading of ‘work hours’. They also reported if their work was shift, night, casual, from home, self-employment or in more than one job (not mutually exclusive). ‘Work pattern’ was categorized as “shift or night” if participants reported they worked either shifts or nights; and all others were categorized as “not shift or night”.Area of residence, highest education level, marital status, number of children ≤16 years, smoking, alcohol intake, self-rated health, height and weight were self-reported. Body mass index (BMI; kg/m2) was calculated using self-reported weight and height and classified using the World Health Organization categories [27].Analyses were conducted in 2015 using SPSS version 22.0 (IBM Corporation, Armonk, NY, USA) with statistical significance set at p < 0.01 (two-tailed). Mean work hours, domain specific sitting time and overall sitting, and sleep variables were calculated separately for occupation (manager/professional, clerical/sales, trades/production/labourer), work hours (full or part-time) and work pattern (night/shift or regular hours) categories. Differences in mean values of sitting and sleep among groups of occupation, work hours and work pattern were tested using analysis of variance (ANOVA), adjusting for variables that were significantly associated with work variables and sitting or sleep in univariate analyses. Differences between time spent in workday and non-workday sitting and sleeping were assessed using ANOVA, with adjustment for demographic variables associated with workday and non-workday sitting or sleep (BMI, children (young cohort)), marital status, area of residence, education, smoking, alcohol intake and other work categories). Two-way ANOVAs were conducted to examine the interaction effect of working eight hours a day or more with occupation, work hours and work pattern on sitting time and sleeping on workdays. Time spent sitting in specific domains was reported as median (25th and 75th percentile). Physical activity hours per week were reported as medians (interquartile range) for each category of occupation, work hours and work pattern, with differences between categories tested using Kruskal-Wallis tests. The odds of reporting: (1) short and long sleep duration (≤6 h/day and ≥8 h/day) and (2) meeting physical activity guidelines, in categories of occupational, work hours and work pattern regularity variables, were examined using logistic regression models, both unadjusted and with adjustment for variables that were significantly associated with work variables, sitting time, sleep and PA (i.e., BMI, number of children (young cohort only), marital status, area of residence, education, smoking, alcohol intake, self-rated health and other work categories). Only findings significant at the 0.01 level in adjusted analyses were discussed in text. Data were included for participants who were working, answered sitting, sleeping and physical activity questions with realistic values and had demographic data of interest in the 2009 young cohort survey and 2010 mid aged cohort survey. A flow chart for inclusion is included in Supplementary Data Figure S1.The sociodemographic characteristics of the young and mid-aged women, separated into categories for occupation, work hours and work pattern, are presented in Table 1. Women in the three occupational groups were significantly different according to all measured characteristics in the young cohort, and all except BMI and alcohol intake in the mid-aged cohort. Similarly, women who worked part-time were different from those who worked full-time, on all measured characteristics in the young cohort and all except smoking, children under 16 years living at home and alcohol intake in the mid-aged cohort. Women who worked regular hours were more likely to be married and had a lower BMI than shift or night workers in both the young and mid-aged cohorts, and were less likely to have a post-high school qualification and be a current smoker in the mid-aged cohort only.Findings are presented for adjusted models except where unadjusted models showed disparate findings.Self-reported hours worked on work days, sleep (work and non-work days), sitting (work and non-work days) and MVPA are presented in Table 2. Hours spent at work on work days differed among all occupational, work hours and work pattern categories in both the young and mid-age cohorts; manager/professionals worked longer hours than clerical/sales and trades/production/ laborer (not in young cohort), full-time longer than part-time and shift/night longer than those who worked regular hours.There were differences in sitting time on work days and non-work days in occupational categories, with sitting being greater on work days than non-work days for manager/professionals, clerical/sales, full-time workers, and those working regular hours, and less for trades/production/laborers and shift or night workers (compared with alternate categories, Table 2). This pattern was consistent for both the young and mid-aged cohort. Part-time workers in the mid-aged cohort reported less sitting on work days than non-work days but there was no difference between these days in the young cohort. Differences between sitting time in various groups were mainly seen on work days, with trades/production/laborers sitting less on work days than clerical/sales and manager/professionals (differences around 3–4 h); shift/night workers sitting less than those working regular hours, and full-time workers sitting more than part-time, in both cohorts. There was an interaction between working longer hours and full-time/part-time status and with regular/shift night for the young cohort but not for the mid-aged cohort. In the young cohort those who work full time and those who work regular hours and work longer than 8 h/day reported longer sitting times. There was no interaction between working eight hours or more and occupation, with sitting longer for either cohort.Figure 1 shows the time spent sitting in specific domains. The main differences between work categories were in sitting for work on work days. Unsurprisingly, in both cohorts those who worked regular hours (compared with night/shift workers), full-time (compared with part-time) and were managers/professionals or sales clerical workers (compared with trades/production/laborers) sat more for work on work days. This difference was between two and six hours per day. Times reported for individual domain specific items are included in Supplementary Table S1.For all groups, sleep duration was about half an hour longer on non-work days than work days. On work days, shift/night workers slept less than regular workers in both cohorts; in the young cohort, trades/production/laborers slept less than other occupational categories and in the mid-aged cohort full-time workers slept less than part-time. On non-work days, full-time workers slept more than part-time workers in both cohorts and in the young cohort trades/production/laborers slept less than manager/professionals. Interactions between working eight hours a day or more and work characteristics, with sleeping were not significant. In the young cohort, the prevalence of shorter and longer sleep was 6.2% and 12.4% respectively on workdays and 2.8% and 35.9% respectively on non-work days. Similarly, in the mid-age cohort, the prevalence of shorter and longer sleep was 10.1% and 9.3% respectively on work days and 5.8% and 22.4% respectively on non-work days in the mid-aged cohort. The odds of reporting shorter and longer sleep are presented in Table 3. The odds of reporting shorter sleep duration on workdays were higher in shift/night workers than regular hours workers, for both the young and mid-aged cohorts.Time spent in MVPA also differed among various work characteristic groups and in the young and mid-aged cohorts. In the young cohort, clerical/sales workers reported less MVPA than manager/professionals and trades/production/laborers; and full-time workers reported more MVPA than part-time workers. In the mid-aged cohort, trades/production/laborers reported more MVPA than clerical/sales and manager/professionals; and full-time workers reported less MVPA than part-time workers. The odds of achieving PA guidelines are presented in Table 4. In adjusted analyses, the odds of achieving guidelines were lower for clerical/sales workers (compared to manager/professionals) in the young cohort and for full-time workers (compared to part-time) in the mid-aged cohort. There was a difference between the adjusted and unadjusted findings for the odds of achieving guidelines in the young cohort, when examining by work status. In adjusted models, there was no longer a significant difference between those who work part-time or full-time; the main modifier was having children or not.The workplace is an opportune setting for health promotion programs, which aim to improve employee health status and productivity [28]. Given that many women are now in the paid workforce [29], and are juggling their work with unpaid family and caring responsibilities [15], it is likely that time pressure might impact on the time available for sleep and physical activity. As many women are also in occupations such as office work, which require them to sit for long hours [6,30], they may be exposed to a trifecta of behavioural risk factors, too much sitting, too little sleep and insufficient physical activity. This study examined how these behaviours vary in women with different occupations and work patterns, as improved understanding of the patterns of these behaviours in working women could help to identify ‘at risk’ groups for health promotion intervention. Differences in self-reported sitting time, sleep and PA were found between groups of women with different work characteristics in both the young and mid-aged cohort. As hypothesised, these differences occurred on work days, particularly. Notably on work days, full-time workers were most likely to report high sitting time, as were those in the managerial/professional or clerical/sales occupations. Work days were, however, different from non-work days in many ways. Not surprisingly, sleep durations were lower on work days than non-work days, most notably amongst those who worked shifts or nights, but also in full-time workers across occupational groups. Physical activity was not reported separately for work and non-work days, but over the week. Nonetheless, work related differences were still noted, with mid-age full-time workers and younger clerical/sales workers least likely to meet the PA guidelines. Average sitting times were comparable with other Australian data [31] yet higher than those reported internationally [32]. The higher levels of sitting reported in the current study are likely to reflect the multiple domain measure, which provides estimates that are about 2 h higher (per day) than when a single item sitting estimate, such as the International Physical Activity Questionnaire is used [33]. On workdays, differences in sitting time followed expected patterns in both the young and mid aged cohorts; managers/professionals and clerical/sales workers, full time workers and regular pattern workers reported higher durations of sitting time, as has been previously shown [31,34,35] and this appears to be driven by high work sitting on work days (Figure 1). The amount of sitting time reported by younger women who work full-time with regular hours appeared to be particularly influenced by working longer hours, which often equate with more work-based sitting [36]. Non-work day sitting did not differ among any of the work-related groups, even for those who sat less on work days (trades/production/labourers). Differences in non-work day sitting have not been consistently observed in previous studies [31,34]. It appears that differences in sitting time in employed women are largely due to sitting for work, although this should be confirmed by studies using objective measures of sitting time.In both cohorts, average sleep duration was consistent with previous Australian reports [37,38], which have shown that shift-work is typically associated with shorter sleep durations and poorer sleep quality [39]. This was borne out by our findings that shift and night workers in both cohorts were mostly likely to report shorter sleep on work days. In both cohorts, average sleep times were longer on non-work days than workdays, which may be a strategy for trying to catch up on sleep.Previous reviews have identified that blue collar workers report higher levels of occupational PA and lower levels of leisure time PA than white collar and professional workers [11,40]. However, in the current study, after adjusting for covariates, it was not the ‘blue collar’ (used to describe the trades/production/labourer category in this study) workers, but the clerical/sales workers, who were significantly less likely to be physically active, when compared with managers/professionals in the young cohort. This finding is consistent with a previous report which found that managers/professionals were more likely to be active than other occupational groups, in women of similar age [41]. In the mid-aged cohort, there was no association between occupational category and physical activity, although full-time workers were less likely (than part-time) to achieve physical activity recommendations. This might be expected as there is potentially more leisure-time available with part-time work. As the nature of the association between occupational category and physical activity is inconsistently observed in other studies [7,11,13], it may be that life-stage is important. Even the mid-age women who worked full time were surprisingly active; this may reflect their lesser involvement in child care and household work than the younger women. Others have reported that MVPA declines with age [20], but previous ALSWH reports have shown that women tend to increase their activity levels when children leave home [42]. The presence of young children may be a factor in the differences noted in achieving physical activity guidelines between the young and mid-aged cohorts. In contrast to the mid aged cohort, full-time workers in the young cohort were more likely to achieve guidelines in unadjusted analyses, but this finding was diminished after adjustment for demographic variables. This may have been due to the fact that part-time workers in the young cohort were more likely to have children in the household (85% compared with 30% for full-time workers; Table 1). Previous research has shown that having young children decreases participation in physical activity [43,44].In terms of which work groups reported more sitting, short or long sleep, and not meeting physical activity guidelines, the two cohorts were remarkably similar. For both cohorts, manager/professionals and clerical/sales workers reported more sitting than trades/labourers, full-time more than part-time and those who worked regular hours more than shift/night workers. For sleep, those working shifts or at night were more likely to report short sleep on work days in both cohorts. Clerical/sales workers in the young cohort and the full-time workers in the mid-aged cohort were less likely to achieve physical activity guidelines. Interestingly, both these groups also reported higher sitting times, so it is possible that these groups are at even greater health risk as high sitting time and low physical activity have detrimental effects on health and mortality [45]. Patterns of sleeping and sitting across work and non-work days were also similar in the two cohorts. For most groups, participants reported sitting less and sleeping longer on non-work days than work days. It is possible that away from the restrictions imposed by work conditions and weekday schedules that women engage in less sitting time and attempt to increase sleep duration on non-work days. However, attempts to catch up sleep on non-work days by sleeping for longer may also present other challenges to maintaining a healthy sleep schedule and wellbeing [46]. Twenty-four-hour data collection would clarify the very variable movement and sleep patterns on non-work and work days.Several limitations should be acknowledged when interpreting the findings of this study. As with all cohort studies, there has been attrition over time, so that the included women were less representative of the general population than the starting cohort in 1996 and particularly in the young cohort, there is a bias towards more educated women [47]. However, the very large sample size and inclusion of women from across all occupational categories and work patterns is a strength. Another limitation is that all the data were self-reported, though the measures of PA and sitting have acceptable psychometric properties. In relation to sleep, we did not include measures of sleep quality, timing, chronotype or presence of sleep disorders. As the study only examines the associations cross-sectionally we cannot infer causality.The categories used for occupation type were broad and this may lead to some dilution or masking of effects. For example, clerical, sales and retail were collapsed into one category; however, we know that office workers (likely to include clerical workers) sit for longer at work than retail workers [6].While most of the findings reported here appear to reflect ‘conventional thinking’, they do provide insight for targeting public health campaigns. Occupational characteristics were associated with higher sitting, shorter sleep and less activity, suggesting an important role for workplaces in promoting healthier behaviours. Differences in sitting time are largely driven by sitting for work and those groups that experience high sitting time at work (manager/professionals, clerical/sales and full-time workers) should be targeted. Workplace modifications including sit-stand desks have been shown to be effective at reducing workplace sitting [48]. Those at health risk from short sleep are the shift and night workers, who may benefit from cognitive behavioural sleep programs [49,50]. The groups at risk from low physical activity varied in the two cohorts. The combination of low physical activity and high sitting time in young clerical sales workers and mid-aged full-time workers was of particular concern. These workers could benefit from a two-pronged approach to increase physical activity and decrease sitting time, perhaps replacing one behaviour with the other, which has been shown to have a beneficial effect in studies that used 24 h monitor data [51].These results indicate that the patterns, amounts and types of women’s paid work impact on three key health behaviours, which are known to have long term effects on the development of a range of chronic diseases. If women are to remain sufficiently healthy to stay in the workforce as they age, it will be important to identify groups who are most ‘at risk’ of poor health behaviours and develop tailored approaches for prevention. Those most at risk of either too much sitting, too little sleep, too little physical activity or a combination of these appear to be women who work shifts and nights, full-time hours, and those in clerical/sales and managerial roles. These results suggest multiple behaviour interventions tailored to work patterns and occupational category may be useful to improve sleep, sitting and activity behaviours in working women.The following are available online at www.mdpi.com/1660-4601/14/3/290/s1, Figure S1: Flow Charts for inclusion of participants’ data young and mid-aged women in the Australian Longitudinal Study on Women’s Health. Table S1: Time reported on the individual sitting time questions on work days and non-work days (hours/day) in young (n = 4650) and mid-aged (n = 3185) women in the Australian Longitudinal Study on Women’s Health (ALSWH).The research on which this paper is based was conducted as part of the Australian Longitudinal Study on Women’s Health, The University of Newcastle and The University of Queensland. We are grateful to the Australian Government Department of Health and Ageing for funding and to the women who provided the survey data. B.K.C. was supported by an Australian National Health and Medical Research Council (NHMRC) Program Grant (NHMRC no. 569940). B.K.C. was also supported by NHMRC Early Career Fellowship #APP1107168. M.J.D. is supported by a Future Leader Fellowship (ID 100029) from the National Heart Foundation of Australia. No financial disclosures were reported by the authors of this paper. All authors made significant contributions to the manuscript. Contributions were: the initial concept and design (B.K.C., T.L.K.A., M.J.D., W.B.), data acquisition (W.B.), statistical analyses and interpretation (B.K.C.), preparation of manuscript (B.K.C.), preparation of tables and figures (B.K.C.) and revision of manuscript B.K.C., T.L.K.A., M.J.D., W.B.).The authors declare no conflict of interest.Time (h/day) reported for the individual sitting time questions on work days and non-work days by (A) young and (B) mid-aged women. Adjusted for marital status, living with children (young cohort only), education, work hours, body mass index, drinking, smoking, area of residence and other work factors. Young cohort: in 2009 aged 31–36 years (n = 4650); Mid-aged cohort: in 2010 aged 59–64 years (n = 3185).Demographic characteristics of the young (n = 4650) and mid-aged (n = 3185) women by occupational category, work hours and work patterns; Australian Longitudinal study on Women’s Health (ALSWH).Young cohort: in 2009 aged 31–36 years; Mid-aged cohort: in 2010 aged 59–64 years; a: three or more standard drinks per day; N: number; SD: standard deviation; p: level of significance.Total sleep, sitting and physical activity duration on work and non-work days reported by young (n = 4650) and mid-aged (n = 3185) women in the ALSWH.Data are mean (standard deviation) except where indicated. Testing for differences in mean values are adjusted for body mass index (BMI), number of children, marital status, area of residence, education, smoking, alcohol intake, self-rated health, work hours per day and other work categories a: different to top category; b: different to second category.Odds ratios (and 95% CIs) for reporting short and long sleep duration young (n = 4650) and mid-aged (n = 3185) women in the ALSWH.Analyses adjusted for BMI, children (young cohort), marital status, area of residence, education, smoking, alcohol intake, self-rated health and other work categories. Shorter sleep compared to >6 h/night, longer sleep compared to <8 h/night. *: p < 0.01; **: p < 0.001; CI: confidence interval; Ref: reference category.Odds of young (n = 4650) and mid-aged (n = 3185) women reporting meeting physical activity guidelines in the ALSWH.Adjusted analyses included BMI, marital status, children <16 years in household (young cohort), area of residence, education, smoking, alcohol intake, self-rated health and other work categories. *: p < 0.01; **: p < 0.001; Ref: reference category. 

Med-MDPI/ijerph_2/ijerph-14-03-00291.txt

@@ -0,0 +1,37 @@
+Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).The study explores spatial data processing methods and the associated impact on the characterization and quantification of a combined health risk indicator at a regional scale and at fine resolution. To illustrate the methodology of combining multiple publicly available data sources, we present a case study of the Lorraine region (France), where regional stakeholders were involved in the global procedures for data collection and organization. Different indicators are developed by combining technical approaches for assessing and characterizing human health exposure to chemical substances (in soil, air and water) and noise risk factors. The results permit identification of pollutant sources, determinants of exposure, and potential hotspot areas. A test of the model’s assumptions to changes in sub-indicator spatial distribution showed the impact of data transformation on identifying more impacted areas. Cumulative risk assessment permits the combination of quantitative and qualitative evaluation of health risks by including stakeholders in the decision process, helping to define a subjective conceptual analysis framework or assumptions when uncertainties or knowledge gaps operate.Humans are exposed daily to multiple chemical and non-chemical (e.g., biological, physical, or psychosocial) stressors. However, toxicological and epidemiological studies typically examine individual stressor-response relationships. Ideally, direct measures of exposure (e.g., biomarkers or personal monitoring data) would be available for all key stressors related to a common health effect throughout the critical time period of exposure and in the population of interest [1]. The exclusive use of biomarker data in cumulative exposure assessment efforts is currently not practicable when considering a large number of diverse chemicals due to analytical and resource limitations [2], especially when the assessment should cover a large territory. Environmental quality data are often available at a fine administrative or resolution level, and enable the building of environmental indicators on a regional scale. The definition of indicators for the identification and characterization of environmental inequalities depends on the reutilization of this type of data, which is very diverse by nature, with regard to its initial intended objectives. In France, this kind of data has already made it possible to highlight important regional disparities in the distribution of environmental quality [3,4]. To date, geographical information systems (GIS) technology has proven to be a powerful tool for dealing with various types of environmental data. Some studies integrate georeferenced measure monitoring or modeling data to estimate the exposure dose, and may include studies on various single environmental media, such as soil [5], water [6], and air [7,8], or a multimedia approach [9]. There will be cases where risk cannot be quantified in any meaningful or reliable way due to lack of representative data or missing source contributions. In order to reduce the spatial data representativeness problem (based on the lack of available data) and characterize associated uncertainty, more sophisticated methods of spatial analysis have been developed [10,11]. Qualitative approaches could also be used to overcome the complexity and data deficiencies that hinder quantitative approaches. Broad indicators using geographically based measures of exposure are used as an indicator of cumulative exposures from all of the potential chemicals associated with that site.Cumulative risk assessment (CRA) is defined as a science policy tool for organizing and analyzing relevant scientific information to examine, characterize, and quantify the combined adverse effects on human health from exposure to a combination of environmental stressors [12]. The ultimate goal of cumulative risk assessment is to provide answers to decision-relevant questions based on organized scientific analysis, even if the answers, at least for the time being, are inexact and uncertain [13]. Cumulative risk assessment therefore involves the quantitative or qualitative evaluation of risks to health and/or the environment from multiple exposures, sources, and routes, while considering differential susceptibility or vulnerability of population subgroups [14]. Due to the limited availability of integrated data on multiple stressors, analytical complexity, and method limitations, exposure assessment is one of the main challenges for CRAs.Assessing risk that includes multiple different risk factors is considerably more complex methodologically and computationally than aggregate risk assessments or single-effect cumulative risk assessments. The advantage of a decision index is the ease in converting highly multivariate technical information into a single number. The approach involves developing a composite score—or index—from measures of various risk dimensions [14]. Various environmental risk indexes have been developed and applied to ranking and comparative analyses [15,16,17]. Often, those indexes use surrogate measures for risk rather than actual calculations of the probability of adverse effects. There is relatively little experience in combining different types of risk. A key issue seems to be the need for method development in this area. Some approaches require synthesizing a risk estimate (or risk indication) by “adding up” risks from different parts of the risk dimension [18]. In these cases, risk assessment requires a common metric such as an exposure dose or hazard quotient. For example, emissions of both carcinogens and non-carcinogens are weighted by a toxicity factor, so they can be combined in a risk-based screening “score” for a particular geographic area by the Environmental Protection Agency’s (EPA) Office of Pollution Prevention and Toxics [18]. Finding a common metric for dissimilar risks is not a strictly analytic process, because judgments must be made as how to link two or more separate scales of risks. These judgments could involve a subjective conceptual analysis framework defined during a deliberative process, including stakeholders to make good decisions and generate operational actions adapted to the policy objectives. Unfortunately, methods used to combine indicators are sometimes selected in an arbitrary manner, with little particular attention paid to the data standardization procedure. This can lead to indices which overwhelm, confuse and mislead decision-makers and the general public.The present study aims to explore spatial data processing methods and the associated impact on the characterization and quantification of a combined health risk indicator. To illustrate the methodology of combining multiple publicly available data sources, we present a case study of the Lorraine region (France), where regional stakeholders were involved in the global procedures for data collection and organization. We also explore technical approaches for assessing and characterizing human health risks associated with a subset of cumulative risk issues.To achieve those objectives, combining a data process and transfer modeling with a spatial approach is fundamental, a prerequisite that implies the need to first overcome different scientific limitations:
+      
+        
+          selecting and processing interest variables that could be built to associate and partly describe the source-effect chain;
+        
+        
+          developing indicators that permit the combination of risk factors.
+        
+      selecting and processing interest variables that could be built to associate and partly describe the source-effect chain;developing indicators that permit the combination of risk factors.We describe here a CRA case study characterizing combined exposures to noise with chemical contaminations of water, air, and soil. The study’s aims are (1) to present an approach that utilizes existing data for comparisons across exposures and populations that could be useful for identifying at-risk populations; and (2) to explore the advantages and disadvantages of using data standardization methods.In France, environmental health inequalities are understood as the unequal geographic distribution of multiple exposure. No statewide data are available that provide direct information on exposures. Exposures generally involve transfer of chemicals from a source through the environment (air, water, soil, food) to an individual or population. For the purposes of the study, data relating to pollutant sources, releases, and environmental concentrations are used to build indicators of potential human exposures to pollutant. Pollution burden indicators should relate to issues that may be potentially actionable by stakeholders. Based on the regional context and data availability, four dimensions were identified and found consistent with criteria for exposure composite indicator development: water, air, soil, and noise. Then, four subindicators should provide a measure that is relevant to the dimension it represents in the context of the study objectives. The subindicators used should also represent widespread concerns related to pollution in Lorraine and provide a good representation of each component.A composite indicator has to identify cumulative risk factor areas rather than hotspot areas of only one risk factor. In this cumulative risk assessment, the key aspect was to highlight areas where multiple stressors act together in contributing to risks. In this way, the strategy defined by the working group was to apply equity constraints for each risk factor. That means that each subindicator had to have a similar weight (equal average and range on the modeling domain) to build the composite indicator based on the sum of each standardized subindicators. The conceptual model is presented in Figure 1.Aggregation of the different factor risks was made using different methodologies for discussing the impact of weighting and aggregation procedures on the effectiveness of risk maps for taking decisions for safeguarding citizen health. The subjective conceptual analysis framework was adopted using a deliberative process to define the common metric that would permit calculation of the composite indicator for dissimilar risks.Data proceeding methods emerge from basic risk assessment concepts and is sufficiently expansive to incorporate multiple factors that reflect population impacts that have not been included in traditional risk assessments. A GIS-based modeling platform for quantifying human exposure (PLAINE: environmental inequalities analysis platform [9,19]) was used to build health risk indicators within the Lorraine region (France). The GIS-based platform permits researchers to:
+        
+          
+            gather emission sources, environmental and population databases;
+          
+          
+            discretize variables on a referent grid (data mapping);
+          
+          
+            transform variables into exposure indicators (exposure transformation);
+          
+          
+            derive subindicators by combining exposure indicators weighted by toxicological data or threshold values (data processing);
+          
+          
+            build the composite indicator from standardized subindicators (indicator development).
+          
+        gather emission sources, environmental and population databases;discretize variables on a referent grid (data mapping);transform variables into exposure indicators (exposure transformation);derive subindicators by combining exposure indicators weighted by toxicological data or threshold values (data processing);build the composite indicator from standardized subindicators (indicator development).The case study concerns Lorraine, the north-eastern region of France, bordering Germany and comprising the departments of Meurthe-et-Moselle, Meuse, Moselle and Vosges. It is a moderately densely populated region, with an area of 23,547 km2 and 2.5 million inhabitants—4% of France’s total population—making it the 11th most populous region in the country. It ranks eighth in GDP (gross domestic product) among the 26 regions of France, placing it per capita among the top economic producing regions in the country, along with Alsace and Île-de-France (Paris). From the end of the 19th century to the 1960s, the economic development of Lorraine was built on two dominant industries: coal mining and steel production. The logistics and service sectors have experienced the strongest growth in recent years, while traditional industries have undergone a decline. Consequently, the region has experienced major difficulty with increasing unemployment, although it is still below the national average. In 1997 the last iron ore mine in Lorraine, which once produced over 50 million tons of iron, was closed.Each region drew up a Regional Environment and Health Action Plan to implement the main objectives of the French National Action Plan according to its own specific needs. Different regions in France, including Lorraine, have included environmental health inequalities reduction in their planning, and need assessment to guide priorities for voluntary action.Transparency of decision making and policy development is the cornerstone of environmental inequalities reduction action. In that spirit, a working group consisting of regional representatives of environmental database managers, thematic experts, and environment and health regional stakeholders (see acknowledgements) was created to define the study objectives definition and the conceptual framework. This group was particularly involved in data collection, selecting stressors, and ensuring the adequacy of the assessment results with potential action implementations.A wide range of data on different sources, agents, pathways and media might potentially be required for integrated assessment of environmental health risks. These data might be used as inputs for models or proxies for other exposure metrics. The decision was made to focus on sources relevant either directly (as measures of exposure) or indirectly (as potential input variables for modeling). In addition, population data were included, since this provides important proxies for source activity in many instances and is, of course, an essential component of exposure assessment.A data inventory was made reflecting the following main themes: soil, land cover, air, drinking water quality, atmospheric emissions/concentrations, polluted sites, and soil and exposure factors. In order to make the task manageable, attention was initially focused on data available at the regional scale that are gathered on a routine basis. Nevertheless, geographic coverage or extent, for example, is inevitably ambiguous. Indeed, most environmental data are samples and do not provide complete area coverage. Hence, in many cases, approaching full regional coverage is possible if different datasets are combined. Some data for Lorraine is a subset of the national monitoring network, so the density of sampling across Lorraine may be sparse. In this study, data sources have generally been included when they were considered to represent a potential basis for assessing exposures across populations at the regional level, either directly or by interpolation. From this inventory, the selection of the database was made based on the interests of and uses for this study (Table 1).A share-of-population census, monitoring, and modeling of environmental quality data production were conducted independently of each other in accordance with specific needs and constraints. This discrepancy implies that the different data types from different sources and support databases cannot be directly represented under a common denominator, namely their spatial location or distribution. Representation is therefore achieved by depicting the different data types as layers and superposing those layers in the same geographical reference grid. The problem of linking data sets derived from incompatible spatial frameworks (for example, linking point- and pixel-based environmental data) has attracted considerable attention. A referent grid of 1 × 1 km was generated for the study, and all the spatial variables were discretized on this grid. Tools have been developed using modeling, spatial analysis, and geostatistic methods to build and discretize interest variables from different supports and resolutions on the 1 km2 regular grid within the Lorraine region. For example, surface soil concentrations were estimated by developing a kriging method able to integrate surface and point spatial supports [11]. For water, distribution unit serve maps were used to spatialize water data measured at water treatment plants. Modeled or estimated noise and air variables were aggregated from their initial grids to the referent grid using surface ratios. Buffer zones around potentially contaminated sites and soils were generated using a distance (300 m) defined by the working group. GIS was used to partition the proximity data assigned to the areal unit of the referent grid that is only partially within the distance buffer into “inside the buffer” and “outside the buffer” portions based on the percentage of the areal unit that lies within and without the distance buffer, respectively.Different methods were used to transform environmental spatial datasets into exposure variables. An exposure model developed by INERIS (MODUL’ERS [22]) was used to assess the transfer from soil to individual exposure through ingestion pathways (soil and vegetation pathways). This model was used to estimate population age class hazard quotients (HQ) from interpolated topsoil trace metal concentrations and for estimating non-cancer risk. For the ingestion pathway, the HQ is the ratio of the average daily dose (ADD; milligrams per kilogram per day) of a chemical to the reference dose (RfD, milligrams per kilogram per day), defined as the maximum tolerable daily intake of a specific pollutant that does not result in any deleterious health effects.Generally, to combine HQs, stressors need to have a common target organ [14]. We assumed independence of action and we summed the HQs to build the topsoil concentration indicator. Use of this exposure model to map exposure indicators can be seen in detail in Caudeville et al. [9]. The air concentration indicator was estimated using the sum of the ratio between the annual average pollutant atmospheric concentrations and the European air quality standards [23]. Broad indicators were built using geographically based measures of hazard as a cumulative measure. For example, we used distance from a polluted soil site to build a proxy based on the density of the potentially contaminated site by areal unit. A score was used to estimate the relative risks of direct emissions by combining total pollutant emissions (sum of pollutants) and toxicity-weighted pollutant emissions for cancer or respiratory non-cancer effects. Weighting emissions by toxicity does not take into account fate, transport, or location and behavior of receptor populations. It is often desirable to aggregate indicators into broader thematic indices. The air risk factor indicator combined modeled concentrations and estimated emissions following the equity principle to give similar weights to the two dimensions with similar area numbers, global indicator averages and indicator ranges. Site proximity and topsoil concentration databases were also combined into a higher-level soil indicator simply by adding them.For water, drinking water concentrations were compared to European drinking water standards (chosen previously by a different working group) in a tool developed by the Regional Health Agency of Lorraine. The four-year averaged number of substance exceedance thresholds permitted us to build a score. An elevated score indicates that drinking water supplied in those areas could have concentrations that could lead to chronic disease in the population. The link between exposure and outcome (other terms: endpoint, reaction, response) was given by reasonably well-established exposure-response curves which are derived from research into noise effects. The Lden indicator (developed in the context of the noise European framework) was used to map noise around road infrastructures. It corresponds to the average sound pressure level over all days, evenings and nights in a year.The standardization procedure described here subjects subindicators to two different transformations that yield dimensionless and comparable figures. These can readily be aggregated to a higher-level thematic indicator simply by adding them. Aggregation of the different risk factors was made using different methodologies to discuss the impact of weighting and aggregation procedures on the effectiveness of risk maps used for making decisions safeguarding citizen health. Two methods were explored to build a homogeneous metric that permitted us to respect the equity constraint defined by the working group.The first method used a normal score function applied to transform each dataset into a normal distribution varying between 0 and 1. A score was assigned for each geographic unit derived from the ranks of the observations within the dataset. For each individual grid, a value was assigned which either expressed exactly or approximated the order statistic expectation of the same rank in a sample of standard normal random variables with the same size as the observed data set. The second method assigned a percentile, varying between 0 and 1, for each subindicator and geographic unit, based on the rank order of the value. A percentile was calculated from the ordered values for all areas that have a score.When a geographic area had no indicator value (for example, an area that had no noise estimation) or had exposure values equal to zero (for example, an area with no water exposure hotspot), a background exposure value was assigned corresponding to the mean of the first missing transformed ranks. This approach permitted us to obtain data independent of the chosen unit and scale with a similar average and range for each subindicator. Those scores allow comparison of one geographic area to other localities in the region where hazard effect data or population characteristics are present. Thus each area’s score for a specific indicator is relative to the ranks of that indicator in the rest of the region.The mathematical formula for calculating the composite indicator of the two methods used addition of the normalized or percentile-ranked subindicators. The method used existing environmental data to create a screening score for the population across the area. The population size at fine resolution was used to weight the composite indicator spatial aggregation at the French census block level. An area with a high score would be expected to experience much higher impacts than areas with low scores.One distinctive aspect of CRAs is their ability to examine multiple stressors that may affect health outcomes. Excluding non-chemical stressors from analysis may underestimate cumulative exposure and/or risk [1]. We illustrate a method that utilizes publicly available data sources and existing analytical methods to examine chemical and non-chemical stressor exposures to inform screening-level CRAs in order to identify subpopulations that may have a higher level of concern. Our method uses a combination of toxicological/threshold values and data transformation methods to characterize the unequal geographic distribution of environmental risks.Subindicators are presented here as regional maps (Figure 2).The air concentration indicator variations are weak throughout the studied area due to the background exposure concentration (Figure 2a). The area with the most elevated values corresponds to urban agglomerations. For drinking water, the map presents several hotpots corresponding to one or a combination of different pollutant concentrations above defined thresholds (Figure 2b). For example, the most elevated area of concentration (2.25) corresponds to natural arsenic and fluoride exceedances averaged during the four-year period of the study. The atmospheric emission indicator (Figure 2c) presents a similar pattern to that of the air concentration, but it also integrates district-level data in the higher-value district where an industrial site is located. The most elevated value (9.5) corresponds to polycyclic aromatic hydrocarbon and benzene emissions from steel industry activity. The noise map (Figure 2d) presents no value (as 0) for 90% of the studied area. Existent values are located on the region’s principal roads, based on available modeled noise levels. Principal contaminated sites and soil are located around a north-south axis called the Lorrain furrow (Figure 2e). The soil concentration indicator map presents two areas in which the value is greater than four (Figure 2f). These correspond to well-known contaminated sites. The largest value corresponds to a topsoil contamination of Hg, Cd, Cu and Zn, and the other to elevated concentrations of Cd in the topsoil. Specific spatial patterns are influenced by data spatialization methods and exposure variable transformation. Spatial resolution could also have an impact on individual area indicator values. In contrast to water risk factor indicators, where spatialization corresponds to a surface ratio of the initial spatial layer, the distribution of the emission indicators depends on the size of the geographic support aggregation.Maps of the combined exposure variable indicators for air and soil are presented in Figure 3. Exposure variables of contaminated sites and soil and topsoil concentration data were combined in order to integrate the soil contribution of rural areas (in this database, topsoil concentration samples are mainly located in forests and agricultural fields) and urban areas (contaminated sites and soil are historically located in urban areas). The atmospheric emissions and concentration dataset were combined in order to take into account three conventionally estimated pollutant concentrations (O3, PM10, NO2) and the emissions of 24 pollutants, for better taking into account industrial sources.The largest risk expressed by the composite risk indicator, obtained using the normal transformation method, corresponds to an industrial site. Spatial patterns of hotspot exposure are localized on the Lorrain furrow, reflecting the association between regional industrial and organized urban space dynamics (Figure 4).The composite indicator of spatial patterns depends on the local combination of the individual subindicators and their local interrelationships. Table 2 presents the correlation coefficients (r) obtained between the different estimated risk factor indicators.Weak correlations were found between the subindicators. The highest correlation (r = 0.365) corresponds to noise and air, due to similar environmental sources (automobile transport). The regression analysis revealed low negative correlations between water and the other subindicators. This subindicator is therefore less implicated in the highest composite indicator values.Figure 5 shows subindicator contributions above the 90th percentile for the composite indicators estimated using the two transformation methods. The noise and water risk factor contributions are quite similar. The variation of soil and air risk factors may be explained by the slope curve impact on the resulting composite indicator. Air and soil curves have similar forms for the percentile rank method (see Appendix A, Figure A1b,d), in contrast to the normal transformation for which the slope on the maxima range is more flattened for the soil than for the air risk factor (see Appendix A, Figure A1a,c). This results in a stronger contribution of the air risk factor in the composite indicator based on normal transformation compared to uniform transformation.Efforts were made to select complete, accurate and current datasets for inclusion. Nonetheless, there are different kinds of uncertainty that are likely to be introduced in the development of this type of approach. Those uncertainties mainly depend on: (1) the combination method’s impact on the capacity of the selected indicator metric to reflect the considered phenomena; (2) data representativeness, which controls the degree to which data gaps or omissions influence the results. The latter mainly concerns missing spatio-temporal aspects, source contributions, and the characterization level of the exposure scheme (from the source to the external exposure modeling with the integration or not of transfer/transport phenomena and population behavior).Empirical methods could be set up, driven by the will to characterize other contributions not integrated in an initial database. Those choices are guided to reach the best compromise between data representativeness and method robustness, consistent with the objectives of the study.In chemical mixtures risk assessment, exposure addition, more commonly called dose addition, assumes a common toxic mode of action across compounds, or at least evidence of toxicologically similar responses, so that the “total dose” is of concern for the assessment. Where only qualitative data is available, proxy indicators can be built, but are more difficult to use for measuring exposure quantitatively and for combining with exposure assessment variables. In our study different options were proposed and adopted by the working group.Two data transformation methods were applied to provide a common metric for each subindicator (a single function applied to each X or each Y data value) with respect to the equity constraint. The indicators used in this analysis have varying underlying distributions, and distribution normalization or percentile rank calculations provide a useful way to transform data. Nevertheless, the choice of a transformation implies the making of assumptions about those distributions (normal for the normalization or uniform for the percentile rank transformation method) that control the degree to which the data that are included in the model are correct.Therefore, each area’s value for a specific indicator is relative to the distance to the average in the data space in the rest of the places in the region. The distribution form used will impact the weight of an individual area in the resulting composite indicator. In our case, where cumulative hotspot exposure is the desired measure, better characterization of the highest values is researched.The transformation needed to reproduce the relative distance between each point of the original subset reflects the efficiency of this function to limit the over- or underestimation of a range of points. As the sigmoid form shows (see Appendix A, Figure A1b), rank percentile transformation smooths the extremum values. Because the composite indicator objective is used to highlight potential hotspot exposure, the highest value distance respects are the most important/critical. In the test case, the normal function permits a better description of the outliers. An over- or underestimation will impact the global ranking of other individual areas or subindicator weightings. The adequacy of the expressed relative distance between points on a specific curve range could be characterized by the curve slope (a low slope implies a potential underestimation).Our environmental noise estimates were only based on modeled noise levels from road traffic. Since industries, railroads and an airport also exist within the study area, it is likely that road traffic is not the only main contributor to human activity–related environmental noise in this region. While environmental noise may be the primary source of background noise in communities, non-environmental sources of noise may also be present and influence individual-level noise estimates. In this study, background exposure was not taken into account due to data deficiency. For water-related data, background exposure was also not integrated, due to the subindicator calculation mode where water concentrations below substance-specific thresholds were not considered.Ranking the data involved putting the values in numerical order and then assigning new values to denote where in the ordered set they fall. In those two datasets there are ties in the data where no value or no hotspot exposure is considered, expressed as a zero value. This means that several values are the same, so that there is no strictly increasing order. For the considered background exposure, we averaged the ranks for the tied values (see Appendix A, Figure A1e–h). This processing resulted in a heteroscedasticity creation (unequal variances) that impacted the local contributions of associated subindicators on the resulting composite indicator. Affecting a background exposure value will generate a gap between the initial null value and the rest of the distribution. It tends to flatten the curve slope, reduce the distance between points in the maxima value range, and finally decrease subindicator sensibility in the resulting composite indicator.In order to choose a calculation design in the context of environmental inequalities, certain queries need to be answered. One of these is to provide a uniform basis for mapping that is fine enough to reflect local variations in exposures, both to aid visual representation and interpretation of the data and to facilitate analysis of spatial patterns. Regular grid systems generally best satisfy that criterion and permit us to reduce the so-called “small number problem”, which can lead to highly unstable estimates of risk and large variations in uncertainty between zones [24].More specifically, we need to define a calculation mode in an attempt to overcome scientific knowledge gaps in combining quantitative and qualitative approaches. A subjective conceptual analysis framework was set up during a deliberative process including stakeholders. This included the need to traduce the working group’s adopted “rules” in terms of calculation assumptions and designs. For example, the equity constraint proposed here requires for each subindicator a common metric with a similar mean and range. Different distributions are permitted with respect to this constraint, such as family symmetrical distributions (uniform, normal, logistic, etc.), which can be used during the standardization procedure. The selection of the distribution must be led, in order to reduce the distance between point distortions generated during the standardization procedure as much as possible. This can be achieved by estimating the best fitting distribution by ranking goodness-of-fit statistics using Anderson–Darling or Kolmogorov–Smirnov tests. If better precision is desired at a specific range of the distribution (as the extremum values), a sliding window regression could be used to compute slope estimates along with the curve and help interpret the potential impact of the transformation on the individual subindicator area value. The distribution selection process must also take into account the impact of heteroscedasticity that arises by assigning default values for background exposure or areas where values are missing. This might be measured by comparing the ratio between the assigned and the first unassigned default values. Higher variance differences will decrease the sensibility of the subindicator in the composite indicator. In our test case, normal transformation was preferred over uniform standardization. However, the choice of calculation design is sometimes a compromise between potential competing needs.This pilot study successfully applied a composite risk indicator using a cumulative screening method at a fine resolution in Lorraine. The issues confronted when considering such a wide range of different data sources provided insight into ways to improve data generation and collection. However, we encountered several limitations in regards to specific indicators. Indicators are surrogates for the characteristics being modeled, so a certain amount of uncertainty is inevitable. That means this model, comprised of a suite of indicators, is considered useful in identifying places burdened by multiple sources of pollution. Qualitative approaches may be used to overcome the complexity and data deficiencies that hinder quantitative approaches. Cumulative risk assessment permits the combination of quantitative and qualitative evaluation of health risks by integrating stakeholders in the decision process of defining the subjective conceptual analysis framework or assumptions when uncertainty or knowledge gaps operate. Engaging stakeholders associated with the development, review, and use of exposure-science information contributes to formulating problems, collecting data, accessing data, and developing decision-making tools.Using a limited data set, a test of the model’s assumptions to changes in subindicator spatial distribution showed the impact of data transformation in identifying more impacted areas.Our results permitted us to identify pollutant sources, determinants of exposure, and potential hotspot areas. A diagnostic tool was developed for stakeholders to visualize and analyze the composite indicators in an operational and accurate manner. The designed support system will be used in many applications and contexts:
+      
+        
+          mapping environmental disparities throughout the Lorraine region;
+        
+        
+          identifying vulnerable populations and determinants of exposure, to set priorities and target for pollution prevention, regulation, and remediation;
+        
+        
+          providing exposure databases to quantify spatial relationships between environmental, socioeconomic and health indicators.
+        
+      mapping environmental disparities throughout the Lorraine region;identifying vulnerable populations and determinants of exposure, to set priorities and target for pollution prevention, regulation, and remediation;providing exposure databases to quantify spatial relationships between environmental, socioeconomic and health indicators.Over the next few years, we plan to refine the method by using spatial models to combine the global source–exposure–effect chain and to integrate additional indicators more adapted for agricultural or urban contexts (such as pesticide substances or radiofrequency exposure). In addition, we will look for new ways to integrate population mobility into exposure estimations. Exposure indicators and data processing algorithms will be integrated in the French coordinated integrated environment and health platform PLAINE to map and analyze environmental health inequalities at the national scale.This work was supported by the Lorraine Regional Environment, Planning and Housing Agency (DREAL). Our particular thanks go to P. Borr (DREAL) and K. Theaudin (Regional Health Agency, RHA) and A. Gill (SGAR: prefecture service) for their vision regarding the importance of this work. We would also like to thank all members of the working group, including G. Fourniquet (Bureau de Recherche Géologiques et Minières, France's leading public research institution for the earth sciences), J.P. Schmitt (AASQA, Official Air Quality Monitoring Associations), P. Vannier (RHA), M. Fleury (DREAL), H. Boulanger (RHA), C. Aubrège (SGAR), N. Mehira (SGAR), E. Boiselet (DIRECCTE), J. Gabe (DREAL, and T. Gambini (Health Regional Service) for their contributions to this work.The work presented here was conceived of, carried out and analyzed by J.C. D.I., E.M. and R.B. gave important suggestions. All authors read and revised the manuscript and approved the final version.The authors declare no conflict of interest.Linear regressions on the original data were used to test the assumption efficiency of the underlying transformation (Figure A1). For air and soil, the normal transformation (Figure A1a,c, R2 = 0.96 and 0.93) presents a better representation than the rank percentile transformation (Figure A1b,d, R2 = 0.88 and 0.84). This is due to the original distribution form of the subset, which better fits a normal distribution than a uniform distribution.Correlation between subindicator before and after data transformation: (a) air and normal transformation; (b) air and percentile rank transformation; (c) soil and normal transformation; (d) soil and percentile rank transformation; (e) water and normal transformation; (f) water and percentile rank transformation; (g) noise and normal transformation; (h) noise and percentile rank transformation.Conceptual framework of the data proceeding for determining composite indicators.Maps of exposure variables for: (a) air concentration; (b) water exposure hotspots; (c) air emissions; (d) LDen noise; (e) potential site and soil contamination; and (f) soil concentration.Maps of combined exposure variables for (a) air and (b) soil.Composite indicator map (SN method) aggregated at the French census block level.Histograms of the percentage contribution of the risk factor to the composite indicator above the 90th percentile for the normal transformation (SN = blue) and percentile rank method (Perc = red).Data sources included in the study.Spearman correlation coefficients (r) between the different risk factor indicators estimated.

Med-MDPI/ijerph_2/ijerph-14-03-00292.txt

@@ -0,0 +1,15 @@
+Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).The health impact on populations residing in industrially contaminated sites (CSs) is recognized as a public health concern especially in relation to more vulnerable population subgroups. The aim of this study was to estimate the risk of congenital anomalies (CAs) in Italian CSs. Thirteen CSs covered by regional CA registries were investigated in an ecological study. The observed/expected ratios (O/E) with 90% confidence intervals (CI) for the total and specific subgroups of CAs were calculated using the regional areas as references. For the CSs with waste landfills, petrochemicals, and refineries, pooled estimates were calculated. The total number of observed cases of CAs was 7085 out of 288,184 births (prevalence 245.8 per 10,000). For some CSs, excesses for several CA subgroups were observed, in particular for genital and heart defects. The excess of genital CAs observed in Gela (O/E 2.36; 90% CI 1.73–3.15) is consistent with findings from other studies. For CSs including petrochemical and landfills, the pooled risk estimates were 1.10 (90% CI 1.01–1.19) and 1.07 (90% CI 1.02–1.13), respectively. The results are useful in identifying priority areas for analytical investigations and in supporting the promotion of policies for the primary prevention of CAs. The use of short-latency effect indicators is recommended for the health surveillance of the populations residing in CSs.Human health impacts in contaminated sites (CSs) are of great concern worldwide. The European Environment Agency has estimated there to be approximately 2.5 million potentially contaminated sites, many of which are related to industrial activities [1,2]. Census activities regarding CSs were recently carried out [3,4], and monitoring is currently in progress in Europe [5,6]. Industrial activities can cause widespread contamination, and the inadequate environmental management of these sites can potentially affect human health [7,8]. WHO provided a definition of CSs from a public health perspective: “Areas hosting or having hosted human activities which have produced or might produce environmental contamination of soil, surface or groundwater, air, food-chain, resulting or being able to result in human health impacts” [7]. This refers mainly to industrial activities, including waste disposal and treatment.There are several approaches for assessing the impacts on human health most of which use an epidemiological approach [9,10]. A first level of epidemiological investigation is aimed at defining the health profile of the population residing in the proximity to CSs. The SENTIERI (Epidemiological Study of Residents in Italian National Priority Contaminated Sites) project has adopted a useful approach for defining the health status of residents in contaminated sites [11]. The project defines a framework of the health profile of populations residing in Italian contaminated sites defined by ministerial decrees as remediation sites of national interest. Mortality, hospitalization, and cancer incidence of populations residing near CSs have been investigated by epidemiological studies [11,12,13,14]. The health impact on populations residing in CSs is a public health concern, especially in relation to the most vulnerable population subgroups. Infants and fetuses are more susceptible than adults to environmental contaminants [15]. Of the birth outcomes, congenital anomalies (CAs) represent an important topic that is investigated worldwide. Major CAs are estimated to be diagnosed in 2%–4% of births [16]. CAs are considered a major cause of fetal death, infant mortality and morbidity, and long-term disability [17,18]. CAs are diseases with a high impact on affected individuals, their families, and the community in terms of quality of life and healthcare service needs [19]. The etiology of most CAs is multifactorial. WHO estimated that 5% (range 1%–10%) of CAs are attributable to environmental causes [20] and the gene-environment interaction can contribute to the causation process [21].Population-based registries of CAs are the main source of CA prevalence data [21]. CA registries are tools used for identifying genetic and teratogenic exposures, as well as evaluating and planning health care services and prevention policies [19]. CA registries are also effective tools for epidemiological assessments in polluted areas [22]. Research and monitoring focused on the potential environmental causes of CAs provide useful information aimed at generating hypotheses in order to identify teratogens. Following a feasibility assessment on the epidemiological study of adverse reproductive outcomes in small polluted areas, the Joint Action EUROCAT 2011–2013 reported that, “currently, links between reproductive health and contaminated sites are studied in a non-extensive and non-integrated way in Europe” [22]. Studying the impact of environmental causes on CAs should be based on an epidemiological approach and is a very complex field of investigation [16,21]. The association of CAs with industrially CSs has not been adequately investigated. A systematic evaluation of the epidemiological evidence reported the association between CAs and exposure to landfills, petrochemical plants, and/or refineries as being limited, and the association with other industrial sources as inadequate or not evaluable [23]. Few analytical epidemiological studies aimed at investigating selected CAs in industrial areas have been performed [24,25,26] and no evidence of association or only slight association in specific subgroups were detected. Limited to the exposure to air pollution, evidence of association has been reported in two reviews, particularly in relation to specific congenital heart defects [27,28]. A few epidemiological studies, including geographic studies, have investigated the association of CAs with exposure to waste disposals, but results are not consistent, and the evidence of the association is evaluated as limited [29,30,31,32,33,34,35].The aim of our study was to estimate the risk of congenital anomalies in various contaminated sites in Italy. To our knowledge, this is the first multi-site study performed in Europe on the risk of CAs in different types of industrially contaminated areas.Italian CSs have been defined by Ministerial decrees and identified for remediation following a documented contamination with a potential health impact. They are recognized as remediation sites of national interest because of the relevance of the pollution considering health, environmental, and social criteria. The contamination is the result of direct emissions, or indirect releases from waste disposals, controlled or uncontrolled. These areas are characterized by various types of industrial activities and have been classified as: chemicals, petrochemicals and refineries, steel plants, power plants, mines and/or quarries, harbor areas, asbestos or other mineral fibers, landfills, and incinerators [11,12,13,14,23]. CSs have different levels of environmental contamination and information on specific chemical contaminants is not available for all sites [36]. Legislative decrees provide information about the type of industrial activity; in the same CS, multiple types of current and past activities can be present. The decrees have also defined the CS boundaries, which are constituted by a municipality or by aggregates of municipalities [23].In our study, 13 CSs covered by existing CA registries were considered. We used data collected by the following Italian population-based registries: the Congenital Malformations Registry of Mantua; IMER-Emilia Romagna Registry of Congenital Malformations; the Tuscany Registry of Congenital Defects; Birth Defects Registry of Campania; and the Congenital Malformations Registry of Sicily.Data included cases with CAs among live births (up to 1 year of age), fetal deaths (with gestational age of 20 weeks or more), and terminations of pregnancy for fetal anomaly. For the CSs in Sicily and Apulia, since the CA registries were set up recently, a specific algorithm for the detection of live birth cases with CAs by hospital discharge records was used [37,38]. This algorithm filters the hospital discharge record data with ICD9 codes 740-759, applying a priori criteria in order to classify individual records into three groups: “certain CA”, “uncertain CA”, “no CA”. In our study, we used only those cases with CA classified as certain by the algorithm.The study period ranged from 1992 to 2012, with some differences between registries depending on the availability of data (Table 1). The number of births by municipality of mother’s residence and year of birth were extracted from the National Institute of Statistics [39].We analyzed the total CAs and seven subgroups of CAs, defined according to the classification used by EUROCAT [40]. Isolated minor CAs were excluded according to the EUROCAT definitions. Cases with multiple anomalies were considered as a single case in the definition of total CAs. The following subgroups of CAs with relative ICD10-BPA codes were analyzed: nervous system (Q00–Q07), congenital heart defects (Q20–Q26), oro-facial clefts (Q35–Q37), digestive system (Q38–Q45, Q790), genital (Q50–Q52, Q54–Q56), urinary system (Q60–Q64, Q794), and limb (Q65–Q74).For each CS, the number of observed cases for each subgroup of CAs was compared to the number of expected cases. Expected cases were calculated using the prevalence in the total area covered by the corresponding registry. We calculated the observed/expected ratio (O/E) with a 90% confidence interval (90% CI). We used a 90% CI in order to reduce the critical use of the CI as a surrogate of the hypothesis test [41] and in line with the previous studies on cancer, mortality and hospitalization performed in the same study areas. For the CSs with petrochemicals and/or refineries, and waste landfills, we calculated a pooled estimate of the risk of the total and the most frequent CA subgroups. The pooled estimates were calculated using a random effects meta-analysis applied to the single estimates calculated for each CS.Data analysis was performed using SAS version 9.2 (SAS Institute Inc., Cary, NC, USA) and STATA version 13 (StataCorp LP, College Station, TX, USA).The study area is represented by 13 CSs covered by regional registries of CAs. Table 1 reports the sources of environmental exposure included in each CS. The total number of resident cases with CAs during the study period was 7085 out of 288,184 births, thus accounting for a prevalence of 245.8 per 10,000. In the reference areas, 51,501 cases with CAs were registered during the same time period out of 2,368,031 total births, accounting for a prevalence of 217.48 per 10,000. The number of cases of total CAs was higher than the expected value in the following CSs: Laghi di Mantova (O/E 1.15; 90% CI 1.03–1.29), Livorno (O/E 1.33; 90% CI 1.25–1.41), Piombino (O/E 1.51; 90% CI 1.31–1.73), Manfredonia (O/E 1.36; 90% CI 1.22–1.51), and Taranto (O/E 1.10; 90% CI 1.02–1.18) (Table 2).Considering the subgroup of CAs, several excesses were observed in some CSs:
+      
+        
+          in Laghi di Mantova for congenital heart defects (CHD) (O/E 1.37; 90% CI 1.15–1.62), digestive system (O/E 1.80; 90% CI 1.16–2.67), and genital (O/E 1.54; 90% CI 1.07–2.14);
+        
+        
+          in Livorno for CHD (O/E 1.56; 90% CI 1.42–1.72), limb (O/E 1.60; 90% CI 1.37–1.85), and genital (O/E 1.44; 90% CI 1.19–1.73);
+        
+        
+          in Manfredonia for CHD (O/E 1.55; 90% CI 1.34–1.79), digestive system (O/E 1.91; 90% CI 1.30–2.70), and urinary system (O/E 1.62; 90% CI 1.14–2.25);
+        
+        
+          in Gela for CHD (O/E 1.22; 90% CI 1.02–1.45), genital (O/E 2.36; 90% CI 1.73–3.15), and urinary system (O/E 3.21; 90% CI 2.55–3.99).
+        
+      in Laghi di Mantova for congenital heart defects (CHD) (O/E 1.37; 90% CI 1.15–1.62), digestive system (O/E 1.80; 90% CI 1.16–2.67), and genital (O/E 1.54; 90% CI 1.07–2.14);in Livorno for CHD (O/E 1.56; 90% CI 1.42–1.72), limb (O/E 1.60; 90% CI 1.37–1.85), and genital (O/E 1.44; 90% CI 1.19–1.73);in Manfredonia for CHD (O/E 1.55; 90% CI 1.34–1.79), digestive system (O/E 1.91; 90% CI 1.30–2.70), and urinary system (O/E 1.62; 90% CI 1.14–2.25);in Gela for CHD (O/E 1.22; 90% CI 1.02–1.45), genital (O/E 2.36; 90% CI 1.73–3.15), and urinary system (O/E 3.21; 90% CI 2.55–3.99).The CA subgroups for which the highest number of excesses were observed were genital (in 6 CSs) and CHD (in 5 CSs).For CSs characterized by the presence of landfills, petrochemical, and refinery plants, Table 3 reports the pooled estimates of total CAs and selected CA subgroups. For the pool of seven CSs including petrochemical and/or refinery plants, the pooled risk estimate was 1.10 (90% CI 1.01–1.19) for total CAs, 1.15 (90% CI 1.00–1.30) for CHD, and 1.24 (90% CI 1.02–1.47) for genital. The pooled estimate in the 11 CSs containing a landfill was 1.07 (90% CI 1.02–1.13) for total CAs, and 1.08 (90% CI 1.00–1.16) for CHD.The health impact assessment of industrially polluted areas is important for improving scientific knowledge and public health decision-making. It is a very complex process due to the multiple and heterogeneous sources of pollution, the role of other non-environmental risk factors, and the multifactorial etiology of the diseases [7]. We compared the prevalence of CAs in populations residing in the neighborhood of polluted areas with larger reference areas. We analyzed data from the CA registries. The disease registries are considered to be the most accurate data source to provide a better definition of health outcomes also related to polluted areas [11]. The CA prevalence estimates are being highly variable across the different registries due to different diagnostic practices and methods of gathering and coding data [21]. In our study, the prevalence of CAs in each CS was compared with the prevalence in the region including the CS. Thus, we used the same data sources for the study area and the reference area, thus avoiding bias due to differences in registration activities.In some study areas, we used hospital discharge data filtered by applying a specific algorithm [37]. In a sensitivity study aimed at testing and validating this algorithm, a positive predictive value of “certain cases” equal to 93.5% was observed [38]. The algorithm produces a valid estimation of CA cases, but limited to live birth cases. The terminations of pregnancy due to fetal anomaly are estimated in about 17% of the total cases [42]. We performed a sensitivity analysis excluding FD and TOPFA cases in the study areas where all the types of events were collected. The results were very similar and some slight differences were observed only for the subgroup of the nervous system. This result was expected, as the percentage of TOPFA varies between CA subgroups, with higher values for CAs of the nervous system and of the abdominal wall, in addition to chromosomal anomalies [43]. Thus, for the subgroup of nervous system, the results relating to areas where only live births cases were considered should be interpreted with caution.For some CSs, a high risk of CAs emerged from the data analysis: in Laghi di Mantova, Livorno, Gela, and Manfredonia, excesses for several subgroups of CAs were observed. In some of these areas, the occurrence of CAs had been investigated in previous epidemiological studies. A study on CAs performed in the CS of Laghi di Mantova from 2002 to 2006 detected an overall risk of 1.47 (95% CI 0.77–1.82), using the surrounding municipalities as reference area. In the CS of Gela, excesses of genital and urinary CAs were detected by two epidemiological studies [44,45]; in particular, the rates of hypospadias (56.7 and 46.7 per 10,000 respectively) were more than double compared to Italian and European references. More recently, an even higher prevalence of hypospadias (88.5 per 10,000) a 3.5-fold excess compared to the reference, was detected using data from the Congenital Malformations Registry of Sicily [46].We also estimated risks considering the pool of CSs including landfills, petrochemical, and/or refineries. In order to limit the variability among the single estimates, we used a meta-analytical approach. Pooled risk estimates considering the CSs with petrochemical and/or refineries showed excesses for total, heart, and genital CAs. Since many CSs contain multiple sources of exposure other than those selected, the pooled estimates should be interpreted with caution. For the pool of eleven CSs including waste landfills, the meta-analytical estimate of risk showed excesses of total CAs and CHD.The two CSs in Campania, namely Area Litorale Vesuviano and Litorale Domizio Flegreo e Agro Aversano, are large areas characterized by the widespread presence of heterogeneous waste disposal sites (controlled landfills, illegal dumps, illegal and uncontrolled waste disposal sites with hazardous waste). Epidemiological studies at municipality level performed in these areas have detected excesses for various CAs [47,48,49]. In our study, no excesses of CAs were observed in the two CSs. The wide geographical and demographic dimension of these areas, the heterogeneity of the sources of exposure, and the limitation due to the ecological design, need to be considered in the interpretation of the results.We used an approach previously adopted for mortality, hospitalization, and cancer incidence studies in Italian CSs [11,12,13,14]. The methodology used in these studies is defined by WHO as a first level investigation at the population level within the wider framework of a funnel approach [7]. The ecological approach at the minimum aggregation level of the municipality determines the known limitations of this kind of design, mainly the ecological fallacy. Despite the limitations, the ecological design is consistent with the aim of describing the health profile of the population in industrially contaminated areas through a gradual investigation [7], as well as providing preliminary information, which is useful for the evolution of the research, especially when prior knowledge and evidence is limited [50]. Ecological studies also play a major role in the investigation of associations of public health importance [51]. The ecological design did not enable us to consider possible individual confounders related to maternal and paternal risk factors. Population living in the neighborhood of the CSs have tendentially a more disadvantaged socio-economic condition. We did not provide risk estimates adjusted for the socio-economic deprivation index at the municipality level due to the limitation of representing the socio-economic status for municipalities with more than 10,000 inhabitants [52]. The unavailability of quantitative assessments of population exposure is a limitation shared by other studies in the SENTIERI project [14]. A more accurate characterization of the exposure is needed in order to better fit the epidemiological investigation in the CSs.As CAs are rare events, CAs were grouped in order to ensure an adequate statistical power. This was to hide a potential association with a specific anomaly, especially for subgroups with an etiological and pathogenetic heterogeneity [16]. Our study detected excesses in particular for the heart and genital subgroups. Further studies, with a sufficient number of cases, are needed to investigate the association of selected subgroups or single anomalies.The present study aimed to improve the description of the health profile of residing populations in CSs previously investigated through other health outcomes, most of which were long-term. The results do not support the potentially causal role of the risk of CAs in terms of pollution in contaminated areas; however, the detected risks suggest etiological hypotheses and support further studies.Finally, the study consolidates the recommendations that emerged as part of the European Joint Action EUROCAT, which identified CA registries as a primary information source for epidemiological surveillance in contaminated areas through a linkage with environmental pollution data [22].The results of this multi-site study identified areas with excesses of different subgroups of CAs. The results are useful for identifying priority areas for analytical investigations, aimed at etiological research and intervention studies. The study supports planning for primary prevention of CAs, thus mainly the remediation of contaminated areas. The results on CAs complete the epidemiological framework in areas where different environmental stressors produce different health effects. A multi-outcome approach (mortality, cancer incidence, hospitalization, and congenital anomalies) contributes to understanding the role of environmental factors in the health risk profile of populations residing in contaminated areas. Fetuses and pregnant women comprise vulnerable subgroups sensitive to environmental insults and require specific surveillance. Our findings reinforce the recommendation of using short-latency effect indicators, such as the prevalence of CAs, for the health surveillance of populations residing in contaminated areas. It would be desirable to expand the surveillance of CAs to a wider international context in populations living in proximity to hazardous areas.This work was performed within the framework of the “RISCRIPRO-SENTIERI” project, funded by the Italian Ministry of Health’s Project CCM 2012, aimed at describing the reproductive health profile of populations living in Italian Contaminated Sites. RiscRipro_Sentieri Working Group members: Gianni Astolfi, Silvia Baldacci, Fabrizio Bianchi, Elisa Calzolari, Pietro Carbone, Matteo Cassina, Achille Cernigliaro, Maurizio Clementi, Susanna Conti, Paolo Contiero, Gabriella Dardanoni, Francesca Gorini, Ivano Iavarone, Valerio Manno, Sonia Marrucci, Fabrizio Minichilli, Amanda Julie Neville, Roberto Pasetto, Anna Pierini, Federica Pieroni, Vanda Pironi, Paolo Ricci, Michele Santoro, Gioacchino Scarano, Salvatore Scondotto, Giovanna Tagliabue, Domenica Taruscio, Andrea Tittarelli.Michele Santoro, Fabrizio Minichilli, Anna Pierini, and Fabrizio Bianchi conceived and designed the study; Michele Santoro performed the statistical analysis and drafted the manuscript. Fabrizio Minichilli, Anna Pierini, and Gianni Astolfi contributed to the data analysis. Gianni Astolfi, Lucia Bisceglia, Pietro Carbone, Susanna Conti, Gabriella Dardanoni, Ivano Iavarone, Paolo Ricci and Gioacchino Scarano contributed to the study design, the acquisition of data, and the interpretation of the results. Fabrizio Bianchi critically revised the manuscript for important intellectual content. Finally, all authors read and approved the final manuscript.The authors declare no conflict of interest.Contaminated sites, sources of environmental exposure, registry of congenital anomalies, type of cases, study period, and number of births.C = production of chemical substance/s; P&R = petrochemical plant and/or refinery; S = steel industry; E = electric power plant; H = harbor area; A = asbestos/other mineral fibres; L = landfill; I = incinerator; LB = live births; FD = fetal deaths; TOPFA = termination of pregnancy for fetal anomaly.Number of cases (N), observed/expected ratio (O/E) with 90% confidence interval (90% CI) by contaminated site and subgroup of congenital anomalies.Pooled estimate (PE) with 90% confidence interval (90% CI) by source of exposure and subgroup of congenital anomalies.CS = contaminated site.

Med-MDPI/ijerph_2/ijerph-14-03-00293.txt

@@ -0,0 +1,19 @@
+Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).Stroke is a leading cause of disability in the United States and disproportionately affects minority populations. We sought to explore the quality of life in urban, minority stroke survivors through their own photos and narratives. Using the Photovoice method, seventeen stroke survivors were instructed to take pictures reflecting their experience living with and recovering from stroke. Key photographs were discussed in detail; participants brainstormed ways to improve their lives and presented their work in clinical and community sites. Group discussions were recorded, transcribed, and coded transcripts were reviewed with written narratives to identify themes. Participants conveyed recovery from stroke in three stages: learning to navigate the initial physical and emotional impact of the stroke; coping with newfound physical and emotional barriers; and long-term adaptation to physical impairment and/or chronic disease. Participants navigated this stage-based model to varying degrees of success and identified barriers and facilitators to this process. Barriers included limited access for disabled and limited healthy food choices unique to the urban setting; facilitators included presence of social support and community engagement. Using Photovoice, diverse stroke survivors were able to identify common challenges in adapting to life after stroke and important factors for recovery of quality of life.Stroke is one of the leading causes of disability in the United States and can result in a wide spectrum of effects on both physical and mental health. Recovery can be a long and difficult process for stroke survivors to navigate. A 2010 study of Northern Manhattan stroke survivors showed that functional independence was lost over a period of five years, independent of recurrent stroke and other risk factors [1]. Stroke survivors are also more likely to internalize and marginalize themselves socially from others [2].The risk of recurrent stroke disproportionately affects minority populations, with African-Americans twice as likely and Hispanics one and a half times as likely to experience recurrent stroke compared to Caucasians [3]. In addition, there is limited insight into the full degree of impact this disease has in minority populations and the barriers they may face in recovery.Photovoice is a qualitative research method that was developed to learn about personal experiences through the viewpoint of the study participant. Subjects are given cameras and are asked to present their own experience through photographic expression and narratives, guided by targeted questions [4]. This technique places data collection directly in the hands of the subject and allows them to capture, present, and narrate complex emotional and physical obstacles that may be otherwise difficult to verbalize. The results can then be used to highlight needs, inform, and influence policy to help effect social change.In the past, the Photovoice technique has been used to study the experiences of marginalized communities, including homeless single mothers in Detroit and HIV-positive youth in Africa [5,6]. In these studies, Photovoice provided its participants with the opportunity to display their own day-to-day experiences and reach an audience to which they may have been otherwise isolated. Using Photovoice, we aimed to develop a greater understanding of recovery from stroke through photos and narratives grounded in the perspective of the stroke survivor. Specifically, we aimed to examine quality of life of urban, minority stroke survivors and identify barriers and facilitators that affect emotional and physical recovery.Our team recruited participants from a cohort enrolled in a randomized controlled secondary stroke prevention trial: Prevent Return of All Inner City Strokes through Education (PRAISE) [7,8]. The inclusion criteria for the PRAISE trial were the following: age 40 and older and history of stroke or transient ischemic attack (TIA) within the past five years. Subjects were also required to be able to physically and cognitively participate in educational group sessions. For participation in Photovoice, patients were required to be able to operate a digital camera.Subjects already enrolled in the PRAISE trial were approached for recruitment in the Photovoice study after the initial baseline enrollment survey, or as they returned for follow-up interviews and data collection at 6-month and 12-month visits for the larger trial. We enrolled 17 study participants in four different Photovoice groups. Each group participated in three separate sessions held between November 2010 and July 2011. Of these, we provided 15 with digital cameras and two chose to use their own cameras. We transcribed written notes and recorded audio during each meeting. We allowed participants to keep the cameras as a gift, and presented those who used personal cameras with a gift card equivalent to the camera value ($50). The Mount Sinai Medical Center Institutional Review Board (IRB) approved the study (GCO# 02-0515 0001 03). Trained moderators held one session to introduce the Photovoice method, provide brief photography training, and asked participants to use photographs to answer a question: What has made it easier or harder to live your daily life, be part of your community, and prevent having another stroke? During the second session, participants presented their top five photos to the group using the SHOWeD method to explore their personal experiences: (1) What do you See here? (2) What is Happening here? (3) How does this relate to Our lives? (4) Why does this situation or concern exist? (5) What can we Do about it? [9] During the third session, participants presented their photos and shared narratives with the group.At the completion of the PRAISE trial in 2013, we administered follow-up phone calls to the Photovoice participants in order to obtain feedback focused on their experience within the Photovoice study; all were contacted with the exception of one who was deceased at the time of follow up.Using a grounded theory approach and content analysis method [10], data were reviewed and analyzed after each session and again after all sessions were complete. The content analysis method involved initial identification of meaningful text, then more in depth interpretation through coding, and categorization and identification of themes and subthemes [11]. Audio recordings from the second and third Photovoice sessions were transcribed and combined with notes from the sessions.We utilized iterative review of the transcripts, session notes, written text, and coder triangulation in order to thematically analyze the data. We created a master list of themes and codes that included deductive themes developed from the literature review and the interactive group sessions and inductive themes from the session notes, transcripts, and text accompanying the photographs [12]. Relevant themes were identified by grouping similar codes. Manual methods for analysis were used. Two authors (who acted as group session facilitators) independently coded transcripts using the master list and a third author (who was not present during group sessions) reviewed the coded transcripts and found an inter-rater agreement of 76%. The three authors then discussed and resolved any disagreements through consensus. Member checking with participants was conducted at the end of the third group session to validate themes that were deduced between sessions two and three. The study was IRB approved and we de-identified all documents (removing participant names and replacing them with numbers) and kept all files in password-protected folders protected by a firewall.Overall demographics of the participants can be seen in Table 1. The mean age was 64 years (±9.8) and mean time from stroke was 2.0 years (±1.5). Participants were mostly female (65%), and Black (65%). Utilizing modified Rankin scores collected from the PRAISE trial, it was found that 41% had moderate to moderately severe post-stroke disability; the remainder had slight or no disability. The score ranges from 0 to 6, where 0 is no disability and 6 is dead, and 5 is severe disability requiring continuous care [13].The photos and accompanying narratives varied in content but shared a common theme: a journey of recovery and adaptation to life after stroke. We present this major theme through a three-stage conceptual framework composed of the subthemes that emerged during the Photovoice sessions. As outlined in Figure 1, these stages include: (1) The initial stroke experience and reaction to the immediate effects of the stroke (acknowledgement versus avoidance); (2) coping strategies (integration versus isolation); and (3) long-term adaptation. Transitioning between the first and second stages was facilitated by acknowledgement of new challenges and hindered by avoidance of these same challenges. Transitioning between the second and third stages was facilitated by positive reflection on the personal experience and increased social integration, and was hindered by negative reflection and social isolation. Here, we explain each stage and transition of this framework as described by the participants, as well as the sub-themes that characterize each.During the group discussions (second session), nearly all participants shared the immediate emotional and physical impact of the stroke experience, focusing on the “How does this relate to our lives?” question of the SHOWeD method. Many felt a strong sense of shame and disappointment upon realizing the impact of their new disability. One younger woman described the initial impact, describing a photo of the entrance to her building:
+          
+            “The day I came home it took me almost half an hour to get to the front door to the lobby. I was crushed. I wanted to just cry, crawl under a rock or just be somewhere there was no one... It wasn’t until then that I realized my life was over…”
+          “The day I came home it took me almost half an hour to get to the front door to the lobby. I was crushed. I wanted to just cry, crawl under a rock or just be somewhere there was no one... It wasn’t until then that I realized my life was over…”After the initial stroke, most participants described a process of acknowledging the new emotional and physical barriers. Those able to cope with these barriers did so by either overcoming or accepting their newfound limitations. Others described a process of avoiding situations that would reveal their limitations to themselves and others, which was associated with sustained feelings of frustration.• Environmental barriersAs all participants lived in densely populated urban neighborhoods, many of their photos depicted local environmental features that made living with a physical disability difficult. During the group discussions, participants identified these barriers through photographs, highlighting a lack of public benches which discouraged walking (no place to sit and rest) and a lack of subway elevators, which limited transportation options. Local disrepair, such as cracks in sidewalks, further discouraged outdoor exercise, posing threats to physical safety. Some participants found these obstacles isolating and avoided going outside, while others engaged them as motivational tools. One emphasized,
+          
+            “This is an obstacle course... I found out you have to be twice as cautious and more tentative walking…and I didn’t want people to help me, I was sweatin’ bullets, but I had to take that challenge.”
+          “This is an obstacle course... I found out you have to be twice as cautious and more tentative walking…and I didn’t want people to help me, I was sweatin’ bullets, but I had to take that challenge.”The discussion and identification of environmental barriers revealed the common struggles that those who were physically disabled faced in navigating the urban environment, and fostered dialogue among the group to answer the “What can We Do about it?” question of the SHOWeD method.• Emotional barriersNearly all participants expressed difficulty in coping with emotional barriers and openly shared during the group discussions. Feelings of depression, social isolation, and frustration were frequently expressed in narratives. One woman compared herself to a single tree between two large buildings visible from her window; to her, the picture she took represented feeling trapped between her current disabled state and where she wanted to be (Figure 2).For some, avoidance of these emotional barriers delayed the coping process. An African American woman initially hid the extent of her disability from others, and internalized her struggle:
+          
+            “… I was helping a lot of people…because of their personality I didn’t want them to know I was sick. So I refused talking to them at all, because in my mind I felt that it was better that they be angry with me than to feel that I was sick.”
+          “… I was helping a lot of people…because of their personality I didn’t want them to know I was sick. So I refused talking to them at all, because in my mind I felt that it was better that they be angry with me than to feel that I was sick.”In contrast, those who described successfully coping with new physical and emotional challenges identified family, friends, and even pets as key facilitators of this process through pictures.After the initial coping stage, participants reflected on their pre- and post-stroke experiences and described milestones by which they measured their own progress in recovery. To convey the difficulty of learning to walk again, one 46-year-old African American woman photographed a wheelchair, “to remind me of how far I came. I used to depend on that wheelchair so much.” Similarly, a 62-year-old African American man photographed subway stairs, explaining that climbing them “was one of the first things I wanted to overcome, and it’s something that I chose to overcome and I did it.” (Figure 3)Positive reflection allowed many participants to re-evaluate their experiences as part of a coherent, purposeful journey. This perspective allowed some participants to accept their newfound limitations and adapt to new routines. After acknowledging her physical limitations, one survivor explained,
+          
+            “You know, I’m not going to be the same person and I don’t really care… I’m still who I am, I’m just a little bit awkward now… Sometimes you got to go through bad…in order [to] really appreciate who you are.”
+          “You know, I’m not going to be the same person and I don’t really care… I’m still who I am, I’m just a little bit awkward now… Sometimes you got to go through bad…in order [to] really appreciate who you are.”Aided by religion and/or spirituality, other participants integrated their negative experiences and limitations into a larger narrative of personal growth.Many participants identified an increase in social activity and bonds within families as a facilitator in the reflective process. One African-American female described a sense of self-worth and empowerment from giving back to the community through the distribution of homemade juices at a local gym. Another participant explained, “I don’t think that I’m different, but…my difference is gonna make a bigger impact on somebody else. I’m a walking business card, ‘hello, your cholesterol!’”However, several participants detailed a process of negative reflection accompanied by self-isolation and frustration, spending large amounts of time alone. One Hispanic female lived alone and described the negative emotions resulting from having outlived all her friends and family and shared feelings of helplessness and loss. Another participant shared a picture of a broken chair on the sidewalk and compared it to being a stroke survivor without any support (Figure 4).“I see this broken chair and I think about how many people in my position, who had a stroke or have an illness are broken, pushed aside, discarded… if I didn’t have my family and my friends, I could have been sent to a nursing home somewhere and sit in a corner and nobody ever comes to visit, and that’s the most disheartening thing in the world.”Ultimately, a change in perspective facilitated by positive reflection allowed many survivors to adapt to a physically and emotionally challenging environment. This allowed them to develop positive habits and lifestyle improvements. This was most often addressed during group discussions when answering “What can We Do about it?”, while discussing pictures that represented challenges faced.• Adapting to disabilityAlmost all participants shared pictures of assistive devices, such as wheelchairs, shower chairs, and canes, to emphasize how a once-easy task became very difficult after the stroke. Participants framed transitions between devices (for example, from a wheelchair to a cane) as milestones and symbols of progress. Each group emphasized the importance of learning to do “basic” self-care activities to reestablish a sense of independence. One participant recalled, “If you want to take a shower or go outside, you don’t have to wait for anyone else to do it.” Participants described adapting to new modes of transportation. Discussing a photograph of subway stairs, one participant recalled, “I think now, I could do it by myself, but I’m nervous because there are so many stairs and you have to walk so much… I used to take the subway everywhere and now I can’t as much, now I rely on buses.”• Diet and health awarenessSeveral participants described new focus on adhering to a daily medication which some initially resisted, but ultimately adapted. Describing a picture of a plate filled with pills, one 57-year-old African American man stated,
+          
+            “[The] negative is that, that’s it for life…we got to take that forever now; we don’t give it up just cause we feel better…the pills are our life… It is a positive that you’re still alive…you are living.” (Figure 5)
+          “[The] negative is that, that’s it for life…we got to take that forever now; we don’t give it up just cause we feel better…the pills are our life… It is a positive that you’re still alive…you are living.” (Figure 5)In addition, some participants identified difficulties with making healthy food choices in their neighborhood. One 62-year-old African-American male photographed snacks at a corner store, “…As soon as you walk in…they have this experience of salt…a wall full of salt…[but] with the knowledge that it will hurt you, that overcomes the desire to have it.” (Figure 6)• HobbiesHobbies served as sources of motivation and physical rehabilitation; one Hispanic female became an urban gardener:
+          
+            “If you had issues with yourself or with not accomplishing, you can get into gardening and forget all the negativity… When I garden I talk to myself…it’s almost like you can look at yourself and see that you’re coming along, like each day you get better.”
+          “If you had issues with yourself or with not accomplishing, you can get into gardening and forget all the negativity… When I garden I talk to myself…it’s almost like you can look at yourself and see that you’re coming along, like each day you get better.”Aside from facilitating physical recovery, gardening and other hobbies, such as knitting, also provided survivors with opportunities to reflect on their own physical and emotional progress. In this way, hobby-based adaptation facilitated continued reflection.• Photovoice exhibitThe results of the Photovoice sessions culminated in a gallery show highlighting pictures and allowing participants to discuss their photographs and experiences at various community venues including a premier and presentation at a legislative breakfast including ten local and state elected officials and representatives. In response to the display, officials agreed to have the local community board take up the issues highlighted such as traffic light timing, road repairs, and increasing benches on commonly used streets so people could rest along their routes. They also invited the community to display future Photovoice exhibits in the state office building, and they did, eventually displaying projects in regards to women with gestational diabetes, and people with diabetes and visual impairment. • Follow-up interviewAll participants expressed a therapeutic value to the act of taking photographs and sharing them in a group setting. For these participants, taking part in the Photovoice project itself helped facilitate their emotional recovery. Hearing other survivors’ stories allowed them to feel less isolated in their experience. The social interactions required of group discussions gave them confidence to participate more in their communities and make new connections. One elderly African-American participant, who initially spoke of feeling very alone, became more involved in local community theater after completing the study. Another participant who initially expressed feelings of isolation stated that she developed a friendship with another participant in the group that lasted long after the sessions were over.Using the multifaceted technique of Photovoice, we investigated the effects of stroke on quality of life in an urban, predominantly minority population. Our study revealed a largely consistent model of major themes of the stroke experience, coping, and adaptation emerging as stages in recovery. While prior qualitative studies have produced similar models, this is the first to utilize Photovoice in a predominantly minority population of stroke survivors living in an urban area [2,14,15,16].Individual participants navigated the immediate emotional and physical effects of stroke in different ways, but those who described a strong support network of friends and family were able to transition from initial stages of coping to long-term adaptation more easily. Isolation resulted from a failure to acknowledge physical and emotional barriers, resulting in extended coping mechanisms. Our findings are consistent with prior studies utilizing different techniques that identified social relationships as an important facilitator of physical and emotional recovery. One meta-analysis of 25 qualitative studies of stroke survivors identified social isolation as a major theme and found the majority of participants experienced feelings of increasing social withdrawal [16]. In a prior focus group study on quality of life after stroke, discussions focused on changes in social relationships post-stroke and frustration when there was a lack of social support [2].Participants also identified environmental obstacles that uniquely affect stroke survivors living in an urban setting, such as sidewalk cracks and potholes and limited public transportation options that prevent independence. This is consistent with results from a prior study that also found lack of physical access was one of the most frequently documented barriers to recovery [17].Several studies have explored the positive impact of creative therapy in marginalized populations, including war veterans with Post-Traumatic Stress Disorder and trauma victims; a previous Photovoice study on patients with aphasia highlighted the positive effects of the experience on its participants [18,19,20,21,22,23]. We found the Photovoice technique itself facilitated reflection on the stroke experience and the challenges in recovery. Given the positive impact of self-reflection that was seen through Photovoice, we advocate for its use as a beneficial intervention in the stroke recovery process.Photovoice has capabilities for empowerment and self-advocacy above those found with traditional qualitative methods [24]. Several participants expressed a sense of purpose and reported feeling empowered to help others in different ways after participating; some reported increased community volunteering, while others used their stories as educational tools.This technique also improved the relationship between the researchers and the community by allowing active engagement of the participants in the process; therefore, making the study a collaborative effort. The method was utilized in a partnership with a community action board that was focused more on action than process. For the PRAISE trial, board members reviewed every step of the process and provided input, and worked with researchers to develop the peer-education groups which were the essence of the intervention. The Photovoice intervention itself helped build a community and reduce isolation by bringing together stroke survivors from a given community. Recognizing the need to help people access positive reflection, increase social interaction, and reduce negative reflection and isolation, we changed the planned intervention in the PRAISE trial. To address isolation, we developed an introductory exercise in which we had people break into pairs, share a time when they accomplished something and share each other’s stories with the rest of the group, and we included simple exercises the group practiced regularly (i.e., ones that could be done with limited mobility or in wheelchairs) [7]. We also decided to explore whether stroke survivors were experiencing post-traumatic stress disorder (PTSD) and added specific PTSD scales to the surveys in the PRAISE trial; we were the first to find an association between stroke and PTSD and are currently developing interventions to address this [25,26].Photovoice helped to identify areas in which urban, minority stroke survivors could benefit from targeted interventions. Examples include increasing outlets for social support, improving accessibility to public transportation, and providing healthier food choices in urban areas. With our participants, the active engagement fostered social interactions and facilitated empowerment that allowed participants to advocate for changes to barriers that stroke survivors face in an urban setting. In addition, the technique allowed for collaboration between researchers and participants that furthered the overall goals of the study and allowed for innovation of additional research questions. Photovoice is useful as both a tool for research and for self-reflection. Upon completion of the Photovoice study, participants expressed a sense of accomplishment and desire to share their experiences to a wider audience. This echoes the importance of reflection and socialization in transforming a potentially isolating experience into one of growth and self-realization; it illuminates the unique positive attributes this method provides in its ability to enlighten researchers and participants at the same time.This work was supported by the National Institute on Minority Health and Health Disparities (5P60MD000270) and the National Center for Research Resources (UL1TR000067) of the National Institutes of Health. The study sponsor did not have any role in study design, collection, analysis, or interpretation of data, writing, or decision for publication. Special thanks to Shannon Gearheart for her contribution to idea development and data collection. Research reported in this publication was supported by the National Institute on Minority Health and Health Disparities of the National Institutes of Health under Award Number P20MD006899. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.R.B. and C.R.H. conceived and designed the experiments. B.K., R.N. and J.Z.G. conducted the group interview sessions and contributed to the manuscript. R.B., B.K., K.F. and R.N. analyzed the data. R.B., B.K. and C.R.H. wrote the manuscript.The authors declare no conflict of interest. The founding sponsors had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, and in the decision to publish the results.Conceptual model.The tree.Stairs.The broken chair.Plate of pills.The wall of salt.Photovoice participant demographics (n = 17)TIA—Transient Ischemic Attack.

Med-MDPI/ijerph_2/ijerph-14-03-00294.txt

@@ -0,0 +1 @@
+Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).Integration of health education and welfare services in primary care systems is a key strategy to solve the multiple determinants of chronic diseases, such as Human Immunodeficiency Virus Infection and Acquired Immune Deficiency Syndrome (HIV/AIDS). However, there is a scarcity of conceptual models from which to build integration strategies. We provide a model based on cross-sectional data from 168 Community Health Agents, 62 nurses, and 32 physicians in two municipalities in Brazil’s Unified Health System (UHS). The outcome, service integration, comprised HIV education, community activities (e.g., health walks and workshops), and documentation services (e.g., obtainment of working papers and birth certificates). Predictors included individual factors (provider confidence, knowledge/skills, perseverance, efficacy); job characteristics (interprofessional collaboration, work-autonomy, decision-making autonomy, skill variety); and organizational factors (work conditions and work resources). Structural equation modeling was used to identify factors associated with service integration. Knowledge and skills, skill variety, confidence, and perseverance predicted greater integration of HIV education alongside community activities and documentation services. Job characteristics and organizational factors did not predict integration. Our study offers an explanatory model that can be adapted to examine other variables that may influence integration of different services in global primary healthcare systems. Findings suggest that practitioner trainings to improve integration should focus on cognitive constructs—confidence, perseverance, knowledge, and skills.Across the globe, Human Immunodeficiency Virus Infection and Acquired Immune Deficiency Syndrome (HIV/AIDS) continues to disproportionately burden low-income ethnic/racial groups and sexual minorities who face myriad other chronic diseases [1,2,3]. In 2015, 830,000 people were living with HIV in Brazil, the prevalence for adults aged 15 to 49 was 0.6%, with the highest rates in populations facing low educational attainment and economic inequality [4]. HIV risk behaviors arise within the context of socioeconomic determinants of health, such as the physical, familial, cultural, organizational, economic, policy/legal, and social environments in which those affected live [5,6]. Given the scope of these determinants of health, governments worldwide, including Brazil, have developed workforces to provide HIV/AIDS prevention education while also assisting low-income individuals to obtain documentation (e.g., birth certificates, working papers, identity cards, etc.) for receiving primary care and welfare services (e.g., provision of nutritious food, etc.) [7,8].Lack of consensus on a definition for service integration (e.g., in this case, the integration of HIV education with community activities and documentation services) has delayed research on the inclusion of different services into primary care [9,10]. “Service integration” by various authors has been used to refer to integrated care, continuity of care, coordinated care, managed care, comprehensive care, and patient-centered care [11]. There have been a number of studies that have confirmed the feasibility of integrating different services with primary care, such as mental health and drug treatment [12,13]. Integration has been shown to be a cost-effective tool [14] with the potential to improve disease treatment response, increase the likelihood of remission [15,16] and access to services [17], and improve patient well-being [17]. What remains unknown is how best to integrate HIV education with welfare-related services within primary care systems [10].This paper advances the literature by providing an explanatory framework within which to build further research on service integration. We define “service integration” as the coordinated behaviors of practitioners in the Estratégia Saúde da Família (FHS; Family Health Strategy), the primary care program of Brazil’s Sistema Único de Saúde (UHS; Unified Health System), which offers multiple services to individuals at risk for multiple disorders and diseases. The UHS employs physicians, nurses, and Agentes Comunitários da Saúde in Portuguese or in English as community health agents (CHAs), all charged with the integration of welfare and public health services within primary-care units across the country [18]. The purpose of our study was to identify significant predictors of service integration at the practitioner and organizational levels along with practitioners’ job characteristics. The authors believe that the framework can be adopted by other researchers to study the integration of welfare services with other chronic diseases, such as tuberculosis (TB), cancer, etc.The FHS has institutionalized the provision of health and welfare services to comply with Brazil’s 1988 federal constitution and the 1990 Lei Orgânica da Saúde (Organic Health Law) [19]. The FHS offers free primary health care through a decentralized system, deploying interprofessional teams, each comprised of one physician, a nurse, two or three nurse assistants, and four to six CHAs. Each team provides care to between 800 and 1000 families, representing some 4000 individuals [20]. Collectively, the FHS teams provide primary care to more than 60 million Brazilian citizens [19]. According to Brazil’s National Standards, physicians, nurses, and CHAs, are involved in the provision of services in three key domains, as follows. HIV education comprises teaching community residents about how to access HIV testing and free condoms, how to use condoms, and how to practice safe needle exchange [21]. The FHS is involved in helping patients obtain documentation, such as birth certificates, working papers, identity cards, etc., needed for receiving primary care and welfare services. One must be a Brazilian citizen to qualify for services, for example, the cesta básica (food basket), a program that offers nutritious food for families living below the poverty line. Documentation is a crucial issue in a universal healthcare system to track Brazil’s most prevalent health problems and prioritize interventions to address them [20,22]. FHS teams also involve residents in sociocultural activities, including education campaigns to improve HIV knowledge, address stigma, and debunk HIV myths [23].Knowledge base, skills, perseverance, confidence, perception of team efficacy, and familiarity with the communities all influence how practitioners integrate different services. Whereas nurses and physicians possess biomedical knowledge (i.e., etiology and epidemiology of diseases) [24], CHAs (or Community Health Workers (CHWs), as known globally) are trained in basic medical practices and are specifically hired to use their lived experiential knowledge to impart health-promoting behaviors [25]. CHAs may acquire their lived experiential knowledge through having greater  familiarity with communities’ lifestyles, traditions, and habits than physicians and nurses are intended to have [26]. Physicians and nurses often lack training in building partnerships with consumers to stimulate change within local communities [22,27]. Nonetheless, CHAs may assist medical practitioners with socio-emotional counseling skills [25]. The process of sharing their skills and knowledge can enhance practitioners’ confidence to provide both social and medical services [28].Work conditions and available resources, such as office space, medical supplies, and data management systems, all comprise organizational factors that influence how different practitioners in a health team contribute to integrating services [29,30]. FHS uses discrete, formalized assignments for staff, while encouraging all practitioners to work together to combine different services. Though the roles and responsibilities of each FHS practitioner might be specific, FHS practitioners are not precluded from providing those services. While physicians are charged with developing care plans, nurses perform nursing care, request laboratory tests, dispense medications, supervise CHAs, and engage in health promotion activities [31]. CHAs collect and manage household data on births, deaths, disease incidence, and immunization status of children—in addition to offering health promotion activities [32,33]. Therefore, FHS practitioners integrate services by pooling their knowledge, skills, and cognitive competencies discussed above [34].Several job-related variables, practitioners’ assignments and their discretion to perform such assignments, may influence their decisions concerning when and how to integrate different services [35,36]. For example, federal guidelines encourage FHS practitioners to account for the input of consumers as co-creators in their own health care [37,38]. Interprofessional collaboration, another important influence in integration, is characterized by practitioners working side by side, applying diverse knowledge/solutions to health issues. Other job-related factors include “practitioner autonomy” and “skill variety” [35]. Autonomy gives practitioners confidence that they are able to deal immediately with a consumer’s health issues [39]. Skill variety is a concept that arises from practitioner autonomy; in order for practitioners to offer consumers a wider range of services, they must possess diverse skill sets, such as the ability to integrate health and welfare services into their practices [40].CHAs are hired from the communities in which they live, but most nurses and physicians reside outside the poor communities that they serve [20]. Yet the practitioners’ socio-geographic relationships to the communities they serve has received little attention in the literature. In this study, we examined whether or not practitioners living closer to the communities they served were better able to tailor prevention services to the needs of those communities.Our theoretical framework (Figure 1) reflects concepts of Cognitive-Behavioral Theory, Job Characteristics Theory [41], Modern Organizational Development Theory [42], and Structural Contingency Theory [43,44]. Cognitive Behavioral Theory reflects intrapersonal factors that influence practitioners’ service integration. Job Characteristics Theory suggests that practitioners’ integration of services can be influenced by their prescribed roles, by expected organizational norms, and by the presence or absence of organizational structure and of human and material resources. Practitioners are also influenced by their organization’s size/capacities, climate, and culture. Organizational theories, such as modern organizational developmental theory and structural contingency theory, explain how practitioners behave within community-based organizations, such as the FHS, that are systematized around provision of medical services.This study was approved by the Institutional Review Boards of Columbia University, New York, NY, USA, (IRB-AAAC4674 (Y2M01)) and Universidade Católica, Rio de Janeiro, RJ, Brazil. This study followed a community-engaged strategy that included university partners, health care administrators, and CHAs who developed the study’s aims and methods and plans for disseminating findings [45]. Written consent was sought from all study participants. Collaborating with FHS administrators helped us to establish study aims that would produce results that would be beneficial to the communities in which the study would take place [46].We enrolled 168 CHAs, 62 nurses, and 32 physicians from 30 Unidades Básicas de Saúde (UBS), also known as community-based primary health care clinics, in two Brazilian municipalities. Through the UBS, FHS teams provide primary health care services. Each UBS had at least one physician (range = 1–2), one nurse (range = 1–5), and one CHA (range = 1–23). The average length of employment was 40 months (SD = 31; range = 4–156). Participation was voluntary. Brazil’s policy on research does not permit financial incentives; however, refreshments were provided during data collection.Eight master’s level Brazilian interviewers were trained in research methods and procedures, and administered the survey using password-protected mobile computers. Data were downloaded into a password-protected database, DatStat Illume 4.6 (DatStat, Seattle, WA, USA) [47]. All data were kept in password-secured computer files, to which only relevant research personnel had access. There was no documentation linking respondents assigned ID numbers to the UBS for which they worked. The survey was administered verbally and lasted from 45 to 75 min. Approximately 85% of staff from all clinics participated.Survey questions addressed participants’ perceptions of and attitudes toward their knowledge, skills and confidence, and job characteristics. We piloted the survey with 42 practitioners. CHAs showed difficulty understanding questions that tapped opinions and attitudes toward scientific research, and physicians found the survey too long. We used this input to modify the survey. Survey questions were then translated from Portuguese to English and iteratively back-translated into Portuguese [48] for accuracy.All predictor variables appear in Table 1. The outcome, service integration, was measured by combining three dichotomous (yes/no) variables: (1) “I teach consumers how to prevent HIV and AIDS”; (2) “I help my consumers obtain documents, such as voter registration, working papers, and birth certificates”; and (3) “I help my consumers get involved in community activities, such as health walks and workshops.” Though each of the three integration variables may align more closely with the roles and job descriptions of one or another FHS provider, these services can be provided by all providers. Moreover, our measure of “integration” follows our conceptual framework in that integration is conceptualized as the combination of different services performed by different providers.Demographics included “age” measured as a continuous variable; “gender” included male or female; “race” included Black, White, and Pardo (Pardo refers to mixed races, such as mulattos [49]).We summarized descriptive frequencies of demographics and job context variables. We investigated influences on FHS teams’ provision of HIV/AIDS education along with their involvement of consumers in community-level activities and helping them to obtain documentation/registration, “service integration,” as a latent variable underlying three measures: HIV/AIDS education; community activities; and documentation services (Figure 1). We used MPlus 7.1 software (Muthén & Muthén, Los Angeles, CA, USA) to fit the structural equation model following the form on Figure 1 using the weighted least squares estimation appropriate for categorical outcomes. The χ2 goodness of fit test, the ratio of the χ2 to the degrees of freedom, and the root mean square error of approximation (RMSEA) were used to assess fit. An RMSEA value of ≤0.05 signifies a good fit [51], and as the χ2 statistic tends to be over-sensitive to minor misfit, the ratio of the χ2 to the d.f (degrees of freedom) <2 is often considered good fit [52]. The percent of variability in the latent service integration variable explained by all the predictors in the model was used to quantify explanatory power of the model. Modification indices (MIs) were tested to examine any possible direct effects between the predictors and the three observed measures of service integration (above and beyond the effects through the latent service integration variable).Table 2 shows that most respondents were CHAs (n = 169; 64%); 62 were nurses (24%); and 31 were physicians (12%). The highest proportion of respondents identified as Pardo (n = 123; 47%); 82 were White (32%); and 54 (21%) were Black. The majority were female (n = 214; 82%). Average age was 33.60 (SD = 9.99, range = 20–70). The majority of practitioners (n = 175; 67%) reported one to five years of working with FHS. 73% of CHAs and 60% of nurses had 1–5 years of experience. 46% of physicians reported spending 1–5 years with the FHS. The highest proportion of practitioners reported that they had ≤250 cases per month (n = 133; 51%); 91 (35%) reported their caseload was >501 cases per month, and 38 practitioners stated that their caseload was between 251 and 500 cases per month (14%). Seventy-seven percent of CHAs had ≤250 cases per month; 95% of nurses and 94% of physicians reported having >501 cases per month. Half of the practitioners reported that their commute to work was 0–10 min (n = 129; 50%); 88 (34%) stated their commute ranged from 11–30 min; and 43 (16%) over 30 min. Sixty-seven percent of CHAs reported that their length of commute was 0–10 min, 48% of nurses reported commutes of 11–30 min, and 55% of physicians reported commutes over 30 min. Ninety-one percent of CHAs said “yes” to the questions asking if they lived in proximity to their work; however, 55% of nurses and 84% of physicians answered “no”.Of the 262 practitioners, 217 (83%) reported offering HIV-prevention services; 212 (81%) engaged in community mobilization; and 118 (41%) engaged in documentation services. Of the 169 CHAs, 142 (83%) reported offering HIV-prevention services; 138 (82%) reported mobilizing communities; and 85 (51%) stated they did documentation services. Fifty-three of 62 nurses reported offering HIV-prevention services (85%); 51 (81%) stated they mobilized communities; and 14 (23%) engaged in documentation services. Of 31 physicians, 22 (71%) offered HIV prevention services; 24 (77%) mobilized communities; and 8 (26%) offered documentation services.Table 3 shows significant predictors of service integration. None of the demographic or job context variables were significant predictors of service integration, but practitioners with more experience trended toward higher service integration (B = 0.258; p = 0.060), and CHAs appear to be more involved in helping consumers obtain documentation/registrations.Among the individual level factors, confidence (B = 0.322; p = 0.020), knowledge and skills (B = 0.448; p = 0.006), and perseverance (B = 0.237; p = 0.036) had a significant positive effect on integration. Among job characteristics, skill variety had a positive and significant association with service integration (B = 0.355; p = 0.017). Work-methods autonomy (B = −0.222; p = 0.097) and decision-making autonomy (B = −0.237; p = 0.075) trended toward a decreased association with service integration. Organizational factors did not show significant associations with service integration. Perseverance was a strong predictor (MI = 9.345) of the provision of HIV-prevention services, even above its effect on overall service integration. The summary model fit indices for the structural equation model were χ2 = 35.9 with degrees of freedom = 46, and the RMSEA was 0.001, 90% Confidence Interval (CI) (0.000–0.024). An RMSEA value of ≤0.05 signifies a good fit. The factor loadings of the three measures of service integration were all significant: HIV-prevention services –0.46, civil registration –0.43, community mobilization –0.48. Overall, the predictors in the model explained 62% of the variability in service integration.Integrating HIV prevention while helping consumers to obtain documentation and participate in community activities appears to be commonplace among FHS teams. Results show that greater FHS experience may predict better integration because practitioners who maintain long-term relationships with FHS consumers are likely to know their needs well and thus integrate the most needed services. The longer the time of their employment in the FHS, the better practitioners will be able to identify individual consumer behavior patterns that might lead to disease transmission, and the faster they will act to curtail those behaviors. CHAs are closest to the consumers they serve geographically and have greater knowledge of their community’s lifestyles, traditions, and culture than do other members of the FHS team. In making household visits, CHAs study their patients vis-à-vis their sociocultural practices [53]. In doing so, CHAs are the ones best able to identify individual HIV risk behaviors. They appear to be consistently engaging residents in community activities to learn HIV prevention [26,54].Presumably, FHS team members’ awareness of one another’s expertise encourages them to consult one another about tasks or services needed for consumers and communities; this generates new knowledge for them. Such consultations ensure that services are integrated in accordance with consumers’ needs, and this, in turn, boosts consumer confidence. Service integration requires practitioners to use a combination of clinical, interpersonal, and advocacy skills, along with empathy and compassion toward consumers [18]. It has been established that physicians are given more autonomy than nonmedical practitioners, such as CHAs, [55], due to the perception that their medical and scientific knowledge is superior to the experiential knowledge of CHAs. Indeed, in Brazil, CHAs have reported having less autonomy in making decisions on how to integrate services [56]; nonetheless, our findings suggest that perhaps when motivated and with a sense of social closeness to the communities they serve, CHAs are able and willing to integrate services.Our results showed that no organizational factors predicted service integration. This may be due to practitioners’ perception in recognizing Brazil’s poorly managed health system. Their exposure on a daily basis to corruption, lack of governance, and internal bureaucracy, all of which delays the process of buying and delivering supplies and medications, could result in practitioners perseverance in dealing with limitations of a universal health system and in seeking support of FHS team members while delivering integrated services to consumers [57]. However, it is important to note that these results stem from a secondary analysis of FHS data; our choices of quantitative measures were thus limited. Other organizational variables, such as practitioners’ job security, remuneration, and job satisfaction, may have improved the explanatory power of our model; however, it is also important to note that no realistic empirical model or dataset could contain every variable that might influence service integration.Recall and information bias may have affected the reliability of this study’s findings. Practitioners who participated were asked to offer information about the previous six months; however, their attitudes might have changed during that time. Data were cross-sectional, making it impossible to determine causal associations; a longitudinal design would allow for a more comprehensive understanding of the associations between variables of interest. Our findings are not generalizable beyond the two municipalities where we collected data. Terms such as “HIV prevention services,” “work conditions,” and “work resources” were open to interpretation by respondents. The authors also identify that certain survey items, such as “I am able to make treatment plans which fit the needs and abilities of my patient”, may be double barreled and hence may have resulted in practitioners agreeing to the statement. We thus recommend further research to assess the best ways for measuring these variables, rephrasing survey items to avoid response bias, and include other variables that we did not examine. The authors also recognize that the survey instrument excluded questions on practitioners’ training and way of working through the stigma and discrimination faced by people living with HIV, especially those who are members of key populations such as Men who have sex with men (MSM), sex workers, etc. Variables that may improve the model include meeting minutes, assessments, reports, policies, inventory lists of medical equipment, and ways on how practitioners deal with the socio-structural factors impacting consumers. Job-related variables may include practitioners’ job security, remuneration, and job satisfaction. Future research may also include a larger sample, including FHS teams from multiple municipalities. Finally, we recommend longitudinal studies to examine relationships between the multi-level factors facilitating service integration and its effect on consumer outcomes. This will assist in creating sophisticated, evidence-based guidelines for service integration.Our explanatory model underscores key variables that might inform training to help physicians, nurses, and CHAs improve service integration. Practitioner trainings should focus on harnessing cognitive constructs, such as confidence, perseverance, and knowledge and skills. We recommend training that espouses active, problem-based, and action learning in order to reflect real-world practices. The need to address complex health promotion in the context of community factors calls for recognizing the limits of individual practitioner expertise so that practitioners can leverage the expertise of colleagues to provide comprehensive care to consumers. Diversity of expertise within FHS teams is important so that clinical expertise (physicians and nurses) may merge with experiential knowledge (CHAs) to solve complex health issues. We recommend that FHS health care managers/administrators work together to provide curricula/trainings that focus on service integration. Academic and FHS collaboration curricula ought to include the input of both CHAs and medical staff. The purpose of including CHAs in curricular design is to ensure that community-based skills are included in training/education initiatives.This study shows that practitioners in Brazil’s FHS are likely to provide HIV education while helping individuals obtain documentation for receiving primary care and welfare services, and to participate in health promotion community activities. Our major contribution is an explanatory model that can be adapted to examine the impact of other variables that might influence integration of other services. This model can also be adapted to study service integration in primary care systems in other countries.The authors thank the Columbia University School of Social Work and the Institute for Latin American Studies at Columbia University for supporting this research with pilot and dissemination funding, respectively. We thank research participants who graciously provided the data for this research. We also thank the following: Estratégia Saúde da Família providers and administrators critical to the execution of this research Ivanete Hindriches da S. Torres, Roselí Monteiro Silva, Gilberto da Silva Dorneles, Paulo de Tarso Machado Assis, Maria do Carmo de Castro Tófani, and Nádia Cristina Dias Duarte; our partners at Universidade Católica Rio de Janeiro Sueli Bulhões da Silva, and Luíza Helena Nunes Ermel; and individuals who provided overall support—Susan Witte, Margareth Zanchetta, and Barbara L. Simon.R.R. participated in the analysis and interpretation of data, and writing of the manuscript. R.M.P. participated in the design and execution of the study, and contributed to critical revisions of the manuscript. M.W. contributed analysis tools. All authors read and approved the final manuscript.The authors declare no conflict of interest. The founding sponsors had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, and in the decision to publish the results.Service Integration Framework. FHS: Estratégia Saúde da Família (Family Health Strategy); Human Immunodeficiency Virus Infection and Acquired Immune Deficiency Syndrome (HIV/AIDS).Predictors of Service Integration.I know how to ask questions to help client/patients discuss their healthI know how to ask questions about health risksI know how to ask questions about medication side effectsI know exactly what my client/patient needs areI am able to make treatment plans which fit the needs and abilities of my patientI am able to address client/patient needsI am committed to delivering the best services possible to the families in my catchment area, even when they are difficultThe existence of FHS teams has improved the quality of health in my catchment areaI know the latest news in my catchment area affecting client/patientsI utilize other colleagues in deciding interventionsI have access to colleagues when I need helpTeam meetings are importantMy client/patient values and preferences are very importantMy client/patient goals are very importantMy client/patient and I work together to address needsWith client/patients’ help, I monitor client/patient outcomesI am able to understand and use protocolsI have the knowledge/skills to bring together information from different sources to address my client/patient’s needsI know how to use new information to treat my client/patientI can tailor my work based on the information I gathered from my client/patient and from researchI am able to change or alter treatment based on changes in the needs of the client/patientHow many client/patients do you serve?Tell us in years the length of time you worked for the FHSDo you live near your job? (yes or no)Tell us your length of commute (0–10 min; 11–30 min; and greater than 30 min).Poor work conditions interfere with my ability to address needs of client/patientLack of resources interfere with my ability to address needs of client/patientAll Cronbach alphas greater than 0.5, considered “reasonably good” when the subject matter under examination is novel (newly developed measures used in the analysis) ([50], p. 70); FHS: Family Health Strategy.Demographics and Job Context (Practitioner type).a: indicates significant effects when p < 0.05; CHAs: Community Health Agents; Pardo: refers to mixed races, such as mulattos [49].Standardized estimated direct effects from fully-adjusted structural equation model.a: indicates significant effects when p < 0.05; b: indicates significant effects when p < 0.10; B: Path Coefficients.

Med-MDPI/ijerph_2/ijerph-14-03-00295.txt

@@ -0,0 +1,2585 @@
+Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).Brucellosis, the most common zoonotic disease worldwide, represents a great threat to animal husbandry with the potential to cause enormous economic losses. Brucellosis has become a major public health problem in China, and the number of human brucellosis cases has increased dramatically in recent years. In order to evaluate different intervention strategies to curb brucellosis transmission in China, a novel mathematical model with a general indirect transmission incidence rate was presented. By comparing the results of three models using national human disease data and 11 provinces with high case numbers, the best fitted model with standard incidence was used to investigate the potential for future outbreaks. Estimated basic reproduction numbers were highly heterogeneous, varying widely among provinces. The local basic reproduction numbers of provinces with an obvious increase in incidence were much larger than the average for the country as a whole, suggesting that environment-to-individual transmission was more common than individual-to-individual transmission. We concluded that brucellosis can be controlled through increasing animal vaccination rates, environment disinfection frequency, or elimination rates of infected animals. Our finding suggests that a combination of animal vaccination, environment disinfection, and elimination of infected animals will be necessary to ensure cost-effective control for brucellosis.Brucellosis, a bacterial disease caused by various Brucella species, is one of the most common zoonotic infections globally [1,2,3]. Four Brucella species are mainly responsible for the disease: B. melitensis in goats and sheep, B. abortus typically found in cattle, B. canis in dogs, and B. suis in swine [4]. Even though these four species of Brucella can infect humans, B. melitensis remains the major cause of human disease worldwide (and may account for up to 90% of all brucellosis cases [5]). The remaining illnesses are caused by B. abortus and B. suis, with rare but persisting cases of B. canis infections in humans [6]. Although there are a small number of reports of vertical and horizontal transmission between humans [7,8], it is generally acknowledged that human-to-human transmission of the infection is a very rare event [9].Humans become infected typically through consumption of the unpasteurized products contaminated by the bacterial agent, and to a lesser extent, contact with infected animals. Consequently, brucellosis in humans is strongly linked to the management of infected animals and ingestion of unpasteurized dairy products [10]. According to the length and severity of symptoms, the disease in humans is arbitrarily classified as acute (less than 8 weeks), subacute (from 8 to 52 weeks), or chronic (more than 1 year) [9]. The disease is commonly underreported, misdiagnosed and once chronic disease develops, it is resistant to treatment, which consists of antibiotics for long periods [11,12,13]. Mortality is reported to be negligible, but the illness can persist for several years. Though less severe in animals, brucellosis can cause economic losses by adversely affecting reproduction and fertility, survival of newborns, and milk yields [14,15].Human brucellosis matches the regions of the world with high levels of animal infection endemicity: the Mediterranean basin, Middle East, Western Asia, Africa, and South America [3] where hundreds of thousands of new cases are reported annually [9,16]. In developed countries, control of animal brucellosis has been successfully achieved. However, many of these control options are less achievable in developing countries [17]. In China, human brucellosis is a class B notifiable infectious disease, and information regarding each confirmed case must be reported to the Chinese Center for Disease Control and Prevention (CCDC) through the National Notifiable Disease Surveillance System (NNDSS) since 2004 [18]. Although many measures based on the control programs for brucellosis have been set up, the brucellosis-positive rate in humans has increased significantly in recent years [19]. It is believed that reduced awareness of the disease and concordant reductions in surveillance and vaccination rates has led to an overall rise in human cases [20]. Hence, reconsidering the use of animals vaccination to reduce susceptibility to brucellosis infection has become urgent.Mathematical modeling has the potential to analyze the mechanisms of transmission and the complexity of epidemiological characteristics of infectious diseases, and can indicate new approaches to prevent and control future epidemics [21,22]. In recent years, several mathematical modeling studies have reported on the transmission of brucellosis [23,24,25,26,27,28,29,30]. However, these earlier models have mainly focused on the spread of brucellosis through using statistical methods or the theoretical brucellosis dynamic model. Only Hou et al. [25] and Li et al. [26] studied the underlying dynamics of brucellosis transmission between sheep and human in Inner Mongolia, China, and explicitly quantified levels of transmission between Brucella of environment and individuals. But model-based evaluations do not yet exist for strategising the control of brucellosis in mainland China and other provinces. In this paper, we explored the utility of three different dynamic mathematical models to explain the incidence of brucellosis from 2004 to 2014, and the best-fit model was selected by using Akaike information criterion. The best-fit model is then used to investigate the potential for future outbreaks, and to estimate the national-level basic reproduction number as well as the more local-level outbreak thresholds for the eleven provinces with high case numbers (including Inner Mongolia, Shanxi, Heilongjiang, Hebei, Xinjiang, Jilin, Henan, Liaoning, Shaanxi, Shandong and Ningxia), and to provide new insights into the epidemiology of brucellosis, including province-level control targets required to achieve elimination.The annual and cumulative reported human brucellosis cases of mainland China and the selected 11 provinces with high case numbers from 2004 to 2014 were obtained from the National Notifiable Disease Surveillance System (NNDSS). The spatiotemporal distribution map presented in Figure 1 showed that human brucellosis was widely distributed in northern, northeastern, and western China. Elsewhere, cases are more sporadic. The 11 provinces with high case numbers include Inner Mongolia, Shanxi, Heilongjiang, Hebei, Xinjiang, Jilin, Henan, Liaoning, Shaanxi, Shandong and Ningxia; these provinces accounted for > 
+          
+            
+              
+                98.5
+                %
+              
+            
+          
+         of the national reported cases. During 2004–2014, a total of 344,351 new cases were reported in mainland China, with the annual incidence of human brucellosis increasing. In 2004, only 21 provinces reported 11,472 new cases of human brucellosis, but the epidemic regions expanded to 30 provinces of China (except for Hong Kong, Macao, Taiwan, and Tibet) with 57,222 cases reported in 2014.Brucellosis can be transmitted to individuals through direct contact with infected animals or indirect transmission by Brucella in the environment, which contains contaminated forage, water, grass, liquids, products and the uterine fluids from infected animals [26]. Environment-to-individual transmission is defined as a indirect transmission process with a high infectious dose, resulting from contact with Brucella in the environment. In contrast, individual-to-individual transmission is assumed to be a direct transmission process, with a lower infectious dose. In mainland China, B. melitensis (sheep-type Brucella) is the predominant pathogen associated with large outbreaks [31]. The result of Li et al. [20] also indicated that sheep and goats probably were the main animal hosts transmitting the diseases to humans in northern, northeastern and western China. Hence, sheep and goats were considered and called sheep population in the model. There are some assumptions for our model: (1) Brucellosis in the exposure period is hardly detected, and animals in this period can also infect susceptible sheep and humans. Hence, we ignored the exposed period in the sheep population; (2) In China in the 20th century, the control measure for brucellosis was mainly vaccinating livestock [31], but there are few reports on animal vaccination in the first fifteen years of 21st century. Therefore, animal vaccination is also ignored; (3) There is no data reporting human-to-human transmission of brucellosis, so the rate of human-to-human and human-to-animal transmission was ignored. There are also some other assumptions on our model, which are demonstrated in the flowchart (see Figure 2).In this study, the brucellosis model classified the human population (denoted by 
+          
+            
+              
+                N
+                h
+              
+            
+          
+        ) into susceptible 
+          
+            
+              
+                S
+                h
+              
+            
+          
+        , exposed 
+          
+            
+              
+                E
+                h
+              
+            
+          
+        , acute infectious 
+          
+            
+              
+                I
+                h
+              
+            
+          
+        , and chronic infectious 
+          
+            
+              
+                C
+                h
+              
+            
+          
+         compartments. Human population birth and death rates were included as 
+          
+            
+              
+                b
+                h
+              
+            
+          
+         and 
+          
+            
+              
+                d
+                h
+              
+            
+          
+        , respectively. In China, the human birth rate has been greater than the death rate in recent years. The human host incubation period is 
+          
+            
+              
+                1
+                /
+                σ
+              
+            
+          
+         years, and the incubation period of human brucellosis is about two weeks [32], so the clinical outcome rate of exposed people is 
+          
+            
+              
+                σ
+                =
+                26
+              
+            
+          
+         per year. Acute infections progress at the rate p into chronic infections and revert to susceptible at rate m. The population of sheep, denoted 
+          
+            
+              
+                N
+                s
+              
+            
+          
+        , were separated into either susceptible 
+          
+            
+              
+                S
+                s
+              
+            
+          
+         and infectious 
+          
+            
+              
+                I
+                s
+              
+            
+          
+         compartments. From the China Animal Husbandry Statistical Yearbook [33], one could calculate that the yearly breeding stock deviation of the sheep population in mainland China was less 5 percent, so sheep recruitment and slaughter rates, listed as b, were set to balance each other. Let W denote the density of Brucella in the environment. Animals with brucellosis shed Brucella in the environment at rate k and environmental Brucella had a mortality rate of δ, and environmental shedding mainly includes the contaminated products by infected animals. Susceptible animals acquire brucellosis either through direct transmission at rate 
+          
+            
+              
+                
+                  β
+                  s
+                
+                
+                  S
+                  s
+                
+                
+                  
+                    I
+                    s
+                  
+                  
+                    N
+                    s
+                  
+                
+              
+            
+          
+         or by ingesting environmental Brucella at transmission rate 
+          
+            
+              
+                
+                  β
+                  
+                    s
+                    w
+                  
+                
+                
+                  S
+                  s
+                
+                f
+                
+                  (
+                  W
+                  )
+                
+              
+            
+          
+        . Susceptible humans acquire brucellosis through direct contact with infected animals or indirect transmission by environmental Brucella at rates 
+          
+            
+              
+                
+                  β
+                  h
+                
+                
+                  S
+                  h
+                
+                
+                  
+                    I
+                    s
+                  
+                  
+                    N
+                    h
+                  
+                
+              
+            
+          
+         and 
+          
+            
+              
+                
+                  β
+                  
+                    h
+                    w
+                  
+                
+                
+                  S
+                  h
+                
+                g
+                
+                  (
+                  
+                    N
+                    h
+                  
+                  ,
+                  W
+                  )
+                
+              
+            
+          
+        , respectively. In this study, the standard incidence and general incidence rates were used for direct contact transmission and indirect transmission, respectively. Hence, the brucellosis model is described by the following ordinary differential equations:
+		
+          (1)
+          
+            
+              
+                
+                  
+                    
+                      
+                        
+                          
+                            d
+                            
+                              S
+                              s
+                            
+                          
+                          
+                            d
+                            t
+                          
+                        
+                        =
+                        b
+                        
+                          N
+                          s
+                        
+                        −
+                        
+                          β
+                          s
+                        
+                        
+                          S
+                          s
+                        
+                        
+                          
+                            I
+                            s
+                          
+                          
+                            N
+                            s
+                          
+                        
+                        −
+                        
+                          β
+                          
+                            s
+                            w
+                          
+                        
+                        
+                          S
+                          s
+                        
+                        f
+                        
+                          (
+                          W
+                          )
+                        
+                        −
+                        b
+                        
+                          S
+                          s
+                        
+                        ,
+                      
+                    
+                  
+                  
+                    
+                      
+                        
+                          
+                            d
+                            
+                              I
+                              s
+                            
+                          
+                          
+                            d
+                            t
+                          
+                        
+                        =
+                        
+                          β
+                          s
+                        
+                        
+                          S
+                          s
+                        
+                        
+                          
+                            I
+                            s
+                          
+                          
+                            N
+                            s
+                          
+                        
+                        +
+                        
+                          β
+                          
+                            s
+                            w
+                          
+                        
+                        
+                          S
+                          s
+                        
+                        f
+                        
+                          (
+                          W
+                          )
+                        
+                        −
+                        b
+                        
+                          I
+                          s
+                        
+                        ,
+                      
+                    
+                  
+                  
+                    
+                      
+                        
+                          
+                            d
+                            W
+                          
+                          
+                            d
+                            t
+                          
+                        
+                        =
+                        k
+                        
+                          I
+                          s
+                        
+                        −
+                        δ
+                        W
+                        ,
+                      
+                    
+                  
+                  
+                    
+                      
+                        
+                          
+                            d
+                            
+                              S
+                              h
+                            
+                          
+                          
+                            d
+                            t
+                          
+                        
+                        =
+                        
+                          b
+                          h
+                        
+                        
+                          N
+                          h
+                        
+                        +
+                        m
+                        
+                          I
+                          h
+                        
+                        −
+                        
+                          β
+                          h
+                        
+                        
+                          S
+                          h
+                        
+                        
+                          
+                            I
+                            s
+                          
+                          
+                            N
+                            h
+                          
+                        
+                        −
+                        
+                          β
+                          
+                            h
+                            w
+                          
+                        
+                        
+                          S
+                          h
+                        
+                        g
+                        
+                          (
+                          
+                            N
+                            h
+                          
+                          ,
+                          W
+                          )
+                        
+                        −
+                        
+                          d
+                          h
+                        
+                        
+                          S
+                          h
+                        
+                        ,
+                      
+                    
+                  
+                  
+                    
+                      
+                        
+                          
+                            d
+                            
+                              E
+                              h
+                            
+                          
+                          
+                            d
+                            t
+                          
+                        
+                        =
+                        
+                          β
+                          h
+                        
+                        
+                          S
+                          h
+                        
+                        
+                          
+                            I
+                            s
+                          
+                          
+                            N
+                            h
+                          
+                        
+                        +
+                        
+                          β
+                          
+                            h
+                            w
+                          
+                        
+                        
+                          S
+                          h
+                        
+                        g
+                        
+                          (
+                          
+                            N
+                            h
+                          
+                          ,
+                          W
+                          )
+                        
+                        −
+                        σ
+                        
+                          E
+                          h
+                        
+                        −
+                        
+                          d
+                          h
+                        
+                        
+                          E
+                          h
+                        
+                        ,
+                      
+                    
+                  
+                  
+                    
+                      
+                        
+                          
+                            d
+                            
+                              I
+                              h
+                            
+                          
+                          
+                            d
+                            t
+                          
+                        
+                        =
+                        σ
+                        
+                          E
+                          h
+                        
+                        −
+                        p
+                        
+                          I
+                          h
+                        
+                        −
+                        m
+                        
+                          I
+                          h
+                        
+                        −
+                        
+                          d
+                          h
+                        
+                        
+                          I
+                          h
+                        
+                        ,
+                      
+                    
+                  
+                  
+                    
+                      
+                        
+                          
+                            d
+                            
+                              C
+                              h
+                            
+                          
+                          
+                            d
+                            t
+                          
+                        
+                        =
+                        p
+                        
+                          I
+                          h
+                        
+                        −
+                        
+                          d
+                          h
+                        
+                        
+                          C
+                          h
+                        
+                        ,
+                      
+                    
+                  
+                  
+                    
+                      
+                        
+                          N
+                          s
+                        
+                        =
+                        
+                          S
+                          s
+                        
+                        +
+                        
+                          I
+                          s
+                        
+                        ,
+                        
+                          N
+                          h
+                        
+                        =
+                        
+                          S
+                          h
+                        
+                        +
+                        
+                          E
+                          h
+                        
+                        +
+                        
+                          I
+                          h
+                        
+                        +
+                        
+                          C
+                          h
+                        
+                        .
+                      
+                    
+                  
+                
+              
+            
+          
+        
+        All parameters of Equation (1) are assumed to be nonnegative. The mathematical properties of the brucellosis model including calculations for the basic reproduction number and global stability of the equilibrium are given in Section 2 Dynamical behavior (Supplementary Material).Due to the general incidence rates of indirect transmission in Equation (1), multiple models can be created. To evaluate the situation of brucellosis in some provinces and the whole country (mainland China), we need to choose the most appropriate functional form for indirect brucellosis transmission i.e., a comparison could be made between transmission determined by a mass action incidence rate, saturating incidence rate and standard incidence rate. Case 1. Standard incidence rate: 
+        
+          
+            
+              
+                β
+                
+                  s
+                  w
+                
+              
+              
+                S
+                s
+              
+              f
+              
+                (
+                W
+                )
+              
+              =
+              
+                β
+                
+                  s
+                  w
+                
+              
+              
+                S
+                s
+              
+              
+                W
+                
+                  N
+                  s
+                
+              
+              ,
+              
+                β
+                
+                  h
+                  w
+                
+              
+              
+                S
+                h
+              
+              g
+              
+                (
+                
+                  N
+                  h
+                
+                ,
+                W
+                )
+              
+              =
+              
+                β
+                
+                  h
+                  w
+                
+              
+              
+                S
+                h
+              
+              
+                W
+                
+                  N
+                  h
+                
+              
+            
+          
+        
+      . Case 2. Saturating incidence rate: 
+        
+          
+            
+              
+                β
+                
+                  s
+                  w
+                
+              
+              
+                S
+                s
+              
+              f
+              
+                (
+                W
+                )
+              
+              =
+              
+                β
+                
+                  s
+                  w
+                
+              
+              
+                S
+                s
+              
+              
+                W
+                
+                  ε
+                  +
+                  W
+                
+              
+              ,
+              
+                β
+                
+                  h
+                  w
+                
+              
+              
+                S
+                h
+              
+              g
+              
+                (
+                
+                  N
+                  h
+                
+                ,
+                W
+                )
+              
+              =
+              
+                β
+                
+                  h
+                  w
+                
+              
+              
+                S
+                h
+              
+              
+                W
+                
+                  ε
+                  +
+                  W
+                
+              
+            
+          
+        
+      . Case 3. Mass action incidence rate: 
+        
+          
+            
+              
+                β
+                
+                  s
+                  w
+                
+              
+              
+                S
+                s
+              
+              f
+              
+                (
+                W
+                )
+              
+              =
+              
+                β
+                
+                  s
+                  w
+                
+              
+              
+                S
+                s
+              
+              
+                W
+                M
+              
+              ,
+              
+                β
+                
+                  h
+                  w
+                
+              
+              
+                S
+                h
+              
+              g
+              
+                (
+                
+                  N
+                  h
+                
+                ,
+                W
+                )
+              
+              =
+              
+                β
+                
+                  h
+                  w
+                
+              
+              
+                S
+                h
+              
+              
+                W
+                M
+              
+            
+          
+        
+      . Here, ε and M are scaling factors for the brucella concentration in the environment.To determine the most appropriate functional form, we firstly fitted each Case of model to the annual infection data on numbers of national human brucellosis cases from 2004 to 2014. Literature reviews facilitated epidemiological parameter values for brucellosis (including extrinsic incubation period of human brucellosis, transfer rate from acute infections to chronic infections and susceptible populations, Brucella shedding rate by infected animals and the decaying rate of Brucella in the environment) are given in Table S1 (Supplementary Material). From the China Animal Husbandry Statistical Yearbook [33], we can find that the value of sheep recruitment and slaughtering rate is 
+        
+          
+            
+              b
+              =
+              0.9026
+            
+          
+        
+      , (0.8531–0.9521) for mainland China. From the China Statistical Yearbook [34], one can obtain the average values and corresponding 95% confidence intervals of the human birth rate 
+        
+          
+            
+              b
+              h
+            
+          
+        
+       and death rate 
+        
+          
+            
+              d
+              h
+            
+          
+        
+       from 2004 to 2014 for mainland China, which are 
+        
+          
+            
+              
+                b
+                h
+              
+              =
+              0.012098
+            
+          
+        
+      , (0.0119863–0.0122097) and 
+        
+          
+            
+              
+                d
+                h
+              
+              =
+              0.006937
+            
+          
+        
+      , (0.0067422–0.0071318), respectively.For the initial values of the model in mainland China, the susceptible sheep and human population, the number of cases of human brucellosis infection can be directly obtained from [33,34], which are 
+          
+            
+              
+                
+                  S
+                  s
+                
+                
+                  (
+                  0
+                  )
+                
+                =
+                2.85
+                ×
+                
+                  10
+                  8
+                
+              
+            
+          
+        , 
+          
+            
+              
+                
+                  S
+                  h
+                
+                
+                  (
+                  0
+                  )
+                
+                =
+                1.29998
+                ×
+                
+                  10
+                  9
+                
+              
+            
+          
+        , and 
+          
+            
+              
+                
+                  I
+                  h
+                
+                
+                  (
+                  0
+                  )
+                
+                =
+                11472
+              
+            
+          
+        , respectively. However, the number of infected animals, the density of Brucella in the environment, and the number of cases of exposure need estimates. The number of annual cases of human brucellosis infection in mainland China were used to estimate the transmission rate of 
+          
+            
+              
+                β
+                s
+              
+            
+          
+        , 
+          
+            
+              
+                β
+                
+                  s
+                  w
+                
+              
+            
+          
+        , 
+          
+            
+              
+                β
+                h
+              
+            
+          
+         and 
+          
+            
+              
+                β
+                
+                  h
+                  w
+                
+              
+            
+          
+        . We firstly fixed the human indirect transmission rate and assumed that 
+          
+            
+              
+                
+                  β
+                  
+                    h
+                    w
+                  
+                
+                =
+                0.5
+              
+            
+          
+        , because there might be high correlations between 
+          
+            
+              
+                β
+                s
+              
+            
+          
+        , 
+          
+            
+              
+                β
+                
+                  s
+                  w
+                
+              
+            
+          
+        , 
+          
+            
+              
+                β
+                h
+              
+            
+          
+         and 
+          
+            
+              
+                β
+                
+                  h
+                  w
+                
+              
+            
+          
+        . Then, we estimated the initial values of model and transmission rate of 
+          
+            
+              
+                β
+                s
+              
+            
+          
+        , 
+          
+            
+              
+                β
+                
+                  s
+                  w
+                
+              
+            
+          
+        , 
+          
+            
+              
+                β
+                h
+              
+            
+          
+         by using the least-squares fitting routine fminsearch in MATLAB. The least-square estimation is adopted here to find the parameter values to minimize the objective function:
+		
+          
+            
+              
+                L
+                =
+                
+                  1
+                  n
+                
+                
+                  ∑
+                  
+                    t
+                    =
+                    1
+                  
+                  n
+                
+                
+                  
+                    (
+                    Y
+                    
+                      (
+                      t
+                      )
+                    
+                    −
+                    y
+                    
+                      (
+                      t
+                      )
+                    
+                    )
+                  
+                  2
+                
+                ,
+              
+            
+          
+        
+        where 
+          
+            
+              
+                Y
+                (
+                t
+                )
+              
+            
+          
+         is the theoretical number of human infection cases, 
+          
+            
+              
+                y
+                (
+                t
+                )
+              
+            
+          
+         is the actual reported human cases at year t, and n is the number of reported data. Finally, using the obtained values from MATLAB, the resulting estimates for parameters 
+          
+            
+              
+                β
+                s
+              
+            
+          
+        , 
+          
+            
+              
+                β
+                
+                  s
+                  w
+                
+              
+            
+          
+        , 
+          
+            
+              
+                β
+                h
+              
+            
+          
+         of mainland China could be generated (independently for all three incidence functions) through DEDiscover software (DEDiscover is a general-purpose tool to perform simulation, parameter estimation and statistical analysis for any problem that can be expressed as a set of differential Equations [35]), and listed in Table 1.Figure 3 shows the comparison between theoretical and annual cases of human brucellosis infections in mainland China are plotted along with the 95% percent interval for 1000 simulation outputs from the alternative models Cases 1–3. From Figure 3, we can conclude that in three cases, there is no obvious difference in the simulation results. In order to check the adequacy of the model, we use the method of Akaike information criterion (AIC) [36,37,38], to compare the validity of the three results and choose the best model from multiple competing models. Since the ratio of the number of data points to the number of parameters fitted is less than 40, we need to compute the 
+          
+            
+              
+                A
+                I
+                
+                  C
+                  c
+                
+              
+            
+          
+         for each model. 
+          
+            
+              
+                A
+                I
+                
+                  C
+                  c
+                
+              
+            
+          
+         values corresponding to three Cases are 
+          
+            
+              
+                A
+                I
+                
+                  C
+                  
+                    c
+                    1
+                  
+                
+                =
+                181.4021
+              
+            
+          
+        , 
+          
+            
+              
+                A
+                I
+                
+                  C
+                  
+                    c
+                    2
+                  
+                
+                =
+                186.8334
+              
+            
+          
+        , 
+          
+            
+              
+                A
+                I
+                
+                  C
+                  
+                    c
+                    3
+                  
+                
+                =
+                186.6243
+              
+            
+          
+        , and the model with the lowest 
+          
+            
+              
+                A
+                I
+                
+                  C
+                  c
+                
+              
+            
+          
+         is Case 1. Moreover, Δ 
+          
+            
+              
+                A
+                I
+                
+                  C
+                  
+                    c
+                    2
+                  
+                
+                =
+                A
+                I
+                
+                  C
+                  
+                    c
+                    2
+                  
+                
+                −
+                A
+                I
+                
+                  C
+                  
+                    c
+                    1
+                  
+                
+                =
+                5.3004
+                >
+                4
+              
+            
+          
+         and Δ 
+          
+            
+              
+                A
+                I
+                
+                  C
+                  
+                    c
+                    3
+                  
+                
+                =
+                A
+                I
+                
+                  C
+                  
+                    c
+                    3
+                  
+                
+                −
+                A
+                I
+                
+                  C
+                  
+                    c
+                    1
+                  
+                
+                =
+                5.2222
+                >
+                4
+              
+            
+          
+        , which imply that assumptions Case 2 and Case 3 are considered as the model with less support. In a word, the model selection results show that Case 1, the standard incidence function, is consistently the best model to be used for inference.For 11 selected provinces with high case numbers, we also fitted each model for the yearly data to determine the most appropriate functional form. For Inner Mongolia and Jilin, we noted significant reductions in case numbers following 2011 and 2010, respectively. The causes for the decrease in these two provinces are prevention and control strategies including animal vaccination and elimination of the infected animals. For the Heilongjiang and Shannxi provinces, the cumulative number of brucellosis cases was used to give the estimation. Also from the China Statistical Yearbook [34] and China Animal Husbandry Statistical Yearbook [33], one can obtain the demographic parameter values (including human population birth rate 
+          
+            
+              
+                b
+                h
+              
+            
+          
+         and death rates 
+          
+            
+              
+                d
+                h
+              
+            
+          
+        , sheep recruitment and slaughtering rate b) and their corresponding 95% confidence intervals for these 11 provinces, listed in Table S2 (Supplementary Material). Fixing the human indirect transmission rate as 
+          
+            
+              
+                
+                  β
+                  
+                    h
+                    w
+                  
+                
+                =
+                0.5
+              
+            
+          
+        , the estimates of 
+          
+            
+              
+                β
+                s
+              
+            
+          
+        , 
+          
+            
+              
+                β
+                
+                  s
+                  w
+                
+              
+            
+          
+        , 
+          
+            
+              
+                β
+                h
+              
+            
+          
+         for these 11 provinces are also obtained by using DEDiscover software, and shown in Table S3 (Supplementary Material). The alternative model structures were further compared through assessing their fit to the eleven provinces with high case numbers, which also show that the standard incidence function was the best model to be used. Hence, the following model will be used for further simulation:
+		
+          (2)
+          
+            
+              
+                
+                  
+                    
+                      
+                        
+                          
+                            
+                              
+                                
+                                  d
+                                  
+                                    S
+                                    s
+                                  
+                                
+                                
+                                  d
+                                  t
+                                
+                              
+                              =
+                              b
+                              
+                                N
+                                s
+                              
+                              −
+                              
+                                β
+                                s
+                              
+                              
+                                S
+                                s
+                              
+                              
+                                
+                                  I
+                                  s
+                                
+                                
+                                  N
+                                  s
+                                
+                              
+                              −
+                              
+                                β
+                                
+                                  s
+                                  w
+                                
+                              
+                              
+                                S
+                                s
+                              
+                              
+                                W
+                                
+                                  N
+                                  s
+                                
+                              
+                              −
+                              b
+                              
+                                S
+                                s
+                              
+                              ,
+                            
+                          
+                        
+                        
+                          
+                            
+                              
+                                
+                                  d
+                                  
+                                    I
+                                    s
+                                  
+                                
+                                
+                                  d
+                                  t
+                                
+                              
+                              =
+                              
+                                β
+                                s
+                              
+                              
+                                S
+                                s
+                              
+                              
+                                
+                                  I
+                                  s
+                                
+                                
+                                  N
+                                  s
+                                
+                              
+                              +
+                              
+                                β
+                                
+                                  s
+                                  w
+                                
+                              
+                              
+                                S
+                                s
+                              
+                              
+                                W
+                                
+                                  N
+                                  s
+                                
+                              
+                              −
+                              b
+                              
+                                I
+                                s
+                              
+                              ,
+                            
+                          
+                        
+                        
+                          
+                            
+                              
+                                
+                                  d
+                                  W
+                                
+                                
+                                  d
+                                  t
+                                
+                              
+                              =
+                              k
+                              
+                                I
+                                s
+                              
+                              −
+                              δ
+                              W
+                              ,
+                            
+                          
+                        
+                        
+                          
+                            
+                              
+                                
+                                  d
+                                  
+                                    S
+                                    h
+                                  
+                                
+                                
+                                  d
+                                  t
+                                
+                              
+                              =
+                              
+                                b
+                                h
+                              
+                              
+                                N
+                                h
+                              
+                              +
+                              m
+                              
+                                I
+                                h
+                              
+                              −
+                              
+                                β
+                                h
+                              
+                              
+                                S
+                                h
+                              
+                              
+                                
+                                  I
+                                  s
+                                
+                                
+                                  N
+                                  h
+                                
+                              
+                              −
+                              
+                                β
+                                
+                                  h
+                                  w
+                                
+                              
+                              
+                                S
+                                h
+                              
+                              
+                                W
+                                
+                                  N
+                                  h
+                                
+                              
+                              −
+                              
+                                d
+                                h
+                              
+                              
+                                S
+                                h
+                              
+                              ,
+                            
+                          
+                        
+                        
+                          
+                            
+                              
+                                
+                                  d
+                                  
+                                    E
+                                    h
+                                  
+                                
+                                
+                                  d
+                                  t
+                                
+                              
+                              =
+                              
+                                β
+                                h
+                              
+                              
+                                S
+                                h
+                              
+                              
+                                
+                                  I
+                                  s
+                                
+                                
+                                  N
+                                  h
+                                
+                              
+                              +
+                              
+                                β
+                                
+                                  h
+                                  w
+                                
+                              
+                              
+                                S
+                                h
+                              
+                              
+                                W
+                                
+                                  N
+                                  h
+                                
+                              
+                              −
+                              σ
+                              
+                                E
+                                h
+                              
+                              −
+                              
+                                d
+                                h
+                              
+                              
+                                E
+                                h
+                              
+                              ,
+                            
+                          
+                        
+                        
+                          
+                            
+                              
+                                
+                                  d
+                                  
+                                    I
+                                    h
+                                  
+                                
+                                
+                                  d
+                                  t
+                                
+                              
+                              =
+                              σ
+                              
+                                E
+                                h
+                              
+                              −
+                              p
+                              
+                                I
+                                h
+                              
+                              −
+                              m
+                              
+                                I
+                                h
+                              
+                              −
+                              
+                                d
+                                h
+                              
+                              
+                                I
+                                h
+                              
+                              ,
+                            
+                          
+                        
+                        
+                          
+                            
+                              
+                                
+                                  d
+                                  
+                                    C
+                                    h
+                                  
+                                
+                                
+                                  d
+                                  t
+                                
+                              
+                              =
+                              p
+                              
+                                I
+                                h
+                              
+                              −
+                              
+                                d
+                                h
+                              
+                              
+                                C
+                                h
+                              
+                              ,
+                            
+                          
+                        
+                        
+                          
+                            
+                              
+                                N
+                                s
+                              
+                              =
+                              
+                                S
+                                s
+                              
+                              +
+                              
+                                I
+                                s
+                              
+                              ,
+                              
+                                N
+                                h
+                              
+                              =
+                              
+                                S
+                                h
+                              
+                              +
+                              
+                                E
+                                h
+                              
+                              +
+                              
+                                I
+                                h
+                              
+                              +
+                              
+                                C
+                                h
+                              
+                            
+                          
+                        
+                      
+                    
+                  
+                
+              
+            
+          
+        The provinces with the highest risk of brucellosis incidence in the country are very distinct in many ways, for example with respect to animal habitat and animal and human population densities. The process was initially conducted for mainland China (total) and then repeated for each of the 11 provinces with high case numbers. Through using the available parameters in Tables S1 and S2 (Supplementary Materials), and the estimated parameters in Table S3 (Supplementary Materials), Monte Carlo simulation runs were then conducted to assess the data fitting of the cases of brucellosis infection with mainland China and 11 provinces with high case numbers, and the plots are shown in Figure 4. The 95% percent interval for all 1000 passing simulation trajectories show that there is a good match between the reported data and the theoretical prediction of Equation (2).The basic reproduction number provides useful guidelines for the prevention and control strategies for epidemics. For Equation (2), the basic reproduction number is:
+		
+          (3)
+          
+            
+              
+                
+                  
+                    
+                      
+                        R
+                        0
+                      
+                      =
+                      
+                        
+                          β
+                          s
+                        
+                        b
+                      
+                      +
+                      
+                        
+                          k
+                          
+                            β
+                            
+                              s
+                              w
+                            
+                          
+                        
+                        
+                          b
+                          δ
+                        
+                      
+                      =
+                      
+                        R
+                        
+                          0
+                        
+                        i
+                      
+                      +
+                      
+                        R
+                        
+                          0
+                        
+                        e
+                      
+                    
+                  
+                
+              
+            
+          
+        
+        where 
+          
+            
+              
+                
+                  R
+                  
+                    0
+                  
+                  e
+                
+                =
+                
+                  
+                    k
+                    
+                      β
+                      
+                        s
+                        w
+                      
+                    
+                  
+                  
+                    b
+                    δ
+                  
+                
+              
+            
+          
+         and 
+          
+            
+              
+                
+                  R
+                  
+                    0
+                  
+                  i
+                
+                =
+                
+                  
+                    β
+                    s
+                  
+                  b
+                
+              
+            
+          
+         are partial reproduction numbers due to environment-to-individual transmission and individual-to-individual transmission, respectively. Using 
+          
+            
+              
+                k
+                =
+                15
+                ,
+                δ
+                =
+                3.6
+              
+            
+          
+        , the average value of b and relevant parameter values in Table S3 (Supplementary Materials), and the corresponding values of 
+          
+            
+              
+                R
+                
+                  0
+                
+                i
+              
+            
+          
+        , 
+          
+            
+              
+                R
+                
+                  0
+                
+                e
+              
+            
+          
+         and 
+          
+            
+              
+                R
+                0
+              
+            
+          
+         for the 11 selected provinces and mainland China are given in Table 2. These quantities of 
+          
+            
+              
+                
+                  R
+                  0
+                
+                >
+                1
+              
+            
+          
+         obtained for the 11 selected provinces and mainland China imply that future epidemics are highly likely, unless effective control measures are put in place. The local basic reproduction numbers of the provinces with an obvious increase in incidence (including Xinjiang, Henan, Liaoning, Shandong, Jilin and Ningxia) are much larger than average for the whole country. Also in these six provinces, the partial reproduction numbers for environment-to-individual transmission are much larger than for individual-to-individual transmission.In recent years, control of animal brucellosis has been successfully achieved in the developed world through the combination of vaccination and test-and-slaughter programs, coupled with effective disease surveillance and animal movement control [3]. In developing countries, however, control by test-and-slaughter and animal vaccination is hardly achievable because of limited resources to indemnify farmers [16]. However, in China, the vaccination against brucellosis of domestic animals has resulted in a rapid decline in the incidence of brucellosis in animals and humans in the 1980s and 1990s [31]. Hence, we believe that brucellosis can be controlled in China if the coverage of vaccination is broad enough. The same conclusion was made in [14], and also claimed that if vaccination rate is too high it may be impossible to eradicate brucellosis as it has been the case in low-income and middle-income countries like India, Mongolia, Mexico and Russia et al. We will study the impact of animal vaccination, elimination of infected animals, and environment disinfection for brucellosis control in China.Animal vaccination can cause a rapid decline in the incidence of brucellosis in animals and humans [31]. For the control measure of animal vaccination, the assumption is that s is the vaccine efficacy level, and v is the coverage of the vaccination programme (i.e., the proportion of the population vaccinated). Then, the vaccination programme reduces the reproduction number to the value 
+          
+            
+              
+                
+                  R
+                  0
+                
+                
+                  (
+                  1
+                  −
+                  v
+                  ·
+                  s
+                  )
+                
+              
+            
+          
+        . From this it follows that the minimum coverage of animals vaccination rate for eradication brucellosis is given by:
+          (4)
+          
+            
+              
+                
+                  
+                    
+                      v
+                      ≥
+                      
+                        s
+                        
+                          −
+                          1
+                        
+                      
+                      
+                        (
+                        1
+                        −
+                        
+                          R
+                          
+                            0
+                          
+                          
+                            −
+                            1
+                          
+                        
+                        )
+                      
+                      .
+                    
+                  
+                
+              
+            
+          
+        A recent study [25] conducted in Inner Mongolia demonstrated that the vaccine effectively protects approximately 82% of sheep i.e., we assume 
+          
+            
+              
+                s
+                =
+                0.82
+              
+            
+          
+        .The control measure of quarantine, separation and elimination of the infected animals for brucellosis can also be incorporated depending on the animals and regions. The assumption is that α is the infected-animal removal rate, and the control reproduction number is 
+          
+            
+              
+                
+                  R
+                  0
+                
+                
+                  b
+                  
+                    b
+                    +
+                    α
+                  
+                
+              
+            
+          
+        . Hence, the brucellosis epidemic can be controlled only with this measure with a minimum coverage of:
+		
+          (5)
+          
+            
+              
+                
+                  
+                    
+                      α
+                      ≥
+                      b
+                      (
+                      
+                        R
+                        0
+                      
+                      −
+                      1
+                      )
+                      .
+                    
+                  
+                
+              
+            
+          
+        In our model, the density of brucella in environment is denoted by W. Hence, the disinfection of environment can be considered another measures for brucellosis. The assumption is that l is the disinfection frequency and the unit is time, m is the effective disinfection rate each time, and the control reproduction number of disinfection is 
+          
+            
+              
+                
+                  R
+                  
+                    0
+                  
+                  i
+                
+                +
+                
+                  R
+                  
+                    0
+                  
+                  e
+                
+                
+                  δ
+                  
+                    δ
+                    +
+                    l
+                    m
+                  
+                
+              
+            
+          
+        . Hence, the minimum disinfection frequency for controlling brucellosis is described as:
+		
+          (6)
+          
+            
+              
+                
+                  
+                    
+                      l
+                      ≥
+                      
+                        δ
+                        m
+                      
+                      
+                        
+                          
+                            R
+                            
+                              0
+                            
+                            i
+                          
+                          +
+                          
+                            R
+                            
+                              0
+                            
+                            e
+                          
+                          −
+                          1
+                        
+                        
+                          1
+                          −
+                          
+                            R
+                            
+                              0
+                            
+                            i
+                          
+                        
+                      
+                      .
+                    
+                  
+                
+              
+            
+          
+        
+		There is no study on the effective disinfection rate each time. In the following calculations we assume that 
+          
+            
+              
+                m
+                =
+                0.5
+              
+            
+          
+        .The minimum vaccination coverage rate, removal rate, and disinfection frequency required to control brucellosis epidemics for the 11 selected provinces and mainland China are presented in Table 3. The outbreak of brucellosis in mainland China could be controlled by animal vaccination, or elimination of the infected animals, or environment disinfection, as long as the measure was strict: the minimum vaccination coverage rate was 0.1478, the minimum removal rate of infected animals was 0.1245, or the minimum disinfection frequency was three times per year. However, this was highly heterogeneous among the 11 provinces with high case numbers: the vaccination coverage necessary for control ranged from 0.0271 to 0.5549 and the removal rate ranged from 0.0154 to 0.7108, while the disinfection frequency ranged from 1 to 19 times per year. Furthermore, we conclude that brucellosis may be easily controlled by a combination of animal vaccination, environment disinfection, and elimination of infected animals.The re-emergence of brucellosis in recent years represents one of the major public health threats in mainland China. Since the beginning of the 21st century, the number of human brucellosis cases has dramatically increased throughout China and the number of human cases has reached a historic high with 57,222 cases reported in 2014. In the period 2004–2014, the human brucellosis epidemic was largely confined to the northern provinces (mainly including Inner Mongolia, Shanxi, Heilongjiang, Hebei, Xinjiang, Jilin, Henan, Liaoning, Shaanxi, Shandong and Ningxia; see Figure 1) where animal husbandry is commonly practiced, and thus the risk of infection is higher. Epidemiological reports on human brucellosis have previously been associated with animal habitat, occupation, host density, socioeconomic status, travel and immigration [39,40,41,42,43]. Although many measures based on the control programs for brucellosis have been set up, the brucellosis-positive rate in animals has increased significantly, along with an increase in human brucellosis cases in recent years. Increase in animal feeding, lack of immunization and animal quarantine, and frequent trading have been implicated as the key risk factors for the dramatic increase in brucellosis incidence in the past decade [19].To investigate the underlying dynamics of brucellosis transmission in mainland China including its provinces with high prevalence, a mathematical model was constructed and fit to human incidence data. Under the general biological assumptions, we have given the formula of the basic reproduction number. We have also proven the global stability of the disease-free equilibrium when 
+        
+          
+            
+              
+                R
+                0
+              
+              <
+              1
+            
+          
+        
+      , and the global stability of the endemic equilibrium when 
+        
+          
+            
+              
+                R
+                0
+              
+              >
+              1
+            
+          
+        
+      . The model was then used to explore the magnitude of control and prevention measures necessary to block brucellosis transmission. By comparing the results of three models using national human disease data and 11 provinces with high case numbers, the best-fitted model with standard incidence was used to investigate the potential for future outbreaks. The estimation of the reproductive numbers (Table 2) shows that future epidemics in 11 selected provinces and mainland China are highly likely, unless effective control measures are put in place. The local basic reproduction number of the provinces with an obvious increase in incidence (including Xinjiang, Henan, Liaoning, Shandong, Jilin and Ningxia) are much larger than average for the whole country. Our model suggests that brucellosis can be controlled through animal vaccination, environmental disinfection, or elimination of infected animals. Hence, much remains need to be done for the local and provincial CDCs aiming to reach the goal of controlling human and animal brucellosis in China.The current study suffers from several limitations. One limitation of the current model is that the mainland China and 11 provinces with high case numbers were treated as homogenous and well-mixed whereas in reality, this is likely not the case. To try to control for this heterogeneity, we did separately model the 11 provinces with the highest incidence, but even within those provinces there will be areas with higher local transmission than others. There also exists a strong age structure, sex structure and occupational association with human brucellosis [20], and therefore a spatially explicit age-structured model would be studied in the future. Finally, the economic costs of different control measures also needed to be considered for future intervention targeting with different regions.Nevertheless, the brucellosis dynamic model developed in this study could increase the understanding of the spread and control of the disease, identify the mechanisms influencing transmission dynamics, and reflect the current trend in the incidence of human brucellosis in mainland China. Our estimates of the basic reproduction numbers for selected 11 provinces (in addition to the country on a whole) can quantify the magnitude of human brucellosis in mainland China, with these initial estimates conveying important information about the prospects for effective control measures. Our investigation may be also potentially helpful to strategise control for the prevention and surveillance of brucellosis beyond China, in other endemic countries.The following are available online at www.mdpi.com/1660-4601/14/3/295/s1, Table S1: Constant parameters description and values (
+          
+            
+              
+                y
+                e
+                a
+                
+                  r
+                  
+                    −
+                    1
+                  
+                
+              
+            
+          
+        ), Table S2: Values of b, 
+          
+            
+              
+                b
+                h
+              
+            
+          
+         and 
+          
+            
+              
+                d
+                h
+              
+            
+          
+         and 95% confidence intervals (
+          
+            
+              
+                y
+                e
+                a
+                
+                  r
+                  
+                    −
+                    1
+                  
+                
+              
+            
+          
+        ), Table S3: Estimated values of 
+          
+            
+              
+                β
+                s
+              
+            
+          
+        , 
+          
+            
+              
+                β
+                
+                  s
+                  w
+                
+              
+            
+          
+         and 
+          
+            
+              
+                β
+                h
+              
+            
+          
+         with their 95% confidence intervals (
+          
+            
+              
+                y
+                e
+                a
+                
+                  r
+                  
+                    −
+                    1
+                  
+                
+              
+            
+          
+        ). Dynamical behavior.We thank Juan Zhang, Laith Yakob and Thomas Alex Weppelmann for helpful discussions. The project is funded by the National Natural Science Foundation of China under Grants (11331009, 11671241, 11501338, and 11601293), Graduate Students’ Excellent Innovative Item of Shanxi Province No. 2015BY01. Natural Science Foundation of Shanxi Province Grant No. 201601D021002, 131 Talents of Shanxi University, Program for the Outstanding Innovative Teams (OIT) of Higher Learning Institutions of Shanxi.Mingtao Li, Guiquan Sun and Zhen Jin conceived and designed the experiments; Mingtao Li performed the experiments; Mingtao Li analyzed the data; Mingtao Li and Wenyi Zhang contributed reagents/materials/analysis tools; Mingtao Li, Guiquan Sun, Wenyi Zhang and Zhen Jin wrote the paper.The authors declare no conflict of interest.Spatiotemporal distribution of annual human brucellosis cases, by province, in China, 2004–2014.Transmission diagram on the dynamical transmission of brucellosis.Brucellosis model fitting for the annual cases of human brucellosis infection with different Cases. The light grey shaded area shows the 95% confident interval (CI) for all 1000 simulations, and the blue circles mark the reported data for human brucellosis cases. Let 
+            
+              
+                
+                  x
+                  (
+                  t
+                  )
+                
+              
+            
+           represent annual cases of brucellosis infection, and 
+            
+              
+                
+                  x
+                  (
+                  t
+                  )
+                  =
+                  X
+                  (
+                  t
+                  )
+                  −
+                  X
+                  (
+                  t
+                  −
+                  1
+                  )
+                
+              
+            
+          , where 
+            
+              
+                
+                  
+                    
+                      d
+                      X
+                      (
+                      t
+                      )
+                    
+                    
+                      d
+                      t
+                    
+                  
+                  =
+                  σ
+                  
+                    E
+                    h
+                  
+                  
+                    (
+                    t
+                    )
+                  
+                
+              
+            
+          .Brucellosis model fitting for cases of human brucellosis infection in mainland China and 11 selected provinces. The light grey shaded area shows the 95% CI for all 1000 simulations, and the blue circles mark the reported data for human brucellosis cases.Model selection table for mainland China.Estimated values of 
+            
+              
+                
+                  R
+                  
+                    0
+                  
+                  i
+                
+              
+            
+          , 
+            
+              
+                
+                  R
+                  
+                    0
+                  
+                  e
+                
+              
+            
+          , 
+            
+              
+                
+                  R
+                  0
+                
+              
+            
+           and their 95% confidence intervals.Estimates of minimum vaccination coverage rate v, removal rate α, and disinfection frequency l.

Med-MDPI/ijerph_2/ijerph-14-03-00296.txt

@@ -0,0 +1,363 @@
+Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).Introduction: Efforts have been made to reduce HIV/AIDS-related mortality by delivering antiretroviral therapy (ART) treatment. However, HIV patients in resource-poor settings are still dying, even if they are on ART treatment. This study aimed to explore the factors associated with HIV/AIDS-related mortality in Southwestern Ethiopia. Method: A non-concurrent retrospective cohort study which collected data from the clinical records of adult HIV/AIDS patients, who initiated ART treatment and were followed between January 2006 and December 2010, was conducted, to explore the factors associated with HIV/AIDS-related mortality at Jimma University Specialized Hospital (JUSH). Survival times (i.e., the time from the onset of ART treatment to the death or censoring) and different characteristics of patients were retrospectively examined. A best-fit model was chosen for the survival data, after the comparison between native semi-parametric Cox regression and parametric survival models (i.e., exponential, Weibull, and log-logistic). Result: A total of 456 HIV patients were included in the study, mostly females (312, 68.4%), with a median age of 30 years (inter-quartile range (IQR): 23–37 years). Estimated follow-up until December 2010 accounted for 1245 person-years at risk (PYAR) and resulted in 66 (14.5%) deaths and 390 censored individuals, representing a median survival time of 34.0 months ( IQR: 22.8–42.0 months). The overall mortality rate was 5.3/100 PYAR: 6.5/100 PYAR for males and 4.8/100 PYAR for females. The Weibull survival model was the best model for fitting the data (lowest AIC). The main factors associated with mortality were: baseline age (>35 years old, AHR = 3.8, 95% CI: 1.6–9.1), baseline weight (AHR = 0.93, 95% CI: 0.90–0.97), baseline WHO stage IV (AHR = 6.2, 95% CI: 2.2–14.2), and low adherence to ART treatment (AHR = 4.2, 95% CI: 2.5–7.1). Conclusion: An effective reduction in HIV/AIDS mortality could be achieved through timely ART treatment onset and maintaining high levels of treatment adherence.HIV/AIDS continues to be a major global public health issue and thus far, has claimed the lives of more than 34 million people worldwide. In 2014, approximately 1.2 (1.0–1.5) million people died from a HIV-related causes [1]. Sub-Saharan Africa was the most affected region, with 25.8 (24.0–28.7) million people living with HIV in 2014; the region accounts for nearly 70% of new HIV infections globally [1]. In Ethiopia, it has been estimated that approximately 45,200 (36,500–55,200) deaths were related to AIDS and that 793,700 (716,300–893,200) people were living with HIV in 2013 [1]. Effective treatment with antiretroviral (ARV) drugs can control the infection and the disease, and allow HIV-infected people to enjoy healthy and productive lives. Antiretroviral therapy (ART) reduces HIV replication and the infection of new cells, and it improves the immune system function. Therefore, ARV therapy positively influences the quality of life and the survival of seropositive HIV carriers [2].Different clinical, demographic, socio-economic, and behavioural factors have been reported to be related to the survival of HIV-infected patients under ART [3,4]. In Sub-Saharan Africa, studies have reported an important correlation between mortality in HIV-patients under treatment and late diagnosis of HIV, the delay in ART initiation, an advanced World Health Organization (WHO) clinical stage, low CD4 counts, high viral loads, a low body weight, low haemoglobin levels, and poor socio-economic conditions [5,6,7,8]. Gender and migration have also been associated with the risk of dying in seropositive individuals in South Africa [4]. In Ethiopia, factors such as severe anemia, a history of co-infection with tuberculosis (TB), marital status, WHO stage, low CD4 counts, poor adherence to ART, substance use, and opportunistic infections, were also found to be important determinants of HIV/AIDS-related deaths [9,10,11,12].Both semi-parametric and parametric survival models have been used to predict the effects of different potentially influential factors on the time until HIV/AIDS-related death (survival time) [3,9]. Although the Cox regression model has been widely used for this purpose due to its minimal requirements of assumptions for predicting the prognostic factors associated with survival [3,13], parametric models have been demonstrated to be more accurate in creating projections regarding the risk of mortality beyond the observed follow-up period [14,15]. In this paper, we report the results obtained using three of the most popular parametric survival models (i.e., exponential, Weibull, and log-logistic) and one semi-parametric (Cox proportional hazard model) model; these models analyzed non-concurrent retrospective cohort data from patients with HIV/AIDS, who initiated ART treatment between January 2006 and December 2010 at the Jimma University Specialized Hospital (JUSH) in Ethiopia, in order to identify the predictive factors of HIV/AIDS-related mortality.Jimma University Specialized Hospital (JUSH) is one of the oldest public referral hospitals in Ethiopia, which was established in 1930 by Italian invaders to provide medical services to their soldiers. JUSH is located in Jimma City, 352 km southwest of Addis Ababa. According to the census conducted by the Central Statistics Agency of Ethiopia (CSA) in 2007, Jimma City had a total population of 120,960 inhabitants living in an area of approximately 50.5 km2 (population density of 2394.3 inhabitants/km2), with males (50.3%) slightly outnumbering females. Following the Ministry of Health (MoH) policy for the control of the HIV/AIDS epidemic in Ethiopia, JUSH has implemented the HIV/AIDS prevention and control programme since 2002, providing the administration of ART regimens in a separate unit as the main component of the programme. ART regimens in Ethiopia consist of a generic low-cost fixed-dose combination (FDC) of two nucleoside reverse transcriptase inhibitors (NRTIs) and one non-nucleoside reverse transcriptase inhibitor (NNRTI), with first line regimens of lamivudine (3TC) combined with stavudine (d4T) or zidovudine (AZT), and either nevirapine (NVP) or efavirenz (EFV) [16]. The ART regimens in this study included the following: 1a (3TC+D4T+NVP), 1b (3TC+D4T+FFV), 1c (3TC+AZT+NVP), and 1d (3TC+AZT+EFV).All medical cards of HIV/AIDS patients aged ≥ 18 years who initiated ART treatment between January 2006 and December 2010 at JUSH were revised, and available data of potential predictors of survival were collected by trained nurses. Each patient had a chart/record with a distinctive identification number, known as the ART unique identification number.According to the national HIV testing and counselling guideline, HIV patients should be followed-up routinely [17]. Once the HIV diagnosis was confirmed using rapid HIV antibody tests (KHB/STAT/PAK®/Unigold™ tiebreaker algorithm) at JUSH, patients were clinically examined, and laboratory tests such as the CD4 count, total WBC count, haemoglobin measurement, transaminases ALT/AST ratio, and TB screening were performed. During the study period, the viral load measurement was not available at JUSH. Thus, ART onset was primarily based on the HIV disease stage and on the degree of immune damage (CD4 counts). The following criteria were considered to initiate ART at JUSH: WHO Stage 4 disease irrespective of the CD4 cell count, a WHO Stage 3 disease with a CD4 cell count < 350/mm3, and a WHO Stage 1 or 2 disease with a CD4 cell count < 200/mm3. After ART onset, patients were evaluated within the next two weeks, and then every one or two months thereafter, during scheduled medical appointments at JUSH. The evaluation included the assessment of drug side effects, the disease progression, and clinical improvements/deterioration, including the identification of opportunistic infections (such as Pneumonia and TB) or recurrent problems. According to the programme guidelines, when a patient does not turn up for scheduled appointments and/or does not pick up the ART medicines, he/she is contacted by telephone and/or visited at home by health workers.The outcome variables were the survival time in months and the HIV/AIDS-related events (i.e., dead or censored). The survival time was calculated in months, taking into account the dates of onset of ART and the occurrence of the event (death) or censoring. The censoring time was measured for individuals who were on ART until December 2010 or failed to follow-up. The evaluated risk factors were gender (male or female), baseline age, baseline CD4 count, baseline weight in Kg, baseline TB status (negative or positive), baseline opportunistic infection (no or yes), ART regimen (1a, 1b, 1c or 1d), and level of adherence to the ART regimen (low or high). Death was defined as confirmed HIV/AIDS-related death with the certification of death by a medical practitioner, or a verbal or telephone confirmation of death from a relative or friend. High adherence was defined as a 95% adherence based on pill counts at clinic visits, and poor or low adherence was defined as the failure to achieve this criterion [18].The mortality rate in this study is expressed as the number of deaths per 100 people per year at risk (PYAR). While the numerator of the mortality rate is the number of deaths identified during the follow-up period, the denominator is the sum of the total years that each person was followed-up. Kaplan-Meier survival curves were used to estimate the probability of death during the study period. The Log-rank test was used to compare the estimated survival curves according to gender. Semi-parametric (Cox regression) and parametric (exponential, Weibull, and log-logistic) [13] survival models were first applied to identify the best-fitting model for the time-to-HIV/AIDS death data, and then to calculate the hazard ratios (HRs) of the incidence of death.The hazard function at time t for a particular patient with a set of p covariates (x1, x2, … xp) is given as follows [19]:
+        
+          (1)
+          
+            
+              
+                h
+                
+                  (
+                  
+                    t
+                    |
+                    X
+                  
+                  )
+                
+                =
+                 
+                
+                  h
+                  0
+                
+                
+                  (
+                  t
+                  )
+                
+                exp
+                
+                  (
+                  
+                    
+                      β
+                      1
+                    
+                    
+                      x
+                      1
+                    
+                    +
+                    
+                      β
+                      2
+                    
+                    
+                      x
+                      2
+                    
+                    +
+                    …
+                    +
+                    
+                      β
+                      p
+                    
+                    
+                      x
+                      p
+                    
+                  
+                  )
+                
+                =
+                
+                  h
+                  0
+                
+                
+                  (
+                  t
+                  )
+                
+                exp
+                
+                  (
+                  
+                    
+                      β
+                      ′
+                    
+                    X
+                  
+                  )
+                
+              
+            
+          
+        
+        where 
+          
+            
+              
+                
+                  β
+                  j
+                
+              
+            
+          
+         is the estimated parameter for the jth covariate, 
+          
+            
+              
+                
+                  h
+                  0
+                
+                
+                  (
+                  t
+                  )
+                
+              
+            
+          
+         is the baseline hazard function, and x is the vector of the covariates. The baseline hazard function was assumed to follow a specific distribution when a fully parametric proportional hazard model was fitted to the data, whereas the semi-parametric (Cox proportional hazard) model had no such constraint.The graphical evaluation method was used for the appropriateness of the Weibull model. The 
+          
+            
+              
+                log
+                
+                  {
+                  
+                    −
+                    l
+                    o
+                    g
+                    
+                      S
+                      ^
+                    
+                    
+                      (
+                      t
+                      )
+                    
+                  
+                  }
+                
+              
+            
+          
+         versus log(t) line is a straight line when the Weibull distribution is appropriate or reasonable. The exponential regression model is a special case of the Weibull model with a shape parameter equal to 1, which leads to a constant hazard function. For the exponential model, the 
+          
+            
+              
+                log
+                
+                  S
+                  ^
+                
+                
+                  (
+                  t
+                  )
+                
+              
+            
+          
+         versus time plot should yield a straight line [19]. For the log-logistic model, a plot of log(1−
+          
+            
+              
+                
+                  S
+                  ^
+                
+                
+                  (
+                  t
+                  )
+                
+              
+            
+          
+        )/(
+          
+            
+              
+                
+                  S
+                  ^
+                
+                
+                  (
+                  t
+                  )
+                
+              
+            
+          
+        ) versus log(t) with a positive slope “p”, or log(
+          
+            
+              
+                
+                  S
+                  ^
+                
+                
+                  (
+                  t
+                  )
+                
+              
+            
+          
+        )/(1−
+          
+            
+              
+                
+                  S
+                  ^
+                
+                
+                  (
+                  t
+                  )
+                
+              
+            
+          
+        ) versus log(t) with a negative slope ‘p’, should be linear [19,20]. The relevant functions for the different parametric models for this study data set are plotted in Supplementary file Figure S1.The Akaike information criterion (AIC) was used to compare the three parametric models. The model with the lowest AIC value was considered to be the best model for fitting the data. The AIC was calculated using the following formula [19]:
+        
+          (2)
+          
+            
+              
+                A
+                I
+                C
+                =
+                −
+                2
+                *
+                l
+                o
+                g
+                
+                  (
+                  
+                    l
+                    i
+                    k
+                    e
+                    l
+                    i
+                    h
+                    o
+                    o
+                    d
+                  
+                  )
+                
+                +
+                2
+                
+                  (
+                  
+                    p
+                    +
+                    1
+                    +
+                    s
+                  
+                  )
+                
+              
+            
+          
+        
+        where p denotes the number of covariates in the model, not including the constant terms; s = 0 stands for the exponential model; and s = 1 represents the Weibull and log-logistic models.The goodness-of-fit of the semi-parametric and parametric survival models was tested by plotting the cumulative hazard rate against the Cox-Snell residuals of each model. The cumulative hazard rates of the models that fell closer to the referent line were considered to indicate the models with a better adherence to their assumptions. The Cox-Snell residual for the ith individual at observed time ti was defined as [19]:
+        
+          (3)
+          
+            
+              
+                
+                  r
+                  
+                    c
+                    i
+                  
+                
+                =
+                
+                  H
+                  ^
+                
+                
+                  (
+                  
+                    
+                      t
+                      i
+                    
+                    |
+                    
+                      x
+                      i
+                    
+                  
+                  )
+                
+                =
+                −
+                l
+                o
+                g
+                
+                  [
+                  
+                    
+                      S
+                      ^
+                    
+                    (
+                    
+                      t
+                      i
+                    
+                    |
+                    
+                      x
+                      i
+                    
+                    )
+                  
+                  ]
+                
+              
+            
+          
+        
+        where ti is the observed survival time for individual i, xi is the vector of covariate values for individual i, and 
+          
+            
+              
+                
+                  S
+                  ^
+                
+                
+                  (
+                  
+                    
+                      t
+                      i
+                    
+                  
+                  )
+                
+              
+            
+          
+         is the estimated survival function on the fitted value.Ethical approval for the data collection of this study was obtained from the Jimma University Health Science Research Office (Reference number: RPGS/520/2011). The trained nurses collected the data with the administered questionnaire from standard medical registration cards of patients in the record office at the ART unit.The majority of the 456 adult HIV/AIDS patients included in the study who attended JUSH between 2006 and 2010 were female (312, 68.4%) (Table 1), and had initiated ART treatment before the age of 35 (median: 30 years, inter-quartile range (IQR: 23–37 years). The median baseline body weight and CD4 count of the patients were 51.0 kg (IQR: 45.0–57.0 kg) and 151.5 cells/µL (IQR: 90.75–217.2 cells/µL), respectively. The follow-up until December 2010 accounted for 1245 person-years at risk (PYAR) (857 and 388 PYAR for females and males respectively), and resulted in 66 (14.5%) deaths (41 females and 25 males) and 390 censored individuals, representing a median survival time of 34.0 months (IQR: 22.8–42.0 months). The overall mortality rate was 5.3/100 PYAR, with rates for males (6.5/100 PYAR) being higher than for females (4.8/100 PYAR). Forty deaths (60.6%) occurred early in the first year after the onset of ART. Most deaths were in patients that, at the time of the treatment onset, were older than 35 years (45.5%), had a TB infection confirmed (59.1%), and showed WHO clinical III or IV stages (53.1%). Deaths were more likely to occur among individuals who maintained a low adherence to the ART regimen than among those who maintained high adherence (p < 0.001). Deaths during the follow-up period were also more frequent among individuals who began ART in the late clinical stage IV (p < 0.001), compared to those who began ART in WHO stage I.According to the Kaplan Meier curves, the survival of the HIV-patients decreased quickly during the first seven months and then tailed off gradually, to reach its minimum value (75% survival) at the end of the follow-up (month 60th; Figure 1A). Survival was higher in females than in males during nearly the entire study period (except in the final months), but this difference was not significant (p = 0.26) (Figure 1B). The estimated survivorship functions do not reach zero, which indicates that the greatest observed survival time in the study was a censored value.The Weibull regression model was the best model for fitting the data among the parametric models, since it has the lowest AIC value (492). Additionally, this was the model with the best adherence to the model assumptions (cumulative hazard closer to the reference line), as illustrated in the Cox-Snell residuals plots (Figure 2).Figure 3A indicates that the Weibull survival estimates are also in good agreement with the observed survival estimates, while Figure 3B indicates the evolution of a death rate pattern with time. A high risk of death occurred at the beginning of the ART regimen, and this risk decreased with time (decreasing hazard rate; shape parameter < 1).In the multivariate Weibull regression model, the age of the patient at the beginning of the treatment, baseline body weight, baseline disease stage, and adherence to the ART regimen during the follow-up, were significantly associated with the time to HIV/AIDS-related mortality (Table 2). The patients who began ART at ages above 35 years exhibited a significantly higher death hazard than those who began treatment at an age below 25 years (AHR = 3.8, 95% CI: 1.6–9.1), and the baseline weight in kilograms at the beginning of the treatment was inversely associated with the time to death (AHR = 0.93, 95% CI: 0.90–0.97). Additionally, the patients in the advanced stage IV of the disease at the beginning of ART exhibited an increased risk of dying compared to those who began in the early stage I (AHR = 6.2, 95% CI: 2.2–14.2). Moreover, the individuals with low adherence to the ART regimen were more likely to present with fatal events than those with high adherence (AHR = 4.1, 95% CI: 2.5–7.1). Gender, baseline CD4 count, history of TB co-infection, history of opportunistic infections, and type of ART regimen, were not statistically significantly associated with the time until HIV/AIDS-related death.The Weibull survival regression model allowed for a more accurate identification of the risk factors for HIV-related mortality incidence rates in a non-concurrent retrospective cohort of patients with HIV/AIDS, who initiated ART regimen and were followed at Jimma University Specialized Hospital in Ethiopia between January 2006 and December 2010. The death rates decreased with time and were associated with an increased age and an advanced stage of the disease at the beginning of the treatment, as well as with poor adherence to the ART regimen during the study period.Mortality in the first year of follow-up in our study (60%) was consistent with findings of a study in South Omo, Ethiopia, where 62.9% of patients died in the first year after the onset of ART [11]. Other retrospective cohort studies in Northern and North-western parts of Ethiopia have also reported similar death figures, with between 56% and 59% of total individuals dying before completing the first year of treatment [21,22]. High mortality rates during the first months after beginning the ART regimen were also reported through prospective cohort studies in Tanzania and South Africa [4,23], and were strongly associated with anaemia, thrombocytopenia, and severe malnutrition [4,23,24]. In Ethiopia, it was further found that early high death rates mainly occurred in patients with an advanced disease stage [25], as also reported in studies conducted in sub-Saharan Africa countries [26].The mortality rates in our study (5.3/100 PYAR) are slightly lower than those reported in countries like Uganda and South Korea [27,28], but higher than those reported in other regions of Ethiopia (e.g., the Arbaminch city, the Amhara region, and the capital city Addis Ababa) [22,27,29,30]. For instance, the mortality rate estimated at Arbaminch hospital was 9.1 per 100 PYAR [27], and was 3.4 per 100 PYAR at University of Gonder Hospital, in North-western Ethiopia [22].Comparisons of survival models under different distributions of the hazard function provide the best model for fitting the specific data with appropriate inference [31]. In our study, the Weibull survival model exhibited the smallest AIC, indicating its ability to fit the data. Previous longitudinal studies in Australia and England have also recognized the Weibull regression model as the best model for fitting the time until HIV/AIDS-related death data [3,32]. Our findings also agree with a study that compared parametric models for breast cancer survival data in India [14].One of the weaknesses of the semi-parametric Cox model is that it makes the analyst focus on the regression coefficients, without considering the underlying distribution [15]. However, investigators have developed parametric survival models that lead to more precise estimations of survival probabilities and a better understanding of the event evolution during the time of study [3,14,15,28,33]. Hence, parametric models may be superior to semiparametric models in this setting because they allow for explicit modelling of the underlying death risk (baseline hazard) [15].The significance of gender in determining the survival time until death is variable in many studies [7,24,25]. Although our study did not find any association between gender and the survival time until HIV/AIDS-related death, other studies in Ethiopia and abroad have reported that the mortality rate seems to be higher in males than in females [7,26]. Among possible reasons for the gender difference, it has been suggested that female patients tend to know about their HIV status at an earlier stage and begin antiretroviral therapy with better CD4 cell counts relative to males [7].In contrast to other studies conducted in the Northern and Somali region of Ethiopia [10,22,29], our study did not demonstrate an association between baseline TB infection and death hazard rate, possibly because all of the HIV-TB co-infected patients received opportune TB treatment, in concordance with the DOTs treatment guidelines. A study in Northern Ethiopia showed that HIV/AIDS patients who developed TB had shorter survival times than TB-negative patients [29]; and in North-western Ethiopia, a study showed that the presence of a tuberculosis co-infection at ART onset was significantly associated with HIV/AIDS-related mortality [22] (Hazard ratio = 2.91; 95% CI: 2.11–4.02).Our study found that an increased age (>35 years old) was associated with an increased death hazard rate. These results are consistent with findings from a study in China showing a strong association of the age at the beginning of ART treatment with HIV/AIDS-related death [24], but are in conflict with a study conducted in Addis Ababa, the capital of Ethiopia [9], and to other studies revealing no relationship of the age at the onset of ART treatment with the mortality rate [10,21,34]. Moreover, our findings showed that individuals who began the ART regimen with a low baseline weight and who were in the late clinical stage IV exhibited a greater risk of death. Those results support the recommendations to start ART at earlier stages, as previously stated in the Ethiopian Guidelines for the HIV counselling and testing (i.e., in all symptomatic persons at WHO stage IV, irrespective of CD4 cell counts [17]). Previous studies in Ethiopia, Tanzania, and China have also revealed that an advanced clinical stage at the initiation of ART is a significant predictor of mortality in HIV/AIDS patients under ART [3,7,9,24,29]. A study in Tanzania reported that patients who started ART in the advanced disease stage (WHO clinical stage IV) were four times more at risk of dying than early clinical stage patients [7]. Likewise, several studies showed that low body weight at the initiation of ART was significantly associated with HIV/AIDS-related mortality [7,10,21]. People who started ART with weights below 40 kg were dying at a 2.37 times higher rate than people with weights above 60 kg [21]. A study composed of two hospitals in “Shashemene” and “Assela”, in Ethiopia, however, found that the baseline weight was not a significant predicator of mortality [35]. Conversely, the WHO clinical stage at the beginning of ART was a significant predictor of mortality in this study.The adherence level of people who undergo ART treatment was also significantly associated with HIV/AIDS-related death in this study. Patients with low ART treatment adherence had a four- times higher risk of dying compared to those with high adherence. Similar findings were found in studies conducted in the Somali region and the Northern Province of Cameron, reporting that low adherence to ART treatment is a significant predictor of mortality [10,36]. Successful antiretroviral therapy depends on sustaining high rates of adherence (i.e., correct dosage, taken on time, and in the correct manner, i.e., either with or without food). Educating a community regarding HIV testing services is essential for obtaining an earlier diagnosis, earlier initiation of ART, successful enrolment in ART treatment services, and efficient adherence counselling [27]. Complementary studies including quantitative and qualitative methodologies would help to better identify the determinants of the adherence of HIV patients to the ART, as well as the factors that influence the compliance of HIV case-management guidelines by health workers.Our findings should be interpreted considering the nature of the study design. As a non-concurrent retrospective cohort, the study required data on exposure status at a specific earlier time-point (i.e., potential risk factors at the time of treatment onset), but also the identification of outcome events (i.e., death or censored) during the study period. Information on two important potential risk factors was not available in the healthcare records of HIV patients, i.e., JUSH: body mass index (BMI) [8,36,37,38] and viral load [39,40]. While the BMI could not be calculated due to the lack of the registered heights of patients in clinical records, measurements of viral loads were not available at the hospital during the study period. On the other hand, the outcome assessment of HIV patients and time at risk are highly dependent on the accomplishment of scheduled appointments by patients, as well as on the follow-up efforts and their registration on medical records by health workers at JUSH. Follow-up efforts were not quantified in our study, however, health workers at JUSH always attempt to follow the standard guidelines for the case-management of HIV patients, and those guidelines promote an appropriate follow-up.Although study findings should be interpreted in consideration of the study design (i.e., retrospective collection of cohort data at a specialized hospital), our findings revealed high mortality rates in the earlier months after ART onset. The Weibull model was found to be the best fitting parametric model for the HIV/AIDS-related mortality data, allowing for the identification of the following factors associated with the mortality rate: age group older than 35 years, low baseline weight and advanced clinical stage IV at the beginning of ART, and low adherence to ART. The timely onset of ART treatment, and the promotion and monitoring of the adherence to ART treatment, should be important components of HIV/AIDS programmes.The following are available online at www.mdpi.com/1660-4601/14/3/296/s1, Figure S1: Different parametric models for data set in this study.We gratefully acknowledge the financial support and the data obtained from Jimma University. We would also like to thank Jimma University Specialized Hospital ART center for its assistance and help in facilitating the data collection.Dinberu Seyoum Shebeshi conceived and designed the study, performed the data cleaning and statistical analysis, analyzed the data, and drafted the manuscript. Ayele Taye was involved in designing the study and critically reviewed the manuscript. Niko Speybroeck supervised and critically reviewed the manuscript. Jean-Marie Degryse, Belay Birlie, Mulualem Tadesse, Akalu Banbeta, Angel Rosas-Aguirre, Luc Duchateau, and Yehenew Getachew Kifle critically reviewed and approved the manuscript. All authors reviewed and gave input to the subsequent manuscript drafts.The authors declare that they have no conflicting interests.(a) Kaplan-Meier survival function curve for all individuals; (b) Kaplan-Meier survival function curve by gender with Log-rank test p-value.Cox-Snell residuals plot (Black line is the cumulative hazard, and Red line is the reference line with slope=1.0 and intercept=0) to evaluate the model fits of the semi-parametric and parametric survival models.(a) The observed and estimated proportion of alieved individuals based on parametric survival models; (b) The hazard rate of the selected Weibull survival model with the shape and scale parameters.Summary of HIV/AIDS-related mortality by different baseline characteristics of adult HIV patients included in the study, Jimma University Specialized Hospital, Southwest Ethiopia.Multivariate analysis of factors associated to HIV/AIDS mortality in southwestern Ethiopia: Parameter estimate with the standard error, acceleration factor, and 95% confidence interval using the four potential survival models.AHR: Adjusted Hazard Ratio; 
+        
+          
+            ϕ
+          
+        
+      : Acceleration factor, it relates the factor with the survival time in log-logistic survival model; Significance at p-value < 0.05.

Med-MDPI/ijerph_2/ijerph-14-03-00297.txt

@@ -0,0 +1,80 @@
+Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).Yixing, known as the “City of Ceramics”, is facing a new dilemma: a raw material crisis. Cadmium (Cd) exists in extremely high concentrations in soil due to the considerable input of industrial wastewater into the soil ecosystem. The in situ technique of diffusive gradients in thin film (DGT), the ex situ static equilibrium approach (HAc, EDTA and CaCl2), and the dissolved concentration in soil solution, as well as microwave digestion, were applied to predict the Cd bioavailability of soil, aiming to provide a robust and accurate method for Cd bioavailability evaluation in Yixing. Moreover, the typical local cash crops—paddy and zizania aquatica—were selected for Cd accumulation, aiming to select the ideal plants with tolerance to the soil Cd contamination. The results indicated that the biomasses of the two applied plants were sufficiently sensitive to reflect the stark regional differences of different sampling sites. The zizania aquatica could effectively reduce the total Cd concentration, as indicated by the high accumulation coefficients. However, the fact that the zizania aquatica has extremely high transfer coefficients, and its stem, as the edible part, might accumulate large amounts of Cd, led to the conclusion that zizania aquatica was not an ideal cash crop in Yixing. Furthermore, the labile Cd concentrations which were obtained by the DGT technique and dissolved in the soil solution showed a significant correlation with the Cd concentrations of the biota accumulation. However, the ex situ methods and the microwave digestion-obtained Cd concentrations showed a poor correlation with the accumulated Cd concentration in plant tissue. Correspondingly, the multiple linear regression models were built for fundamental analysis of the performance of different methods available for Cd bioavailability evaluation. The correlation coefficients of DGT obtained by the improved multiple linear regression model have not significantly improved compared to the coefficients obtained by the simple linear regression model. The results revealed that DGT was a robust measurement, which could obtain the labile Cd concentrations independent of the physicochemical features’ variation in the soil ecosystem. Consequently, these findings provide stronger evidence that DGT is an effective and ideal tool for labile Cd evaluation in Yixing.Yixing, an ancient “City of Ceramics” and a modern industrially developed city, is facing a new dilemma: serious soil ecosystem contamination. Cadmium (Cd) is relatively rare in the Earth’s crust; however, due to electroplating, wastewater irrigation, mining, smelting and the abuse of pesticides and fertilizers, the Cd contamination in soils seems increasingly severe in Yixing [1,2,3]. Cd could cause serious nervous system damage to humans through bioaccumulation and bioamplification in members of the food chain, such as cereals and vegetables [2,3]. In view of this, Cd has been classified as a Group 1 human carcinogen by the International Agency for Research on Cancer (IARC) [2,3,4,5]. Due to its high mobility and widespread occurrence, it is of great urgency to monitor the Cd bioavailability in the soil ecosystem of Yixing.At present, there exist numerous methods for Cd bioavailability evaluation: the microwave digestion, dissolved concentration in soil solution, as well as the traditional extractants, i.e., acid extractant HAc, chelat extractant EDTA and neutral extractant CaCl2 [4,5,6]. All of these methods, as the ex situ approaches, were widely applied due to their cheapness and simplicity [3,4,5,6]. However, many investigations indicated that not all of the Cd fractions have toxicity to biota; its bioavailability and mobility depend on its specific forms of fractions [6,7]. Previous research had traditionally divided cadmium into three categories: (1) effective form; (2) potential effective form; and, (3) inert form [6,7,8]. Although the different components could interconvert into each other in certain conditions, the bioavailability of Cd in the ion phase which belongs to the effective form is 20-fold higher compared to the Cd in the complex phase that belongs to the potential effective form [9,10,11,12,13]. Consequently, Zhang et al. [14] revealed that the ex situ measurement always neglected the process of plant uptake. All these approaches were static measurements and sabotaged the original growing environment of plants [15]. These methods involved just a simple step which extracted the metal fractions by applying different intensity reagents; the extracted fractions do not have an inevitable immanent connection with the bioavailable components. Previous investigations have indicated that the physicochemical parameters, i.e., redox potentials variation and the pH variation, could significantly influence the fate and the labile of the Cd fractions in the soil ecosystem [15,16,17]. Consequently, optimizing the in situ, dynamic method to simulate the plant uptake for soil Cd bioavailability monitoring is of significant importance.The diffusive gradients in the thin film (DGT) technique, as the passive sampling technique, have drawn the attention of many researchers who have seen this technique as a determining mechanism since its creation [16,17]. Based on the Fick’s first law, DGT monitored the diffusion of the dissolved species of Cd through adding a membrane-diffusive layer, which could also control and restrict the flux accumulated in an ion-exchange resin [14,16]. The membrane filter used to protect the inner diffusive gel and resin gel builds a federation, and such a joint federation could form the diffusive layer. The Cd concentration measured by DGT depends on the thickness of the diffusive layer (Δg, 0.93 cm), its diffusion coefficient of Cd2+ (D, the value obtained from the http://www.dgtresearch.com/) in the diffusive layer, the exposure window area (A, 3.14 cm2), the deployment time (t, 86,400 s) and the accumulated mass of Cd2+ over the deployment time (M, ng) [15,16,17].
+      
+        (1)
+        
+          
+            
+              
+                C
+                
+                  D
+                  G
+                  T
+                
+              
+              =
+               
+              
+                
+                  M
+                  Δ
+                  g
+                
+                
+                  D
+                  A
+                  t
+                
+              
+            
+          
+        
+      The CDGT-obtained Cd fractions not only included the dissolved amount in the soil concentration, but also contained the components diffused from the further solution and the fractions released from the solid phase; all these components constituted the labile fractions of the biota. Consequently, the CDGT could be considered as a chemical surrogate for the biota. Moreover, it could eliminate the accidentalia because its measured concentration was the mean concentration of elements during the deployment time. In view of the distinct feature of the dynamic process, the DGT-measured concentration can truly reflect the bioavailability of elements because of the different ingredient of the resin gel. The related study showed that the AG50W-X cation-exchange resin gel could be applied for Cs and Sr evaluation [18]; silver iodide distributed in the resin gel could adsorb sulphide fractions [19]; the resin gel containing ferrihydrite could monitor the labile phosphorus [17]; the resin with thiol groups could fix the flux of Hg [20,21,22]; and the resin with common chelex-100 particles could be applied to predict the bioavailability of bivalent and trivalent metal ions toxicity [23].There existed good correlations between the DGT-measured concentrations and the accumulated concentrations in plant tissues; however, researchers have not reached a consensus on the good performance of the DGT technique. A number of researchers contend that DGT-measured concentrations of Cd in soils were not well correlated with Cd concentrations in wheat [24], lettuce [25] and ryegrass [26] and the previous experimental results indicated that the DGT was not a robust tool for evaluating the Cd bioavailability; its measured labile concentrations were easily influenced by the species of accumulated biota and the physicochemical properties of soils [27,28,29]. As we know, different growth periods accounted for the different rate of uptake, but the DGT mimic the uptake processes mechanically. The optimized results achieved by DGT probably change according to the mean rate of biota uptake. Therefore, further investigation of the DGT for Cd bioavailability in the extreme soil ecosystem such as Yixing is essential to environmental research.Five traditional ex situ methods including soil solution concentration, microwave digestion Cd concentration and three chemical extraction methods (chelating extractant-EDTA, acid extractant-HAc and salt solutions-CaCl2), as well as the in situ technique of DGT, were applied as the Cd evaluation tool in this study. The typical cash crops in Yixing—paddy and zizania aquatica—were supplied for the selection of the plant with tolerance to labile Cd fractions. Linear correlations between the different indicator-measured Cd concentrations and plant-accumulated Cd concentrations were used to identify the best-suited indicator for Cd evaluation in this study. Moreover, the multiple regression models were also applied to select the ideal method, which is independent of soil physicochemical properties’ variation, for Cd bioavailability evaluation. The objective of this study is to select the suitable cash crop with high tolerance to Cd contamination. Furthermore, we aim to provide an ideal and robust tool for Cd contamination evaluation in Yixing.Yixing is a city famous for its special Zisha ceramic products. However, due to the anthropogenic impact on the soil environment, the raw material of Zisha might be significantly affected by Cd pollution. To gain an overall understanding the Cd contamination, samples were collected from 12 sites in Yixing, Jiangsu China (Figure 1). Sampling sites 1 and 2 were located in Yifeng town, where the soil ecosystem might have little influence of Cd contamination. Consequently, sampling sites 1 and 2 were selected as the control group. The other sampling sites were located in the town of Dingshu. Sampling sites 3 and 4 were located in the Yangan, which has many non-ferrous metal smelters. Sampling sites 5, 6 and 7 were the field for the paddy-grown soil. Sampling sites 8–12 are located in the Fangxi, where there are many pottery workshops. Moreover, sampling sites 11 and 12 were the fields where zizania aquatica was grown. The collected soil samples were frozen under −80 °C in the ultra-low temperature freezer, then freeze-dried by the vacuum freeze dryer (B16-333-8811, LABCONCO Ltd., Kansas, USA), and finally sieved with 2 mm stainless steel mesh. Prior to the pot experiment, the total Cd for each crop was digested by the commonly used aqua regia method and determined by flame atomic absorption spectrophotometry (Z-81001, Hitachi Ltd., Hitachi, Japan) with the method detection limit (0.002 mg·L−1 Cd). Quality control of the analytical method was conducted every ten samples using a certified standard solution to ensure accuracy and precision with experimental errors (<5%). The physiochemical properties of the samples are showed in Table 1.To ensure the repeatability and accuracy for Cd evaluation, two typical cash crops—paddy and zizania aquatica—were selected as the supplied plants. The fifteen seeds of paddy and six seeds of zizania aquatica were sown in the pots filled with 0.75 kg of the collected soils and three replicate tests were carried out for every soil site, and 55% and 75% maximum moisture content were kept in paddy- and zizania aquatica-grown pots, respectively. After germination, in order to keep an equal number of seedlings per pot, we reduced the number of seedlings to 10 paddy and four zizania aquatica seeds, respectively. All plants were under the natural day–night cycle and grown with no nutrient addition in a greenhouse. During the plant growth process, deionized water was added every day to maintain the soil moisture to reach the 55% and 75% maximum moisture content for paddy- and zizania aquatica-grown soil, respectively. Given that the growth of zizania aquatica requires large quantities of water, the pots with the zizania aquatica-grown soil were hydrated twice a day. After four weeks of growth, smut fungus was applied to stimulate the fresh stem of zizania aquatica. After eight weeks of growth, all plants were harvested and separated into shoots and roots. The collected shoots of the plants were rinsed with tap water and further cleaned with deionized water in order to sweep away the fine particles adsorbed on the root surface. Then, the roots were soaked in the 20 mmol∙L−1 EDTA for 15 min and then washed with deionized water; all these procedures were to exclude the interference of Cd fractions adsorbed onto the root surface. The shoots and roots of the plants were dried in an oven at 70 °C for 4 h to wipe off chlorophyll, after which the temperature was reduced to 50 °C to a constant weight. Dry weights were successively recorded. Cadmium concentrations in plant tissues were determined by flame atomic absorption spectrophotometry (Hitachi Z-81001, Hitachi Ltd., Hitachi, Japan) with the method detection limit (0.002 mg·L−1 Cd). To ensure accuracy and precision with experimental errors (<5%), quality control of the analytical method was conducted on ten samples at a time using a certified standard solution. After harvest, the remaining soils were air-dried at room temperature and sieved with 2 mm stainless steel mesh for the following analysis of various parameters.Single extraction methods: Three widely used single extraction methods were selected to measure the bioavailable Cd fractions in soil pools because of their different characters and extracted intensity. The extractants were 0.11 mol·L−1 HAc [27,29], 0.05 mol·L−1 EDTA and 0.01 mol·L−1 CaCl2 [28,30]. Among the three methods, the 0.11 mol·L−1 HAc was the first step in a three-step sequential extraction procedure recommended by the European Community Bureau of Reference (BCR) [27,31,32,33]. EDTA is a commonly used chelating agent with a strong bonding force, which can complex with most metal ions. CaCl2 is a typical neutral agent, which, based on the principle of ion exchange, was used to measure the target elements.All extraction procedures were conducted in triplicate. The extracted solutions were centrifuged at 3000× g for 20 min at 25 °C (avoid the extracted process overheating due to the high-speed rotation). The supernatants were filtered with the strainer and then transferred to leach solution in 10 mL centrifuge tubes. The solution was acidified with nitric acid and stored in a refrigerator at 4 °C prior to analysis. The detailed operations of the three main procedures for metals extraction are listed in Table 2. The procedure just described is shown as the first listed method (HAc1).Total Cd concentration in soil: The total Cd concentration in soil was determined by the microwave digestion of aqua regia according to the published methods [29,30,31]. The total amount of Cd in soil solutions was measured using atomic absorption spectrophotometry (Hitachi Z-81001, Hitachi Ltd., Hitachi, Japan).The dissolved Cd concentration in soil solution: The concentrations of Cd in soil solutions (Csol) were measured according to the traditional centrifugation method. The soil solution was obtained at the 80% maximum moisture content, and the soil solution was collected by centrifuging (10,000× g) the paste soils for 20 min at 25 °C. To ensure the purity of the analytes, the supernatants were filtered through a 0.45-μm pore size cellulose nitrate filter membrane and then were acidified by nitric acid [31,34,35]. The Cd concentrations in soil solutions were measured using atomic absorption spectrophotometry (Hitachi Z-81001, Hitachi Ltd., Hitachi, Japan).DGT technique: Figure 2 is the schematic view of a DGT device applied in the soil. The piston-type DGT device assembled from bottom to top consisted of a plastic base, a resin gel, a diffusive gel, a protective membrane filter and a plastic cap. The exposure window with the determined soils is 2 cm in diameter (the plastic base and cap were purchased from DGT Research Limited, Manchester, UK). The diffusive gel (0.8 mm thickness) was made by adding 15% acrylamide and 0.3% agarose-derived cross-linker following a published procedure [15,16,17,34]. The resin gel (0.4 mm thickness) was made by inserting Chelex-100 into diffusive gel. In order to protect the inner gel, in the DGT assembly, a 0.13 mm cellulose-nitrate filter membrane (0.45-µm pore size; Whatman, Maidstone, UK) was placed above the resin gel [14]. The application of DGT in soils could be simply divided into three steps as follows [36]:Pretreatment of the soil subsample: Firstly, the maximum water holding capacity (MWHC) of the soil subsample was measured under specified temperature conditions. The subsample was then placed in a 100 mL plastic pot while, at the same time, the soil was kept at 60% MWHC using deionized water under a fixed saturation for 48 h. Then, moisture content was modulated up to 80% MWHC for 24 h before DGT deployment.Deployment of DGT device: To ensure complete diffusion between the soil paste and DGT device, the top of the DGT was gently pressed to guarantee the touch area of the DGT device. The assembled DGT devices were deployed for 24 h at 25 ± 1 °C during the deployment of the DGT, and the temperature and moisture content were controlled at the constant state (Figure 3).Retrieval and elution of DGT: Filter membranes were given 24 h to accumulate soil particles, which were removed by deionized water. After carefully disassembling the device, resin gels were transferred into a micro vial filled with 1 mL of 1 mol·L−1 nitric acid for at least 16 h. The concentrations of cadmium in the eluent were measured by atomic absorption spectrophotometry (Hitachi Z-81001).The relationships between various bioavailable indicators of Cd measured by the chosen methods and the concentration of Cd in plant tissues were investigated using Word. Statistical analyses were performed using the SPSS statistical package (version 10.0 for Windows, IBM, New York, NY, USA). The multiple regression models were established to describe the performance of the indicators.As Figure 4 showed, the paddy and zizania aquatica seemed to have significant differences in their biomass. The biomass of paddy has an extremely obvious variation in the different sampling sites. However, the biomass of zizania aquatica seemed not to be influenced by the Cd contamination concentration. The zizania aquatica, as the widely cultivated vegetable, has the stringent requirement on the water quantity compared to the paddy. Consequently, in theory, the zizania aquatica might show inhibited performance in biomass due to more leaching mass of the labile Cd fractions compared to the paddy. The zizania aquatica indicated that the tolerance to Cd contamination might be due to its extremely developed root and abnormal growth stem. The zizania aquatica biomass of roots showed significant superiority for shoots during generation and the seedling period, but its fresh stem will be enlarged by infected smut fungus at the last phase of the seedling period. All these phenomena arise in the extremely large biomass of zizania aquatica shoots compared to zizania aquatica roots, as well as the paddy shoots and roots. However, the investigation by Yao et al. (2015) indicated that the biomass of typical terrestrial plants, wheat and maize were significantly restrained by the additional Cd in the Cd-contaminated soil (Figure S1 in the Supplementary Information) [37]. The Cd fractions applied in the simulation experiment showed stronger bioavailability in the biota compared to that in Yixing. Due to the strong aging effect on the Cd-contaminated soil in the simulation experiment, the role of the biota auxin and the soil organisms and microorganisms, which leeched into biota growth, merited our attention [17,35]. Further investigation will support this speculation.To demonstrate the accumulation effects of the zizania aquatica on the Cd-contaminated soil, the accumulation coefficient (the ratio of plant-accumulated Cd concentration to soil total Cd concentration) and the transfer coefficient (the ratio of shoot-accumulated Cd concentration to root-accumulated Cd concentration) of paddy and zizania aquatica were exhibited in Figure 5. The accumulation coefficients of paddy and zizania aquatica varied from 0.23 to 0.52, and 0.31 to 1.02, respectively. This phenomenon demonstrated that zizania could effectively accumulate the Cd fractions and reduce the total Cd concentrations in soil. Furthermore, the zizania aquatica tend to transport the Cd fractions upward from paddy, which is indicated by the zizania aquatica transfer coefficients from 2 to 5.38, compared to that of paddy from 1.1 to 1.47. The zizania aquatica showed an extremely outstanding performance in Cd contamination improvement. However, due to the extremely high transfer coefficients of the zizania aquatica, the zizania aquatica stem, as the typical edible cash crop, might have a potential risk to human health through bioaccumulation and biomagnification. Consequently, the zizania aquatica could not be selected as the accumulation biota for Cd-contamination improvement. Moreover, the zizania aquatica is not recommended as the main cash crop in Yixing.The shoot was selected for Cd accumulation because the shoots were easier to collect compared to the roots. Furthermore, since the shoot was the edible part of paddy and zizania aquatica, it might have a direct relation with human health. The relationship between the different indicator-measured Cd concentration and Cd accumulation concentration in the shoot of the paddy and zizania aquatica were shown in Figure 6 and Figure 7, respectively. The DGT-measured Cd concentration and the total Cd concentration in soil solution showed a significant linear correlation with the accumulated Cd concentration in plant shoots. However, the EDTA, HAc and CaCl2 measured Cd, as well as the total Cd concentration in soil, seemed to have no relationships with the Cd accumulation concentration in plant tissues.It has been recognized that the DGT technique only measured labile species, and excluded kinetically inert organic species, large colloids and strong organic-metal complexes [25,26,38]. As a dynamic, biota simulation tool, it could obtain Cd concentration including the fractions in pore water and those related to the dynamic resupply of elements from complexes in soil solutions and solid phases [26,39]. Consequently, DGT can be used as an ideal tool for Cd bioavailability evaluation in soils. The soil solution also showed obvious advantages compared with the other ex situ measurements. Due to the fact that plant roots directly take up mineral elements from the soil solution, the concentration of Cd in soil solution was considered a good indicator for plant availability prediction [25,26,27,39,40]. Previous studies showed that total dissolved Cd in soil solutions is a more direct means of estimating the potential for Cd uptake by plants [26,41,42,43]. Similar studies have reported that the concentration of Pb, Zn and Ca in soil solution could reflect the bioavailable fractions for plants uptake [41,42,43,44]. Accordingly, the Cd concentration in soil solution as a typical ex situ approach might be feasible to evaluate the bioavailable Cd fraction in soil.In this study, given the poor performance of the single extraction methods for the ex situ measurements, they could not reflect the bioavailable level of Cd. Among them, EDTA, as a chelating agent, extracted the largest amounts of Cd from soils compared with the other two extractants (HAc and CaCl2); it extracted target elements including the organically bounded fractions and components in oxides or secondary clay minerals [26,40]. HAc indicated a larger Cd extraction ability than CaCl2; HAc-extracted Cd was the mixture of organic matter-bounded Cd and calcium carbonate/minerals-bounded Cd. CaCl2 as the neutral extractants, based on the ion-exchange principle, measured the target elements, but the exchange capacity was limited by the concentration of the Ca2+, and it was easily affected by the soil texture [41,45,46,47,48,49]. We came to the conclusion that single extraction methods have laid the foundation for analysis error, due to their extraction intensity and extraction time; moreover, the concentration of the extractant restrained the empirical value, and led to the mismatching of the labile component in the soil. In addition, the static extraction could not establish an equal relationship with the dynamic uptake process. Consequently, all these deficiencies hindered the accuracy analysis of ex situ methods for Cd bioavailability evaluation.The ex situ technique-measured Cd indicated a significant correlation with the accumulated Cd concentration in the plant tissue in previous simulation experiments (Table S1 in the Supplementary Information) [37]. Excluding the ageing effect, the supplied biota, paddy and zizania aquatica might result in the poor performance of the ex situ measurement. Accordingly, the traditional ex situ measurement has non-ideal performance in Cd-contaminated soil, which has the cultivation process of the Cd tolerance biota.Upon further analysis of the pros and cons of the selected methods, multiple regression models were applied for the optimization [49,50]. Physicochemical properties which might affect the process of plant uptake in both the solid phase and solution, including pH, OM, CEC, and soil texture were selected as the index for analysis. In order to simplify the soil properties from multidimensional to lower-dimensional parameters, principal components analysis (PCA) was used before establishing the multiple regression models. Regarding the mechanical composition of the soil, with the particle size of clay, it has economic value for Zisha products. In addition, compared to other particle sizes, the clay soil particles were effectively proportioned to bind to metals. Consequently, the silt and sand proportion were neglected in PCA.Firstly, the variance of the data set with interrelated variables (pH, DOC, CEC, clay proportion, the concentration of the TP, TN and K as well as the labile Zn and Pb concentrations in the soil ecosystem) was classified to the independent variables which were called principal components (PC); then, taking the eigenvalues >1 as the extraction criterion, two PCs were extracted. The first (PC1) and second PC (PC2) accounted for 69% and 27%, respectively, and this simplification could explain 90% of the total variance of the analysis data. The factors of PC1 for CEC, OM and clay proportion were 0.811, 0.675 and 0.524; the factors of PC2 for pH, TP and labile Zn were 0.414, 0.219, and 0.792. The CEC, OM and clay proportion were obviously correlated with the PC1 which was representative of the “organic matter”; likewise, the pH, TP and labile Zn which primarily correlated with PC2 were representative of “inorganic matter”. More specifically, the “inorganic matter”, as the competing constituent, interfered in the labile fractions complex of the “organic matter” during the uptake process of the plant, and these mechanisms dictated the bioavailability and liquidity of Cd.Multiple regression models were built to explain how PCs influence the relationships between various bioavailable indicators and the accumulations of Cd by wheat and maize. The designated bioavailable indicator and the PCs were chosen as the input and the Cd concentration in the plant tissue of wheat or maize was chosen as the output. The performance condition of the DGT technique, soil solution, HAc, CaCl2, EDTA and total amount were intuitively reflected through the regressions (Equations (1)–(12)). The results of the evaluation indicated that wheat-grown and maize-grown soil of DGT were significantly influenced by PC1 and PC2 (Equations (1) and (7)); the traditional chemical methods were only influenced by PC2 (Equations (2)–(6) and (7)–(12)). The investigated results showed that the correlation coefficients of paddy-grown soil obtained by HAc, CaCl2, EDTA, soil solution and total amount concentration were from 0.101, 0.034, 0.076, 0.791 and 0.018 and rose to 0.51, 0.46, 0.42, 0.83 and 0.56, respectively; the zizania aquatica-grown soil obtained by HAc, CaCl2, EDTA, soil solution and total amount concentration were from 0.087, 0.051, 0.083, 0.808 and 0.135 and rose to 0.33, 0.45, 0.42, 0.86, and 0.59, respectively. The intuitive understanding of the numerical amounts, with the aid of the models, took the principal impacts of plants’ uptake into account, so the multiple regressions were considered to embrace more variance compared with the simple regression. The results indicated that the correlation coefficients of HAc, CaCl2, EDTA, soil solution and total amount concentration of paddy-grown and zizania aquatica-grown soil all showed marked increases. Taking the complex physicochemical properties into account, the phenomenon of the great improvement in the correlation coefficients between the traditional ex situ Cd bioavailability indicators and accumulated Cd in biota shoot might be the result of the weakened disturbance factors. The ex situ approaches could not reflect the influence of the relevant properties in the soil ecosystem; they are just the embodiment of the equilibrium state of the components of target elements. In addition, the multiple regression models could eliminate the disadvantages of the ex situ measurements, and simplify and optimize the correlation between the indicators and bioaccumulation. However, the DGT-measured Cd did not seem to be optimized after the multiple regression models. The mobile Cd fractions transferred from the soil surface to pore water, through the DGT membrane filter and diffusive gel, then bound to the resin gel during the DGT device accumulation. Meanwhile, the physicochemical properties relevant to this dynamic transfer process, which occur during uptake were embraced by the DGT accumulation. Accordingly, the DGT technique was the simulation process of plant uptake. Consequently, the values after the optimization by the multiple regression models for DGT show no excellent performance; this phenomenon also demonstrates that the DGT technique was a robust tool for Cd bioavailability evaluation in the Cd-contaminated soil of Yixing.
+		
+		paddy-Cd = 0.13DGT + 0.353(I) + 0.167(II) + 0.83 R2 = 0.89**
+		(1)
+		
+		
+		paddy-Cd = 0.71HAc − 0.638(II) + 2.723 R2 = 0.51*
+		(2)
+		
+		
+		paddy-Cd = 0.241EDTA − 0.701(II)+2.42 R2 = 0.46*
+		(3)
+		
+		
+		paddy-Cd = 2. 12CaCl2 − 0.615(II) + 1.55 R2 = 0.42*
+		(4)
+		
+		
+		paddy-Cd = 0.22total − 0.847(II) + 0.985 R2 = 0.56*
+		(5)
+		
+		
+		paddy-Cd = 0.1soil solution − 0.39(II) + 1.416 R2 = 0.83**
+		(6)
+		
+		
+		zizania-Cd = 0.037DGT + 0.33(I) + 0.723(II) + 0.158 R2 = 0.93**
+		(7)
+		
+		
+		zizania-Cd = 0.665HAc − 1.456(II) + 7.857 R2 = 0.33*
+		(8)
+		
+		
+		zizania-Cd = 10.072 − 0.843EDTA − 1.95(II) R2 = 0.41*
+		(9)
+		
+		
+		zizania-Cd = 16.83CaCl2 − 2.818(II) + 14.26 R2 = 0.45*
+		(10)
+		
+		
+		zizania-Cd = 16.243 − 1.154total − 2.18(II) R2 = 0.59*
+		(11)
+		
+		
+		zizania-Cd = 0.068soil solution + 0.589(II) + 0.56 R2 = 0.86**
+		(12)
+		The previous investigation also revealed the advantages of DGT for Cd bioavailability evaluation. Nolan et al. [11], Zhang et al. [14] and Tian et al. [50] have demonstrated the good performance of Cd on Lepidium sativum; Zn on wheat; and Cu on Lepidium heterophyllum of DGT. They indicated that DGT could fully dynamically reflect the actual condition as well as the resupply intensity of metals attached on the solid surface, which then diffused to pore water. Furthermore, all these advantages were precisely what traditional ex situ measurements could not achieve. Compared with other in situ measurement techniques, the technique of DGT was also shown to be superior in measuring Cu bioavailability in an evaluation of barley [17]. Luo et al. [47] also found that DGT predicted Cd toxicity better than either free Cu2+ activity or soil solution. However, as the DGT technique could not embrace all the processes of plant growth and neglected the rhizospheric microorganism, we thought that the DGT technique may sometimes fail to accommodate all the factors that influence plant uptake [24,26]. Further investigation will demonstrate the performance of DGT for different growth periods.This study showed that the typical local cash crop, zizania aquatica, was tolerant to and could effectively reduce the total Cd concentration, which was indicated by the high accumulation coefficients. However, owing to its extremely high transfer coefficient, it could not be selected as the accumulation biota for Cd-contamination improvement. Moreover, the zizania aquatica, as the main cash crop in Yixing, might be an unreasonable agricultural structure.Furthermore, the DGT technique showed better performance for the Cd bioavailability prediction in paddy and zizania aquatica soil compared to the typical ex situ methods i.e., soil solution, HAc, EDTA, CaCl2 extractants and microwave digestion. The DGT measurement embraced the principle that physicochemical properties components influence the labile Cd fractions, as indicated by the inconspicuous improvement of the correlation coefficient obtained by multiple regression models compared to linear regression models. Consequently, the DGT technique might be a promising and robust tool for Cd bioavailability evaluation in Yixing. As an in situ, dynamic technique, DGT performed excellently for the time-weighted average concentrations obtained through the pre-concentration process, independent of the physicochemical properties. The advantages of the DGT technique were noticeable in this study. However, quantification of the target labile fraction is not straightforward during accumulation by the DGT device. The DGT process was not sufficiently sensitive to reflect the rhizosphere microorganism and the organism’s effects due to the process’s comparatively long deployment time for the target element in a low-contaminated ecosystem. Consequently, the accurate accumulation of DGT for extremely low concentrations of the target elements should be improved.The following are available online at www.mdpi.com/1660-4601/14/3/297/s1, Figure S1: The biomass of wheat (A) and maize (B) grown in soils added with different levels of Cd. Each value is the mean ± SD (standard deviation, n = 3), Table S1: Linear correlation coefficients (r) between Cd concentrations in the plant tissues and bioavailable concentrations of Cd measured by eight methods in soils.We are grateful for the grants for the project supported by National Science Funds for Creative Research Groups of China (No. 51421006), the Key Program of the National Natural Science Foundation of China (No. 91647206), the National Key Plan for Research and Development of China (2016YFC0502203), Program for Changjiang Scholars and Innovative Research Team in University (No. IRT13061), the Fundamental Research Funds for the Central Universities (2017B40414), Jiangsu Province Ordinary University Graduated students Scientific Research Innovation Plan (No. B1605332) and PAPD.The following statements should be used “Yu Yao and Peifang Wang conceived and designed the experiments; Yu Yao and Lingzhan Miao performed the experiments; Jun Hou and Chao Wang analyzed the data; Yu Yao contributed reagents/materials/analysis tools; Yu Yao wrote the paper.The authors declare no conflict of interest.The distribution of sampling sites, (a), (b) were the distribution of the sampling sites and the (c) was the located of the Yixing.Schematic view of a diffusive gradients in thin film (DGT) device.Schematic view of DGT deployment in soil.The biomass of paddy (A); and zizania (B) in the 12 sampling sites of Yixing.The accumulation coefficients and transfer coefficients of paddy (A); and zizania (B).Correlation between Cd concentrations in plant tissues and bioavailable concentrations of Cd measured by six methods in paddy-grown soils.Correlation between Cd concentrations in plant tissues and bioavailable concentrations of Cd measured by six methods in zizania-grown soils.The basic physical and chemical properties of sampled soils.The MC, OM and CEC represent the moisture content, organic matter and cation exchange capacity, respectively.The procedures of the three extraction methods adopted in this study.

Med-MDPI/ijerph_2/ijerph-14-03-00298.txt

@@ -0,0 +1 @@
+Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).Limited information is available on the perceptions of stakeholders concerning the health co-benefits of greenhouse gas (GHG) emission reductions. The purpose of this study was to investigate the perceptions of urban residents on the health co-benefits involving GHG abatement and related influencing factors in three cities in China. Beijing, Ningbo and Guangzhou were selected for this survey. Participants were recruited from randomly chosen committees, following quotas for gender and age in proportion to the respective population shares. Chi-square or Fisher’s exact tests were employed to examine the associations between socio-demographic variables and individuals’ perceptions of the health co-benefits related to GHG mitigation. Unconditional logistic regression analysis was performed to investigate the influencing factors of respondents’ awareness about the health co-benefits. A total of 1159 participants were included in the final analysis, of which 15.9% reported that they were familiar with the health co-benefits of GHG emission reductions. Those who were younger, more educated, with higher family income, and with registered urban residence, were more likely to be aware of health co-benefits. Age, attitudes toward air pollution and governmental efforts to improve air quality, suffering from respiratory diseases, and following low carbon lifestyles are significant predictors of respondents’ perceptions on the health co-benefits. These findings may not only provide information to policy-makers to develop and implement public welcome policies of GHG mitigation, but also help to bridge the gap between GHG mitigation measures and public engagement as well as willingness to change health-related behaviors.China is facing major challenges from both climate change and air pollution. Over the past 100 years, China has experienced noticeable climate changes, with the annual average air temperature increasing by 0.5–0.8 °C, a trend that is projected to intensify in the future [1]. The World Health Organization (WHO) has estimated, considering only the well understood impacts of dangerous climate change, and assuming continued progress in economic growth and health protection, that climate change is likely to cause approximately 250,000 additional deaths annually around the world between 2030 and 2050 [2]. According to the Global Climate Risk Index 2016, in 2014, China was identified as the 18th most affected country by weather-related extreme events among the total of 187 countries, and the ranking was 31th during the period 1995–2014 [3]. For air pollution, the Asian Development Bank reported that fewer than 1% of the 500 largest cities in China meet the air quality standards for PM2.5 recommended by the WHO, and seven Chinese cities are ranked among the 10 most polluted cities around the world [4]. In 2004, more than three-quarters of the Chinese urban population was exposed to air that did not meet the national air quality standards (GB3095-1996) [5]. Air pollution is the fourth leading risk factor for disease burden in China, leading to about 1.2 million premature deaths in 2010 [6].In light of the situation and potential far-reaching health consequences and disease burden attributable to climate change and air pollution, further substantial actions are necessary to reduce greenhouse gas (GHG) and air pollutants emissions [7,8]. On 12 November, 2014, a joint China-US announcement vowed that China will stabilize its GHG emissions by 2030. The goal is “relatively ambitious”, since it means that at least 20% of China’s power will come from sources other than fossil fuels by 2030, up from around 10% in 2014, which could potentially reduce China’s gross domestic product (GDP) by 1% to 3.7% [9]. The reductions goal would make China’s agenda to mitigate air pollutants, GHG emissions and climate change more urgent and challenging. How to balance environmental and public health threats against near term economic prosperity may be the crucial challenge faced by developing countries like China.One way to help China address this challenge and achieve the climate change mitigation goal is to fully account for the health co-benefits of GHG emission reductions. According to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change (IPCC), appropriate GHG mitigation measures aimed at curbing climate change will themselves have additional effects on public health, irrespective of the net effect on overall health gains and most of them beneficial [7,10]. One of the mechanisms for these so-called “co-benefits” is that GHGs and air pollutants to a large extent stem from the same sources and are linked in terms of their atmospheric formation and evolution as well as the effects on ecosystem and human beings [7,11,12]. Besides, some air pollutants such as black carbon and ozone are also climate-warming agents (namely greenhouse pollutants) with higher radiative forcing per unit than CO2 [12,13]. Thus, in addition to mitigating climate change, GHGs reductions may also deliver other improvements in public health simultaneously (Figure 1). For instance, GHG mitigation actions like reduced fossil-fuel combustion and improved energy efficiency can provide additional health benefits from reduced air pollution and related ill-health, especially in relation with the decrease in short-lived climate pollutants (e.g., black carbon and ozone), and the air quality benefits and health gains are often realized on a local scale and in the near-term [13,14,15]. These features of the ancillary health benefits may make GHG abatement measures much more attractive to local and national stakeholders (e.g., urban residents, key emitters and policy-makers), and can help motivate attempts and policies to put them into practice preferentially.Although these health “co-benefits” of GHG mitigation strategies have been modeled or quantified in an increasing number of studies published in the scientific literature around the world in recent years [7,16,17,18], to date, no investigation has been conducted specifically to assess the perceptions of stakeholders about the ancillary health benefits of GHG emission reductions. In addition, despite the numerous studies focusing on the role of governments and economic sectors in curbing climate change, relatively limited attention has been paid to the contributions of individual factors (e.g., awareness, attitude and behavior) to GHG controls [8,19]. As the causes of climate change lie ultimately in human behavior, the collective effect of individual behavior change by many individuals may result in appreciable reductions in GHG emissions [7,17,20]. Engaging the public can also help generate support for effective climate change mitigation actions by businesses, and local and national governments [8,21]. A significant volume of studies have demonstrated the associations between perceptions and individuals’ willingness to change behavior and support for climate change mitigation, and mitigation policies may risk being ineffective or rejected when public lacking an understanding of the issue [21,22,23]. Besides, public health co-benefits of GHG abatement make impacts of climate change mitigation more geographically, temporally, and personally relevant and context-specific, which would be useful in encouraging public behavior change and support for climate policy. It is thus reasonable to assume that understanding the perceptions of stakeholders about the health co-benefits related to carbon emission reductions may contribute to the improvement in public engagement with mitigation measures, and addressing the problems of both climate change and air pollution as well as provide a driver for further mitigation efforts [8,24,25].Aiming to fill this knowledge gap, we assess the perceptions of urban residents on the health co-benefits of GHG emission reductions in three cities in China. We firstly investigate the awareness level of the ancillary health benefits related with GHG abatement among subgroups within the survey population, taking into account demographic profiles. Then we analyse the perceptions of respondents about the health co-benefits in different economic or social sectors. Lastly, the potential influencing factors of respondents’ perception status are explored.A cross-sectional study “perceptive assessment of health risks caused by climate change, air pollution and health co-benefits of low carbon transition in China” was designed to investigate stakeholders’ knowledge of, attitudes toward, and perceptions of environmental issues in three Chinese cities. In this project, experts in the fields of climate change and air pollution, policy-makers, environmental scientists, and policy researchers collaborated closely. In order to achieve our objectives, literature review, existing policies analysis, workshops, questionnaire survey, and focus group discussions were conducted. As a part of this interdisciplinary project, the present study reports on a subset of the cross-sectional survey relating to questions on the health co-benefits of GHG emission reductions.Three cities, Beijing, Ningbo and Guangzhou, were selected as the study settings, because they are representative of the different climatic zones and socio-economic areas in China. The three cities are typical of the Beijing-Tianjin-Hebei region (BTH), the Yangtze River Delta (YRD), and the Pearl River Delta (PRD), respectively, where there is a high degree of economic development, as well as high levels of air pollution and large carbon emissions [26]. Thus, information from these settings may mirror the current perceptions status of urban residents about climate change, air pollution and GHG emissions in China.With the hope of ensuring good geographical coverage and obtain information that best represents the diverse socio-economic and demographic characteristics of the local urban population in each city, four districts were targeted to perform the field questionnaire survey (Figure 2). The selection was based on multiple factors, including old and new districts, downtown and its surroundings, population size and density, as well as function and coverage of specific district (according to the newest statistical yearbook of each city). For example, in Beijing, there are four administrative divisions (Core Functional Area, Urban Function Extension Area, New Area of Urban Development, and Ecological Conservation Area) with different socio-demographic patterns, and we selected one district from each of the division (Figure 2). For each city, the survey was carried out in four Community Committees selected randomly from each district. The committees are usually located at the geographic centers of the specific communities and are well known among local residents.The sample size was estimated using Kish-Leslie formula for descriptive studies with single proportions [27]. There was no previous information on the level of people’s perceptions about the health co-benefits of GHG reductions. Hence, in order to obtain a relatively larger sample size, we hypothesized a prevalence of adult residents’ awareness of the ancillary benefits in each of the target city to be 30%. The sample size required, n, was generated as n = pqz2/d2, where, p was the assumed knowledge prevalence of the health co-benefits related to GHG abatement (p = 0.3); q = 1 − p; z was 1.96 (for 5% alpha error); and d was the level of precision (5%). This gave a required sample size of 323 participants, which was increased by 10% to compensate for potential non-respondents, resulting in the final study population of 360 for each city. The total required sample size for the three cities was 1080.For each specific city, the pre-estimated sample size of 360 was stratified by gender and age groups (15–24, 25–44, 45–59 and over 60) representative of the distributions determined according to the latest statistical yearbook of each city, and then was proportionally distributed in each selected district. Participants were recruited purposely from the chosen committees, and the sample was based on quotas for gender and age in proportion to the population shares. Eligible respondents were permanent residents, aged over 14 years, who had lived in that community for at least six months. Only one eligible member was selected from each household to complete the survey. The final respondent sample from each city was compared to the characteristics of the corresponding target total population, and the population samples reflected similar gender balance and age distributions (Table 1).A semi-structured questionnaire was drafted following an extensive review of the scientific literature on health co-benefits in relation to GHG reductions. Discussions and deliberations among the researchers, and consultation with experts were also performed during the development of the preliminary questionnaire. An international workshop (attended by experts in the fields of climate change, air pollution, low carbon transition and policy research, public health officials and policy-makers from the three selected cities, interviewers, data collectors and administrators for the field survey) was organized to pre-test and review the draft questionnaire. The feedback and comments from attendees were then incorporated into the revised questionnaire. In order to assess the validity and reliability issues, before its use in the main investigation, the questionnaire was field-tested with a convenience sample of 21 urban residents from Changping district of Beijing. Based on the pilot study, participants’ comments on the content appropriateness, question clarity, and administration format were considered, and minor modifications were integrated to produce the final questionnaire and related interview guide.The full questionnaire comprised the following five sections: Section “A”, included seven questions regarding awareness of health risks associated with climate change; Section “B”, included 10 questions about perceptions of health risks related to air pollution; Section “C”, included 18 questions regarding knowledge of, attitudes toward and perceptions of carbon emission reductions and the corresponding health co-benefits; Section “D”, included five questions about participants’ policy related concerns and recommendations. Socio-demographic information of the respondents was documented at the end of the interview with 12 questions (Section “E”). In total, the questionnaire contained 52 close-ended or open-ended questions. 3–5 point Likert-type items, i.e., 1 (Strongly Disagree), 2 (Disagree), 3 (Uncertain), 4 (Agree) to 5 (Strongly Agree), or other categorical items such as “Yes”, “No”, “Don’t know” or “Others” were employed to assess respondents’ knowledge of, attitude toward and perception of the mentioned environmental issues. This paper only reports on a subset of questions (Supplementary Materials), those related to health co-benefits of GHG abatement.After the sampling protocol was developed, in order to ensure community support and engagement, the objectives and activities of the study were firstly delivered from our partners, the Centre for Disease Control and Prevention (CDC) of Beijing, Ningbo and Guangdong Province, to the selected District CDCs, then to the corresponding community health service centers (stations), and lastly to the randomly chosen Community Committees, and their participation were invited. All the selected committees showed their interest in and support of this study, and committed staff resources to the field investigation. 3–5 days before the surveys started, staff from each committee informed all permanent residents living in the community about the aims, contents, date and location of the survey, and encouraged them to participate. As an incentive to participate, all respondents were offered an approximately 30 Chinese Yuan (about 4 US dollars) gift (a laundry detergent sample) upon completion of their interview. Those who showed their willingness to take part in the interview then came to the committee to fill out the questionnaire, and participants were selected purposely according to the quotas for gender and age in proportion to the population shares.In each city, before the field investigation, the questionnaires were administered firstly by interviewers involved in data collection to facilitate their understanding about the contents and questions. All interviewers received a half day intensive and systematic training. The training consisted of general information about the project, the purposes of the study, a review of the study’s interviewer manual, question-by-question review of the questionnaire, interview techniques and skills for approaching the participants, as well as social and cultural sensitivity issues during data collection, and an additional session for practicing the survey scripts with a partner. The first author and corresponding author reviewed each investigator’s initial interview, and pointed out inappropriate issues concerning the survey such as manners and choice of words.From October to November of 2015, participants were interviewed by the trained investigators using the semi-structured questionnaire. In each site, one to four senior researchers were always present in the field to monitor the process, coordinate the questionnaire collection and, examine the quality of the data collected. The staff involved were familiar with the survey procedure documents, and communicated with the first author or corresponding author for any questions and clarification when needed. All completed questionnaires were reviewed and checked for completeness and validity by the field supervisors (senior researchers). Incomplete questionnaires were returned to the respondents immediately to figure out the potential reasons (e.g., difficulties in understanding the questions, omission, refuse to cooperate, or unexpected exigency), and the completion was asked for if possible (in the case of permission and cooperation). The reasons found out (with the help of the staff from the local community health service centers or Community Committees) will be delivered to the investigators as feedback in order to prepare for or avoid similar situations. The survey in each site was carried out continuously until the required sample size was met.The collected data were double entered into a database by trained personnel using EpiData 3.1 software (EpiData Association, Odense, Denmark). The data were then transferred to statistical software and analyzed according to different variables. Firstly, descriptive statistics were used to illustrate demographic characteristics of participants and percentages of categorical variables. Secondly, a series of Chi-square tests were employed when appropriate to examine the associations between socio-demographic variables and perceptions of participants on the health co-benefits of GHG mitigation (1 = never heard, 2 = only heard, and 3 = familiar); otherwise, Fisher’s exact test (expected cell frequencies less than or equal to five) was used. Thirdly, the distributions and percentages of perception variables in different economic or social sectors were summarized. The primary purpose for this series of questions was to further investigate the awareness of respondents on how GHG reductions can bring about health co-benefits, as well as re-check the perceptions of participants about the contents of the health co-benefits.In addition, unconditional logistic regression analysis was performed to assess the associations of demographic variables, attitude (from 1 = Strongly Disagree, to 5 = Strongly Agree) and practice factors involving environmental issues (independent variables) with perceptions of respondents on the health co-benefits related to GHG abatement (dependent variable, coded as 0 = not familiar, and 1 = familiar). To determine which factors were associated with the perception variable, logistic regression analysis was done at both bivariable and multivariable level. Factors that had a p-value < 0.05 and those identified in the literature review to have potential effects on the perceptions of co-benefits were included in the final multivariable model, and the odds ratios (OR) with 95% confidence intervals (CI) were calculated. All statistical analyses were performed using IBM SPSS 19.0 (SPSS Inc., Chicago, IL, USA) and Stata 12.1 (Stata Corporation, College Station, TX, USA). All statistical tests were two-sided and a p value less than 0.05 was considered to be statistically significant.Ethical approval was granted by the Ethics Committee of the Chinese Center for Disease Control and Prevention (No. ICDC-2015005). The survey was anonymous, and verbal informed consent was obtained from all participants prior to each interview, after explaining the objectives of the study. Respondents were assured that the privacy and confidentiality of the data was to be maintained. Sufficient time was given to participants to complete the questionnaire, and it was emphasized that the participation was voluntary and they had the right to refuse participation or withdraw from the survey at any time.A total of 1166 participants took part in the survey, but seven respondents did not complete the survey after having started it, because of difficulties in understanding the questions, distractions, or unexpected exigency, as reported by the interviewers, so finally 1159 completed questionnaires were included in this analysis, representing an overall response rate and survey completion rate of 99.4%.Table 2 presents the demographic characteristics of the total sample. Of the 1159 participants, individuals aged from 25 to 59 years were the majority (66.7%), and women accounted for 51.3%. Ethnically, 97.4% of the participants were Han Chinese, similar to the actual ethnicity distributions in each city. The sample covered various education levels and occupations, and more than half of the respondents (55.1%) with family monthly average income between 2000 and 5000 Chinese Yuan (about 290–725 US dollars, and the national monthly per capita disposable income of urban households in 2015 was 2649.17 Yuan). Participants from urban and rural (according to registered residence, namely Hukou) accounted for 75.5% and 24.5%, respectively.Participants were asked about their perceptions of the health co-benefits in relation to GHG emission reductions. We found widespread agreement (91.9% participants) that GHG abatement do not only mitigate climate warming, but also improve public health in various ways. Table 3 shows the responses of participants regarding the health co-benefits of GHG reductions. Despite the high recognition of the ancillary health benefits relating to GHG mitigation, relatively few (15.9%) respondents reported that they were familiar with the specific types of health co-benefits, while the majority (67.8%) admitted that they had only heard about the concept but could not figure out the details. Although all three cities presented low awareness level about the ancillary health benefits, participants from Ningbo showed a statistically significant (p = 0.002) higher perceptions level (21.4%) than in the other two cities.The levels of awareness of the health co-benefits varied significantly across different age groups of respondents (χ2 = 59.94, p < 0.001). In general, younger participants tended to have higher awareness level. The gender of respondents did not seem to play a role in the level of awareness of health co-benefits (p = 0.25). There was an association between individual education and the belief that GHG reductions would bring improvement in public health (χ2 = 91.67, p < 0.001), as respondents with higher education level tended to be more aware of the ancillary health benefits of GHG mitigation. Family income and registered residence played a statistically significant role as well. The perceptions of participants increased with family income, except for the group of <1000 Chinese Yuan (χ2 = 22.79, p = 0.012). While 17.0% of respondents from urban areas reported they were aware of the health co-benefits, fewer participants (12.3%) from rural areas claimed to be familiar with the concept (χ2 = 6.26, p = 0.044). Respondents’ marital status seemed to be associated with the level of awareness of health co-benefits (χ2 = 14.56, p = 0.017), but the factor of participants’ health status did not appear to have an effect. Figure 3 shows the distribution of the respondents who were familiar with the health co-benefits by occupation. Among all participants, students, medical personnel, staff of commerce or service trade and company employees presented relatively higher proportion in the awareness of the health co-benefits.Participants were asked about the potential pathways through which carbon emission reductions can bring about additional health gains. As shown in Figure 4, a large majority of respondents (87.8%) indicated that GHG mitigation measures can reduce indoor air pollution and improve outdoor air quality, and consequently protect public health from the hazards of air pollution. Roughly two thirds (67.9%) reported that the low carbon transition could improve living, producing (built) and ecological environment, which improves human health and wellbeing. This could be achieved by “increasing the amount of physical activities” and “reducing the intake of unhealthy or junk food (e.g., food with high fat content)” according to 32.1% and 55.0% of the respondents, respectively. Of special note is that 33.4% of the participants indicated that a low carbon lifestyle can improve their mental outlook.Table 4 summarizes the perceptions of respondents on the health co-benefits of GHG mitigation in different economic or social sectors. Participants generally showed high levels of awareness, but with some confusion to different degrees. For instance, to the question “In energy production and use, through what ways could carbon emission reductions bring about health co-benefits?” Almost all respondents (95.9%) selected “decrease air pollutants, improve air quality, and reduce diseases caused by air pollution”; and “mitigate climate change, decrease the burden of climate-sensitive diseases” was selected by 88.6% of the respondents. However, “increase physical activity, reduce obesity and cardiovascular diseases” and “encourage scientific innovation and facilitate social development” which are exactly not the pathways that GHG mitigation policies bring about health co-benefits in energy generation, were also indicated by 75.8% and 73.8% respondents, respectively. Similar relatively high perception levels, along with some misunderstanding about the pathways of GHG abatement measures that can generate health co-benefits, were also observed in the transport, agriculture and household sector (Table 4).A series of bivariate logistic regression analyses were conducted firstly to examine the associations between the dependent variable (perceptions of the health co-benefits) and each independent variable. After the variables filter, factors with statistically significant impacts on the dependent variable were included in the final multivariable logistic model, including “Age”, “Education level”, “Attitudes toward the current urban air pollution”, “Attitudes toward governmental policy attempts and progress to deal with the problems of air pollution and climate change”, “Have respiratory diseases”, and “Choose low carbon lifestyle in daily life or work”. The only exceptions were two variables (“Gender” and “Family monthly average income”), which did not have significant bivariate associations with the perception variable but were still included in the final model, since they were plausible and were considered a priori to be important variables.Table 5 presented the final logistic regression analysis for the associations between independent variables and respondents’ perceptions on the health co-benefits of GHG reductions. The age of participants played a negative role in awareness of individuals about the health co-benefits (OR = 0.98, 95% CI: 0.97–0.99). By contrast, these four variables—attitudes of participants toward air pollution and governmental policy efforts, suffering from respiratory diseases, or following low carbon lifestyle in usual—exerted significant positive influence on respondents’ perceptions about the ancillary health benefits. The effect of gender, education level and family income on the perceptions of respondents appeared statistically non-significant (p > 0.05).Given the role of the so-called “health co-benefits” in motivating emitters to put GHG mitigation measures into practice at individual, social, national and even global level, the potential associations between perception of the ancillary health benefits and individuals’ willingness to change behavior and support for low carbon transition, and the challenges of air pollution and climate change faced China, it is important to assess the perceptions of the health co-benefits of GHG emission reductions among stakeholders in different regions in China. In this study, the perceptions of urban residents on the health co-benefits of GHG mitigation, and the relevant influencing factors were investigated. To our knowledge, this is the first study to specially assess the awareness of the health co-benefits in relation to GHG emission reductions around the world. Therefore, findings from the study have the potential to fill an important knowledge gap.Evidence from health belief models, risk perception attitude frameworks and social cognitive theory have proposed significant, although sometimes small, associations between knowledge of the issue, risk perceptions and the likelihood of actions [21,28,29]. If properly executed, under specific conditions, and with certain target audiences, information may lead to increased awareness and this may lead to behavior changes [30]. For example, studies have shown that response of population to natural hazard warnings is partly determined by perceived urgency of the threat [31], and the risk perceptions need to be accompanied by efficacy beliefs and proximity to the hazard to promote action [21,28]. However, for climate change, individual perception of risk is a necessary but far from sufficient factor on its own, to contribute to motivating individuals to change their behavior [22]. Instead, it is emotions—the feelings along with thinking—that are central [32]. It should be noted that negative emotions such as fear, guilt and pessimism, are likely to produce passive and defensive responses, and hardly do much to encourage people to change their behaviors or to press for wider social action [8]. Several studies have focused on risk perceptions of climate change, but less attention has been paid to the awareness of the health benefits associated with mitigation actions [33,34]. In addition to providing engaging messages about how to address the problem, ancillary health gains of GHG reductions may also represent climate change mitigation in ways that connect with people’s core ideologies and identities and then anchor it in positive emotions, which is one of the crucial determinants of behavior and behavioral change [8,35,36]. Besides, the ancillary health benefits make GHG abatement measures more “down to earth” and personally-relevant issues, which is an evolutionary tendency for people to pay attention to and appears to be particularly compelling [24,37,38]. Reframing climate change from an environmental to a public health issue and linking GHG mitigation policies to beneficial health gains (positive vision) may bring climate change mitigation closer to home thereby increasing its relevance to the public, and potentially encouraging public engagement in mitigative behavior change [25,37,39]. Thus, this perceptive assessment of the health co-benefits in relation to GHG abatement may fill some of the evidence gaps and provide additional motivation to help change individuals’ climate-affecting behaviors and habits.Similar with previous research on risk perceptions of climate change [25,33,34], we found that awareness of the multiple benefits of GHG emission controls on climate and public health was very high (91.9%) in the three cities of this study, likely due to the extensive media coverage, internet penetration and government advocacy on this topic in China. In recent years, awareness raising of low carbon development and green growth has been carried out continuously across the whole country, and it seems that the idea of low carbon transition has been absorbed by the general public [26]. This information increase is reflected in growing numbers of scientific reports, newspaper articles, and increased awareness of and interest in international negotiations devoted to the issue. Though the knowledge was widespread, (more concrete) awareness of the health co-benefits of GHG reductions was less common, only 15.9% respondents claimed they were familiar with the specific concept. When the three cities were compared, despite some statistically significant differences between cities, we found large proportions of the study populations with low awareness level of the health co-benefits, which indicates that further public health outreach campaigns on the topic of the health co-benefits of GHG mitigation are needed.Results showed that young participants tended to show higher recognition of the health co-benefits, and this relationship was also observed in the final multivariable logistic model. The reasons for this may be that the health co-benefits of GHG mitigation is a relatively new concept raised in recent years [7,17], and young individuals usually have the advantage of learning and absorbing emerging information through more channels and faster than older individuals [40]. For example, young individuals may access information through channels such as school education, the internet, smart phone, television, radio, and newspaper, while older individuals mostly obtain information through only traditional media like television and newspaper [25,40]. According to China Internet Network Information Center, new media based on internet and smart phones are becoming a key pathway to spread information in China, with 74.7% of the internet users aged 10–39 years [41]. Besides, the younger age groups are reportedly more likely to be concerned about environmental issues such as air pollution and climate change [25,42]. Currently, no studies have assessed gender differences with respect to the perceptions of health co-benefits relating to GHG reductions. While evidence from risk perceptions assessment of climate change or heatwave is still not uniform, women seem to be generally more concerned about environmental issues and are aware of environmental risks than men [25,40,43,44]. In our study, although women tended to demonstrate higher awareness level of the health co-benefits, the difference between genders was statistically non-significant (p = 0.192).We found a relationship between education and belief that carbon emission reductions could generate health co-benefits; higher education levels were associated with better awareness. Education is associated also with a greater probability of expressing concerns about environmental threats such as climate change, heatwaves and air pollution [25,45,46]. It has been suggested that modern lifestyles largely disconnect individuals from directly experiencing changes and instead make them more dependent on mediated information about environmental issues (e.g., internet, smart phones or television) [22,47]. It was reported that compared with those with lower education levels, individuals with higher education may be more willing to search health-related information, and more likely to access climate change information through the internet [34,43]. However, the education variable did not have a significant effect in the final logistic model (p = 0.757), partly due to the adjustment for age, gender and income, which were often interlinked and overlapping with education [40]. Respondents with high household income were more likely to report awareness of the health co-benefits. Similarly, individuals with urban registered residence were more aware of the benefits of GHG reductions. Compared with rural participants, urban respondents were relatively higher-income earners who could access information through more channels like internet, smart phones, television or newspapers [25,40,41]. Higher income is often associated with higher education levels, while is predicted to make people more aware of environmental issues [48,49]. Besides, it was suggested that there was a higher probability for urban residents to be concerned about climate change than their rural counterparts [45].Also as expected, respondents who are students, medical personnel, staff of commerce or service trade and company employees were more likely to be aware of the health co-benefits, which could be explained by the fact that they were mostly younger individuals who are the main internet users in China [41]. This may also explain the different awareness levels among participants with different marital status. It is not surprising perhaps that, medical personnel who are trained in public health were more likely to be interested in environment-related information [34,43]. Surprisingly, teaching staff and technicians showed relatively lower awareness levels, implying further research is needed to establish whether this is related to lack of information or personal motivation that may have led to the lower awareness levels. For the perceptions on the ways GHG mitigation can improve public health, although the majority of respondents recognized reducing air pollution and improving the environment as ways of carbon emission reductions that can bring additional health gains, relatively few individuals indicated “increase physical activity” or “reduce the intake of unhealthy or junk food” as the potential approaches. 33.4% participants insisted that GHG abatement measures can improve people’s mental outlook, suggesting the additional value of relevant low carbon measures (e.g., sustainable urban design and urban green space) in the improvement of public emotional and mental health, such as enhancing cultural and aesthetic values, increasing social contact and neighborhood cohesion, and offering restorative experiences [8,15,18,38,39]. However, in terms of the relatively low awareness level, these aspects require more investigation and health education campaigns.Overall there was close to consensus about the statements on the health co-benefits of GHG mitigation in different economic or social sectors (Table 4). Plausible explanations for this phenomenon are summarized as following. First, widespread awareness of the risks of climate change or temperature warming have been observed by numerous studies around the world [28,33,34,40,43,49,50]. As a similar environmental issue, and in terms of the increasing governmental and social media awareness raising activities in recent years in China, widespread approval of the statements of the health co-benefits related to GHG reductions may be expected; Second, though there is a substantial awareness, it is likely that there is some extent of confusion or misunderstanding among the respondents in terms of differentiating the concept of “health co-benefits” from “social benefits or welfare benefits”; Third, social desirability may be a factor influencing responses to the survey. As an emerging concept, when participants know little or understanding not well about the health co-benefits of GHG abatement, then a convenient answer may be provided. Or individuals may have been worried about giving “wrong” answers, and the socially desirable ones were chosen [51]. For instance, in the context of the increasing governmental and social policy advocacy and publicity of low carbon transition and green growth in recent years, various positive effects of GHG reductions have been gradually becoming a socially-acceptable consensus in China, especially in regions that the most economically developed while undergoing the most serious air pollution and largest carbon emissions, such as Beijing, Ningbo and Guangzhou. Thus, when some participants from the three studied cities do not understanding the specific content of the health co-benefits well, they may be to relate some positive outcomes (especially outcomes with the words like “improvement”, “development”, “innovation”, or “decrease”) to the impacts of GHG abatement measures according to social expectations (courtesy bias). In light of the aborative development of the study instrument (questionnaire) and quality control of survey process, another less possible reason may be that too many messages were delivered or the questionnaire questions were too complicated. To save time, some respondents might have only absorbed the simplest of these pieces of the questions, and checked similar answers for each question rather than spending the necessary time to think carefully about each of them [42,49].Participants who agreed the air pollution in their city was a serious problem and has caused health impacts, or approved that government has taken a package of measures to address climate change, air pollution and carbon emissions and the situations is improving, were more likely to be aware of the health co-benefits of GHG mitigation. Consistent with the discussion above, a higher level of concern about environmental issues such as climate change, air pollution or carbon emissions is more likely to be associated with the awareness of the health co-benefits. Similarly, respondents with respiratory diseases may be more concerned about air pollution and relevant mitigation measures, which increases their chances to access the information involving the health co-benefits of GHG reductions, making this variable a significant predictor of the perceptions status (OR = 1.50, 95% CI: 1.01–2.25). The final logistic regression model also revealed that the individuals who followed low carbon lifestyles tended to be more aware of the health co-benefits. This association may be explained by a low carbon lifestyle increasing their chances to learn about the health co-benefits, or by the recognition of the health co-benefits in relation to GHG mitigation actions motivating them to follow a low carbon lifestyle.The present study has several strengths. Firstly, to the best of our knowledge, it is the first study to assess specially the perceptions of urban residents on the health co-benefits in relation to GHG reductions around the world. Findings from this study, together with other research initiatives currently under way in the UK-China research project “China Prosperity Strategic Programme (SPF 2015-16)” (perceptive assessment of health risks caused by climate change, air pollution and health co-benefits of low carbon transition in China), will, it is hoped, fill some of the knowledge gaps on the topic of GHG mitigation. Secondly, the three studied cities were selected from the BTH, YRD and PRD regions, which are typical of three different climatic zones in China, where the socio-economic fields achieve the greatest success while undergoing the most serious air pollution and facing greatest challenge of carbon emissions in China. Thus, findings based on these cities may are representative of the perception status of urban residents about the health co-benefits of GHG abatement in most developed parts of China. Thirdly, our field survey involving three cities and twelve districts, activities such as objective publicity, local investigators training and focus groups during the surveys, may play a role in improving the understanding of the health co-benefits of low carbon transition among local experts, CDC officials, staff of each committee, and interviewers. Improved perceptions making it is possible for the stakeholders to take the health co-benefits into consideration in their daily life, routine activities or institutional operations, which ultimately help to facilitate the development and implementation of low carbon policies in China. For example, during the focus group discussions, after discussing and understanding the health co-benefits of mitigation measures in different economic sectors, most participants claimed that given the individually and environmentally beneficial gains (personally-relevant health benefits and air quality improvement) of GHG reductions, they would like to consider alternative options (low carbon and green) in their routines and habits, such as choosing active travel (walking, cycling, and public transport) for short city trips, limiting or changing westernized dietary patterns that mainly from animal sources and rich in saturated fat and sugar, and shifting carbon-dependent consumption customs to environmentally-friendly lifestyles. In general, respondents regarded these low-carbon alternative options as mitigation actions that could be realistically achieved at the local and personal level (details will be discussed in a forthcoming companion paper). Fourthly, our survey revealed that the awareness levels of respondents about the health co-benefits of carbon emission reductions are very low (15.9%), while certain segments of the population were more likely to be aware of the health co-benefits than others, such as those who were younger, more educated, with higher family mean income, and more concerned about air pollution, climate change and government efforts. These findings could provide helpful information to policy-makers to develop and implement win-win policies for air pollution and climate change mitigation. They may also improve public acceptability of GHG mitigation measures, and the necessity to conduct extensive GHG reductions-related health education campaigns, at both national and local level. Besides, in light of the appreciable health co-benefits of GHG abatement, our findings also hold the potential to help bridge the gap between present and future GHG abatement measures and public support and engagement, as well as the willingness to change their behaviors. For instance, although public health co-benefits of GHG mitigation are plausible and attractive to those people interviewed, participants indicated that their behavior change was often constrained by the lack of enabling infrastructures and mechanisms. To be specific, some respondents pointed to a lack of acceptable and reliable built environment (e.g., urban green space) and public transport in their locality for active travel, unaffordable low carbon goods and household appliances, and intractable social norms and expectations that requiring carbon-dependent consumption customs and lifestyles (e.g., animal sources foods, private motor vehicles, and high-emission electronic goods). This information may inspire governments to undertake further practice change and actions to reduce or remove the barriers gradually, through health education campaigns creating environmental citizenship, in combination with a framework of incentives and regulations (Carrots and Sticks) [38,39] (details will be presented in a forthcoming companion paper).Several limitations of this study should also be acknowledged. First, the cross-sectional nature of this study does not allow for the assessment of trajectories [44,51]. By relying on survey-based evidence from a single point in time, we are unable to uncover the causal relationships between independent variables and the perceptions of respondents on the health co-benefits of GHG mitigation. For instance, in the final multivariate regression model, we cannot establish whether it is the low carbon lifestyle influencing the awareness of respondents, or individuals’ perceptions of the health co-benefits lead to low carbon lifestyles. Second, the study was conducted on a convenience basis, did not involve a random sample (selected purposely). Hence this study provides only a snapshot of urban residents’ perceptions on the health co-benefits, the results may not representative of the entire population, and are not necessarily generalizable to other regions. However, there were four districts with different socio-economic features in each city were investigated, and our findings are based upon a sample of sufficient size and with similar socio-demographic characteristics of the target population of each city. Thus, a reasonable reflection of the awareness of urban residents on the health co-benefits of GHG reductions in the three studied cities could be provided by this survey. Third, all information about the perceptions of the health co-benefits are based on respondents’ self-reports, which is, by definition, subjective. As with all studies of this kind, the answers may be affected by the biases of social desirability and courtesy, meaning that participants are prone to report the answers that they perceive to be socially acceptable, or that they thought the interviewer wanted to hear and in the favor of the survey, which may be different from their real perceptions [21,33,40,41,52]. However, a series of measures including but not restricted to investigator training, explanation of survey objectives, anonymity of the survey, and assurance of privacy were taken to deal with these potential biases. Fourth, the present survey focuses on the permanent residents in the three cities, newcomers from other regions (less than 6 months) were not included, so our results cannot be extrapolated to these population sub-groups. Another limitation of this study pertains to the fact that the shortage of similar studies carried out in China or other regions of the world makes the comparison and interpretation difficult. Therefore, information from surveys about risk perceptions of climate change or heatwaves conducted around the world were borrowed as references [21,22,25,33,34,40,42,43,44].The perceptive assessment of the health co-benefits of GHG emission reductions carried out in this study sheds some lights on the existing knowledge gaps. Overall, individuals’ awareness of the health co-benefits of GHG reductions is still limited, as only 15.9% of the participants stated that they were familiar with the specific contents of the health co-benefits. Considering the potential confusion or misunderstanding among the respondents about the “health co-benefits” versus the “social benefits or welfare benefits” presented in Table 4, the actual awareness level of participants may be even lower. Perceptions of the health co-benefits related to GHG mitigation are influenced by socio-demographic characteristics of participants; those who are younger, more educated, with higher family average income, and with urban registered residence, were more likely to be aware of the ancillary health benefits than others. The final logistic regression model revealed that age, attitudes toward the air pollution and governmental efforts to address the problem, suffering from respiratory diseases, and following low carbon lifestyle in daily life or work, were significant predictors of respondents’ awareness about the health co-benefits of GHG abatement.The influencing factors of respondents’ perceptions on the health co-benefits identified in this study, may help bridge the gap between GHG mitigation measures and individuals’ knowledge of, attitudes toward and perceptions of carbon emission reductions. With the advance of the low carbon transition in China, which is likely to encounter resistance from parts of the society, the health co-benefits of GHG emission controls will continue to be an important driver for strategies aiming to remove barriers to mitigation in the context of both climate change and air pollution. Acknowledging insufficiency in both the scientific information production and public interpretations of the health co-benefits of GHG reductions is a necessary step in improving public awareness, follow and support of GHG abatement measures in China. This compliments moves towards a more comprehensive framework for public education campaigns explaining the health co-benefits of carbon emission reductions, accompanied by well-resourced and sustained policies enforcement, which could stimulate more productive engagement between scientific knowledge, individuals’ perceptions and changes in behaviors [8,24]. Individuals are the actors who ultimately initiate, inspire, guide and enact the reductions in GHG emissions to curb climate change [22], and public perceptions may be of considerable interest to policy-makers, as they can drive policy as much as scientific risk assessments [44,53]. Therefore, the public may be the strongest ally in the battle against carbon emissions. In order to let the health co-benefits of GHG reductions become a mobilising, engaging and effective instrument in regional or even global health thinking, the participation, commitment and ownership of individuals (“grassroots”) is needed [7,8,54].Our findings demonstrate that people’s perceptions and the influencing factors are complex and are often difficult to predict, pointing to the value of further work on public understanding of the health co-benefits of GHG emission reductions in China. Public health awareness campaigns and information dissemination are important to deliver the health co-benefits, but need to be accompanied by careful inquiry into attitudes on the merits of emission controls and behavioral change towards low carbon lifestyles [8,26]. More research is needed to understand which communications methods are the most effective to make this happen. In this context, the call for “more information” means that attractive and public acceptable scientific knowledge with the potential to motivate people to engage in behaviors that reduce GHG emissions is needed, rather than just evidence filling an “information-deficit” related to the health co-benefits of climate change mitigation. Of special note that it is often the “no concern” or “value-action gap” that impede public participation and action, which may eventually undermine policy implementation of low carbon transition [7,8,26,55], requiring further awareness raising and behavioral research.The following are available online at www.mdpi.com/1660-4601/14/3/298/s1.This work was supported by the China Prosperity Strategic Programme Fund (SPF) 2015-16 (Project Code: 15LCI1), the National Basic Research Program of China (973 Program) (Grant No. 2012CB955504), and the Medical Technology Program Foundation of Zhejiang (2014KYA202), Science and Technology Program of Ningbo, China (2014C50027). The funders played no role in the design, development, or interpretation of the present work. The views expressed in the article are those of the authors and do not necessarily reflect the position of the funding bodies.J.H.G. and Q.Y.L. conceptualized, designed and initiated the study. J.H.G. and Q.Y.L. drafted the initial manuscript. G.Z.X., T.F.H., H.L.L., S.H.G., J.W., J.L., J.Y. and J.L. involved in the development of methodology and discussion of article structure, A.W., S.V., S.K., P.W., Y.Z., W.J.M., T.L., X.B.L. and H.X.W. reviewed and revised the manuscript. All authors read and approved the final manuscript.The authors declare no conflict of interest.Pathway of the health co-benefits of greenhouse gas mitigation measures.Study settings selected to conduct the field questionnaire survey.Percentage of respondents who self-reported that they were aware of the health co-benefits in relation to GHG reductions among different occupations.Perceptions of respondents on the pathways of GHG emission reductions to provide ancillary public health benefits.Demographic information of the sample versus the target total population of each studied city.Demographic characteristics of the study participants (N = 1159).Perceptions of respondents (by total and subgroups) on the health co-benefits of GHG emission reductions.Perceptions of respondents on the health co-benefits in relation to GHG mitigation measures in different economic or social sectors.Multivariable logistic regression analysis for the significant predictors of respondents’ perceptions on the health co-benefits in relation to GHG mitigation.OR: odds ratio; SE: standard error; CI: confidence interval; p: p-value; NA: not applicable (for reference).

Med-MDPI/ijerph_2/ijerph-14-03-00299.txt

@@ -0,0 +1 @@
+Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).Numerous studies have unmasked the deleterious effects of particulate matter less than 2.5 μm (PM2.5) on health. However, epidemiologic evidence focusing on the effects of PM2.5 on skin health remains limited. An important aspect of Asian dust (AD) in relationship to health is the amount of PM2.5 contained therein. Several studies have demonstrated that AD can aggravate skin symptoms. The current study aimed to investigate the effects of short-term exposure to PM2.5 and AD particles on skin symptoms in schoolchildren. A total of 339 children recorded daily skin symptom scores during February 2015. Light detection and ranging were used to calculate AD particle size. Generalized estimating equation logistic regression analyses were used to estimate the associations among skin symptoms and the daily levels of PM2.5 and AD particles. Increases in the levels of PM2.5 and AD particles were not related to an increased risk of skin symptom events, with increases of 10.1 μg/m3 in PM2.5 and 0.01 km−1 in AD particles changing odds ratios by 1.03 and 0.99, respectively. These results suggest that short-term exposure to PM2.5 and AD does not impact skin symptoms in schoolchildren.Ambient particulate matter (PM) is a mixture of solid particles and liquid droplets originating from various natural and anthropogenic sources and comprises an important source of air pollutants [1]. Clinical, mechanistic, and epidemiological evidence have demonstrated adverse health effects from short-term and long-term exposure to ambient PM; these adverse influences are of concern to governments of various countries as well as the World Health Organization [2,3,4]. However, most epidemiological studies regarding the negative health effects of ambient PM have focused on respiratory and cardiovascular injuries.The skin is the outermost barrier and is directly exposed to various environmental pollutants. Ultraviolet radiation from sunlight has been the most studied environmental hazard, and its consequences on skin are well established [5]. While a number of clinical and epidemiological studies highlight the associations with ambient PM and health effects, very little research is available to date concerning skin effects [5,6]. Furthermore, the exact mechanism of skin damage by ambient PM has yet to be elucidated. However, diesel-exhaust particles, which are an important source of PM less than 2.5 μm in diameter (PM2.5), have been shown to induce a strong inflammatory response in human skin cells [7,8]. Asian dust (AD), which originates in the deserts of East Asia, is the second strongest source of dust emissions worldwide (accounting for about 20% of the global total) and increases the concentrations of air pollutants including PM2.5 [9,10]. In addition, recent studies reported that AD was associated with an increase in mortality as well as emergency treatment for cardiovascular and respiratory diseases [11,12,13,14]. Moreover, Otani et al. found an adverse effect of AD on skin health in Japan [15]. These results suggest that short-term exposure to PM2.5 can aggravate skin symptoms.Few studies have investigated the association between ambient PM and skin symptoms in children. Therefore, this study aimed to investigate the association between skin symptoms in children and short-term exposure to PM2.5 in western Japan. We also studied the relationship between skin symptoms in children and AD because AD is one of the important sources of PM2.5 in Japan.In this panel study, skin symptoms of schoolchildren were monitored daily in the morning during February 2015. The study was performed in Matsue, the capital city of Shimane Prefecture, in southwest Japan. This city houses approximately 200,000 individuals and covers an area of 530.2 km2. Four elementary schools were selected from a total of 35 in Matsue City because these four schools are located near principal roads in the central part of Matsue City and are within 10 km of each other as well as from the observatory for monitoring air pollution in Matsue City. All elementary schools agreed to participate in the study. All subjects lived within a 1-km radius of the schools. A total of 345 students aged 10 to 12 years in 2015 were enrolled. The study was approved by the institutional ethics committee (Ethics Committee of the Faculty of Medicine, Tottori University, approval number 2473). The study was also approved by the Matsue City Board of Education. The children and their parents were informed by teachers and provided written consent.In January 2015, the schoolteachers provided information regarding the sex, height, and weight of each child based on physical measurements obtained by each school and recorded this information in a logbook for recording skin symptoms by teachers. The parents filled their children’s data concerning presence of asthma, allergic rhinitis, allergic conjunctivitis, atopic dermatitis, and food allergies in the logbook. The subjects were considered to have asthma if they met any of the following criteria in the past 12 months: diagnosis of asthma by a pediatrician, presence of wheezing, use of asthma medication, or a visit to a hospital for asthma. The subjects were considered to have allergic rhinitis, allergic conjunctivitis, atopic dermatitis, and/or food allergy if they met any of the following criteria in the past 12 months: diagnosis of any of these conditions by a pediatrician, use of medication for any of these conditions, or a visit to a hospital for any of these conditions. From 1 February 2015 to 28 February 2015, each child recorded daily scores for itchiness and rash as skin symptoms; scores were recorded as 0 (no symptoms), 1 (mild symptoms), or 2 (severe symptoms). Scores were recorded between 8:00 a.m. and 9:00 a.m. All children went to school on foot and were potentially exposed to any air pollutants.Daily average concentrations of PM2.5, sulfur dioxide (SO2), nitrogen dioxide (NO2), and ozone in Matsue City were calculated based on data from the Japanese Ministry of the Environment, which monitors and reports hourly concentrations of these air pollutants. Meteorological variables, such as daily average levels of temperature, humidity, wind speed, and atmospheric pressure, were obtained from the Japan Meteorological Agency. The data were used to examine the associations between changes in skin symptoms and air pollutant levels.Light Detection and Ranging (LIDAR) depolarization measurements performed simultaneously at two wavelengths can be used to identify non-spherical dust particles such as airborne sand dust particles and spherical particles such as air pollution aerosols in real time [16,17]. The LIDAR system can be used to measure the level of AD [16,17]; recently, a few studies from Japan used LIDAR data to estimate the effects of AD on health [18,19,20]. Daily particle levels were determined based on the median value of 96 measurements collected over a 24-h period from midnight of one day to midnight of the next day. The daily levels were only calculated when the number of available measurements exceeded 50% of the total number of measurements. This study used values measured from 120 m to 150 m above ground, which is the minimum altitude required by LIDAR systems to measure non-spherical and spherical particles. Data for AD concentrations from LIDAR were obtained from the Matsue observatory.To adequately address correlations among repeated measurements within a subject, generalized estimating equation (GEE) logistic regression analyses were used to estimate the associations among the daily skin symptoms of children and the daily average levels of PM2.5, SO2, NO2, and ozone and median levels of AD particles [21,22]. A skin symptom event was defined as a daily skin symptom score ≥2. The GEE logistic regression models included individual characteristics (sex, height, weight, asthma, allergic rhinitis, allergic conjunctivitis, atopic dermatitis, and food allergies) and meteorological variables (daily temperature, humidity, and atmospheric pressure) [23,24,25,26]. Estimates are given as the odds ratio in skin symptom events per interquartile range (IQR) change of PM2.5, SO2, NO2, ozone, and AD concentrations, with 95% confidence intervals (CIs). The working correlation matrices were set to exchangeable, and robust variance estimators were adopted for constructing the CIs for the odds ratios. Multiple imputations with chained equations were used for treating missing data; this adequately addresses the uncertainty of multiply generated prediction values for missing data [27]. The two-pollutant models were applied to different combinations of air pollutants (SO2, NO2, and ozone) to assess the stability of the effect of PM2.5 and AD on skin symptoms after adjustment for individual characteristics (age, sex, height, weight, and presence of asthma, allergic rhinitis, allergic conjunctivitis, atopic dermatitis, and food allergies), and meteorological variables (temperature, humidity, and atmospheric pressure). The GEE analyses were performed using R version 3.2.2 (R Foundation for Statistical Computing, Vienna, Austria).Of the 345 children who were recruited, six were excluded because they failed to maintain a daily record of skin symptoms. The characteristics of the remaining 339 children are shown in Table 1. Data were missing for sex (n = 2), age (n = 3), height (n = 6), and body weight (n = 8).Table 2 shows the daily levels of temperature, humidity, atmospheric pressure, wind speed, PM2.5, NO2, ozone, SO2, and AD particles from 1 February 2015 to 28 February 2015.Table 3 shows the odds ratios for an IQR increase in skin symptoms based on levels of PM2.5, NO2, ozone, SO2, and AD particles. Increases in the levels of PM2.5, NO2, ozone, SO2, and AD particles were not related to an increased risk of skin symptom events in children either with or without atopic dermatitis. The cumulative total summary of the symptom score was 5395 (56.8%) for a score of 1, 800 (8.4%) for 2, 84 (0.9%) for 3, and 3213 (33.8%) for non-observed. In a two-pollutant model adjusted for NO2, ozone, and SO2, PM2.5 and AD particles were not associated with risk of skin symptom events (Table 4 and Table 5).A number of studies have unmasked the deleterious effects of ambient PM on internal organs [2,3,4]. It has gradually become clear that there are four potential mechanisms by which ambient PM exerts adverse effects on the skin, including generation of free radicals, induction of cutaneous inflammatory cascades, activation of an aryl hydrocarbon receptor (AhR)-dependent mechanism, and alteration of cutaneous microflora [5,28]. However, to the best of our knowledge, the epidemiologic evidence focusing on the effects of ambient PM on skin health remains limited. Especially, in the relationship between short-term exposure to ambient PM and skin health, current evidence is lacking. Therefore, the present study investigated associations among daily PM2.5 and skin symptoms in schoolchildren, but found no significant relationship. Similarly, NO2, ozone, SO2, and AD particles were also not associated with skin symptoms. These results suggest that short-term exposure to air pollutants does not have an impact on skin symptoms of schoolchildren in Japan.Kim et al. recently completed a small longitudinal study elucidating a significant association between outdoor levels of PM2.5 and skin symptoms in children with atopic dermatitis [29]. They also concluded that NO2 and volatile organic compounds may aggravate atopic dermatitis; thus, patients with skin disease may be more sensitive to exposure to PM2.5 than are subjects without skin disease. In the current study, our statistical analysis was adjusted for the presence of atopic dermatitis as an individual characteristic. Notwithstanding, irrespective of the presence of atopic dermatitis, the present study found no association with PM2.5 and skin symptoms in schoolchildren.Kim et al. estimated the seasonal associations among skin symptoms and PM2.5 for winter, spring, summer, and autumn. They found a significant association with skin symptoms and PM2.5 only in winter, which exhibited the highest PM2.5 level among the four seasons. The levels of PM2.5 in Seoul, South Korea, where research by Kim et al. was conducted, exceeded 20 μg/m3 and were higher than those reported in the present study. Accordingly, the difference in results between the study of Kim et al. and the present study may depend on differences in levels of PM2.5. Alternatively, low-concentration or short-term exposure to PM2.5 may be insufficient to aggravate skin symptoms.There are many reports suggesting that PM2.5 contributes to various skin diseases, such as inflammatory skin disease, androgenetic alopecia, and skin cancer [30,31,32]. Furthermore, Vierkötter et al. found a strong association of premature skin aging with exposure to PM2.5 in the elderly [33]. Similarly, other air pollutants, such as NO2, ozone, and volatile organic compounds, also have adverse effects on skin health [31,34]. Recent studies suggest the possible mechanisms by which air pollutants such as PM2.5 cause adverse skin effects; such mechanisms include increasing inflammatory cytokines, reducing the barrier function of skin, and activating reactive oxygen species and AhR [2,3,4,6,33]. Thus, PM2.5 may adversely affect skin health in schoolchildren. To better estimate the effects of PM2.5 on skin health and skin symptoms in children, objective evaluation methods such as score of intrinsic and extrinsic skin aging may be needed.An important aspect of dust storms in relationship to health issues is the amount of PM2.5 that they contain. AD is also a serious health concern because of the associated heavy pollution, and it has been associated with increased mortality, emergency treatment for cardiovascular diseases, and hospitalization for pneumonia [14,35,36,37]. Recently, several studies from Japan demonstrated an association between AD and skin symptoms in healthy adult subjects [10,11,38,39]. Therefore, the present study also estimated the daily levels of AD particles and skin symptoms in schoolchildren using LIDAR data but found no such relationship. However, the studies reporting a significant association with AD and skin symptoms estimated the effect of heavy AD on skin health. In these studies, heavy AD was defined as a density greater than 0.6 km−1 or 1.0 km−1. In the present study, the average level of AD particles was 0.02 km−1. Thus, low-concentration exposure to AD particles may be insufficient to aggravate skin symptoms. Alternatively, the effects of AD on skin health may differ between adults and children.There are several limitations in the current study. First, the study duration, which was one month, may not have been long enough to estimate the effects of short-term exposure to ambient PM on skin health. However, our previous one-month study found a significant negative association with ambient PM and respiratory function in schoolchildren [23], suggesting that one month may be sufficient time. Second, this also restricted any assessment of seasonal variation. Regardless, according to the results of Kim et al. [29], PM2.5 during winter was the most concerning with respect to the deleterious effects of ambient PM on skin health. Thus, although the present study was conducted during the most suitable season, such effects may depend on the composition of PM2.5. Third, skin symptom scores were not validated because validated skin symptom scores were not available in Japan at the time of the study. Therefore, as in previous studies which estimated the association with PM2.5 and respiratory symptoms, skin symptom scores were reported as 0 (no symptoms), 1 (mild symptoms), or 2 (severe symptoms) [40]. Finally, we were unable to measure the individual amount of exposure to particulate air pollutants.Short-term exposure to PM2.5 was not associated with skin symptoms in schoolchildren. AD particles are an important source of PM2.5 in Japan. However, there was no relationship between AD particles and skin symptoms. Neither PM2.5 nor AD particles aggravated the skin symptoms of schoolchildren.We would like to thank Atsushi Shimizu (National Institute for Environmental Studies) for providing LIDAR data. This research was supported by the Environmental Research and Technology Development Fund (5-1453) of the Japanese Ministry of the Environment and Tottori prefecture. We would like to thank Editage (www.editage.jp) for English-language editing.Masanari Watanabe, Jun Kurai, Hiroyuki Sano, and Yuji Tohda conceived the study. Masanari Watanabe, Hisashi Noma, Jun Kurai, Hiroyuki Sano, Yuji Tohda, and Eiji Shimizu designed the study. Masanari Watanabe, Hisashi Noma, and Jun Kurai wrote the manuscript. Masanari Watanabe, Jun Kurai, Kyoko Iwata, and Degejirihu Hantan contributed to data collection. Masanari Watanabe, Jun Kurai, Kyoko Iwata, and Degejirihu Hantan performed the laboratory work. Masanari Watanabe, Hisashi Noma, and Kyoko Iwata performed the statistical analysis and interpretation of the results. Masanari Watanabe, Hisashi Noma, and Jun Kurai contributed to critical revision of important intellectual content. All authors read and approved the final manuscript.The authors declare no conflict of interest.Characteristics of the 339 children included in this study.Data are shown as mean ± standard deviation or n (%).Average meteorological and air contaminant levels from 1 to 28 February 2015.Data are shown as mean ± standard deviation.Multivariate analysis using generalized estimating equation (GEE) logistic regression models to assess the association between skin symptoms and interquartile range (IQR) changes in the air contaminant concentrations.CI, confidence interval; NS, not significant.Estimated effects of PM2.5 on skin symptoms in the two-pollutant model after adjustment for NO2, ozone, and SO2.CI, confidence interval; NS, not significant.Estimated effects of AD particles on skin symptoms in the two-pollutant model after adjustment for NO2, ozone, and SO2.CI, confidence interval; NS, not significant.

Med-MDPI/ijerph_2/ijerph-14-03-00300.txt

@@ -0,0 +1,336 @@
+Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).The present work is the first systematic and large scale study on radioactive materials and heavy metals in surface soil around the Bayanwula prospective uranium mining area in China. In this work, both natural and anthropogenic radionuclides and heavy metals in 48 surface soil samples were analyzed using High Purity Germanium (HPGe) γ spectrometry and inductively coupled plasma-mass spectrometry (ICP-MS). The obtained mean activity concentrations of 238U, 226Ra, 232Th, 40K, and 137Cs were 25.81 ± 9.58, 24.85 ± 2.77, 29.40 ± 3.14, 923.0 ± 47.2, and 5.64 ± 4.56 Bq/kg, respectively. The estimated average absorbed dose rate and annual effective dose rate were 76.7 ± 3.1 nGy/h and 83.1 ± 3.8 μSv, respectively. The radium equivalent activity, external hazard index, and internal hazard index were also calculated, and their mean values were within the acceptable limits. The estimated lifetime cancer risk was 3.2 × 10−4/Sv. The heavy metal contents of Cr, Ni, Cu, Zn, As, Cd, and Pb from the surface soil samples were measured and their health risks were then assessed. The concentrations of all heavy metals were much lower than the average backgrounds in China except for lead which was about three times higher than that of China’s mean. The non-cancer and cancer risks from the heavy metals were estimated, which are all within the acceptable ranges. In addition, the correlations between the radionuclides and the heavy metals in surface soil samples were determined by the Pearson linear coefficient. Strong positive correlations between radionuclides and the heavy metals at the 0.01 significance level were found. In conclusion, the contents of radionuclides and heavy metals in surface soil around the Bayanwula prospective uranium mining area are at a normal level.Radiation exposure and heavy metal pollution around uranium mining areas have captured worldwide public attention for several decades [1,2,3,4]. The intensive uranium exploitation and the inappropriate management of the residues have had a harmful impact on the environment [4,5,6]. In recent decades, the dose contribution from technologically enhanced naturally occurring radioactive materials is increasing [7,8]. The worldwide annual effective dose to the public from natural radiation exposure is 2.4 mSv [9], while it is 3.1 mSv in China, which increased from 2.3 mSv in 1990s [10,11,12]. The demand for uranium resources in China is increasing with the development of nuclear power industries [9,13,14] and the rising price of uranium internationally [15]. Consequently, the activities on exploiting uranium ores and their hydrometallurgy processes were heavily strengthened and there are also some reports concerning the environmental contamination around uranium mines [16]. However, there have been few specific studies related to radionuclides and heavy metals assessment from uranium mining areas in China, especially around prospective uranium mining areas. A pre-mining study on radiation levels and heavy metals around uranium mining areas could establish a baseline database on the environmental radiation levels and become an essential reference guide for the future [17]. The aim of this study was to establish the radioactive materials and heavy metals contents from the surface soil around the Bayanwula uranium pre-mining area in China due to the lack of published environmental data, to assess the radiation and heavy metals risk for local residents, and to investigate the correlations between the radionuclides and heavy metals.The Bayanwula uranium mining region is located in the central part of the Sonid Left district, which is in the northwest part of the XilinGol prairie of Inner Mongolia in China. The study area is about 30 km north of the capital of Sonid Left. The altitude ranges from 1040 m to 1255 m. There are around 34,000 residents in Sonid Left. This region has a continental climate with a warm summer and cold winter. The average annual precipitation is less than 200 mm. The study area is characterized by grassland, not cultivated, and no industries. The sampling was carried out in June 2015 prior to the uranium mining activities. The map of the mining area and sampling locations are shown in Figure 1, in which the sampling locations were mapped using the software ESRI Arc GIS desktop 10.1 (Environmental Systems Research Institute, Inc., Redlands, CA, USA) based on the coordinates determined by the Global Positioning System (GPS).As shown in Figure 1, a total of 48 surface soil samples were collected within about a 30 km radius from the center of the mining area. At each sampling location, a square area of 1 m2 was marked out. Then four samples were collected from the surface layer (up to 10 cm) of the four corners of the square area (1 m × 1 m) using a stainless steel cylindrical sampler (Ø10 cm × H10 cm), mixed, and placed in a labeled polythene bag after removing impurities such as stones, gravels, and roots. In the laboratory, each sample was dried in an oven at 100 °C for more than 24 h to remove the moisture content, homogenized, and was separated into two parts. One of them was sieved through a 0.25 mm mesh. A sample of 338.0 g was weighed and sealed in an airtight polythene (Ø75 cm × H70 cm) cylindrical container and left for more than 30 days to allow 226Ra and its decay products to reach secular equilibrium before further gamma-ray measurement. The concentrations of 238U, 226Ra, 232Th, 40K, and 137Cs were determined by a HPGe γ-ray spectrometry system (Oak Ridge Technology & Engineering Cooperation, Oak Ridge, TN, USA).The other part was sieved through 0.150 mm mesh and weighed 0.2–0.5 g with accuracy up to 0.1 mg. They were then digested with a concentrated acid mixture (HNO3, HF, and HClO4) (Analytical reagent, EMD Millipore Corporation, Darmstadt, Germany). The solution was transferred to a 25 milliliter volumetric flask. The content of 7 elements (Cr, Ni, Cu, Zn, As (non-metal trace element), Cd, and Pb) was determined by inductively coupled plasma-mass spectrometry (ICP-MS) (Agilent Technologies Inc., Santa Clara, CA, USA). Lower limits of detection (LLDs) were determined as 10 μg/kg for Cr, 13 μg/kg for Ni, 13 μg/kg for Cu, 8 μg/kg for Zn, 3 μg/kg for As, 0.3 μg/kg for Cd, and 7 μg/kg for Pb in dry soil weight.HPGe γ-ray spectrometry system employed to carry out the radioactivity measurements was based on a high-purity germanium p-type coaxial photon detector made by Oak Ridge Technology & Engineering Cooperation (ORTEC). The detector relative efficiency exceeded 32% while the resolution was better than 1.82 keV at 1.33 MeV 60Co. The γ spectrum of 40 keV–3 MeV was acquired and analyzed using the software Gamma vision (6.01) (Oak Ridge Technology & Engineering Cooperation, Oak Ridge, TN, USA) and a 8192 multichannel analyzer (Oak Ridge Technology & Engineering Cooperation, Oak Ridge, TN, USA). The whole detector system was placed inside a 10 cm lead layer shield. Before and after all sample counting, the backgrounds were measured and were subtracted from the corresponding photopeaks. The energy and efficiency calibrations of the counting system were performed using γ sources of 238U, 234Th, 226Ra, 40K, and 137Cs with the same size of each sample. It took 86,400 s to reduce the counting statistical error for each measurement. The activity concentrations of 238U, 232Th, 226Ra, 40K, 137Cs in the soil samples were determined in Bq/kg dry weight. The activity concentration of 238U was derived from 234Th (63.3 keV). The 232Th in the soil samples was derived from 212Pb (238.6 keV), 208Tl (538.2 keV), and 228Ac (911.2 keV). The 226Ra activity was determined by its daughter radionuclides 214Pb at 351.9 keV and 214Bi at 609.3 keV. The activity concentrations of 40K and 137Cs were derived from the photopeaks of 1460.8 and 661.7 keV, respectively. The minimum detectable activity for each radionuclide was determined from the HPGe γ-ray spectrometry system and samples for the counting time of 86,400 s, and was estimated to be 3.7 Bq/kg for 238U, 0.1 Bq/kg for 232Th, 0.1 Bq/kg for 226Ra, 1.7 Bq/kg for 40K, and 0.01 Bq/kg for 137Cs.The natural radioactivity of building materials is mainly from the 238U series, 232Th series, and 40K. As 98.5% of the radiological effects of the uranium series are produced by radium and its daughter products, the contribution from 238U has been replaced with the decay product 226Ra [18,19]. Therefore, the radiation hazard indices are usually determined by the activity concentrations of 226Ra, 232Th, and 40K.The absorbed γ dose rate (nGy/h) in air at 1 m above the ground for radionuclides (238U series, 232Th series, and 40K) uniformly distributed on the ground was computed by following Equation (1) [9].
+          
+            (1)
+            
+              
+                
+                  D
+                  =
+                  0.462
+                  ×
+                  
+                    A
+                    
+                      Ra
+                    
+                  
+                  +
+                  0.604
+                  ×
+                  
+                    A
+                    
+                      Th
+                    
+                  
+                  +
+                  0.0417
+                  ×
+                  
+                    A
+                    K
+                  
+                
+              
+            
+          
+          where ARa, ATh, and AK are the activity concentrations of 226Ra, 232Th, and 40K (Bq/kg), respectively.The annual effective dose is presented to express the irradiated dose of the human body from natural existing radionuclides in the earth’s crust soil. It is expressed [9] by following Equation (2).
+          
+            (2)
+            
+              
+                
+                  AED
+                  =
+                  D
+                  ×
+                  8760
+                  ×
+                  0.2
+                   
+                  
+                    (
+                    
+                      or
+                       
+                      0.8
+                    
+                    )
+                  
+                  ×
+                  0.7
+                  ×
+                  
+                    
+                      10
+                    
+                    
+                      −
+                      3
+                    
+                  
+                
+              
+            
+          
+          where AED is annual effective dose (μSv/y); D is γ dose rate (nGy/h); the coefficient 0.7 Sv/Gy is for the conversion coefficient from the absorbed dose in air to the effective dose received by adults; 0.2 for the outdoor occupancy factor; 8760 hour/year is equal to 365 days × 24 h per year.Both the radium equivalent activity (Raeq) and the external hazard index (Hex) were equally used to evaluate the effect of the external γ radiation on human beings. The radium equivalent activity and external hazard index were calculated by Equations (3) and (4). The Raeq should not exceed 370 Bq/kg and the Hex should be less than unity [10].
+          
+            (3)
+            
+              
+                
+                  
+                    
+                      Ra
+                    
+                    
+                      eq
+                    
+                  
+                  =
+                  
+                    A
+                    
+                      Ra
+                    
+                  
+                  +
+                  1.43
+                  ×
+                  
+                    A
+                    
+                      Th
+                    
+                  
+                  +
+                  0.077
+                  ×
+                  
+                    A
+                    K
+                  
+                
+              
+            
+          
+          
+            (4)
+            
+              
+                
+                  
+                    H
+                    
+                      ex
+                    
+                  
+                  =
+                  
+                    A
+                    
+                      Ra
+                    
+                  
+                  /
+                  370
+                  +
+                  
+                    A
+                    
+                      Th
+                    
+                  
+                  /
+                  259
+                  +
+                  
+                    A
+                    K
+                  
+                  /
+                  4810
+                
+              
+            
+          The internal hazard index (Hin) was introduced to describe the hazard of radon and its short-lived products in building materials, given by Equation (5) and recommended to be less than unity [10].
+          
+            (5)
+            
+              
+                
+                  
+                    H
+                    
+                      in
+                    
+                  
+                  =
+                  
+                    A
+                    
+                      Ra
+                    
+                  
+                  /
+                  185
+                  +
+                  
+                    A
+                    
+                      Th
+                    
+                  
+                  /
+                  259
+                  +
+                  
+                    A
+                    K
+                  
+                  /
+                  4810
+                
+              
+            
+          The lifetime cancer risk (LTCR) was obtained by Equation (6) [11,12]:
+          
+            (6)
+            
+              
+                
+                  LTCR
+                  =
+                  AED
+                  ×
+                  DL
+                  ×
+                  RFSE
+                
+              
+            
+          
+          where DL is the duration of lifetime, 70 years; and RFSE is the risk factor for stochastic effects of the common population, 0.055/Sv [12].Human beings are exposed to soil metals through the ingestion and inhalation of dust particles through the mouth and nose, and dermal contact [20,21]. The health risk assessment model used in this study was developed by the US Environmental Protection Agency [22,23]. The doses are calculated as follows:
+        
+          (7)
+          
+            
+              
+                
+                  D
+                  
+                    ing
+                  
+                
+                =
+                
+                  
+                    C
+                    ×
+                    IngR
+                    ×
+                    EF
+                    ×
+                    ED
+                  
+                  
+                    BW
+                    ×
+                    AT
+                  
+                
+                ×
+                
+                  
+                    10
+                  
+                  
+                    −
+                    6
+                  
+                
+              
+            
+          
+        
+        
+          (8)
+          
+            
+              
+                
+                  D
+                  
+                    inh
+                  
+                
+                =
+                
+                  
+                    C
+                    ×
+                    InhR
+                    ×
+                    EF
+                    ×
+                    ED
+                  
+                  
+                    PEF
+                    ×
+                    BW
+                    ×
+                    AT
+                  
+                
+              
+            
+          
+        
+        
+          (9)
+          
+            
+              
+                
+                  D
+                  
+                    dermal
+                  
+                
+                =
+                
+                  
+                    C
+                    ×
+                    SA
+                    ×
+                    SL
+                    ×
+                    ABS
+                    ×
+                    EF
+                    ×
+                    ED
+                  
+                  
+                    BW
+                    ×
+                    AT
+                  
+                
+                ×
+                
+                  
+                    10
+                  
+                  
+                    −
+                    6
+                  
+                
+              
+            
+          
+        
+        where Ding, Dinh, Ddermal are the average daily intake through ingestion, inhalation, and dermal absorption in mg/(kg·day), C is the concentration of metal in the soil (mg/kg), IngR and InhR are the ingestion and inhalation rate of soil, respectively (mg/day, m3/day), EF is the exposure frequency (day/year), and ED is the exposure duration (year). SA is the exposed skin area (cm2), SL is the skin adherence factor, ABS is the dimensionless dermal absorption factor, PEF is the particle emission factor in m3/kg, BW is the average body weight (kg), and AT is the average time (day). The doses calculated for each element and exposure pathway are subsequently divided by the corresponding reference dose (RfD) (mg/(kg·day)) to yield a hazard quotient (HQ) (or non-cancer risk), whereas for carcinogens, the dose is multiplied by the corresponding slope factor (SF) (mg/(kg·day))−1 to produce a level of cancer risk. The hazard index (HI) is then the sum of HQ [24]. If HI < 1, it is believed that there is no significant risk of non-carcinogenic effects and the magnitude of risk increases as HI increases [23]. Carcinogenic risk is used to estimate the probability of an individual developing any type of cancer from the lifetime exposure to carcinogenic hazards. The acceptable risk for regulatory purposes is in the range of 10−6–10−4 [20]. These values indicate that one additional case in a population of 1 million to one in 10,000 people is acceptable. In this study, hazard index methods and cancer risk methods were used to assess health risks of metal exposure to children and adults in the Bayanwula uranium pre-mining area. The detailed corresponding parameters are presented in Table 1 [20,21,23,25].The activity concentrations of radionuclides (238U, 232Th, 226Ra, 40K, and 137Cs) in 48 surface soil samples around the Bayanwula prospective uranium mining area are presented in Table 2. The mean values of 238U, 232Th, and 226Ra are lower than the China and world mean values. However, the mean value of 40K was around two times higher than both the worldwide and China’s average of 412 [6] and 580.0 Bq/kg [26], respectively. The activity concentration of 137Cs was 5.64 Bq/kg, which was the anthropogenic radionuclide from nuclear weapon tests or nuclear power accidents. The absorbed γ dose rate in air, annual effective dose, hazard indices, and lifetime cancer risk calculated from radionuclides in soil samples are shown in Table 3. The calculated mean outdoor γ dose rates was 76.7 nGy·h−1, which was higher than the world average of 60 nGy/h [6] and the Chinese mean value of 62.8 nGy/h [26]. The mean value of radium equivalent activity was 138.0 Bq/kg, lower than the reference value of 370 Bq/kg. The external and internal hazard indices did not exceed unity, which indicates that the γ radiation from the soil was at a safe level. The lifetime cancer risk was 3.2 × 10−4/Sv, which was also at a very low level.The contents of heavy metals (Cr, Ni, Cu, Zn, As, Cd, and Pb) in surface soil from the prospective uranium mining area, background values of Inner Mongolia, mean values of China [27,28], and China soil guidelines [29] are also given in Table 2. The mean concentrations of Cr, Ni, Cu, Zn, As, and Cd are much lower than both the national mean backgrounds and the grade І soil quality standard (This level is mainly applicable to the national nature reserve except for the high background areas). However, the concentration of Pb is much higher than China’s background value and within the grade II soil quality standard (The level is mainly applied to general farmland, vegetable land, tea garden, orchard, pasture, and other soil; the soil quality basically could not cause harm and pollution to plants and the environment). The results of the health risk assessment of the heavy metals in the soil around the study area are listed in Table 4 for children and Table 5 for adults. For non-cancer risk, the ingestion dose of the heavy metals is significant for children and adults. The non-cancer risk of the heavy metals for children is higher than that for adults. The hazard indices (HIs) decrease in the order of Pb > Cr > As > Ni > Cu > Cd > Zn for both children and adults are all lower than unity. For cancer risk, Cr, Ni, As and Cd were considered in this study. The cancer risk from the heavy metals is much lower than the acceptable range of 10−4. It can be clearly seen from the tables that the non-cancer risk is more important than the cancer risk for both children and adults. These results indicate that both the cancer and non-cancer risks for the children and adults living around the Bayanwula prospective uranium mining region are all at acceptable levels.The correlations between the natural radionuclides and the heavy metals in the surface soil samples were performed using the SPSS computer package, Version 19 for Windows. The statistical significance of the Pearson correlation was determined by the t test [30,31]. If a value was close to zero, there was no association between the two elements. The terms “weak”, “moderate”, and “strong” were presented for correlation coefficients of 0.2–0.4, 0.4–0.6, and >0.6, respectively [31]. The alpha level for testing significance was set at 0.01 and 0.05. The Pearson correlations of the heavy metals and radionuclides are shown in Table 6. It was found that 238U was weakly positively correlated with 232Th and 226Ra at the 0.05 significance level. A strong positive correlation between 232Th and 226Ra at the 0.01 significance level was present. Both the radionuclides 232Th and 226Ra moderately positively correlated with Cr and Zn, and weakly correlated with 40K and Ni. There were also strong positive correlations between heavy metals: Cr and Zn, Ni and Cu, and Cu and Zn. These strong correlations among metals and radionuclides suggest their common origin. However, there are observed moderate or strong negative correlations between the radionuclide 40K with Ni, Cu, and Zn at the 0.01 significance level. Additionally, it was found that no correlations exist between the radionuclides and heavy metals, i.e., Cr and 40K. The absence of correlations could be explained by the mutual independence or different behavior of the elements.The radionuclides (238U, 232Th, 226Ra, 40K, 137Cs) and heavy metals were measured in 48 surface soil samples from the Bayanwula prospective uranium mining area in China. Activity concentrations of 238U, 232Th, and 226Ra were lower than the world average except for 40K. The values obtained were within the acceptable limits. The annual effective dose and various radiation hazard indices indicate that there is low radiological risk to the local populations around the uranium mining area. The contents of the heavy metals Cr, Ni, Cu, Zn, As, and Cd were within the Chinese soil guidelines Grade І except for Pb, which was about three times higher than the average of China. The non-cancer risk index and cancer risk were estimated to be less than the acceptable limits. The risks of heavy metals for children are all higher than that for adults. A strong positive correlation between radionuclides and heavy metals at the 0.01 significance level was found which suggests their common origin. The correlation study also indicated negative and weak correlations between the radionuclides and heavy metals. This study established the baseline information regarding the natural, artificial radioactivity, and heavy metals status around the Bayanwula prospective uranium mining area in China. To the best of our knowledge, this is the first systematic and large scale study on radiation levels around prospective uranium mining areas in China. These background data could be an important reference for public environmental concerns.The authors would like to thank Guilin Bai from the XilinGol center for disease control and prevention, and Huhejiletu from the Sonid Left center for disease control and prevention for their help with the sampling work.Haribala Bai, Bitao Hu, and Yuhong Li wrote the article; Haribala Bai, Chengguo Wang, Gerilemandahu Sai, Xiao Xu, and Shuai Zhang performed the sampling work, conducted the experiments, and processed the data; Shanhu Bao mapped Figure 1.The authors declare no conflict of interest.The map of the Bayanwula uranium mining area and sampling locations.Parameters used to evaluate the exposure risk of soil metals.The contents of radionuclides (Bq/kg) and heavy metals (mg/kg) in surface soil samples around the Bayanwula prospective uranium mining area.a Activity concentration ± expanded uncertainty, b SD represents standard deviation; MVC: Mean values in China; CSG І: Chinese soil guidelines Grade І; CSG II: Chinese soil guidelines Grade II; WAV: world average values.The radiation hazard indices and lifetime cancer risk.Daily doses, hazard quotients, hazard indices, and cancer risks of heavy metals for children.Daily doses, hazard quotients, hazard indices, and cancer risks of heavy metals for adults.The pearson correlation matrix for the natural radionuclides and the heavy metals.a Correlation is significant at the 0.05 level; b Correlation is significant at the 0.01 level.

Med-MDPI/ijerph_2/ijerph-14-03-00301.txt

@@ -0,0 +1 @@
+Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).Clopidogrel has significantly reduced the incidence of recurrent atherothrombotic events in patients with acute coronary syndrome (ACS) and in those undergoing percutaneous coronary intervention (PCI). However, recurrence events still remain, which may be partly due to inadequate platelet inhibition by standard clopidogrel therapy. Genetic polymorphisms involved in clopidogrel’s absorption, metabolism, and the P2Y12 receptor may interfere with its antiplatelet activity. Recent evidence indicated that epigenetic modification may also affect clopidogrel response. In addition, non-genetic factors such as demographics, disease complications, and drug-drug interactions can impair the antiplatelet effect of clopidogrel. The identification of factors contributing to the variation in clopidogrel response is needed to improve platelet inhibition and to reduce risk for cardiovascular events. This review encompasses the most recent updates on factors influencing pharmacokinetic and pharmacodynamic responses to clopidogrel.Dual antiplatelet therapy with aspirin and P2Y12 inhibitors is crucial for patients with acute coronary syndrome (ACS) and post percutaneous coronary intervention (PCI) to prevent future thrombotic events [1]. Clopidogrel, an oral irreversible P2Y12 receptor antagonist, is widely used in clinical practice in comparison to other P2Y12 antagonists such as ticagrelor or prasugrel. After intestinal absorption, approximately 85% of the clopidogrel prodrug is hydrolyzed by esterase into an inactive form, leaving only 15% of clopidogrel transforming to the active metabolite by the hepatic cytochrome P450 (CYP450) system, of which CYP2C19 is a crucial enzyme. However, vast studies have shown broad inter-individual variability in the antiplatelet effect of clopidogrel. Impaired platelet responsiveness to clopidogrel may result in increased risk of cardiovascular events [2,3]. Numerous studies have demonstrated the association between CYP2C19 polymorphisms and the antiplatelet effect of clopidogrel. Also, other factors including epigenetics, demographics, concurrent diseases, and drug-drug interactions may contribute to the poor response. Our review attempts to demonstrate the comprehensive components affecting pharmacodynamics and pharmacokinetics that can explain the mechanisms underlying clopidogrel response variabilities.Polymorphisms in genes responsible for the drug efflux (ABCB1), metabolic activation or inactivation (CES1, CYP2C19, CYP3A4/5), and biological activity (P2RY12, PEAR1) of clopidogrel may account for a part of the interindividual variability in clopidogrel response (Figure 1, Table 1).Clopidogrel functions only when absorbed by the intestine after oral administration. Evidence has shown that clopidogrel absorption is limited by the intestinal efflux transporter P-glycoprotein (P-gp) encoded by ABCB1 gene. Taubert et al. firstly demonstrated that variable intestinal clopidogrel absorption was influenced by the ABCB1 C3435T polymorphism in 60 patients with coronary artery disease [4]. Then Simon et al. found that carriers of the ABCB1 3435 TT genotype had a higher rate of cardiovascular events than CC homozygotes in patients with acute myocardial infarction (AMI) [5]. Subsequent clinical studies and meta-analysis verified the association between ABCB1 3435TT genotype and impaired platelet response as well as the higher risk of major adverse cardiovascular events [6,7]. Several trials have shown that ABCB1 3435T was associated with lower levels of plasma clopidogrel and its active metabolite [8,9,10]. However, there are also inconsistent reports on the association of ABCB1 polymorphism and clopidogrel response. For example, a recent meta-analysis including six studies with 10,153 subjects failed to show an association between the ABCB1 C3435T polymorphism and the risk of overall recurrent ischemic events, stent thrombosis, or bleeding in clopidogrel treated patients [11], which was further confirmed by Jaitner et al. in patients undergoing PCI [12].Carboxylesterase (CES) is the most predominant hydrolytic enzyme in the human body. CES catalyzes the hydrolysis of numerous ester- and amide-containing endogenous compounds, toxins, and medications to their respective free acids. The vast majority of absorbed clopidogrel is shunted by CES1 to inactive carboxylic metabolites [41]. Therefore, genetic variations affecting CES1 expression or its activity are supposed to be important determinants of clopidogrel response. CES1 has two isotypes, CES1A1 (often called CES1) and CES1P1. Previous studies have identified several single nucleotide polymorphisms (SNPs) in the coding region of CES1, including rs71647871 (G143E), rs71647872 (D260fs), and the intronic variant rs8192950 [13,42]. There is a study showing that the rs8192950 polymorphism is associated with a decreased risk of clinical events in clopidogrel treated patients with extracranial or intracranial stenosis [13]. The frameshift mutation rs71647872 is extremely rare. The G143E results in the non-conservative amino acid substitution of Glycine 143 to Glutamic acid, which decreases CES1 catalytic activity [42]. Lewis et al. found that carriers of the CES1 143E allele have higher levels of the clopidogrel active metabolite and better clopidogrel response than the 143G allele (wild-type) in healthy people [14]. Meanwhile, in patients with coronary heart disease treated with clopidogrel, the lower ADP-induced platelet aggregation and lower risk of cardiovascular events were found in 143E allele carriers. Tarkiainen et al. also reported in healthy volunteers that CES1 143E carriers have a larger AUC of clopidogrel and the active metabolite and lower P2Y12-mediated platelet aggregation [15]. In addition, CES1P1 rs3785161 was found to be associated with attenuated antiplatelet effect of clopidogrel in 162 coronary heart disease patients [16].Paraoxonase-1 (PON1) is an esterase synthesized in the liver and associated with HDL (high density lipoprotein)-cholesterol. A previous study has shown that PON1 is a crucial enzyme for clopidogrel biotransformation to the active metabolite through hydrolytic cleavage of the c-thiobutyrolactone ring of 2-oxo-clopidogrel [17]. The PON1 Q192R (rs662) variant was initially reported to activate clopidogrel more efficiently [17]. However, the following investigations did not replicate the results of Bouman et al. [18]. Mega et al. reported that the Q192R genetic variant was not associated with the pharmacologic or clinical response to clopidogrel, and their results of the meta-analysis including 13 studies also demonstrated no statistically significant association between the 192Q variant and major adverse cardiac events (MACE) during clopidogrel therapy [19]. Moreover, the work by Dansette et al. showed that the second step enzymatic conversion mainly depends on the P450 pathway from 2-oxo-clopidogrel to 4b cis, but also depends on PON1 to minor metabolite 4b “endo”, whose antiplatelet activity was not determined yet [43]. These results suggest that the role of PON1 in clopidogrel resistance may not be important.The CYP2C19 isoenzyme is involved in the two-step reaction of clopidogrel activation. Hulot et al. firstly found the association between the CYP2C19 loss-of-function (LOF) allele (*2) and a marked decrease in platelet responsiveness to clopidogrel in young healthy male volunteers in 2006 [20]. Several following studies have demonstrated the association between CYP2C19 genetic variants (*2, *3, and *17) and the risk of adverse cardiovascular outcomes in clopidogrel-treated patients [21,22,23,24,25]. A genome-wide association study (GWAS) demonstrated that the CYP2C19*2 variant was associated with poor clopidogrel response (p = 4.3 × 10−11) in healthy people and increased ischemic events (p = 0.02) during a 1-year follow-up in patients [23]. Wei et al. confirmed that CYP2C19*2 was associated with higher rate of clopidogrel resistance and ischemic events [21]. Although Bhatt et al. failed to observe the association in patients with stable angina [22], two large meta-analyses have demonstrated the significant association between CYP2C19 LOF and recurrent cardiovascular events in different ethnic patients [44,45]. Meanwhile, the CYP2C19 LOF had a significant reduction of AUC0–t, the concentration of clopidogrel, or active metabolite [24,26,46]. In March 2010, the U.S. Food and Drug Administration (FDA) even announced a boxed warning on clopidogrel, stating that CYP2C19 LOF which harbors two reduced function alleles (*2 and *3) reduces CYP2C19 catalytic activity and attenuates the efficacy of clopidogrel. After that, the American College of Cardiology Foundation and the American Heart Association published a consensus document addressing this FDA warning [47]. However, CYP2C19 LOF allele carriage accounts for only 5% to 12% of the overall variability of the clopidogrel response [23].A CYP2C19 gain-of-function (GOF) allele (*17) in the 5-flanking region of the gene is observed to be associated with increased CYP2C19 transcription [48]. This GOF allele confers a rapid metabolism of CYP2C19 substrates, which may lead to a higher concentration of clopidogrel active metabolite, an enhanced antiplatelet activity, and an increased risk of bleeding events during clopidogrel therapy [27]. Hamsze et al. also confirmed that the CYP2C19*17 polymorphism was associated with decreased on-treatment platelet reactivity and increased risk of major bleedings [27,28]. In the 2013 updated Clinical Pharmacogenetics Implementation Consortium Guidelines for CYP2C19 Genotype and Clopidogrel Therapy, CYP2C19 genotype-guided clopidogrel therapy was recommended to ACS patients managed with PCI [49].CYP3A consists of the 3A4 and 3A5 isoenzymes and is responsible for the conversion of 2-oxo clopidogrel into active clopidogrel metabolites. Therefore, reduced CYP3A4/5 activity is supposed to decrease clopidogrel response. Between the two isoenzymes, CYP3A4 is the dominant form and CYP3A5 acts as a so-called “backup system” in situations where drugs may act as inhibitors of CYP3A4 [50].CYP3A4*1G has been reported to be associated with decreased CYP3A4 expression [29]. However, Danielak et al. reported that influence of CYP3A4*1G was not found on either the pharmacokinetics or pharmacodynamics of clopidogrel [51].The CYP3A5*3 allele, a functional SNP located in intron 3, results in a premature truncated protein associated with null enzymatic activity, and CYP3A5*3/*3 homozygotes lack CYP3A5 protein expression and activity in the liver [52]. The influence of CYP3A5*3 polymorphism on clopidogrel response may be dependent on CYP2C19 genetic status and CYP3A4 inhibitors. Patients with the CYP3A5*3/*3 genotype exhibited higher platelet reactivity compared to carriers of the CYP3A5*1 allele in CYP2C19 poor metabolizers [30]. This may help to explain that when reduced CYP2C19 activity and CYP3A4 substrates or inhibitors occur, the CYP3A5 backup system for CYP3A4 would play a role. Nakkam N et al. also reported that the impact of CYP3A5*3 on clopidogrel response is pronounced in subjects carrying CYP2C19 LOF [31]. In patients treated with amlodipine, a CYP3A4 inhibitor, CYP3A5 non-expressers (*3/*3 homozygotes) showed higher on-treatment platelet reactivity [32].Activation of P2Y12 receptor leads to sustained platelet aggregation via the phosphoinositide 3-kinase (PI3K) pathway to activate glycoprotein IIb/IIIa. Fontana et al. identified three SNPs and one nt insertion in the P2RY12 (i-C139T, i-T744C, G52T, i-ins801A), and two haplotypes called H1 and H2 [33] were inferred from the four polymorphisms; carriers of the H2 haplotype exhibited enhanced platelet activity. In another study, Rudez et al. identified haplotype F was associated with higher on-clopidogrel platelet reactivity [34]. However, a subsequent study failed to find an association between P2RY12 polymorphism (T774C) and clopidogrel responsiveness in 597 ACS patients [35]. Therefore, more studies are necessary to corroborate the relation between P2RY12 genetic polymorphisms and clopidogrel response.Platelet Endothelial Aggregation Receptor-1 (PEAR1), a platelet transmembrane protein, is associated with platelet aggregation and endothelial function. PEAR1 polymorphisms have been shown to influence platelet reactivity after antiplatelet therapy [36]. PEAR1 rs41273215 and rs57731889 were independent predictors of high on-treatment platelet reactivity and low on-treatment platelet reactivity, respectively, in Chinese coronary heart disease after PCI [37]. Other SNPs, including rs2768759 and rs11264579, were also reported to increase platelet activity [38,39]. Lewis et al. also evaluated the impact of PEAR1 rs12041331 polymorphism on platelet aggregation and clinical outcomes in three studies. They found that carriers of the rs12041331 A allele were more likely to experience cardiovascular events or die compared to those of GG homozygotes [40]. However, this observation suggested the effect of rs12041331 on post-aspirin platelet aggregation is mediated through collagen receptor pathways and not ADP-dependent pathways. Therefore, further studies are necessary to define the precise role of PEAR1 polymorphism in antiplatelet therapy.Though pharmacogenetics studies have found several SNPs associated with clopidogrel response, the effect of most polymorphisms on the individual variation of platelet activity has not been fully confirmed except for CYP2C19 LOF. Recent studies have indicated that the epigenetic modification of genes involved in drug disposition or effects can also affect the drug response. Epigenetic modification can affect gene expression and chromatin structure without altering the nucleotide sequence, and is influenced by physiological and pathological conditions and environmental factors as well. Interest in the epigenetic study of clopidogrel response has also increased in recent years. Most of these studies regarding clopidogrel are focused on microRNA and DNA methylation.MicroRNAs (miRNAs) are single stranded, short, and small noncoding RNA of ~22 nucleotides in length [53]. MiRNAs can reduce mRNA expression by binding to the target mRNA directly and interfering with protein translation [54]. Numerous studies have explored the link between specific mRNAs and miRNAs to platelet reactivity and activation [55,56,57]. For example, Kondkar et al. have demonstrated that miR-96 regulates the expression of platelet vesicle-associated microtubule protein 8 (VAMP8), a critical component of platelet granule exocytosis [56]. Girardot et al. also indicated that miR-28 regulates the expression of the thrombopoietin receptor directly [57].In 377 miRNAs observed in human platelets, miR-223 was the most differentially expressed miRNA in platelet-rich plasma compared with platelet-poor plasma and serum [58]. Circulating platelet miRNAs are also supposed to act as indicators for tailoring antiplatelet therapies. MiR-223 is in the highest level among platelet miRNAs and could suppress P2RY12 mRNA level in HEK293 cells [59]. Decreased miR-223 expression in platelet and plasma predicted high on-treatment platelet reactivity in clopidogrel treated patients [60,61], which indicated that the miR-223 level might serve as a potential biomarker to predict clopidogrel response. MiR-26a was found to participate in the regulation of platelet reactivity by clopidogrel via regulating the expression of vasodilator-stimulated phosphoprotein [62].DNA methylation is specially observed in the context of the cytosine phosphate guanine (CpG) dinucleotide and is mostly studied in promoter regions or gene bodies and represses gene transcription [63].ABCB1 promoter methylation was reported to suppress ABCB1 mRNA and protein expression in tumor cells [64,65,66]. Hypomethylation of ABCB1 promoter was associated with poor response to clopidogrel in Chinese ischemic stroke patients with CYP2C19*1/*1 genotype [67]. ABCC3, another member of the ABC family, was associated with the efflux of clopidogrel and its antiplatelet activity [68,69]. However, ABCC3 promoter methylation and down-regulation of ABCC3 mRNA had no significant association with clopidogrel response [70]. Different detection methods of platelet activity and different subjects result in different conclusions. Therefore, further studies are needed to corroborate the conclusion. Hypomethylation of P2RY12 promoter was associated with clopidogrel resistance in coronary artery disease (CAD) patients with smoking, albumin <35 g/L, and alcohol abuse [71]. A recent epigenome-wide study revealed that lower methylation of cg03548645 within TRAF3 body was associated with increased platelet aggregation and vascular recurrence in ischemic stroke patients treated with clopidogrel [72]. They hypothesized that higher TRAF3 expression due to decreased methylation may lead to an increase in the CD40 signal pathway interfered platelet-platelet interactions [73,74].Variation in platelet inhibitory effects of clopidogrel has been shown to be associated with the genetic factors, including polymorphisms and epigenetics. However, those data of genetic variations are insufficient to explain the varieties in clopidogrel response. Non-genetic factors, such as demographic characteristics, concurrent diseases, and drug interactions, are also observed to have an influence on the antiplatelet effect of clopidogrel.Age and BMI were significantly associated with clopidogrel response [75,76]. Older age was independently associated with a higher rate of high residual on-treatment platelet reactivity with both clopidogrel and ticagrelor in 494 patients on dual antiplatelet therapy [77]. Additionally, obesity (BMI ≥ 30 kg/m2) was independently associated with higher residual platelet reactivity in clopidogrel-treated patients [78].Several trials have demonstrated that smokers show better clopidogrel responsiveness than nonsmokers, which is called the smokers’ paradox. Gurbel et al. assessed the effect of smoking on clopidogrel and prasugrel therapy in patients with CAD. Lower clopidogrel active metabolite exposure and decreased antiplatelet effects of clopidogrel were observed in nonsmokers compared to smokers, but the same phenomenon was not observed for prasugrel, another P2Y12 antagonist [79]. Park et al. found that better clopidogrel response could be reversed after discontinuation of smoking, which further confirmed the causal relationship between smoking and clopidogrel [80]. An explanation for the smokers’ paradox is CYP1A2 and CYP2B6 induction by cigarette smoking, which results in greater formation of the clopidogrel active metabolite [79,81]. Also, CYP1A2 rs762551 polymorphism showed influence on clopidogrel response only in smokers [81]. However, a recent study observed a significant inverse correlation between the VerifyNow P2Y12 reaction unit and hemoglobin levels in current smokers receiving clopidogrel therapy, and the difference in the platelet reaction unit (PRU) between nonsmokers and current smokers disappeared after adjusting for the effect of hemoglobin on PRU [82]. The exact mechanisms for these observations require further study. However, the results of Ferreiro et al. and Zhang et al. are contrary to these findings [83,84]. Their studies showed that cigarette smoking is an independent risk factor for adverse ischemic outcomes with a single antiplatelet agent or in non-cardioembolic ischemic stroke patients. Moreover, in smokers there was a trend of lower composite vascular events in clopidogrel-treated patients as compared with the aspirin-treated patients, whereas no such trend was observed in never-smokers. It is necessary to confirm under which circumstances the effect of the smokers’ paradox on clopidogrel will be established.Patients with diabetes mellitus (DM) account for the most proportion of the population worldwide. The prevalence of cardiovascular diseases (CVDs) rises to as high as 55% in DM patients, and CVDs account for 65% of deaths in DM patients [85]. Diabetic patients have a poor antiplatelet effect of clopidogrel treatment compared to non-diabetic patients [86,87]. The mechanism of this phenomenon has been explored by some investigators but is inconsistent. Insulin can regulate platelet activity through the insulin receptor substrate-1 (IRS-1) on platelets. Type 2 DM is characterized by reduced insulin sensitivity which may lead to increased platelet activity and decreased antiplatelet activity. IRS-1 rs956115 and rs13431554 polymorphisms were associated with high platelet activity and increased risk of adverse events in type 2 DM CAD patients [88,89]. Moreover, Angiolillo et al. explored the mechanism of clopidogrel response variability in 60 diabetic and non-diabetic patients treated with aspirin and clopidogrel [90]. The maximal plasma concentration of the clopidogrel active metabolite and the area under the concentration-time curve was lower in the diabetic group. In addition, significantly higher platelet activity was observed in vitro incubation of clopidogrel active metabolites with platelets from DM patients as compared to those from non-DM patients. These results suggested that poor clopidogrel response in DM is mainly due to the decreased concentration of the active metabolite but also partly attributed to the upregulation of the P2Y12 signaling pathway [90]. When the antiplatelet effect of clopidogrel is influenced by DM, switching to other potent antiplatelet drugs may be considered. Results from the TRITON-TIMI 38 trial have also revealed that prasugrel tends to provide a greater reduction in ischemic events than clopidogrel in diabetic patients compared with non-diabetic patients with ACS [91]. In the PLATO trial, a consistent benefit with ticagrelor over clopidogrel, including reduced mortality, was observed irrespective of diabetic status [92]. The cilostazol plus clopidogrel therapy achieved greater platelet inhibition compared with clopidogrel alone in T2DM and CYP2C19 LOF variants [93]. Therefore, the novel and more potent P2Y12 receptor inhibitors (prasugrel and ticagrelor) or cilostazol may be a better strategy to overcome clopidogrel poor response in patients with DM.Chronic kidney disease (CKD, creatinine clearance <60 mL/min) is a common comorbidity of patients with atherosclerotic vascular disease. CKD was associated with increased risk of recurrent cardiovascular and bleeding events in PCI-treated patients [94,95]. Siddiqi et al. also reported a higher risk of death, myocardial infarction, and bleeding in patients with CKD [96]. It is suggested that prolonging clopidogrel beyond the standard guidelines of 12 months after PCI may help to reduce this long-term increased risk in patients with CKD [96]. However, other researchers drew inconsistent conclusions. Mangiacapra et al. found no association between residual platelet reactivity and CKD when the VerifyNow P2Y12 assay was used, but some association between CKD and ischemic and bleeding events [97]. Baber et al. found that the association between residual platelet reaction and CKD was attenuated after multivariable adjustment, which implied that confounding risk factors, rather than renal dysfunction itself, account for the high platelet reactivity (HPR) [98]. Moreover, the HPR could result in increased risk of ischemic and bleeding events irrespective of CKD status. A large meta-analysis also failed to show the definite association between antiplatelet therapy and CKD [99], suggesting that this association needs more studies to be proven.Numerous studies are focused on the drug-drug interactions between the proton pump inhibitor, calcium channel blocker, statin, morphine, or caffeine and clopidogrel.Both clopidogrel and aspirin can show the symptoms of gastrointestinal bleeding. Patients with clopidogrel treatment are usually recommended for proton pump inhibitors (PPIs) to prevent gastrointestinal complications such as ulceration and bleeding. Omeprazole and esomeprazole, two moderate CYP2C19 inhibitors, have been reported to decrease clopidogrel efficacy and increase risk of adverse clinical outcomes [100,101]. The weak CYP2C19 inhibitor pantoprazole has less effect on the clopidogrel response [101]. In 2010, the FDA and the European Medicines Agency issued a warning on concomitant administration of clopidogrel and PPIs, especially for omeprazole and esomeprazole. Clinically, more patients are recommended for pantoprazole treatment. However, some observational studies also showed discordant findings for the concomitant use of PPIs. Chiara et al. suggested that PPIs as a class were associated with worse clinical outcomes in observational studies of patients, with unstable angina/non-ST-segment elevation myocardial infarction patients receiving clopidogrel and aspirin therapy, but omeprazole showed no difference in ischemic outcomes as compared with the placebo [102]. The COGENT trial demonstrated that omeprazole therapy did not lead to an increased risk of cardiovascular events but significantly attenuate the gastrointestinal risk [103]. These controversial conclusions reveal that further prospective research is still needed to determine the clinical significance of clopidogrel-PPI interactions.As some calcium channel blockers (CCBs) can inhibit CYP3A4, a key enzyme in the conversion of clopidogrel, the concomitant use of CCBs might compete with clopidogrel for the CYP3A4 enzyme, which can also result in an impaired response to clopidogrel. Several studies have reported the impacts of CCBs on the platelet reactivity and clinical outcomes of clopidogrel [104]. CYP3A4*1G allele decreased the CYP3A4 enzyme activity, and was associated with an increased vulnerability to the effects of CCBs on clopidogrel response variation [105]. Park et al. indicated that taking a 600 mg loading dose of clopidogrel may reduce this impact [105]. Meanwhile, some CCBs also show strong inhibitory effects on P-gp and lead to decreased intestinal efflux of clopidogrel, thereby increase plasma concentration of clopidogrel. Therefore, the coadministration of P-gp inhibiting CCBs (such as verapamil, nifedipine, diltiazem, and barnidipine) can counteract the reduced effect of CCBs on the metabolic activation of clopidogrel, while non-P-gp inhibiting CCBs such as amlodipine still diminish the efficacy of clopidogrel [106].Statins, known as HMG-CoA reductase inhibitors, are widely used to treat hypercholesteremia. As CCBs, it is possible that CYP3A4-metabolized statins (atorvastatin, simvastatin) also influence clopidogrel response. Lau et al. firstly reported that atorvastatin other than pravastatin reduced the antiplatelet effect of clopidogrel by controlling CYP3A4 activity in a dose-dependent manner [107]. Horst et al. also reported that pre-treatment with atorvastatin and simvastatin reduced the inhibitory effects of clopidogrel significantly in patients with CAD [108]. Additionally, the platelet inhibitory effect of clopidogrel was enhanced by replacement of atorvastatin with a non-CYP3A4-metabolized statin in patients with high platelet reactivity [109]. However, these results did not conform to several subsequent studies [110,111,112], which revealed no significant difference between atorvastatin and other statins affecting the clinical efficacy in ACS patients receiving clopidogrel therapy. Saw et al. also did not observe adverse interaction between clopidogrel and statins after long-term use [112]. Therefore, clinical significance of this putative drug-drug interaction is still controversial. These conflicting observations may be due to differences in the assessment assay of platelet activity, use of different doses or classes of statin, and patient selection.Morphine is sometimes recommended for the treatment of pain from myocardial infarction. Hobl et al. performed a randomized, double-blind, placebo-controlled, cross-over trial in 24 healthy volunteers receiving a 600 mg loading dose of clopidogrel simultaneously with 5 mg morphine or placebo. They found that morphine delayed the Tmax of clopidogrel and decreased the AUC0−n of clopidogrel active metabolite by 34%. In addition, the time to maximal inhibition in platelet aggregation was two-fold longer than the placebo group. As morphine can delay gastric emptying, it is possible that morphine might delay clopidogrel absorption, decrease peak plasma levels, and decrease its antiplatelet activity [113,114]. Interestingly, this phenomenon was also seen in both prasugrel and ticagrelor, suggesting that morphine and oral antiplatelet interaction is non-drug specific and is most likely explained by morphine’s gastrointestinal side effects [115]. The randomized, double-blind, placebo-controlled IMPRESSION trial showed that morphine lowered the total exposure to ticagrelor and its active metabolite by 36% and 37%, respectively, with a concomitant delay in maximal plasma concentration of ticagrelor. Moreover, a higher platelet reactivity was found in the group that received morphine [116]. Therefore, the use of morphine should be careful during the use of oral antiplatelet medications.Tea and coffee that contain caffeine are the most widely consumed beverages in daily life. A few investigations have exhibited the influence of caffeine intake on the cardiovascular system [117]. Clopidogrel inhibits the platelet P2Y12 receptor, leading to increased intracellular cyclic adenosine monophosphate (cAMP) level. It is observed that caffeine can also increase cAMP accumulation and alter the aggregation of platelets by upregulating the adenosine A2A receptor [118]. Acute caffeine administration was reported to increase the platelet inhibition effect of clopidogrel in healthy subjects and CAD patients [119]. The extent of influence of caffeine on clopidogrel responsiveness requires further examination.In recent years, the adverse cardiovascular events resulting from clopidogrel resistance have raised increasing concerns. Therefore, it is meaningful to determine the specific factors contributing to the variability in clopidogrel clinical treatment. However, the mechanisms underlying poor clopidogrel response have not yet been fully elucidated. In this review, we systematically summarized numerous factors affecting the pharmacodynamics and pharmacokinetics of clopidogrel. Most of these factors (e.g., CYP2C19*2/*3 and DM) may contribute to the low active metabolite of clopidogrel and/or high on-treatment platelet reactivity, which predict increased risk of thrombosis, ischemic events, and cardiovascular death. While other factors (CYP2C19*17 and smoking) may have the opposite effect, increasing clopidogrel active metabolite exposure and improving clopidogrel response. The impact of some factors (e.g., PON1 and CKD) cannot be fully determined on the response to clopidogrel treatment. In short, it is difficult to determine only one factor or only a few factors at a time that result in clopidogrel resistance. Increasing drug dosage or switching to alternative drugs such as ticagrelor or prasugrel is preferred for those patients with impaired antiplatelet effects of clopidogrel.This project was supported by the National Science and Technology Major Project (2013ZX09509107), National Natural Science Foundation of China (No.81373489, No.81422052, and No. 81403018), Hunan Provincial Natural Science Foundation of China (13JJ1010 and 2015JJ3169), Funds of Hunan Provincial Education Department (12K006), and the Fundamental Research Funds for the Central Universities of Central South University (2014zzts073).Yan-Jiao Zhang and Xiao-Ping Chen were responsible for the concept of the review; Yan-Jiao Zhang drafted the manuscript; Mu-Peng Li, Jie Tang, and Xiao-Ping Chen provided a critical review of the manuscript. All authors contributed to and agreed on the final version of the manuscript.The authors declare no conflict of interest.The metabolic pathway of clopidogrel and its targeted receptor. Intestinal absorption of the prodrug clopidogrel is limited by P-glycoprotein (P-gp). After absorption, the clopidogrel (inactive) is oxidized to 2-oxo clopidogrel (still inactive) by CYP450 enzymes. The 2-oxo clopidogrel is then transformed into active metabolites that will bind to P2Y12 receptor on platelet surfaces.Genetic polymorphisms observed to be associated with clopidogrel response.ACS: acute coronary syndrome; MI: myocardial infarction; AMI: acute myocardial infarction; CHD: coronary artery disease; clop-AM: clopidogrel active metabolite; CLP: clopidogrel; NA: not available.

Med-MDPI/ijerph_2/ijerph-14-03-00302.txt

@@ -0,0 +1,6 @@
+Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).Artisanal and small-scale gold mining (ASGM) has been an important source of income for communities in the Madre de Dios River Basin in Peru for hundreds of years. However, in recent decades, the scale of ASGM activities in the region has increased dramatically, and exposures to a variety of occupational and environmental hazards related to ASGM, including mercury, are becoming more widespread. The aims of our study were to: (1) examine patterns in the total hair mercury level of human participants in several communities in the region and compare these results to the 2.2 µg/g total hair mercury level equivalent to the World Health Organization (WHO) Expert Committee of Food Additives (JECFA)’s Provisional Tolerable Weekly Intake (PTWI); and (2), to measure the mercury levels of paco (Piaractus brachypomus) fish raised in local aquaculture ponds, in order to compare these levels to the EPA Fish Tissue Residue Criterion of 0.3 µg Hg/g fish (wet weight). We collected hair samples from 80 participants in four communities (one control and three where ASGM activities occurred) in the region, and collected 111 samples from fish raised in 24 local aquaculture farms. We then analyzed the samples for total mercury. Total mercury levels in hair were statistically significantly higher in the mining communities than in the control community, and increased with increasing geodesic distance from the Madre de Dios headwaters, did not differ by sex, and frequently exceeded the reference level. Regression analyses indicated that higher hair mercury levels were associated with residence in ASGM communities. The analysis of paco fish samples found no samples that exceeded the EPA tissue residue criterion. Collectively, these results align with other recent studies showing that ASGM activities are associated with elevated human mercury exposure. The fish farmed through the relatively new process of aquaculture in ASGM areas appeared to have little potential to contribute to human mercury exposure. More research is needed on human health risks associated with ASGM to discern occupational, residential, and nutritional exposure, especially through tracking temporal changes in mercury levels as fish ponds age, and assessing levels in different farmed fish species. Additionally, research is needed to definitively determine that elevated mercury levels in humans and fish result from the elemental mercury from mining, rather than from a different source, such as the mercury released from soil erosion during deforestation events from mining or other activities.Artisanal and small-scale gold mining (ASGM) is a term broadly used to describe the gold mining by individuals, families, or groups with minimal mechanization, often in the informal or illegal sector of the market [1]. ASGM occurs in over 70 developing countries [2]. It is estimated to employ 13 million people globally, and an additional 80–100 million people are directly reliant upon or impacted by ASGM [3].In the Peruvian Amazon, where gold deposits are typically alluvial in nature, ASGM has been a source of income for local populations for centuries [4]. However, the Peruvian region of Madre de Dios has seen a drastic increase in ASGM, and a commensurate increase in the population, over the last three decades [4]. In 2011, there were estimated to be more than 80,000 ASGM workers in Peru, plus an additional 300,000 workers employed in peripheral services [5]. For example, the population of the region is estimated to have roughly doubled every 20 years from 1940 to 2015 [6]. Madre de Dios produced close to 70% of Peru’s artisanal gold in 2001 [7]. However, in 2011, an estimated 97% of mining concessions in the region were illegal (in [8]). The environmental impacts of these illegal ASGM activities are substantial, and include deforestation [9,10], habitat loss and desertification [11], and air and water pollution from gasoline and oil spills and combustion [12]. The region is also touted as one of the world’s greatest biodiversity hotspots [13], further increasing the impacts of the environmental degradation associated with ASGM.ASGM activities involve the extensive use, and subsequent environmental release, of inorganic elemental mercury, which is used to amalgamate fine alluvial gold particles and increase gold recovery. Elemental mercury is burned off of the gold amalgam before sale, to ensure a pure gold product [14]. Although some larger operations use retorts or fume hood condensers to capture and recycle mercury vapors, it is common for ASGM workers to burn mercury indoors, often in homes, without ventilation, which can result in substantial inhalation exposure [14]. After elemental mercury is vaporized, it can enter into aquatic ecosystems, where it may be biomethylated by bacteria into an organic form, methylmercury [15]. Fish and macroinvertebrates in aquatic ecosystems accumulate methylmercury in their tissues, with increasing concentrations in higher trophic levels. At least 95% of the total mercury present in fish tissue is methylmercury [16]. Elemental mercury can also directly enter aquatic ecosystems, without undergoing vaporization; an estimated 45% of the mercury used in informal and illegal ASGM activities in the Amazon region of Brazil is directly lost as a result of spills or dumping [17]. Human populations in Madre de Dios depend upon fish as a dietary staple, and fish consumption is the main route of exposure to methylmercury [18].Global estimates cite ASGM as contributing approximately 17% of anthropogenic mercury emissions, and it is responsible for further mercury contamination through direct spillage into the environment [2,19]. In 2010, ASGM was the major source of global mercury emissions to air, releasing 727 tons per year [20]. Peruvian mercury imports exponentially increased from 2003 to 2009 [9], and from 2006 to 2009 it was estimated that 95% of this imported mercury was used in ASGM (in [7]). In October 2013, Peru signed the United Nations’ Minamata Convention, an international treaty to reduce anthropogenic mercury emissions [21]. However, Peru’s mercury imports have increased in recent years, largely for use in ASGM [9]. A distinction is made in Peru between informal and illegal mining. Informal mining is used to describe mining operations that take place in legally-designated mining zones, but without a formal government-issued permit. Illegal mining—that occurring outside of legally-designated mining zones—makes up the majority of mining in the Madre de Dios region [22]. The Peruvian government has been unable to control the spread of illegal mining over the last decade. A reported 450 hectares have been deforested within the Tambopata National Reserve as of September 2016 [23]. In May 2016, the government of Peru issued a 60-day public health emergency after studies showed that up to 48,000 people across more than 85,000 hectares are affected by mercury exposure due to mining [24].Mercury exposure is associated with a wide range of adverse human health effects, including neurological, cardiac, motor, reproductive, genetic, renal, and immunological conditions [25,26,27,28,29,30,31]. Exposure to elemental mercury vapors can result in tremors, a decrease in memory performance, and a decrease in autonomic nervous system functioning [32]. Exposure to methylmercury can cause damage to the central nervous system [25,33]. Developing fetuses and young children are at a particularly high risk of adverse neurological effects from methylmercury [34], and maternal methylmercury can be transferred to a fetus in the womb [35].Human mercury exposure may be assessed through assays of urine (useful for evaluating elemental mercury exposure) and blood (useful for measuring methylmercury exposures from dietary sources and elemental exposures) [36,37]. The total mercury in human hair is a good indicator of long-term mercury exposure, particularly for methylmercury [38,39]. The concentration of hair methylmercury is proportional to the blood concentration at the time of the formation of the hair strand [40], with the concentration measured in hair being approximately 250 times that of the corresponding blood level. The Joint Food and Agriculture Organization of the United Nations (FAO) and the World Health Organization’s (WHO) Expert Committee of Food Additives (JECFA) set a Provisional Tolerable Weekly Intake (PTWI) of 1.6 µg methylmercury per kg of body weight. The PTWI was set at a level considered sufficient to protect developing fetuses, the subgroup most vulnerable to the effects of mercury exposure [41]. The PTWI is associated with a total hair mercury concentration of approximately 2.2 µg mercury/g dry hair; we used this level as a comparison for our adult populations [42].Several previous studies have evaluated human exposure to mercury in Madre de Dios. While ASGM comprises a significant source of mercury in the environment in this region, there are also non-anthropogenic sources and geochemical effects to consider. For example, hair mercury levels have been shown to be positively correlated with river pH and dissolved organic carbon in the Amazon basin [43]. Elevated mercury levels in the Madeira River (into which the Madre de Dios River flows, via the Beni River) are largely due to natural sources and biogeochemical processes [44]. However, for the purposes of this study we will review the literature that has investigated mercury in the region as related to ASGM activities.One unpublished study demonstrated that the average hair mercury levels exceeded the U.S. EPA Reference Concentration for human hair of 1 µg/g [45], and a second study showed that hair mercury levels were higher in ASGM areas than in non-mining areas [46]. High levels of mercury have also been found in the blood and urine of ASGM miners in Madre de Dios [47]. Additionally, hair mercury levels have been associated with a higher fish consumption [46,47]. The U.S. EPA set a methylmercury Tissue Residue Criterion for fish tissue intended for human consumption at 0.3 µg MeHg/g fish tissue (wet weight) [48]. The EPA estimates that about 95% of the total mercury present in fish tissue is methylmercury, yielding a reference dose for total mercury in fish tissue of 0.333–0.4 µg Hg/g fish tissue (wet weight) [48]. Studies of mercury and methylmercury in fish from rivers in Madre de Dios have shown ambiguous results, with some indicating levels below the EPA criterion [49], and some indicating elevated exposures [45,50,51], depending on the fish species sampled and the location of capture. A relatively recent development in Madre de Dios is the introduction of pond aquaculture practices [52] to improve natural fluvial stocks [53] and provide income alternatives to ASGM [54]. One of the most commonly farmed and consumed fish species in Madre de Dios is a native species known locally as “paco” (Piaractus brachypomus), favored for its performance in fish farms [55]. Mercury biomagnifies in aquatic food webs, as methyl mercury is retained by lower trophic levels and passed on to higher trophic levels after predation or consumption [56]. Paco is an omnivorous fish and, assuming that the fish raised in ponds on pelleted diets mimic omnivorous diets, we did not expect to find high mercury concentrations under natural environmental conditions.Our study examined the impacts of ASGM in the Madre de Dios river basin of Peru, and appears to be the first to examine the relationship of the distance of various communities from the river headwaters, as well as to assess male and female exposure, without a particular occupational focus. The first aim of our study was to identify the risk factors for elevated total hair mercury levels of human participants in several communities along the length of the Madre de Dios River or located on tributaries, and to compare the observed levels of mercury in hair to the total hair mercury concentration of 2.2 µg/g, corresponding to the PTWI. We hypothesized that hair mercury levels would increase with increasing geodesic distance from the headwaters of the river. We used the geodesic distance, even for communities located on tributaries, to see if there was a trend of mercury concentration in human populations as one moved farther away from the headwaters, where little or no mercury contamination from ASGM should be present. Our second aim was to measure the mercury levels of the paco raised in local fish ponds in Madre de Dios, and to compare these levels to the EPA Fish Tissue Residue Criterion of 0.3 µg Hg/g fish (wet weight), which is the concentration of methylmercury in fish tissue that can be consumed without the expectation of adverse health outcomes (based on the assumption of 0.0175 kg fish/day) [48]. We hypothesized that the fish in the pisciculture ponds would have a mercury concentration above the tissue residue criterion.Data were collected from May to July 2014 in four Peruvian communities along the 1060 km long Madre de Dios River, which originates in Peru [57] and empties into the Beni River in Bolivia (Figure 1). All research protocols involving human subjects were approved by the University of Michigan IRB (HUM00086592). All subjects gave their informed consent for inclusion before participating in the study.Field sites for human subjects were chosen to represent a range of distances from the headwaters of the Madre de Dios River (Table 1). The four sites were Bajo Madre de Dios, Boca Amigo, Mazuco, and Pilcopata. Pilcopata, the community closest to the headwaters, was chosen as a control site, due to the lack of nearby ASGM activities; as determined through an evaluation of satellite images from 2005 to 2014. Landsat data are distributed by the Land Processes Distributed Active Archive Center (LP DAAC), located at USGS/EROS, Sioux Falls, SD, USA. The presence of illegal mining in the other three communities was determined through an evaluation of satellite images and through conversations with community members. Formal demographic and population data were not available for all of the communities at the time of the study. The communities were generally similar, with the exception of proximities to mining and the degree of urbanization; Boca Amigo and Bajo Madre de Dios were not urban. Mazuco was the largest community sampled.In the small communities of Pilcopata, Bajo Madre de Dios, and Boca Amigo, households near the main town center were visited for recruitment via convenience sampling. In the larger town of Mazuco, a section of the community was selected based on convenient access for researchers and all houses in that section were approached for recruitment. At all four sites, one (and, in rare cases, two) adult participated in each sampled household, with a range of 12 to 19 households visited per site. Our a priori sampling goal was 20 subjects from 20 households, in each of the four communities.The survey completed by all subjects addressed demographic factors (age, sex, study site, education, duration of residence, household size, pregnancy, etc.); the frequency (number of meals per week including fish) and source of different types of fish consumed; the number of servings of fish consumed in the three days prior to the survey; and the frequency and source of other types of protein consumed. Note that, due to the often-illegal nature of mining in the area, we chose not to ask participants about their mining activities. The study inclusion criteria were an age of at least 18 years and residence in the study site for at least six months.After completing the survey, a collection of roughly 200 strands of hair (the approximate number needed to obtain the necessary mass of hair for subsequent analysis) were cut from the occipital region of the skull of each participant. Samples were taken with stainless steel scissors as close to the scalp as was safely possible. The hair was stored between sheets of adhesive paper, and the sample end closest to the scalp was marked. The samples were labeled, stored individually in double plastic Ziploc bags, and frozen after transport to the University of Michigan and prior to analysis.Human hair samples were analyzed at the University of Michigan using a Milestone Direct Mercury Analyzer (DMA-80, Milestone Inc., Shelton, CT, USA), using EPA method 7473 [59]. Samples were trimmed so that the 4 cm length of hair closest to the scalp would be analyzed. Hair grows at an average rate of one cm per month [60], so our analysis was intended to estimate the total mercury exposure for the four-month period preceding analysis. A 50–55 mg hair sample was measured and placed in the DMA for analysis. Three different readings were taken for each sample, and the three runs were averaged to estimate the total mercury concentration per sample. Quality control (QC) measures included the random testing of reference materials (IAEA-086), blank measurements every 10th reading, and random checks of previously measured samples. Recovery rates of 95%–100% were considered acceptable for reference material tests. The Limit of Detection (LOD) was 0.003 ng Hg.Fish samples were collected from aquaculture farms throughout the Madre de Dios region (Table 2). Farms were selected based on convenience and awareness of their existence by local governmental and non-governmental agencies. We purchased recently-caught and killed fish that were to be consumed by the farm owner, directly from fish farmers. Fish from farms were targeted because heavy rainfall during the study period prevented fisherman from obtaining fish from the rivers. The selection of fish farm locations was not consistent with the sites where the human participants were sampled, due to constraints in access to refrigeration. The Iberia site, farthest from the Madre de Dios River and near the Brazilian border, served as a reference site, because it was believed to be similar to other study sites, but not subject to ASGM mercury. At the time of the study, no large-scale ASGM activities were believed to occur in Iberia. This study only focused on the fish species “paco”, as it is easy to find in fish farms throughout the region, allowing for a spatial comparison within one fish species.Five paco specimens were purchased from pisciculture farmers on the same day as the fish had been caught and killed for their own consumption. All five samples were collected from a single pond from each fish farm, using standard fishing practices. Fish are purchased from NGOs and governmental agencies as fingerlings by the farmer. All fish are added to ponds at the same time, and are harvested in nets one year later. Therefore, farmers were able to tell us exactly how long fish had been in their respective ponds. The information gathered for each fish sample included the number of months in the pond, length (cm), approximate weight (kg), and GPS coordinates of the pond. A tissue sample weighing approximately 50 g was taken from the side of the deceased fish, over the lateral line just anterior to the tail. The sample was placed in a Ziploc bag and labeled; the remainder of the fish was either purchased or returned to the pond owners. All of the fish samples were frozen locally and then shipped to the University of Michigan on dry ice.At the University of Michigan, fish samples were measured on a balance (to ±0.0001 g) while still frozen, then placed into individual Whirl Pak bags, labeled, and dehydrated for at least 48 h using a vacuum freeze dryer dehydrator. The final dry weights (to ±0.0001 g) were recorded after dehydration; these weights were, on average, 22%–25% of the wet weight. The dried samples were then pulverized to a fine powder and sampled using a disposable spatula. Samples were analyzed using the DMA-80 and EPA method 7473 [59]. As with the human hair samples, three different readings were taken and averaged for each fish sample. QC was similar to that used for human hair; for reference materials (IAEA-407, IAEA-436), recoveries between 90% and 110% were considered acceptable. The LOD was again 0.003 ng Hg.The mercury concentration in fish tissue was converted from the dry weight concentration (i.e., the DMA output) to the wet weight concentration using Equation (1).The conversion of dry weight [Hg] to wet weight [Hg] was as follows:
+        
+		Wet weight [Hg] = (Dry weight [Hg] × (1 − Δweight))/100,
+		(1)
+		
+        where Δweight is the change in weight for each sample between dry and wet weight measurements.Statistical analysis was performed using RStudio (RStudio, Boston, MA, USA) and Stata 14.1 (StataCorp LLC, College Station, TX, USA). Descriptive statistics were computed for all variables, and univariate and bivariate relationships were examined visually, through the use of histograms, scatterplots, and quantile plots, in combination with correlation coefficients. Non-normally-distributed variables were log-transformed prior to parametric statistical analyses; for these variables, the geometric mean and geometric standard deviation (GSD) are reported, in addition to the arithmetic mean and standard deviation. For inferential statistical tests, results were considered significant when p < 0.05.To analyze the measured total mercury in human hair, ANOVA tests were used to test differences in mercury levels by categorical variables, using both the original response scales (two to four possible response categories, depending on the survey item) and the results after collapsing them into binary categories. ANOVA tests were also run on categorical variables and mercury levels after log-transforming the mercury levels. Chi-squared tests were used to test differences in the fraction of hair mercury measurements exceeding the 2.2 μg/g PTWI total hair mercury equivalent level by site and sex. Linear regression analyses were used to evaluate the bivariate association between the measured human hair total log-transformed mercury levels and each of the independent variables (i.e., demographic factors, community site, fish consumption, etc.), and logistic regression analyses were used to evaluate the odds ratio (OR) for total hair mercury levels >2.2 µg/g associated with each of the individual independent variables. A forward stepwise regression approach was used to develop multivariate linear and logistic regression models; variables were retained where p-values of 0.10 or less were observed. To account for the few cases where multiple adults were sampled in a single household, regression models accounted for the intragroup correlation of data by household (Stata “cluster” option).For the fish samples, the total mercury (Hg mg/kg dry weight) was plotted against the fish specimen weight mercury, and the R-squared value and slope of the relationship were assessed. Descriptive statistics were computed for the mean wet weight mercury concentration (mg/kg), both overall and by study site. The fraction of samples exceeding the EPA tissue residue criterion of 0.3 µg MeHg/g fish tissue was computed, both overall and by sampling site.Study participation rates were not formally assessed. The most common reason given for non-participation was a lack of interest in providing a hair sample, either due to aesthetic considerations or cultural beliefs. It is possible that some individuals may have declined to participate because they were engaging in mining themselves; however, this was never stated as a reason to the researchers. Table 3 shows information about the 81 participants, both overall and by site. Of the 81 adult participants, 12 (14.8%) were sampled from six households; all other participants were the only adults sampled in their households. Just under half of the participants (39%, or 48.2%) were male. Participants were roughly equally distributed across the four sites. The mean age, mean residence time, and total household size differed among the four sites, but were not statistically significant. The number of children per household significantly differed among sites, with Pilcopata having the highest value, and Boca Amigo the lowest value. Education levels differed by site: Pilcopata had by far the largest fraction of participants that had completed more than secondary school (data not shown). Mazuco and Pilcopata had the largest fraction of participants that had completed secondary school, and Bajo Madre de Dios had by far the largest fraction of participants that had not completed primary school.The frequency of fish consumption was correlated with the number of servings of fish in the three days prior to the survey (Spearman correlation coefficient, 0.41, p = 0.002), suggesting that the number of meals containing fish in the prior three days was a useful surrogate measure of fish consumption. Fish and chicken consumption were not correlated (spearman correlation coefficient, −0.12, p = 0.89). Mean servings of fish in the past three days did significantly differ among sites, with Boca Amigo having the highest mean value and Pilcopata the lowest. Finally, the consumption of different fish species (reported in free-text responses to the questionnaire) indicated that the fish species paco and bagre did not differ significantly among sites, but that the consumption of boca chico, sabalo, doncella, and zungaro did. Participants in Pilcopata were the least likely to consume any of these four types of fish; participants in Mazuco were the most likely to consume boca chico and sabalo, while participants in Boca Amigo were the most likely to consume doncella and zungaro.Figure 2 shows the distribution of the reported frequency of fish consumption by study site. Boca Amigo and Pilcopata had the greatest fraction of participants reporting an infrequent (i.e., two times per month or less) consumption of fish, while Bajo Madre de Dios and Pilcopata had the greatest fraction of participants reporting very frequent consumption (i.e., greater than or equal to seven times per week).Table 4 shows the descriptive statistics for the total mercury levels in human hair. All of the collected hair samples had a sufficient mass for laboratory analysis. The control site, Pilcopata, had the lowest arithmetic and geometric mean and standard deviation, and there was only a single individual at Pilcopata that had mercury readings above the PTWI total hair mercury equivalent level of 2.2 µg/g. The highest arithmetic and geometric mean total mercury levels were found at the downriver mining sites Boca Amigo and Bajo Madre de Dios; these two sites also had the highest measured values of any of the four sites. At both Boca Amigo and Bajo Madre de Dios, two-thirds or more of the hair samples exceeded the reference level. ANOVA results indicated that the arithmetic and geometric mean levels of total hair mercury differed among the four sites. No significant differences in arithmetic or geometric mean levels of total hair mercury were noted when considering sex, either overall or within any of the four sites. The fraction of samples exceeding the reference level significantly differed among sites. No significant differences in the fraction of samples exceeding the EPA limit for total hair mercury were noted between sexes, either overall or within a site.Unadjusted linear regression models using a single dependent variable regressed on the log-transformed total hair mercury level, yielded p-values > 0.05 for nearly all variables. However, the coefficients for three of the four sites (reference site: Pilcopata) yielded p-values < 0.0001. Each of these site indicator variables showed positive coefficients for log-transformed total hair mercury levels, when compared to the reference site. These variables were selected for inclusion in an adjusted linear regression model.A similar approach was used for logistic regression, with the dependent variable being the total mercury hair level in excess of the 2.2 µg/g PTWI total hair mercury equivalent level. The majority of independent variables assessed did not reach statistical significance. However, as with the linear regression models, the coefficients for three of the four sites (reference site: Pilcopata) reached statistical significance. These variables were selected for inclusion in an adjusted logistic regression model.Table 5 shows the results of the multivariable adjusted linear and logistic regression models. Linear regression model one included the only variable (site) that reached statistical significance. The two farthest downriver ASGM communities (Bajo Madre de Dios and Boca Amigo) both had substantially larger coefficients than the upriver community (Mazuco), and all three of these communities had elevated levels of mercury compared to the reference community (Pilcopata). Adjustment for other factors, including age, sex, education level, frequency and type of fish consumption, and frequency and type of other protein consumption, only resulted in marginal improvements to the model. Model two is adjusted for age and sex, and had an essentially identical model fit (0.51 R2).Logistic regression model three shows the only variable (site) that reached significance in the logistic regression analysis. Consistent with the results of linear regression models one and two, logistic regression model three showed that the two farthest downriver ASGM communities (Bajo Madre de Dios and Boca Amigo) both had substantially larger ORs than the upriver community (Mazuco), and all three of these communities had elevated levels of mercury compared to the reference community (Pilcopata). Participants in Boca Amigo were 5.9 times more likely to have mercury levels in excess of the 2.2 µg/g reference level than those in Pilcopata. As with the linear regression models, adjustment for other factors (e.g., age, sex, education level, frequency and type of fish consumption, and frequency and type of other protein consumption) resulted in negligible improvements to the model. Logistic regression model four, which adjusted for age and sex, achieved a virtually identical pseudo-R2 (0.40).The weight of the fish specimen was positively correlated with mercury (mg Hg/kg dry fish tissue) (Figure 3). The R2 value of the relationship between fish specimen weight in kg and total mercury concentration in dry fish tissue, was 0.932. The fish sampled were between approximately four and eight months old at the time of sacrifice (i.e., fingerling transplant to aquaculture pond plus four to eight months of growth).The arithmetic and geometric mean total mercury concentration for all 111 samples across the four study sites are shown in Table 6. All samples were below the EPA Tissue Residue Criterion of 0.3 mg/kg. Of all the samples, only two fish from Virgen de la Candelaria had mercury concentrations near this value (0.23 and 0.22 mg Hg/kg wet weight fish tissue). As shown in Table 6, Mazuco had the lowest mean mercury concentration for both wet weight and dry weight (0.028 mg/kg and 0.106 mg/kg, respectively). The site with the highest mercury concentration was Virgen de la Candelaria (Ww = 0.12 mg/kg Hg and Dw = 0.40 mg/kg Hg). The control site of Ibería had a mean wet weight of 0.04, making it the third lowest concentration. Figure S1 in the supplementary materials provides a visual representation of these data.Finally, when mercury levels were modeled by approximate age of fish, the average mercury levels were seen to increase by approximately 0.01 mg/kg per year (data not shown). Fish of approximately four months of age had mean levels of 0.04 mg/kg wet weight, increasing to 0.06 mg/kg at six months of age, and 0.08 mg/kg at eight months of age. Thus, it appears that, as the fish increased in size and had increasing residence time in aquaculture ponds, their mercury levels increased in a roughly linear manner. If this linear trend were to continue through one year of age (the typical age at harvest), the total mercury content would be approximately 0.12 mg/kg, which is still well below the EPA tissue residue criterion of 0.3 mg/kg.The results of this study suggest that people living in communities in the Madre de Dios region of Peru where ASGM activities occur are subjected to higher mercury exposure than those living in a non-mining community. Residence in a ASGM community was associated with highly increased odds of total hair mercury levels exceeding the reference level of 2.2 µg Hg/g hair (ORs of 5.8, 3.8, and 2.2 for participants in Bajo Madre de Dios, Boca Amigo, and Mazuco, respectively). These ORs suggest a dose-response relationship, with the communities furthest from the headwaters of the Madre de Dios River (Bajo Madre de Dios and Boca Amigo) having the highest odds of exceeding the 2.2 µg/g PTWI total hair mercury equivalent level. Measurements of paco purchased from aquaculture farms in the region did not identify any samples with levels of mercury that exceeded the EPA tissue residue criterion of 0.3 μg Hg/kg wet weight. The site used as a control, Iberia, did not have the lowest levels of mercury in sampled fish, but levels from this site were among the lowest of all sites evaluated. Collectively, these results suggest that ASGM activities are associated with higher human mercury exposure, and that farmed fish contribute little to mercury exposure.Our human hair analysis results are generally consistent with several prior studies that have evaluated human exposure to mercury in Peru. Among 226 adults from the capital city of Madre de Dios, Puerto Maldonado, the average hair mercury concentration was two to three times higher than the reference level of 2.2 µg Hg/g hair. [45]. A 2012 study on the mercury levels in human hair among 104 participants in the town of Puerto Maldonado in Madre de Dios and 100 participants in an ASGM mining zone in the region, found that residence location and sex were correlated with higher mercury levels, and that the total levels of mercury in hair were significantly higher in the mining zones [46]. A study of 103 ASGM miners in Madre de Dios in 2010 found that all participants had detectable levels of mercury in urine, and 91% had detectable levels of blood methylmercury [47]. As with the study by Ashe (2012) [46], Yard et al. (2012) [47] found that higher fish consumption was associated with increased hair mercury levels; Yard et al. further found that exposure to heated gold-mercury amalgam was correlated with higher hair mercury levels.Our results do not match those of similar studies in concluding that the amount of fish consumed correlates with increased mercury levels. It is possible that, due to the timing of this study, fish was not currently in season and had not been for several months. Therefore, the level of total mercury may be higher in the study populations during the dry season, when fishing is more lucrative and diets depend more heavily upon river fish as a staple food. Additionally, the metric for fish consumption used in this study relied on self-reported dietary behaviors, and may be susceptible to recall or other reporting biases. Other studies have also shown a correlation between sex and total hair mercury levels [45,46]; we did not identify such a relationship here.The fish sample results were consistent with two of four previous studies of mercury concentrations in fish in the Madre de Dios region. A 1997 study measured the level of mercury and methylmercury in samples from seven fish families in the Manu River in Madre de Dios. The total mercury levels in two samples of paco were 0.053 and 0.067 µg Hg/gram wet weight [49]—i.e., below the EPA tissue residue criterion, and similar to those found here. The total mercury concentrations in paco purchased from markets in the capital of Madre de Dios, Puerto Maldonado, were below the EPA criteria for methylmercury (mean 0.24 mg/kg) [45], though it is important to note that fish sold at this market may not have been locally caught or raised. Two other studies of species of fish which are common in the Madre de Dios river found average total and methylmercury concentrations above the EPA tissue residue criterion [45,51]. Future research needs to include an investigation on different fish species, as well as different fish sources (farmed versus river caught, etc.). In addition, future research should consider the characteristics of individual ponds, such as pond age, source of water used, distance of pond from mining activities, fish fry source, etc., to eliminate possible confounding.Our study has several important limitations. First, the illegal and informal nature of mining and political circumstances may have influenced participation rates among households in the four research sites. If so, this would have negatively biased the total hair mercury results presented here, as households actively engaged in mining may have been less likely to participate. Our methods of recruitment meant that we were not able to compute quantitative participation rates. Second, our study utilized convenience sampling instead of random sampling. This limits the generalizability of our findings. Third, the fish samples analyzed were well below the one-year age at which aquaculture fish are typically harvested for market sale, which may have negatively biased their measured mercury levels. Fourth, there was a potential temporal mismatch between our survey items related to fish and other protein consumption, which were intended to assess consumption in a typical week, and the measurements of hair mercury, which reflected mercury exposure over the several months prior to sample collection. Finally, our regression model development approach used a conventional statistical approach (i.e., binary cutoff p-values for stepwise model building). This approach may have resulted in some potentially important risk factors being discarded during our model building efforts [61,62], but was considered appropriate given our small sample size and the exploratory nature of our analyses. Nevertheless, our finding that the total hair mercury levels were statistically significantly higher in communities where ASGM activities occur, supports the idea that anthropogenic activities increase the risk of exposure to mercury and suggests that further study incorporating a consideration of occupational, residential, and changing nutritional exposures is needed.Low levels of total hair mercury in the control study population near the headwaters, where ASGM mining was not occurring during the study period, suggest that higher total hair mercury levels observed in communities were associated with the presence of ASGM mining. We did not observe the levels of total hair mercury to be related to levels of mercury in the local fish. Our inability to assess the ASGM status among individual subjects limits our ability to evaluate the impact of ASGM activities on individual doses of total hair mercury, and may have resulted in an underestimation of mercury exposure in the subjects of the three ASGM communities assessed. The frequency of the consumption of fish and the number of fish consumed did not prove to be predictive of observed levels of total mercury in hair. More studies are needed to determine what species of fish, and what sources of fish, are lowest in mercury concentration for human consumption.In all 111 observations of farmed fish in the region of Madre de Dios, there were no levels of mercury above the EPA Tissue Residue Criterion, though it is important to note that the fish sampled were only, on average, about one-half the age of harvest in the region of Madre de Dios, and so our estimates of mercury levels in fish at the time of harvest for sale or consumption are likely to be low. Also, the fish ponds sampled as part of our study were relatively new (a few years old on average), and it is likely that, as the ponds age, methylmercury will continue to accumulate through transformation after atmospheric deposition in the stagnated pond water. Finally, paco are omnivorous fish, and so it is expected that they would have a lower mercury concentration under natural conditions than other carnivorous fish.Our study has shown that there may be an increased risk of mercury exposure in some populations in the region of Madre de Dios; this risk appears to be dependent upon location. Further research on historical mercury levels in sediments throughout the river basin, localized biotic and abiotic factors affecting environmental chemical processes, and the analysis of similar populations along the main channel of the Madre de Dios River, as well as its tributaries, would help to definitively quantify the impact that anthropogenic releases of mercury from ASGM has on the human population and ecosystem.The following are available online at www.mdpi.com/1660-4601/14/3/302/s1, Figure S1: Comparison of mercury in paco fish tissue by study site: mean wet weight mercury concentration (mg/kg), mean dry weight mercury concentration (mg/kg).Funding for this research was provided in part by the University of Michigan International Institute’s Tinker Field Research Grant, the University of Michigan Rackham Graduate School, and the University of Michigan School of Natural Resources and the Environment (SNRE). The authors wish to thank Joel Blum and Allen Burton for their contributions to this manuscript, to Paul Drevnick and Marcus Johnson for assistance in conducting laboratory analyses, and to Yessenia Apaza and the staff at La Asociación para la Conservación de la Cuenca Amazónica (ACCA), as well as the Ministry of Health in Mazuco, for their assistance in data collection. Raúl Loayza-Muro, head of the Laboratory of Ecotoxicology—LID, Faculty of Sciences and Philosophy, Universidad Peruana Cayetano Heredia in Lima, Peru coordinated fish tissue sample storage and shipping to the U.S. and offered project support. Aubrey L. Langeland would personally like to thank her cohort at SNRE for their support and collaborations. Finally, the authors wish to thank the participating subjects, without whom the research described here would not have been possible. Landsat images were used to identify study sites. These data are distributed by the Land Processes Distributed Active Archive Center (LP DAAC), located at USGS/EROS, Sioux Falls, SD, USA.Aubrey L. Langeland and Rebecca D. Hardin conceived and designed the experiments; Aubrey L. Langeland performed the experiments; Aubrey L. Langeland and Richard L. Neitzel analyzed the data; Rebecca D. Hardin contributed reagents/materials/analysis tools; Richard L. Neitzel and Aubrey L. Langeland wrote the paper.The authors declare no conflict of interest.Map of the study site locations in the Madre de Dios river basin of Southeastern Peru. Map created with ESRI ARCGIS software, with layers downloaded from Atrim Biodiversity Information Systems [58].Percent of participants reporting categories of fish consumption frequency at four study sites in the Madre de Dios river basin of Southeastern Peru.Scatterplot showing the association between total mercury concentration (dry weight) and total fish weight for all fish tissue samples, at 24 sampling sites in the Madre de Dios river basin of Southeastern Peru.Geographic and mining-related characteristics of four human study sites in the Madre de Dios river basin of Southeastern Peru.1 Calculated in ArcGIS10 software using a Haversine formula to calculate the great-circle distance between the study site coordinates and those of the Madre de Dios river origin (ignores topography and river bends; “As the crow flies”). Although these numbers are not the actual distance downriver, they did correctly rank-order the villages (i.e., closest to furthest from the headwaters), which is consistent with the variable created in our analysis. The community of Mazuco is located on the Inambari River, a tributary to the Madre de Dios River. Similarly, the community of Pilcopata is located on the Kosñipata River, a tributary to the Upper Madre de Dios River, which flows into the main Madre de Dios River. Thus, the purpose of the “distance to headwaters” measurement is not intended to be a measurement of distance along the Madre de Dios River for a community, but rather to highlight the distance from the point of origin of the Madre de Dios River.Characteristics of 24 study sites for Piaractus brachypomus (Paco) sample collection in the Madre de Dios river basin of Southeastern Peru.Demographic characteristics of 81 human participants at four study sites in the Madre de Dios river basin of Southeastern Peru.1 Significant difference among sites, ANOVA, p-value < 0.05; 2 Significant difference among sites, ANOVA, p-value < 0.01; 3 Significant difference among sites, χ2, p-value < 0.01.Descriptive statistics for the mercury (µg Hg/g hair) in human hair samples by site and sex, at four study sites in the Madre de Dios river basin of Southeastern Peru.1 Boca Amigo had one observation with an average total mercury content of 30.12 µg Hg/g hair. This outlier was not included in any of the above statistical calculations except for the toxic level count; 2 Significant difference in geometric mean total hair mercury levels between sites, ANOVA, p < 0.01; 3 Significant difference in fraction of samples ≥reference level between sites, χ2, p < 0.01.Results of multivariable linear and logistic regression models evaluating differences in total hair mercury at four study sites in the Madre de Dios river basin of Southeastern Peru.Descriptive statistics for measured mercury concentration in paco fish tissue samples at 24 sites in the Madre de Dios river basin of Southeastern Peru.1 One outlier was removed from the samples collected from this study site; 2 Significant difference in geometric mean levels by site, ANOVA, p < 0.001.

Med-MDPI/ijerph_2/ijerph-14-03-00303.txt

@@ -0,0 +1 @@
+Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).A multi-omics approach was applied to an urban river system (the Brisbane River (BR), Queensland, Australia) in order to investigate surface water quality and characterize the bacterial population with respect to water contaminants. To do this, bacterial metagenomic amplicon-sequencing using Illumina next-generation sequencing (NGS) of the V5–V6 hypervariable regions of the 16S rRNA gene and untargeted community metabolomics using gas chromatography coupled with mass spectrometry (GC-MS) were utilized. The multi-omics data, in combination with fecal indicator bacteria (FIB) counts, trace metal concentrations (by inductively coupled plasma mass spectrometry (ICP-MS)) and in-situ water quality measurements collected from various locations along the BR were then used to assess the health of the river ecosystem. Sites sampled represented the transition from less affected (upstream) to polluted (downstream) environments along the BR. Chemometric analysis of the combined datasets indicated a clear separation between the sampled environments. Burkholderiales and Cyanobacteria were common key factors for differentiation of pristine waters. Increased sugar alcohol and short-chain fatty acid production was observed by Actinomycetales and Rhodospirillaceae that are known to form biofilms in urban polluted and brackish waters. Results from this study indicate that a multi-omics approach enables a deep understanding of the health of an aquatic ecosystem, providing insight into the bacterial diversity present and the metabolic output of the population when exposed to environmental contaminants.Understanding the complex interaction between microbial communities and environmental changes, natural or anthropogenic, is a major research challenge [1]. Free-living microorganisms, such as planktonic bacteria in a water column or the diverse microbial community found in soils, would be predicted to display large inter-individual metabolic variability as a result of their individual state, interaction with their surrounding environment and exposure to pollution sources. However, through a systems biology approach, harnessing the depth of knowledge provided by each microorganism can be aggregated and investigated in greater detail, and with increased throughput [2,3]. For example, microbial communities are known to colonize most environments, including contaminated sites or harsh climatic environments [4,5,6]. These communities have been demonstrated to metabolize recalcitrant contaminants (via bioremediation) and have been observed to thrive under extreme conditions [7]. The recent development of omics-based techniques has enabled environmental microbiologists to recognize and characterize such microbial communities as an imperative ecological parameter in monitoring polluted and extreme environments, either by detecting community shifts in response to contaminant(s) or their resilience towards climatic or physico-chemical disturbances [7]. In this context, the application of metagenomics and meta-transcriptomics has been vital in detailing the microbial community present and its potential activity, and techniques such as meta-proteomics and metabolomics have provided insight into the specific proteins and metabolites that are expressed [2].Desai et al. [7], amongst others, proposed that the simultaneous analyses of multiple omics-based approaches would lead to a system-wide assessment of site-specific microorganisms and their underlying physico-chemical disturbances; they focused on characterizing the protein synthesis and carbon and carbohydrate metabolism of the sampled population at xenobiotic/anthropogen contaminated sites. To an extent, many researchers have already started employing multi omics-based techniques to their research. Bullock et al. [8] investigated the microbial activity in relation to organic matter degradation and turnover at various locations in the Mid-Atlantic Bight. Date et al. [9] used metagenomics and 13C labelled metabolomics to monitor the metabolic dynamics of fecal microbiota. Hook et al. [10] investigated contaminated sediments using transcriptomics and metabolomics in order to better understand the modes of toxic action within contaminated ecosystems. Specifically, the function of transcripts with altered abundance of Melita plumulosa (an epibenthic amphipod) was investigated following whole-sediment exposure to a series of common environmental contaminants. Such contaminants included pore-water ammonia, bifenthrin and fipronil (pesticides), diesel and crude oil (petroleum products), and metals (Cu, Ni, and Zn). Subsequent data integration and hierarchical cluster analysis demonstrated grouped transcriptome and metabolome expression profiles that correlated with each specific contaminant class. Many of the transcriptional changes observed were consistent with patterns previously described in other crustaceans [11]. Likewise, Hultman et al. [5] undertook a similar study investigating the microbial metabolism of permafrost. They used several omics approaches, combined with post-data analysis, to determine the phylogenetic composition of microbial communities of intact permafrost, the seasonally thawed active layer and thermokarst bog (surfaces of marshy hollows). The multi-omics strategy revealed good correlation of process rates for methanogenesis (the dominant process), in addition to providing insights into novel survival strategies for potentially active microbes in permafrost [5].The inclusion of metabolomics in (meta)transciptomics and metagenomics investigations has enabled researchers to assess biochemical profile variations of entire microbial communities living in contaminated sites [6,12]. Metabolomics is a well-established scientific field that focuses on the study of low molecular weight metabolites (typically <1000 Da) within a cell, tissue or bio-fluid [13,14,15]. Furthermore, the application of environmental metabolomics is an expanding field within the metabolomics platform. Environmental metabolomics assesses and characterizes the interactions of living organisms within their environment [4] and is traditionally used as a tool to investigate environmental factors, either physical or chemical, and their impact to a specific organism. For example, Gómez-Canela et al. [16] used targeted environmental metabolomics to investigate Gammarus pulex (a freshwater amphipod crustacean) following controlled exposures to selected pharmaceuticals in water. Similarly, Cao et al. [17] studied the bioaccumulation and metabolomics responses in Crassostrea hongkongensis (an oyster) impacted by different levels of metal pollution; and Ji et al. [18] studied the impact of metal pollution on Crangon affinis (a shrimp). In addition, community metabolomics extends the application of environmental metabolomics even further through the investigation of all metabolites expressed from an entire microbial community, thus enabling a meta-metabolomics approach [6].The advancement of omics-based techniques and their integration (coined multi-omics) have contributed towards the fields of environmental and molecular biology, thereby pushing the boundaries of our understanding of microbial physiology [19]. To date, such studies have focused on specific pollution events (e.g., the Deepwater Horizon oil spill [20]), the assessment of biotechnology/bioremdiation (e.g., bioremdiation of steriods in the enviornment [21]) or used to characterize well-controlled engineered systems (e.g., anerobic bioreactors [22,23]). To the best of our knowledge, such an approach has not been used to characterize a system as part of a water quality monitoring survey. The application of metagenomics or metabolomics in isolation has been applied with some success [24]. For example, metagenomics has been applied to assess drinking water microbial populations after various treatment methods [25] and assess river microbiomes across various land use types [26]. Beale et al. [27] used metabolomics with physico-chemical data to assess water pipeline infrastructure and water pipe biofilms, characterizing biofilms based on pipe material and the excreted metabolites that pass from the biofilm into the water stream. A similar study was used to investigate impacts of exposure to chemicals of emerging concern relative to other stressors in fathead minnows, which was used as a model species [28].The current study herein merges bacterial metagenomics and community metabolomics with additional phyico-chemico data, thus using a multi-omics based approach to investigate an urban river system. It is anticipated that such an approach would provide an additional layer of information on top of traditional water quality monitoring parameters that will ultimately result in a deeper understanding of the the diverse microbial population present, enabling researchers to characterize environmental systems, not based on inferred water quality data but as an interconnected complex system. Furthermore, it is anticipated that a multi-omics approach will enable a better appreciation of the system’s resilience to urban physical and/or chemical changes and stress.Water samples were collected from five sites along the Brisbane River (BR), Queensland (Qld), Australia during the low outgoing tide in December 2013. For reference, the BR sampling sites are presented in Figure 1, are annotated with stars and designated BR1 to BR5. Triplicate samples were collected from each site at one sampling event, giving a total of 15 water samples. The in-situ measurements of temperature (°C), conductivity (mS/cm), pH, salinity (ppt), turbidity (NTU), and dissolved oxygen (mg·L−1) were made at the time of collection using a calibrated AQUAprobe AP-300 water quality probe (Aquaread, Broadstairs, Kent, UK). Moreover, each site was matched, by location and date, with sample sites from the Healthy Waterways “Ecosystem Health Monitoring Program (EHMP)” water quality monitoring program [29]. Of note, Healthy Waterways is a not-for-profit, non-government, membership-based organization working to protect and improve waterway health in South East Qld. Data from the EHMP matched sites were included in the analysis in order to expand the breadth of physico-chemical data collected.At each site, triplicates of 10-liter water samples were collected from 30 cm below the water surface in sterile carboy containers. The water samples were then transported on ice to the laboratory, and processed within 6–8 h. Sample site characteristics and GPS coordinates are provided in Table 1. In addition, complementary data from the matched EHMP sites are provided in Table 2. Furthermore, site BR1 is located on the upper reaches of the BR. This site receives overflow of water from the Wivenhoe Reservoir. Site BR2 is located in a peri-urban non-sewered catchment. Site BR3 is at a major tributary of the Brisbane River. The catchment where site BR3 is located has residential and industrial developments and is serviced by a wastewater treatment plant (WWTP). Sites BR5 and BR6 are located on the lower reaches of the river, in highly urbanized areas and is tidally influenced. The catchment had not received any rainfall 7 days prior to sampling.Samples were filtered through a 0.45 μm pore size (47 mm diameter) hydrophilic membrane (Durapore polyvinylidene difluoride, PVDF, Millipore, Tokyo, Japan), and the dissolved organic carbon in 30 mL of sample was determined in triplicate using a 820 TOC (total organic carbon) analyzer (Model: Sievers, GE Analytical Instruments Inc., Boulder, CO, USA).A 10 mL aliquot of water from each site was filtered through a 0.45 µm pore size nitrocellulose membrane (Millipore) prior to acidifying to 2% with concentrated nitric acid (AR grade; Sigma Aldrich, Castle Hill, NSW, Australia). Each sample was prepared in triplicate, with three samples collected per site (n = 45, for all five EHMP sites). Aluminum (Al), cadmium (Cd), cobalt (Co), chromium (Cr), copper (Cu), iron (Fe), lead (Pb), nickel (Ni) and zinc (Zn) were determined using an Agilent 7700x quadrupole-type ICP-MS (Agilent Technologies, Mulgrave, VIC, Australia) equipped with an Agilent ASX-520 auto sampler. The instrument was operated in He-mode. The integration time was 0.3 s per mass, 1 point per mass, 3 replicates and 100 sweeps per replicate. All samples and standards were stored in Falcon tubes (15 mL) that were rinsed with 2% nitric acid, and kept at 4 °C until analyzed. All consumables were soaked in 10% nitric acid for at least 24 h and rinsed repeatedly with MilliQ and 2% nitric acid (in MilliQ water) before use.The membrane filtration method was used for the isolation and enumeration of FIB. Serial dilutions of water samples were made in sterile MilliQ water, and filtered through 0.45-µm pore size nitrocellulose membrane (Merck Millipore, Bayswater, VIC, Australia). Dilutions were placed on modified membrane-thermotolerant Escherichia coli agar medium (modified mTEC agar, Difco, Detroit, MI, USA) and membrane-Enterococcus indoxyl-d-glucoside (mEI) agar (Difco) for the isolation of E. coli and Enterococcus spp., respectively. Modified mTEC agar plates were incubated at 35 °C for 2 h to recover stressed cells, followed by incubation at 44 °C for 22 h, while the mEI agar plates were incubated at 41 °C for 48 h [30,31].Four sub-samples (2 L) of the 10 L parent sample were filtered through 0.45 µm pore size nitrocellulose membrane. Multiple membranes were used in case of membrane clogging due to sample particulates. The membrane(s) were immediately transferred into a sterile 15 mL Falcon tube containing phosphate buffered saline (Sigma-Aldrich, St. Louis, MO, USA). The sample tube was vortexed for 5 min to detach the microbial biomass from the membrane, followed by centrifugation at 4500 g for 15 min at 4 °C to obtain a pellet [32]. One sample tube was used for DNA extraction and subsequent metagenomics analysis. The remaining three tubes were used for metabolite extraction and subsequent community metabolomic analysis.DNA was extracted from the pellet obtained from each water sample using the MO BIO PowerSoil® DNA Isolation Kit (MO BIO Laboratories, Carlsbad, CA, USA) as per Ahmed et al. [33]. The extracted DNA samples were quantified using a ND-1000 spectrophotometer (NanoDrop Technology, Wilmington, DE, USA).The V5 and V6 regions of the 16S rRNA gene were amplified using the primer set described previously [34]. Noting that the V5 and V6 region has been shown to provide equivalent diversity estimates to the V4 region, and is similar to full length 16S rRNA sequencing [35]. Amplicons from each sample were pooled in equal amounts. All samples were paired-end sequenced at a length of 300 nucleotides [nt] in each direction by the University of Minnesota Genomic Center (Minneapolis, MN, USA), using version 3 chemistry on the MiSeq platform. Raw data were deposited in the NCBI Sequence Read Archive under BioProject accession number SRP062949.Sequence processing was performed using mothur software ver. 1.33.3 (http://www.mothur.org) [36]. Sequences were first trimmed to 150 nt and paired-end joined using fastq-join [37]. Quality trimming was performed to remove sequences with average quality scores <35 over a window of 50 nt, homo-polymers >8 nt, ambiguous bases, or mismatches to primer sequences. High-quality sequences were aligned against the SILVA database ver. 115 [38]. Sequences were further quality trimmed using a 2% pre-cluster [39,40], and chimera removal using UCHIME [41]. Assignment of OTUs was performed at 97% identity using the furthest-neighbor algorithm. Taxonomic assignments were made against the Ribosomal Database Project database ver. 9 [42]. For comparisons among BR sampling sites, sequence reads for each replicate were rarefied by random subsample to 25,000 (75,000 sequence reads per site).Overall, 45 samples were collected from five sample sites. Of note, each sample site was collected in triplicate and then subsequently analyzed in triplicate. All samples were derivatized prior to analyses by gas chromatography-mass spectrometry (GC-MS) as mentioned in previous studies [43,44,45]. Briefly, a 1.0 mL aliquot of ice cold methanol (LC grade, ScharLab, Sentemanat, Spain) and MilliQ water (50:50 v/v) were added to each sample pellet, then vortexed briefly before centrifugation at 572.5 g for 15 minutes at 4 °C. Adonitol (20.0 µg/mL, HPLC grade, Sigma-Aldrich, Castle Hill, NSW, Australia) was added as an internal standard. A 100.0 µL aliquot of the supernatant was then transferred to a fresh tube and dried in a centrifugal evaporator at 210 g and 37 °C (Model number: RVC 2-18; Martin Christ Gefriertrocknungsanlagen GmbH, Osterode, Germany).For GC-MS analysis, dried samples were derivatized using 40.0 µL methoxyamine HCl (20 mg/mL in pyridine) followed by 70.0 µL N,O-bis(trimethylsilyl)trifluoroacetamide (BSTFA) in 1% trimethylchlorosilane (TMCS). Samples were then briefly vortexed before being transferred to GC-MS vials and microwaved for 3 min at 120 °C/600 W using a Multiwave 3000 system (Perkin Elmer Inc., Melbourne, VIC, Australia).Single quadrupole GC-MS was performed as previously reported [46,47]. Briefly, an Agilent 7890B gas chromatograph (GC) oven coupled to a 5977A mass spectrometer (MS) detector (Agilent Technologies) was used. The GC-MS system was fitted with a 30 m HP-5MS column, 0.25 mm internal diameter and 0.25 µm film thickness. Injections (1.0 µL) were performed in 1:10 split mode, with the oven held at an initial temperature of 70 °C for 2.0 min. The temperature was then ramped-up to 300 °C at 7.5 °C·min−1 and the final temperature (300 °C) was held for 5.0 min. The transfer line was held at 280 °C. Total ion chromatogram (TIC) mass spectra were acquired within a range of 45–550 m/z, with a 2.89 spectra·s−1 acquisition frequency. A solvent delay time of 7.5 min ensured that the source filament was not saturated or damaged. Data acquisition and spectral analysis were performed using the Qualitative Analysis software (Version B.07.00) of MassHunter workstation. Qualitative identification of the compounds was performed according to the Metabolomics Standard Initiative (MSI) chemical analysis workgroup [48] using standard GC-MS reference metabolite libraries (NIST 14, Fiehn and Golm) and with the use of Kovats retention indices based on a reference n-alkane standard (C8-C40 Alkanes Calibration Standard, Sigma-Aldrich, Castle Hill, NSW, Australia). For peak integration, a 5-point detection filtering (default settings) was set with a start threshold of 0.2 and stop threshold of 0.0 for 10 scans per sample.The metagenomics, metabolomics, trace metal and water quality data were subjected to further statistical analysis involving multivariate analyses of Principal Component Analysis (PCA) and Partial Least Square-Discriminant Analysis (PLS-DA) using SIMCA 14.1 (MKS Data Analytics Solutions, Uméa, Sweden). For the merging of these multiple datasets, the data was first imported, matched by sample location identifiers and log transformed. The analyses provided spatial distribution of various bacteria, their intracellular and extracellular metabolic activity and the relationships between the water quality parameters investigated.Analysis of the physico-chemical parameters measured during sampling and the data available on the EHMP website revealed the level of degradation of the sampled system as a temporal snapshot. The salinity measurements of the sites indicate that BR4 and BR5 are above the recommended threshold for a freshwater ecosystem in the “Australian and New Zealand Guidelines for Fresh and Marine Water Quality (AWA)” [49] (threshold value of ca. 1.1 ppt). However, it is noteworthy to mention that these sites are influenced by the tide and the nearby estuary, although samples were collected during low tide. All the other sites were within the acceptable threshold for salinity (Table 1).The AWA guidelines also state that lowland river systems have turbidity limits between 6–50 NTU. For the sites sampled, BR3 was observed to be above this threshold (137 NTU); the tidal influenced sites (BR4 and BR5) were also within this range but are considered high for estuaries/marine environments (which have a threshold of 0.5–10 NTU).All sites were within the acceptable range stated in the guidelines with respect to pH (6.5–8.0). However, site BR2 was observed at the maximum range of the threshold (pH 8.0). For the EHMP derived data, chlorophyll a levels (5 μg·L−1) were observed to be within acceptable limits with the exception of site BR1 (6.3 μg·L−1). The total phosphorous (TP) (acceptable level ≤50 μg·L−1) was high for all the sites (above 50 μg·L−1), with sites BR2 (320 μg·L−1), BR3 (270 μg·L−1), and BR4 (140 μg·L−1) observed to contain exceptionally high TP when compared to the threshold value. Likewise for filterable reactive phosphate (FRP) (acceptable level ≤20 μg·L−1), all sites were observed above the guideline value, with sites BR2 (240 μg·L−1) and BR3 (260 μg·L−1) the highest. For total nitrogen (TN) (acceptable levels ≤500 μg·L−1), sites BR2 (880 μg·L−1) and BR3 (810 μg·L−1) were above the guideline value. Nitrogen oxide(s) (NOx) (acceptable level ≤60 μg·L−1) was high for all sites with the exception of BR1 (31 μg·L−1). Sites BR2 (500 μg·L−1), BR3 (550 μg L−1), BR4 (210 μg·L−1) and BR5 (100 μg·L−1) were observed to have significantly higher NOx content than the threshold. Lastly, all sites were below the threshold regulatory limit for ammonium based nitrogen (NH4+) (<20 μg·L−1).Heavy metal pollution of waterways is typically associated with mining activities or discharges from manufacturing industries. Heavy metal pollution in water and sediments can have serious effects on the aquatic ecosystem and can make water unsuitable for livestock and/or human consumption. Furthermore, some animals (i.e., fish, shellfish and oysters) can also “bio-accumulate” metals [17], making them unsafe for consumption. As such, the concentration of metals in an urban stream is of interest, more so when the stream has the potential to further impact downstream fisheries in estuaries and marine environments.In the context of this study, soluble metals in the sampled river were analyzed because they would most likely impact the planktonic biota (in terms of abundance and diversification) and their metabolism (i.e., metabolic output). However, metals bound within sediments and biofilms, although important, were considered outside the scope of this investigation and is the focus of future work.All trace metals analyzed were below the trigger values set for freshwater aquatic ecosystems at the 90% level of protection of species in the guidelines. The trigger values were set at 80 μg·L−1 for aluminum, 0.4 μg·L−1 for cadmium, 1.8 μg·L−1 for copper, 5.6 μg·L−1 for lead, 13 μg·L−1 for nickel, and 15 μg·L−1 for zinc. The guidelines have no set limit for chromium, cobalt and iron [49]. However, as indicated in Table 3, site BR3 was observed to have elevated levels for all the metals analyzed. In particular, site BR3 was observed to have significantly higher concentrations of aluminum (2.01 μg·L−1) and iron (4.53 μg·L−1) compared to the up-stream and down-stream sampling sites. It also was observed to have slightly higher concentrations of cobalt (2.38 ng·L−1), chromium (1.0 ng·L−1), copper (8.1 ng·L−1), lead (7.6 ng·L−1) and nickel (36.9 ng·L−1). It is noteworthy to mention that site BR3 was located near a wastewater treatment plant and is located at the junction of a side stream that enters into the BR.Among the 15 water samples analyzed from the five sites, all samples yielded culturable E. coli and Enterococcus spp. The concentrations of E. coli and Enterococcus spp. in water samples ranged from 15–307 colony forming units (CFU) and 4–544 CFU per 100 mL of water, respectively (Table 4).The concentrations of E. coli and Enterococcus spp. were much greater in water samples from sites BR3 and BR4 compared to the other sites. Furthermore, sites BR3, BR4 and BR5 were all classed as poor quality (Class D) using the Microbial Water Quality Assessment Category (MAC) framework outlined in the Australian recreational water guidelines. It should be noted that sites BR1, BR2, and BR3 all had evidence of wildlife within the sampling sites. Sites BR4 and BR5 are also known human recreational locations and may have wildlife present that was not observed at the time of sampling. These wildlife factors may influence the FIB measurements obtained [50,51]. Interestingly, in a parallel investigation by Ahmed et al. [33], it was observed that human fecal contamination was present at site BR3 using a microbial source tracking toolbox (MST) approach targeting human-wastewater-associated Bacteroides HF183 molecular markers [33]. In addition, Aves (avian) fecal contamination was observed using the avian-associated GFD molecular marker at sites BR1, BR3, BR4, and BR5; Bovinae (cow) fecal contamination was observed using the cattle-associated CowM3 molecular marker at site BR2 and Equidae (horse) fecal contamination was observed using a horse-associated molecular marker at site BR4.The estimated Good’s coverage of the sample groups (75,000 sequence reads per site, with 3 independent samples subsampled to 25,000 sequence reads) ranged from 97% to 100%, with an average of 96% ± 1.2% among all samples. An average Shannon diversity index of 5.00 ± 0.24, OTU richness value of 1725 ± 172, and abundance-based coverage estimate of 4024 ± 1035 richness were observed among all samples. Table 5 provides a summary of the bacterial metagenomics data based on observed and unique features per order, family and genera for each sampled site. Figure 2 illustrates the bacterial order profile summary of the sampled sites and Figure 3 provides an overview of the site features in terms of similarity and uniqueness as presented as a Venn diagram.The unique Family features for the sites were Clostridiales incertae sedis, Desulfonatronaceae, Corynebacteriaceae, GpX, and Dermatophilaceae for Site BR1; Incertae Sedis, Aquificaceae for site BR2; Ktedonobacteraceae, Herpetosiphonaceae, Aerococcaceae, and Spirochaetales incertae sedis for site BR3, Brevinemataceae, Euzebyaceae, Dietziaceae, Thermoactinomycetaceae 1, Saccharospirillaceae, Cohaesibacteraceae, Dermacoccaceae, Clostridiaceae 2, Psychromonadaceae, Rubrobacteraceae, and Thiohalorhabdus for site BR4; and, Aquificales incertae sedis, Thermosporotrichaceae, Desulfarculaceae, Congregibacter, Cellulomonadaceae, Acholeplasmataceae, Bartonellaceae, Thermolithobacteraceae, Lactobacillaceae, Leuconostocaceae, Micromonosporaceae, Promicromonosporaceae, and Sphaerobacteraceae for site BR5.The GC-MS analysis of the samples indicated a presence of 289 peaks per chromatogram, of which 54 were considered statistically significant (S/N ratio ≥50 with an adjusted p-value ≤ 0.05). Univariate and multivariate statistical tools such as t-test, Principal Component Analysis (PCA) and Partial Least Square-Discriminant Analysis (PLS-DA) were used to analyze the distribution and classification of the various metabolites. Due to the unsupervised nature of the data and the number of sample sites, PCA was observed as a less satisfactory method to discriminate between the metabolite distributions. As such, samples were processed further using Partial Least Square-Discriminant Analysis (PLS-DA). PLS-DA is used to examine large datasets and has the ability to measure linear/polynomial correlation between variable matrices by lowering the dimensions of the predictive model, allowing easy distribution between the samples and the metabolite features that cause the distribution.The data quality of PLS-DA model was assessed by the linearity (R2X) and predictability (Q2), which were observed at 0.8294 and 0.565, respectively. These are indicative of a model that reasonably fits the data and has a weak/moderate predictive capability (~0.5). Figure 4A illustrates the PLS-DA score scatter plot of the metabolomic dataset groups (sample sites), and Figure 4B illustrates the loading scatter plot of the observed metabolites. The majority of the identified metabolites were sugars, fatty acids and amino acids. Secondary metabolites such as perillyl alcohol, lithocholic acid and phytol were also observed. As biological datasets tend to significantly vary from sample to sample, a distance of observation (DModX) analysis was also used to identify and eliminate any outliers. DModX is the normalised observational distance between variable set and X modal plane and is proportional to variable’s residual standard deviation (RSD). “DCrit (critical value of DModX)”, derived from the F-distribution, calculates the size of observational area under analysis. The DModX plot (not shown) data indicate that no samples exceeded the threshold for rejecting a sample. The threshold for a moderate outlier is considered when the sample DModX value is twice the DCrit at 0.05, which, in this instance, was 2.897 (DCrit  =  1.435). Table 6 lists the ‘identified’ significant metabolites after Benjamini-Hochberg adjustment. The unique metabolite features for the sites were Unknown Compound 13 (MW = 218.2) for site BR2; Xylitol (dTMS), l-Arabinose (4TMS), Unknown Compound 4 (MW = 325.2), and Unknown Compound 15 (MW = 189.1) for site BR3; Phytol mixture of isomers, Erythritol (4TMS) and d-Fructose (5TMS) for site BR4; and, Unknown Compound 18 (MW = 278.2) and Unknown Compound 9 (MW = 325.2) for site BR5. Figure 5 provides an overview of the site metabolite features in terms of similarity and uniqueness as presented as a Venn diagram.As illustrated in the summary table (Table 7), an assessment of the water quality parameters in isolation is often difficult and tedious to decipher in terms of the system’s health and resilience; not to mention looking at the metagenomics and metabolomics data in isolation, due to the volume of data. An elevated result or a breach of the guidelines may not necessarily mean that the site or system is degraded. For example, the microbial indicators of the sites sampled suggest that sites BR3, BR4, and BR5 may pose a risk to human health (and were indeed classed as low quality). However, as illustrated in the study by Ahmed et al. [52] of the same samples, only site BR3 was observed to have a human wastewater signature. Likewise, sites BR4 and BR5 had elevated salinity levels according to the guidelines but it was noted that these sites were heavily influenced by the tide. As such, it is important to note that such data only provide a snapshot of the system at the time of sampling and may not represent the characteristics of the overall system at all times. One approach to overcome such problems is to sample the system more frequently (both temporarily and longitudinally). However, this will significantly increase the cost of analysis. An alternative approach that requires fewer samples to be collected is a multi-omics approach. Environmental multi-omics relies on a deeper analysis of the system being sampled in terms of bacterial diversity and metabolic output. Furthermore, it combines metadata to investigate relationships between sites. While it is ideal to do such an analysis over a period of time in order to establish seasonal trends, the study presented herein demonstrates its application and illustrates the added value of such an approach.As such, in order to assess the entire system (from site BR1 through to BR5 from the perspective of heavy metals, physical and chemical parameters, metabolites and bacterial diversity), first the multiple datasets collected need to be collated and analyzed using a multi-omics approach in order to see if the data provide insight into the river system’s health. Investigating complex systems in isolation, whether it be analyzing measurements or sites in isolation, without consideration of upstream and downstream conditions, can result in an incorrect assessment of overall health or degradation. To this end, a series of PLS-DA plots were created in order to combine the multiple datasets presented herein. Each dataset was first matched by site name and log transformed in SIMCA to normalize the data. This enabled the data to be interrogated and provided a greater depth of analysis compared to investigating each site and parameter in isolation. The following section details such an assessment using the MAC characterization (i.e., Class A and D; which is also the same grouping as turbidity), the salinity data (i.e., high and low salinity) and MAC classification in combination with low salinity site data to categorize sites for comparison.After the data were uploaded into SIMCA individually, matched by sample location identifiers and log transformed, they were then grouped based on the MAC category of ‘Class A’ and ‘Class D’. The resulting PLS-DA model was assessed by the linearity (R2X and R2Y) and predictability (Q2), which were observed at 0.584, 0.987 and 0.750, respectively. This is indicative of a model that reasonably fits the data and has a good predictive capability (>0.7). Figure 6A illustrates the PLS-DA score scatter plot of the combined datasets grouped based on MAC values (i.e., Class A and Class D), and Figure 6B illustrates the loading scatter plot of the observed parameters.Using the MAC Class PLS-DA model, the dominant significant taxa classified at the class level for the ‘Class A’ pooled samples were Acidobacteria, Alphaproteobacteria, Anaerolineae, Bacilli, Betaproteobacteria, Chlamydiae, Chloroflexi, Elusimicrobia, Fusobacteria, Gammaproteobacteria, Gemmatimonadetes, Holophagae, Ignavibacteria, Ktedonobacteria, Mollicutes, Negativicutes, Nitrospira, Opitutae, Spartobacteria, Spirochaetes, Thermodesulfobacteria, Verrucomicrobiae, and Zetaproteobacteria. The dominant significant metabolic features were metabolites relating to carbohydrate metabolism (l-gulose, l-arabinose), glucagon signaling pathway (α-d-Glucose-1-phosphate, dipotassium salt dihydrate), and starch and sucrose metabolism (d-Cellobiose). Furthermore, no trace metals were correlated with the pool ‘Class A’ sample cohort.In contrast, the dominant significant taxa classified at the class level for the ‘Class D’ pooled samples were Armatimonadetes, Chlorobia, Chloroplast, Chrysiogenetes, Chthonomonadetes, Clostridia, Cyanobacteria, Deferribacteres, Dehalococcoidetes, Deinococci, Epsilonproteobacteria, Fibrobacteria, Flavobacteria, Lentisphaeria, Planctomycetacia, Sphingobacteria, Synergistia, Thermolithobacteria, Thermomicrobia, and Thermotogae. The dominant significant metabolic features were metabolites relating to secondary bile acid biosynthesis (Lithocholic acid), carbohydrate metabolism (3,6-anhydro-d-galactose), fatty acid biosynthesis (capric acid), fructose and mannose metabolism (d-mannose), biosynthesis of unsaturated fatty acids (erucic acid methyl ester), and pentose and glucuronate interconversions (d-ribulose), in addition to chemical markers commonly found in human waste stream such as phytol mixture of isomers (manufacture of synthetic forms of vitamin E and vitamin K1), and osteoarthritis medication (d-glucosamine hydrochloride). Lastly, the trace metals of Al, Cr, Fe, Co, Ni, Cu, Zn and Pb were associated with the pooled ‘Class D’ sample cohort.This suggests that ‘Class D’ pooled samples are correlated based on a number of factors, primarily bacteria that are known to cause or influence algae blooms (such as Cyanobacteria), organisms that lack aerobic respiration (Clostridia, Synergistia) and a number of green sulfur and non-sulfur bacteria (Chlorobia, Thermomicrobia), which are exacerbated due to the presence of pollutants (such as the presence of human waste stream indicators and heavy metals). Furthermore, bacteria capable of dehalogenating polychlorinated aliphatic alkanes and alkenes (Dehalococcoidetes) and organisms highly resistant to environmental hazards (Deinococci) were more abundant in Class D pooled samples. The presence of such organisms suggest the organisms within the sites are resistant to pollutants. However, the presence of Fibrobacteria suggests that commensal bacteria and opportunistic pathogens may also be present. In contrast, ‘Class A’ pooled samples were found to have organisms more commonly found in soil and aquatic environments, with no significant human waste-derived contaminants or metals present.The data were grouped based on salinity data which was classed as ‘Low’ (<1.0 ppt) and ‘High’ (~30 ppt). The resulting PLS-DA model was assessed by the linearity (R2X and R2Y) and predictability (Q2), which were observed at 0.450, 0.983 and 0.911, respectively. This is indicative of a model that reasonably fits the data and has an excellent predictive capability (>0.9). Figure 7A illustrates the PLS-DA score scatter plot of the combined datasets grouped based on Salinity values (Low and High), and Figure 7B illustrates the loading scatter plot of the observed parameters.Using the salinity class PLS-DA model, the dominant significant taxa classified at the class level for the ‘Low’ salinity sample sites were: Acidobacteria, Alphaproteobacteria, Anaerolineae, Bacilli, Betaproteobacteria, Chlamydiae, Chloroflexi, Elusimicrobia, Fusobacteria, Gammaproteobacteria, Gemmatimonadetes, Holophagae, Ignavibacteria, Ktedonobacteria, Negativicutes, Nitrospira, Opitutae, Spartobacteria, Spirochaetes, Thermodesulfobacteria, Verrucomicrobiae, and Zetaproteobacteria. The dominant significant metabolic features were metabolites relating to carbohydrate metabolism (l-gulose, butanoic acid), alanine metabolism (propanedioic acid), Biosynthesis of secondary metabolites (glycerol). Furthermore, Al, Cr, Zn and Pb were correlated with the ‘low’ salinity pooled cohort.In contrast, the dominant significant taxa classified at the class level for the ‘High’ salinity sample sites were: Armatimonadetes, Caldilineae, Chlorobia, Chloroplast, Chrysiogenetes, Chthonomonadetes, Clostridia, Cyanobacteria, Deferribacteres, Deinococci, Epsilonproteobacteria, Fibrobacteria, Flavobacteria, Planctomycetacia, Sphingobacteria, Synergistia, Thermolithobacteria, Thermomicrobia, and Thermotogae. The dominant significant metabolic features were metabolites relating to secondary bile acid biosynthesis (lithocholic acid), carbohydrate metabolism (3,6-anhydro-d-galactose), fatty acid biosynthesis (capric acid), fructose and mannose metabolism (d-mannose), biosynthesis of unsaturated fatty acids (erucic acid methyl ester), osteoarthritis medication (d-glucosamine hydrochloride), and pentose and glucuronate interconversions (d-ribulose). In addition to chemical markers commonly found in human waste streams, such as Phytol, and perillyl alcohol (a monoterpene isolated from the essential oils of lavandin, peppermint, spearmint, cherries, celery seeds, and several other plants) were also detected. Lastly, Fe was associated with the pooled high salinity sample cohort. Like the previous assessment, the addition of salinity as a grouping highlights the presence of photosynthetic bacteria in addition to bacteria that are resilient to pollution sources.As illustrated in Figure 6, sites BR1 and BR2 were grouped apart from BR3. In order to further analyze this sub-grouping, the ‘High’ salinity based sites (BR4 and BR5) were removed and a subsequent PLS-DA comparison was undertaken. Figure 8 illustrates the PLS-DA comparison based on Microbial Water Quality Assessment Category class and ‘Low’ salinity. The resulting PLS-DA model was assessed by the linearity (R2X and R2Y) and predictability (Q2), which were observed at 0.560, 0.964 and 0.657, respectively. This is indicative of a model that reasonably fits the data and has an average predictive capability (≥ 0.5).This comparison highlights the increased presence of metals, short-chain fatty acids (SCFA) and sugars in site BR3 when compared with sites BR1 and BR2. Furthermore, the increased abundance of bacteria belonging to Acidobacteria, Actinobacteria, Armatimonadetes, Chloroflexi, Chloroplast, Chrysiogenetes, Chthonomonadetes, Dehalococcoidetes, Fibrobacteria, Sphingobacteria, and Thermolithobacteria suggests an environment that is capable of dehalogenating polychlorinated aliphatic alkanes and alkenes (Dehalococcoidetes) and the presence of Fibrobacteria suggests that commensal bacteria and opportunistic pathogens may also be present.Bacterial populations were observed to be affected by the nature of the sites sampled. BR1 site was observed to be rich in soil- and water-based bacteria, Ralstonia and Bordotella from Burkholdericeae family, mostly of wildlife/domestic animal sources and plant origin. Also, expectedly, a large number of Actinomycetes were also observed at BR1 and BR2 sites. The Burkholderiaceae population dropped at the agriculturally prominent site, BR2. Greater values for parameters such as temperature, pH, phosphates and nitrogen compounds resulted in a greater abundance of photosynthetic populations such as Cyanobacteria and other chloroplast-containing bacteria. The presence of cyanobacterial family II is possibly indicative of increased occurrence of sulphur compounds at BR2. It was observed that the populations of Alteromonas and related families within the Alteromonadales increased at BR2. This was unexpected as these bacteria are generally found in marine environments. However, it is probable that the tidal nature of the Brisbane River (86 km from the mouth) combined with high dissolved oxygen content [53] (9.3 mg·L−1) and added phosphorous salts (320 μg·L−1) might have resulted in higher Alteromonas, possibly due to the high phosphate metabolizing ability of organophosphorus acid anhydrolase (OPAA) (EC3.1.8.2) expression systems [54]. Due to the vicinity around wastewater treatment plants and other similar activities, site BR3 was expected to have greater abundances of Actinomycetes, Streptomyces, and Frankia and other related facultative anaerobic bacteria. The site was observed to have considerably greater turbidity with respect to previously reported values of about 50–60 NTU [53]. Actinomycetes are known to decay organic matter, especially in nutrient-rich environments such as wetlands [55] and various river samples [56], especially in association with Sphingobacteria around areas of human and animal activities. The co-occurrence of Actinomycetes with Spingobacteria was expected as it has been reported that the latter are efficient degraders of geosmin and 2-methylisoborneol, the compounds produced by Actinomycetes. These compounds have been reported to be responsible for increasing turbidity and odor of the riverine water system [57]. The current findings are therefore in line with previous reports, as greater turbidity was observed at site BR3 relative to sites BR1 and, especially, BR2. Similarly, greater turbidity at downstream sites of BR4 and BR5 may be partially attributed to less abundant sphingobacterial populations at those sites. It is also important to note that fecal coliforms, which comprises such organisms as Escherichia coli, Klebsiella pneumoniae and Enterobacter aerogenes (order Enterobacteriales) and faecal streptococci such as Enterococcus (order Lactobacillales) were not observed in significant numbers, although the study by Ahmed et al. [33] detected the presence of human and non-human fecal molecular markers.The sites BR4 and BR5 are located very close to BR mouth, at the distances of about 22 km and 13 km, respectively. The area is considered as an inter-tidal zone, with a decreased flow-rate. The inter-tidal nature was associated with decreased levels of turbidity (15.3 NTU at BR4 with respect to 137.3 NTU at BR3) and dissolved oxygen. Due to the contribution of downstream oceanic and upstream sedimentation salts (from silts, soils, human and animal activities), inter-tidal zones, especially tidal flats have reportedly higher amounts of salinity. The higher content of deposited clay in this region also contributes towards increased salinity as compared to low clay soils or sands [58]. The high nutrient deposition combined with low flow rate and slightly anoxygenic nature of water at site BR4 and BR5 sites possibly resulted in increased populations of Rhodobacteria and Rhodospiralles such as Rhodobacter, Acetobacter, and Azosirillium spp. among others. Most of these bacteria are known to be chemo- and photo-autotrophic in nature and reportedly occur around Mangrove ecosystems (which are common around Brisbane). A significant decrease in nitrogen content at site BR4 and BR5 could be attributed to the nitrogen fixing and phosphate solubilisation abilities of these bacteria [59]. Other bacterial classes with a population increase were from Gamma-proteobacteria, with the order Incertae sedis members concentrated at site BR4 and Pseudomonadales (4%) at site BR5. It is very likely that the significant decrease in phosphates, nitrogen and turbidity was caused by the solubilization facilitated by Pseudomonas and related species at site BR5 [59].It was noticed that the lower abundances of Burkholderiales at sites BR2 and BR3 were associated with decreased levels of sugars and sugar alcohols. Similarly, the greater abundances in Cyanobacteria at BR2 and Actinomycetes at BR3 are likely to influence the observed greater concentrations of SCFAs, such as butanoic and propanedioic acid. Such SCFAs enhance the biofilm formation ability of bacteria. In particular, exposure to lower concentrations of SCFAs, such as ca. 6 mM, enhances the biofilm formation ability of Actinomycetes bacteria [60]. Similarly, an increased level of erythritol could also be linked to bacterial biofilm formation. Although the contribution of erythritol to biofilm formation was observed to be less than SFCA, its greater concentration across all the sites may compensate for the deficit. Furthermore, it has been shown that erythritol, along with amino acids such as aspargine and phenylalanine, act as major contributory metabolites towards biofilm formation [61]. The abundance of SCFAs may also be indicative of mixotroph organisms able to utilize different SCFAs as organic carbon sources, either during growth or nutrient stress lipogenic phases [62]. Lastly, the sugar levels increased again at sites BR4 and BR5, very likely due to the increased fermentation caused by Rhodobacteria, Rhodospirilli, Gamma-proteobacteria and, especially, photosynthetic bacteria. As such, based on the metabolites observed, sites BR2 and BR3 have metabolites present that suggest a system that has potential to form biofilms and/or a community of mixotrophs that utilize SCFAs and that sites BR4 and BR5 are environments undergoing degradation.Multiple characterizations of the BR system were performed by various genomic, ionic and metabolic methods. The metagenomics output indicated a presence of high levels of freshwater bacteria such as Burkholdariales and lower levels of Actinomycetes and Rhodospirillae in the upstream sites. In contrast, the population levels reversed in downstream sites, affected by salinity, pH and oxygen availability changes. Human interference was indicated by the increasing populations of Actinomycetes (BR3), including fecal bacteria and Pseudomonadales (BR4 and BR5). Greater abundances of these populations in downstream areas was also possibly caused by the increased levels of sugar alcohols, such as erythritol, SCFAs and aromatic amino acids, contributing heavily towards biofilm production. Overall, the multi-omics approach presented herein was able to provide a deeper insight into water quality contamination and riverine health in terms of metabolic and microbial properties that are not possible using traditional water quality methods. As such, a multi-omics based approach should be considered when characterizing complex environmental systems, in particular when assessing the impacts of agricultural practices, sewage treatment and environmental endpoints.The authors would like to acknowledge the assistance and support provided by the CSIRO’s Microbiology Water Quality Sciences Group, in particular the support of J. Sidhu and A. Palmer for their assistance in collecting water samples used in this study. The authors would also like to thank and acknowledge the financial support of the CSIRO Land and Water business unit. Amplicon sequence data were processed and analyzed using the resources of the Minnesota Supercomputing Institute.D.J.B. conceived, designed, and performed the metabolomics experiments, metals, statistical analysis and wrote the paper. A.V.K. assisted with the metabolomics experiments and in interpreting the metabolomics data. W.A. conceived, designed, and performed the metagenomics experiments. S.C. assisted with collating and analyzing the complimentary EHMP and metadata. P.D.M. performed the ICP-MS analysis and assisted in the interpretation of data. C.S. processed and analyzed amplicon sequencing data and provided review of the manuscript. M.J.S. provided review of the manuscript. E.A.P. assisted in interpreting the data and preparing the manuscript.The authors declare no conflict of interest.Map of the Brisbane River (BR) and the selected sampling sites (BR1–BR5).Bacterial order (top 17) profile of the BR sample sites. Note: ‘others’ represent orders less than 2% of the total sequence abundance.Bacterial metagenomics similarity and uniqueness characterization based on (A) order; (B) family and (C) genus.PLS-DA plot of the identified metabolites. (A) PLS-DA Score Scatter plot; (B) PLS-DA Loading Scatter plot.Metabolite similarity and uniqueness characterization.PLS-DA plot of the metadata and multi-omics datasets based the Microbial Water Quality Assessment Category class assessment. (A) PLS-DA Score Scatter plot; (B) PLS-DA Loading Scatter plot.PLS-DA plot of the metadata and multi-omics datasets based the Microbial Water Quality Assessment Category class assessment. (A) PLS-DA Score Scatter plot; (B) PLS-DA Loading Scatter plot.PLS-DA plot of the metadata and multi-omics datasets based the Microbial Water Quality Assessment Category class and low salinity assessment. (A) PLS-DA Score Scatter plot; (B) PLS-DA Loading Scatter plot.Sample site description and in-situ water quality characteristics. Water quality parameters were analyzed in triplicate (n = 3) and the relative standard deviation as a percentage (%RSD) are presented in the parentheses.Note: EHMP is defined as “Ecosystem Health Monitoring Program”. * DOC is defined as dissolved organic carbon.Sample site description and EHMP site matched water quality characteristics.Note: EHMP is defined as “Ecosystem Health Monitoring Program”. * FRP is defined as filterable reactive phosphorus.Concentrations of metals in the water sourced from different sites and attributed to different sources. Values in the parenthesis denote standard deviations between the samples (n = 9).Microbial water quality based on the various coliform counting methods. Values in the parenthesis indicate standard deviations (n = 9).a Determined using the geometric mean of nine samples (n = 9); b Recreational Microbial Water Quality Assessment calculated using Enterococcus spp. data following the NH&MRC “Guidelines for Managing Risks in Recreational Water”; c Calculated using the ranked method (n = 9).Summary of site bacterial metagenomics characterization.Most significant metabolites from the sampled river sites identified by based on their fold change (FC), p-value and Adjusted (Adj.) p-values.Physico-chemical and microbial water quality summary.Note: * MAC is defined as the Microbial Water Quality Assessment Category.

Med-MDPI/ijerph_2/ijerph-14-03-00304.txt

@@ -0,0 +1 @@
+Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).Educational posters are used to enhance knowledge, attitudes and self-confidence of patients. Little is known on their effectiveness for educating health care professionals. As these professionals may play an important role in suicide prevention, the effects of a poster and accompanying evaluation and triage guide on knowledge, self-confidence and attitudes regarding suicidal thoughts and behaviours, were studied in a multicentre cluster randomised controlled trial, involving staff from 39 emergency and 38 psychiatric departments throughout Flanders (n = 1171). Structured self-report questionnaires assessed the knowledge, confidence and beliefs regarding suicidal behaviour management, and attitudes. Data were analysed through a Solomon four-group design, with random assignment to the different conditions. Baseline scores for knowledge and provider confidence were high. The poster and accompanying evaluation and triage guide did not have an effect on knowledge about suicide and self-confidence in suicidal behaviour management. However, the poster campaign appeared to be beneficial for attitudes towards suicidal patients, but only among staff from mental health departments that were assigned to the un-pretested condition. Given the limited effects of the poster campaign in the studied population with a relatively high baseline knowledge, the evaluation of this poster as part of a multimodal educational programme in a more heterogeneous sample of health care professionals is recommended.Suicide is a major global public health problem accounting for more than 800,000 deaths each year. The annual mortality rate is estimated at 11.5 deaths per 100,000 people, which equates to one death every 40 [1].Accumulating evidence shows that training and educating gatekeepers is a worthwhile investment in suicide prevention. Gatekeepers may play a pivotal role in the early identification, management and referral of suicidal patients [1,2,3,4]. They can be among the first to screen and intervene for suicide risk as they may be in close contact with suicidal individuals and therefore have the opportunity to interrupt an ongoing suicidal process [5,6]. Front-line health professionals, such as general practitioners, mental health professionals and emergency department staff, report that targeted education and training in suicide prevention would be helpful [3,7,8,9,10]. Furthermore, a broad range of both health and community professionals appear to benefit from education and training interventions [3,11,12,13,14]. In particular, such interventions may provide gatekeepers with better knowledge about suicide, promote more adaptive attitudes towards suicidal patients, and increase provider confidence in assessing and managing suicide risk [15,16,17,18,19,20,21,22].The field of suicide prevention has made great strides in developing education and training interventions for various key groups, e.g., gatekeeper training, workshops with actors role-playing patients, e-learning, 1-day train-the-trainer programmes, and educational posters campaigns [2,23,24,25,26]. Educational poster campaigns have been used for a long time in various health domains. Targeting a wide range of health promotion issues they may well be a promising tool to raise awareness, increase knowledge and elicit behaviour change among patients [27,28].Educating health professionals about suicide prevention is a component of many national suicide prevention strategies, but the effects of educational poster campaigns regarding early detection of suicide risk, intervention, follow-up and referral of suicidal patients have hardly been studied. Currier et al. [3]) suggested that an educational poster and accompanying evaluation triage guide may be a simple and cost-effective tool for emergency department staff in increasing provider awareness and improving provider perception of knowledge and skills regarding the identification and management of suicidality. However, the interpretation of the findings of this study is hampered by methodological shortcomings including the use of non-validated measurements and the inclusion of only one comparator site.Therefore, the current study aimed at developing and evaluating a poster campaign using validated measurements and multiple controls. It was hypothesized that the poster and accompanying evaluation and triage guide will improve knowledge regarding suicidality, will increase provider confidence in assessing and treating at-risk individuals, and will lead to more adaptive attitudes towards suicidal patients for staff of both emergency and psychiatric departments in Flanders.We used a Flemish adaptation of the “Is Your Patient Suicidal?” poster that was originally developed by the Suicide Prevention Resource Center (SPRC) in the United States [3]. This is a four-color A3 size poster entitled “Is Your Patient Suicidal?”, which is accompanied by a 1-page, double-sided clinical triage guide. The poster offers information on identifying and responding to high-risk patients, including (1) the most common and manifest signs of acute suicide risk; (2) facts and figures; (3) questions that can be used to detect and discuss suicidal ideation and attempt history when signs are noticed of suspected; and (4) referral for additional suicide prevention services. The accompanying guide “Suicide risk: A guide for evaluation and triage” provides further guidance to identify suicidal ideation and suicidal intent, triage criteria to evaluate the level of risk (including interventions concerning high-risk patients, moderate-risk patients and low-risk patients), and checklists for discharge and documentation.The educational materials were adapted to the Flemish context of this trial and field-tested in different focus groups, including the Flemish task force of suicide prevention and clinical staff of emergency and psychiatric departments. The poster and guide were displayed for four weeks in strategic staff-only sites such as meeting rooms, lunchrooms and staff toilets.Emergency and psychiatric departments were recruited from July 2013 until January 2014. In total, 49 Flemish hospitals agreed to participate, accounting for 64.5% of the total number of hospitals in the Flanders region. At the individual level, 2364 health professionals from emergency and psychiatric departments throughout Flanders were invited to participate, of whom 1171 (49.5%) agreed to participate. The study population included 638 (54.5%) emergency department (ED) staff and 533 (45.5%) mental health professionals.In order to evaluate the impact of the educational poster campaign on knowledge, confidence and attitudes of staff of these two types of hospital departments, a Solomon four-group design was used. This design allows for the control of pretesting effects by including both experimental and control conditions with and without initial pretesting [29]. Therefore, the subjects of the two department types were randomly assigned to one of the following four groups: (1) one experimental group of health professionals being assessed before and after exposure to the poster campaign (n = 212; 14 departments); (2) one control group of health professionals being assessed twice over a time frame comparable to the experimental group (n = 338; 22 departments); (3) one experimental group of health professionals being assessed only after exposure to the poster campaign (n = 298; 21 departments); and (4) one control group of health professionals assessed only once (n = 323; 18 departments).As the poster campaign was conducted at the department level, and not at the individual level, a cluster design was adopted with department being the unit of randomisation. In order to fulfil the sample size requirements, no restrictions on cluster size were imposed to the hospital departments. Since some hospital departments covered over 60 potential subjects, while others only identified 15 eligible participants, cluster sizes vary. Consequently, there was a slight difference in the numbers of subjects of the two hospital departments assigned to each condition.As the power of a Solomon four-group design is supposed to be greater than that of a post-test-only control group design [30], the power was calculated based on a two-sample t-test on the post-test-only groups. To account for the cluster design, an intra-cluster coefficient (ICC) of 0.05 was assumed. As no values for ICC under this setting were available in the literature, an assumption was made based on general practice and medical trials in which ICC values were reported between 0.01 and 0.05 [31,32]. In order to demonstrate an effect-size of 0.4 between the intervention groups (1 and 3) and the control groups (2 and 4) with a power of 80%, a statistical significance of 5% and an ICC of 0.05, 16 departments of at least 15 health professionals were needed in both the intervention groups (1 and 3) and the control groups (2 and 4).Staff of both emergency and psychiatric departments in Flanders were eligible to participate in the study if they (a) were 18 years or older; (b) were in close contact with suicidal patients (e.g., physicians, physician assistants, psychiatrists, psychologists, and nurses) and (c) provided informed consent to participate.Non-clinical hospital staff (e.g., administrative personnel, and ambulance drivers) were not eligible for participation. Study coordinators of participating departments were intensively informed about the educational intervention. In order to avoid bias of results, they were excluded from participating in the study.Ethical approval was obtained from the ethical board of the University Hospital of Ghent in accordance with the ethical principles expressed in the Declaration of Helsinki (ethical approval code EC/2013/473). Given the multicentre character of the study, the study protocol was also approved by the institutional review board of each participating site.Consistent with the study procedure of the SPRC, the study involved three phases including (1) completion and collection of baseline questionnaires (lasting 3 weeks; Questionnaire S1); (2) exposure to the educational poster campaign (displayed for 4 weeks); and (3) completion and collection of follow-up questionnaires (lasting 3 weeks).The study coordinators (mostly head nurses) of the departments were asked to facilitate the study by distributing the paper-and-pencil surveys during regular staff meetings and encouraging personnel to participate. In order to match baseline and follow-up questionnaires without compromising the confidentiality of staffs’ responses, each participant created his or her own unique identification number.After giving informed consent, clinical hospital staff anonymously completed self-reports, covering sociodemographics and the following questionnaires.A subscale of the Dutch translation [33] of the 14-item Question, Persuade and Refer questionnaire (QPR) was used to assess self-perceived knowledge about suicide [34]. Levels of knowledge were assessed using questions such as ‘How do you rate your knowledge about suicide warning signs?’. Answers were given on a Likert scale ranging from 1 (very low) to 5 (very high). Responses were summed to provide a total score ranging from 7 to 35, with higher scores representing greater levels of self-perceived knowledge. The QPR has been shown to reliably assess effects of training on self-perceived knowledge of suicide prevention [23,35,36].The eight items of the Suicide Information Test (SIT) asking about warning signs and risk factors [37] was used to assess knowledge about suicide more objectively. The original questionnaire is comprised of 28 true-false items and was translated into Dutch and adjusted for a Flemish randomized controlled trial [38]. The questionnaire includes statements such as ‘Suicidal tendencies are inherited, and suicide runs in families’. Clinical staff could agree (score 1) or disagree (score 0) with the eight statements, resulting in total scores ranging from 0 (disagreed with all statements) to 8 (agreed with all statements). Higher scores thus reflect greater knowledge about warning signs and risk factors of suicide.A subscale of the Confidence and Beliefs Questions (CBQ) was used to measure provider confidence in suicidal behaviour management [22]. The questionnaire was translated into Dutch for the clinical trial of De Beurs and colleagues [33]. The subscale consists of three items. Example: ‘I am confident in my ability to successfully treat a suicidal patient’. Scoring occurs on a 5-point Likert scale ranging from ‘strongly disagree’ to ‘strongly agree’. The subscale is summed to derive a total subscore ranging from 3 to 15, with higher scores indicating higher levels of provider confidence. The CBQ has been found to measure differences in confidence regarding suicide care [22]. Attitudes towards suicidal behaviour and suicidal patients were assessed using an adjusted version of the Attitudes Towards Suicide Questionnaire (ATTS) [39]. The original 37-item instrument was translated into Dutch and was reduced to 29 items on the basis of a confirmatory factor analysis reported by De Clerck and colleagues [40]. In this study, only the three items concerning the factor willingness to help were considered relevant. Example: ‘It is a humane duty to try to stop someone from dying by suicide’. Responses are scored on a 5-point Likert scale from 1 (completely disagree) to 5 (completely agree). A sum score was calculated from 3 to 15, with higher scores reflecting more adaptive attitudes, i.e., greater willingness to help suicidal patients. The ATTS is a valid and reliable measure in clinical and community samples for determining attitudes towards suicidal behaviour, demonstrating high internal consistency and test-retest reliability [39,41,42].This study was a multicentre clustered randomized controlled trial that was conducted at emergency and psychiatric departments in Flanders. Departments were randomized to one of the four conditions based on a block design (4 per block) and stratified by department type using random allocation software. Main analyses were conducted separately for staff of emergency and psychiatric departments. The methods used are those described in ‘Statistical treatment of the Solomon four-group design: a meta-analytic approach’ [30]. However, they were adapted to the clustered design by applying mixed models with cluster (combination of hospital and department) as a random effect.Preliminary analyses included chi-square analysis and t-tests to assess baseline differences between the pretested conditions. The initial phase of the analysis associated with the Solomon four-group design, started with a 2 × 2 between-groups analysis of variance (ANOVA) on the post-test scores, with the two main effects being pre-test vs. no pre-test and intervention vs. no intervention. If no significant outcomes were observed for both the main effect of pre-test and the interaction of pre-test by poster campaign, the data from the two pretested groups were reanalysed by using a two-group analysis of covariance (ANCOVA) with pre-test scores as control variables (covariate) and post-test scores as criterion variables (dependent variables). SPSS version 22 (Chicago, IL, USA) was used for all analyses. Data were presented as means with 95% confidence intervals (CI). The level of significance was set at p < 0.05.Table 1 shows the baseline characteristics for the experimental and the control group at the department level and the staff level. At the department level, there were no differences in the number of emergency and psychiatric departments assigned to each pretested condition (7 and 11 respectively).The baseline sample consisted of 307 (55.8%) mental health professionals and 243 (44.2%) ED staff. The majority were women (n = 358, 65.1%). All age groups between 18 and 65 years were well represented, with a mean age of 38.25 years (SD = 11.23). Participants comprised predominantly general nurses (n = 269, 48.9%) and psychiatric nurses (n = 135, 24.5%). Years of professional experience as ED or mental health care provider ranged from 0 years to 40 years (M = 14.5 years, SD = 10.6). Frequency of contact with suicidal patients varied according to department, with 64.2% of mental health professionals reporting daily contact with suicidal patients compared to 16.1% of ED providers (χ2(1) = 118.436, p < 0.001). About half (54.9%) reported having additional education or training in suicide prevention within 6 months prior to the baseline assessment of this study, with significant differences between staff of psychiatric and emergency departments (70.8% vs. 35.0% respectively; χ2(1) = 63.878, p < 0.001). A significant amount of health professionals reported experience with suicide or suicidal behaviour in their personal environment, i.e., a family member (n = 100, 37.0%), close friend (n = 141, 31.2%), colleague or acquaintance (n = 227, 47.6%).Analysis regarding the subscale of the QPR questionnaire revealed a mean score of 24.1 (SD = 3.8; Min = 7, Max = 35). Almost half of the baseline sample (n = 241, 43.8%) rated their general understanding about suicide and suicide prevention as ‘high’ or ‘very high’. Significant differences in self-perceived knowledge level were found between staff of psychiatric and emergency departments, with mental health professionals reporting higher baseline knowledge scores than ED providers (M = 25.2 vs. M = 21.6; t(507) = −12.42, p < 0.001).Analysis regarding the SIT showed that baseline levels of knowledge regarding risk factors and warning signs of suicide were considerably elevated. With regard to the 3 items asking about risk factors, 81.6% of the health professionals answered at least 2 items correctly. Similar results were found for the 5 items asking about warning signs, with 75.5% of the subjects answering at least 4 items correctly. There were no significant differences in level of knowledge regarding risk factors between staff of emergency and psychiatric departments (M = 2.1 vs. M = 2.1; t(470) = −0.026, p = 0.98). However, mental health professionals appeared to have greater knowledge about warning signs than ED providers (M = 4.1 vs. M = 3.8; t(468) = −4.019, p < 0.001).With regard to attitudes, the mean ATTS subscore was 12.4 (SD = 2.0; Min = 6, Max = 15) indicating that baseline attitudes of clinical staff are quite adaptive. However, mental health professionals and ED staff significantly differed in their willingness to help suicidal patients (M = 12.8 vs. M = 11.8; t(510) = −5.695, p < 0.001). Staff of psychiatric departments more readily endorsed that ‘they are prepared to help a person in s suicidal crisis by making contact’ (93.5% vs. 76.9%; χ2(1) = 29.141, p < 0.001) and reported more disagreement with the statement ‘if someone wants to commit suicide, it is their business and we should not interfere’ (85.9% vs. 77.0%; χ2(1) = 29.141, p = 0.009).At baseline, the vast majority of clinical staff reported no hesitancy in asking about patients’ current suicidal ideation (n = 390, 76.5%) and about half reported feeling confident or very confident in its ability to successfully assess (n = 282, 54.8%) and treat (n = 256, 49.9%) suicidal patients. There were significant differences in provider confidence among staff of emergency and psychiatric departments. ED providers were less confident in the assessment and treatment of suicidal behaviour and were more hesitant to ask a patient if he or she is suicidal compared to mental health professionals (M = 10.0 vs. M = 11.7; t(507) = −9.581, p < 0.001).There was no statistically significant difference between the two pretested conditions in terms of demographics and outcome measures at baseline.First, a 2 × 2 mixed analysis of variance (ANOVA) was conducted on the four total post-test scores of the QPR questionnaire. The two factors were pre-test (yes vs. no) and poster campaign (yes vs. no). For both staff of emergency and psychiatric departments, the ANOVA showed no significant interaction effect of pre-test by poster campaign (F = (1, 44) = 1.463, p = 0.23; F = (1, 28) = 0.164, p = 0.69, respectively). As an interaction effect could not be identified (referred to as ‘Test A’) [30], an examination of the main effect of poster campaign followed (referred to as ‘Test D’) [30]. For ED providers as well as mental health professionals, this main effect was not significant (F = (1, 34) = 0.306, p = 0.58; F = (1, 28) = 0.818, p = 0.37 respectively). Furthermore, a two-group analysis of covariance (ANCOVA) was performed on the total post-test scores, covarying the total pre-test scores (referred to as ‘Test E’) [30]. “Test E is the preferred test of these three (Tests E, F, and G), however, primarily because of its greater power or ability to detect the treatment effect” [30] (p. 151). For staff of both departments, the ANCOVA showed no significant effects of poster campaign in Condition 1 and 2, the two pretested groups (F = (1, 7) = 0.117, p = 0.74; F = (1, 14) = 0.199, p = 0.66, respectively). Because significance of the ANCOVA was lacking, a t-test was performed on the scores of Condition 3 and 4, the post-test only groups (referred to as ‘Test H’) [30]. Again, for both staff of emergency and psychiatric departments the results of the t-test were not significant (t(19) = −1.14, p = 0.27; t(13) = 0.38, p = 0.71 respectively). Finally, the results of Test E and H were combined with a meta-analysis (referred to as ‘Test I’) [30]. For both ED and mental health providers, the meta-analysis was not significant (zmeta = 1.01, p = 0.31; zmeta = 0.57, p = 0.57, respectively).For staff of both emergency and psychiatric departments, no significant results were found for the SIT scores.For staff of both emergency and psychiatric departments, the results of the analyses of the CBQ scores did not achieve levels of significance.For ED providers, a 2 × 2 ANOVA on the four ATTS post-test scores revealed that the interaction effect of pre-test by poster campaign was substantial but not significant by conventional standards (F = (1, 50) = 3.130, p = 0.08). Subsequent ANOVA could not identify a significant main effect of the poster campaign on attitude (F = (1, 40) = 1.936, p = 0.17). In addition, the ANCOVA was not significant (F = (1, 9) = 0.123, p = 0.50). The t-test that was performed on the scores of the post-test only groups, seemed not to be significant (t(17) = −0.15, p = 0.88). The meta-analysis was also not significant (zmeta = 0.58, p = 0.56). Among mental health professionals, the ANOVA analysis could, however, reveal a significant interaction between pre-test and poster campaign (F = (1, 27) = 6.139, p = 0.02). Therefore, a main effects analysis was performed on the pretested groups (referred to as ‘Test B’) [30] but no significant simple effect of the poster campaign could be identified (F = (1, 9) = 0.492, p = 0.50). Subsequently, a simple main effects test was conducted on the post-test only groups (referred to as ‘Test C’) [30]. This result was significant, indicating that the poster campaign affected attitudes of mental health professionals towards suicidal patients, but only for those professionals that were assigned to the un-pretested condition (M = 12.0 vs. M = 12.6; t(14) = 2.58, p = 0.02). The corresponding between group effect size was 0.33.The present study was conducted to evaluate the impact of a brief educational suicide prevention poster campaign on knowledge, self-confidence and attitudes towards suicidal behaviour among staff of emergency and psychiatric departments. The educational poster and accompanying triage guide appears to have no effect on knowledge about suicide and self-confidence in suicide care in the studied population. However, the findings demonstrate that the poster campaign may positively affect attitudes, that is, lead to a greater willingness to help suicidal patients among staff of psychiatric departments.The results do not accord with what was expected based on the evaluation of the SPRC “Is Your Patient Suicidal?” poster campaign in the United States [3], as less powerful evidence for the effectiveness of the educational poster campaign for health care professionals was found in Flanders. In the United States, approximately half of the ED staff members exposed to the poster and guide reported improvements in their self-perceived knowledge and skills regarding detection and treatment of suicidality, although it must be added that the US study suffered from some methodological shortcomings that could interfere with the interpretation of findings.A possible explanation for the lack of effect of the poster campaign on knowledge levels can be found in the fact that in Flanders staff’s knowledge scores were already near its maximum at the start of the study. More specifically, almost half of the study group perceives their general understanding about suicide and suicide prevention as good or very good. High baseline levels regarding risk factors and warning signs are also observed, indicating that self-perceived knowledge level appears to reflect actual knowledge levels. Due to this pre-study knowledge, participants may experience a ceiling effect, making it more difficult to increase their knowledge about suicide and suicide prevention any further.At baseline, clinical staff report not only a high level of knowledge, but also express a high level of willingness to help suicidal patients. A possible explanation can be found in the reliance on voluntary participation and, as a consequence, a relative homogenous sample that may have skewed the findings to be more positive than they otherwise might have been.Further, at pre-test, the vast majority of mental health professionals perceive themselves as highly skilled in dealing with suicidal patients as they report no hesitancy in asking about patients’ suicidality and feel (very) confident in their ability to successfully detect and manage suicidal behaviour. A possible explanation can be found in the high rate of additional education or training in suicide prevention that staff of psychiatric departments received within 6 months prior to the study, which also may explain the high knowledge level at baseline. Compared to staff of psychiatric departments, ED staff report less confidence in identifying and intervening with suicidal patients and more hesitancy in discussing current suicidal ideation. This may be due to the low rate of education or training in suicide prevention among this occupational group, as only one in three ED providers were trained or educated in suicide prevention in the 6 months prior to the baseline assessment.This study provides insights into the effect of suicide prevention poster campaigns as there is a clear lack of scientifically sound evaluation of suicide prevention strategies in terms of early detection of suicide risk, intervention, follow-up and referral of suicidal patients. The clustered randomized controlled Solomon four-group design is rare in this field of research, but most certainly represents a strength of this study. A cluster randomized controlled trial of this size provides a large amount of evidence.The present study contributes meaningfully to the understanding of the effectiveness of an educational suicide prevention poster campaign as a small significant effect on attitudes of mental health care providers could be observed. However, this result should be interpreted with caution due to the increased chance of a type I error caused by the high number of comparisons made associated with the Solomon four-group design.Possible reasons for the little effect of the poster campaign may be lack of simple design impact, content and location of the posters due to information overload that often characterizes the clinical workspace.Several occupational groups were underrepresented in the study, such as physicians and psychiatrists. It is recommended to evaluate the effect of this poster as a part of a multimodal educational programme in a more heterogeneous sample thus targeting other gatekeepers as well.In this trial, the use of an educational poster campaign does not lead to improved knowledge and self-confidence and has little beneficial impact on attitudes of studied health care providers, most probably due to high levels of pre-study knowledge and experience. However, a poster campaign may be an effective tool in raising awareness when embedded in a broader suicide prevention strategy. In addition, it is recommended that pre-study knowledge is assessed and that preventive efforts using a poster particularly target care providers with limited knowledge.Advancing training to detect, intervene and follow-up individuals at risk for suicide is widely recommended as suicide prevention policy for all health professionals and gatekeepers. Therefore, the poster and accompanying triage guide will be used as an additional tool as part of broader suicide prevention training programmes that are provided in Flemish hospitals.This randomised controlled trial provides limited evidence for the effectiveness of an educational poster campaign for suicide prevention. As the studied population appears to have relatively high baseline knowledge about suicide, further research is needed in a more heterogeneous sample targeting other gatekeepers in various health domains with limited knowledge as well. Furthermore, it is recommended to evaluate the effect of this poster and accompanying triage guide as a part of a multimodal educational programme.The following are available online at www.mdpi.com/1660-4601/14/3/304/s1, Questionnaire S1: SHORT SURVEY ON KNOWLEDGE, SELF-CONFIDENCE AND ATTITUDES TOWARDS SUICIDAL BEHAVIOUR.This study was funded by the Flemish Government.Gwendolyn Portzky and Kees van Heeringen conceived and designed the experiments; Renate van Landschoot performed the experiments and analyzed the data; Renate van Landschoot, Gwendolyn Portzky and Kees van Heeringen wrote the paper.The authors declare no conflict of interest.Baseline characteristics of the intervention and control group at the cluster level and the individual level in n (%) unless otherwise stated. (Totals do not always equal 212 (intervention group) or 338 (control group) due to missing data.)

Med-MDPI/ijerph_2/ijerph-14-03-00305.txt

@@ -0,0 +1 @@
+Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).Introduction: The prevalence of adolescent electronic cigarette (e-cigarette) use has increased in most countries. This study aims to determine the relation between the frequency of e-cigarette use and the frequency and intensity of cigarette smoking. Additionally, the study evaluates the association between the reasons for e-cigarette use and the frequency of its use. Materials and Methods: Using the 2015 Korean Youth Risk Behavior Web-Based Survey, we included 6655 adolescents with an experience of e-cigarette use who were middle and high school students aged 13–18 years. We compared smoking experience, the frequency and intensity of cigarette smoking, and the relation between the reasons for e-cigarette uses and the frequency of e-cigarette use. Results: The prevalence of e-cigarette ever and current (past 30 days) users were 10.1% and 3.9%, respectively. Of the ever users, approximately 60% used e-cigarettes not within 1 month. On the other hand, 8.1% used e-cigarettes daily. The frequent and intensive cigarette smoking was associated with frequent e-cigarette uses. The percentage of frequent e-cigarette users (≥10 days/month) was 3.5% in adolescents who did not smoke within a month, but 28.7% among daily smokers. Additionally, it was 9.1% in smokers who smoked less than 1 cigarette/month, but 55.1% in smokers who smoked ≥20 cigarettes/day. The most common reason for e-cigarette use was curiosity (22.9%), followed by the belief that they are less harmful than conventional cigarettes (18.9%), the desire to quit smoking (13.1%), and the capacity for indoor use (10.7%). Curiosity was the most common reason among less frequent e-cigarette users; however, the desire to quit smoking and the capacity for indoor use were the most common reasons among more frequent users. Conclusions: Results showed a positive relation between frequency or intensity of conventional cigarette smoking and the frequency of e-cigarette use among Korean adolescents, and frequency of e-cigarette use differed according to the reason for the use of e-cigarettes.The electronic cigarette (e-cigarette) is a battery-operated device that vaporizes a solution of nicotine, glycerol, and flavoring agents [1]. It is often advertised as a healthier alternative to conventional cigarettes or as a smoking cessation aid [1,2]. The prevalence of e-cigarette use among adolescents has been increasing in many countries [3,4,5,6,7]. Several studies have found that adolescents who used only e-cigarettes were more likely to initiate the use of combustible cigarettes later [8,9,10,11,12].Current e-cigarette users are defined as adults and adolescents who have used them once or more during the past 30 days. It is important to distinguish established users from non-established users because there is a difference in the reason for e-cigarette initiation in addition to the extent of its use [13]. According to previous studies considering current cigarette smokers, the successful smoking cessation rate was higher among daily or intensive e-cigarette users compared with non-daily or intermittent users [14,15,16].The general concept of current e-cigarette use is broadly defined to differentiate experimenters from regular users. Studies have argued that most adolescent e-cigarette users are experimenters and only conventional cigarette smokers tend to use e-cigarettes on a regular basis [17]. A recent study on the distribution of the frequency of e-cigarette use among adults tried to define regular users among current adult smokers and suggested more than 5 out of the past 30 days as a definition of a “persistent user” [13]. Only a few studies have reported that the frequency of e-cigarette use was particularly associated with conventional cigarette smoking among adolescents. Warner demonstrated that, even though e-cigarette use frequency rose with the amount of ever smoking in US adolescents, the frequency of e-cigarette use was not associated with the amount of conventional cigarette smoking [18].It is important to recognize the reasons for e-cigarette initiation of adolescents because it can help us better understand what attracts adolescents to e-cigarettes. Furthermore, several reasons for initiating e-cigarettes may have increased the risk of continued use of e-cigarettes [19]. A study has reported that the reasons for e-cigarette use were related to future smoking cessation among adults [20]. Smokers who used e-cigarettes for smoking cessation showed a higher smoking cessation rate, whereas those with other reasons showed lower smoking cessation rates than non-e-cigarette users [20]. Some studies have reported predictors of continued e-cigarette use [19] and the reasons for experimentation among adolescents [21]; however, few have investigated the reasons for initiation regarding the intensity of e-cigarette use. This study aims to determine the relation between the frequency of e-cigarette use and the frequency and intensity of conventional cigarette smoking among Korean adolescents based on a nationally representative cross-sectional sample. In addition, the study identifies the association between the reasons for e-cigarette use and the frequency of its use.The Korean Youth Risk Behavior Web-Based Survey (KYRBWS) is a nationally representative cross-sectional survey of Korean middle and high school students. The KYRBWS was established in 2005 for assessing health-risk behaviors of adolescents and has provided data for the development and evaluation of school health policies and programs in Korea. The survey was approved by the institutional review board of the Korea Centers for Disease Control and Prevention (2014-06EXP-02-P-A). Written informed consent was received from all participants and their parents or legal guardians. The KYRBWS was described in detail in a previous study [22]. In brief, the KYRBWS data were collected anonymously using a multistage, stratified, cluster-sampling method. The stratification was performed on the basis of 44 provinces and types of schools according to geographic accessibility, the number of schools and population, living environment, smoking rate, and alcohol consumption. The 2015 survey included 70,362 students (ages 13–18 years) in 2400 classrooms (secondary sampling units), which included three classes considering the three year school-term (one class for each year) from each of 400 middle schools and 400 high schools (primary sampling units). Of these, 68,043 adolescents from 797 schools participated in the survey (96.7% response rate). From these participants, we analyzed 6656 adolescents who had an e-cigarette experience. Our study is based on the public use dataset (https://yhs.cdc.go.kr/new/pages/main.asp).Ever conventional cigarette smokers were defined as those who responded “yes” to the question, “Have you ever tried a cigarette, even one puff, in your life?” Among ever-smokers, current conventional smokers were defined as those who replied from “1 and 2 days” to “every day” for the question, “During the past 30 days, how many days did you smoke cigarettes, even one cigarette?” Intensity of conventional cigarette smoking was defined by the following question: “How many cigarettes did you smoke a day on average in the past 30 days?” Response options were “fewer than 1 per day,” “1 per day,” “2 to 5 per day,” “6 to 9 per day,” “10 to 19 per day,” and “20 or more per day.”The reasons for e-cigarette use were collected from the following question: “Which of the following is your main reason for using e-cigarettes?” Response options were “it seems to be less harmful,” “for smoking cessation,” “for indoor use,” “it is easy to obtain,” “it tastes better,” “it has a good flavor,” “it does not smell like tobacco,” “curiosity,” and “other.”We included several sociodemographic variables, such as age, gender, and school grade that might be associated with conventional cigarette smoking or e-cigarette use. Ever e-cigarette use was defined by a “yes” answer to the following question: “Have you ever tried e-cigarettes?” Current e-cigarette use was defined as those who replied from “1 and 2 days” to “every day” to the question, “During the past 30 days, how many days did you use e-cigarettes?” The number of days that used e-cigarette was re-grouped into 0–2 days/month, 3–9 days/month, and ≥10 days/month. All data were analyzed by considering both sample weights and the complex sample design of the survey. The number of participants was presented as unweighted samples of individuals who participated in the KYRBWS 2015 survey. The prevalence of demographic characteristics and e-cigarette use was presented on the basis of the weighted percentage of participants with 95% confidence intervals to represent Korean adolescents. The chi-square test was used for categorical variables. A threshold for statistical significance was set at a two-tailed p < 0.05 level. All data were analyzed using SPSS version 21.0 (SPSS Statistics Inc., Chicago, IL, USA).Table 1 shows the sociodemographic characteristics and the health-risk behaviors for all participants in the 2015 KYRBWS, with 52.1% being boys and 48% being girls. The prevalence of ever conventional cigarette smokers was 17.1%, and 3.7% of participants were current daily conventional cigarette smokers. Among the participants, 9.6% had not smoked cigarettes in the last month, although they were ever conventional cigarette smokers. Additionally, 1.3% of participants had smoked conventional cigarettes only for 1–2 days within month. The prevalence of ever and current e-cigarette use was 10.1% and 3.9%, respectively. Out of the total participants, 6.0% were ever e-cigarette users but had not used e-cigarettes within a month, while 1.3% had used them for only 1–2 days per month. Daily e-cigarette users were 0.7%. Out of ever e-cigarette users, approximately 60% used e-cigarettes not within 1 month. Otherwise, 16.1% used them for more than 10 days per month (Table 2). Compared with e-cigarette users for 0–2 per month, frequent users were older (16.2 years vs. 15.8 years) and were more prevalent among 12th graders than 7th graders (21.9% vs. 9.7%). Current use of e-cigarettes was more prevalent among ever conventional cigarette smokers (41.5%) than among never conventional cigarette smokers (25.3%), and e-cigarette users for more than 10/month were 2 times more prevalent among ever smokers than among never smokers (17.2% vs. 9.5%). A positive correlation was observed between the frequency of conventional cigarette smoking and the frequency of e-cigarette use. Percentage of frequent e-cigarette use was 9 times greater among daily smokers than among conventional cigarette users for <1 per month (28.7% vs. 3.5%). Smoking amount was also positively correlated to the frequency of e-cigarette use. Frequent e-cigarette use was 6 times more prevalent among smokers going for ≥20 cigarettes/day than among smokers going for <1 cigarettes/month (55.1% vs. 9.1%).Among ever e-cigarette users, the most common reason for e-cigarette use was curiosity (22.9%), followed by the belief that they were less harmful than conventional cigarettes (18.9%), the desire to quit smoking (13.1%), and the desire to smoke indoors (10.7%) (Table 3). For infrequent e-cigarette users (<3 per month), curiosity was the most frequent reason for e-cigarette use (28.8%), whereas for more frequent e-cigarette users (>10 per month), the desire to quit smoking (21.0%) and the capacity for indoor use (19.5%) were the most frequent reasons for e-cigarette use. The belief that e-cigarettes are less harmful was a common reason for use among both less (<3 per month) and more (≥10 per month) frequent users of e-cigarettes (19.3% and 17.9%, respectively).This study, similar to previous studies [5,23,24], found that e-cigarette use was more prevalent among ever conventional cigarette smokers than among never conventional cigarette smokers. In addition, as the frequency and intensity of cigarette smoking increase, the percentage of frequent use of e-cigarette increases. This is different from the study conducted for U.S. adolescents, where the frequency e-cigarette use was not associated with the frequency and amount of conventional cigarette smoking [18]. This finding is plausible because even heavier adolescent smokers tend to be more nicotine-dependent [25] and would use e-cigarettes more heavily. It has been suggested that intensive or daily use of e-cigarettes, especially the tank type, may increase the chances of quitting smoking among adults [14,15]; however, no cessation aid, including e-cigarettes, has proved to be effective among adolescents [26]. Moreover, considering the potential adverse effect of nicotine on adolescent brain development [27], heavier e-cigarette use would likely cause more potential harm than benefit among adolescents.In our study, most ever e-cigarette users were infrequent users. Even among daily smokers, more than 50% used e-cigarettes for <3 per month. This suggested that many adolescent e-cigarette users could be experimenters [17]. However, 9.5% of never smokers were frequent e-cigarette users (≥10 per month), and 3.3% were daily e-cigarette users. This means that non-smoking adolescents may be using e-cigarettes as new nicotine supplementary devices. This could be a serious issue because nicotine may delay brain development, causing problems with cognition and emotional regulation [27,28], and e-cigarette use can be a gateway to combustible tobacco use and other substance use [8,9,10]. Also a recent study reported that the frequent e-cigarette vaping increased a risk of frequent and heavy conventional cigarette smoking 6 months later [29].Overall, the most common reason for e-cigarette use was curiosity, followed by the belief that they are less harmful than conventional cigarettes, the desire to quit smoking, and the capacity for indoor use. The reasons for use differed according to the frequency of e-cigarette use. As expected, for frequent e-cigarette users, common reasons for use were the desire to reduce/quit smoking and the capacity for indoor use. This is worrisome because adolescents who use e-cigarettes as a method to avoid an indoor smoke-free policy may lose out on the chance of quitting smoking, which can interrupt a smoke-free policy. For infrequent e-cigarette users, curiosity, better taste, and good flavor comprised half of the reasons for e-cigarette use. This group initiates e-cigarette use for recreational purposes [30], and this can increase the probability of their becoming conventional cigarette smokers or other substance users [9].Reasons for using e-cigarettes can be important in public health implications because they influence the continued use of e-cigarettes, and nicotine use is not safe especially among adolescents [10]. A previous study showed that adolescents who first tried to use e-cigarettes because they were cheap, because they wanted nicotine, or because they wanted to quit smoking tended to continue using e-cigarettes [19]. Additionally, adults who tried e-cigarettes for goal-oriented reasons (e.g., to use e-cigarettes where smoking was not allowed or to quit) were more likely to report continued use [31,32]. In our study, many e-cigarette users replied that they used e-cigarettes for the cessation of smoking. Additionally, we found that goal-oriented use of e-cigarettes (e.g., for smoking cessation, for indoor use, or because they were less harmful, had a better taste, or had less tobacco smell) were more prevalent among frequent users of e-cigarettes (e.g., ≥3 per month). This can be a possible explanation for the association between reasons for use of e-cigarettes and the continued use of them. On the other hand, e-cigarettes can be used as a recreational device in terms of the purpose of e-cigarette uses, even in non-smoking adolescents [30]. Our findings show e-cigarettes were used as a smoking cessation aid among adolescent smokers and for recreational purposes even among non-smokers, although there is lack of evidence for their safety and efficacy in adolescents [26,27]. Therefore, we suggest the need for more education for adolescents about e-cigarettes and stricter control by the government.Our study has several limitations. First, we could not establish any causal relations between the frequency of e-cigarette use and smoking frequency or intensity because our analysis was cross-sectional and the directionality of our findings could not be determined. Second, self-reports of e-cigarette use and smoking may differ from the actual rates. This is particularly true for smoking behavior among women, which may have been under-reported. Third, we could not evaluate why participants did not use e-cigarettes, especially among conventional cigarette only users. However, when we compared conventional cigarette only users with dual users of conventional cigarettes and e-cigarettes, dual users showed a significantly higher frequency and intensity of conventional cigarette smoking (Supplementary Table S1). Finally, we did not adjust the confounding variables to evaluate the frequency of e-cigarette use according to smoking behaviors or the reasons for e-cigarette use. Despite these limitations, however, this study has the advantage of using a large amount of nationally representative data. Our results can help in understanding the behaviors of adolescent e-cigarette users and may be usefully adopted in a future campaign to prevent adolescents from using e-cigarettes. In conclusion, the frequency of e-cigarette use was positively associated with the frequency or intensity of conventional cigarette smoking. In terms of the reasons for e-cigarette use, curiosity was the most frequent reason among less frequent e-cigarette users, whereas the desire to quit smoking and the capacity for indoor use were more frequent reasons among frequent e-cigarette users. With the rapidly increasing prevalence of e-cigarette use among adolescents, this study can further our understanding of behaviors related to e-cigarette use and aid in the appropriate control of e-cigarettes. The following are available online at www.mdpi.com/1660-4601/14/3/305/s1. Table S1. Comparison between cigarette only users and dual users.Jung-Ah Lee wrote the original draft, reviewed the data, and performed statistical analysis; Sungkyu Lee wrote the original draft and critically reviewed the final manuscript; Hong-Jun Cho conceived and coordinated the study, reviewed the data, wrote the original draft, and critically reviewed the final manuscript. The authors declare no conflict of interest.Sociodemographic and smoking-related characteristics of participants (N = 68,043).SE: Standard error.Frequency of e-cigarette use according to smoking behaviors among ever e-cigarette users of Korean adolescents (N = 6656).* The percentage and confidence interval denotes column %. † The number in parenthesis denotes column %. SE: Standard error; CI: Confidence interval.Frequency of e-cigarette use according to reason for e-cigarette use among ever e-cigarette users of Korean adolescents (N = 6656).* The percentage and confidence interval denotes row %. † The number in parenthesis denotes row %. ‡ The number in parenthesis denotes column %. CI: Confidence interval.

Med-MDPI/ijerph_2/ijerph-14-03-00306.txt

@@ -0,0 +1,244 @@
+Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).Resuspension of sediment-borne microorganisms (including pathogens) into the water column could increase the health risk for those using river water for different purposes. In the present work, we (1) investigated the effect of sediment disturbance on microbial resuspension from riverbed sediments in laboratory flow-chambers and in the Apies River, Gauteng, South Africa; and (2) estimated flow conditions for sediment-borne microorganism entrainment/resuspension in the river. For mechanical disturbance, the top 2 cm of the sediment in flow-chambers was manually stirred. Simulating sudden discharge into the river, water (3 L) was poured within 30 s into the chambers at a 45° angle to the chamber width. In the field, sediment was disturbed by raking the riverbed and by cows crossing in the river. Water samples before and after sediment disturbance were analysed for Escherichia coli. Sediment disturbance caused an increase in water E. coli counts by up to 7.9–35.8 times original values. Using Shields criterion, river-flow of 0.15–0.69 m3/s could cause bed particle entrainment; while ~1.57–7.23 m3/s would cause resuspension. Thus, sediment disturbance in the Apies River would resuspend E. coli (and pathogens), with possible negative health implications for communities using such water. Therefore, monitoring surface water bodies should include microbial sediment quality.In recent years, the accumulation of microorganisms including pathogens in riverbed sediments and possible resuspension during human-induced activities or natural processes has received increased research attention worldwide [1,2,3,4,5]. Once in the aquatic environment, a number of processes could lead to the settling of microorganisms and their subsequent resuspension from the bed sediments. Such processes could include natural gravitation, hyporheic exchange with the waterbed, attachment to suspended particles and aquatic vegetation, and filtration within the bed sediments [6,7,8,9,10]. Attachment to sediment particles could be through weak reversible van der Waals forces or strong irreversible forces due to the secretion of extracellular polymeric substances (EPS) and other bacterial appendages [7,11]. The bacteria could also be found in inter-sediment spaces within the riverbed sediments [12,13]. Increased flow within a water catchment may lead to erosion of riverbed sediments [14,15,16]. As reviewed by Chang and Scotti [17], sediment resuspension from the riverbed is determined by the intensity of the river-flow above the bed. When the river-flow exerts stress on the riverbed above a certain critical value, bed materials in contact with the water begin a rolling or jumping motion along the bed surface (entrainment). When the stress exerted by the flowing river increases further above the entrainment conditions, this may lead to turbulence at the sediment–water boundary layer, resulting in bed particles being lifted away from the riverbed (resuspension) and transported as suspended load [17,18,19]. This movement results in the re-suspension of bacteria from the sediment layer [20,21,22,23,24]. Once resuspended, the microorganisms could exist in a free-living state [25,26] or could resettle through the processes mentioned earlier.Extreme storm events cause sediment resuspension due to increase in flow, while prolonged drought events may lead to discontinuity of water supply from the main distribution systems in many water-scarce countries. As a result, there is an increased use of alternative water bodies such as rivers (e.g., for cleaning, bathing, and drinking). The increased use of rivers may result in the disturbance of riverbed sediments and consequently increase the human health risk associated with exposure to resuspended enteric waterborne pathogens [27,28,29]. This resuspension could also increase the cost involved in treating such water for human use.South Africa is not excluded from the current tremendous world population growth. With a current population of over 54 million, the country has experienced an increase in annual population growth rate from 1.27% in 2002 to 1.58% in 2014 [30] with tremendous pressure being placed on resources and infrastructure. Rural communities have been the most affected by this increase as most of them are located in areas with limited or complete lack of potable water supply, adequate sanitation, and waste management facilities [31]. As in many developing countries, these communities resort to surface water (especially rivers) as main or alternative water sources despite their low microbial quality [32,33,34,35]. South Africa’s surface waters are also used for recreational activities, especially during the hot summer months. Recent studies have demonstrated the presence of higher concentrations of Escherichia coli in riverbed sediments compared to the water column [36]. Genes for pathogenic organisms like Salmonella spp., Shigella spp. and Vibrio cholerae have also been detected in riverbed sediments in South Africa, indicating the possible presence of these pathogenic organisms [37]. In a study conducted by Abia et al. [38], the authors demonstrated through laboratory experiments that these pathogens could survive in the sediments of the Apies River for up to 30 days. Despite this finding, current monitoring of the country’s water resources (as with many developing countries) for microbial quality does not take into consideration sediment quality [39].Although previous studies have demonstrated riverbed resuspension, this has mostly been in relation to recreational activities in developed countries. Sediment resuspension in rivers that are directly and extensively used for personal and household hygiene as well as drinking, has not been given much attention. In addition, most studies have employed complex approaches to demonstrate the process of resuspension of sediment-borne microorganisms [1,20,23,24]. As such, this paper reports on simple, but effective, experiments on hydraulic and mechanical disturbances of sediment in both flumes and at reach scale in a river. The aim of the study was to (1) investigate the possible impact of mechanical disturbance and increased river flow on the resuspension of E. coli from riverbed sediments of the Apies River, Gauteng, using laboratory and field experiments; and (2) to estimate the flow conditions necessary for entrainment and resuspension of particles from the riverbed. To our knowledge, this study is the first to assess the impact of sediment disturbance on the microbial quality of river water in South Africa. The results of this study will be a useful contribution and could be quite influential in promoting increased awareness of the microbial health hazard associated with riverbed sediments—a hazard that can come to occur with mobilisation by flood waves or disturbance by people or livestock. Such information could be of great significance to water governing bodies within South Africa and other Sub-Saharan countries, as disturbance of microbially-contaminated river sediments could have negative impacts on public health.This study was carried out in the Apies River that flows across the city of Pretoria in the Province of Gauteng, South Africa (Figure 1).The study site (Apies river) has previously been described [36,40]. Located in a catchment characterised by high erodibility, the river receives high sediment loads, especially during run-off in the wet season. Extensive industrialisation and increasing human settlements along the river has led to the destruction of the vegetation around the river and canalisation of most parts of the river within urban areas. The presence of weirs and a dam on the river’s course affects the flow of the river in some areas. However, other parts of the river, especially as it passes through the rural and agricultural areas, have been maintained in their natural state. These morphological alterations along the river, therefore, lead to erosion in narrow areas of high flow and deposition in broader areas with low flow. The river water is used by surrounding communities for irrigation, watering of cattle, household and personal hygiene, spiritual cleansing and fishing. In some informal settlements, the river is the only water source available for the inhabitants and at the same time serves as a point for human waste disposal.The experiments were performed in flow-chambers (length 42 cm, width 19 cm and height 13 cm) constructed in the laboratory using periplex glass (Figure 2).The design of the chambers was adapted from Shelton et al. [41]. The chambers were connected to aquarium pumps (maximum capacity of 600 L/h) that were submerged in 45-L plastic containers. Sediment cores (length 41.5 cm, width 18.5 cm and height 12.5 cm) were collected from three different locations in the Apies River and placed into the chambers and transported to the laboratory. These sites were selected due to the difference in their particle size distribution as previously described [42]. Site 1 (AP5) had a high clay content, Site 2 (AP6) had a high silt content while Site 3 (AP7) was mostly coarse sand. Once in the laboratory, river water collected from the same sites was added into the sediment chambers to an additional height of 5 cm. The 45-L plastic containers were also filled with river water from each site. The chambers were then inoculated by adding E. coli (ATCC 25922) to the overlaying water in the sediment chamber to a final concentration of ca. 107 CFU/mL. The water (and approximately the top 2-cm layer of sediments) in the sediment chamber was manually stirred using a plastic hockey stick (Thermo Fisher Scientific, Edenvale, South Africa) to allow equal distribution of the organisms in the water column and the top sediment layer. The chambers were then allowed to stand overnight at room temperature to permit settlement of the bacteria. On the sampling day, water was collected from each chamber prior to disturbing the sediments.In the laboratory flow-chambers, the top 2 cm of sediments was manually stirred using a sterile disposable plastic hockey stick, ensuring that the entire length and width of the chamber was stirred to allow for proper resuspension. A new hockey stick was used for each chamber and care was taken to avoid contamination from hands during the stirring. This was to ensure that the microbial count observed in the water column was from the experiment and not from external contamination. Following the sediment disturbance, water samples were collected once the plume reached the surface. This was time zero (0 min) sampling. Thereafter, samples were collected after 10, 30 and 60 min. The water samples were analysed for E. coli counts using the Colilert® 18/Quanti-tray® 2000 system (IDEXX Laboratories (Pty) Ltd., Johannesburg, South Africa) following the manufacturer’s instructions [43]. Experiments were carried out in triplicate.For the flow experiments, a metal tray was inclined at an approximate angle of 45° perpendicular to the flow direction in the chambers (Figure 2). A sudden surge was produced by pouring approximately 3 L of river water within 30 s on the tray, thus causing a sediment disturbance in the chambers as the water from the tray reached the bottom of the chamber. The aim of this experiment was to mimic the effect of the direct discharge of large volumes of treated water from wastewater treatment works into the river. The water column was then sampled and analysed in the same way as with the mechanical disturbance experiments. The increased flow experiments were duplicated.For both types of disturbance experiments, water turbidity was measured before and after sediment disturbance using a T100 portable turbidity meter (EUTECH Instruments, Aachen, Germany).The manual resuspension experiment was then conducted at two selected sites on the Apies River (AP6 and AP9; Figure 1). Site A6 was selected due to the ease of accessibility and the fact that it was used by some inhabitants of the area for irrigation and personal hygiene. This was to simulate an area where community members could enter the water for recreational activities, to bathe, do laundry or for religious rituals. Site AP9 was selected because it was a point at which cows always crossed the river. Site AP9 had similar sediment particle size distribution as Site 3 (AP7) used in the laboratory experiments. At site AP6, the sediment was manually disturbed by raking an approximate 1 m2 area of the riverbed using a garden rake as previously described [44]. Water samples were then collected about 2 m away, downstream from the raked area to observe whether the disturbed sediments were actually carried downstream by the water current. The experiment at site AP9 was carried out at a time when cows usually cross the river. Observations made during several field visits to this sampling site showed that the farmers always watered their cows between 8:30 and 9:30 a.m. daily. It was also observed that the river water became more turbid downstream from the crossing point. Thus, the sampling point was approximately 3 m away from the point at which the cows crossed. Due to the fast-flowing nature of the water in the field experiments compared to the laboratory experiments, samples were collected at 20-s intervals from the time the plume reached the sampling point (2 m and 3 m away from the point of sediment disturbance for AP6 and AP9, respectively). Initial samples were also collected prior to mechanical disturbance. Samples were transported to the laboratory at 4 °C on ice and analysed immediately using the Colilert reagent to enumerate E. coli. The turbidity of the water was measured on site before and after sediment disturbance using the same instrument as for the laboratory experiments. The field experiments were performed on a single day. No direct defaecation into the water by the cows was observed during the field experiment as Site AP9.Mechanical disturbance of the riverbed is not always necessary for the mobilization of sediment and associated bacteria. Natural flows above some threshold are sufficient to initiate movement and suspend the sediment. Conventional methods have been used to determine approximate threshold conditions for the six selected sites along the Apies River where limited information has been obtained. Sediment samples were sent to the South African Agricultural Research Council for particle size analysis. Results of the sand particle sizes at these sites have previously been published [42]. The channel widths, flow depths and near-bed flow velocities were measured on one day during the dry season using a digital water velocity flow meter, the Global Water Flow Probe, Model FP211 (Global Water Instrumentation, Xylem Inc., Dallas, TX, USA).Entrainment (sometimes referred to as incipient motion), implies the mobilisation or setting in motion of riverbed material [45]. The flow condition at which entrainment would occur at each of the sites was estimated using the Shields criterion. The critical shear stress was determined for the median grain size (d50) in the bed material using the Shields diagram as compiled by Vanoni [46]. The corresponding flow depth (D) was then calculated from the formulation for the bed shear stress (τ),
+          
+            (1)
+            
+              
+                
+                  τ
+                   
+                  =
+                   
+                  ρ
+                   
+                  g
+                   
+                  R
+                   
+                  S
+                
+              
+            
+          
+          in which ρ is the density of water, g is gravitational acceleration, R is the hydraulic radius and S is the energy gradient. For wide channels, R can be approximated by D and, assuming uniform flow, S is equal to the channel gradient (it is acknowledged that higher energy gradients would occur during natural unsteady, non-uniform flows). The average flow velocity (V) corresponding to this flow depth was calculated using the Manning equation,
+          
+            (2)
+            
+              
+                
+                  V
+                   
+                  =
+                   
+                  
+                    1
+                    n
+                  
+                   
+                  
+                    R
+                    
+                      
+                        2
+                        3
+                      
+                    
+                  
+                   
+                  
+                    S
+                    
+                      
+                        1
+                        2
+                      
+                    
+                  
+                
+              
+            
+          
+          in which n is an empirical coefficient, estimated as 0.020 for these conditions. The discharge, or flow rate, is the product of the average velocity and the cross-sectional flow area, defined by the flow depth and channel width.Once mobilised, the sediment can be transported along the bed or in suspension within the body of the flow. Suspension requires greater flow conditions than those required for entrainment, and were estimated using the criterion proposed by Nino et al. [18]. This criterion is an advance on earlier ones, such as those proposed by Bagnold [47] and van Rijn [48], being developed from van Rijn’s criterion but taking into consideration the hiding effect that sometimes occurs when smaller-sized particles are about to undergo entrainment from a rough bed [18]. The ratio of the shear velocity (u*) to the particle settling velocity (w) at the onset of suspension is determined as:
+          
+            (3)
+            
+              
+                
+                  
+                    
+                      
+                        u
+                        ∗
+                      
+                    
+                    w
+                  
+                   
+                  =
+                   
+                  21.2
+                   
+                  
+                    D
+                    ∗
+                    
+                      −
+                      1.8
+                    
+                  
+				   
+				   
+				   
+				  for
+				  1
+				  <
+				   
+				   D
+				   ∗
+				   
+				    <
+					9
+                
+              
+            
+          
+          
+            (4)
+            
+              
+                
+                  
+                    
+                      
+                        u
+                        ∗
+                      
+                    
+                    w
+                  
+                   
+                  =
+                   
+                  0.4
+				   
+				   
+				   
+				  for
+				  
+				   D
+				   ∗
+				   
+				   >
+				   9
+                
+              
+            
+          
+          with,
+          
+            (5)
+            
+              
+                
+                  
+                    u
+                    ∗
+                  
+                   
+                  =
+                   
+                  
+                    
+                      g
+                       
+                      D
+                       
+                      S
+                    
+                  
+                
+              
+            
+          
+          
+            (6)
+            
+              
+                
+                  
+                    D
+                    ∗
+                  
+                   
+                  =
+                   
+                  
+                    d
+                    s
+                  
+                   
+                  
+                    (
+                    
+                      
+                        
+                          
+                            (
+                            
+                              
+                                S
+                                s
+                              
+                               
+                              −
+                               
+                              1
+                            
+                            )
+                          
+                           
+                          g
+                        
+                        
+                          
+                            ν
+                            2
+                          
+                        
+                      
+                    
+                    )
+                  
+                  
+                     
+                    
+                      
+                        1
+                        3
+                      
+                    
+                  
+                
+              
+            
+          
+          where ds is the representative grain diameter, Ss is the specific gravity of the sediment and ν is the kinematic viscosity of water. D* = 9 corresponds to a sand size of 0.38 mm, which is smaller than all the maximum sand sizes measured at the river sites, so u*/w = 0.4 is the appropriate suspension condition. The settling velocity was estimated from the graph presented by Graf [49] for natural quartz grains. The flow depth corresponding to the suspension shear velocity was calculated from Equation (5) and the corresponding velocity and discharge then calculated as for the entrainment condition. The critical shear stress for suspension was calculated from the critical shear velocity using the formula:
+          
+            (7)
+            
+              
+                
+                  τ
+                   
+                  =
+                   
+                  ρ
+                   
+                  
+                    u
+                    ∗
+                    2
+                  
+                
+              
+            
+          Data analysis was performed using SPSS 20.0 (IBM Corporation, Armonk, NY, USA). Correlation between E. coli count and turbidity was investigated for each sediment type using the non-parametric Spearman’s rank correlation test. The Mann–Witney U test was used to compare the turbidity as well as the E. coli counts after resuspension between the various sediment types. A regression analysis was used to verify if a change in turbidity (independent variable) could explain a change in E. coli counts (dependent variable) in water following sediment disturbance. All tests were considered statistically significant at p < 0.05.Mechanical disturbance of the sediment in the laboratory experiment showed a corresponding increase in water E. coli count and water turbidity (Figure 3).The increase in E. coli in the water column following mechanical sediment disturbance ranged between 3.6 and 35.8 times higher than the initial concentration prior to sediment disturbance. The concentration of E. coli in the water column then gradually decreased over time to approximately the initial concentration prior to sediment disturbance. There was equally a decrease in turbidity and depending on the sediment type, the decrease was either faster or slower than the decrease in water E. coli counts.The results obtained with the disturbance of bottom sediments in the chambers by increased flow are shown in Figure 4.For the flow experiments, a 2.4- to 17.4-fold increase in E. coli count was observed in the water column following the induced water surge. The induced water surge also caused a peak in water turbidity as with the mechanical disturbance experiments. There was equally a strong correlation (p < 0.05) between the water E. coli counts and turbidity for all sediment types.The field experiments consisted of raking an approximate 1 m2 portion of the river bed and the crossing of cows in the river. Figure 5a,b show the results of the raking and cow-crossing experiments respectively. Both experiments resulted in an increase in E. coli counts and turbidity within the water column following sediment disturbance. In the field experiment, the increase in E. coli was 7.9 times and 6.5 times higher than the concentration before resuspension for the raking and the cow-crossing respectively. The increase in E. coli concentration after sediment disturbance was positively correlated to turbidity for both experiments.Results of the regression analysis between E. coli and turbidity during the laboratory and field experiments are presented in Table 1.A positive correlation was observed between E. coli and turbidity for sediments of AP6 and AP7 in the laboratory experiments and at Site AP9 in the field experiment.Due to the limited flow and river geometry data for the Apies River, the Shields criterion was used to estimate the flow conditions at which sediment particles could be entrained, and Nino et al.’s criterion [18] was used to estimate the conditions required for the sediment suspension. The measured channel widths, gradients, flow depths and near-bed flow velocities are shown in Table 2.The sand sizes ranged from 0.18 mm (AP6) and 0.52 mm (AP8). Of all the sites studied, Site AP9 was the shallowest (0.25 m; measured at the centre of the channel width, while Site AP6 was the deepest (0.78 m). In terms of the channel width, Site AP7 was the widest (23.5 m) while Site AP2 (11.5 m) was the narrowest of all the sites. Estimated flow conditions needed for entrainment and suspension of sediments from the riverbed in the Apies River are given in Table 3.Discharges required to just mobilise bed particles ranged between 0.004 m3/s (AP7) and 0.021 m3/s (AP8). Greater discharges ranging between 0.04 m3/s (AP6) and 7.23 m3/s (AP7) were needed to cause the particles to completely resuspend from the riverbed.Although several studies have been conducted on sediment resuspension within aquatic ecosystems, these have been in beach environments, mostly used for recreational activities in developed countries like the USA [50,51,52] and Portugal [53,54]. Studies reporting on sediment resuspension, especially in developing Sub-Saharan countries, are limited or not available. As per the latest United Nations’ Sustainable Development Goals, close to 700 million people (the majority of whom are in Sub-Saharan Africa) still lack access to safe portable drinking water. In these parts of the world, people still have to partially or entirely depend on surface water bodies for their daily water requirements (for drinking and other household uses). With the increased droughts experienced in many of these countries due to changing climates, the need to use these surface water bodies is expected to increase. The increased access to these water bodies could lead to increased waterborne disease outbreaks due to resuspension of microbial pathogens from sediments of such polluted water bodies. Therefore, the current study is of significance to water regulating bodies and scientists in such countries as it provides information on the need to include sediment monitoring when designing aquatic ecosystem management programmes. This would in turn help improve the quality of these aquatic environments and thus protect aquatic life in general and the health of human populations using these waters in particular.Once in the aquatic environment, microorganisms are exposed to several factors that can affect their ability to survive in that environment. Of these, sedimentation has been recognised as an important factor [55,56]. Although several processes have been identified as contributing to the settling of bacteria into bed sediments, attachment to suspended sediments enhances the transport of these microorganisms from the water column into the bed sediments [57,58,59]. In the sediments, the microorganisms are protected from multiple stressors like UV light and predation; and are provided with higher nutrient concentrations [24,60,61,62]. In a recent study conducted in New Zealand, Devane et al. [63] concluded that disturbing bottom sediments in an urban river resulted in the resuspension of sediment-borne microorganisms. Similar findings were obtained in our study. The disturbance of sediments in the laboratory experiments resulted in an increase in the concentration of E. coli in the water column (Figure 3 and Figure 4).During the wet season, heavy rainfall may lead to erosion of riverbed sediments due to increased river flow [64]. In the laboratory flow experiments in our study, creating a sudden water surge in the flow-chambers induced a rise in E. coli concentration and turbidity in the water column. This observation ties in with findings of previous sediment resuspension works [23,44,65]. The correlation between the water turbidity and E. coli concentration suggests that in the absence of external sources, sediments could serve as a reservoir of high concentrations of microorganisms that could be resuspended, thus affecting the microbial quality of surface water bodies. The correlation between E. coli and turbidity also suggests that turbidity might be used as a rough, but potentially useful (local) indicator of the likelihood of microbial contamination. As such, visually observing the water for turbidity could be an immediate indication that people should avoid contact with the water. While communities may not have access to tools for testing the water for microbial quality, a visual assessment would therefore make turbidity an acceptable approach to continuously monitor the water, thus preventing possible exposure to microbial pollutants that may be present in the water.Mechanisms that lead to weak primary bacterial attachment include attachment through van der Waals forces, electrostatic and hydrophobic interactions, hydrodynamic forces and steric hindrance [66]. When these forces are electrostatic, the negative charge of the bacterial surface and that on the non-living surface lead to a repulsive force that makes the attachment easily reversible. On the other hand, hydrophobic forces appear to be stronger [67]. Thus, during riverbed disturbances, bacteria loosely attached through the weak forces, together with bacteria in the inter-grain spaces are released into the water column and can be detected. Also, it has been reported that bacteria will attach faster to sediments with smaller grain sizes [68]. This could explain the higher increase in E. coli concentration in the water column of chambers containing sediments from Site 1 (AP5) with higher clay content (14.5%) compared to Site 2 (AP6; 7.8%) and Site 3 (AP7; 2.6%) which had higher medium to coarse sand particles. The colloidal nature of the clay particles also allowed E. coli to stay in the water column for a longer period of time. The larger sizes and thus heavier particles of sediments from site AP6 and AP7 favoured a rapid settling of the particles together with the attached bacteria. This, however, indicates that in some places that are dominated by clay materials, smaller particles may settle slower than bound bacteria, thus resulting in low correlation between microbial counts and turbidity.The Apies River has several uses. Towards the northern end of the river before it joins the Pieenaars River, the Apies River is the main source of water for many farmers that use its waters for their animal farms. It is also used by households as a point for laundry usually involving the women and children getting into the water to do washing. The field study was conducted to mimic sediment disturbance during such water uses. Samples collected approximately 2 m from the sediments disturbance point revealed an increase in E. coli concentration in the water column. The increase in the E. coli concentration obtained in this study following sediments disturbance are similar to those reported by Orear and Dalman [44] who recorded a 7.5-fold increase in the concentration of E. coli in the water column following a 30-s raking of a 1-m2 plot. Recreational activities have been reported to cause resuspension of sediment-borne microorganisms [69]. It should be noted, however, that the degree of resuspension during recreational or other in-stream activities may also depend on the number of persons involved and the total area of the riverbed disturbed.During the dry winter season, most farmers walk their cows to the river so that they can drink directly from the river water. The crossing of the cows in the rivers results in the disturbance of the sediments in the riverbed which leads to the resuspension of sediment-borne organisms as shown in the results of the field experiments reported in this study. The effect of cows crossing a river on the concentration of E. coli in the water column has previously been reported [70,71]. However, contrary to the studies of Davies–Colley et al. [62] and of McDaniel and Soupir [63], no direct defaecation by the cows was observed during the current study. This therefore means that the increase in E. coli concentration and the turbidity recorded during the cows-crossing experiment was solely due to the resuspension of the sediments from the riverbed. The Apies River is also used for rituals by some religious groups. These religious groups believe that the water and sediments from the river are a source of protection and so use the river for spiritual cleansing and spiritual fortification of their homes (personal communication with a villager who came to collect water and sediments from the river). Such activities and other recreational activities could therefore lead to the resuspension of microorganisms from sediments and could represent a possible health risk to the users.Some studies have reported a strong positive correlation between E. coli concentrations and turbidity in water bodies [72,73]. The increase in E. coli concentration in the water column both in the laboratory and field sediment disturbance experiments in the current study, coincided with an increase in the water turbidity, demonstrated by a strong correlation between the two parameters. These results support the findings of Muirhead et al. [74] and Walters et al. [24]. In both their studies, they observed that following sediment disturbance, E. coli concentrations and turbidity (or total suspended solids) in the water column showed a positive correlation. In a study conducted in the city of Las Vegas, NV, USA, Huey and Meyer [73] reported that faecal indicator bacteria (E. coli and Enterococci spp.) were strongly correlated to turbidity. The authors concluded that turbidity could be used as an indicator of the microbial quality of water bodies. Although this has been shown, George et al. [75] reported that the relationship between E. coli and turbidity could be affected by the size of the watershed involved. In the current study, although sediment disturbance resulted in an increase in E. coli concentrations and turbidity values, this did not always result in a positive correlation between the two parameters (Table 1). While a strong E. coli/turbidity correlation was observed in some sediment types (AP6 and AP7) in the laboratory experiments and AP9 in the field experiments, E. coli/turbidity correlation was not statistically significant at Site AP5 in the laboratory and AP6 in the field. This suggests that sediment characteristics are important factors with regard to the correlation between E. coli and turbidity. In the sediments of Site AP5 that had the highest clay composition, attached E. coli could have settled faster while free colloidal clay particles remained in the water column for a longer period. This could have therefore led to the turbidity remaining higher even when the E. coli concentration had reduced to approximately initial values. On the other hand, in the sediments of AP7 (and AP9 in the field) that was made up of over 90% sand particles, the bound bacteria settled together with the heavier sand particles, resulting in the strong positive correlation observed between turbidity and E. coli. In a real river setting where flow is not controlled as in the laboratory experiments, other factors could interfere with the turbidity beyond the point of sediment disturbance, thus affecting the correlation [21]. Thus, Tornevi et al. [76] observed that the ease of measuring turbidity makes it a suitable first line parameter for estimating levels of microbial contamination. The authors, however, pointed out that because turbidity could be affected by organic and inorganic particles loads, it was necessary to also measure faecal indicator densities to complement the turbidity results.The flow regime of a river as well as its geometry and sediment properties affect the deposition and resuspension of sediments from the riverbed. However, sediment resuspension (and attached microorganisms) from a riverbed is a complex process and has been shown to occur even during normal or base flow conditions [69,70]. In the present study, all the sites for which estimation (entrainment and suspension conditions) was done were characterised by a high (72%) to a very high (86%) proportion of sandy material of various sizes [42] and thus it was assumed that the largest particle sizes (represented by the largest sieve size of 0.5 mm for site AP6 and 2 mm for all the other sites) needed to be suspended. Also, river measurements (width, depth and near-bed velocity) were only taken on one day during the dry season, i.e., on 6 June 2014. The threshold flow velocities, as calculated above, are depth-averaged values (approximately equal to the local velocity at 60% of the total depth below the water surface). The water velocity varies in an approximately logarithmic manner along a vertical profile, from almost zero at the riverbed to a maximum close to the surface [77]. The measured velocities in the current study were close to the bed, and therefore cannot be compared directly with the average threshold velocities calculated for entrainment and suspension. A more reliable comparison is between the actual bed shear stress (calculated from the measured flow depth and slope through Equation (1)) with the threshold values. These have been included in Table 2 and Table 3 and show that entrainment would be expected at all sites on the measurement day, with suspension at sites AP1, AP2 and AP6 and near suspension at sites AP7, AP8 and AP9. Considering that the measurements were taken in the dry season, mobilisation and resuspension of sediment and associated bacteria are likely to occur during normal flow conditions in the Apies River.Using simple laboratory and field-based experiments, the present study investigated the impact of riverbed sediment disturbances on the resuspension of Escherichia coli from riverbed sediment into the water column of the Apies River. The study also estimated flow conditions under which sediment particles within the Apies River could be resuspended. We conclude that sediment disturbance leads to resuspension of riverbed sediments and associated microorganisms and that this resuspension is likely to have a negative effect on the microbial quality of the overlying water. The strong correlation between E. coli and turbidity observed in this study suggests that turbidity could be a suitable proxy which communities may use as a first line of evidence of the possible poor quality of the river and therefore not use the water during such times. Under appropriate flow conditions, sediments in the Apies River would be resuspended and could represent a potential health risk to populations using the river directly without treatment for recreation and other purposes. Although the criteria used and hence the flow conditions obtained for entrainment and suspension in this study are empirical and assumptions have been made in their estimation, the results provide simple evidence of the effect of sediment resuspension on the microbial quality of the water in the Apies River. It is recommended that a complete profiling (e.g., the actual discharge, cross-sectional velocity, the river gradient) of the Apies River be undertaken. Obtaining measured values of these parameters may help in a better understanding of the sediment (and hence sediment-borne bacteria) dynamics within the catchment leading to more appropriate water quality monitoring strategies for the river.The authors would like to thank the Water Research Commission (WRC), South Africa (WRC Projects K5/2169 and K5/2147), the National Research Foundation (NRF) of South Africa and the Tshwane University of Technology (TUT) for funding. However, opinions expressed and conclusions arrived at are those of the authors and are not necessarily to be attributed to the NRF, the WRC or TUT. The funding supplied does not cover open access publication.Akebe Luther King Abia designed the study, performed the laboratory and field experiments and drafted the manuscript. Chris James performed the hydrological calculations and contributed in drafting the manuscript. Eunice Ubomba-Jaswa participated designing the experiments and in performing the field experiments. Maggy Ndombo Benteke Momba was the project leader and supervised the entire study. All authors read and approved the final manuscript.The authors declare no conflict of interest.Map of study sites (Source: Google Earth). DAS: wastewater treatment plant; AP1–AP4, AP8: sites included in entrainment/resuspension parameters estimation; AP5, AP6, AP7: sediment and water collection sites for laboratory experiments; AP6 (raking site) and AP9 (cow-crossing site): field experiments.Schematic representation of sediment chambers used for the laboratory resuspension experiments (Figure modified from Abia et al. [38]).Escherichia coli (solid lines) and turbidity (dashed lines) results in the water column for the mechanical sediment disturbance experiment in the laboratory for the three experimental rounds (TR1–TR3); BR: before resuspension.E. coli (solid lines) and turbidity (dashed lines) results in the water column for the sediment disturbance experiment through increased flow in the laboratory for the two experimental rounds (TR1–TR2).E. coli concentration (solid lines) and turbidity (dashed lines) of the water column before resuspension (BR), followed by (a) after raking and (b) after cow-crossing.Correlation coefficients and significance values for the regression analysis between E. coli and turbidity in the laboratory and field experiments.TR1–TR3: experimental rounds (replicates); *: p < 0.05; AP5, AP6, AP7: Sediment and water collection sites for laboratory experiments; AP6 (raking site) and AP9 (cow-crossing site): Field experiment.River parameters measured at the sampling sites on the Apies River.* Measured at the centre of the river cross-section. AP1, 2, 8: Additional river sites which were accessed by the inhabitants for bathing and other purposes.Estimated sediment mobilisation and suspension flow conditions for sampling sites on the Apies River.

Med-MDPI/ijerph_2/ijerph-14-03-00307.txt

@@ -0,0 +1 @@
+Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).Although the suicidal and self-harming behaviour of individuals is often associated with similar behaviours in people they know, little is known about the impact of perceived social norms on those behaviours. In a range of other behavioural domains (e.g., alcohol consumption, smoking, eating behaviours) perceived social norms have been found to strongly predict individuals’ engagement in those behaviours, although discrepancies often exist between perceived and reported norms. Interventions which align perceived norms more closely with reported norms have been effective in reducing damaging behaviours. The current study aimed to explore whether the Social Norms Approach is applicable to suicidal and self-harming behaviours in adolescents. Participants were 456 pupils from five Scottish high-schools (53% female, mean age = 14.98 years), who completed anonymous, cross-sectional surveys examining reported and perceived norms around suicidal and self-harming behaviour. Friedman’s ANOVA with post-hoc Wilcoxen signed-ranks tests indicated that proximal groups were perceived as less likely to engage in or be permissive of suicidal and self-harming behaviours than participants’ reported themselves, whilst distal groups tended towards being perceived as more likely to do so. Binary logistic regression analyses identified a number of perceived norms associated with reported norms, with close friends’ norms positively associated with all outcome variables. The Social Norms Approach may be applicable to suicidal and self-harming behaviour, but associations between perceived and reported norms and predictors of reported norms differ to those found in other behavioural domains. Theoretical and practical implications of the findings are considered.A number of social factors have been identified which may impact upon individuals’ risk of engaging in suicidal and self-harming behaviour (SSHB hereafter for brevity), including social support and connectedness [1], socioeconomic deprivation [2], and media reporting of suicide [3,4]. Clustering of SSHB has been repeatedly observed in young people [5,6,7], and in those with mental health issues [6,7,8], suggesting these groups may be particularly susceptible to a contagion-like spread of SSHB. Self-harming behaviours are more prevalent in young people than in older groups [9,10], reportedly occurring in children as young as 5 [11] with a peak between 14 and 18 years [12]. Given that self-harm represents a major risk factor for future suicide [13,14,15], young people represent a particularly high-risk group for SSHB, and social factors contributing to that risk may be especially significant within this group.A systematic review of the literature suggests that the SSHB of family, friends, peers and others in children’s and adolescents’ social networks is strongly associated with their own risk of engaging in SSHB [16]. However, research has tended to assume individuals’ accurate knowledge of the behaviour of others, often based on factors such as attendance at the same school or membership of the same family. The potential for inaccuracy of perceptions or the impact of more general beliefs about others’ typical engagement in SSHB has generally not been considered.Perceived social norms surrounding a given behaviour are evidenced to influence an individual’s own engagement in and permissiveness towards that behaviour. Both the normative rates of engagement in a particular behaviour (descriptive norms) and the normative attitude or level of permissiveness towards a given behaviour (injunctive norms) are associated with self-reported behaviour and attitudes. Despite this, individuals tend to believe that others behave in more damaging or negative ways than reported norms would suggest [17,18,19]. Perceptions of group norms surrounding a particular behaviour consistently show strong positive associations with self-reports of engagement in and permissiveness towards that behaviour, with those with higher reported norms perceiving norms to be particularly high [20]. These effects have been identified in a broad a range of behavioural domains, including alcohol consumption [17,18,19], gambling [21], risky sex [22], seatbelt use [23], substance use [24], and snacking behaviour [25]. A relatively small number of studies have found no direct relationship between perceived norms and individuals’ own behaviour, but where this is the case, samples may be small and the reference groups unspecific and distal [26], or norms do predict behaviour but only under particular conditions [27]. There have been wider criticisms of the methodological basis of both social norms research and the Social Norms Approach [28], although these criticisms have themselves been challenged [29]. A recent systematic review of the Social Norms Approach concluded that there may be some issues with how social norms campaigns are implemented, but that there is sufficient evidence to support the central premise of the approach that people do overestimate harmful behaviours and attitudes in their peers [30].Interventions have been developed using the Social Norms Approachin a number of behavioural domains. Social norms surveys are used to measure a target group’s reported behaviour and attitudes (reported norms), and their perceptions of the average behaviour and attitudes of others (perceived norms), and where perceived norms differ (in a negative/unhealthy direction) from reported, reported norms are fed back to the target group, with the aim of aligning perceived norms more closely with reported norms; thereby reducing damaging behaviours or increasing positive ones. The Social Norms Approach has effectively reduced a number of damaging behaviours/attitudes, including alcohol consumption [31], substance use [32], drink-driving [33], bullying [34], and rape-supportive attitudes [35], and increased a number of positive behaviours/attitudes, such as the reporting of bullying [34], sun-protection behaviours [36], HIV-prevention behaviours [37], environmental conservation behaviours [38], and positive sexual attitudes [39]. Again, whilst social norms interventions can be very effective in changing behaviour, a small number of campaigns have failed to show any effect of altering perceptions on subsequent behaviour [40].The current study aimed to extend the literature around social norms to include SSHB. In addition to associations consistently found between SSHB and many of the behaviours to which the Social Norms Approach has already been applied (including alcohol consumption and substance use; e.g., [41,42]), previous work by the current authors (submitted manuscript) suggests that undergraduate students’ SSHB is predicted by their normative perceptions of SSHB. Given that research has identified relationships between children and adolescents’ own SSBH and that of people they know [16], it was considered likely that adolescents’ engagement in SSHB may also be related to their normative perceptions of those behaviours—regardless of the accuracy of those perceptions. Further, young people are particularly susceptible to influence from their social environment [43,44], and evidence suggests that those who are most prone to influence may be at an already heightened risk of engaging in dangerous or risky behaviours [45]. If heightened perceptions of social norms around SSHB are positively associated with adolescents’ own engagement in SSHB, this could have dangerous implications in terms of increasing their risk of harm. The current study therefore aimed to determine whether the Social Norms Approach might be applicable to SSHB in adolescents, through identifying whether discrepancies exist between perceived and reported norms, and whether perceived norms predict adolescents’ reported norms. It was hypothesised that discrepancies would be observed between perceived and reported norms around SSHB, and that perceptions of normative behaviour and attitudes around SSHB would be associated with individuals’ own reported behaviour and attitudes. Should this be the case, it might provide professionals working with young people (e.g., in schools or other socially-oriented settings) with opportunities for intervention and prevention, via the Social Norms Approach.Participants were 456 pupils from five mainstream Scottish high-schools, spanning four different Local Education Authorities (LEAs). Two schools were situated in large urban areas and three in other urban areas (based on the Scottish Government 6 Fold Urban Rural Classification), and they ranged from 5 to 10 in Scottish Index of Multiple Deprivation deciles (where 1 = most deprived, 10 = least deprived). 52.63% of participants were female, with ages ranging 11–17 years (mean = 14.98, SD = 1.09). One participant did not report their gender, and two did not report their age. More detailed breakdown of these characteristics can be found in Supplemental Table S1.The self-report survey instrument was developed through consultation of the literature and previous social norms surveys [21,24]. Surveys were paper-based and responses were obtained using a multiple choice, tick-box-style format. As with previous social norms research, questions were tailored specifically to suit the population and the behaviours under investigation. Participants were asked about their thoughts of self-harm, acts of self-harm, thoughts of suicide, and suicide attempts (descriptive norms), and their permissiveness towards self-harm and suicide attempts (injunctive norms). Self-harm was defined as “deliberately taking an overdose (e.g., pills or other medication), or trying to harm yourself in some other way (such as cutting yourself)”, as per the CASE study [46]. Suicidal ideation was defined as “thinking about attempting to end your life (i.e., deliberately die by suicide)”, and suicide attempt was defined as “attempting to end your life (i.e., deliberately die by suicide)”. Their responses to these items represented the reported norms. Each reported norm question was paired with a question examining participants’ normative perceptions of that same behaviour or attitude in their close friends, their parents, their extended family, high-school pupils the same age and sex as them, pupils at their high-school, high-school pupils in general, people their age in general and people in general. Their responses to these questions represented the perceived norms.Descriptive norms: Self-reported descriptive norms questions were posed in the format “Have you ever (thought about harming yourself/harmed yourself/thought about ending your life/attempted to end your life)?”, and matching items regarding participants’ perceptions of norms were posed in the format “Do you think the following people ever (think about harming themselves/harm themselves/think about ending their lives/attempt to end their lives)?”. Responses to both were given on a 5-point scale: “never”, “have done occasionally in the past”, “have done regularly in the past”, “do so occasionally” and “do so regularly”.Injunctive norms: Self-reported injunctive norms questions were posed in the format “Which statement about (harming oneself/attempting to end one’s life) do you feel best represents your own attitude?”, and matching normative perception questions were posed in the format “Which statement about (harming oneself/attempting to end one’s life) do you feel best represents the attitudes of the following people?”. Responses to each were given on a 3-point scale: “completely wrong”, “understandable under certain circumstances” or “completely OK”. These SSHB questions were embedded within the context of a larger social norms study investigating risky health behaviours in general (including such behaviours as smoking, seatbelt use and help-seeking), the findings of which are reported elsewhere. This larger study also had the advantage of avoiding the over-emphasis of SSHB in a vulnerable population.Full ethical approval for the study was obtained from the University of Strathclyde Ethics Committee (project ID code: UEC13/26; approval date: 5 June 2013). All 32 LEAs in Scotland were contacted in the first instance and provided with study information and permission was requested to contact schools directly. Head-teachers were then contacted, and arrangements for collection of parental assent (for those under 16 years of age) were made with the schools. Those under 16 s for whom parental assent was obtained, and over 16 s (for whom parental assent was not required by the institutional ethics committee), were then invited to participate. All pupils were given full study information and asked to provide written informed consent. They were then invited to complete the survey privately and anonymously, in a classroom setting during an ordinary lesson. Participation took 20–40 min, and surveys were collected in sealed envelopes by the researcher.Reported norms (self-reports of adolescents’ own behaviours and attitudes) were compared with perceived norms (perceptions of others’ behaviour and attitudes) for each of the reference groups to determine if there were significant differences between the behaviour/attitude of respondents and what they perceived to be the norm. In keeping with the terminology used in previous social norms research, this refers to a self-other discrepancy. Given that data was ordinal and had no clear numerical interpretation, and that SSHB does not tend to be normally distributed, non-parametric analyses were performed. Friedman’s ANOVA was selected for this analysis as it detects differences between groups across multiple tests (using analysis of the variance of ranks) where data is ordinal. In order to determine where any differences were found, post-hoc Wilcoxen signed-ranks tests (with Bonferonni corrections) were used.Whether or not there were any associations between perceived and reported norms was also examined. Due to the relative rarity with which some behaviours were reported (e.g., “I attempt suicide regularly/often”), descriptive norms responses were re-coded into a binary variable denoting ever having engaged in a given behaviour (1) and never having engaged in that behaviour (0), and injunctive norms responses were recoded into believing a behaviour to be completely wrong (0), and believing it be OK/understandable at least in some circumstances (1). Binary logistic regression was then used to determine predictors of reported norms, odds ratios and 95% confidence intervals. Separate regressions were run for each of the four descriptive norms and the two injunctive norms. For all self-harm-related outcome variables, predictors entered into the regression were age, sex, perceptions of all others’ thoughts of self-harm, self-harm and permissiveness of self-harm. For all suicide-related outcome variables, age, sex, perceptions of all others’ thoughts of suicide, suicide attempts, and permissiveness of suicide were entered into the regression. Collinearity diagnostics revealed some multicollinearity in three of the models, with tolerance levels <0.1 and VIF >10 for models 1, 2, and 4 (thoughts of self-harm, self-harm and permissiveness of self-harm). It is advised that multicollinearity should be acknowledged and potential bias considered, but all variables should be maintained in the model in order to avoid further complications associated with removing them [47]. All variables were therefore preserved in their respective models.Table 1 illustrates the reported norms of the sample. Figures refer to the number of participants reporting ever having engaged in a particular behaviour, or believing that a particular behaviour is ever “OK”.All six Friedman’s ANOVAs indicated differences between reported and perceived norms. Table 2 presents the results, with significantly different reference groups identified through post-hoc Wilcoxen signed-ranks tests. Full results (including non-significant differences) can be found in Supplemental Table S2.Results of the six binary logistic regressions (Models 1–6) are presented in Table 3, with model statistics. One model estimates predictors of each reported norm. Only variables identified as significantly associated with outcome variables are reported, but full results with all model variables can be found in Supplemental Table S3.Reported thoughts of self-harm were significantly different to perceived thoughts of self-harm. Post-hoc tests indicated parents and extended families were perceived as less likely to have thoughts of self-harm than participants reported, whilst all other groups were perceived as more likely. 27.0%–44.8% of the variance in reported thoughts of self-harm was explained by Model 1. Those who believed their friends had thoughts of self-harm or engaged in self-harm were around one and a half times more likely to report having thoughts of self-harm themselves, whilst this increased to almost six times more likely for perceptions that family members engaged in self-harm. The belief that pupils from the same school have thoughts of self-harm was associated with a decrease by approximately half in reported thoughts of self-harm. Inspection of collinearity diagnostics indicated that there may be some multicollinearity, so the model should be interpreted with caution.For self-harm, reported norms also differed from perceived norms. Parents and extended families were perceived as less likely to engage in self-harm than participants reported. All other groups were perceived as more likely. Model 2 accounted for 24.2%–46.8% of the variance in reported self-harm. Females were almost four times more likely than males to report engaging in self-harm. Those who believed their friends had thoughts of self-harm or that high-school pupils the same sex as them had thoughts of self-harm were around twice as likely to report self-harming, but believing that pupils at the same school had thoughts of self-harm was negatively associated with reported self-harm. As the variables in the current model are identical to those in the previous, multicollinearity was again indicated.Reported thoughts of suicide significantly differed from perceptions of others’ thoughts of suicide. Parents and extended families were perceived as less likely to have thoughts of suicide than participants’ own reported thoughts. High-school pupils of the same age and sex, high-school pupils attending the same school, high-school pupils in general, and people in general were all perceived as more likely. 19.4%–37.8% of the variance in reported thoughts of self-harm was explained by Model 3. Those who believed their close friends had thoughts of suicide were over three times more likely to report thoughts of suicide, whilst believing friends had made suicide attempts was associated with a decrease in own thoughts of suicide.Significant differences were again found between reported and perceived norms for suicide attempts. High-school pupils of the same age and sex, high-school pupils attending the same school, high-school pupils in general, and people in general were all perceived as more likely than reported norms, to make suicide attempts. The only significant predictor in Model 4 was perceptions of close friends’ permissiveness towards suicide attempts, which was associated with an almost thirty times increased likelihood of reporting suicide attempts, but the overall model was not significant.Reported norms for permissiveness of self-harm differed significantly from perceptions of permissiveness of self-harm. Close friends, parents and extended families were all perceived as less likely to be permissive of self-harm than reported norms. Model 5 explained 38.3%–51.2% of the variance in reported permissiveness. Females were more than twice as likely as males to report permissive attitudes. Believing one’s friends hold permissive attitudes was associated with a more than four times increased likelihood of reporting permissive attitudes, and those who believed people in general held permissive attitudes were more than three times more likely to report permissive attitudes. Inspection of collinearity diagnostics indicated that again there may be some multicollinearity between variables, so the model should be interpreted with caution.Finally, it was also found that significant differences existed between reported and perceived permissiveness of suicide attempts. Close friends, parents and extended families were all perceived as less likely to be permissive of suicide attempts than participants reported themselves. 46.5%–62.5% of the variance in reported permissiveness was accounted for by Model 6. Females were about three times as likely as males to report permissive attitudes. Those who believed pupils the same age and sex as them had thoughts of suicide were around one and a half times more likely to hold permissive attitudes, while this increased to just over six times for those who believed their friends held permissive attitudes, and almost thirty times for those who believed their family held permissive attitudes.The aim of the current study was to explore the social norms of SSHB in adolescents and to examine whether discrepancies exist between perceived and reported norms for SSHB. The study also aimed to identify whether adolescents’ own SSHB could be predicted by their normative perceptions of a number of reference groups. Significant self-other discrepancies were indeed observed for all four behavioural outcome variables (thoughts of self-harm, self-harm, thoughts of suicide and suicide attempts) and for both attitudinal outcome variables (permissiveness towards self-harm and towards suicide attempts) but discrepancies were only significant for certain reference groups and were not always in the predicted direction.Approximately 10% of males, and almost 25% of females reported having had thoughts about self-harm, which is comparable to previous European studies [46]. Similarly, our 4% of males reporting having engaged in self-harm is in line with previous studies, and the 20% of females in our sample reporting having engaged in self-harm is comparable to the 17% reported in England and Australia in the CASE study [46] (though it is somewhat higher than rates reported in other Scottish studies, such as [48]). Our sample reported rates of suicidal ideation and attempts at approximately 13% and 4% (respectively), which are also comparable to previous studies in Europe and the US [49,50].There was an overall tendency for participants to believe that proximal groups were less likely, and distal groups more likely, to engage in SSHB than self-reported norms. Parents and family members were perceived as less likely than reported norms to think about self-harm, engage in self-harm, or think about suicide, whilst more distal groups were perceived as more likely to do so. Interestingly, close friends were perceived more similarly to distal groups; with participants reporting that close friends were more likely to engage in SSHB than reported norms.That proximal groups were perceived as less likely to engage in SSHB than reported norms is contrary to previous social norms research, but perceived norms for distal groups followed a similar pattern to that generally observed in social norms research (i.e., that others are perceived as more likely than oneself to engage in negative or damaging behaviours). It is possible that adolescents simply have access to more accurate knowledge about the behaviour of those close to them, and that having never seen any evidence of SSHB amongst their friends and family, they correctly surmise that they do not engage in SSHB. It has previously been shown that self-other discrepancies are smaller for proximal groups than for distal groups, for this reason [51]. However, as previous social norms studies have still found discrepancies for proximal groups (albeit smaller ones), this explanation may not be sufficient. An alternative is that our differing findings are indicative of inherent differences in the way that SSHB is perceived in comparison with previously studied behaviours (e.g., alcohol consumption) in terms of rightness/wrongness, positivity/negativity, personal responsibility, and their impact on others. For example, links with psychological distress and mental ill-health may mean that as adolescents do not believe many of their loved ones to be mentally unwell, SSHB is perceived as unlikely to occur in proximal groups. They know that mental illness does exist, so it is perceived as something that must occur elsewhere (i.e., in more distal groups) by default. Similarly, adolescents may believe that due to the impact that others’ SSHB would have on themselves (e.g., distress caused by a loved one suffering) they would be aware if those close to them had harmed themselves or attempted suicide, whereas they perhaps would not be so personally affected if loved ones engaged in more traditionally studied behaviours (such as alcohol consumption, for example). As they have not been impacted in this way, they may assume it must not have happened. These potential effects of perceptions around SSHB in comparison to other behaviours would suggest that in order to interpret the findings of any social norms research appropriately, a better understanding is required of the ways in which participants perceive and understand the behaviour of interest. The discrepancy in the way that different groups were perceived in the current study also argues for the inclusion of a range of diverse reference groups in future social norms research.Suicide attempts were perceived slightly differently to the other behaviours, with all groups (except parents) perceived as more likely to make suicide attempts than reported norms—although these discrepancies were only significant for distal group norms. Whilst previous social norms research has examined a number of risky and damaging behaviours, the ultimate aim of those behaviours tends not to be inflicting harm/death, as it (usually) is with SSHB, so it is perhaps unsurprising that our findings in this regard were so different to those of previous studies. It is interesting that parents were perceived as different to all other groups and uniquely immune to SSHB. This may be accounted for by the fact that the number of parents any individual has tends to be a relatively small number in comparison to larger reference groups (e.g., extended family, people your age), necessarily reducing the likelihood of any behaviour occurring within that limited group. Alternatively, it may represent a form of optimism bias, whereby the thought of a parent being hurt represents the kind of negative event that individuals believe is unlikely to happen to them [52]. Differential effects on an individual of others dying by suicide, compared to their engagement in other SSHBs, has been noted elsewhere [16].For both attitudinal outcome variables (permissiveness of self-harm and of suicide attempts), there were significant discrepancies between perceived and reported norms, as predicted. However, again these discrepancies were not in the expected direction, with proximal groups perceived as less likely than reported norms to be permissive of both self-harm and suicide attempts. Distal groups tended to be perceived as more permissive of self-harm than reported norms—consistent with previous social norms literature—but not significantly. As previously mentioned, findings may demonstrate adolescents’ enhanced knowledge of the attitudes of those close to them, in comparison to more distal, unfamiliar groups. However, the differences in the direction of discrepancies between the current findings and previous social norms findings suggest that SSHB is perceived somewhat differently to behaviours studied previously, perhaps in terms of their ethical or moral status, their links with mental illness, and their effect on others (particularly in the case of suicide).The finding that distal groups tended towards being perceived as more likely to be permissive of self-harm may be indicative of a belief that permissiveness is a negative feature, reflecting similar results to those found in previous social norms research (i.e., individuals believe that others behave in worse ways than they do themselves). Evidence suggests that whilst attitudes towards SSHB vary across individuals [53,54], many people hold very negative and “blaming” views [55,56,57,58], which may shape their perceptions of others’ attitudes. The distinction between proximal and distal groups in this regard may reflect in-group/out-group biases [59,60], with those deemed part of participants’ in-group (e.g., friends, family) perceived as behaving differently (better) to those in out-groups (e.g., people not known to the individual).Significant predictor models were generated for five out of the six outcome variables (thoughts of self-harm, self-harm, thoughts of suicide, permissiveness of self-harm and permissiveness of suicide attempts), and each model had a number of significant predictors. The models accounted for a substantial proportion of the variance in the independent variables (e.g., as much as 62.5% in the case of permissiveness of suicide attempts). Again, associations with perceived norms were not always in the predicted direction, based on previous social norms research (i.e., that greater/more permissive perceived norms will predict greater/more permissive reported norms). As already described, gender predicted some of the outcome variables, but age was not associated with any, and descriptive norms tended to more often predict self-reported norms than did injunctive norms. Generally speaking, perceived proximal group norms were more often positively than negatively associated with reported norms, whilst distal group norms were roughly as equally likely to be positively associated with reported norms as negatively. Aside from these features, there was no clear, discernible pattern in the variables associated with outcomes.One clear pattern that did emerge was that the perceived norms of close friends were particularly important in predicting reported norms, with at least one close friends-related norm associated with each of the six outcome variables. The wide-reaching impact of peers on self-harming behaviour in adolescents has been well-documented [61], and our findings appear to support this; suggesting that perceived norms pertaining to close friends may be particularly influential in increasing adolescents’ engagement in and permissiveness towards SSHB. Stronger associations between reported norms and the perceived norms of proximal groups (relative to distal groups) have been shown in previous social norms research [62,63]. However, the need for peer approval during adolescence has been highlighted previously [64], as have socialisation and modelling effects [6], so the mechanisms of peer influence over SSHB may be multiple [61]. A review of peer influences on alcohol consumption [17] identified 3 specific processes through which influence occurs; overt encouragement, modelling, and perceived social norms. The latter 2 in particular may be relevant in this context; although it is not impossible that all 3 play a part in the case of close friendships.The finding that perceived descriptive norms more often predicted reported norms than did injunctive norms (with a ratio of 2:1) is contrary to patterns shown in previous social norms research, which generally finds injunctive norms to be better predictors of reported norms [18]. There are several possible reasons for this. Firstly, given that a number of relationships increase in intimacy and perceived significance as a function of age [65,66], the mutual holding of shared values deemed important in the maintenance of successful relationships in older individuals [67,68] may be considered less so in adolescents, such that the (perceived) beliefs and attitudes of those around them are less influential in shaping their own. Secondly, given that adolescents may be particularly prone to egocentrism [69], making inferences about others’ attitudes and beliefs may simply not be of interest to them, whilst actual behaviours are more visible and salient, and require less outward-focused thought. Finally, a relative lack of knowledge of other people on account of adolescents’ age and inexperience may render them unfamiliar with other people’s thoughts and attitudes, such that they are less likely to use them as a source of information or guidance. Despite this however, some previous research exists which supports the current findings that descriptive norms are better predictors of reported norms than are injunctive norms [62].The current findings suggest that as found in previous social norms research, there were discrepancies between perceived and reported norms for adolescent SSHB, and certain perceived norms predicted individuals’ own behaviour and attitudes. As such, social norms interventions which feed back normative information relating to SSHB in order to align perceived norms more closely with reported norms, thereby reducing any related increase in individuals own behaviour, may be applicable within this population. As the impact of perceived on reported norms differs slightly from previous social norms research in the current study, any intervention would need to be carefully designed, taking into consideration the specific behaviour or attitude in question, and the nature of the reference groups used. As descriptive norms appeared particularly salient to this age group, an appropriate social norms intervention might be one which feeds back information regarding the relatively low reported incidence of SSHB in this sample, framing messages in terms of “people your age/sex” or “people in general”. As friends’ norms also appeared particularly salient, these messages could be shaped to infer that their friends constitute part of this group, and therefore have similarly low rates of engagement in SSHB.More generally, the observation that certain perceived norms—particularly those of friends—were positively associated with individuals’ own reported behaviour and attitudes, has important practical implications for families and professionals caring for or working with adolescents, and suggests that perception may play a significant part in eliciting an imitative or contagion-like spread of SSHB. It is important that measures are taken to ensure that the social environments in which adolescents function are healthy, supportive and open, and that efforts are made to prevent the development of a culture in which SSHB is perceived as ubiquitous or normalised. It is also vital that young people are provided with psychoeducation on issues around psychological distress, alternative coping strategies, and the importance of seeking support.One vital consideration when designing interventions based on the Social Norms Approach—or indeed social influence more generally—is the importance of striking a balance between the avoidance of “normalising” particular behaviours, and the exacerbation of stigma and/or feelings of isolation in those who engage in them. This is particularly important with regard to SSHB, given that engagement therein may be increased by feelings of social isolation [70,71,72,73] and experiences of stigma [74,75]. The thoughtful and sensitive design of interventions conveying supportive, non-judgemental messages is thus imperative in order to avoid inadvertent increases in harm. That social norms interventions are not always effective in changing behaviour should also be considered [26,40], and additional support and preventative measures should continue to be employed to help protect young people from harm.The study is subject to some limitations, including challenges associated with appropriate wording of the novel survey instrument, and issues around defining meaningful reference groups (perhaps accounting for some of the multicollinearity observed). The survey was not piloted in an adolescent sample (only on undergraduate students sampled in the authors’ previous work), so there may have been issues around comprehensibility and relevance to the sample, although attempts were made to adapt the wording and reference groups for increased age-appropriateness. Despite efforts to recruit as widely as possible, only five schools agreed to participate, and although they were relatively varied in their socioeconomic characteristics, they were all from urban or semi-urban settings, so representativeness cannot be guaranteed. Finally, there are limits to the inferences that can be made from the current findings given that the data was cross-sectional (so assumptions about the causal direction of effects cannot be confirmed) and that some potential multicollinearity was indicated for some of the models. Future research should aim to replicate the current findings using larger samples (particularly including schools in rural settings), longitudinal methods (in order to address causality), and carefully defined reference groups.The current findings differed quite substantially from those obtained in previous social norms research. Discrepancies between perceived and reported norms appear consistent, but different reference groups appear to be perceived differently and vary in their predictive power over reported norms. Although cross-sectional in design (and therefore unable to confirm causal direction), a number of perceived social norms appear predictive of adolescents’ behaviour and attitudes in this domain, and the regression models generated appear to account for substantial proportions of the variance in reported norms. The fact that perceived norms are at all predictive of reported norms argues for the relevance of the Social Norms Approach within the domain of SSHB, and the potential for the development of interventions based on the approach for the reduction of SSHB. Descriptive norms appear to be particularly important to this population, such that interventions which aim to utilise normative information to reduce SSHB in this group may do well to employ a particular focus on behaviour, relative to attitudes. A number of issues remain unclear, and further research is required to help explain in more detail some of the unexpected findings of the current study, and to continue to explore whether and how the Social Norms Approach might be applicable to the reduction of SSHB in young people.The following are available online at www.mdpi.com/1660-4601/14/3/307/s1, Table S1: Characteristics of schools and participants from each school; Table S2: Full results of Friedman’s ANOVA with post-hoc Wilcoxen signed-ranks to determine difference between perceived and reported norms; Table S3: Binary logistic regression analyses of all variables tested for associations with reported norms.This study was funded by a University of Strathclyde Ph.D. studentship. Open access publishing costs covered by the University of Stirling Article Processing Charge Fund.Jody Quigley conceived of and designed the study, conducted the study, analyzed the data, and wrote the paper. Susan Rasmussen provided advice and guidance, contributed to study design, and commented on drafts of the paper. John McAlaney provided advice and guidance, contributed to study design and data analysis, and commented on drafts of the paper.The authors declare no conflict of interest.Descriptive data relating to reported norms.Results of Friedman’s ANOVA with post-hoc Wilcoxen signed-ranks to determine difference between perceived and reported norms.1 SH = self-harm; 2 SA = suicide attempt; 3 Perm = permissiveness (injunctive norms); 4 S = self (reported norm); 5 O = other (perceived norm).Binary logistic regression analyses of variables significantly associated with reported norms.1 SH = self-harm; 2 SA = suicide attempt; 3 Perm = permissiveness (injunctive norms).

Med-MDPI/ijerph_2/ijerph-14-03-00308.txt

@@ -0,0 +1 @@
+These authors contributed equally to this work.Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).Background: With the rapid increase in the incidence and mortality of lung cancer, a growing number of lung cancer patients and their families are faced with a tremendous economic burden because of the high cost of treatment in China. This study was conducted to estimate the economic burden and patient responsibility of lung cancer patients and the impact of this burden on family income. Methods: This study uses data from a retrospective questionnaire survey conducted in 10 communities in urban China and includes 195 surviving lung cancer patients diagnosed over the previous five years. The calculation of direct economic burden included both direct medical and direct nonmedical costs. Indirect costs were calculated using the human capital approach, which measures the productivity lost for both patients and family caregivers. The price index was applied for the cost calculation. Results: The average economic burden from lung cancer was $43,336 per patient, of which the direct cost per capita was $42,540 (98.16%) and the indirect cost per capita was $795 (1.84%). Of the total direct medical costs, 35.66% was paid by the insurer and 9.84% was not covered by insurance. The economic burden for diagnosed lung cancer patients in the first year following diagnosis was $30,277 per capita, which accounted for 171% of the household annual income, a percentage that fell to 107% after subtracting the compensation from medical insurance. Conclusions: The economic burden for lung cancer patients is substantial in the urban areas of China, and an effective control strategy to lower the cost is urgently needed.Lung cancer has been a major public health problem in most countries for several decades. In 2012, an estimated 1.8 million new cases of lung cancer were diagnosed (12.9% of the total new cancer cases); 58% of these cases occurred in the less developed regions of the world. Lung cancer was the most common cause of death from cancer worldwide, responsible for an estimated nearly one in five cancer deaths (1.59 million deaths, 19.4% of the total cancer deaths) [1].In China, lung cancer has been the leading cancer diagnosis and the leading cause of cancer deaths for many years. According to the Chinese National Central Cancer Registry (NCCR), the year 2010 saw 605,946 new lung cancer diagnoses (19.59% of the total cancer cases) in China with a crude incidence rate of 46.08 per 100,000. Of these diagnoses, the number of cases from urban areas was 348,107 (57.45% of the total new lung cancer cases). With respect to mortality from cancer, an estimated 486,555 people in China died of lung cancer in 2010 with a crude mortality rate of 37.00 per 100,000, and 279,919 (57.53%) of these patients were from urban areas [2]. The World Health Organization (WHO) predicted that by 2025, the annual number of new cases of lung cancer mortality in China will be over one million, and the number of lung cancer patients will be the highest in the world [3].As the population ages, the expenditure for lung cancer treatment will become even more burdensome to the entire society [4,5,6]. To date, a relatively large amount of research has been done on the economic burden of lung cancer worldwide [7,8,9,10,11,12,13,14,15,16]. However, the economic burden of lung cancer patients varies by country, due to differences in economic development, health systems, exchange rate, purchasing power, and other factors [5].Studying the economic burden of lung cancer in China is attracting the attention of more and more scholars. However, most studies to date have been limited in measuring the direct costs or hospitalization expenses of lung cancer patients by using data from specific hospitals [17,18,19,20,21]. Moreover, most previous studies also were limited in calculating short-term costs without considering the cost of long-term treatment in the years following diagnosis.According to previous reports, the five-year survival rate of lung cancer patients in China is reported at 10%–15% [3,18,22]. Estimating the expenditure of lung cancer patients in the first five years is necessary to grasp the burden of the whole course of lung cancer treatment. Therefore, the aim of this study is to calculate the total cost of lung cancer treatment for lung cancer survivors in China within five years from the date of diagnosis. The total cost includes direct costs and indirect costs. In addition, we analyzed the payment of health insurance schemes on the direct medical costs and the economic impact of lung cancer on the family income, information that could provide an objective basis for health care policymaking.In this study, the sample was identified through the Nangang District Registry information system of the Chinese National Central Cancer Registry [23]. From 2010 to 2014, 396 individuals survived a diagnosis of lung cancer in 10 communities of Nangang District. Among them, 195 lung cancer patients were finally enrolled in the study; the remainders were removed from the potential sample because of the researchers’ inability to get follow-up data and/or patients’ refusal to participate. The retrospective questionnaire survey was conducted with 195 surviving lung cancer patients or with their main family caregivers to collect the information on sociodemographic characteristics, utilization of medical services, and costs during illness. All of the respondents provided informed consent in writing.Cases included in the study were surviving patients who had been diagnosed with lung cancer defined according to the International Classification of Disease 10th revision (ICD-10) codes from C33-C34 [24].In this study, the economic burden of lung cancer was estimated by calculating direct medical costs, direct nonmedical costs, and indirect costs covered in the five years following diagnosis. Direct medical costs were expenditures on medical services and drugs associated with diagnosis and treatment performed in hospitals, clinics, and pharmacies. These direct costs included three components: inpatient cost, outpatient cost, and purchased drug cost for lung cancer treatment paid by insurance, co-payment, and non-insurance [25]. Direct nonmedical costs included transportation costs for a round trip to visit health service providers and costs associated with accommodations, extra nutrition, and hired escorts for lung cancer patients during the period of treatment in medical institutions. Indirect costs were defined as days of lost productivity for both patients and their family caregivers resulting from outpatient visits and hospitalization due to lung cancer [16].The direct cost was calculated by summing the cost of the episodes of care in the period considered in this study. Because the costs of treatments varied in different years, all the costs were adjusted to 2014 to eliminate the effects of inflation using the annual consumer price index (CPI) of Heilongjiang Province [26,27].Indirect costs were measured using the human capital approach. The indirect costs were calculated by multiplying the average daily wage income for an urban resident in Heilongjiang Province from 2010 to 2014 [16,28,29]. In China, the mandatory retirement age for men is 60, and, for women, it is 50. Therefore, the productivity loss for patients and their families beyond this age range were not considered in this study. Since the lost wages were considered over five years, a 3% inflation index was used to convert past monetary values into present value in 2014 [6,30]. The Medical Ethics Committee of Harbin Medical University (Daqing) agreed with this study and examined the project for related medical ethics problems. The ethical project identification code is 16HMUSCI032.As noted, 195 patients with lung cancer were eventually included in this study. Study participants at the time of the survey were aged 29 to 89; 122 (62.56%) of them were male; 125 (64.10%) had retired, 22 (11.28%) were employed, and about a quarter were unemployed. No statistical significance was found at the 5% level between respondents and identified patients not included in the sample by age (p = 0.898) and gender (p = 0.311). Of the 195 patients included in the study, 157 respondents (80.51%) were covered by various health insurance schemes and 38 (19.49%) were uninsured (Table 1).The direct medical costs to the lung cancer patients in the study within five years after diagnosis were $40,650 per patient, making up 93.80% of the total costs. The average total cost of outpatient visits for those who made such visits was $1679 per patient, accounting for 3.87% of the average total costs. In addition, the direct medical cost for hospitalization was $28,307 per lung cancer patient, representing 65.32% of the total costs of treatment. The cost of drugs associated with lung cancer treatment that were purchased from a pharmacy was $10,664 per patient, comprising 24.61% of the total costs. In summary, hospitalization was the main component of the direct medical costs for lung cancer patients, followed by the cost of purchased drugs (Table 2).The average direct nonmedical cost was $1890 per patient, which accounted for 4.36% of total costs (Table 2). The largest direct nonmedical cost was the nutrition costs, $1300 per patient, higher than the accommodation costs and transportation costs.The average indirect cost calculated as the lost productivity of patients and family caregivers was $795 and represented the smallest share of overall costs at 1.84%. The average cost of lost productivity was considerably higher for family caregivers ($704) than for the patients themselves ($92) (Table 2).A total of $7,926,775 was spent as direct medical costs for the 195 patients in the sample, $780,085 (9.84%) of which was paid by 38 noninsured patients. The direct medical costs for patients covered by medical insurance were paid jointly by insurance and as copayments by insured patients. Of the total direct medical expenditures for lung cancer, medical insurance paid for 22.98% of the outpatient costs, 48.86% of the hospitalization costs, and 2.59% of the costs of purchased drugs, respectively (Table 3).Of the 195 lung cancer patients in the study, 157 participated in medical insurance schemes, and the average reimbursement rate of direct medical costs for insured patients was 39.55%. The highest reimbursement rate for direct medical costs was other types of insurance, which included commercial medical insurance, two or more forms of medical insurance, and so forth. Of the 195 participants, 141 patients were covered by Urban Employees Basic Medical Insurance (UEBMI), with a reimbursement rate of 39.39%, followed by the reimbursement rate of 38.14% from Urban Residents Basic Medical Insurance (URBMI) (Table 4).The proportion of the economic burden of lung cancer treatment to household income during the first year after diagnosis decreased from 171% to 107% after health insurance reimbursement. The impact of the economic burden on the family economic situation varied with the quintile household income. The results showed that the higher the household income, the lower the economic burden of lung cancer relative to household income. For the low- and middle-income households, the economic burden of the first year’s treatment was still higher than the annual household income after the reimbursement of medical insurance (Table 5).This paper was designed to estimate the economic burden of lung cancer treatment within five years from the date of diagnosis for lung cancer survivors in urban China. The total economic burden was $43,336 per patient and was concentrated mainly in the first year after diagnosis. The study showed that direct medical costs were the major component of the total economic burden of patients with lung cancer. The average direct medical cost in the first year after diagnosis was $30,277. Prior studies that estimated the direct economic burden of lung cancer inpatients showed that the average direct cost ranged from $16,276 to $20,076 during the period of hospitalization in other areas of China [32,33,34]. These estimates are considerably lower than those identified in our study because their direct medical costs were collected from medical records in specific hospitals only. Of the direct medical costs, the proportion of hospitalization costs to the total economic burden of lung cancer was higher than other costs. The average hospitalization cost that we estimated was $28,307, accounting for 65.32% of the total economic burden, followed by purchased drugs, which accounted for 24.61%. The study by Vasiliki et al. indicated that hospitalization costs ranged from 31% to 71% of total costs based on findings from several other countries [16]. The results of our study suggest that more attention should be paid to the management of hospitalization costs and to the cost of drugs from pharmacies for lung cancer patients.The average indirect cost found in this study was $795, accounting for 1.84% of the total economic burden. This figure is probably an underestimate for several reasons. First, lost working days were considered a time loss only during the treatment period; the days lost by patients and their families during the recovery period were not included in our study due to lack of reliable data; Second, productivity losses for males older than 60 and for females over 50, the official retirement ages in China, were not calculated in the indirect cost. The relatively high ages of lung cancer patients might account for the small indirect costs. Third, we estimated productivity lost by using the unified wage standard without considering the different level of wages in different professions.Although some costs are paid by medical insurance, the economic burden of the first year still accounted for 107.49% of the average family annual income after the insurance compensation. All of the families suffered catastrophic health expenditures, defined as out-of-pocket spending for health care that exceeds 40% of a household’s income [35,36]. A study by Park et al. reported that the cost for five-year lung cancer survivors constituted 44.7% of the per capita income during the same period in a tertiary care hospital in South Korea in 2002 [15]. The results of our study showed that the economic burden of lung cancer remains heavy, especially for low-income families in urban China. Thus, the government needs to strengthen the Fiscal Medical Assistance for these lung cancer patients.Medical insurance played an important role in reducing the proportion of out-of-pocket expenses in direct medical costs. In our study, more than one-third of the direct medical costs were paid by medical insurance, and the costs paid by copayments of insured patients and noninsured patents were 54.5% and 9.84%, respectively. In China, urban residents are covered mainly by two social basic medical insurance systems: UEBMI, which covers employed urban residents, and URBMI, which is designed to cover nonemployed urban residents, children, and students [37]. Combined, UEMBI and URBMI covered about 71.35% of the urban population, with coverage rates of 39.30% and 32.04%, respectively, in Heilongjiang Province in 2014 [38]. The results of this study showed slightly different reimbursement ratios for lung cancer patients: 39.39% for UEBMI and 38.14% for URBMI patients. Of the 157 patients with health insurance, the reimbursement rate of health insurance was 39.55%, and the remaining 60.45% of direct medical costs were paid Out-of-Pocket. The high copayment for current health insurance schemes needs to be reduced to avoid the occurrence of catastrophic health expenditures.This analysis has several limitations. First, the recall bias of respondents in reviewing the costs could not be avoided due to the retrospective investigation in the community; Second, because this study included only survivors of lung cancer, the indirect costs consisted of only the productivity loss caused by delays, and not the cost of lost productivity due to premature mortality, which could greatly undervalue the indirect economic burden of illness [39,40]. The lack of data on productivity lost by patients and caregivers during the recovery period could also underestimate the indirect cost estimates; Third, intangible economic costs to the psychological and mental health of lung cancer patients and their caregivers, such as the pain, sorrow, and inconvenience due to the decline in quality of life, were not included because they are difficult to convert into a monetary value [41]; Finally, the variety of clinical types and stages of lung cancer were not taken into account because of a lack of data, but they comprise a topic worthy of further studies.From the perspective of the survival of individuals with lung cancer, this study showed that the costs of lung cancer are substantial compared to the patients’ household income. The economic burden of lung cancer patients is attributed mainly to direct costs in the urban areas of China. Furthermore, the largest component of the total economic burden was the cost of hospitalization. Indirect costs were considerably higher for family caregivers than for patients themselves and represented a relatively small proportion of the total economic burden. Although medical insurance paid on average 40% of the total direct medical cost for insured patients, the proportion of out-of-pocket expenses was still too high. The economic burden for lung cancer patients was heaviest in the first year following diagnosis and was likely to induce catastrophic health expenditures for households, especially those low- and middle-income families. To further reduce the health care economic burden, increased and well-targeted subsidies for low-income patients are needed.This study was supported by grants from the National Institutes of Health, Fogarty International Center and National Cancer Institute, Bethesda, MD, USA (Grant: RO1TW009295), and the National Nature Science Foundation of China (71673071).Zhengzhong Mao and Teh-wei Hu conceived and designed the study. Guoxiang Liu, Jian Du, Wenqi Fu, Xiaowen Zhao, and Weidong Huang collected data. Yang Liu and Xianming Zhao interpreted data. Xin Zhang and Shuai Liu wrote the first draft. All authors approved the final version of the draft for publication.The authors declared no conflict of interest.Sociodemographic characteristics of 195 lung cancer patients.Purchasing power parity (PPP) was used in this study to convert RMB to dollars: $1 = 3.567 RMB (The PPP values of RMB against the dollar in 2014) [31].Total economic burden of 195 lung cancer patients.* The cost of purchased drugs was the expense of the drugs that patients bought in pharmacies for their own lung cancer treatment after the initial diagnosis. Expenditures for medicine prescribed by the doctors and purchased in medical institutions were included in outpatient or inpatient costs; ** The other costs were estimated including the cost of obtaining copies of medical records, the cost of special diets, special clothes, and wheelchairs for patients, etc. $1 = 3.567 RMB (the PPP values of RMB against the dollar in 2014).Medical expenditures payment by type of medical service.Unit: USD; * Total cost paid by insurer or patients as a percent of total direct medical costs; ** Paid by the copayment of insured patients; *** Non-covered refers to the costs for services paid by patients without any medical insurance.Direct medical costs paid by type of medical insurance for insured patients.Unit: USD.Influence of the financial burden in the first year after diagnosis on the family economic situation.Note: Column 4 = column 2/column 1; column 5 = (column 2 − column 3)/column 1.

Med-MDPI/ijerph_2/ijerph-14-03-00309.txt

@@ -0,0 +1,1599 @@
+Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).Leishmaniasis is the third most common vector-borne disease and a very important protozoan infection. Cutaneous leishmaniasis is one of the most common types of leishmaniasis infectious diseases with up to 1.2 million occurrences of new cases each year worldwide. A dynamic transmission multivariate time series model was applied to the data to account for overdispersion and evaluate the effects of three environmental layers as well as seasonality in the data. Furthermore, ecological niche modeling was used to study the geographically suitable conditions for cutaneous leishmaniasis using temperature, precipitation and altitude as environmental layers, together with the leishmaniasis presence data. A retrospective analysis of the cutaneous leishmaniasis spatial data in Afghanistan between 2003 and 2009 indicates a steady increase from 2003 to 2007, a small decrease in 2008, and then another increase in 2009. An upward trend and regularly repeating patterns of highs and lows were observed related to the months of the year, which suggests seasonality effect in the data. Two peaks were observed in the disease occurrence—January to March and September to December—which coincide with the cold period. Ecological niche modelling indicates that precipitation has the greatest contribution to the potential distribution of leishmaniasis.The World Health Organization estimates that there are 0.9–1.2 million annual cases of Leishmaniasis worldwide with approximately 0.2 to 0.4 million visceral leishmaniasis (VL) cases and 0.7 to 1.2 million cutaneous leishmaniasis (CL) cases each year [1]. Leishmaniasis is a vector-borne parasitic disease contracted through bites from infected female phlebotomine sand flies, which may result in mild self-healing cutaneous lesions to fatal visceral cases [2]. It is the third most common vector-borne disease and a very important protozoan infection [3]. The burden of the disease is overwhelming and the psychological effect can be disturbing. In some societies, women infected with this disease are stigmatized and may be deemed unsuitable for marriage and motherhood [4].There are over 30 leishmania species known to infect humans [5]. The three main types of leishmaniasis disease are (i) visceral leishmaniasis (VL), often known as kala-azar and is the most serious form of the disease; (ii) cutaneous leishmaniasis (CL), which is the most common; and (iii) mucocutaneous [5]. Over 98 countries and territories are endemic for leishmaniasis with more than 70%–75% of the global CL cases occurring in ten countries: Afghanistan, Algeria, Brazil, Colombia, Costa Rica, Ethiopia, Iran, Peru, Sudan and Syria [1].Afghanistan is a cutaneous leishmaniasis endemic country and the disease is of a serious health concern with about 200,000 estimated new cases of CL infection nationwide and 67,500 cases in Kabul alone [6]. There are two forms of cutaneous leishmaniasis (CL): zoonotic cutaneous leishmaniasis (ZCL) and anthroponotic cutaneous leishmaniasis (ACL). Most cases of leishmaniasis in Afghanistan are caused by Leishmania tropica, which is transmitted anthroponotically (i.e., humans are the reservoir) by the sand fly (Phlebotomus sergenti) [7,8,9]. On the other hand, ZCL is caused by Leishmania major with bites by Phlebotomus papatasi and occurs indigenously in rural northern Afghanistan [10]. However, in Afghanistan, public surveillance data are often reported without clear distinction between ZCL and ACL but are rather primarily focused on ACL [4,10,11].A study of the risk factors for leishmaniasis incidence in Afghanistan found that household construction materials, design, density and presence of the disease in the neighborhoods are significant risk factors for ACL in Kabul [8]. Intervention measures such as simple window screens were demonstrated to reduce sand fly human vector contact [8]. Furthermore, evidence suggests that rodents may serve as natural hosts for CL and their infestations may increase the outbreak of the ZCL disease [11]. Similarly, the seasonality in the occurrence of ZCL in humans can be attributed to seasonal activity of the vector [11]. Many of the studies have not considered a spatial analysis approach to leishmaniasis in Afghanistan except for Adegboye [12]. For example, in Adegboye [12], a spatial hierarchical Bayesian model was used to analyze the spatial pattern of provincial level geo-referenced data of leishmaniasis in Afghanistan while Adegboye and Kotze [3] used random effects to assess spatial dependencies. They found excess risks of leishmaniasis in the northeastern and southeastern parts of Afghanistan [3,12].The goal of this study is twofold—firstly, it is to explore the effect of time-varying environmental variables such as temperature, precipitation and elevation on the prediction of cutaneous leishmaniasis (CL) in Afghanistan. In surveillance counts data, underreporting or reporting delays of disease incidence are very common, which may give rise to overdispersion and blurred dependencies [13]. In this case, the usual Poisson model assumption of equal transmission incidence rates is doubtful; therefore, an alternative model that allows for possible overdispersion was used to achieve the first goal via negative binomial multivariate time series model. This model allows for overdispersion and seasonality. It is not unusual to discover this association to spread over a few time periods, and the association between environmental variables and disease occurrence over time may lead to bias unless the relationship is adequately modelled. We incorporated lag variables in the model to account for delay effects that spread over time.The second goal of this study is to estimate the ecological niche and potential distribution of CL in Afghanistan using ecological niche modelling. It is not unusual to have some diseases more frequent in certain geographical areas, which could be attributed to environmental suitability of the region. Ecological niche modelling (ENM) provides a way to relate the geographical distributions of species to ecologic niches [14]. ENM is a growing area in spatial analysis of disease epidemiology [14,15], characterizing the distribution of the disease in a space defined by environmental parameters [15]. The use of ENM under the framework of geographic information system in spatial disease modelling cannot be overemphasized. For example, Du et al. [16] used ENM to identify the potential high risk areas for severe fever with thrombocytopenia syndrome (SFTS) in China. Adegboye and Kotze [15] used ENM to explore the geographical constraints that may favour the outbreak of the H5N1 virus in Nigeria. Similarly, ENM was used to estimate the current niche and potential distribution of mycetoma in Sudan and South Sudan [17], and to predict the zoonotic transmission niche of Ebola covering countries across Central and West Africa [18]. Recently, Chalghaf et al. [19] used ENM to study the effects of environmental variables on the geographical distribution of P. papatasi and to estimate environmental suitability index for CL in Tunisia.The rest of the paper is structured as follows. Section 2 introduces the data set that motivated this study and the methods used. The results of the analyses are reported in Section 3. In Section 4, the concluding remarks will be presented.Afghanistan is a landlocked country that is located in South Asia. The country is characterized by mountainous regions, and it is a leishmaniasis disease endemic country. Monthly data of cases of leishmaniasis reported to the Afghanistan Health Management Information System (HMIS) under the National Malaria and Leishmaniasis Control Programme (NMLCP) were obtained from the Ministry of Public Health (MoPH), Kabul, Afghanistan. The leishmaniasis infections were confirmed clinically or by examining the leishmania parasites in the skin lesion biopsy using calibrated ocular micrometre supported binocular light microscopy. There were 148,945 new cases of leishmaniasis recorded in a total of 20 provinces across Afghanistan between 2003 and 2009. Province level population denominators were obtained from the central statistics organization (CSO) of Afghanistan. Figure 1 shows the nationally aggregated monthly number of cases per 100,000 for the years 2003 to 2009.Three environmental layers data sets were used in this study. The satellite-derived environmental-land surface temperature (LST) was obtained from the Moderate Resolution Imaging Spectroradiometer (MOD11 L2 version 6, USGS/Earth Resources Observation and Science (EROS) Center, Sioux Falls, South Dakota) at 1 km spatial resolution [20] while the elevation data was obtained from the Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) Global Emissivity Dataset (GED) version 3, USGS/Earth Resources Observation and Science (EROS) Center, Sioux Falls, South Dakota [21]. LST and elevation environmental layers were retrieved from the online data pool, courtesy of the National Aeronautics and Space Administration (NASA) Earth Observing System Data and Information System (EOSDIS) Land Processes Distributed Active Archive Center (LP DAAC), United States Geological Survey/Earth Resources Observation and Science (USGS/EROS) Center, Sioux Falls, South Dakota. The monthly accumulated rainfall data measured by the Tropical Rainfall Measuring Mission (TRMM: TMPA/3B43) [22] jointly conducted by NASA and the Japan Aerospace Exploration Agency (JAXA) was obtained from NASA.Suppose 
+          
+            
+              
+                y
+                
+                  i
+                  ,
+                  t
+                
+              
+            
+          
+         is the monthly counts of leishmaniasis cases in province 
+          
+            
+              
+                i
+                =
+                1
+                ,
+                2
+                ,
+                .
+                .
+                .
+                ,
+                S
+              
+            
+          
+         at time series 
+          
+            
+              
+                t
+                =
+                1
+                ,
+                .
+                .
+                .
+                ,
+                T
+              
+            
+          
+        . The disease counts are assumed to follow a negative binomial distribution, 
+          
+            
+              
+                
+                  y
+                  
+                    i
+                    ,
+                    t
+                  
+                
+                ∣
+                
+                  y
+                  
+                    t
+                    −
+                    l
+                  
+                
+                ∼
+                NegBin
+                
+                  (
+                  
+                    μ
+                    
+                      i
+                      ,
+                      t
+                    
+                  
+                  ,
+                  ψ
+                  )
+                
+              
+            
+          
+        , with conditional mean
+        
+          (1)
+          
+            
+              
+                
+                  μ
+                  
+                    i
+                    ,
+                    t
+                  
+                
+                =
+                
+                  ν
+                  
+                    i
+                    ,
+                    t
+                  
+                
+                +
+                
+                  λ
+                  
+                    i
+                    t
+                  
+                
+                
+                  y
+                  
+                    i
+                    ,
+                    t
+                    −
+                    l
+                  
+                
+                +
+                
+                  ϕ
+                  
+                    i
+                    t
+                  
+                
+                
+                  ∑
+                  
+                    j
+                    ≠
+                    i
+                  
+                
+                
+                  w
+                  
+                    j
+                    ,
+                    i
+                  
+                
+                
+                  y
+                  
+                    j
+                    ,
+                    t
+                    −
+                    l
+                  
+                
+                ,
+              
+            
+          
+        
+        and conditional variance
+          
+            
+              
+                
+                  
+                    σ
+                    
+                      i
+                      ,
+                      t
+                    
+                    2
+                  
+                  =
+                  
+                    μ
+                    
+                      i
+                      ,
+                      t
+                    
+                  
+                  +
+                  ψ
+                  
+                    μ
+                    
+                      i
+                      ,
+                      t
+                    
+                    2
+                  
+                  ,
+                
+              
+            
+          
+        where 
+          
+            
+              
+                y
+                
+                  i
+                  ,
+                  t
+                  −
+                  l
+                
+              
+            
+          
+         denotes the disease counts in province i at time 
+          
+            
+              
+                t
+                −
+                l
+              
+            
+          
+         with lag 
+          
+            
+              
+                l
+                ∈
+                
+                  1
+                  ,
+                  2
+                  ,
+                  .
+                  .
+                  .
+                
+              
+            
+          
+        ; 
+          
+            
+              
+                (
+                
+                  ϕ
+                  
+                    i
+                    t
+                  
+                
+                ,
+                
+                  ν
+                  
+                    i
+                    t
+                  
+                
+                ,
+                
+                  λ
+                  
+                    i
+                    t
+                  
+                
+                >
+                0
+                )
+              
+            
+          
+         and ψ is the overdispersion parameter. The parameters in 
+          
+            
+              
+                μ
+                
+                  i
+                  ,
+                  t
+                
+              
+            
+          
+         is decomposed into two parts: 
+          
+            
+              
+                log
+                
+                  (
+                  
+                    ν
+                    
+                      i
+                      ,
+                      t
+                    
+                  
+                  )
+                
+                =
+                log
+                
+                  (
+                  
+                    n
+                    
+                      i
+                      ,
+                      t
+                    
+                  
+                  )
+                
+                +
+                
+                  a
+                  i
+                
+                +
+                
+                  α
+                  0
+                
+                +
+                
+                  α
+                  1
+                
+                t
+                +
+                
+                  α
+                  2
+                
+                sin
+                
+                  
+                    
+                      2
+                      π
+                    
+                    12
+                  
+                  t
+                
+                +
+                
+                  α
+                  3
+                
+                cos
+                
+                  
+                    
+                      2
+                      π
+                    
+                    12
+                  
+                  t
+                
+              
+            
+          
+         is the random "endemic" component that incorporates trend parameter, and sinusoidal wave of frequency to capture seasonality, and 
+          
+            
+              
+                n
+                
+                  i
+                  ,
+                  t
+                
+              
+            
+          
+         as the offset population size, whereas 
+          
+            
+              
+                log
+                
+                  (
+                  
+                    λ
+                    
+                      i
+                      ,
+                      t
+                    
+                  
+                  )
+                
+                =
+                
+                  β
+                  0
+                
+                +
+                
+                  x
+                  
+                    i
+                    ,
+                    t
+                  
+                
+                
+                  β
+                  1
+                
+              
+            
+          
+         is the autoregressive component with explanatory variable(s) termed the "epidemic" component(s) [23]. The epidemic component is used to capture the occasional outbreaks while the endemic component explains the baseline rate of cases that is persistent with a stable temporal pattern [23].In addition, 
+          
+            
+              
+                log
+                
+                  (
+                  
+                    ϕ
+                    
+                      i
+                      t
+                    
+                  
+                  )
+                
+                =
+                
+                  γ
+                  0
+                
+                +
+                
+                  c
+                  i
+                
+              
+            
+          
+         is the random neighbor-driven component that enables interdependency exploration between provinces with 
+          
+            
+              
+                w
+                
+                  j
+                  ,
+                  i
+                
+              
+            
+          
+         representing the spatial weight matrix (spatio-temporal component). The random effects 
+          
+            
+              a
+            
+          
+         and 
+          
+            
+              c
+            
+          
+         are assumed to be normally distributed with a mean 0 and variance Σ. Several models with varying structures and complexity were explored.These models were validated using the proper scoring rules based on probabilistic one-step-ahead predictions to assess the best model. The smaller the score, the better the predictive quality [13]. A scoring rule 
+            
+              
+                
+                  S
+                  (
+                  P
+                
+              
+            
+          :
+            
+              
+                
+                  y
+                  )
+                
+              
+            
+           measures the predictive quality of a stated predictive distribution P by comparing it with the actual observed value y. When the expected value under P becomes minimal given that the observed value y is indeed a realization from P, then we have a proper scoring rule. For this study, we shall use the logarithmic score 
+            
+              
+                
+                  Log
+                  S
+                  (
+                  P
+                
+              
+            
+          :
+            
+              
+                
+                  y
+                  )
+                
+              
+            
+           and the ranked probability score 
+            
+              
+                
+                  RPS
+                  (
+                  P
+                
+              
+            
+          :
+            
+              
+                
+                  y
+                  )
+                
+              
+            
+           for random effects model and Akaike information criterion (AIC) for models without random effects (see, for example, [13]).These models are implemented in “R” [24] using the package “surveillance” (version 1.12.1) [25,26] as function “hhh4” [13], which can be downloaded from “R” website.The main point of ecological niche modelling (ENM) is to relate the geographical distributions of species to ecologic niches [14]. The geographical variations in the species occurrences are often profoundly favoured by certain climatic and environmental constraints [14,15]. In ENM, the observed presence data (and pseudo-absence data) together with ecological variables at the sample region are used to provide a reasonable likelihood of the species being present at all other locations [15,27]. Such projections assume that a species distribution is mainly determined by its environmental requirements and not by other factors [27] (see, for example, [28] for more details on ENM).MaxEnt version 3.3.3k [29] was used to estimate the ecological niches of leishmaniasis in Afghanistan via maximum entropy algorithm [30,31]. Although the MaxEnt software package [29] is particularly designed for species distribution/environmental niche modeling [32], the software can be easily adapted to disease niche modelling. The usual input of “species” presence locations will be replaced by the “disease” presence locations together with the set of environment variables. The analysis was carried out by dividing the occurrence data into 80% training data and 20% test data. The default convergence threshold of 0.00001 and 500 maximum iterations were used while 10 replicated models based on independent random partitions were used to get a robust model. The jackknife test of variable importance was used to estimate the relative contributions of the environmental variables to the model. Furthermore, the logistic output was used to estimate the suitability index (predicted probability of presence), which gives values between 0 and 1, indicating impossible conditions to highly suitable conditions.The receiver operating curve (ROC) was used to investigate high predictive power of the model. A model with high predictive power will have a large area under the ROC curve (AUC), that is, the model accurately predicts the presence and absence. Following Hosmer and Lemeshow [33], a model with a AUC value between 0.7 and 0.8 is considered to provide acceptable discrimination while an AUC value above 0.8 is excellent (see [30,31,34] for further details on MaxEnt and [32] for an excellent practical guide to MaxEnt).We begin the analysis by exploring the observed monthly cases of leishmaniasis in Afghanistan between 2003 and 2009. The top panel of Figure 2 shows the distribution of the monthly cases of the disease and a map of the aggregated monthly data across the seven years under study. The second panel is the trend component, which indicates a steady increase from 2003 to 2007, a small decrease in 2008, and then another increase in 2009. The third panel shows that the seasonal factor is the same for each year. An upward trend and regularly repeating patterns of highs and lows was observed related to the months of the year, which suggests seasonality in the data. The monthly profile indicates two peaks in the disease occurrence in Afghanistan between 2003 and 2009—January to March and September to December—which coincides with the cold period, while July is the hottest month and March is the wettest month. The largest seasonal factor is for March, and the smallest is for September, suggesting a peak in the cases of leishmaniasis in March and a trough in September of each year.Similarly, the map in Figure 1 indicates higher disease incidence around the Kabul area (northeastern). Kabul City accounted for more than 50% of the total new cases in 2009, and this may be attributed to availability of health care facilities, which is crucial for data gathering in public health studies. The rate of infections is about the same for both male and female (around 50% each), while those in age group 4–14 years are more likely to be infected than other ages.The autocorrelation plot (Figure 3) shows a significant autocorrelation at lag 1 with 
+          
+            
+              
+                ρ
+                =
+                0.7977
+              
+            
+          
+         (D − W = 0.3719, p-value = 0). A further look using the partial autocorrelation function (PACF) shows a significant spike only at lag 1, which attests to the ACF. The autocorrelation at lag 1 is sufficient to explain the higher order autocorrelations, which implies that we can discard the longer lags.The model described in Section 2.3 has been fitted to the leishmaniasis count data to account for overdispersion and spatial dependency. The following model formulations were fitted to account for variability in the incidence of the disease as well as to investigate the effects of several risk factors:
+		
+          
+            M1 
+			
+                
+                  
+                    
+                      
+                        μ
+                        
+                          i
+                          ,
+                          t
+                        
+                      
+                      =
+                      
+                        ν
+                        
+                          i
+                          ,
+                          t
+                        
+                      
+                    
+                  
+                
+              ; only seasonal variation in the endemic component, where 
+                
+                  
+                    
+                      log
+                      
+                        (
+                        
+                          ν
+                          
+                            i
+                            ,
+                            t
+                          
+                        
+                        )
+                      
+                      =
+                      log
+                      
+                        (
+                        
+                          n
+                          
+                            i
+                            ,
+                            t
+                          
+                        
+                        )
+                      
+                      +
+                      
+                        α
+                        0
+                      
+                      +
+                      
+                        α
+                        1
+                      
+                      t
+                      +
+                      
+                        α
+                        2
+                      
+                      sin
+                      
+                        
+                          
+                            2
+                            π
+                          
+                          12
+                        
+                        t
+                      
+                      +
+                      
+                        α
+                        3
+                      
+                      cos
+                      
+                        
+                          
+                            2
+                            π
+                          
+                          12
+                        
+                        t
+                      
+                    
+                  
+                
+              
+          
+          
+		  M2 
+            
+                
+                  
+                    
+                      
+                        μ
+                        
+                          i
+                          ,
+                          t
+                        
+                      
+                      =
+                      
+                        ν
+                        
+                          i
+                          ,
+                          t
+                        
+                      
+                      +
+                      
+                        λ
+                        
+                          i
+                          ,
+                          t
+                        
+                      
+                      
+                        y
+                        
+                          i
+                          ,
+                          t
+                          −
+                          1
+                        
+                      
+                      +
+                      
+                        ϕ
+                        
+                          i
+                          ,
+                          t
+                        
+                      
+                      
+                        ∑
+                        
+                          j
+                          ≠
+                          i
+                        
+                      
+                      
+                        w
+                        
+                          j
+                          ,
+                          i
+                        
+                      
+                      
+                        y
+                        
+                          j
+                          ,
+                          t
+                          −
+                          l
+                        
+                      
+                    
+                  
+                
+              ; seasonal variation in the endemic component, autoregressive in the epidemic component, spatiotemporal component, but without adjusting for covariates in the epidemic component, where 
+                
+                  
+                    
+                      log
+                      
+                        (
+                        
+                          ν
+                          
+                            i
+                            ,
+                            t
+                          
+                        
+                        )
+                      
+                      =
+                      log
+                      
+                        (
+                        
+                          n
+                          
+                            i
+                            ,
+                            t
+                          
+                        
+                        )
+                      
+                      +
+                      
+                        α
+                        0
+                      
+                      +
+                      
+                        α
+                        1
+                      
+                      t
+                      +
+                      
+                        α
+                        2
+                      
+                      sin
+                      
+                        
+                          
+                            2
+                            π
+                          
+                          12
+                        
+                        t
+                      
+                      +
+                      
+                        α
+                        3
+                      
+                      cos
+                      
+                        
+                          
+                            2
+                            π
+                          
+                          12
+                        
+                        t
+                      
+                    
+                  
+                
+              , 
+                
+                  
+                    
+                      λ
+                      =
+                      
+                        β
+                        0
+                      
+                    
+                  
+                
+               and 
+                
+                  
+                    
+                      log
+                      
+                        (
+                        
+                          ϕ
+                          
+                            i
+                            ,
+                            t
+                          
+                        
+                        )
+                      
+                      =
+                      
+                        γ
+                        0
+                      
+                    
+                  
+                
+              , using the weights 
+                
+                  
+                    
+                      w
+                      
+                        j
+                        ,
+                        i
+                      
+                    
+                  
+                
+              =1/# neighbors of region j in the spatiotemporal component.
+          
+          
+		  M3 
+            
+                
+                  
+                    
+                      
+                        μ
+                        
+                          i
+                          ,
+                          t
+                        
+                      
+                      =
+                      
+                        ν
+                        
+                          i
+                          ,
+                          t
+                        
+                      
+                      +
+                      
+                        λ
+                        
+                          i
+                          ,
+                          t
+                        
+                      
+                      
+                        y
+                        
+                          i
+                          ,
+                          t
+                          −
+                          1
+                        
+                      
+                      +
+                      
+                        ϕ
+                        
+                          i
+                          ,
+                          t
+                        
+                      
+                      
+                        ∑
+                        
+                          j
+                          ≠
+                          i
+                        
+                      
+                      
+                        w
+                        
+                          j
+                          ,
+                          i
+                        
+                      
+                      
+                        y
+                        
+                          j
+                          ,
+                          t
+                          −
+                          l
+                        
+                      
+                    
+                  
+                
+              ; seasonal variation in the endemic component with random effects, autoregressive in the epidemic component with covariate adjustment and spatiotemporal component with random effects,
+          
+        
+		where 
+          
+            
+              
+                log
+                
+                  (
+                  
+                    ν
+                    
+                      i
+                      ,
+                      t
+                    
+                  
+                  )
+                
+                =
+                log
+                
+                  (
+                  
+                    n
+                    
+                      i
+                      ,
+                      t
+                    
+                  
+                  )
+                
+                +
+                
+                  a
+                  i
+                
+                +
+                
+                  α
+                  0
+                
+                +
+                
+                  α
+                  1
+                
+                t
+                +
+                
+                  α
+                  2
+                
+                sin
+                
+                  
+                    
+                      2
+                      π
+                    
+                    12
+                  
+                  t
+                
+                +
+                
+                  α
+                  3
+                
+                cos
+                
+                  
+                    
+                      2
+                      π
+                    
+                    12
+                  
+                  t
+                
+              
+            
+          
+        , 
+          
+            
+              
+                λ
+                =
+                
+                  β
+                  0
+                
+                +
+                
+                  β
+                  1
+                
+                ∗
+                log
+                A
+                L
+                T
+                +
+                
+                  β
+                  2
+                
+                ∗
+                T
+                e
+                m
+                p
+                e
+                r
+                a
+                t
+                u
+                r
+                e
+                +
+                
+                  β
+                  3
+                
+                ∗
+                P
+                r
+                e
+                c
+                i
+                p
+                i
+                t
+                a
+                t
+                i
+                o
+                n
+                +
+                
+                  β
+                  4
+                
+                ∗
+                W
+                i
+                n
+                d
+              
+            
+          
+        , 
+          
+            
+              
+                log
+                
+                  (
+                  
+                    ϕ
+                    
+                      i
+                      ,
+                      t
+                    
+                  
+                  )
+                
+                =
+                
+                  γ
+                  0
+                
+                +
+                
+                  c
+                  i
+                
+              
+            
+          
+        . The weights 
+          
+            
+              
+                w
+                
+                  j
+                  ,
+                  i
+                
+              
+            
+          
+        =1/# neighbors of region j in the spatiotemporal component.
+                
+                  
+                    
+                      
+                        μ
+                        
+                          i
+                          ,
+                          t
+                        
+                      
+                      =
+                      
+                        ν
+                        
+                          i
+                          ,
+                          t
+                        
+                      
+                    
+                  
+                
+              ; only seasonal variation in the endemic component, where 
+                
+                  
+                    
+                      log
+                      
+                        (
+                        
+                          ν
+                          
+                            i
+                            ,
+                            t
+                          
+                        
+                        )
+                      
+                      =
+                      log
+                      
+                        (
+                        
+                          n
+                          
+                            i
+                            ,
+                            t
+                          
+                        
+                        )
+                      
+                      +
+                      
+                        α
+                        0
+                      
+                      +
+                      
+                        α
+                        1
+                      
+                      t
+                      +
+                      
+                        α
+                        2
+                      
+                      sin
+                      
+                        
+                          
+                            2
+                            π
+                          
+                          12
+                        
+                        t
+                      
+                      +
+                      
+                        α
+                        3
+                      
+                      cos
+                      
+                        
+                          
+                            2
+                            π
+                          
+                          12
+                        
+                        t
+                      
+                    
+                  
+                
+              
+                
+                  
+                    
+                      
+                        μ
+                        
+                          i
+                          ,
+                          t
+                        
+                      
+                      =
+                      
+                        ν
+                        
+                          i
+                          ,
+                          t
+                        
+                      
+                      +
+                      
+                        λ
+                        
+                          i
+                          ,
+                          t
+                        
+                      
+                      
+                        y
+                        
+                          i
+                          ,
+                          t
+                          −
+                          1
+                        
+                      
+                      +
+                      
+                        ϕ
+                        
+                          i
+                          ,
+                          t
+                        
+                      
+                      
+                        ∑
+                        
+                          j
+                          ≠
+                          i
+                        
+                      
+                      
+                        w
+                        
+                          j
+                          ,
+                          i
+                        
+                      
+                      
+                        y
+                        
+                          j
+                          ,
+                          t
+                          −
+                          l
+                        
+                      
+                    
+                  
+                
+              ; seasonal variation in the endemic component, autoregressive in the epidemic component, spatiotemporal component, but without adjusting for covariates in the epidemic component, where 
+                
+                  
+                    
+                      log
+                      
+                        (
+                        
+                          ν
+                          
+                            i
+                            ,
+                            t
+                          
+                        
+                        )
+                      
+                      =
+                      log
+                      
+                        (
+                        
+                          n
+                          
+                            i
+                            ,
+                            t
+                          
+                        
+                        )
+                      
+                      +
+                      
+                        α
+                        0
+                      
+                      +
+                      
+                        α
+                        1
+                      
+                      t
+                      +
+                      
+                        α
+                        2
+                      
+                      sin
+                      
+                        
+                          
+                            2
+                            π
+                          
+                          12
+                        
+                        t
+                      
+                      +
+                      
+                        α
+                        3
+                      
+                      cos
+                      
+                        
+                          
+                            2
+                            π
+                          
+                          12
+                        
+                        t
+                      
+                    
+                  
+                
+              , 
+                
+                  
+                    
+                      λ
+                      =
+                      
+                        β
+                        0
+                      
+                    
+                  
+                
+               and 
+                
+                  
+                    
+                      log
+                      
+                        (
+                        
+                          ϕ
+                          
+                            i
+                            ,
+                            t
+                          
+                        
+                        )
+                      
+                      =
+                      
+                        γ
+                        0
+                      
+                    
+                  
+                
+              , using the weights 
+                
+                  
+                    
+                      w
+                      
+                        j
+                        ,
+                        i
+                      
+                    
+                  
+                
+              =1/# neighbors of region j in the spatiotemporal component.
+                
+                  
+                    
+                      
+                        μ
+                        
+                          i
+                          ,
+                          t
+                        
+                      
+                      =
+                      
+                        ν
+                        
+                          i
+                          ,
+                          t
+                        
+                      
+                      +
+                      
+                        λ
+                        
+                          i
+                          ,
+                          t
+                        
+                      
+                      
+                        y
+                        
+                          i
+                          ,
+                          t
+                          −
+                          1
+                        
+                      
+                      +
+                      
+                        ϕ
+                        
+                          i
+                          ,
+                          t
+                        
+                      
+                      
+                        ∑
+                        
+                          j
+                          ≠
+                          i
+                        
+                      
+                      
+                        w
+                        
+                          j
+                          ,
+                          i
+                        
+                      
+                      
+                        y
+                        
+                          j
+                          ,
+                          t
+                          −
+                          l
+                        
+                      
+                    
+                  
+                
+              ; seasonal variation in the endemic component with random effects, autoregressive in the epidemic component with covariate adjustment and spatiotemporal component with random effects,Following the results from the exploratory analysis (See Figure 2), we fitted a parameter driven Model M1 with only fixed parameters, including only a trend parameter and sinusoidal wave of frequency to capture seasonality in the ’endemic’ part of Equation (1). Table 1 displays the results from our model applied to the leishmaniasis data. Model M1 was fitted as a baseline model to assess the appropriateness of negative binomial model rather than a Poisson model. Model M1 provides a better fit (AIC = 133, 03.79) when compared with the Poisson model with an AIC of 404,954.22. The negative binomial baseline model has a significant overdispersion parameter, 
+          
+            
+              
+                ψ
+                =
+                7.596
+              
+            
+          
+         (
+          
+            
+              
+                95
+                %
+              
+            
+          
+         Confidence Interval (CI) 7.027–8.166), implying evidence for residual overdispersion.It is not unusual to discover this association to spread over a few time periods. Leishmaniasis incidence for any given month was autoregressed on the previous month’s infections. From exploratory analysis, one month lag autoregression was found to have a significant contribution. Additional parameters were introduced in the second Model M2. One month lag was included in the ’epidemic’ autoregressive effect and spatiotemporal component. The inclusion of these two terms improve the model via smaller AIC, LogS and RPS as shown in Table 1. Model (M2) provides an improvement over M1 with an AIC of 12,815.86 and overdispersion parameter 
+          
+            
+              
+                ψ
+                =
+                5.204
+              
+            
+          
+         (
+          
+            
+              
+                95
+                %
+              
+            
+          
+         CI 4.789–5.617). Finally, in Model M3, in addition to adjusting for covariates (Altitude, Precipitation, Temperature and Wind), each region gets its own fixed parameter in the endemic component as well as random neighbour-driven components that enable interdependency exploration between region.The predictive capabilities of the models were assessed through one-step-ahead predictions of the last five months. Model M3 was chosen as the best model based on smaller RPS and LogS among the three models. The interpretation of the parameter values are on the log scale, for example, in M3 
+          
+            
+              
+                e
+                
+                  α
+                  1
+                
+              
+            
+          
+         = exp(Trend) = 1.063 represents the seasonality-adjusted factor by which the basic endemic incidence increases per month. The estimates for trend and the seasonal components in the three models are quite similar and significant overdispersion parameter. The map in Figure 4 displays estimated spatially correlated random effects. It revealed that three provinces—Takhar, Baghlan and Laghman—show a low endemic incidence of CL in Afghanistan using variables from Model M3. The results of the best model confirmed the significant effect of precipitation and temperature. Likewise, the trend and the sinusoidal wave of frequency exerted a significant effect on leishmaniasis. However, no significant relationship of wind was detected as shown in Table 1.The fitted values for six provinces with the highest Leishmaniasis incidence in Afghanistan from Model 3 were presented in Figure 5. The plot is split into three components: the spatiotemporal, the autoregressive and the endemic components.Figure 6 presents the map of the suitability habitat for leishmaniasis in Afghanistan estimated using MaxEnt ecological niche modeling. The eastern and northeastern regions of the country are potential high risk areas for leishmaniasis represented by warmer colors in Figure 6. These areas represent previously known occurrences of leishmaniasis. The relative contribution of each environmental layers to the final training MaxEnt model using the average of the 10 replicates is presented in Table 2. Similarly, in Table 2, the results of the jackknife test of variable importance were presented in terms of ranking of training model gains when using each variable separately. The environmental layer with highest gain when used in isolation is the afg-prec4, which appears to have the most useful information by itself. The environmental layer that decreases the gain the most when it is omitted is afg-prec4, which appears to have the most information that is not present in the other variables. The nine environmental layers shown in Table 2 most favour the occurrence of leishmaniasis. The mean AUC value for the training models was 0.929, and that of the testing model was 0.756, which is an indication of good performance.The impact of environmental variables on leishmaniasis cannot be ruled out and human activities play a significant role in the dispersion of the vectors, thereby changing the geographical distribution of the disease. Previous studies have investigated the effects of environmental variables on the occurrence of leishmaniasis [19,35,36,37,38,39,40,41]. Several authors have previously used times series analyses to study infectious disease incidences [13,42,43,44,45,46,47,48]. These approaches were used to investigate the dynamic pattern of the disease over different temporal and spatial scales. Geospatial and spatio-temporal techniques as well as ecological niche analysis are very common approaches to study the association between environmental layers and leishmaniasis (see, for example, [38,49,50,51]).In this study, the dependent variable is the monthly time series of leishmaniasis in the provinces of Afghanistan between 2003 and 2009. The explanatory variables are the environmental layers: precipitation, temperature and altitude. Although cutaneous leishmaniasis is a worldwide health issue [52], Afghanistan, Algeria, Brazil, Colombia, Iran (Islamic Republic of) and the Syrian Arab Republic accounted for over 95% of the cases. There are several challenges in studying the effect of environmental factors on the transmission of infectious diseases in Afghanistan due to high variation in the landscape and climate. In the same vein, the association between environmental layers and disease occurrence over time may lead to bias unless the relationship is adequately modelled. When the association between disease incidence and environmental layers spans into the future, modelling techniques that incorporate autoregressive components is deemed necessary.Spatial time series analysis was used to investigate the effect of time-varying environmental layers on the occurrence of cutaneous leishmaniasis in Afghanistan by using a flexible NegBin model that allows for overdispersion and varying components’ specification. In this model, different types of variation and correlation were incorporated within a single model. Our results show two significant peaks—January to March and September to December—with the highest peak in March, suggesting a peak in the cases of leishmaniasis in March and a trough in September of each year. It is probable that the seasonality observed in the cases of leishmaniasis could be attributed to the abundance of sandflies (vector carrying the leishmania parasite), which vary between species, location and year of collection [53]. In the Mediterranean context, two peaks of species P. perniciosus were recorded in July and September in Tunis, Almeria, Algarve and Catania [53]. A tri-modal peak of P. tobbi was recorded in Cyprus, and a bi-modal peak was observed in Turkey, while a single peak of P. ariasi was observed in France and Georgia [53]. Gálvez et al. [39] reported that the seasonal abundance of P. perniciosus sandflies in Central Spain had two peaks in July and September, while it reported a single peak of P. ariasi in August. El-Shazly et al. [40] reported peaks of P. perniciosus in July and October in Egypt.Rodents may serve as natural hosts for cutaneous leishmaniasis [11]. The seasonality in the occurrence of zoonotic cutaneous leishmaniasis in humans was attributed to seasonal activity of the rodents, which are natural reservoir hosts of the disease [11,54]. Regions with good vegetation for plant growth have been described to provide shelter and food for rodents, thereby providing ideal sand fly habitats [55].Previous studies have described the occurrence of the disease as seasonal [38,41,55]. The seasonality in our data was also confirmed by the significance of trend and sinusoidal wave term as well as precipitation and temperature in our final model. These findings are similar to other studies, indicating a positive association between precipitation, temperature and leishmaniasis [41,55].Furthermore, we explored the environmental constraints that favoured the occurrences of leishmaniasis via ecological niche models (ENM) to predict suitable environmental conditions that favour the transmission of leishmaniasis in Afghanistan. We used the ENM technique to predict the areas of potential high risk for leishmaniasis disease using maximum entropy modelling (MaxEnt model) [30,31,34]. Previous authors have used ENM to estimate environmental suitability index for leishmaniasis [19,49]; however, to the best of our knowledge, this the first study that provides a suitability index for leishmaniasis in Afghanistan.The ENM estimates indicate that the eastern and northeastern regions are potential high risk areas for leishmaniasis in Afghanistan. Precipitation and temperature provide significant contribution in predicting the suitable condition for leishmaniasis, with average precipitation in April contributing the greatest. This is similar to studies that found precipitation to contribute significantly to the predictive probability of the presence of leishmaniasis [19,41] and sandflies [55].The impact of environmental influences on leishmaniasis cannot be ruled out and human activities play a significant role in the dispersion of the vectors, thereby changing the geographical distribution of the disease. In our study, precipitation in the month of April provided the greatest contribution to the ecological niche of leishmaniasis in Afghanistan. The use of geographical information system and remotely sensed satellite derived environmental layers to infectious disease modelling have been used in this study. The major limitation of this study is the use of aggregated leishmaniasis surveillance data in which we were unable to distinguish between ACL and ZCL entities. Actually, ACL lesions have long incubation and chronic persistence, whereas ZCL lesions are characterized by shorter incubation and rapid healing, which may have influence on seasonality of diagnosis. Moreover, the disease incidences are calculated from passive case detection data, and, therefore, the spatial distribution of the disease could be significantly influenced by the distribution of public health services. This study is solely based on the ecological aspect of disease transmission. Our results could be validated by incorporating sandfly and rodent distribution data. However, findings in this study serve as a versatile tool for studying and understanding transmission and spread of leishmaniasis. The model may be useful in different frontiers of the disease upsurge and possibility of its containment or eradication.We thank the three anonymous reviewers for their insightful comments and suggestions which helped improve the manuscript. The authors are grateful to Tara Pylate for her editorial feedback and carefully reading the manuscript. We also thank Qatar National Library for providing Open Access Author Fund.Oyelola A. Adegboye conceived and designed the study; Majeed Adegboye extracted the environmental layers; Oyelola A. Adegboye and Majeed Adegboye analyzed the data; and Oyelola A. Adegboye and Majeed Adegboye wrote the manuscript.The authors declare no conflict of interest.Map of Afghanistan showing aggregated incidence per 100,000 in all provinces in the study period (2003–2009). The white dotted areas indicate provinces without data.Decomposition of leishmaniasis time series into additive components. From the top panel to the bottom panel, time series plots of the disease, trend components, and seasonal components are indicated.Autocorrelation plot of the “lag” (time span between observations) and the autocorrelation. The blue lines indicated 95% bounds for statistical significance.Map of random intercepts in the endemic component. The colour indicates the level of endemicity in each provinces after adjusting for variables such population and environmental layers. The bi-chromatic range from dark blue to yellow indicates low relative endemic incidence to high endemic incidence.Plots of predicted values for cutaneous leishmaniasis in six selected provinces in Afghanistan showing the relative contributions of the three components (spatiotemporal, autoregressive and endemic) in Model 3. The dots represent the observed disease counts.The ecological niche modelling of leishmaniasis in Afghanistan using MaxEnt: Predicted occurrence probability map of cutaneous leishmaniasis in Afghanistan bases on outbreak data with environmental variables in Table 2. Dots represent the administrative headquarters points used for the ENM. Warmer colors in the habitat suitability map show areas with better predicted conditions.Parameter estimates (95% 
+            
+              
+                
+                  
+                  ‡
+                
+              
+            
+          CI) from the five models fitted to the leishmaniasis data.
+                
+                  
+                    
+                      
+                      ‡
+                    
+                  
+                
+              CI: Confidence Interval.Estimates of relative contributions of the environmental variables to the Maxent model.

Med-MDPI/ijerph_2/ijerph-14-03-00310.txt

@@ -0,0 +1 @@
+DeceasedLicensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).This study examines aspects of prediction of suicide and death of undetermined intent. We investigated all consecutive, autopsied patients between 1993 and 1997 who had been in contact with the Addiction Centre in Malmö from 1968 onwards. The staff was asked, shortly after autopsy but before they knew of the manner of death, if they thought the patient had committed suicide. The case records were blindly evaluated, and toxicological autopsy findings for alcohol in blood samples investigated. The specificity of prediction was 83% and significantly more often correct than the sensitivity, which was only 45% for suicide and for suicide/death of undetermined intent (93% versus 39%). Suicidal communication was more often considered non-serious before death of undetermined intent than before suicide. The former could be predicted by ideation but not by suicide attempt reported in case records, unlike suicide, which was predicted by both. The undetermined group also showed higher levels of alcohol in the blood at autopsy. We concluded that more serious clinical investigation of suicidal feelings, which may be hidden and not taken seriously, and treatment of alcohol use disorders with active follow-up appear urgent in the efforts to prevent suicide.Suicide is a major health problem, with more than 800,000 people killing themselves every year [1]. The prediction of accomplished suicide in order to prevent it is urgent but difficult. Different rating scales have been developed, such as the Scale of Suicidal Ideation (SSI) [2], the Suicidal Intent Scale (SIS) [3], and the Sad Persons Scale [4]. However, a recent review of the predictive value for future suicide attempts has not been encouraging [5], with the conclusion that none of the known rating scales reached a sensitivity of 80%, nor was a specificity of 50% reached. Clinical judgement as a complement to rating scales was strongly recommended. Typical scales for suicide prediction can be less effective when a respondent has alcohol or other substance use disorders, so additional clinical judgement is warranted.Alcohol use disorders are also commonly found among suicide victims [6,7,8]. These are usually regarded as the second most common diagnosis among suicide victims after depression, and have been related to deaths of undetermined intent [8,9,10]. In population-based surveys, suicide and death of undetermined intent are usually combined in the analyses [11,12], and similarities have been shown between the two groups in several studies [13,14,15]. However, although there are similarities between these manners of death, differences in background variables have been highlighted recently [10], as well as similarities between accidental overdoses and death with undetermined intent rather than with suicide in substance use disorders [16].Alcohol and other substance use disorders also show a high risk of suicide [17,18]. Suicidal ideation and suicide attempts are also common among people with alcohol use disorders [19]. However, predictors such as suicide attempts and suicidal ideation are often considered less serious among people with alcohol use disorders, especially if occurring during drinking and intoxication. Though attempted suicide has also been shown to be a highly significant risk factor for completed suicide in young males with alcohol use disorders, this group was found to have a significantly lower risk of completed suicide than other suicide attempters [20]. Likewise, the suicide risk in suicide attempters who had recently consumed alcohol was assessed as less severe, and they were less often referred to a psychiatrist compared with those who had not [21]. Other investigators have found that a correlation between suicidal intent and the lethality of the suicide attempt was seen only among patients without a diagnosis of alcohol dependence [22]. Low scores on the SIS when alcohol was used prior to self-harm have also been shown [23].However, other authors have pointed out the triggering effect of alcohol on suicidal behaviour. According to one study, 50% of attempted suicides happened within one hour of alcohol use [24]. Attempted suicide can be triggered by alcohol [25], and alcohol use has been associated with a faster transition from suicidal impulse to action [26]. High levels of alcohol at autopsy have been found in people with brittle/sensitive personalities [27], and a dose-response relationship between the number of drinks and suicidal behaviour has also been shown [28].The present study is based on a sample of patients who had been treated at the Addiction Centre in Malmö from 1968 onwards, who died in the period from 1993 to 1997, and who were autopsied at the Department of Forensics. The sample has been presented in a previous study on unnatural death and drugs used in life [29]. This special sample is now revisited, with the study inspired by recent findings and discussions of the limited sensitivity and specificity of known rating scales for suicide risk. Complementary clinical evaluation was recommended.The first aim was to investigate the accuracy of the staff’s clinical prediction of suicide and death of undetermined intent, and possible differences between these manners of death. A second aim was to investigate, using case record evaluation, the predictive value of suicidal ideation and attempt of suicide and death by undetermined intent. Finally, alcohol levels at autopsy by those manners of death were compared.A forensic examination sampling procedure was used. The procedure was carried out on all consecutive autopsies of patients who had been in contact with the Addiction Centre in Malmö University Hospital. In Sweden, a forensic examination is carried out on most people who have died outside hospitals by suspected natural causes (disease) but with no medical history that can explain the death, or by unnatural manner (trauma including homicide, suicide, death of undetermined intent, and accidental fatal intoxications).In all, 388 consecutive forensic autopsies on previous patients at the Department of Forensic Medicine in Lund from 1993 to 1997 inclusive were investigated (Figure 1), as well as an investigation of case records from 1968 onwards. The sampling was carried out in the 1990s, but there have been no significant changes in methodology since then, and changes in the epidemiology of deliberate self-harm do not apply to suicidal ideation and attempt investigated in the present sample.Substance use was diagnosed according to International Classification of Disease (ICD 9 and 10) [30,31] 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. It is safe to conclude that they all fulfilled the criteria for alcohol dependence and/or had a drug problem. From 1968 to 1994, all the patients treated at the Department of Clinical Alcohol Research were admitted for alcohol problems; after that date, the clinic became an Addiction Centre, which also received patients with narcotic misuse. Some of them may not have had an alcohol problem, but in only seven cases could alcohol use disorder not be confirmed (1.8% of the total sample), though it could be suspected. A total of 89 patients had used illegal drugs (some legal drugs as well) and another 73 had used legal drugs. A previous study had shown that the number of drugs used in life increased the risk of death of undetermined intent but not suicide [29]. However, number of drugs detected at autopsy showed similar rates for undetermined intent and suicide.The interviews were performed within a few days of the patient’s death, with nurses and nursing assistants who had previously had contact with the patient. Most of the services provided involved psychosocial interventions by nurses and registered nurses; the group of physicians was small and the younger physicians had often only worked at the department for a few months. We therefore decided to interview nurses/registered nurses to ensure that we acquired the most reliable information.As the interviews were performed shortly after death, the interviewer and the interviewees did not know the manner of death. The staff remembered 157 patients, as expected more often those with a recent contact (Table 1). However, there was no difference in remembering manner of death. The suicidal outcome as judged by staff was dichotomized into ‘yes’ or ‘no’, but they sometimes stated that it was ‘only a threat’ or ‘just when drunk’, which made up a third category ‘intent not considered serious’.The entire records were evaluated for those who had been in- or outpatients at the Addiction Centre in Malmö University Hospital from first admission and onwards. Consequently, these ratings were not biased by any knowledge of the manner of death. The ratings of suicidal ideation and attempt were used in the present study.There were similar rates of patients who had been in contact with the clinic within the previous three months, regardless of manner of death. Furthermore, there was no evidence of more people seeking help in later contact in future suicide victims or those who died by undetermined intent (27% of the suicide cases, 30% of the undetermined cases, and 23% of the others, see Table 2. There were also similar numbers for other time intervals, one year, five years, etc. We chose to include all remembered patients, regardless of time passed since last contact, and then compared those with recent contact with more distant.Femoral blood samples of alcohol concentration had been collected for in 43/45 cases of suicide and in 86/91 of deaths by undetermined intent.A logistic regression with odds ratio (OR) and confidence interval (CI) was used to relate suicidal ideation and attempt to undetermined death and completed suicide. Fisher’s exact test was used for comparison between groups. Student’s t-test was used for comparison of continuous variables.Ethical approval was not required for deceased persons in Sweden at that time. However, the National Board of Forensic Medicine approved the study by oral confirmation number 1993.Suicide predicted by staff and ‘intent not considered serious’ were compared for suicides and death of undetermined intent (Table 3). A significantly higher incidence of ‘intent not considered serious’ was found in the undetermined group (8/17 versus 0/9, Fisher’s exact test p = 0.023).Sensitivity was 9/20 (45%) for the prediction of future suicide, but specificity was high 114/137 (83%). The latter prediction was significantly more often correct as compared to the prediction of suicide (p < 0.0001).The sensitivity was 18/67 (26%) if suicide and death of undetermined intent were aggregated. If ‘intent not considered serious’ was included, the sensitivity was higher 26/67 (39%), more similar to suicide, and the specificity was also high, 84/90 (93%).The prediction of no suicide/death of undetermined intent was significantly more often correct, as compared to the prediction for suicide/undetermined intent (84/90 versus 26/67, p < 0.0001).The figures were similar for those with recent contact within a year and for the total group.The prediction of neither suicide nor death of undetermined intent was 20/20 (100%), while the prediction of suicide/death of undetermined intent was 7/18 (39%) (p < 0.0001).The association between suicidal behaviour (ideation and attempt) reported in the case records and manner of death is presented in Table 4. The sensitivity of the prediction of suicide by suicide ideation was 36%, and the specificity was 84%. The sensitivity of prediction by suicide attempt was 33% and the specificity was 83%. The sensitivity of prediction of suicide/undetermined death by suicidal ideation was 29% and the specificity was 88%. The sensitivity of prediction of suicide/undetermined death by suicide attempt was 26% and the specificity was 83%. In all cases the sensitivity was lower than the specificity.The sensitivity ranged from 26% (suicide attempt in case records for suicide/undetermined intent) to 45% (prediction by staff, regardless of time span since last contact.)The specificity ranged from 83% (suicide attempt in case records for suicide) to 100% (prediction by staff within a year before suicide). Thus, a somewhat better prediction was made by staff.A logistic regression was performed to assess the impact of suicidal ideation and suicide attempt on suicide and undetermined death. Suicidal ideation was related to both suicide (OR: 2.95, CI = 1.50–5.81, p = 0.002) and death of undetermined intent (OR: 1.92, CI = 1.09–3.37, p = 0.023). However, suicide attempt was only related to suicide (2.41, CI = 1.22–4.75, p = 0.011), and there was no significant correlation with death of undetermined intent.The blood levels of alcohol were higher in death of undetermined intent (1.79 per mille) as compared to suicide (0.72 per mille) (t-test p = 0.0001).In the suicide group, alcohol was detected in 25/43 (58%) versus 60/82 (73%) in the undetermined group, a non-significant difference. One person with a positive test in urine was included.The present study considers clinical prediction without the use of any rating scales. This prediction did not seem to be very accurate for suicide, less than 50% with or without the inclusion of death of undetermined intent. On the other hand, the specificity was high, and the prediction that persons would not commit suicide was often correct (83%–93%). When suicide attempt or ideation were used as predictors, a similar sensitivity and specificity showed the same relationship with a higher specificity.However, the highest specificity was found by staff intuition within a year before death.This contradicts the conclusion by the Swedish Council on Health Technology Assessment (SBU) review of questionnaires [5], which showed a low specificity and higher sensitivity, mostly including repeated suicide attempts rather than completed suicide. A meta-analysis of suicide risk within a year after discharge showed that 60% of future suicide victims were considered to be at low risk at discharge, so 40% were high risk, similar to the present finding (39%–45%) [32]. Prediction by staff intuition could be a complement to rating scales.The poor prediction of suicide may reflect unawareness of life events occurring after last contact that may trigger suicide. The staff could judge resilience only, which corresponds to trait factors in suicide risk but not state as described by Goldston et al. [33]. Better follow-up may improve the possibility of providing support in the event of distressing life events.The better prediction of survival may also reflect better contact. The staff seemed to know who would survive, but they did not know who would commit suicide, and prediction had escaped their attention. Ringel [34] proposed a presuicidal syndrome, which included ‘Einengung’ or constriction of human relationships and values. In this state, the person may be very much alone and not communicative with others. Furthermore, in an investigation of the long-term course of depression after a suicide attempt [35], some subjects pointed out that the decision to continue living was a very private one, not necessarily communicated with others. Therefore, the decision to commit suicide may very well be a decision taken in a lonely state.The present findings support both views, which may be an explanation for the poor sensitivity and a need to be more open to exploring suicidal feelings and existential issues among the patients. Suicidal communication was sometimes not taken seriously. This type of communication was related to death of undetermined intent rather than suicide. These patients may themselves have been less serious in their intent, and jeopardised their lives with a fatal outcome. Worthy of note is that, overall, they also had higher levels of alcohol in their blood samples at autopsy. Heavy drinking leads to loss of inhibitions and risk-taking, which may trigger self-inflicted death despite less serious intent. It has been shown that those who die by undetermined intent more frequently have alcohol in the blood at autopsy, 62% versus 35% [36], as compared to 73% versus 58% in the present sample, a non-significant difference. Only people with alcohol use disorder were included in the present study, which may explain the higher rates in the suicide group.The present findings support the hypothesis of the triggering effect of alcohol on suicide and death of undetermined intent [25,26], especially the latter.Suicide attempt was related to completed suicide, but no relation could be shown with death by undetermined intent. This is in agreement with a recent study [10], which showed a correlation between hospitalisation for self-harm and later suicide, but not undetermined death, in the female group. Other investigators have found suicidal threats (34%) and previous suicide attempts (31%) in cases of undetermined intent, but the sample only included 31% with an alcohol problem [37]. The present findings from the case records are compatible with the results from the staff interviews and autopsy findings of higher blood levels. It indicates a less serious intent in undetermined cases, though self-inflicted death may be triggered by alcohol.The present study supports the view of a continuum from more ambivalent suicidality in the case of death of undetermined intent to less ambivalent suicidality in the case of suicide, as proposed by other authors in a multicentre study of a general population of self-inflicted deaths [38]. Alcohol use seems to trigger more serious suicidal behaviour. More knowledge is needed about suicidal behaviour as a predictor of death by undetermined intent, and a more thorough clinical investigation of suicidal intent.The staff knew that the patient was dead, which may have impacted their judgement. The impact of the knowledge is not known. Furthermore, anyone using a rating scale will probably undertake preventative measures, which was not possible in the present study, reducing that confounder.Another limitation (also often inherent in rating scales) is that the staff did not know the life events that may occur and trigger an accomplished suicide. The impact of traits (such as disease, personality) and states (such as life events) have been discussed in the context of suicide [33]. In most cases only traits or resilience could be judged.Some people were not remembered, possibly because they had not been to the clinic shortly before death. There were at least similar rates for those who had been in contact within a year and after more than a year, so the effect of recall bias due to time lapse since last contact does not appear to affect the results. However, there is some recall bias due to closer relationships with the staff or more serious illness, which we cannot control for. No scales were used in the present study, so the comparison between staff judgement was made against findings in literature.The major strength in the present study was the pseudo-prospective design and the use of multiple sources of data, clinically from interviews with staff and case records, as well as autopsy findings.The implications of the present study are that clinical predictions by people who know the patient are good without any systematic inquiry (83%–93%) in the case of deciding who is not going to commit suicide. This reflects their intuition about the patient’s resilience. However, the prediction of suicide was poor, less than chance (39%–45%). Suicide scales have also appeared to be inadequate. We propose more active clinical inquiry of suicidal tendencies, especially as people tend to be very private about their suicidal feelings, and also more active follow-up.Suicidal ideation should be taken more seriously among people with substance use disorders, including those regarded as ‘just a threat’ or ‘only when drunk’. Suicidal ideation of all levels may be predictive of death of undetermined intent, and alcohol appears to trigger fatal suicidal behaviour. Vigorous treatment of alcohol use disorders is also urgent in the ambition to prevent suicide and death by undetermined intent.Governmental funding of clinical research within the Swedish NHS (National Health Service) and Ellen and Henrik Sjöbring’s Memorial Foundation supported the study. Anna Lindgren provided statistical advice. Leslie Walke revised the language.M.B. and P.L. designed the study. A.F. performed the interviews and the case record evaluation. M.B. and L.B. analysed the data. L.B. wrote the paper.The authors declare no conflict of interest.Sampling procedure (modified after [29]).Length of time since last contact with the clinic before death, total sample and patients remembered by staff (%).In three cases the last date of contact was not known.Time since last contact with the clinic by manner of death.In three cases the last date of contact was not known.Suicidal outcome by manner of death according to staff’s judgement.* intent not considered serious versus suicidal outcome; undetermined death versus suicide, p = 0.023.Suicidal ideation and suicide attempt by manner of death according to case record evaluation in the long-term course.

Med-MDPI/ijerph_2/ijerph-14-03-00311.txt

@@ -0,0 +1,5 @@
+Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).Online social networking sites (SNSs) have gained increasing popularity in the last decade, with individuals engaging in SNSs to connect with others who share similar interests. The perceived need to be online may result in compulsive use of SNSs, which in extreme cases may result in symptoms and consequences traditionally associated with substance-related addictions. In order to present new insights into online social networking and addiction, in this paper, 10 lessons learned concerning online social networking sites and addiction based on the insights derived from recent empirical research will be presented. These are: (i) social networking and social media use are not the same; (ii) social networking is eclectic; (iii) social networking is a way of being; (iv) individuals can become addicted to using social networking sites; (v) Facebook addiction is only one example of SNS addiction; (vi) fear of missing out (FOMO) may be part of SNS addiction; (vii) smartphone addiction may be part of SNS addiction; (viii) nomophobia may be part of SNS addiction; (ix) there are sociodemographic differences in SNS addiction; and (x) there are methodological problems with research to date. These are discussed in turn. Recommendations for research and clinical applications are provided.The history of social networking sites (SNSs) dates back to 1997, when the first SNS SixDegrees emerged as a result of the idea that individuals are linked via six degrees of separation [1], and is conceived as “the small world problem” in which society is viewed as becoming increasingly inter-connected [2]. In 2004, Facebook, was launched as an online community for students at Harvard University and has since become the world’s most popular SNS [3]. In 2016, there were 2.34 billion social network users worldwide [4]. In the same year, 22.9% of the world population used Facebook [5]. In 2015, the average social media user spent 1.7 h per day on social media in the USA and 1.5 h in the UK, with social media users in the Philippines having the highest daily use at 3.7 h [6]. This suggests social media use has become an important leisure activity for many, allowing individuals to connect with one another online irrespective of time and space limitations.It is this kind of connecting or the self-perceived constant need to connect that has been viewed critically by media scholars. Following decades of researching technology-mediated and online behaviors, Turkle [7] claims overreliance on technology has led to an impoverishment of social skills, leaving individuals unable to engage in meaningful conversations because such skills are being sacrificed for constant connection, resulting in short-term attention and a decreased ability to retain information. Individuals have come to be described as “alone together”: always connected via technology, but in fact isolated [8]. The perceived need to be online may lead to compulsive use of SNSs, which in extreme cases may result in symptoms and consequences traditionally associated with substance-related addictions. Since the publication of the first ever literature review of the empirical studies concerning SNS addiction in 2011 [3], the research field has moved forward at an increasingly rapid pace. This hints at the scientific community’s increasing interest in problematic and potentially addictive social networking use. In order to present new insights into online social networking and addiction, in this paper, 10 lessons learned concerning online social networking sites and addiction based on the insights derived from recent empirical research will be presented. These are: (i) social networking and social media use are not the same; (ii) social networking is eclectic; (iii) social networking is a way of being; (iv) individuals can become addicted to using social networking sites; (v) Facebook addiction is only one example of SNS addiction; (vi) fear of missing out (FOMO) may be part of SNS addiction; (vii) smartphone addiction may be part of SNS addiction; (viii) nomophobia may be part of SNS addiction; (ix) there are sociodemographic differences in SNS addiction; and (x) there are methodological problems with research to date. These are discussed in turn.Social networking and social media use have often been used interchangeably in the scientific literature. However, they are not the same. Social media refers to the web 2.0 capabilities of producing, sharing, and collaborating on content online (i.e., user-generated content, implying a social element). Accordingly, social media use includes a wide range of social applications, such as collaborative projects, weblogs, content communities, social networking sites, virtual game worlds, and virtual social worlds [9], each of which will be addressed below.Collaborative projects can be shared and worked on jointly and simultaneously using cloud-based computing. Two different types can be distinguished: Wikis allow for creating, removing and modifying online content (e.g., Wikipedia). Social bookmarking applications, on the other hand, allow for numbers of people to accumulate and appraise websites (e.g., Delicious). Taken together, collaborative projects may produce a superior end result in comparison to individual projects [9], which can be linked to the concept of collective intelligence, whereby the intelligence in the group is greater than the sum of its parts [10].Weblogs (or “blogs”) can also be considered social media. Blogs allow individuals to share personal online diaries and information (sometimes in the form of images and videos), which may or may not be commented upon by other internet users. Next, there are content communities and video-sharing sites (e.g., YouTube). Content may include videos, but also text (e.g., BookCrossing), photographs (e.g., Instagram), and PowerPoint presentations (e.g., Slideshare), and in most cases, there is no a need for individuals to have personal profiles, and if they do, these tend to include limited personal information. Virtual game worlds allow users to create an online alter ego in the form of an avatar and to play with other players in large gaming universes (and the next section covers gaming in more detail). Kaplan and Haenlein [9] differentiate these from virtual social worlds from virtual game worlds, whereby the former allow individuals to create online characters which live in an alternative virtual world that is similar to their real life environments on the one hand, but defies physical laws. Arguably the best example of these virtual social worlds is Second Life, populated by human-like avatars, who engage in activities users engage in on an everyday basis, such as furnishing houses, going shopping, and meeting friends.Finally, there are social networking sites, which we have previously defined as “virtual communities where users can create individual public profiles, interact with real-life friends, and meet other people based on shared interests” ([3]; p. 3529). Social networking is particularly focused on connecting people, which does not apply to a number of the other social media applications outlined above. Engaging in social networking comprises a specific type of social media use, therefore they are not synonymous. Consequently, studies that have examined social media addiction and social networking addiction may also be using the terms interchangeably, suggesting nosological imprecision.Despite social networking being one type of social media use (as outlined in the previous section), the behavior is inherently eclectic because it includes a variety of apps and services that can be engaged in. For instance, social networking can be the use of traditional social networking sites, such as Facebook. Facebook can be considered an ‘egocentric’ SNS (rather than the previously more common virtual communities that focused on shared interests between members) because it allows individuals to represent themselves using individual profiles and wall posts. These can contain text and audiovisual content, whilst connecting to friends who often appear as real life friends and acquaintances given the main motivation of individuals to use SNSs such as Facebook is to maintain their connections [3].In 2016, the most popular social networking site was Facebook with 1712 million active users [5]. Facebook has long established its supremacy in terms of active members, with membership numbers steadily increasing by 17%–20% annually [11]. Facebook is a very active network. Every minute, 510,000 comments are posted; 293,000 statuses are updated; and 136,000 photos are uploaded, whilst the average user spends approximately 20 min daily on the site [11].Over the past few years, new networks have emerged that have gradually risen in popularity, particularly amongst younger generations. Instagram was launched in 2010 as a picture sharing SNS, claiming to “allow you to experience moments in your friends’ lives through pictures as they happen” [12]. In 2016, Instagram had 500 m active users [5]. Snapchat was launched in 2011 [13] as an SNS that allows users to message and connect with others using a smartphone and to send texts, videos, and make calls. Snapchat is different from other networks in that it has an inherently ephemeral nature, whereby any messages are automatically deleted shortly after the receiver has viewed them, allowing an increased experience of perceived privacy and safety online [14]. However, teenagers are especially aware of the transitory nature of Snapchat messages and therefore take screenshots and keep them stored on their mobile phones or in the cloud, simply to have proof of conversations and visuals spread on this medium. The privacy advantage of the medium is thereby countered. Snapchat had 200 million users in 2016 [5]. In the same year, Snapchat was the most popular SNS among 13–24 year-old adolescents and adults in the USA, with 72% of this group using them, followed by 68% Facebook users, and 66% Instagram users [15]. The popularity of Snapchat—particularly among young users—suggests the SNS landscape is changing in this particular demographic, with users being more aware of potential privacy risks, enjoying the lack of social pressure on Snapchat as well as the increased amount of control over who is viewing their ephemeral messages. However, it could also be the case that this may lead to the complete opposite by increasing the pressure to be online all the time because individuals risk missing the connecting thread in a continuing stream of messages within an online group. This may be especially the case in Snapchat groups/rooms created for adolescents in school or other contexts. This can lead to decreasing concentration during preparation tasks for school at home, and may lead to constant distraction because of the pressure to follow what is going on as well as the fear of missing out. From a business point of view, Snapchat has been particularly successful due to its novel impermanent approach to messaging, with Facebook founder Mark Zuckerberg offering $3 billion to buy the SNS, which has been declined by Evan Spiegel, Snapchat’s CEO and co-founder [13]. These facts suggest the world of traditional SNS is changing.Social networking can be instant messaging. The most popular messaging services to date are WhatsApp and Facebook Messenger with 1000 million active users each [5]. WhatsApp is a mobile messaging site that allows users to connect to one another via messages and calls using their internet connection and mobile data (rather than minutes and texts on their phones), and was bought by Facebook in 2014 for $22 billion [16], leading to controversies about Facebook’s data sharing practices (i.e., Whatsapp phone numbers being linked with Facebook profiles), resulting in the European Commission fining Facebook [17]. In addition to WhatsApp, Facebook owns their own messaging system, which is arguably the best example of the convergence between traditional SNS use and messaging, and which functions as an app on smartphones separate from the actual Facebook application.Social networking can be microblogging. Microblogging is a form of more traditional blogging, which could be considered a personal online diary. Alternatively, microblogging can also be viewed as an amalgamation of blogging and messaging, in such a way that messages are short and intended to be shared with the writer’s audience (typically consisting of ‘followers’ rather than ‘friends’ found on Facebook and similar SNSs). A popular example of a microblogging site is Twitter, which allows 140 characters per Tweet only. In 2016, Twitter had 313 million active users [5], making it the most successful microblogging site to date. Twitter has become particularly used as political tool with examples including its important role in the Arab Spring anti-government protests [18], as well as extensive use by American President Donald Trump during and following his presidential campaign [19]. In addition to microblogging politics, research has also assessed the microblogging of health issues [20].Social networking can be gaming. Gaming can arguably be considered an element of social networking if the gaming involves connecting with people (i.e., via playing together and communicating using game-inherent channels). It has been argued that large-scale internet-enabled games (i.e., Massively Multiplayer Role-Playing Games [MMORPGs]), such as the popular World of Warcraft, are inherently social games situated in enormous virtual worlds populated by thousands of gamers [21,22], providing gamers various channels of communication and interaction, and allowing for the building of relationships which may extend beyond the game worlds [23]. By their very nature, games such as MMORPGs are “particularly good at simultaneously tapping into what is typically formulated as game/not game, social/instrumental, real/virtual. And this mix is exactly what is evocative and hooks many people. The innovations they produce there are a result of MMOGs as vibrant sites of culture” [24]. Not only do these games offer the possibility of communication, but they provide a basis for strong bonds between individuals when they unite through shared activities and goals, and have been shown to facilitate and increase intimacy and relationship quality in couples [25] and online gamers [22,23]. In addition to inherently social MMORPGs, Facebook-enabled games—such as Farmville or Texas Hold “Em Poker”—can be subsumed under the social networking umbrella if they are being used in order to connect with others (rather than for solitary gaming purposes) [26,27].Social networking can be online dating. Presently, there are many online dating websites available, which offer their members the opportunity to become part of virtual communities, and they have been especially designed to meet the members’ romantic and relationship-related needs and desires [28]. On these sites, individuals are encouraged to create individual public profiles, to interact and communicate with other members with the shared interest of finding a ‘date’ and/or long-term relationships, therewith meeting the present authors’ definition of SNS. In that way, online dating sites can be considered social networking sites. However, these profiles are often semi-public, with access granted only to other members of these networks and/or subscribers to the said online dating services. According to the US think tank Pew Research Center’s Internet Project [29], 38% of singles in the USA have made use of online dating sites or mobile dating applications. Moreover, nearly 60% of internet users think that online dating is a good way to meet people, and the percentage of individuals who have met their romantic partners online has seen a two-fold increase over the last years [29]. These data suggest online dating is becoming increasingly popular, contributing to the appeal of online social networking sites for many users across the generations. However, it can also be argued that online dating sites such as Tinder may be less a medium for ‘long-term relationships’, given that Tinder use can lead to sexual engagement. This suggests the uses and gratifications perspective underlying Tinder use points more in the direction of other motives, such as physical and sexual aspirations and needs, rather than purely romance.Taken together, this section has argued that social networking activities can comprise a wide variety of usage motivations and needs, ranging from friendly connection over gaming to romantic endeavors, further strengthening SNS’ natural embeddedness in many aspects of the everyday life of users. From a social networking addiction perspective, this may be similar to the literature on Internet addiction which often delineates between addictions to specific applications on the Internet (e.g., gaming, gambling, shopping, sex) and more generalized Internet addiction (e.g., concerning problematic over-use of the Internet comprising many different applications) [30,31].In the present day and age, individuals have come to live increasingly mediated lives. Nowadays, social networking does not necessarily refer to what we do, but who we are and how we relate to one another. Social networking can arguably be considered a way of being and relating, and this is supported by empirical research. A younger generation of scholars has grown up in a world that has been reliant on technology as integral part of their lives, making it impossible to imagine life without being connected. This has been referred to as an ‘always on’ lifestyle: “It’s no longer about on or off really. It’s about living in a world where being networked to people and information wherever and whenever you need it is just assumed” [32]. This has two important implications. First, being ‘on’ has become the status quo. Second, there appears to be an inherent understanding or requirement in today’s technology-loving culture that one needs to engage in online social networking in order not to miss out, to stay up to date, and to connect. Boyd [32] herself refers to needing to go on a “digital sabbatical” in order not be on, to take a vacation from connecting, with the caveat that this means still engaging with social media, but deciding which messages to respond to.In addition to this, teenagers particularly appear to have subscribed to the cultural norm of continual online networking. They create virtual spaces which serve their need to belong, as there appear to be increasingly limited options of analogous physical spaces due to parents’ safety concerns [33]. Being online is viewed as safer than roaming the streets and parents often assume using technology in the home is normal and healthy, as stated by a psychotherapist treating adolescents presenting with the problem of Internet addiction: “Use of digital media is the culture of the household and kids are growing up that way more and more” [34]. Interestingly, recent research has demonstrated that sharing information on social media increases life satisfaction and loneliness for younger adult users, whereas the opposite was true for older adult users [35], suggesting that social media use and social networking are used and perceived very differently across generations. This has implications for social networking addiction because the context of excessive social networking is critical in defining someone as an addict, and habitual use by teenagers might be pathologized using current screening instruments when in fact the activity—while excessive—does not result in significant detriment to the individual’s life [36].SNS use is also driven by a number of other motivations. From a uses and gratifications perspective, these include information seeking (i.e., searching for specific information using SNS), identity formation (i.e., as a means of presenting oneself online, often more favorably than offline) [37], and entertainment (i.e., for the purpose of experiencing fun and pleasure) [38]. In addition to this, there are the motivations such as voyeurism [39] and cyberstalking [40] that could have potentially detrimental impacts on individuals’ health and wellbeing as well as their relationships.It has also been claimed that social networking meets basic human needs as initially described in Maslow’s hierarchy of needs [41]. According to this theory, social networking meets the needs of safety, association, estimation, and self-realization [42]. Safety needs are met by social networking being customizable with regards to privacy, allowing the users to control who to share information with. Associative needs are fulfilled through the connecting function of SNSs, allowing users to ‘friend’ and ‘follow’ like-minded individuals. The need to estimate is met by users being able to ‘gather’ friends and ‘likes’, and compare oneself to others, and is therefore related to Maslow’s need of esteem. Finally, the need for self-realization, the highest attainable goal that only a small minority of individuals are able to achieve, can be reached by presenting oneself in a way one wants to present oneself, and by supporting ‘friends’ on those SNSs who require help. Accordingly, social networking taps into very fundamental human needs by offering the possibilities of social support and self-expression [42]. This may offer an explanation for the popularity of and relatively high engagement with SNSs in today’s society. However, the downside is that high engagement and being always ‘on’ or engaged with technology has been considered problematic and potentially addictive in the past [43], but if being ‘always on’ can be considered the status quo and most individuals are ‘on’ most of the time, where does this leave problematic use or addiction? The next section considers this question.There is a growing scientific evidence base to suggest excessive SNS use may lead to symptoms traditionally associated with substance-related addictions [3,44]. These symptoms have been described as salience, mood modification, tolerance, withdrawal, relapse, and conflict with regards to behavioral addictions [45], and have been validated in the context of the Internet addiction components model [46]. For a small minority of individuals, their use of social networking sites may become the single most important activity that they engage in, leading to a preoccupation with SNS use (salience). The activities on these sites are then being used in order to induce mood alterations, pleasurable feelings or a numbing effect (mood modification). Increased amounts of time and energy are required to be put into engaging with SNS activities in order to achieve the same feelings and state of mind that occurred in the initial phases of usage (tolerance). When SNS use is discontinued, addicted individuals will experience negative psychological and sometimes physiological symptoms (withdrawal), often leading to a reinstatement of the problematic behavior (relapse). Problems arise as a consequence of the engagement in the problematic behavior, leading to intrapsychic (conflicts within the individual often including a subjective loss of control) and interpersonal conflicts (i.e., problems with the immediate social environment including relationship problems and work and/or education being compromised).Whilst referring to an ‘addiction’ terminology in this paper, it needs to be noted that there is much controversy within the research field concerning both the possible overpathologising of everyday life [47,48] as well as the most appropriate term for the phenomenon. On the one hand, current behavioral addiction research tends to be correlational and confirmatory in nature and is often based on population studies rather than clinical samples in which psychological impairments are observed [47]. Additional methodological problems are outlined below (Section 2.10). On the other hand, in the present paper, the present authors do not discriminate between the label addiction, compulsion, problematic SNS use, or other similar labels used because these terms are being used interchangeably by authors in the field. Nevertheless, when referring to ‘addiction’, the present authors refer to the presence of the above stated criteria, as these appear to hold across both substance-related as well as behavioral addictions [45] and indicate the requirement of significant impairment and distress on behalf of the individual experiencing it in order to qualify for using clinical terminology [49], such as the ‘addiction’ label.The question then arises as what it is that individuals become addicted to. Is it the technology or is it more what the technology allows them to do? It has been argued previously [34,50] that the technology is but a medium or a tool that allows individuals to engage in particular behaviors, such as social networking and gaming, rather than being addictive per se. This view is supported by media scholars: “To an outsider, wanting to be always-on may seem pathological. All too often it’s labelled an addiction. The assumption is that we’re addicted to the technology. The technology doesn’t matter. It’s all about the people and information” [32]. Following this thinking, one could claim that it is not an addiction to the technology, but to connecting with people, and the good feelings that ‘likes’ and positive comments of appreciation can produce. Given that connection is the key function of social networking sites as indicated above, it appears that ‘social networking addiction’ may be considered an appropriate denomination of this potential mental health problem.There are a numbers of models which offer explanations as to the development of SNS addiction [51]. According to the cognitive-behavioral model, excessive social networking is the consequence of maladaptive cognitions and is exacerbated through a number of external issues, resulting in addictive use. The social skill model suggests individuals use SNSs excessively as a consequence of low self-presentation skills and preference for online social interaction over face-to-face communication, resulting in addictive SNS use [51]. With respect to the socio-cognitive model, excessive social networking develops as a consequence of positive outcome expectations, Internet self-efficacy, and limited Internet self-regulation, leading to addictive SNS use [51]. It has furthermore been suggested that SNS use may become problematic when individuals use it in order to cope with everyday problems and stressors, including loneliness and depression [52]. Moreover, it has been contended that excessive SNS users find it difficult to communicate face-to-face, and social media use offers a variety of immediate rewards, such as self-efficacy and satisfaction, resulting in continued and increased use, with the consequence of exacerbating problems, including neglecting offline relationships, and problems in professional contexts. The resultant depressed moods are then dealt with by continued engagement in SNSs, leading to a vicious cycle of addiction [53]. Cross-cultural research including 10,930 adolescents from six European countries (Greece, Spain, Poland, the Netherlands, Romania, and Iceland) furthermore showed that using SNS for two or more hours a day was related to internalizing problems and decreased academic performance and activity [54]. In addition, a study using a sample of 920 secondary school students in China indicated neuroticism and extraversion predicted SNS addiction, clearly differentiating individuals who experience problems as a consequence of their excessive SNS use from those individuals who used games or the Internet in general excessively [55], further contributing to the contention that SNS addiction appears to be a behavioral problem separate from the more commonly researched gaming addiction. In a study using a relatively small representative sample of the Belgian population (n = 1000), results suggested 6.5% were using SNSs compulsively, with this group having lower scores on measures of emotional stability and agreeableness, conscientiousness, perceived control and self-esteem, and higher scores on loneliness and depressive feelings [56].Over the past few years, research in the SNS addiction field has largely focused on a potential addiction to using Facebook specifically, rather than other SNSs (see e.g., [57,58,59,60,61,62,63,64,65]). However, recent research suggests individuals may develop addiction-related problems as a consequence of using other SNSs, such as Instagram [66]. It has been claimed that users may experience gratification through sharing photos on Instagram, similar to the gratification they experience when using Facebook, suggesting that the motivation to share photos can be explained by uses and gratifications theory [66,67]. This may also be the reason for why individuals have been found to be less likely to experience addiction-related symptoms when using Twitter in contrast to Instagram [66]. In addition to the gratification received through photo sharing, these websites also allow to explore new identities [68], which may be considered to contribute to gratification, as supported by previous research [69]. Research has also suggested that Instagram use in particular appears to be potentially addictive in young UK adults [66], offering further support for the contention that Facebook addiction is only one example of SNS addiction.Other than the presence and possible addictive qualities of SNSs other than Facebook, it has been contended that the respective activities which take place on these websites need to be considered when studying addiction [70]. For instance, Facebook users can play games such as Farmville [36], gamble online [71], watch videos, share photos, update their profiles, and message their friends [3]. Other researchers have moved beyond the actual website use that is referred to in these types of addictions, and specifically focused on the main activities individuals engage in, referring to constructs such as ‘e-communication addiction’ [72]. It has also been claimed the term ‘Facebook addiction’ is already obsolete as there are different types of SNSs that can be engaged in and different activities that can take place on these SNSs [70]. Following this justified criticism, researchers who had previously studied Facebook addiction specifically [58] have now turned to studying SNS addiction more generally instead [73], demonstrating the changing definitional parameters of social networking in this evolving field of research.Recent research [74,75] has suggested that high engagement in social networking is partially due to what has been named the ‘fear of missing out’ (FOMO). FOMO is “a pervasive apprehension that others might be having rewarding experiences from which one is absent” [76]. Higher levels of FOMO have been associated with greater engagement with Facebook, lower general mood, lower wellbeing, and lower life satisfaction, mixed feelings when using social media, as well as inappropriate and dangerous SNS use (i.e., in university lectures, and or whilst driving) [76]. In addition to this, research [77] suggests that FOMO predicts problematic SNS use and is associated with social media addiction [78], as measured with a scale adapted from the Internet Addiction Test [79]. It has been debated whether FOMO is a specific construct, or simply a component of relational insecurity, as observed for example with the attachment dimension of preoccupation with relationships in research into problematic Internet use [80].In one study using 5280 social media users from several Spanish-speaking Latin-American countries [74] it was found that FOMO predicts negative consequences of maladaptive SNS use. In addition, this study also found that the relationship between psychopathology (as operationalized by anxiety and depression symptoms and assessed via the Hospital Anxiety and Depression Scale) and negative consequences of SNS use were mediated by FOMO, emphasizing the importance of FOMO in the self-perceived consequences of high SNS engagement. Moreover, other research [75] using 506 UK Facebook users has found that FOMO mediates the relationship between high SNS use and decreased self-esteem. Research with psychotherapists working with clients seeking help for their Internet use-related behaviors also suggested that young clients “fear the sort of relentlessness of on-going messaging (…). But concurrently with that is an absolute terror of exclusion” [34]. Taken together, these findings suggest FOMO may be a significant predictor or possible component of potential SNS addiction, a contention that requires further consideration in future research. Further work is needed into the origins of FOMO (both theoretically and empirically), as well as research into why do some SNS users are prone to FOMO and develop signs of addictions compared to those who do not.Over the last decade, research assessing problematic and possibly addictive mobile phone use (including smartphones) has proliferated [81], suggesting some individuals may develop addiction-related problems as a consequence of their mobile phone use. Recent research has suggested problematic mobile phone use is a multi-faceted condition, with dependent use being one of four possible pathways, in addition to dangerous, prohibited, and financially problematic use [82]. According to the pathway model, an addictive pattern of mobile phone use is characterized by the use of specific applications, including calls, instant messaging, and the use of social networks. This suggests that rather than being an addictive medium per se, mobile technologies including smartphones and tablets are media that enable the engagement in potentially addictive activities, including SNS use. Put another way, it could be argued that mobile phone addicts are no more addicted to their phones than alcoholics are addicted to bottles.Similarly, it has been argued previously that individuals do not become addicted to the Internet per se, but to the activities they engage in on the Internet, such as gaming [50] or SNS use [3]. With the advent and ubiquity of mobile technologies, this supposition is more pertinent than ever. Using social networking sites is a particularly popular activity on smartphones, with around 80% of social media used via mobile technologies [83]. For instance, approximately 75% of Facebook users access the SNS via their mobile phones [84]. Therefore, it can be suggested that smartphone addiction may be part of SNS addiction. Previous research [73] supported this supposition by specifically indicating that social networking is often engaged in via phones, which may contribute to its addictive potential. Accordingly, it is necessary to move towards nosological precision, for the benefit of both individuals seeking help in professional settings, as well as research that will aid developing effective treatment approaches for those in need.Related to both FOMO and mobile phone addiction is the construct of nomophobia. Nomophobia has been defined as “no mobile phone phobia”, i.e., the fear of being without one’s mobile phone [85]. Researchers have called for nomophobia to be included in the DSM-5, and the following criteria have been outlined to contribute to this problem constellation: regular and time-consuming use, feelings of anxiety when the phone is not available, “ringxiety” (i.e., repeatedly checking one’s phone for messages, sometimes leading to phantom ring tones), constant availability, preference for mobile communication over face to face communication, and financial problems as a consequence of use [85]. Nomophobia is inherently related to a fear of not being able to engage in social connections, and a preference for online social interaction (which is the key usage motivation for SNSs [3]), and has been linked to problematic Internet use and negative consequences of technology use [86], further pointing to a strong association between nomophobia and SNS addiction symptoms.Using mobile phones is understood as leading to alterations in everyday life habits and perceptions of reality, which can be associated with negative outcomes, such as impaired social interactions, social isolation, as well as both somatic and mental health problems, including anxiety, depression, and stress [85,87]. Accordingly, nomophobia can lead to using the mobile phone in an impulsive way [85], and may thus be a contributing factor to SNS addiction as it can facilitate and enhance the repeated use of social networking sites, forming habits that may increase the general vulnerability for the experience of addiction-related symptoms as a consequence of problematic SNS use.Research suggests there are sociodemographic differences among those addicted to social networking. In terms of gender, psychotherapists treating technology-use related addictions suggest SNS addiction may be more common in female rather than male patients, and describe this difference based on usage motivations:
+		
+          (…) girls don’t play role-playing games primarily, but use social forums excessively, in order to experience social interaction with other girls and above all to feel understood in their very individual problem constellations, very different from boys, who want to experience narcissistic gratification via games. This means the girls want direct interaction. They want to feel understood. They want to be able to express themselves. (…) we’re getting girls with clinical pictures that are so pronounced that we have to admit them into inpatient treatment. (…) we have to develop strategies to specifically target girls much better because there appears a huge gap. Epidemiologically, they are a very important group, but we’re not getting them into consultation and treatment.
+		  [34]
+		(…) girls don’t play role-playing games primarily, but use social forums excessively, in order to experience social interaction with other girls and above all to feel understood in their very individual problem constellations, very different from boys, who want to experience narcissistic gratification via games. This means the girls want direct interaction. They want to feel understood. They want to be able to express themselves. (…) we’re getting girls with clinical pictures that are so pronounced that we have to admit them into inpatient treatment. (…) we have to develop strategies to specifically target girls much better because there appears a huge gap. Epidemiologically, they are a very important group, but we’re not getting them into consultation and treatment.This quote highlights two important findings. First, in the age group of 14–16 years, girls appear to show a higher prevalence of addictions to the Internet and SNSs, as found in a representative German sample [88], and second, teenage girls may be underrepresented in clinical samples. Moreover, another study on a representative sample demonstrated that the distribution of addiction criteria varies between genders and that extraversion is a personality trait differentiating between intensive and addictive use [89].Cross-sectional research is less conclusive as regards the contribution of gender as a risk factor for SNS addiction. A higher prevalence of Facebook addiction was found in a sample of 423 females in Norway using the Facebook Addiction Scale [58]. Among Turkish teacher candidates, the trend was reversed, suggesting males were significantly more likely to be addicted to using Facebook [90] as assessed via an adapted version of Young’s Internet Addiction Test [79].In other studies, no relationship between gender and addiction was found. For instance, using a version of Young’s Internet Addiction Test modified for SNS addiction in 277 young Chinese smartphone users, gender did not predict SNS addiction [91]. Similarly, another study assessing SNS dependence in 194 SNS users did not find a relationship between gender and SNS dependence [51]. In a study of 447 university students in Turkey, Facebook addiction was assessed using the Facebook Addiction Scale, but did not find a predictive relationship between gender and Facebook addiction [62].Furthermore, the relationships between gender and SNS addiction may be further complicated by other variables. For instance, recent research by Oberst et al. [74] found that only for females, anxiety and depression symptoms significantly predicted negative consequences of SNS use. The researchers explained this difference by suggesting that anxiety and depression experience in girls may result in higher SNS usage, implicating cyclical relationships in that psychopathological symptom experience may exacerbate negative consequences due to SNS use, which may then negatively impact upon perceived anxiety and depression symptoms.In terms of age, studies indicate that younger individuals may be more likely to develop problems as a consequence of their excessive engagement with online social networking sites [92]. Moreover, research suggests perceptions as to the extent of possible addiction appear to differ across generations. A recent study by [72] found that parents view their adolescents’ online communication as more addictive than the adolescents themselves perceive it to be. This suggests that younger generations significantly differ from older generations in how they use technology, what place it has in their lives, and how problematic they may experience their behaviors to be. It also suggests that external accounts (such as those from parents in the case of children and adolescents) may be useful for clinicians and researchers in assessing the extent of a possible problem as adolescents may not be aware of the potential negative consequences that may arise as a result of their excessive online communication use. Interestingly, research also found that mothers are more likely to view their adolescents’ behavior as potentially more addictive relative to fathers, whose perception tended to be that of online communication use being less of a problem [72]. Taken together, although there appear differences in SNS addiction with regards to sociodemographic characteristics of the samples studied, such as gender, future research is required in order to clearly indicate where these differences lie specifically, given that much of current research appears somewhat inconclusive.Given that the research field is relatively young, studies investigating social networking site addiction unsurprisingly suffer from a number of methodological problems. Currently, there are few estimations of the prevalence of social networking addiction with most studies comprising small and unrepresentative samples [3]. As far as the authors are aware, only one study (in Hungary) has used a nationally representative sample. The study by Bányai and colleagues [93] reported that 4.5% of 5961 adolescents (mean age 16 years old) were categorized as ‘at-risk’ of social networking addiction using the Bergen Social Media Addiction Scale. However, most studies investigating social networking addiction use various assessment tools, different diagnostic criteria as well as varying cut-off points, making generalizations and study cross-comparisons difficult [53].Studies have made use of several different psychometric scales and six of these are briefly described below. The Addictive Tendencies Scale (ATS) [94] is based on addiction theory and uses three items, salience, loss of control, and withdrawal, whilst viewing SNS addiction as dimensional construct. The Bergen Facebook Addiction Scale (BFAS) [58] is based on Griffiths’ [45] addiction components, using a polythetic scoring method (scoring 3 out of 4 on each criterion on a minimum of four of the six criteria) and has been shown to have good psychometric properties. The Bergen Social Media Addiction Scale is similar to the BFAS in that ‘Facebook’ is replaced with ‘Social Media’ [95]. The E-Communication Addiction Scale [72] includes 22 questions with four subscales scored on a five-point Likert scale—addressing issues such as lack of self-control (cognitive), e-communication use in extraordinary places, worries, and control difficulty (behavioral)—and it has been found to have a high internal consistency, measuring e-communication addiction across different severity levels, ranging from very low to very high.The Facebook Dependence Questionnaire (FDQ) [96] uses eight items based on the Internet Addiction Scale [97], with the endorsement of five out of eight criteria signifying addiction to using Facebook. The Social Networking Addiction Scale (SNWAS) [51] is a five-item scale which uses Charlton and Danforth’s engagement vs. addiction questionnaire [98,99] as a basis, viewing SNS addiction as a dimensional construct. This is by no means an exhaustive list, but those assessment tools highlighted here simply demonstrate that the current social networking addiction scales are based on different theoretical frameworks and use various cut-offs, and this precludes researchers from making cross-study comparisons, and severely limits the reliability of current SNS epidemiological addiction research.Taken together, the use of different conceptualizations, assessment instruments, and cut-off points decreases the reliability of prevalence estimates because it hampers comparisons across studies, and it also questions the construct validity of SNS addiction. Accordingly, researchers are advised to develop appropriate criteria that are clinically sensitive to identify individuals who present with SNS addiction specifically, whilst clinicians will benefit from a reliable and valid diagnosis in terms of treatment development and delivery.In this paper, lessons learned from the recent empirical literature on social networking and addiction have been presented, following on from earlier work [3] when research investigating SNS addiction was in its infancy. The research presented suggests SNSs have become a way of being, with millions of people around the world regularly accessing SNSs using a variety of devices, including technologies on the go (i.e., tablets, smartphones), which appear to be particularly popular for using SNSs. The activity of social networking itself appears to be specifically eclectic and constantly changing, ranging from using traditional sites such as Facebook to more socially-based online gaming platforms and dating platforms, all allowing users to connect based on shared interests. Research has shown that there is a fine line between frequent non-problematic habitual use and problematic and possibly addictive use of SNSs, suggesting that users who experience symptoms and consequences traditionally associated with substance-related addictions (i.e., salience, mood modification, tolerance, withdrawal, relapse, and conflict) may be addicted to using SNSs. Research has also indicated that a fear of missing out (FOMO) may contribute to SNS addiction, because individuals who worry about being unable to connect to their networks may develop impulsive checking habits that over time may develop into an addiction. The same thing appears to hold true for mobile phone use and a fear of being without one’s mobile phone (i.e., nomophobia), which may be viewed as a medium that enables the engagement in SNSs (rather than being addictive per se). Given that engaging in social networking is a key activity engaged in using mobile technologies, FOMO, nomophobia, and mobile phone addiction appear to be associated with SNS addiction, with possible implications for assessment and future research.In addition to this, the lessons learned from current research suggest there are sociodemographic differences in SNS addiction. The lack of consistent findings regarding a relationship with gender may be due to different sampling techniques and various assessment instruments used, as well as the presence of extraneous variables that may contribute to the relationships found. All of these factors highlight possible methodological problems of current SNS addiction research (e.g., lack of cross-comparisons due to differences in sampling and classification, lack of control of confounding variables), which need to be addressed in future empirical research. In addition to this, research suggests younger generations may be more at risk for developing addictive symptoms as a consequence of their SNS use, whilst perceptions of SNS addiction appear to differ across generations. Younger individuals tend to view their SNS use as less problematic than their parents might, further contributing to the contention that SNS use has become a way of being and is contextual, which must be separated from the experience of actual psychopathological symptoms. The ultimate aim of research must be not to overpathologize everyday behaviors, but to carry out better quality research as this will help facilitate treatment efforts in order to provide support for those who may need it.Based on the 10 lessons learned from recent SNS addiction research, the following recommendations are provided. First, researchers are recommended to consider including an assessment of FOMO and/or nomophobia in SNS addiction screening instruments because both constructs appear related to SNS addiction. Second, it is recommended that social networking site use is measured across different technologies with which it can be accessed, including mobile and smartphones. It is of fundamental importance to study what kinds of activities are being engaged in online (social networking, gaming, etc.), rather than the medium through which these activities are engaged in (i.e., desktop computer, tablet, mobile/smartphone). Third, risk factors associated with problematic social networking need to be assessed longitudinally to provide a clearer indication of developmental etiology, and to allow for the design of targeted prevention approaches. Fourth, clinical samples need to be included in research in order to ensure the sensitivity and specificity of the screening instruments developed. Fifth, in terms of treatment, unlike treating substance-related addictions, the main treatment goal should be control rather than abstinence. Arguably, abstinence cannot realistically be achieved in the context of SNS addiction because the Internet and social networking have become integral elements of our lives [3,8,33]. Rather than discontinuing social networking completely, therapy should focus on establishing controlled SNS use and media awareness [53].This paper has outlined ten lessons learned from recent empirical literature on online social networking and addiction. Based on the presented evidence, the way forward in the emerging research field of social networking addiction requires the establishment of consensual nosological precision, so that both researchers and clinical practitioners can work together and establish productive communication between the involved parties that enable reliable and valid assessments of SNS addiction and associated behaviors (e.g., problematic mobile phone use), and the development of targeted and specific treatment approaches to ameliorate the negative consequences of such disorders.This work did not receive any funding.The first author wrote the first complete draft of the paper based on an idea by the second author. The authors then worked collaboratively and iteratively on subsequent drafts of the paper. The authors declare no conflict of interest.

Med-MDPI/ijerph_2/ijerph-14-03-00312.txt

@@ -0,0 +1,58 @@
+Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).Background: Current levels of traffic-related air pollution (TRAP) are associated with the development of childhood asthma, although some inconsistencies and heterogeneity remain. An important part of the uncertainty in studies of TRAP-associated asthma originates from uncertainties in the TRAP exposure assessment and assignment methods. In this work, we aim to systematically review the exposure assessment methods used in the epidemiology of TRAP and childhood asthma, highlight recent advances, remaining research gaps and make suggestions for further research. Methods: We systematically reviewed epidemiological studies published up until 8 September 2016 and available in Embase, Ovid MEDLINE (R), and “Transport database”. We included studies which examined the association between children’s exposure to TRAP metrics and their risk of “asthma” incidence or lifetime prevalence, from birth to the age of 18 years old. Results: We found 42 studies which examined the associations between TRAP and subsequent childhood asthma incidence or lifetime prevalence, published since 1999. Land-use regression modelling was the most commonly used method and nitrogen dioxide (NO2) was the most commonly used pollutant in the exposure assessments. Most studies estimated TRAP exposure at the residential address and only a few considered the participants’ mobility. TRAP exposure was mostly assessed at the birth year and only a few studies considered different and/or multiple exposure time windows. We recommend that further work is needed including e.g., the use of new exposure metrics such as the composition of particulate matter, oxidative potential and ultra-fine particles, improved modelling e.g., by combining different exposure assessment models, including mobility of the participants, and systematically investigating different exposure time windows. Conclusions: Although our previous meta-analysis found statistically significant associations for various TRAP exposures and subsequent childhood asthma, further refinement of the exposure assessment may improve the risk estimates, and shed light on critical exposure time windows, putative agents, underlying mechanisms and drivers of heterogeneity.Asthma is a chronic inflammatory disease of the airway which has a large impact on quality of life and poses a great burden on health services [1]. In children, asthma is the most commonly reported chronic disease in developed countries [2]. Environmental factors, importantly including improved hygiene, ambient air pollution exposures, and early-life exposures to microbes and aeroallergens, contribute to the development of asthma [2]. In a recent systematic review and meta-analyses, we found statistically significant associations between traffic-related air pollution (TRAP) and the incidence and lifetime prevalence of childhood asthma, although there was significant heterogeneity in some of the risk estimates [3]. These effects are biologically plausible. Britain’s Committee on the Medical Effects of Air Pollutants proposed four mechanisms by which air pollution can affect asthma: (1) oxidative stress and damage; (2) inflamed pathways; (3) airway remodeling; and (4) enhancement of respiratory sensitization to allergens [4]. Oxidative stress relates to common asthmatic traits [5], and was suggested to play a role in asthma pathogenesis [6]. Further, it was previously highlighted as one chief pathway which underpins the adverse health effects of (traffic-related) air pollution on the respiratory systems [7].TRAP is a particularly important and challenging exposure to study given its ubiquity, its dominance in present urban areas, its proximity to human receptors, and its high spatial and temporal variability [8,9,10,11]. For example, the local traffic contribution to ambient nitrogen dioxide (NO2) can be up to 80%, and ranges between 9% and 53% for urban particulate matter less than 10 micrometres in diameter (PM10), and 9%–66% for urban particulate matter less than 2.5 micrometres in diameter (PM2.5) [8].In the epidemiological studies included in the most recent meta-analyses of TRAP and the development of childhood asthma, different exposure assessment methods and indices have been used to characterise the exposure to TRAP, including distance to roads, active measurement of air pollutants, use of routinely measured air pollution data, land-use regression (LUR) modelling, air dispersion modelling and remote sensing [3]. These various methods and indices differ substantially and have advantages and disadvantages in terms of their spatial and temporal resolution, specificity to traffic, data and effort/expertise requirements, transferability and information provided on the actual pollutants. Furthermore, the different epidemiological studies focused on different pollutants and different exposure time windows [3]. The use of different exposure assessment methods in health effects or impacts studies can result in different estimates, partly due to the difference in accuracy and precision of the exposure estimates and the potential differential effects of different pollutants. Although the evidence base is very limited, research has shown differences, for example in the performance of and the results from dispersion models versus LUR [12,13,14] which in two studies translated into small differences in the risk estimates, but in one study translated into differences in the direction of effect estimates of NO2 on birth weight [13]. In a previous meta-analysis on TRAP and childhood asthma, there was some suggestion of a difference between associations with NO2 from within-community studies that used LUR models (five studies, odds ratio (OR) = 1.14, 95% confidence interval (CI) 1.06, 1.23) and those from studies that used dispersion models (five studies, OR = 1.02, 95% CI 0.97, 1.07) [15]. Further, there were differences in estimated health impacts when using different pollutant-specific exposure-response functions. For example, cases of asthma attributable to PM10 and NO2 differ substantially to cases of asthma attributable to black carbon [16].In this paper, we aim to describe and discuss the exposure assessments conducted in studies of TRAP and childhood asthma development, including the methods used in the different regions, the pollutants and exposure assignment and time windows studied. We then highlight research gaps and make suggestions for further research in this rapidly growing area. Our focus is on the exposure assessments and not the effects of TRAP on asthma development per se; which we reviewed in depth elsewhere [3]. Our results and discussion are applicable to other research on TRAP and various health outcomes, beyond childhood asthma, as the exposure assessment methods are often similar [11,17].We conducted a systematic review to synthesize the literature on TRAP exposures and the subsequent risk of childhood asthma development defined as incidence or lifetime prevalence [3]. We followed established guidance published by the University of York’s Centre for Reviews and Dissemination [18]. We registered a protocol (registration number: CRD42014015448) with the international prospective register of systematic reviews (PROSPERO) documenting our methodological approach a priori [19].We performed the searches on 8 September 2016 via the database search interface OvidSP (http://ovidsp.ovid.com/). We searched the following databases for relevant studies: Embase (1996 to week 36, 2016), Ovid MEDLINE (R) (1996 to August 2016), and “Transport Database” (1988 to August 2016). We identified relevant studies by entering four sets of combined keywords in the “Multi-Field Search” option in OvidSP. We searched for the selected keyword combinations in “All Fields”. The keyword combinations were:
+      
+        
+          “Child*” AND “air pollution” AND “asthma”;
+        
+        
+          “Child*” AND “air quality” AND “asthma”;
+        
+        
+          “Child*” AND “vehicle emissions” AND “asthma”; and
+        
+        
+          “Child*” AND “ultra-fine particles” AND “asthma”.
+        
+      “Child*” AND “air pollution” AND “asthma”;“Child*” AND “air quality” AND “asthma”;“Child*” AND “vehicle emissions” AND “asthma”; and“Child*” AND “ultra-fine particles” AND “asthma”.We applied no limits on the initial publication date and no limits on language although we eventually excluded three foreign language studies due to translation difficulties [20,21,22]. We conducted a hand search in the reference lists of all the included studies and of previous relevant reviews we identified [15,17,23,24,25,26,27,28,29,30]. We contacted authors of unpublished studies (abstracts only) and the authors of the most recurrent studies to ensure the inclusion of all relevant published material on the topic and this resulted in the inclusion of two additional studies [31,32]. We searched Google for any other material related to “traffic-related air pollution” AND “childhood asthma” and this resulted in the inclusion of one additional study [33]. One study was also not identified in the searches but by one of the reviewers and this was included [34]. We exported studies into an Endnote X7.4 library and removed duplicates automatically using the Endnote function “Find Duplicates”. For inclusion, we selected studies that met all the following criteria:
+      
+        
+          Were published epidemiological/observational studies;
+        
+        
+          Explicitly specified the term “asthma” as an outcome for investigation;
+        
+        
+          Examined the childhood exposure from birth until 18 years old [35] to any designated TRAP metric or established traffic-related air pollutant including proximity to roads or traffic, carbon monoxide (CO), elemental carbon (EC), nitrogen oxides (NOx), nitric oxide (NO), NO2, hydrocarbons, particles of different aerodynamic diameters (PM2.5, PM10, PMcoarse, UFPs) or PM2.5 absorbance as a marker for black carbon (BC) concentrations [10,36]; and
+        
+        
+          Examined and reported associations between preceding exposure to TRAP and subsequent risk of asthma reported as incidence or lifetime prevalence from birth until 18 years old.
+        
+      Were published epidemiological/observational studies;Explicitly specified the term “asthma” as an outcome for investigation;Examined the childhood exposure from birth until 18 years old [35] to any designated TRAP metric or established traffic-related air pollutant including proximity to roads or traffic, carbon monoxide (CO), elemental carbon (EC), nitrogen oxides (NOx), nitric oxide (NO), NO2, hydrocarbons, particles of different aerodynamic diameters (PM2.5, PM10, PMcoarse, UFPs) or PM2.5 absorbance as a marker for black carbon (BC) concentrations [10,36]; andExamined and reported associations between preceding exposure to TRAP and subsequent risk of asthma reported as incidence or lifetime prevalence from birth until 18 years old.All titles and abstracts were reviewed against the inclusion criteria by one researcher (Haneen Khreis) with a random 20% independently reviewed by another researcher. All potentially relevant studies were then retrieved and the available full-papers reviewed against the inclusion criteria by one researcher (Haneen Khreis) with a random 50% independently reviewed by another researcher (Mark J. Nieuwenhuijsen). Screening was undertaken manually and differences were resolved by consensus. The following data items were extracted from each included study:
+      
+        
+          Study reference and setting;
+        
+        
+          Study design;
+        
+        
+          Age group;
+        
+        
+          Number of participants;
+        
+        
+          Exposure assessment method(s);
+        
+        
+          Pollutant(s) studied;
+        
+        
+          Exposure assessment place;
+        
+        
+          Exposure assessment time; and
+        
+        
+          Air pollution estimates validation, if any.
+        
+      Study reference and setting;Study design;Age group;Number of participants;Exposure assessment method(s);Pollutant(s) studied;Exposure assessment place;Exposure assessment time; andAir pollution estimates validation, if any.Data was primarily extracted from the main papers of the included studies. Where necessary, data items were missing from the main papers, data was extracted from the supplementary materials [31,37,38,39,40,41,42,43,44,45,46,47,48,49,50,51,52], and the associated publications [53,54,55,56,57,58,59,60,61,62,63,64,65,66]. Data extraction was undertaken manually by one researcher (Haneen Khreis). A random 50% was independently reviewed by another researcher (Mark J. Nieuwenhuijsen). A fuller detail of the screening methodology can be found in Khreis et al. (2017) [3].The databases searches yielded 4276 unique articles, from which 95 were selected for detailed assessment of the full text, one of which was identified by a peer reviewer. Figure 1 shows the flow of papers. A total of 42 studies met our inclusion criteria [31,32,33,34,36,37,38,39,40,41,42,43,44,45,46,47,48,50,51,52,58,67,68,69,70,71,72,73,74,75,76,77,78,79,80,81,82,83,84,85,86,87] (Table 1).A summary of the included studies’ key characteristics is shown in Table 1. Ages of participants ranged from 1 to 18 years old and sample sizes ranged from 184 [69] to 1,133,938 [85]. Follow-up periods ranged from 1 to 16 years [47]. Eighteen studies were conducted in Europe, 11 in North America, 5 in Japan, 3 in China and 1 in each of Korea and Taiwan. Thirty-two studies were cohort studies (25 of which were birth cohorts), 6 studies were case-control studies (2 of which were nested in a birth cohort), and 4 studies were cross-sectional.The exposure to TRAP was assessed using different methods, sometimes in isolation and other times in combination with each other (Table 1). Most studies (N = 22) used LUR models, 16 studies used TRAP surrogates (e.g., proximity to roadways), 11 studies used traffic-related air pollutant concentrations measured at fixed-site monitoring stations, 8 studies used air dispersion modelling, 1 study used remote sensing and 1 study used diffusion tubes at the residence to measure NO2. These methods vary substantially in terms of their spatial and temporal resolution, specificity to traffic, data and effort/expertise required, transferability and information provided on the actual pollutants (Table 2). These are key criteria important in studies of TRAP and asthma (and other health effects).In the literature, it was also apparent that the use of the different exposure assessment methods varied by region (Table 1). For example, 8 out of the 11 studies using pollutant measurements at fixed-site monitoring stations only used this exposure method (i.e., not in combination with other methods or metrics), 7 of which were from Japan, Taiwan, Korea and China. Also, 12 out of the 22 studies using LUR model used this exposure method only, 9 of which were from Europe (predominantly from the PIAMA cohort in The Netherlands), while the remaining 3 were from Canada and the USA. The remaining USA studies showed the most variability in the exposure assessment methods choice and used residential diffusion tube monitoring, dispersion modelling, fixed-site monitoring stations, proximity measures and multiple novel TRAP surrogates (see Patel et al. 2011 who used some new surrogates, including “four-way street intersection density” and “number of New York City transit bus stops”).NO2 was the pollutant most studied (31 studies), followed by PM2.5 (18 studies), BC or PM2.5 absorbance (15 studies), and PM10 (14 studies). Other pollutants including NOx (7 studies), EC (4 studies), CO (3 studies), PMcoarse (3 studies), NO (2 studies) were less frequently studied. Only two studies assessed particulate matter composition elements, considered as non-exhaust road traffic emissions, including copper (Cu), iron (Fe), zinc (Zn), nickel (Ni), sulfur (S), and vanadium (V). These studies exclusively originated from the Dutch PIAMA cohort [48,75]. One study assessed oxidative potential, which is a measure of the inherent capacity of particulate matter to oxidise target molecules, [31], and no studies assessed ultra-fine particles.Table S1 in the supplementary material is a summary of where and when the exposure to TRAP was assessed in each included study and whether any validation was undertaken. The assignment of TRAP exposures was almost exclusively based on the residential address of the participating children. Only a few studies considered the impact of moving residence on TRAP exposure levels and undertook additional or sensitivity analyses for movers/non-movers or assigned the exposure at multiple addresses based on the residential history. There were a few studies which assigned the exposure based on school locations instead of residence. Shima and Adachi [79] and Shima et al. [81] used routine measurements from fixed-site stations near school addresses to represent TRAP exposures in Japan, whilst Deng, Lu, Norbäck, Bornehag, Zhang, Liu, Yuan and Sundell [70] and Deng, Lu, Ou, Chen and Yuan [86] used routine measurements from fixed-site stations near children’s kindergartens to represent TRAP exposures in China.The exposure assignment was generally static; i.e., not taking children’s mobility into account. In many cases, this could be argued as reasonable as participants were in their infancy or early life (birth–3 years old), and residential exposure is then thought to be most relevant. Only 10 studies, mostly recent, considered children’s mobility in the exposure assessment and assigned time-weighted exposures at day cares and/or schools [33,34,38,40,42,50,83,84], and other locations where the child spends significant time [46,76], alongside residence. These studies were conducted at ages when exposure at the residential address becomes less relevant due to children’s increased mobility.In terms of the exposure time window investigated, studies differed, but birth year was the most explored time window (Table S1). Very few studies investigated alternative exposure windows such as different years of life, longer duration, cumulative or life-time exposure.Studies using LUR or dispersion modelling validated their modelled exposure estimates against measured concentration using different methods including leave-one-out cross validation procedure (mainly for LUR models) and independent cross validation against fixed-site monitoring stations measurements (mainly for dispersion models). Generally, the validation of the LUR model estimates were not conducted using a separate test validation dataset which significantly limits the comprehensiveness of the validation. No study reported validation against personal exposures.Studies using different TRAP surrogates were the least consistent to show an increased asthma risk associated with TRAP. Studies using dispersion model were more consistent in showing associations. For example, out of 8 studies using dispersion models, 5 showed positive and statistically significant risk estimates. Studies using traffic-related air pollutants concentrations at fixed-site monitoring stations, and studies using LUR modelling generally showed an increased asthma risk associated with TRAP. For example, out of 22 studies using LUR models, 17 showed positive and statistically significant risk estimates. The one study that measured NO2 exposure at the individual residential level also showed statistically significant associations between the exposure and asthma [74]; so did the one study that used remote sensing [85]. Some of the same studies which found no association between roadway proximity and asthma, found increased risks when employing more refined exposure models such as LUR model estimates [36,37,38,42,45,78].We found 42 studies that examined the association between TRAP and the subsequent onset of childhood asthma defined as incidence or lifetime prevalence. Exposures metrics differed in terms of their spatial and temporal resolution and their specificity to traffic. LUR modelling was the most commonly used exposure assessment method and NO2 was the most commonly studied pollutant. Most studies estimated TRAP exposures at the residential address and only a few considered the mobility of the children and/or their residential address changes. Most studies estimated the TRAP exposures at the first year of life (birth year) and only a few studies assessed the effects of cumulative exposures and/or exposures at different time-windows. Validation was undertaken for LUR and dispersion models estimates only and no study has validated exposure estimates against personal monitored exposures. Although our previous meta-analysis found positive and statistically significant associations for various TRAP exposures (black carbon, NO2, PM2.5, PM10) with the onset asthma [3], further refinement of the exposure assessments may improve the exposure-response functions and shed light on associations with other under-investigated pollutants.The prominent focus on NO2 in the literature is probably related to the wide availability of this pollutant measure, the ease and relatively low cost to measure it and its relative specificity to road traffic [30]. The focus on NO2 in air quality guidelines, plans and mitigation strategies in the EU, and beyond, is perhaps reinforcing the study of this pollutant. Fewer studies measured or modelled PM2.5 or particulate components, even though it is more widely implicated in the health effects of air pollution [88]. The cost of measuring and/or modelling PM tends to be higher. The literature, however, suggests that there has been a recent move from studying standard air pollutants to studying other agents, most notably including black and elemental carbon, two agents that are considered as TRAP signatures, but also PM composition elements and other properties such as oxidative potential [31,48,75]. As it stands, there were no studies investigating the impacts of long-term exposure to ultra-fine particles on asthma but there are studies under way to measure ultra-fine particles [89]. The work on PM composition is particularly relevant with the expected wide-spread introduction of electric vehicles and the associated likely reductions of exhaust emissions and increase in non-exhaust emissions [90]. PM composition research could potentially lead to further insight on the putative agents and source of pollutants. For example, Gehring, Beelen, Eeftens, Hoek, de Hoogh, de Jongste, Keuken, Koppelman, Meliefste and Oldenwening [48] suggested that iron, copper, and zinc in PM, reflecting poorly regulated non-exhaust traffic emissions, may increase the risk of asthma and allergy in Dutch schoolchildren. A birth cohort study using oxidative potential measures, particularly using the dithiothreitol assay, found that asthma and other respiratory health outcomes were more strongly related to oxidative potential when compared to PM2.5, suggesting that this exposure metric may be closer to the underlying mechanisms [31]. These different measures are rarely studied and should be further explored in future research, principally in locations where ratios between oxidative potential and other TRAP markers such as NO2 differ; to determine with more confidence which metric predicts respiratory health better.Many studies have used LUR modelling to estimate TRAP exposures, partly because of its relatively low costs, ease of implementation and possibility to consider traffic determinants of exposure such as the road network and traffic density. LUR models also tend to provide a good spatial coverage and resolution for TRAP exposure. The LUR method is an empirical method and uses least squares regression to combine measured data with geographic information system (GIS)-based predictor data reflecting pollutant sources, to build a prediction model applicable to non-measured locations, e.g., residential addresses of cohort members. An advantage of LUR models is that they are stable over time [91,92,93]. However, their validation, most commonly undertaken using leave-one-out cross validation procedure, is incomplete. Relatively few studies used air dispersion models which are based on more detailed knowledge of the physical, chemical, and fluid dynamical processes in the atmosphere. Air dispersion models use information on emissions, source characteristics, chemical and physical properties of the pollutants, topography, and meteorology to model the transport and transformation of gaseous or particulate pollutants through the atmosphere to predict air pollutant concentrations. They allow for a finer temporal and spatial resolution of TRAP exposure and specific source apportionment (beyond TRAP) which is valuable when recommending specific policy interventions targeted at specific sources. Yet, their main drawback is related to the quality of the input data; especially the vehicle emission factors which are highly uncertain [94]. Amongst the encountered exposure methods, these two methods are favorable in terms of their spatial and temporal resolution and their specificity to traffic (Table 2). The preferred method for exposure assessment is not so obvious and depends on available resources, the quality of the input data, expertise, place of study and transferability considerations. For example, de Hoogh, et al. [95] found that the median Pearson R (range) correlation coefficients between LUR and air dispersion model estimates for the annual average concentrations of NO2, PM10 and PM2.5 were 0.75 (0.19–0.89), 0.39 (0.23–0.66) and 0.29 (0.22–0.81) for 112,971 (13 study areas), 69,591 (7) and 28,519 (4) addresses respectively, suggesting a much better agreement for NO2 than for PM, probably because the main source for NO2 is traffic and PM has other sources. The median Pearson R correlation coefficients (range) between air dispersion model estimates and measurements were 0.74 (0.09–0.86) for NO2; 0.58 (0.36–0.88) for PM10 and 0.58 (0.39–0.66) for PM2.5. Wang et al. [96] compared both methods in a study of children’s lung function and found that exposure estimates from LUR and dispersion models correlated very well for PM2.5, NO2, and black carbon, but not for PM10. Health effect estimates did not depend on the type of model used in their population of Dutch children. Yet, with a very limited number of comparison studies, the extent to which estimates of air pollution effects are affected by the choice of exposure model remains unclear. A combination of the LUR and dispersion models may further improve the exposure assessment estimates, possibly accounting for some of the imperfections in the emission databases [97].Compared to estimates from routine monitoring stations LUR and air dispersion model have the advantage that they provide a better spatial resolution, but also require more effort and are costlier. The better spatial resolution may be quite important when the study area is small and clear exposure differences can be observed by detailed exposure assessment. At least, all three methods provide some level of pollutants which may be important for policy reasons, while surrogate measures like distance from roads do not. A relatively new method, remote sensing [85] has the advantage that air pollution estimates can be obtained where there are no or fewer monitors or less resources and expertise is available i.e., medium- and low-income countries, but still needs some further refinement in terms of spatial resolution and the number of pollutants for which good estimation methods are available.We attempted to evaluate the effects of the exposure assessment method on the health risk estimates observed in the included studies, for example with meta-regression, but the number of studies available are still too small to conduct such analyses. Even for NO2 exposure, there were only 20 studies entering the meta-analysis, 12 of which used LUR models and 1 used dispersion modelling. The meta-analyses though suggested considerable heterogeneity, especially in the case of NO2 where most studies where available, and part of this heterogeneity could be caused by different exposure assessment methods. Given the rapid increase in the number of studies in this field, it may become possible to conduct such analysis in the near future.Only a small number of studies considered children’s mobility at ages when exposure at the residential address becomes less relevant and assigned time-weighted TRAP exposures at day care centres and schools and other locations where the child spends significant time alongside residence. Children may spend only around 50%–60% of their time at home, and the rest elsewhere e.g., at school [98]. TRAP exposure levels such as black carbon can be considerably higher when commuting compared to being at home [98], and therefore residential estimates may underestimate the true exposure and bias the exposure-response functions. New tracking technology and portable sensors have now made it possible obtain information on TRAP exposure levels over the day, even though it requires considerable effort and may only feasible for smaller samples [98]. New approaches such as indicating the home and school address and commuting route in geographical information system packages and overlaying this with time adjusted air pollution maps may provide estimates for larger study samples and can be an area of further inquiry. Considering the significant amount of time spent indoors, it may also be beneficial to investigate indoor air pollution exposures and the impact of specifically incorporating these on the exposure–response functions. Currently, all available exposure models, except personal monitors (which have not been used in any of the included studies), estimate outdoor air pollution only and use this as a surrogate for the indoor levels without taking into account indoor-outdoor penetration factors. However, outdoor and indoor TRAP are correlated as there is considerable penetration of outdoor sources to indoor environments. These correlations are may be one rapid and practical method to assign indoor exposures. One study which characterized the indoor–outdoor relationship of PM2.5 in Beijing found that there is a strong correlation between indoor and outdoor PM2.5 mass concentrations, and that the ambient data explained ≥ 84% variance of the indoor data [99]. Another study similarly showed that PM2.5 levels in an Australian primary school were mainly affected by the outdoor PM2.5 (r = 0.68, p < 0.01) [100]. Another study in Germany found that over 75% of the daily indoor PM2.5 and black smoke variation could be explained by daily outdoor variation for those pollutants [101].Further, investigating different exposure time windows may highlight other relevant exposure windows beyond the birth year and early-life that are commonly studied. The differences between effects of early exposure versus later exposures or exposures with greater duration is yet unclear and is difficult to detangle due to the limited number of studies investigating different time windows. Some authors have suggested that exposures of longer duration at elevated TRAP levels may be necessary to generate pathophysiological changes leading to asthma development and therefore may be behind the observed effects [46].Novel approaches to exposure assessment are underway including the use of OMICS technologies that measure biological molecules and/or activity in the body (e.g., transcriptomics, proteomics, metabolomics or methylation) to identify fingerprints of air pollution [89,102]. Although still in their infancy, such approaches may provide a good way of characterising air pollution exposures inside the body and on existing biological samples (that have been stored for a while). Furthermore, they may provide further insight in the underlying mechanisms by which air pollution cause health effects in children and others.Although there appear to be statistical significant associations between TRAP and the development of childhood asthma, there is a further need to improve the exposure estimates, and therefore improve the exposure–response functions and the consistency of the study findings. This is important for example when these exposure–response functions are used for burden of disease and health impact assessment studies, and for better understanding the underlying mechanisms of TRAP and childhood asthma and the potential differential pollutant effects and drivers of heterogeneity. Over the past few years, there has been an epidemic increase in the number of studies in the field, and there are likely to be more studies over the next few years given the importance of the topic. Improvements in exposure assessments, as we discuss in this paper, may well increase the scientific value of these new studies. More refined exposure models are needed, and will arguably produce the most robust associations when investigating the potential health effects of TRAP. Furthermore, we also emphasize the need to incorporate mobility patterns in the exposure estimates and to undertake personal exposure monitoring to cross validate modelling estimates.Although our previous meta-analysis found statistically significant associations for various TRAP exposures and childhood asthma, further refinement of the exposure assessment may improve the risk estimates and shed light on critical exposure time windows, putative agents, underlying mechanisms and drivers of heterogeneity.The following are available online at www.mdpi.com/1660-4601/14/3/312/s1, Table S1: Exposure assessment place, time and validation in the Included Studies.H.K. designed the study, performed the searches, screened and extracted the data, wrote the initial draft and had final responsibility for the decision to submit for publication; M.N. independently reviewed and extracted data for 20%–50% of the studies identified, contributed to the interpretation of data, revised the manuscript, approved the final version and agreed to be accountable for all aspects of the work.The authors declare no conflict of interest.Study screening process.Main characteristics of the included studies.Abbreviations: BAMSE, Barn (children), Allergy, Milieu, Stockholm, an Epidemiology project; BC: black carbon; CAPPS, The Canadian Asthma Primary Prevention Study; CCAAPS, The Cincinnati Childhood Allergy and Air Pollution Study; CCCEH, Columbia Center for Children’s Environmental Health birth cohort study; CCHH, China-Children-Homes-Health study; CEAS, Childhood Environment and Allergic Diseases Study; CHS, The Children’s Health Study; EC, elemental carbon; ESCAPE, The European Study of Cohorts for Air Pollution Effects; GALA II, The Genes–environments and Admixture in Latino Americans; GASPII, The Gene and Environment Prospective Study in Italy; GINIplus, German Infant study on the influence of Nutrition Intervention plus air pollution and genetics on allergy development; ICD, International Classification of Diseases; LISAplus, Life style Immune System Allergy plus air pollution and genetics; LUR, land-use regression; MAAS, The Manchester Asthma and Allergy Study; Medi-Cal, California Medical Assistance Program; NA, not applicable; NO, nitrogen oxide; PM: particulate matter; SAGE II, The Study of African Americans, Asthma, Genes and Environments; SAGE, The Study of Asthma, Genes and the Environment; SORA, Study on Respiratory Disease and Automobile Exhaust; VESTA, Five (V) Epidemiological Studies on Transport and Asthma; y.o., years old.Pros and cons of exposure assessment methods used in the systematic review literature. TRAP: traffic-related air pollution.Ratings: +: good; ++: very good; -: potentially inadequate; --: highly inadequate.

Med-MDPI/ijerph_2/ijerph-14-03-00313.txt

@@ -0,0 +1 @@
+Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).We tested the use of mosquito traps as an alternative to spraying insecticide in Camargue (France) following the significant impacts observed on the non-target fauna through Bti persistence and trophic perturbations. In a village of 600 inhabitants, 16 Techno Bam traps emitting CO2 and using octenol lures were set from April to November 2016. Trap performance was estimated at 70% overall based on mosquitoes landing on human bait in areas with and without traps. The reduction of Ochlerotatus caspius and Oc. detritus, the two species targeted by Bti spraying, was, respectively, 74% and 98%. Traps were less efficient against Anopheles hyrcanus (46%), which was more attracted by lactic acid than octenol lures based on previous tests. Nearly 300,000 mosquitoes from nine species were captured, with large variations among traps, emphasizing that trap performance is also influenced by surrounding factors. Environmental impact, based on the proportion of non-target insects captured, was mostly limited to small chironomids attracted by street lights. The breeding success of a house martin colony was not significantly affected by trap use, in contrast to Bti spraying. Our experiment confirms that the deployment of mosquito traps can offer a cost-effective alternative to Bti spraying for protecting local populations from mosquito nuisance in sensitive natural areas.Bacillus thuringiensis israelensis (Bti) is the most selective and least toxic larvicide currently available to control mosquitoes [1]. However, its sustained use in wetland-dominated areas has revealed strong indirect impacts on animal species that depend on small dipterans and/or their predators for breeding and survival [2]. In Camargue (Rhône delta, southern France), the spraying of 2500 out of 25,000 ha of mosquito larval biotopes with Bti has led to a significant 30%—60% decrease in the breeding success of house martins [3], in the richness and abundance of odonates [4], as well as the invertebrate prey available to reed passerines [5]. Mosquito control in Camargue was initiated 50 years after its implementation on the French Mediterranean coast, on the assumption that Bti use would permit, in contrast to chemical insecticides, a reconciliation of nature protection with human comfort. However, the observed impact on the non-target fauna due both to the mosquito reduction and the collateral effects on benthic chironomids following Bti persistence in the sediments [6,7] is calling for alternative solutions. Various mosquito traps are commercially available for public consumers to reduce the mosquito nuisance and/or decrease the risk of mosquito-borne illness [8]. Could a network of traps be used as a protecting belt around inhabited areas to improve human comfort while preserving wetland biodiversity? Deploying mosquito traps in urban areas appeared as a cost-efficient alternative to traditional mosquito control in Camargue, where small villages and towns are typically surrounded by thousands of hectares of wetlands potentially producing mosquitoes. A first prototype adapted to collective use and inspired from the functioning of traps available for individual consumers was developed and patented by Techno Bam (http://techno-bam.net/fr/), a small local business, in 2014. After some initial tests in 2015, 16 of these traps were deployed in a hamlet of 600 inhabitants and operated during the whole mosquito season in 2016. This study reports on the efficacy and environmental impact of this experiment as an innovative way to control mosquitoes in an area reputed for its high mosquito density during several months of the year. Mosquito traps: Techno Bam traps use octenol-based lures in the form of absorbent beads and release of recycled carbon dioxide from CO2 cylinder to attract female mosquitoes. A power-supplied impellor fan sucks the female mosquito into a net of 1 × 0.5 mm mesh. Because the traps are made for public use, all the material needed for their functioning is concealed into a weather-resistant box locked and riveted to the ground. A total of 16 traps were deployed, covering most of the Sambuc hamlet based on a 60-m attraction radius for mosquitoes (Figure 1). Insect samples: Traps were operated from mid-April through late October 2016, with the nets being emptied from three to five times a week (n = 1380 samples). Fresh samples were brought to the laboratory and weighted. Each week, samples from three traps (n = 86) were examined under a stereoscope to determine the number of species and individuals of biting dipterans, as well as the number of non-target insects identified to taxonomic order. From these samples we calculated a mean bodyweight for mosquitoes (2.55632 mg) that was used to extrapolate their numbers from the weighed samples. Trap performance: While the number of mosquitoes trapped can be a relative measure of trap performance (e.g., for comparing different trap models [9]), the main criteria for assessing absolute performance should be the reduction in biting pressure on humans. Accordingly, trap performance was assessed by comparing the number of mosquitos landing on human baits (calf test) during a 10 min period at three locations in the hamlet (at 10 m and 40 m from trap position) and two locations outside the hamlet (at 550 m and 1130 m from the nearest trap position) before (2015) and during (2016) the experiment (Figure 2). We collected 60 samples in 2015 before trap installation, and 334 samples during the 2016 experiment. Sampling was done at least once a week, should environmental conditions be favorable (low wind, no rain, presence of mosquitoes outside twilight activity peak). Calf tests were made simultaneously by one or several observers, with the same observer(s) covering systematically control and treated points in an alternate manner during each sampling period. All mosquitoes landing on human baits were collected with a mouth aspirator, counted and identified to species. Trap performance was assessed globally and for each mosquito species by estimating the percent decrease in the mean number of biting attempts at treated relative to control areas using Generalized Linear Models with a nested ANOVA design (Statistica V12, Stat Soft Inc. Maisons-Alfort, France), where sites and dates were nested in treatment (fixed factor).Environmental impacts: We estimated direct effects of Techno Bam traps based on the presence of non-target insects captured in the traps, as well as indirect effects based on the breeding success of a colony of house martins (Delichon urbicum) nesting in the treated area (Figure 1). Breeding success was estimated by visiting 21 nests twice a week from 12 May to 27 August to determine the number of fledged young from all breeding attempts in the season. Mean number of young produced by nest was compared to the breeding success observed at the same site prior to the trap experimentation in 2015, as well as to the breeding success observed at two control sites (including the Sambuc colony) and two sites surrounded by Bti-sprayed wetlands that were monitored from 2009 to 20113. These analyses were made using a GLM with a nested ANOVA design where site and year were nested in treatment (fixed factor).The estimated number of mosquitoes trapped daily varied over time, with three peaks observed in June, July and August (Figure 2). Overall, an estimated number of 299,408 mosquitoes was captured, with mean a daily capture rate per trap ranging from one mosquito in early May to 382 mosquitoes in late August.Mean capture rates also varied spatially among traps, ranging from 24 to 399, depending upon their location in the hamlet. The highest number of mosquitoes caught in a single day in one trap was 4300 in late August. Prior to trap installation in 2015, the relative mosquito nuisance was higher at Sambuc (mean 8.6 ± 1.3 SE) than at the control sites (mean 4.1 ± 1.5 SE) located 550 and 1130 m from the hamlet (F (1,32) = 5.21; p = 0.029). After trap installation, however, the relative mosquito nuisance was significantly lower at Sambuc compared to the control sites (F (1,294) = 18.46; p < 0.0001). Overall, the mosquito nuisance was reduced by 70%, with a mean of 4.1 biting attempts/10 min at 10–40 m from the traps, compared to 14.1 at control sites. Calf tests provided similar results when conducted at 10 m and 40 m from the traps (F (1,110) = 0.252, p = 0.62), hence these data were combined in the analyses.On a weekly basis, the mosquito nuisance was kept at very low levels until mid-July in the area covered by traps (Figure 3), despite various peaks in mosquito nuisance obtained at the control sites (calf tests) and confirmed at Sambuc through mosquito captures in the traps (Figure 2). However, three peaks of mosquito nuisance with over 10 biting attempts/10 min were observed in July, and early and late August at Sambuc (Figure 3). In these cases, trap use permitted us to reduce the level and duration of the mosquito nuisance but not to eliminate it completely.Nine mosquito species were captured in traps and on human bait (Table 1). All species present in both the control and treated areas showed a reduced abundance in treated areas, the latter being highly significant for four species. The species mainly responsible for the mosquito nuisance were well controlled by the use of traps, their reduction rate varying from 74% to 98% with the exception of Anopheles hyrcanus, which was responsible for the peak observed near the end of the mosquito season (Table 1). Trap performance was also lower for Culex spp., especially Cx. Modestus, which accounted for 0.14% of captures in traps and 4.16% of captures on human bait. Finally, although there were a few tiger mosquitoes, Aedes albopictus, in the hamlet (two individuals captured in traps and on human bait), the absence of this urban species at control sites makes the calculation of a reduction rate relative to trap use impossible.We counted and identified 39,941 insects in the 86 trap samples that were examined in detail. Of these, 23,098 (57.8%) were mosquitoes, 1499 (3.8%) were Ceratopogonidae and 15,359 (38.4%) were non-target insects (Figure 4). Non-target insects were dominated (85.7%) by non-biting, small Chironomidae, which were occasionally captured by the hundreds, especially in one of the traps located under a street light. Their capture was detected only after adding a second net of a smaller mesh size (1 × 0.5 mm instead of 1.5 × 1 mm) to avoid Ceratopogonidae from escaping from the traps. Fourteen other taxa were also captured in roughly equal proportions, representing globally 5.5% of all the captures in the traps (Figure 4).The house martin Delichon urbicum is a migratory aerial insectivore that breeds colonially in human-inhabited areas. It feeds upon various arthropod species that are caught on the wing within 500 m from the nest [10,11]. In Camargue, breeding extends from early May (laying period) to mid-August (fledging of young from second breeding attempt), with a third of the chick diet being composed of small Nematocera [3]. While mosquito control using Bti spraying had a significant impact on the breeding success of house martins (F(2, 212) = 16.2, p < 0.0001), the use of traps revealed a similar breeding success to the one reported outside the Bti-sprayed area (Figure 5). The mean number of young fledged per nest was 3.3 at sites without mosquito control, 3.1 at the site with Techno Bam traps and 2.2 at sites treated with Bti. According to post-hoc Fisher’s LSD tests, breeding success at Sambuc in 2016 (with traps) was not different (p = 0.24) from that observed in the preceding years at the control sites (including Sambuc), but differed significantly (p = 0.03) from that of sites surrounded by Bti-sprayed wetlands. Although mosquito traps using CO2 and olfactive lures to attract mosquitoes are commonly used in surveillance programs [9,12,13], few studies have experimentally tested their usefulness as a means of mosquito control on a relatively large spatial scale [14]. Considering the high environmental and economic costs of spraying insecticide, this technique appears as the most promising, with a performance similar to traditional methods for controlling mosquitoes [15]. The use of 16 Techno Bam traps spread over 1.5 km within a hamlet of 600 inhabitants allowed us to reduce the mosquito nuisance by 70%. This performance, assessed by comparing the number of mosquitoes landing on human bait within and outside the hamlet, before and during trapping operations, was associated with the catch of nearly 300,000 females from nine mosquito species. Mosquito peaks were nevertheless observed over the six-month sampling season in the controlled areas. These were mostly related to Anopheles hyrcanus, which accounted for 81% of the residual nuisance observed in late August. The lower trap performance against this species (46% reduction) could be related to the type of olfactive lure used [16]. When Anopheles hyrcanus is excluded from our calf-test samples, the performance of the Techno Bam traps reaches 85% in terms of nuisance reduction. An unpublished experiment comparing the performance of the first Techno Bam prototype with Biogent Sentinel traps suggested that lures using lactic acid are more effective against An. hyrcanus than those using octenol. The large discrepancy in the mean number of daily captures among traps (range 24–399) suggests that the performance is influenced by trap placement within the hamlet. Sunlight has been shown to negatively influence the capture probability of Aedes albopictus [17], but literature on this subject is relatively scant. We would also expect wind exposure and the presence of vegetation to influence the mosquito capture rates. Testing this new approach to mosquito control in Camargue was motivated by the significant impacts revealed by Bti spraying on natural predators of mosquitoes and chironomids [2,3,4,5]. In contrast to larvicide spraying of natural areas, the environmental impact of traps is expected to be negligible, being mostly limited to the impoverished fauna found in urbanized areas where the traps are located. Some 86% of the non-target insects captured in the traps were very small chironomids attracted by street lights. Because non-target insects are presumably not attracted by carbon dioxide, only those individuals flying incidentally close to the trap will be caught by the fan aspiration. Hence, although small chironomids accounted for a third of all captures, the proportion caught was presumably negligible relative to their local abundance. Techno Bam traps did not affect the breeding success of house martins nesting colonially at the proximity of traps. These results suggest that the local use of traps has no impact on insects fed to nestlings in contrast to the Bti spraying of wetlands surrounding urban areas where these birds are nesting [3].This study provides the first experimental data on the performance of a public network of mosquito traps as a means of mosquito control to improve human comfort in a locality. The lack of data on the efficacy of Bti spraying, which has been carried out since 2006 in Camargue, does not allow us to quantitatively compare the performance of both techniques. However, traps are qualitatively more versatile as they capture all mosquito species potentially causing a nuisance in human-inhabited areas (Bti spraying targets only Ochlerotatus caspius and Oc. Detritus), they are more economical in spite of their relatively high maintenance costs, and they have a negligible impact on wildlife. Because they are located in human-inhabited areas, mosquito traps could provide a useful complementary tool for the control of container-inhabiting species such as Aedes albopictus and Aedes aegypti, which pose public health problems, and for which traditional integrated mosquito management approaches based on larvae control are inefficient [18,19,20]. Our experiment suggest that the observed 70% reduction in the mosquito nuisance could be increased by combining different olfactive lures, by optimizing the position of traps relative to environmental conditions, and by increasing trap numbers to improve the protecting belt effect. We are indebted to the Parc Natural Régional de Camargue (PNRC) for the coordination of this study, and to the Region Provence Alpe Côte d’Azur, the Departmental Council of Bouches-du-Rhône, and the local authorities of Arles-Crau-Camargue-Montagnette and Metropole Aix-Marseille-Provence for funding the scientific monitoring and use of Techno Bam traps. We are grateful to Loïc Willm for producing the map of the study site and to all volunteers who assisted in the collection of samples and kindly lent their calf to science, in particular Catherine Lavallée-Chouinard, Erika Audry, and Céline Hanzen.B.P. and G.L. conceived and designed the experiments; C.M.-K. collected and managed the data; S.H. supervised insect identification; G.L. analyzed the data; B.P. wrote the paper and obtained the funding.The authors declare no conflict of interest. The founding sponsors and Techno Bam had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, and in the decision to publish the results.Deployment of the 16 Techno Bam traps in the Sambuc hamlet in 2016 with their 60-m attraction radius for mosquitoes relative to location of human bait tests and the breeding colony of house martins (Delichon urbicum).Weekly variation in the mean number of mosquitoes captured daily in each of the 16 Techno Bam traps located at Sambuc from April through October 2016.Temporal variation in the mean number of biting attempts at treated (10–40 m from traps) and control (550–1130 m from traps) sites from April to October 2016.Mean daily captures from each taxonomic group based on 39,941 items identified in 86 Techno Bam trap samples at the Sambuc in 2016.Mean breeding success of house martin in two Bti-sprayed areas and two control areas (including Sambuc) in Camargue between 2009 and 2015 and with Techno Bam traps in 2016.Capture rates and trap performance for the different mosquito species sampled in 2016.

Med-MDPI/ijerph_2/ijerph-14-03-00314.txt

@@ -0,0 +1 @@
+Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).Passive houses and other highly energy-efficient buildings need mechanical ventilation. However, ventilation systems in such houses are regarded with a certain degree of skepticism by parts of the public due to alleged negative health effects. Within a quasi-experimental field study, we investigated if occupants of two types of buildings (mechanical vs. natural ventilation) experience different health, wellbeing and housing satisfaction outcomes and if associations with indoor air quality exist. We investigated 123 modern homes (test group: with mechanical ventilation; control group: naturally ventilated) built in the years 2010 to 2012 in the same geographic area and price range. Interviews of occupants based on standardized questionnaires and measurements of indoor air quality parameters were conducted twice (three months after moving in and one year later). In total, 575 interviews were performed (respondents’ mean age 37.9 ± 9 years in the test group, 37.7 ± 9 years in the control group). Occupants of the test group rated their overall health status and that of their children not significantly higher than occupants of the control group at both time points. Adult occupants of the test group reported dry eyes statistically significantly more frequently compared to the control group (19.4% vs. 12.5%). Inhabitants of energy-efficient, mechanically ventilated homes rated the quality of indoor air and climate significantly higher. Self-reported health improved more frequently in the mechanically ventilated new homes (p = 0.005). Almost no other significant differences between housing types and measuring time points were observed concerning health and wellbeing or housing satisfaction. Associations between vegetative symptoms (dizziness, nausea, headaches) and formaldehyde concentrations as well as between CO2 levels and perceived stale air were observed. However, both associations were independent of the type of ventilation. In summary, occupants of the mechanically ventilated homes rated their health status slightly higher and their health improved significantly more frequently than in occupants of the control group. As humidity in homes with mechanical ventilation was lower, it seems plausible that the inhabitants reported dry eyes more frequently.Very energy-efficient homes, such as passive houses, are those which meet rigorous energy efficiency standards. Because of the air tightness, such buildings need built-in mechanical ventilation [1]. Heat recovery systems are necessary in order to minimize energy loss [1,2]. There are, however, concerns that such ventilation systems may impact health through exposure to excess noise, draughts, and indoor air pollution as a consequence of insufficient cleaning of the air duct system and low levels of indoor air humidity due to an increased volume of outdoor air in winter [2,3,4]. Energy-efficient homes without mechanical ventilation have also been found to be associated with an increased risk of asthma in the United Kingdom [5]. However, a meta-analysis by Maidment and colleagues concluded that energy efficiency interventions led to a small but statistically significant improvement in the health of residents [6].In a study with more than 3000 measurements (chemical pollutants, biological contaminants, indoor climate parameters) we found that the indoor air quality in highly energy-efficient, mechanically ventilated homes was higher than that of conventional homes [7]. Pollutant concentrations in French low-energy school buildings with ventilation systems were lower than in conventional school buildings [8]. A few studies reported that ventilation systems in homes lead to a reduction in reported health symptoms and improvements in overall health [9,10,11], likely attributable to an increased air exchange and thus an improvement in indoor air quality.Leech et al. [9] examined self-reported changes in health status by telephone-administered questionnaires in occupants of new homes in Canada. Occupants of the test group (energy-efficient homes with heat recovery ventilators) provided a health benefit over one year of occupancy.A study in Cornwall, UK (The Breath of Fresh Air Project), investigated the health of asthmatic children in 17 homes [10]. Indoor measurements and health assessments (by questionnaires) were conducted before and after the installation of mechanical ventilation and heat recovery (MHRV) systems. Installations of MHRV systems reduced mite allergen concentrations and children’s asthma symptoms.After “green” renovation (installation of mechanical ventilation, tightening of the building envelope, etc.) of low-income housing in Minnesota, participants’ health and building performance were assessed [11]. Health was assessed via questionnaire. Interviews were administered after residents moved into renovated apartments and approximately 12 to 18 months later. The renovation produced improvements in health, and energy use was reduced by 45% over the one-year period.In this paper we compared the self-rated health and wellbeing of inhabitants of very energy-efficient homes to the health of inhabitants of conventional new houses without mechanical ventilation. In addition, we also evaluated the participants’ perception of the indoor air quality (e.g., stale air as a consequence of increased CO2 or smells) and indoor climate (temperature, humidity, air movement) and their housing satisfaction.Inhabitants of new houses built according to very low energy or passive house standards (Austrian Standard B 8110-1) [12] formed the test group. The houses had no air conditioning. Inhabitants of houses which corresponded to the normal building standards without mechanical ventilation systems formed the control group. It was assumed that in the buildings of the test group the air supply was provided both mechanically and via ventilation through windows (and doors). The study was approved by the ethics committee of the Medical University of Vienna (377/2010).Recruitment of participants is described in [7]. The buildings were located in all provinces of Austria and were built between 2010 and 2012. In both groups, detached houses constituted approximately 70% of the sample. The remaining 30% were apartments in multistory buildings.Interviews were conducted at two different time points (first interview and follow-up interview). Also measurements of indoor parameters (climate, chemical pollutants and biological contaminants) were conducted twice according to standardized analytical methods (e.g., formaldehyde according to ISO 16000-2 and 3 [13,14]). Methods and results of the measurements were reported in [7].The first interview (measurement point T1, n = 293, between October 2010 and May 2012) occurred at approximately three months (±3 weeks) after moving into a new house/apartment, with a follow-up interview (measurement point T2, n = 282, between October 2011 and May 2013) one year later. The drop-out rate was 4%.Interviews were conducted with a structured questionnaire. The questionnaire consisted of the standardized questionnaire SF-36 (36-Item Short Form Survey, [15]) and a section of the wellbeing questionnaire used in AUPHEP (Austrian Project on Health Effects of Particulates, [16]). It consisted of the following parts: Socio-demographic characteristics of the participants, respiratory symptoms and allergies, unspecific symptoms; perception of indoor air quality and climate; satisfaction with the housing situation.Collected data were analysed using SPSS 20.0 for Windows (SPSS Inc., Chicago, IL, USA). Comparisons of categorical data across groups were done by chi-square tests, comparison of time points within groups were done by Bowker’s symmetry tests. Symptom ratings were combined into scores (psychasthenic symptoms, vegetative symptoms). Also air quality ratings were combined into scores with positive attributes (fresh, clean, pleasant, fragrant) into one and negative (stale, stuffy, stagnant, bad smelling, smoky) into another score. These scores as well as climate ratings were McCall transformed (standardized scores: mean 0, standard deviation 1) and subjected to analyses of covariance with group as between subjects and time points as within subjects factor and gender and age as covariates. For the analyses of relationships between ratings and measurements a log transformation of air quality and climate data was performed and linear regression analysis including age and gender as potential confounders was done. For all analyses p-values below 0.05 were considered significant.In total, 575 interviews (test group: 299, control group: 276) were conducted between October 2010 and May 2013 (first time point: n = 293, second time point: n = 282). Of these, 409 interviews were conducted with adults. Parents also filled in questionnaires for the 166 (86 control group, 80 test group) children (<16 years of age) included.The average age of adults in both test and control groups at T1 was virtually the same (37.9 ± 9 years in the test group, 37.7 ± 9 years in the control group); children were, on average, 5.7 years of age in the test group and 7.5 years in the control group. The average household size included 2.8 ± 1.1 participants in both groups; most households consisted of couples and included, on average, 0.8 children <16 years of age (for both the test and control groups).Smokers accounted for 18.4% in the test group and 25.4% in the control group. Of these, only 0.5% (test group) and 2.6% (control group) smoked in their apartment or house.Due to the relatively high costs of such homes, all participants belonged to the upper-middle class (more than 12 years of education, household income above median).There were slight differences by housing type for both adults and children in health ratings. Participants in the test group rated their own health and that of their children higher compared with the control group: 24.9% (average of the ratings at T1 and T2) considered themselves and 50.6% considered their children to be in excellent health, compared with 19.8% and 38.4%, respectively, in the control group (Table 1 and Table 2).Table 1 also shows the health ratings before moving in (recall at T1). In adults, there was a trend towards improvement in the state of health after moving in, but this was not statistically significant. After participants had been living in their new home for more than one year (T2), there was barely any change in the state of health in comparison to T1.The children’s state of health was rated only twice, three months after moving in at T1 and one year later at T2 (Table 2). There was a marked difference over time in the control group (p < 0.05): in this group, parents perceived their children’s health to be considerably better one year after moving in, compared with a downwards shift from “excellent” to “very good” in the test group. In both groups, “good and less good” ratings changed to a more positive perception over time.In addition, the participants rated their change in health status over the last year at T2. At this time (T2), participants had already been living at their new address for approximately one year. Of adults who moved into new housing with a mechanical ventilation system, 19.1% experienced improvements in their health, while 80% saw no change and 0.9% noted some deterioration. In contrast, 13.2% of adults who had moved into buildings without mechanical ventilation perceived that their health had deteriorated, 69.2% felt no change and 17.6% noted some improvement. This difference was statistically significant (p = 0.005).When adult participants were asked to predict their future state of health, 4.3% of the test group and 7.9% of the control group believed that their health would likely deteriorate.In total, 29.9% of all participants reported having allergies; 33.6% of adults and 12.7% of children in the test group and 37.5% of adults and 17.8% of children in the control group were affected. Adults in the test group suffered from an average of two allergies, compared with an average of 2.1 in the control group. The average number of allergies in children was one in the test group and 1.5 in the control group. The most common types were pollen allergies (32.6%), followed by pet hair allergies (23.3%), dust mite allergies (22.2%) and food allergies (16.1%).There was no significant difference in the number or frequency of allergies in residences with mechanical ventilation systems; however, allergies against pollen, pet hair and insects were less frequently observed in occupants with mechanical ventilation (p < 0.05). Prevalence, type or number of allergies did not change over time in either group.In total, 35.5% of the participants experienced, within the four weeks before measurement, dryness of the airways, 22.6% felt a burning sensation in their nose or throat and 13.9% had dry, red or itchy eyes without having a coinciding cold or having visited a swimming pool. Further, 12.5% of participants had been coughing for more than two weeks within the preceding four weeks of the interview.There were no significant differences by housing type in prevalence and number of colds, dryness of the airways, burning sensations in the nose or throat, or coughs; however, adults in the test group had a significantly higher prevalence of dry eyes (19.4%) compared to the control group (12.5%), independent of contact lens usage (p = 0.04). There were no significant changes in the evaluated health complaints in either group over time.Participants were asked how often they experienced the following symptoms in the last four weeks: tiredness, exhaustion, headaches, nausea, dizziness, impaired concentration, anxiety, nervousness, mood changes, and limited performance. The results are shown in Table 3.Neither adults nor children showed any differences by type of residence. Children were generally less affected by the listed health impairments compared with adults.In the test group, there was an increase in tiredness, exhaustion and nervousness after one year; in the control group, increased difficulty to concentrate and nervousness were observed after one year. These differences did not reach statistical significance. The average numbers of health impairments were similar in both groups (Table 3).Negative perceptions of the quality of air (stale, stuffy, stagnant, bad smelling, smoky) were found more frequently in homes without mechanical ventilation. Differences in negative perceptions of indoor air quality between the groups were overall highly significant (p < 0.01), with the exception of “bad smelling” and “smoky”. The results are shown in Table 4.The difference in the positive perception of air quality between groups was highly significant for the attributes “pleasant” and “fresh” (p < 0.01), and significant for the attribute “clean” (p < 0.05). In all these cases, positive perception was more frequent in homes with ventilation systems (Table 5). Air quality ratings did not significantly change in the period between measuring points.The perception of indoor climate, smell and noise is presented in Table 6. No significant differences between measurement points were found, with the exception of satisfaction with humidity in the test group, which 63.9% of participants rated as “just right” at T1, compared with only 53.6% at T2. Therefore, we only report here the average perception (T1 and T2) of indoor climate over one year. Participants who lived in housing with mechanical ventilation (test group) rated temperature and air movement in their homes as significantly more pleasant (p < 0.01) compared with the control group: 77.0% of participants in the test group and 65.2% in the control group rated their room temperature as “just right”; 7.9% and 12.1%, respectively, rated it as “(too) cold”; and 15.2% and 22.7%, respectively, as “(too) warm”. Air movement was considered to be “just right” by 80.6% of participants in the test group and 66.7% in the control group. More participants in the control group compared with the test group complained about draught (29.1% vs. 14.3%, respectively).The control group rated the humidity in their home significantly better compared with the test group: 58.8% of participants in the test group considered the humidity to be “just right”, compared with 67.2% in the control group (p < 0.01); 40.6% of participants in the test group thought that the air at their home was (too) dry, compared with 26.4% in the control group.There were no significant differences regarding annoyance due to smells or noise between groups. Between 48.2% and 56.1% rated smell and noise as “not annoying at all”.Most participants felt that their current housing situation had improved significantly compared with their previous housing situation. Accordingly, 80.6% of participant who lived in housing with mechanical ventilation and 72.0% of participants of the control group felt much more satisfied with their housing conditions; 12.0% and 16.9%, respectively, felt rather more satisfied and 7.4% and 11.0%, respectively, felt neither more nor less satisfied, or dissatisfied.Satisfaction with the housing situation around the time of T1, three months after moving in was completed, was particularly high in the test group: 86.9% were very content with their housing situation, 10.3% were content and only 2.8% were neither content nor not content, or dissatisfied (Table 7). In the control group, 76.5% were very content at the time of T1, 21.4% were content and 2.0% were neither content nor not content, or dissatisfied. Differences in the satisfaction at the time of T1 were not statistically significant.At the point of T2 (one year after T1), there was only a slight reduction in the level of satisfaction (Table 7).With regard to the neighborhood conditions, the following observations were made: At the time of T1, 70.1% of participants in the test group declared that they were very satisfied with their neighborhood conditions, 25.2% were satisfied and 3.7% were dissatisfied or neither satisfied nor dissatisfied, compared with 72.4%, 24.5% and 3%, respectively, in the control group. At the time of T2, 68.2% in the test group and 64.8% in the control group were found to be very satisfied with their neighborhood conditions; 26.4% and 31.9%, respectively, were satisfied and 4.4% and 3.3%, respectively, were dissatisfied or neither satisfied nor dissatisfied. None of these findings indicated statistical significance between groups.Participants in the test group perceived themselves to be significantly more satisfied than their peer group, compared with the control group. Further, 64.1% of participants in the test group and 54.0% participants in the control group estimated that, compared with family and friends, they were much more content with their living situation; 26.3% and 27.5%, respectively, were rather more content, 7.4% and 17.5%, respectively, were equally content and 2.2% and 1.0%, respectively, considered themselves to be more dissatisfied (p < 0.01).There was a weak but statistically significant correlation between the frequency of vegetative symptoms (dizziness, nausea, headaches) and the concentration of aldehydes, in particular formaldehyde, at T2 (Figure 1). This correlation was independent of the type of ventilation, although it has to be noted that indoor formaldehyde concentrations in the test group were significantly lower [7].There was also a significant correlation between the indoor CO2 concentration (especially the highest hourly CO2 mean value = maximum hourly mean) and the perception of stale indoor air (Figure 2). This correlation was also independent of the study group. No other significant correlations could be found.To our knowledge this was, besides, inter alia, the investigations of Leech et al. [9] and Takaro et al. [17], one of the first studies investigating the perceived health of inhabitants of highly energy-efficient homes. Our inspiration to conduct the current study was the Canadian study by Leech et al. [9].Inhabitants of buildings with mechanical ventilation systems in Austria rated their state of health and that of their children slightly higher than participants who lived in dwellings with natural ventilation only.Furthermore, after about 15 months in their new homes, respondents perceived significantly more frequent improvements over the last year if they had lived in housing with mechanical ventilation. This might, in part, be explained by the better air quality [7] in these homes. Leech et al. [9] also found that new occupants of energy-efficient homes (with ventilation systems) reported an improvement over one year in health in comparison with control home occupants.No significant differences by housing type or time points were observed regarding the frequency and number of almost all minor ailments or health complaints. Allergies against pollen, pet hair and insects were less frequent in the test group. However, as percentages did not change over time, it seems rather unlikely that the difference in frequency was due to the housing type.Adults in the test group suffered significantly more frequently from dry eyes compared with adults in the control group. This might be due to the lower humidity in the homes with mechanical ventilation [1]. Accordingly, 40.6% of participants in the test group thought that the air in their home was (too) dry compared with 26.4% in the control group. Measurements also showed that humidity was lower in the test group [7].We found a weak but statistically significant correlation between the frequency of vegetative symptoms (dizziness, nausea, headaches) and the concentrations of formaldehyde. Such symptoms have been described in the literature also at relatively low levels of formaldehyde, even at or below 0.10 mg/m3 [18,19,20]. They may also be associated with other indoor pollutants including CO2 (an indicator of adequate ventilation) and smells. However, no such correlations could be found.Satisfaction with housing and living area in both groups was relatively high: 84.3% of the test group and 76.2% of the control group were very content with their housing; 69.1% and 68.8%, respectively, were very content with the living area. These differences in housing satisfaction between the test and control groups were not significant. However, participants in the test group perceived themselves to be significantly more satisfied with their homes than their peer group, compared with the control group. This may in part be explained by the fact that very energy-efficient homes with energy recovery ventilation systems are still “special” houses in Austria.According to the results of the measurements in the studied homes [7], there were highly significant differences regarding the subjectively perceived quality of air between both groups, with a perceived higher quality of air in the test group. Temperature and air movement were rated significantly more pleasant in the test group. There were no differences between groups regarding smell and noise exposure.In conclusion, inhabitants of new energy efficient buildings with mechanical ventilation generally rated their health and the quality of the indoor air and climate better compared with those who lived in dwellings with window (and doors) ventilation only. However, adults in homes with mechanical ventilation—where humidity was lower [7]—suffered more frequently from dry eyes and found the indoor air (too) dry.We want to thank DI Claudia Schmöger and Gabriela Langer for their assistance and Kristina Standeven for translational help. The study was financially supported by the Austrian Climate and Energy Fund and by the Austrian Research Promotion Agency (FFG).Peter Wallner, Peter Tappler, Ute Munoz, Bernhard Damberger, Anna Wanka and Hans-Peter Hutter performed the experiments and analyzed the data. Peter Wallner, Peter Tappler, Anna Wanka, Michael Kundi and Hans-Peter Hutter designed the experiments and wrote the paper. All authors read and approved the final manuscript.The authors declare no conflict of interest. The founding sponsors had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, and in the decision to publish the results.The following abbreviations are used in this manuscript:Correlation between indoor concentration of formaldehyde and frequency of vegetative symptoms (dizziness, nausea, headaches; standardized score) at T2 (about 1.3 years after moving in) (R2 = 2.3%).Correlation between maximum hourly mean of indoor CO2 concentration (in bedrooms) and perception of stale indoor air (standardized score) at T2 (about 1.3 years after moving in) (R2 = 3%).Subjective health ratings (percentages) before moving in, at T1 (three months after moving in) and at T2 (one year later).Chi2 test: test vs. control group: Before moving: p = 0.472; T1: p = 0.621; T2: p = 0.735.Parental rating of their children’s health (percentages) at T1 (three months after moving in) and at T2 (one year later).Chi2 test: test vs. control group: T1: p = 0.006; T2: p = 0.086.Prevalence of symptoms or health impairments (“always” and “often”) in both groups at both measuring time points (T1: three months after moving in; T2: one year later).Negative perception of indoor air quality in both groups at T1 (three months after moving in) and T2 (one year later).* Answer categories 2–5 (“a little” to “predominantly”) are combined (percentages).Positive perception of indoor air quality in both groups at T1 (three months after moving in) and T2 (one year later).* Percentages are related to answer category 5 (“predominantly”).Perception of indoor climate, smell and noise. Percentage of participants who answered with “just right” regarding room temperature, humidity, air movement or “not annoying at all” regarding smell and noise. T1 (three months after moving in) and T2 (one year later).Satisfaction with the housing situation in both groups T1 (three months after moving in) and T2 (one year later).

Med-MDPI/ijerph_2/ijerph-14-03-00315.txt

@@ -0,0 +1 @@
+Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).The epidemiology of malaria, anaemia and malnutrition in children is potentially altered in mining development areas. In a copper extraction project in northwestern Zambia, a health impact assessment (HIA) was commissioned to predict, manage and monitor health impacts. Two cross-sectional surveys were conducted: at baseline prior to project development (2011) and at four years into development (2015). Prevalence of Plasmodium falciparum, anaemia and stunting were assessed in under-five-year-old children, while hookworm infection was assessed in children aged 9–14 years in communities impacted and comparison communities not impacted by the project. P. falciparum prevalence was significantly higher in 2015 compared to 2011 in both impacted and comparison communities (odds ratio (OR) = 2.51 and OR = 6.97, respectively). Stunting was significantly lower in 2015 in impacted communities only (OR = 0.63). Anaemia was slightly lower in 2015 compared to baseline in both impacted and comparison communities. Resettlement due to the project and migration background (i.e., moving into the area within the past five years) were generally associated with better health outcomes in 2015. We conclude that repeated cross-sectional surveys to monitor health in communities impacted by projects should become an integral part of HIA to deepen the understanding of changing patterns of health and support implementation of setting-specific public health measures.Solwezi district in the Northwestern Province of Zambia has traditionally been a rural, sparsely populated area [1]. However, recent mining developments (i.e., the Kansanshi and Lumwana copper mines) have accelerated in-migration and altered the socioeconomic profile of the district [2,3]. In 2009, the Trident project—a copper mine operated by First Quantum Minerals Limited (FQML)—was launched [4,5]. The mine, which became operational in 2015, is a green field development in a previously remote forested area, covering a lease area of approximately 950 km2. The development included construction of an open pit mine, processing plant, power lines, airstrip, maintenance and administrative infrastructure, access roads and a new residential settlement for the mine workforce and their families. The project development spurred socio-demographic and economic changes in the local community, including physical resettlement, influx of job- and opportunity-seeking migrants, shift in livelihood strategies and urbanization [6,7]. Hence, the direct and indirect ecological, social, economic and health impacts placed on the communities living in this area have been considerable [8,9].Traditionally, studies determining community health impacts associated with mining have focused on HIV and other sexually transmitted infections (STIs), tuberculosis, water and air quality or exposure to hazardous chemical substances [10,11]. Furthermore, malaria is often considered by companies operating in the tropics because of its significant contribution to the local burden of disease and workplace health implications [12]. However, it is less evident how conditions that are especially prevalent in children living in low- and middle-income countries, such as anaemia, diarrhoeal diseases, respiratory tract infections, intestinal parasitic infections or malnutrition are affected by project-related transformations over longer periods of time.Health impact assessment (HIA) is the recommended approach to predict potential effects of industrial projects on the health of affected populations by considering a broad range of social, cultural, economic and ecological determinants of health [13,14]. As part of the Trident project’s feasibility studies, an HIA was commissioned to assist in the identification of potential health impacts and development of a community health management plan to prevent adverse health impacts and maximize health benefits. During the scoping phase of the HIA, a number of health data gaps were identified, which warranted additional primary data collection [15,16]. Hence, a cross-sectional baseline health survey (BHS) was conducted in 2011 [17]. Data from the BHS and secondary data sources (e.g., local health statistics) provided an evidence-base for the subsequent risk assessment phase of the HIA [18]. Therefore, the identified potential health impacts were ranked based on their significance (i.e., impact severity and likelihood of occurrence) using a semi-quantitative risk-ranking matrix [19]. A community health management and monitoring plan was developed that combines continuous and periodic data collection approaches, including district health information system data and repeated cross-sectional health surveys at four-year intervals. While some diseases warrant continuous surveillance depending on their aetiology and significance (e.g., HIV), repeated cross-sectional household surveys at 3–5-year intervals measuring key health indicators are a valid option to observe conditions in communities that may change over longer periods of time (e.g., stunting) and also to allow for assessment of true prevalences as well as knowledge, attitudes and practices (KAP) [20,21,22].Here, we present data from two cross-sectional epidemiological surveys: the 2011 BHS, prior to project development, and the first follow-up health survey completed in 2015, hence, four years into project development. Among the broad spectrum of indicators assessed, Table 1 summarizes the ones selected based on their significance for child health and relevance in the current project setting. The paper specifically discusses trends over the four-year period and makes comparisons between impacted communities (i.e., affected by the project development) and non-impacted comparison communities, and describes associated determinants at household and community levels.The study protocols for the two cross-sectional surveys received approval from the ethics review committee of the Tropical Disease Research Centre (Ndola, Zambia; registration number 00003729). The Solwezi District Health Department supported the studies as a key government partner, with contributions to the study design, community sensitization and fieldwork. At the household level, informed consent (signed or fingerprinted) was obtained from heads of households or mothers/caregivers. At the school level, sensitization activities included school visits prior to the survey. Teachers were informed about the objectives and procedures of the study and consent was obtained by teachers informing parents about the study, who in turn provided written permission to allow their children to participate in the survey. Children assented orally. Children who were found positive for Plasmodium falciparum infection using a rapid diagnostic test (RDT) were treated with an artemisinin-based combination therapy, using artemether-lumefantrine, following national protocols. Children found with mild and moderate anaemia (haemoglobin (Hb) 7–11.0 g/dL) were provided with iron and multivitamin supplements, while severe cases (Hb < 7 g/dL, or those with any signs/symptoms of severe anaemia) were referred to the nearest health facility thereby adhering to the public health referral system followed in Zambia. All children who provided stool specimens for parasitological testing were given a single oral dose of albendazole (400 mg).The Trident project is located about 150 km northwest of Solwezi town, the district capital (Figure 1). Chisasa is the major settlement in the study area, at the junction along the T5 highway connecting Solwezi to Mwinilunga district. At the time of the BHS in 2011, over 60% of the adult population was involved in subsistence agriculture and about 2% employed by the project [6]. In 2015, about 35% of the households in the impacted sites had at least one member employed by, or working as a subcontractor for, the project [7].Two cross-sectional, epidemiological surveys were conducted in July 2011 and July 2015, using the same methodology. Considering the heterogeneity in the distribution of project-related health impacts expected across communities, a stepwise, semi-purposive sentinel site sampling strategy, rather than a fully randomised design, was employed in both surveys [17]. In a first step, all villages potentially affected by the project were identified, whereas “potentially” refers to an impact that may or may not occur and “affected” refers to being affected by either a direct impact (e.g., resettlement or project-sponsored health interventions) or an indirect impact (e.g., in-migration along transport corridors) caused by the project [8]. In a second step, impacted sites were semi-purposively selected based on the magnitude and nature of project-related impacts (e.g., resettled communities and communities along transport corridors). Our approach allowed for sampling of smaller, potentially impacted communities that might otherwise have been excluded, had a random cluster sampling proportional to population size been employed [35]. In a third step, comparison sites were chosen based on their socio-demographic and topographic similarity to the impacted sites as well as proximity to the project area, with two inclusion criteria: (i) located outside the project area; and (ii) no or only limited project-associated impacts such as no project-sponsored health interventions in the community or project employees/contractors residing in the community. In the final step, households were randomly selected within the sentinel sites, with the inclusion criteria at the level of the household requiring the presence of a mother (≥15 years) with at least one child under the age of five years. In parallel, schoolchildren from primary schools in the sentinel sites were sampled to screen for hookworm infection, the most prevalent soil-transmitted helminth in the study area [36].The full list of sentinel sites selected for the 2011 BHS and the 2015 follow-up is shown in Figure 2. In seven sentinel sites, data were collected in both the BHS and follow-up. For an additional seven sentinel sites, data were only available for 2015. This included two impacted sites: the newly developed Kalumbila Town (employee residential area) and Shenengene (a resettlement village). One impacted site was added due to increased importance (Kanzanji became the base of a major mining contractor) and four additional comparison sites were included to augment statistical power for comparison in future surveys. Importantly, findings from Wanyinwa (sampled during the 2011 BHS) are comparable to findings from Northern Resettlement (sampled during the 2015 follow-up) as 97% of the participating households in Northern Resettlement originated from Wanyinwa.The surveys included two main data collection methods: (i) a questionnaire interview with caregivers (≥15 years) in the household; and (ii) an assessment of biomedical indicators in children under the age of five years in a mobile field laboratory. The questionnaire focused on KAP related to issues such as health seeking behaviour, maternal and child health, infectious diseases and participation in health interventions. In addition, basic socio-demographic information was collected, including information on recent in-migration (defined as duration of residency in the current location of less than five years). The questionnaire is provided as a supplementary file S1.On completion of the questionnaire, caregivers together with their under-five-year-old children were asked to visit the field laboratory for the assessment of biomedical indicators. An RDT was used to assess P. falciparum infection from a finger-prick capillary blood sample in children aged 6–59 months (see Table 1). Hb concentration was measured in a capillary blood sample from children aged 6–59 months to determine anaemia (defined as Hb < 11 g/dL). Children aged < 5 years had their weight and height measured.At each school enrolled in the survey, a quota of at least 15 boys and 15 girls was randomly selected. Therefore, all eligible children (i.e., present at the day of the survey; aged 9–14 years) were listed and numbered and the quota was selected using random number sampling. A fresh morning stool sample was collected and subjected to the Kato-Katz technique. A single 41.7 mg thick-smear was examined within 20–40 min for enumeration of hookworm eggs [34]. Eggs were counted and multiplied by a factor of 24 to determine eggs per gram of stool (EPG).In 2011, questionnaire data were entered into EpiData software (EpiData Association; Odense, Denmark). In 2015, data were collected through electronic tablets using the open data kit (ODK) software. Analysis was performed with Stata (StataCorp LP, College Station, TX, USA). Frequencies and odds ratios (ORs) with corresponding 95% confidence intervals (CIs) were determined. Mixed effects logistic regression models were used taking into account clustering at the levels of sentinel sites and of households. The model included a factor for year to capture potential period effects, a factor for type of site (impacted vs. comparison) and an interaction term between the two factors to assess potential differences in changes of prevalence rates from 2011 to 2015 between impacted and comparison sentinel sites. Of note, for 2011 and 2015 comparisons, only sentinel sites that were sampled in both surveys were considered. For analysis with 2015 data only, all 14 sentinel sites were considered.The study populations in 2011 and 2015 are shown in Table 2. In 2011, 289 households were sampled from seven sentinel sites, and in 2015, 516 households were sampled from 14 sentinel sites, with a total sample of 483 and 949 children under the age of five years, respectively. Additionally, 309 (2011) and 477 (2015) children aged 9–14 years were sampled from the selected schools. For 2015 only, the proportions of household with resettlement or migration background and the proportion of households using improved sanitation facilities are shown.At baseline, children in impacted sites showed a lower odds for P. falciparum infection (OR = 0.33, 95% CI 0.05–2.20; Table 3). There was a significantly higher prevalence in 2015 compared to 2011 in all sites and overall in both the impacted and comparison sites, with ORs of 2.51 (95% CI 1.56–4.02) and 6.97 (95% CI 2.20–22.0), respectively, but with no significant different period effect between impacted vs. comparison (OR = 0.36, 95% CI 0.10–1.23).In Figure 3a, the prevalences of 2011 (x-axis) and 2015 (y-axis) are plotted against each other. Communities whose prevalence has increased are plotted in the upper left half of the graph coloured in red and communities whose prevalence has decreased are plotted in the lower right half of the graph coloured in green. Communities whose prevalence has remained stable are located on, or close to the grey line. Wanyinwa/Northern Resettlement and Chisasa were least affected by P. falciparum infection in both 2011 and 2015. In 2015, both communities exhibited high proportions of resettled households and new settlers and, as illustrated in Figure 4a, children with a resettlement or migration background had significantly lower odds of being infected with P. falciparum.At baseline, the stunting rate was slightly higher in the impacted compared to the comparison sites but with no statistical significance (OR = 1.61, 95% CI 0.77–3.35; Table 3). In 2015, stunting was significantly lower in the impacted sites compared to 2011 (OR = 0.63, 95% CI 0.46–0.87), whilst in the comparison sites stunting was higher in 2015 (OR = 1.41, 95% CI 0.58–3.46).Two factors significantly lowered the risk for stunting in the 2015 study population: (i) access to improved sanitation facilities; and (ii) originating from the richest wealth quartile (Figure 4b). In 2015, children from Northern Resettlement were least affected by stunting (Figure 3b).While the difference in anaemia prevalence between impacted and comparison sites was significant at baseline (p = 0.04), there were no significant changes over time in the two categories of sites, although it decreased in both; from 46.6% (194/416) to 41.9% (173/413) in the impacted and from 65.1% (28/43) to 50.8% (33/65) in the comparison sites, respectively (Table 4). Anaemia prevalence was lower in 2015 in all but two sentinel sites (i.e., Chisasa and Musele), where it remained stable (Figure 3c). In 2015, factors significantly associated with anaemia in a child were a concurrent P. falciparum parasitaemia and stunted growth (Figure 4c).The overall prevalence of hookworm infection slightly decreased from 62.5% (172/275) to 60.9% (145/238) in the impacted sites and from 58.8% (20/34) to 50.0% (15/30) in the comparison sites. Hence, the rates of infection did not change significantly over time (p = 0.71 and p = 0.47, respectively; Table 4). Chisasa had the lowest infection rate in 2015, which was however higher than the rate recorded in 2011 (Figure 3d).Presented here is a selection of indicators in children from two cross-sectional surveys spaced by four years within the frame of the Trident copper development project in Zambia. Living in an impacted sentinel site or in a resettled household was associated with better health outcomes for P. falciparum infection, anaemia and stunting in under-five-year-old children. Improved health outcomes were reported in association with distal factors such as employment or relative household wealth, suggesting that the project development may result in positive effects on the health status of children.The most noticeable change observed was the higher prevalence of the P. falciparum infection rate in 2015 compared to 2011 in all sentinel sites. Nkenyawuli, the only comparison site sampled in both 2011 and 2015, showed a markedly higher prevalence in 2015 compared to the impacted sites. Malaria control interventions have been implemented by the project and district health management teams in the impacted sentinel sites, including indoor residual spraying (IRS), distribution of long-lasting insecticidal nets (LLINs), education and awareness and ‘malaria seek and treat’ (i.e., active case detection and treatment performed through house-to-house visits at weekly intervals) [37]. These interventions were generally associated with lower odds for P. falciparum infection. Children in resettled households showed significantly lower P. falciparum infection rates in 2015. In the newly built settlements of Northern Resettlement and Shenengene, prevalences were lowest at 10.9% and 6.3% in 2015, respectively, with the new, solid housing structures having closed eaves and window screens that are associated with lower infection risk as shown before in other malaria-endemic settings [38]. When excluding resettled or migrant households, no other factor was found a determinant for P. falciparum infection (see supplementary Figure S2). Nevertheless, across the entire study area, the 2015 follow-up showed higher P. falciparum infection prevalence compared to the 2011 baseline. This observation is in line with a wider trend in Northwestern Province found during two consecutive Malaria Indicator Surveys (MIS). Indeed, the prevalence in under-five-year-old children, as assessed by RDT, was 17.3% in 2010, while it was almost double in 2012 (32.5%) [23,39]. The strong increase coupled with the absence of significant associations with common risk factors at household and community level point to an environmental influence. As both surveys were conducted in July, we speculate that there were considerable inter-annual fluctuations, such as changes in the average temperature or precipitation [40,41].The stunting rate in children is influenced by a multitude of factors such as recurrent infectious diseases (e.g., hookworm infection), persistent enteropathy, access to improved sanitation and safe drinking water, access to food or children migrating from areas with different rates of stunting [42,43]. Overall, the stunting rates in 2015 in the impacted (39.4%) and comparison sites (47.0%) were similar or higher than the average of the Northwestern Province (36.9%), as determined during the 2013/14 Demographic and Health Survey (DHS) [29]. The improvement of stunting between 2011 and 2015 was significant in the impacted sites but not in the comparison sites. Of all the determinants assessed during the 2015 follow-up, wealth and access to improved sanitation were associated with lower stunting rates. Wealth remained a determining factor when excluding resettled or migrant households as well as households with safe sanitation (see supplementary Figure S2). Access to improved sanitation and reduced environmental contamination has been found previously to avert stunting in children [44]. Among the sentinel sites visited in both surveys, Northern Resettlement, where new houses were built with adjoining ventilated improved latrines, had consequently the highest proportion of households with access to safe sanitation in 2015 (97.1%; Table 2) and at the same time the lowest stunting rate.Anaemia rates in impacted (41.0%) and comparison sites (49.4%) in 2015 were comparable to data obtained during the 2012 MIS for the Northwestern Province, where anaemia was reported in 45.5% in under-five-year-old children [23]. These high rates of anaemia will continue to have long-lasting negative consequences in the study area given that iron deficiency undermines growth, physical fitness and educational performance [30]. Malaria and stunting remained significant determinants for anaemia in multivariate regression models where resettled and migrant households were excluded, respectively (see supplementary Figure S2). This high anaemia rate is a concern, particularly if one considers that health facilities were present in 11 of the 14 sentinel sites (Figure 1) and that health facilities could be most efficient in combating anaemia through the provision of primary health care services, including antimalarial drugs, iron supplementations and growth monitoring [30,45].To our knowledge, no survey data on soil-transmitted helminths for Solwezi district are publicly available. A recent geostatistical analysis by Karagiannis-Voules et al. (2015) estimated the prevalence of soil-transmitted helminth infections at 50% or higher in the general population in that area, which is in line with our findings (50% in the impacted and 60.9% in the comparison sites, respectively) [36]. Hookworm was the predominant soil-transmitted helminth species in both surveys, with similar prevalences in 2011 and 2015. Most children (94.3%) had mild-to-moderate infection intensities (i.e., <4000 EPG; data not shown) and hookworm infections are therefore expected to play an immaterial role in anaemia burden in the current setting [46]. According to the Solwezi District Health Management team, preventive chemotherapy using albendazole was done seven months prior to each survey—in December 2011 and December 2014. However, breaking transmission of hookworm will remain difficult when children continue to walk barefoot, and hence, are in contact with hookworm egg-contaminated soil in this setting [47].Migrant populations can be especially vulnerable to ill-health as they face restricted social cohesion and exclusiveness leading to inequalities [48]. However, in the current setting, children with a recent migrant history were generally found in better health than those from host communities. This can be partly explained by the fact that the migrants in this area were labour- or opportunity-seekers as opposed to involuntarily displaced people. For example, migrant children had significantly lower P. falciparum infection than children who were born and lived in the study area all along. Interestingly, P. falciparum infection prevalence differed greatly between Kalumbila Town (43.2%) and Chisasa (10.4%), the two settings with the highest proportions of migrant households (100% and 65.8%, respectively). While in Chisasa most migrant children came from within Solwezi district (40.9%) or other places in the Northwestern Province (28.8%), most migrant children in Kalumbila Town stem from the Copperbelt Province (43.2%) or Lusaka (8.1%), which are low prevalence areas [23]. For anaemia, however, rates were higher in 2015 compared to 2011 in Chisasa only and remained stable in Kankhozi and Musele, the three sentinel sites with higher proportions of migrants. Potentially new infectious diseases or sudden changes in lifestyle (e.g., feeding habits) coupled with a limited awareness of, and capacity to address, anaemia within the household could explain the slightly higher rates in the migrant population.The noted differences between migrant and host population of children illustrate the importance of understanding the characteristics of migrant populations (e.g., origin, level of skills, health status, economic means and reasons for migration) and their interplay with the local communities. Despite this, they remain often neglected in HIA, especially when planning public health interventions [49].The lack of baseline health data is an inherent limitation for monitoring of health in communities subjected to natural resource development and management projects in low- and middle-income countries [50,51]. For the Trident project, the BHS completed in the frame of the HIA provided a strong evidence-base that reflected the health status of communities prior to project development. Supported by this evidence-base, the HIA identified a wide range of health conditions that warranted management and monitoring throughout the project lifecycle. A priority was given to the control of STIs, including HIV, based on the perceived significant impact, whereas the outcomes of mitigation activities are publicly shared elsewhere [52]. In the absence of a regulation that requires transparent dissemination of HIA outcomes, presenting the findings in the peer-reviewed literature provides an opportunity to adhere to good practice standards such as transparency and the ethical use of evidence, while at the same time producing valuable case studies of HIA practice in the context of natural resource development projects in low- and middle-income countries [53,54]There were no data in 2011 for several sentinel sites that were only added in the 2015 follow-up, which obviously restricts “before–after” comparison. However, the five comparison sites surveyed in 2015 should represent a sufficiently large comparison group for future follow-up surveys. The non-random sentinel site sampling strategy allowed for inclusion of sites considered too important to miss but the resulting non-randomised sample and the results are therefore relevant to the selected sentinel sites only. Due to lower sensitivity of a single compared to duplicate Kato-Katz thick smears, the true hookworm prevalence is likely to be higher than presented here [55]. Furthermore, household characteristics and behavioural aspects (e.g., toilet use at school and footwear) were not determined in children participating in the school survey.Children living in villages considered impacted by a copper mine development in Northwestern Province of Zambia showed generally better health outcomes for P. falciparum infection, anaemia and stunting than children from comparison sites, whereas project-induced changes such as resettlement and employment had a positive influence. These findings though do not infer causality. Through the application of the HIA, health-targets were integrated in a project development that has primarily economic goals, which is in line with the health-in-all sectors approach embraced by the Sustainable Development Goals (SDGs) agenda [56]. Repeated cross-sectional monitoring of key health indicators and determinants of health in communities impacted by projects help to better understand whether and how human health is impacted, which population sub-groups are most vulnerable and help identify underlying risk factors. In collaboration with staff from the local health system, evidence from periodic and longitudinal monitoring generated in the private sector allow for prioritization and adaption of targeted and locally sensitive interventions whereby the public and private sectors share responsibility and synergize efforts in safeguarding human health.The following are available online at www.mdpi.com/1660-4601/14/3/315/s1, supplementary file S1: The questionnaire focused on KAP related to issues such as health seeking behaviour, maternal and child health, infectious diseases and participation in health interventions, supplementary file S2: Wealth remained a determining factor when excluding resettled or migrant households as well as households with safe sanitation.This work was supported by First Quantum Minerals Limited. We thank the Solwezi District Health Management team for the constructive collaboration. In addition, we acknowledge the traditional authorities and communities in the study area for their support, engagement and participation. Particular thanks go to the Trident project health promotion team for their support prior, during and after the surveys. For complementary social and environmental surveillance data, we extend our thanks to Garth Lappeman and Mulenga Musapa from First Quantum Minerals Limited. For statistical support, we are thankful to Christian Schindler and Jan Hattendorf at the Swiss Tropical and Public Health Institute.Astrid M. Knoblauch, Mark J. Divall, Milka Owuor, Colleen Archer and Mirko S. Winkler conceived the study design. Astrid M. Knoblauch, Mark J. Divall, Milka Owuor, Colleen Archer, Kennedy Nduna and Mirko S. Winkler coordinated the fieldwork. Colleen Archer led the laboratory work. Mark J. Divall, Harrison Ng’uni, Gertrude Musunka and Anna Pascall were the overall study coordinators. Astrid M. Knoblauch and Milka Owuor performed the statistical analysis. Astrid M. Knoblauch and Mirko S. Winkler wrote the first draft of the manuscript. Astrid M. Knoblauch, Mark J. Divall, Milka Owuor, Jürg Utzinger and Mirko S. Winkler contributed to the draft development. All authors read and approved the final version of the manuscript for submission. Astrid M. Knoblauch and Mirko S. Winkler are guarantors of the paper.First Quantum Minerals Limited funded the health impact assessment and supported data collection for the baseline (2011) and follow-up health surveys (2015). Astrid M. Knoblauch, Mark J. Divall, Milka Owuor, Colleen Archer and Mirko S. Winkler have supported the Trident project as independent public and occupational health specialists. Gertrude Musunka and Anna Pascall are currently employed by First Quantum Minerals Limited. The corresponding author had full access to all the data in both surveys. The founding sponsor had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, and in the decision to publish the results.Study area and sentinel sites, Trident project, 2011 and 2015, Zambia.Sentinel site selection, Trident project, 2011 and 2015, Zambia.Prevalence rates per sentinel site, Trident project, 2011 and 2015, Zambia: (a) prevalence of P. falciparum in children aged 6–59 months; (b) prevalence of stunting in children aged 0-59 months; (c) prevalence of anaemia in children aged 6–59 months; and (d) prevalence of hookworm in children aged 9–14 years.Determinants of health outcomes during the 2015 follow-up health survey, with adjusted odds ratios and 95% confidence intervals, Trident project, Zambia: (a) determinants of P. falciparum in children aged 6–59 months; (b) determinants of stunting in children aged 0–59 months; and (c) determinants of anaemia in children aged 6–59 months.Selected indicators in children and their relevance in the Trident copper mining project area, Zambia.Study populations, Trident project, 2011 and 2015, Zambia.1 Sentinel site with data for 2011 BHS and 2015 follow-up; NA: not available.Prevalences and period effects for P. falciparum infection and stunting, Trident project, 2011 and 2015, Zambia.1 Describes the change in prevalence between 2011 and 2015; CI: confidence interval; n: sample size; n/a: not applicable; OR, odds ratio.Prevalences and period effects for anaemia and hookworm, Trident project, 2011 and 2015, Zambia.CI, confidence interval; n: sample size; n/a: not applicable; OR, odds ratio; 1 Describes the change in prevalence between 2011 and 2015.

Med-MDPI/ijerph_2/ijerph-14-03-00316.txt

@@ -0,0 +1 @@
+Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).Background: Research has shown that suicide is a phenomenon highly present among the drug dependent population. Different studies have demonstrated an upraised level of comorbidity between personality disorders (PD) and substance use disorders (SUD). This study aimed to describe which PDs are more frequent among those patients with a risk of suicide. Methods: The study was based on a consecutive non-probabilistic convenience sample of 196 bereaved patients attended to in a Public Addiction Center in Girona (Spain). Sociodemographic data, as well as suicide and drug related characteristics were recorded. The risk of suicide was assessed with the Spanish version of “Risk of suicide”. Personality disorders were measured with the Spanish version of Millon Multiaxial Clinical Inventory. Results: The PDs more associated with the presence of risk of suicide were depressive, avoidant, schizotypal and borderline disorders. However, the histrionic, narcissistic and compulsive PDs are inversely associated with risk of suicide even though the narcissistic scale had no statistical correlation. Conclusions: The risk of suicide is a significant factor to take into account related to patients with SUD and especially with the presence of specific PDs. These findings underline the importance of diagnosing and treating rigorously patients with SUD.Approximately 800,000 people die due to suicide around the world every year [1,2]. Suicidal risk is a very complex behavior that is influenced by interacting biological, genetic, psychological, social, environmental and situational factors, as several authors have described [3,4].The link between the risk of suicide and substance use disorders (SUD) is well documented [5,6,7,8]. Li and collaborators [9] showed that the risk of suicide was 7.5 times higher in males and 11.7 times higher in females with a mental or SUD compared to males and females with no disorder. In another recent study among the SUD population, Masferrer et al. [10] found that 61.2% of 196 bereaved SUD patients reported a risk of suicide in a large study focused on describing related variables of risk of suicide among bereaved addicted patients. Taking into account the connection between the risk of suicide and SUD, our interest is to analyze two important variables, personality disorder and risk of suicide, because it could have an important influence on the particular and complex association between these constructs. Our interest is to study the risk of suicide as associated with the comorbid dysfunctional patterns of personality in SUD patients. Personality Disorders (PDs) are defined as inflexible and maladaptive personality traits that are exhibited in a wide range of personal and interpersonal contexts [11]. The Diagnostic and Statistical Manual of Mental Disorders 5 [11], in section II, defines PDs as categorical entities. However, based on the discussion of early research [12,13,14,15,16], PDs have been characterized as dimensional constructs related to a framework that provides a unified model of psychopathology established on shared personality traits. In fact, Diagnostic and Statistical Manual of Mental Disorders DSM-5, in section III, proposes a dimensional alternative model. In the framework of this dimensional approach of psychopathology, Millon’s integrative model of personality disorders [17,18] proposes an explanation of the structure of personality styles on the background of ecological adaptation. Millon classifies personality disorders in accordance with four main dimensions: Personalities with difficulties in taking pleasure (i.e., with schizoid, avoidant or depressive disorders), personalities with interpersonal problems (with dependent, histrionic, narcissistic or antisocial disorders), personalities with intrapsychic conflicts (with sadistic, compulsive, negativistic or masochistic disorders) and personalities with structural deficits (with schizotypal, borderline or paranoid disorders). The latter three pathological personality patterns (schizotypal, borderline and paranoid) represent, in terms of Millon’s theory, more advanced stages of personality pathology and structural impairment. PDs can seriously influence the course, prognosis and the treatment outcomes of SUDs [19]. Reporting a diagnosis of PD is linked with greater impairments as well as a lower quality of life [20,21]. The presence of PD is a greatly prevalent comorbid disorder among substance users [22,23,24,25]. In fact, different studies have demonstrated a high level of comorbidity between PD and SUD [26,27,28,29,30,31]. As Krueger and Eaton [32] stated, comorbidity is the rule not the exception. In this regard, Gonzalez [33] found a prevalence of any personality disorder of 42% among a sample of 53 alcohol and drug dependent inpatients. Colpaert and collaborators [27] reported a rate of 42.6% for at least one PD among 274 patients admitted to a residential substance abuse treatment. Casadio et al. [26] described a rate of 62.2% among addiction outpatients. Moreover, Verheul [19] concluded that rates of PDs among the drug dependent population are four times higher than among the general population. Bearing in mind the negative impact of PDs and the potential risk of suicide, this study aimed to describe which dysfunctional patterns of personality are more frequent among those SUD patients with a risk of suicide and which dysfunctional patterns of personality are more frequent among those without a risk of suicide.The current study is part of wider research. The main goal of this research was to describe the complicated grief symptomatology among a sample of 196 bereaved SUD patients. For more information, see Masferrer et al. [10].The current research was based on a consecutive non-probabilistic convenience sample of individuals (n = 196) attended the Public Addiction Treatment Centre in Girona (Catalonia, Spain). To join the study, patients had to meet the following three inclusion criteria: (a) they had a diagnosis of substance use disorder (SUD) (alcohol, cocaine or heroin dependence) according to the 4th revised edition of Diagnostic and Statistical Manual of Mental Disorders (DSM-IV-TR) criteria; (b) loss of a significant person (family, best friend or partner) at some time in their life, but at least a year previously to the interview; and (c) abstinence during the last month to avoid any toxic effects of drugs. The majority of patients (78.1%) were male, more than a third (37.2%) were married or with a partner. Related to the main drug diagnosis, the majority of the patients (68.9%) reported alcohol dependence, 18.4% heroin dependence and 12.8% cocaine dependence. For the assessment of the risk of suicide, we used the Spanish version of the Risk of Suicide (RS) from Plutchick et al. [34]. The RS could discriminate between individuals and patients with no suicide attempts and those having a history of them. It consists of 15 items with dichotomous responses (yes/no). The RS embraces issues about previous attempts, ideation intensity of current feelings of depression and hopelessness, and other aspects of the attempts. The total score is obtained by summing all items (maximum score 15). The cut-off suggested by the authors of the Spanish version [35] was 6. Internal consistency of the test is 0.90. Personality disorders were measured by the Millon Multiaxial Clinical Inventory [36], the Spanish translation by Cardenal and Sánchez-López [37]. The MCMI-III consists of 175 items with dichotomous answers (true/false), a self-report questionnaire that measures 11 clinical personality patterns, 3 traits of severe personality pathology, 7 syndromes of moderate severity, 3 severe syndromes and a validity scale and 3 modifying indices. The PDs scales cover major diagnostic criteria of DSM-IV-TR. We adopted the most conservative criteria with scores equal or greater than 85 to determine the presence of the PDs. Clinical personality patterns (schizoid, avoidant, depressive, dependent, histrionic, narcissistic, antisocial, sadistic, compulsive, negativistic and masochistic) and 3 traits of severe personality pathology (schizotypal, borderline, and paranoid) were used in the current research.Those patients who met the three inclusion criteria were informed by their therapist about potential participation in the study. The research procedure consisted of a single visit with a psychologist who administered the questionnaires included in the study protocol. All patients were previously informed about the study procedure as well as terms of confidentiality. Informed consent was obtained from all participants and the protocol was approved by the Institutional Ethics and Research Review Board of the Institut Assistència Sanitària (IAS) (No. S041-779). The risk of suicide was measured as the relative frequency and the 95% level of confidence of participants above the RS cut-off point. In order to compare individuals with and without risk of suicide, we performed a bivariate analysis of the PD scores of the patients according to the risk of suicide. According to the nonparametric Kolmogorov Sminov test, the different Millon’s scales did not follow a normal distribution and the significance was below 0.05 (except for the negativist scale which is the only scale with normal distribution with a signification of 0.072). Due to the small group of PDs, we used a non-parametric statistical test. When PDs were defined as categorical variables, we used a Fischer Exact test and when PDs were defined as a dimensional variable, we used U Mann Whitney and Spearman’s correlation. A multiple regression analysis was performed to determine which dysfunctional patterns of personality were associated with the risk of suicide. The results are expressed as absolute numbers, percentages, as well as the mean and standard deviations. A statistical significance of 0.05 was used to compare hypotheses. Data processing and analysis were performed using the SPSS statistical program version 21.0 for Windows (IBM Corp., Armonk, NY, USA).Taking into account that we adopted the most conservative criteria of Millon’s scoring for describing the presence of PD (scores equal or greater than 85), the presence of any PD among the sample was 29.4%. Describing the occurrence according to each PD, those PD with higher frequency were compulsive (7.1%) and narcissistic (7.1%), followed by antisocial (4.6%) and sadistic (3.1%). Twenty-four percent of the patients reported a presence of one PD and only 1.5% two PD. On the other hand, avoidant, dependent and masochistic were not present as a disorder. The first objective of the study was determine which PDs (scoring equal or greater than 85) are more frequent among those patients with a risk of suicide and which PDs are more frequent among those without a risk of suicide. Bearing in mind that the number of PDs was small, we performed a non-parametric statistical test. The results of relationship between PDs as categorical variables and the risk of suicide are set out in Table 1, in which the Fischer Exact test was carried out. What stands out in the first table is that, in the risk of suicide group, there is a higher presence of schizoid, depressive, narcissistic, antisocial, sadistic, schizotypal, borderline and paranoid cases, although the differences between the groups are not statistically significant in any case. The histrionic and compulsive disorders are more present, in a significant way, in the no risk of suicide group. If we compare the direct scores of the different scales of PDs, significant differences are shown in the scores of all scales except in narcissistic through the U Mann Whitney analysis (Table 2). Furthermore, those patients grouped in the risk of suicide presented a higher mean than those without risk, not including histrionic, narcissistic, compulsive and borderline.In addition, the relationship between the risk of suicide and symptomatology of PDs was investigated using Spearman’s correlation coefficient (Table 3). The different scores of PDs showed a significant association with the risk of suicide, with the exception of the narcissistic. The three higher and stronger correlations were borderline (r = 0.703), depressive (r = 0.628) and masochistic (r = 0.529). Otherwise, results indicated an inverse correlation between the risk of suicide and histrionic, narcissistic and compulsive scales, even though the narcissistic scale did not have any statistical correlation.Another main objective of the study was to determine which PDs were associated with the risk of suicide. Thus, in order to describe this, a multiple regression, in which scores of the risk of suicide, as dimensional variables, were the dependent variable, was performed while sociodemographic, suicide-related characteristics (age, gender, education, marital status and patient’s suicide attempt) and dysfunctional patterns of personality were considered independent variables. The results were presented in Table 4. Those dysfunctional patterns of personality defined also as dimensional variables associated with the risk of suicide were avoidant, depressive, schizotypal and borderline.Almost one third of our sample (29.4%) reported some PDs. Therefore, PDs are quite frequent among the current SUD sample, which is in agreement with those results obtained by previous studies [23,24,25]. Comorbid PD and SUD represent a robust determinant of elevated suicide risk [38]. However, the scales in which there are any cases with a score above 85 were avoidant, dependent and masochistic.The primary goal of this study was to determine which dysfunctional patterns of personality are more frequent among those patients with a risk of suicide and which PDs are more frequent among those without a risk of suicide. There are very few patients who reported high scorings of PD in the sample. When we have dichotomized PD variables in “presence of PD” or “absence of PD”, it can be seen that there are very few cases. The differences between the group of “risk of suicide” and “no risk of suicide” are not significant in the different PD, except for histrionic, sadistic and compulsive. However, it should be noted that more PDs were associated with the risk of suicide than without the risk of suicide. Histrionic and compulsive are more frequent in the no risk of suicide group, and sadistic in the risk of suicide group. For this reason and following the theoretical approach of early research [13,15,16], we wanted to analyze more thoroughly the different PD scales in a dimensional way. Consistent with our expectations, reporting a high scoring in PDs’ scales is linked with the risk of suicide, except in the cases of the histrionic, narcissistic and compulsive scales, although histrionic and compulsive are the only scales in which the differences are statistically significant. Specifically, when analyzing how PDs scoring and the risk of suicide perform, borderline, depressive and masochistic are the three scales with a higher association with risk of suicide. These results confirm the important role of PDs as risk factors for suicide as other studies suggested [39,40].An important finding was that the narcissistic, compulsive and histrionic scales had an inverse correlation with the risk of suicide. A potential explanation for these findings might be that stating personality traits of these scales could be defined as protective factors against suicidal risk. Nowadays, some personality characteristics from those PDs are greatly accepted and promoted in Western society [41]. These results lead us to reflect on previous investigations carried out with the MCMI [37], in which a curved model of the narcissistic, histrionic and compulsive scales is considered, meaning that it is the low and the high scores that indicate non-adaptation, whereas intermediate levels on these scales would reflect adaptive patterns, unlike what happens in relation to other scales [41]. Turning next to the narcissistic scale, it was the only PD with no present relationship with the risk of suicide. This finding indicates that the narcissistic PD appears to be a distinct group among cluster B personality disorders related to suicidal risk but this outcome is contrary to that of Pompili and his collaborators [3], who found that individuals with cluster B personality disorders have a greater risk of dying by suicide. At this point, it should be taken into account that presenting specific personality traits does not necessarily entail negative consequences in relation to the presence of mental health problems but reporting many personality traits could be associated to a general dysfunctional pattern of personality, so could be linked to risk of suicide. Therefore, it is important to be aware of differences between dimensional analysis (direct scoring) and categorical analysis (presence–absence of PD) in order to describe the specific characteristics that may be more protective for those who make up the PD. As some authors stated [42], each case has a profile that emerges from quantitative variations and different levels ranging from normal to pathology without the need to cut-off that could be artificial.In order to identify which PDs were linked to the risk of suicide, a multiple regression analysis was conducted. Talking about the sociodemographic variables related to the risk of suicide, marital status was the only relevant characteristic. Being separated or divorced and widowed was statistically significant with the risk of suicide. These results are in accord with previous studies [43]. The current study also found that reporting a previous suicide attempt was associated with the risk of suicide, which was consistent with several examples of previous research [44]. In fact, reporting a previous suicide attempt and being separated or being widow were the variables with a major contribution to risk of suicide according to the Table 4.The regression analysis revealed that the presence of avoidant, borderline, schizotypal and depressive are all associated with the risk of suicide. These results are broadly consistent with previous research [3,45,46,47,48]. A diagnosis of borderline PD doubled the risk of suicide when compared to patients diagnosed with other types of PD [38]. Links and his collaborators [47] found that 25.6% of participants with borderline PD attempted suicide during the course of one year of treatment. Moreover, 60% to 70% of patients with borderline PD reported a history of suicidal behavior [49]. These relationships may partly be explained by the role of impulsivity as a key background factor [3]. PDs are relevant factors to take into account related to the risk of suicide among SUD patients. As a practical implication of the present findings, the results indicate that the identification of comorbidity of SUD is important for improving the treatment among the bereaved drug-dependent population as well as reducing suicidal ideation because, as Schneider et al. noted [40], treating PDs is essential for suicide prevention.The current research presented some limitations that should be considered. It is important to mention that this study was performed using a convenience sample of bereaved substance users, who attended a drug addiction treatment center. Therefore, our sample might be different from the general drug user population. Furthermore, the current research had a cross-sectional design and we relied on self-reporting measures. Thus, we must be cautious due to the small number of PD cases according with the most conservative scoring of Millon [36]. Notwithstanding these limitations, this study provided significant data related to the specificity of the sample. Clarifying the pattern of risk across mental disorders is a necessary step to identify where resources can be most effectively targeted and interventions prioritized [50]. To date, there are no studies that have investigated the association between risk of suicide and dysfunctional patterns of personality among a bereaved SUD sample. As predicted, reporting PDs are linked with the risk of suicide, with the exception of the narcissistic scale. The presence of avoidant, depressive, schizotypal and borderline personality disorders are associated with the risk of suicide. In conclusion, these findings outline the importance of performing therapeutic interventions in order to focus on PD in those bereaved SUD patients and to reduce and prevent the risk of suicide. The authors appreciate the valuable contributions of Garre-Olmo. We thank the support of Arnau Gavaldà and Laia Figueras. We also appreciate the professionals from CAS Teresa Ferrer (IAS) for their help in the collection of data and to the patients for their participation. Laura Masferrer and Beatriz Caparrós conceived and designed the research. Laura Masferrer collected field data and write the manuscript. Beatriz Caparrós contributed to analysis tools and provided critical review. Both authors performed the statistical analysis, interpreted the results and approved the final manuscript.The authors declare no conflict of interest. Relationship between personality disorders (PDs) and risk of suicide.Mann–Whitney U test of PD symptoms related to presence of risk of suicide (M (SD)).Relationship between symptomatology of PD and risk of suicide through the Spearman’ correlation.* Correlation is significant in the level 0.01.Multiple regression analysis of dysfunctional patterns of personality associated with risk of suicide.

Med-MDPI/ijerph_2/ijerph-14-03-00317.txt

@@ -0,0 +1 @@
+Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).Substance use disorders (SUD) and mental health disorders are significant public health issues that co-occur and are associated with high risk for suicide attempts. SUD and mental health disorders are more prevalent among offenders (i.e., prisoners or inmates) than the non-imprisoned population, raising concerns about the risk of self-harm. This cross-sectional study examined the population of a state prison system (10,988 out of 13,079) to identify associations among SUD (alcohol, cannabis, intravenous drugs, narcotics, and tobacco smoking), mental health disorders (anxiety, bipolar, depression, and psychotic disorders), and suicide attempts. The primary aim was to determine which groups (SUD, mental health disorders, and co-occurrences) were strongly association with suicide attempts. Groups with a documented SUD or mental health disorders compared to peers without these issues had 2.0 and 9.2 greater odds, respectively, for attempting suicide, which was significant at p < 0.0001 for both conditions. There were also significant differences within SUD and mental health disorders groups in regard to suicide attempts. Groups with the greatest odds for suicide attempts were offenders with comorbid bipolar comorbid and anxiety, alcohol combined with depression, and cannabis co-occurring with depression. Documentation of suicide attempts during imprisonment indicates awareness, but also suggest a need to continue enhancing screening and evaluating environmental settings.Substance use disorders (SUD) and mental health disorders are significant public health issues that often co-occur and are associated with functional disability [1] and high risk for self-harm, including suicide attempts and completion [2,3,4,5,6,7,8]. The prevalence for SUD and mental health disorders are much greater for the offender population (i.e., prisoners, inmates) than non-imprisoned populations [9,10,11]. Individuals with SUD have a significantly higher probability of being arrested compared to the population without these issues [12,13]. A majority of offenders in the United States (56%) reported that they used at least one substance prior to arrest [14]. Cannabis use was reported by 40% of offenders, followed by 21.4% for cocaine or crack, 12.2% for methamphetamine and amphetamine, and 8.2% for heroin/opioids [14], compared to 3.0% of non-imprisoned adults who indicated taking any substance. Furthermore, offenders who used cannabis, cocaine, methamphetamine, or heroin reported that they used these substances regularly [14].The risk for criminal justice involvement increases when SUD and mental health disorder co-occur [13]. The United States criminal justice system has filled the void created by insufficient mental health services and has become a provider of last resort for some individuals with mental health disorders [15]. A large percent of offenders (24%) in state prison systems reported having recent mental health problems and 49% had symptoms for mental health disorders [16]. The Bureau of Justice Statistics found that 43.2% of the offender population had symptoms for mania disorder, followed by major depressive (23.5%) and psychotic disorder (15.4%) [16], which compares to non-imprisoned population estimates of 0.4%–2.1% for mania (one study reported 6.4% as an upper estimate), 6.0%–7.9% for major depression, and 3.1% for psychoses [3,16,17,18,19]. A large percent of offenders (41.7%) also have symptoms or diagnosed mental health disorders that co-occur with a SUD (e.g., alcohol, cannabis, cocaine, heroin, tobacco smoking) [16].There is a strong relationship between mental health disorders and self-harm (e.g., suicide ideation, attempts, and other self-injurious behaviors) [20,21,22,23,24,25,26,27,28]. In particular, depression and bipolar with anxiety disorders are strongly associated with suicidal ideation, attempts, and completion [3,20,21,22,23,24,26,29,30,31]. Individuals with SUD also have greater incidence of suicidal ideation and attempts compared to the population who do not have a problem with substances [32,33]. Co-occurrence of SUD and mental health disorders increases the risk of suicidal ideation, attempts, and completion [2,3,4,5,6,7,27,28,33]. Cannabis, alcohol, and nicotine (tobacco smoking) use disorders that co-occur with mental health disorders significantly increase the risk of suicides [4,5,29,34,35,36].Offenders greatly exceed the non-imprisoned population in regard to mental health disorders and SUD, particularly cannabis. In addition to a greater prevalence of risk factors, such as SUD and mental health disorders, suicide attempts among offenders exceed the percent for non-imprisoned populations [37,38]. Estimates of suicide attempts among offenders are 2.3% compared to 0.4% among non-imprisoned populations [38]. Among offenders who have depressive or manic symptoms, 13% reported attempting suicide [16]. Despite offenders having greater risk factors for suicide attempts (e.g., SUD and mental health disorders) compared to non-imprisoned populations, there are few investigations that have included this group. This descriptive cross-sectional study of a state department of corrections investigated the association among SUD, mental health disorders, and suicide attempts.The primary aim of this study was to examine risk factors related to suicide attempts in an offender population. We hypothesized (1) there will be sociodemographic differences (gender, race, education, and security level) in regard to suicide attempts, (2) group proportions for suicide attempts will be greater for the population with co-occurring SUD and mental health disorders compared to offenders with either a SUD or a mental illness, and (3) the population with the conditions of bipolar and anxiety disorders and co-occurring cannabis use disorder will have the greatest odds of suicide attempts compared to other SUD that co-occur with other mental health disorders.This cross-sectional descriptive study was approved by an institutional review (IRB) at an academic health center and was conducted in collaboration with a department of corrections (DOC) in the east south central region of the United States. The IRB protocol number is 10-0382F2L.All offenders (men and women) imprisoned between 1 June 2005 and 31 December 2010 and had a record in the electronic health and offender management systems were included. Inclusion criteria also included all race and ethnic groups, and offenders who had date of birth, which was used to calculate age, all of which may be explanatory factors for suicide attempts. This DOC had thirteen facilities, of which two were women only and eleven men only. Facilities were dispersed geographically throughout the state, representing the geographic distribution of the state.The data sources for this study were an electronic health record (EHR) and an offender management system (OMS), which the DOC we collaborated with used to manage offender health and to track and monitor their activities. All offenders receive a complete physical, mental, behavioral health, and dental examination upon arrival at the DOC reception center, also referred to as the intake facility. Health findings are recorded in the EHR by clinic staff (e.g., physicians, advanced registered nurse practitioners, nurses, and medical records specialists). The EHR is comprised of structured data (i.e., clinic staff entered data using standardized data dictionaries). Clinic staff entered diagnoses using the International Statistical Classification of Diseases (ICD-9) and the Systematized Nomenclature of Medicine (SNOMED) at the reception center and throughout offenders’ imprisonment as health conditions are identified and change. Documentation with ICD-9 provides a consistent and internationally recognized way to refer to diseases and disorders. This study extracted all health data (i.e., SUD, diagnoses for mental health disorders, and suicide attempts) from the EHR (See Table 1). Documentation of SUD were made on the basis of screening tools and limitedly on health records prior to imprisonment that offenders provided. Mental health disorders were diagnosed after extensive evaluation by psychologists and psychiatrists. Diagnoses for panic, phobia, and posttraumatic stress syndrome were included with anxiety disorders, as other investigators have grouped these conditions together [19,24,31,39]. Clinic staff documented suicide attempts during imprisonment based on physical evidence and investigations conducted by medical, mental health staff, and correctional officers. The OMS was created electronically from court records and managed by correctional staff (e.g., officers and case managers) throughout offenders’ imprisonment. This study extracted sociodemographic, criminal offense, security level, sentence, earliest potential parole, and release dates from the OMS.We used SAS® 9.4 (SAS Institute, Cary, NC, USA) to conduct all statistical tests. Frequencies and percentages were conducted for race, gender, education, security level, SUD, and mental health disorders (i.e., nominal and categorical variables). Race comparisons were made between African Americans and Whites; there were too few observations for other racial and ethnic groups for meaningful analyses. Means and standard deviations (SD) were calculated for age and date of diagnosis (continuous variables). Differences in population proportions and odds (i.e., offenders with SUD and mental health disorders who attempted suicide) were evaluated using chi-square (χ2). We created mutually exclusive groups for SUD (alcohol, cannabis, intravenous drugs, narcotics, and tobacco smoking) and mental health disorders (anxiety, bipolar, depression or depressive symptoms, and psychotic disorders) to minimize potential statistical noise that polysubstance and comorbid conditions may have on odds for suicide attempts. We created a group for co-occurrence of bipolar and anxiety disorders, since these conditions are highly comorbid and are associated with increased risk for suicide attempts [2,23,27,39,40,41]. Comparisons for suicide attempts in regard to age and duration of imprisonment were made using the Wilcoxon rank sum test. When examining multiple groups, we performed the Fisher exact test and made p-value adjustments (including Bonferroni) to evaluate significance of pairwise comparisons. The p-value adjustment identified which comparisons were significant and decreased the probability of Type I error (false positives).There were 10,988 out of 13,079 offenders (84%) who had valid records in the EHR and OMS (i.e., inclusion criteria for the study). The population was majority White and men (See Table 2) with a mean age of 37.5 (SD = 11.5), 95% CI (37.3, 37.7). The mean age for offenders with documented suicide attempts was 36.6 (SD = 9.6), 95% CI (35.6, 37.6) compared to 37.5 (11.6), 95% CI (37.3, 37.7) who did not attempt suicide. A plurality of the population had completed high school, had a primary offense related to property crimes (e.g., destruction of property, burglary, receiving stolen property), and were classified as medium security, which allows some freedom of movement within a facility, but under greater supervision and restriction than offenders whose status was minimum level (See Table 2). Education was self-reported and a large percent of offenders refused to report this information. We also did not have access to all criminal offense and security level information for all offenders, which resulted in approximately one-third to almost forty percent of these data being unavailable.Substance use disorders were prevalent, in which 6629 offenders (60.3%) had at least one documented issue and 2205 (20.1%) had more than one problem. Tobacco smoking was the most frequently used substance, followed by cannabis and alcohol (See Table 3). A large percent of the population (28.0%) had a diagnosed mental health disorder (i.e., anxiety, bipolar, depression, and psychotic disorders), and 3.7% of offenders had comorbid conditions. Depression was the most frequently diagnosed mental health disorder, and bipolar disorders the least (See Table 3). Suicide attempts occurred in 3.4% of the population (See Table 3).An overwhelming majority of the diagnoses for mental health disorders (92.8%) were made during imprisonment (See Table 3). The mean duration for diagnosing an offender with an anxiety diagnosis was 4.5 years (SD = 6.3), 95% CI (4.1, 4.9); 3.6 years (SD = 5.0), 95% CI (3.0, 4.1) for a bipolar disorder; 4.1 years (SD = 6.1), 95% CI (3.8, 4.4) for depression; and, 4.6 years (6.5), 95% CI (4.0, 5.2) for a psychotic disorder. Most suicide attempts also occurred during imprisonment (See Table 3). The mean number of years imprisoned prior to a suicide attempt was 5.2 years (6.6), 95% CI (4.4, 6.0).Whites, men, and the population with a SUD or mental health disorder had greater odds of suicide attempts compared to peers (See Table 4). A significantly larger percent of offenders classified as maximum security attempted suicide compared to groups with minimum or medium security levels (See Table 4). We conducted the Fisher’s exact test for analyses of variables with more than two groups (i.e., education and security level). A pairwise comparison among security levels indicated that all confinements significantly differed from one another after a p-value adjustment (Bonferroni). The odds for suicide attempts were significantly greater for the maximum level compared to minimum (p < 0.0001) and medium (p = 0.002), and a classification of medium was associated with an increased likelihood for knowingly engaging in behaviors that could be life ending compared to the minimum level (p < 0.0001). Although offenders with post-secondary education (some college, undergraduate, and graduate degrees) had a smaller percent for suicide attempts compared to groups with less educational attainment, the differences were not significant (See Table 4).We created mutually exclusive groups (i.e., offenders who only have one condition) for all SUD and mental health disorders. Some SUD and mental health disorders were associated with increased risk for suicide attempts. There were significant group differences between smokers and non-smokers, as well as the population with a history of intravenous drug use (IDU) compared to those who did not inject drugs. Offenders with a history of IDU had greater odds for suicide attempts compared to their non-IDU peers and smokers compared to non-smokers were more likely to have a documented case of suicidal behaviors, i.e., tries (See Table 5). Offenders with alcohol, cannabis, or narcotic use disorders were not significantly different from peers who did not have an issue with these substances. Offenders with bipolar disorders had significantly greater odds for suicide attempts compared to the population without this mental health problem (See Table 5). Comparisons between offenders with or without depression and the group with or without dual diagnoses of bipolar and anxiety disorders indicated that the population with these mental health disorders were associated with greater odds for suicide attempts (See Table 5). There were no significant differences for groups with or without anxiety or psychotic disorders (See Table 5).There were significant within group differences for suicide attempts in regard to SUD and mental health disorders (See Table 6). Offenders with a history of IDU or tobacco smoking had larger percentages for suicide attempts (See Table 6). A pairwise comparison, using Fisher’s exact test and adjusted p-values, indicated that SUD significantly differed from one another (See Table 7). Offenders who had issues with IDU or tobacco smoking significantly differed from the group that had no SUD history (See Table 7), i.e., a larger percent of individuals with IDU and tobacco smoking issues had instances of suicide attempts compared to the group with no SUD. Only alcohol and cannabis use disorders groups differed significantly from IDU and tobacco smoking (See Table 7). A smaller proportion of the groups with alcohol and cannabis use disorders attempted suicide compared to offenders with histories of IDU or tobacco smoking (See Table 7). However, there were no significant differences detected for suicide attempts when the group without a substance use disorder was compared to offenders who had issues with alcohol, cannabis, or narcotics (See Table 7).In regard to mental health disorders, offenders with bipolar and anxiety had the largest percent of population for suicide attempts (See Table 6). A pairwise comparison for mental health disorders indicated that a significantly smaller percent of the population without a diagnosis had attempted to end their life compared to groups with either anxiety, bipolar, depression, psychoses, or comorbidities (i.e., bipolar and anxiety), see Table 7. The group with bipolar and anxiety disorders had a significantly larger percent of population with suicide attempts compared to groups with anxiety only, depression, or psychotic disorders (Table 7).There were a few significant population differences while controlling for SUD. The population size for this study was not sufficient to conduct analyses for all mental health disorders while controlling for an SUD. The population with depression and the co-occurrence of alcohol, cannabis, or tobacco smoking had greater odds for suicide attempts compared to the groups that had these SUD, but had not been diagnosed as having a mental health disorder (See Table 8). Although alcohol and cannabis use disorders independently were not associated with suicide attempts, their co-occurrence with depression increased the odds of trying to end one’s life (See Table 8). Unlike alcohol and cannabis, tobacco smoking did not increase odds for attempting suicide (See Table 8).Substance use disorders and mental health disorders for this DOC are much more prevalent than the non-imprisoned population. National survey data for alcohol use disorder indicated that 3.5% of adults 18 or older in 2010 had this condition compared to 18.5% in this DOC [42]. Percentages for cannabis (17.4%), cocaine (7.8%), and heroin (0.5%) also exceeded 2010 national averages (1.0%, 0.3%, and 0.1%, respectively) [42]. Further, 35% of offenders in this DOC smoked tobacco compared to 19.3% in the non-imprisoned population [43]. In regard to mental health disorders, the prevalence of depression for this DOC was greater than the 2012 non-imprisoned population (17.1% compared to 7.9%) [19]. The population for this DOC also exceeded the 2012 non-imprisoned population percentages for bipolar (4.9% compared to 2.1%) [3] and psychotic disorders (6.2% compared to 3.1%) [19]. However, diagnoses for anxiety disorders in this DOC were slightly less than the 2005 non-imprisoned population, which included the same conditions as this study (16.4% compared to 18.1%) [44]. The percent of suicide attempts in our population of offenders were 3.4%, which compared to 0.4% for attempts in a national survey of a non-imprisoned population [45] and an estimate of 2.3% across prison systems in the United States [38].Whites compared to African Americans had significantly greater odds for suicide attempts [46,47], which is consistent with other investigations. Men had greater odds for attempting suicide compared to women, which was unexpected. Investigations with non-imprisoned individuals have found that men are less likely to attempt suicide [48,49], but have greater odds for completing suicide compared to women [49]. This finding is surprising, since women in this DOC were significantly more likely to have a mental health disorders, p < 0.0001, OR = 1.4, 95% CI [1.3, 1.6], which are known risk factors for suicide attempts. However, the lower odds for women attempting suicide may be explained by their greater likelihood for utilizing mental health services than men [50]. There were no significant difference (p = 0.12) in suicide attempts in regard to age, which did not coincide with investigations with non-offenders [20,27,29,41,51]. The population that smoked tobacco, as well as the group with bipolar, depression, or bipolar with anxiety disorder, had significantly greater odds for attempting suicide compared to offenders who did not smoke or have these mental health disorders, which investigations of non-imprisoned populations have found [8,33,52,53,54].Surprisingly, we did not find increased odds for suicide attempts for groups that solely had an alcohol [4,6], cannabis [7,32,34], or narcotic use disorder [5]. Further, the population that was solely diagnosed with either anxiety [24,31,55] or a psychotic disorder [29,56] did not have greater odds for suicide attempts compared to peers without these mental health disorders. Although alcohol and cannabis were not associated independently with suicide attempts their co-occurrence with depression resulted in significantly greater odds for trying to commit suicide. The odds for suicide attempts for tobacco smokers who had anxiety or bipolar disorder was not significant, which is not consistent with other investigations [35,36,57,58,59].Tobacco smokers who were diagnosed with depression did not have greater odds for suicide attempts compared to the population with depression only. While alcohol and cannabis use disorders co-occurring with depression were significantly associated with suicide attempts, we were unable to analyze the strength of the associations. However, neither cannabis use disorder nor anxiety independently resulted in greater odds for suicide attempts. Further, cannabis use disorder co-occurring with depression resulted in increased odds for suicide attempts compared to having only a mental health disorder. These findings suggest that past alcohol and cannabis use disorders may be important in regard to suicide attempts, but the mechanism that the substance may have is beyond the scope and aims of this investigation.The findings that maximum security was associated with increased odds for suicide attempts is not well defined, since there are few investigations that have included offenders. However, the increased odds for maximum security was not unexpected, since this security classification is the most restrictive level of imprisonment and is reserved for the most serious offenses, including serious disciplinary charges once imprisoned. In many correctional settings, including the department of corrections that was the source for our data, offenders classified at the maximum level reside in single bed units (i.e., they do not share living quarters), and contact with other offenders is minimal. The association between maximum security and suicide attempts may be related to offenders experiencing a negative environment, increased stress, reinforcement of negative self-worth, and social isolation [28,30,41]. The majority of offenders classified as maximum security were tobacco smokers (64.5%) and a plurality had a diagnosis of depression (38.4), significant risk factors for suicide attempts. Findings from this investigation suggest that greater odds for suicide attempts for offenders with co-occurring SUD and mental health disorders may differ somewhat from non-imprisoned populations and that moderating variables, such as gender and living environment, may have different influences on attempting suicide.Although the aims of this investigation sought to identify associations among sociodemographics, SUD, mental health disorders, and suicide attempts, the strong links among these factors may provide corrections guidance in regard to assessing suicide risk and targeting resources to the population with the greatest odds of attempting to end their life. Tobacco smoking and the population with co-occurring bipolar and anxiety appear to have the greatest odds among SUD and mental health disorders respectively. The population with co-occurring SUD and mental health disorders, particularly tobacco smoking, cannabis, and alcohol use disorders, also had greater odds for suicide attempts. These characteristics are likely the initial risk factors for building a corrections-specific assessment for suicide attempts. Further, security classification, particularly maximum confinements, suggests that the environment of the level may need to be evaluated to identify how corrections can maintain security and safety, while also minimizing risk factors for suicide attempts, such as increased stress and social isolation.The correctional population is a unique environment from non-imprisoned settings, such as security levels, overrepresentation of SUD, mental health disorders, and co-occurrences, limitations of agency to make decisions, large percent for less than a high school graduation and high prevalence of socially and economically disadvantaged backgrounds. Offenders also are separated involuntarily from their family (spouses, significant others, offspring), which may be stressors related to suicide attempts [33]. Despite these differences, suicide attempts is a major issue for the population with SUD and mental health disorders. This descriptive study provides preliminary results regarding the effect that moderating (e.g., race, gender, age, education, security level) and potential mediating variables (e.g., SUD, mental health disorders, co-occurrences) may have on suicide attempts in an offender population.This study is among a few investigations that have included an offender population to access the associations of SUD, mental health disorder, and self-harm (i.e., suicide attempts). Although the results indicated a significant association among the conditions, there were several limitations to this descriptive study. We collected nearly the entire population of this DOC (84%), but our mutually exclusive groups for SUD (alcohol, cannabis, intravenous drug, narcotics, and tobacco) and mental health disorders (anxiety, bipolar, depression or depressive symptoms, and psychotic disorders) resulted in small totals for each subgroup and limited complex comparisons, such as controlling for the presence of some variables. While the primary aims of this investigation were to evaluate risks (SUD and mental health disorders) related to suicide attempts in an offender population, we did not have a complete health history related to these conditions. For example, we neither collected the duration and extent of SUD nor the severity of mental health disorders.With regard to SUD, we did not know the reason for use (i.e., coping, pleasure, or social) that may be explanatory [60]. Information related to treatment history, adherence, or completion (i.e., pharmacological or addiction counseling) was not collected for this investigation, but may be explanatory, particularly in regard to the greater odds for suicide attempts for men compared to women. Men in this DOC may not access mental health services in similar proportions to women. We also did not have a complete history for previous suicidal ideation, attempts, completions, or family history of suicide. Competed suicides were documented in a restricted system and family history, which may be a risk factor [21,27,36,51], was not a routine initial screening question for new arrivals. We also did not have access to clinical notes regarding the details and severity of suicide attempts.The preliminary findings of this investigation suggest further exploration into SUD, mental health disorders, and co-occurrences. In addition to these conditions, there are a growing number of investigations that have found a relationship between chronic conditions (e.g., diabetes, epilepsy, cardiovascular disease) and self-harm, which was not included or examined as a potential mediating variable [8,61]. Despite several limitations, the majority of the data we did not collect are typically documented. These additional data likely will explain more completely the population at risk for attempting suicide, as well as provide an opportunity for analyses that will be more explanatory than descriptive.Corrections is the first and most extensive contact that many offenders have with a health care system. Corrections was mandated by Estelle v. Gamble (429 U.S. 97) (U.S. Supreme Court) to provide unfettered access to health care, which includes mental health services and addiction treatment [62]. Thus, offenders with these conditions are provided access to appropriate services, which they may or may not utilize. In regard to SUD (excluding tobacco smoking in some systems), imprisonment largely minimizes use and access to these substances while also providing treatment. However, this DOC did not provide combined and coordinated treatment for the population with co-occurring SUD and mental health disorders, despite having a shared EHR to facilitate coordination and management of care. Clinical notes documenting mental health care were siloed in the EHR. The protection of mental health notes, provided by psychiatrists and psychologists, is understandable, but treatment and interventions for co-occurrences of SUD and mental health disorders are likely more effective when designed and delivered to address both conditions in a coordinated way [4,63,64] while protecting sensitive confidential health information.The large number of suicide attempts documented during imprisonment compared to the few that occurred prior to prison suggests that corrections recognizes the issues and likely is addressing the problem. The number of suicide attempts that occur during imprisonment is also a call for corrections to continue identifying risk factors, evaluating the correctional environment, and enhancing screenings, especially since prevalence for SUD and mental health disorders changes over time [19]; thus, risk factors, environment, and screenings may also require adjustments. Self-harm in an offender population is not only a corrections issue, but also a public health concern, since the majority of the imprisoned population is released. These condition in released offenders, co-occurring or not, are exacerbated when treatment is unavailable, inaccessible, or too complex to navigate in a non-imprisoned setting. Untreated or poorly addressed SUD, mental health disorders, and co-occurrences that are related to high risk for suicide ideation, attempts, or completion have a large social and economic impact, particularly when individuals become functionally or physically disabled.Future collaborative investigations between corrections and institutions such as academia, public health, addiction treatment services, and mental health, will be essential for enhancing screening and treatment for groups with these conditions. Treatment programs that were developed for non-imprisoned populations may not be the most effective approach for corrections, which has similarities (but also unique characteristics, socially and structurally) to the general population. The risk factors for self-harm are complex and a more complete identification of characteristics associated with self-injurious thoughts and behaviors by subgroups will enhance screening and triaging into appropriate and complete programs. An assessment of current screening practice may identify unknown gaps that are readily addressable. Longer term investigations to identify risk factors will require larger studies with more complete health information and history. Investigations will need to include more expansive sociodemographic variables, such as marital or family status and measures of social connectedness, as well as the potential effects of physical health conditions. Larger and more comprehensive investigations also will facilitate conducting multivariate analyses to adjust the influences of confounding variables [65]. Data-mining methods, such as random forest, will be important for uncovering and explaining moderation effects associated with factors, such as SUD, mental health disorders, physical health, and self-harm [66].Madison L. Gates, Asher Turney, and Michelle Staples-Horne developed the concept for the manuscript; Madison L. Gates and Veronica Walker co-authored the introduction and collected and prepared data for analysis; Madison L. Gates designed and performed statistical analyses; All authors co-authored the results; Elizabeth Ferguson and Madison L. Gates guided authorship of discussion and conclusion; All authors reviewed and edited the final manuscript; Madison L. Gates provided overall guidance for writing the manuscript.The authors declare no conflict of interest.Extracted substance use disorders, mental health disorders, and suicide attempt.ICD-9: International Statistical Classification of Diseases.Population demographics.History of substance use disorders, mental health disorders, and suicide attempt.Group differences for suicide attempt.Percent suicide attempt by substance use and mental health disorders.Within group differences for suicide attempts.p-Value adjustment for multiple comparisons.Odds of suicide attempt for co-occurrence of substance use and mental health disorders.

Med-MDPI/ijerph_2/ijerph-14-03-00318.txt

@@ -0,0 +1,420 @@
+These authors contributed equally to this work.Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).In the USA, little is known about local variation in retail cigarette prices; price variation explained by taxes, bans, and area-level socio-demographics, and whether taxes and hospitality bans have synergistic effects on smoking prevalence. Cigarette prices 2001–2011 from chain supermarkets and drug stores (n = 2973) were linked to state taxes (n = 41), state and county bar/restaurant smoking bans, and census block group socio-demographics. Hierarchical models explored effects of taxes and bans on retail cigarette prices as well as county smoking prevalence (daily, non-daily). There was wide variation in store-level cigarette prices in part due to differences in state excise taxes. Excise taxes were only partially passed onto consumers (after adjustment, $1 tax associated with $0.90 increase in price, p < 0.0001) and the pass-through was slightly higher in areas that had bans but did not differ by area-level socio-demographics. Bans were associated with a slight increase in cigarette price (after adjustment, $0.09 per-pack, p < 0.0001). Taxes and bans were associated with reduction in smoking prevalence and taxes had a stronger association when combined with bans, suggesting a synergistic effect. Given wide variation in store-level prices, and uneven state/county implementation of taxes and bans, more federal policies should be considered.In the U.S., smoking prevalence has declined over time, from approximately 33% in 1980, 26% in 1990, 23% in 2010, and 18% in 2013 [1]. Policies that aim to generate revenue and reduce smoking have likely contributed to the decline [2,3]. Nevertheless, cigarette smoking continues to be a major public health problem in the U.S., resulting in approximately 480,000 deaths each year from smoking-related diseases [3].Work to date suggests that there is considerable heterogeneity in cigarette tax policies between and within states; and suggests that heterogeneity in pricing may be hampering public health tobacco control efforts. Yet, only a few U.S. studies have provided a detailed characterization of cigarette price heterogeneity within states and counties by examining store-level prices. Three U.S. studies analyzed store-level price data [4,5,6] but they all focused on short time periods (≤2 year) and two focused on only a single geographic region [4,6].One potential reason for local heterogeneity is due to differences in how much tobacco taxes are passed onto consumers. In the U.S., tobacco tax policy has used indirect taxes levied on the tobacco producer or vendor (“excise taxes”). Results of prior studies have been mixed regarding whether tobacco excise taxes are in fact fully passed on to the consumer [7,8], only partially passed on [5], or over-shifted [4,6,9,10,11]. Several studies have identified geographic differences in the pass-through rate [4,5] and two studies found differences by socioeconomic status [5,6], factors which may contribute to local variation in cigarette prices. The implementation of smoking bans is another potential contributor to local variability in cigarette prices. If bans decrease demand, then manufacturers may decide to lower prices and reduce their profit margin. Alternately, manufacturers may decide to accept lower volume sales and raise prices to maintain revenue. To date, the relationship between bans and prices is understudied. Only one study addressed this question using state-level price data from 1960–1990, and found significantly lower prices in localities with more stringent anti-smoking laws [9].In general, previous studies have described declining smoking prevalence following the introduction of tobacco tax increases [2,12] and indoor smoking bans in workplaces [13]. Recently, a few studies have analyzed effects of smoking bans in a small number of localities (including hospitality bar and restaurants) and also found them associated with reductions in smoking prevalence [14,15,16,17]. However, no studies have evaluated whether cigarette taxes and smoking bans have a synergistic effect on smoking prevalence.In sum, no study to date has used store-level price data to evaluate associations between hospitality bar and restaurant bans on cigarette prices including a large geographic area over a long period of time. To address this knowledge gap, this study used a large U.S. cigarette price dataset from 2001 to 2011 to document variation in store-level cigarette prices and associations with federal and state cigarette taxes, hospitality bans and area-level socio-demographic characteristics. In addition, we tested whether excise taxes were uniformly passed to consumers irrespective of area-level socio-demographics. We then explored ecologic variation in smoking prevalence by state and county and the contribution of store-level cigarette prices, state and federal cigarette taxes, and hospitality bans to smoking prevalence; and tested whether the presence of hospitality bans changes the impact of cigarette taxes on smoking prevalence.Pricing data 2001–2011 came from Information Resources Inc.’s (IRI) Academic Dataset, a panel of large chain supermarkets and drug stores in 47 U.S. market regions located in 41 states [18]. The current study includes 3084 large chain outlets (from 128 and 13 large chain grocery and drug store companies, respectively) located in 483 counties. Examples of the chain companies are: Albertson’s, A&P, Food Emporium, Pathmark, Walgreens, CVS, and Rite Aid [19].Cigarette prices were available for all Universal Product Codes sold at a store. In order to reduce price variation simply due to varying cigarette size and package size, we restricted the analytic sample to the most popular package size and type of cigarette in the dataset: king-sized and long-sized cigarettes (99% of all cigarette revenue) and single packs (57% of all cigarettes revenue). Cigarette price reflects the “shelf price” and includes excise taxes that the retailer may pass to the customer, store-level promotions and retailer coupons, but does not include sales tax and manufacturer coupons. There was no objective way to classify brands thus we included all types and did not differentiate (premium, standard, generic); note that upwards of 70% of tobacco sales come from premium or standard brands [20,21].State and Federal cigarette taxes for every month of the study period came from the Tax Burden on Tobacco [22]. Because cigarette taxes are applied as excise taxes, the tax may or may not be incorporated into the “shelf price” (at the discretion of the seller). County/municipal cigarette taxes were not used because there was no reliable source for historic data 2001–2011. Data that we manually collected from state websites suggested that the county/municipal taxes were not widespread and were small. For example, available data from 2011 indicated only 13.5% of stores in our dataset were in areas with county/municipal taxes; tax median $0.60 (range 0.05–0.85). The exception to this was New York City and Chicago, where local taxes were very large (1.50 and 2.68, respectively), thus, stores (n = 111) in these counties (n = 6) were excluded (Note that Alaska was not in our dataset). After this exclusion, there were 2973 stores in 477 counties for analyses.State and county workplace smoking bans and date of enactment came from the American Nonsmokers’ Rights Foundation (2001–2011) [23]. A county within a state that adopted a ban was also considered to have adopted the ban [24].The present study only included timing of hospitality indoor smoking bans requiring that all restaurants and free-standing bars be 100% smoke-free. We excluded: (a) restaurant-only or bar-only bans (maximum of only 4% of counties in our dataset in any year); (b) non-hospitality workplace smoking ban policies (in the U.S., most establishments had voluntarily banned indoor smoking by the late 1990s, thus, before the study period and years before the government enacted smoking bans in non-hospitality establishments [25]).Estimated county smoking prevalence for each year came from Dwyer-Lindgren et al. (2014) [26]. Their estimates were derived from the Behavioral Risk Factor Surveillance System (BRFSS) self-reported data on adult smokers [27]. BRFSS does not provide county-level estimates for much of the U.S., thus, Dwyer-Lindgren et al. predicted annual smoking prevalence (daily and any smoking) via a validated small area estimation method that utilized BRFSS, census demographics, state cigarette sales data, and other data (see their publication for more details [26]). We derived non-daily smoking prevalence by subtracting daily smoking prevalence from total smoking prevalence. For models where smoking was an outcome variable, all variables were summarized at the county-level.Data were summarized into annual measures. Adapting methods developed by others [7], at each store location, dollar and unit sales were first aggregated from weeks to years and from Universal Product Codes to standardized brand name to formulate an average yearly price per pack (PPP) for each brand and each store location. These yearly brand-store PPP were then used to create a weighted average price, where weights are based on the proportion of each brand sold during the entire study period. Thus, the weights resulted in yearly PPP independent of temporal brand buying patterns which could have over- or under-estimated cigarette price (prior work has found that some smokers respond to tax increases by switching to more expensive products or by buying less expensive brands, respectively [28]).Because some tobacco regulations are implemented mid-year, annual cigarette state tax at each store was weighted by proportion of packs sold before mid-year and after mid-year: Sum (Weekly Tax × Weekly Packs Sold)/Yearly Packs Sold. In the same manner, bans were calculated to account for mid-year policy changes. All prices and taxes were adjusted to 2010 dollars based on the U.S. Bureau of Labor and Statistics Consumer price index [29].Census data come from the middle of the study period: the American Community Survey (ACS) 5-year summary file 2005–2009 [30]. Each store was assigned to the population-weighted centroid of its block group (n = 2822) and block groups within 1-mile of each store were selected to represent characteristics of residents around each store (This approach was used because many chain stores were located in non-residential areas). Thus, census data in our analyses consist of average socio-demographics of the block groups surrounding each store. Census data were used to construct area-level variables related to age, race, and socio-economic status; these were included due to their potential spatial patterning (thus could confound the tax-price association) as well as their associations with smoking. In addition, SES was included because we were interested in examining differences in tax pass-through by SES. Age (AKA “age”) was represented by four variables: proportion of people aged 10–19, 20–39, 40–64, 65 and over [31]. Race was simplified to proportion of non-Hispanic white (AKA “race”) where a lower proportion means higher proportion of non-white or Hispanic. A socio-economic composite index (AKA “SES”) was derived as others have done [32] using: log per capita income; log median owner value; proportion of residents: with income from interest/dividends/rent, with high-school education, with bachelor’s degree, and with managerial employment. SES variables were normalized, averaged, and then converted into a percentile according to the normal distribution where 0% and 100% represents neighborhoods with the lowest and highest SES, respectively. Region was defined following census categories: Northeast, Midwest, South and West [33]. Urbanicity was based on county population size: large metro area of 1+ million residents, small metro area of less than 1 million residents, micropolitan urban areas (centered on an urban area with population 10,000 to 49,999), and non-core (all other areas smaller than micropolitan) [34].Additional variables were used as control variables in analyses due to their potential to confound the tax-price association. Type of retailer (AKA “store type”), defined as chain supermarket vs. drug store, came from the IRI dataset. Total state tobacco control appropriations 2001–2011 (AKA “tobacco control funding”) came from University of Illinois Chicago’s Bridging the Gap/ImpacTeen Project [35]. They compiled data on public funds allocated by each state for tobacco prevention and control. The allocated funds originate from federal and state sources as well as foundation grants given to states (American Legacy Foundation and Robert Wood Johnson Foundation). Tobacco control funding was included in our analyses to proxy population norms regarding smoking [36]. We converted to dollars per capita (per state population) and adjusted for inflation. We used a generalized linear model (appropriate for normally distributed outcomes) with random intercepts that accounted for clustering by state and county (appropriate for geographically nested data). Time, region, urbanicity, area-level age distributions, state tobacco control funding, store type, were first added to the model to assess variation in price accounted for by temporal and geographical attributes; then variables of interest were added (state taxes, and bans); socio-demographic variables (SES and race); and then interactions between state taxes and bans.The model looks as follows:
+        
+          (1)
+          
+            
+              
+   
+    
+     
+      
+       C
+	   i
+	   g
+	   a
+	   r
+	   e
+	   t
+	   t
+	   e
+	    
+	   P
+	   r
+	   i
+	   c
+	   
+        e
+        i
+       
+       
+     
+     
+      
+       =
+	    
+	   
+        β
+        0
+       
+       +
+	   
+        u
+        
+         jk
+		 
+       
+       +
+	   
+        u
+        j
+       
+       +
+	   
+        β
+        
+         1
+		 …
+		 10
+		 
+       
+       Y
+	   e
+	   a
+        
+        r
+        
+         1..10
+		 
+       
+       +
+	   
+        β
+        
+         11
+		 ,
+		  
+		 12
+		 ,
+		  
+		 13
+		 
+       
+       R
+	   e
+	   g
+	   i
+	   o
+	   
+        n
+        
+         N
+		 E
+		 ,
+		  
+		 S
+		 ,
+		  
+		 M
+		 W
+		 
+       
+       
+     
+    
+    
+     
+      
+      
+       +
+	   
+        β
+        
+         14
+		 ,
+		 15
+		 
+       
+       U
+	   r
+	   b
+	   a
+	   n
+	   i
+	   c
+	   i
+	   t
+	   
+        y
+        
+         r
+		 u
+		 r
+		 a
+		 l
+		 ,
+		  
+		 s
+		 m
+		 a
+		 l
+		 l
+		  
+		 m
+		 e
+		 t
+		 r
+		 o
+		 
+       
+       +
+        β
+        
+         16
+       
+       C
+	   i
+	   g
+	   a
+	   r
+	   e
+	   t
+	   t
+	   t
+	   e
+	    
+	   T
+	   a
+        x
+        j
+       
+       
+     
+    
+    
+     
+	 
+      
+       +
+        β
+        
+         17
+       
+       Hospitality Smoking Ba
+        n
+        
+         jk
+       
+       + 
+        β
+        
+         18
+       
+       SE
+        S
+        i
+       
+       + 
+        β
+        
+         19
+       
+       Rac
+        e
+        i
+       
+       
+     
+    
+    
+     
+      
+      
+       + 
+        β
+        
+         20
+       
+       Cigarettte Ta
+        x
+        j
+       
+        × Hospitality Smoking Ba
+        n
+        
+         jk
+       
+       + …+ 
+        e
+        
+         jki
+       
+       
+     
+    
+   
+   
+   
+          
+        
+        where j is a state, k is a county within that state, and i is a store within that county. Note that “…“ includes four parameters for age, one for tobacco control funding, and one for store type.
+        
+          (2)
+          
+            
+              
+   
+    
+     
+      
+       S
+	   m
+	   o
+	   k
+	   i
+	   n
+	   g
+	    
+	   P
+	   r
+	   e
+	   v
+	   a
+	   l
+	   e
+	   n
+	   c
+	   
+        e
+        k
+       
+       
+     
+    
+    
+     
+      
+               
+	   = 
+	   
+        β
+        0
+       
+       +
+        u
+        
+         jk
+       
+       +
+        u
+        j
+       
+       +
+        β
+        
+         1…10
+       
+       Yea
+        r
+        
+         1..10
+       
+       +
+        β
+        
+         11, 12, 13
+       
+       Regio
+        n
+        
+         NE, S, MW
+       
+       
+     
+    
+    
+     
+      
+	          
+       +
+        β
+        
+         14,15
+       
+       Urbanicit
+        y
+        
+         rural, small metro
+       
+       +
+        β
+        
+         16
+       
+       Cigarettte Ta
+        x
+        j
+       
+       
+     
+    
+    
+     
+      
+	          
+       +
+        β
+        
+         17
+       
+       Hospitality Smoking Ba
+        n
+        k
+       
+       +
+        β
+        
+         18
+       
+       Cigarette Pric
+        e
+        k
+       
+       + 
+        β
+        
+         19
+       
+       SE
+        S
+        k
+       
+       
+     
+    
+    
+     
+      
+	          
+       + 
+        β
+        
+         20
+       
+       Rac
+        e
+        k
+       
+       + 
+        β
+        
+         21
+       
+       Cigarettte Ta
+        x
+        j
+       
+        × Hospitality Smoking Ba
+        n
+        
+         jk
+       
+       
+     
+    
+    
+     
+      
+	          
+       + …+ 
+        e
+        
+         jk
+       
+       
+     
+    
+   
+ 
+          
+        
+        where j is a state and k is a county within that state. Note that “…“ includes four parameters for age, and one for tobacco control funing.In the price model, we included dummy variables for each year rather than a linear time trend in order to account for spurious time correlation between taxes and prices as there are potentially a number of other factors that could have led to changes in the price of cigarettes, most notably the federal tax increase, but also other factors such as operating costs of cigarette vendors, shipping prices and tobacco prices. Price was non-linear over time and a dummy coding approach is consistent with the modeling approach of other studies [5,7,10]. For the smoking prevalence model, smoking was only somewhat non-linear thus, main results are reported using a more parsimonious model that included linear time + quadratic time + dummy variable for federal cigarette tax. Variance decomposition was assessed via intraclass correlation coefficients and state- and county- residual plots (following work by others [37]) in order to illustrate variation of cigarette prices and smoking prevalence across the study area before and after accounting for covariates.Sensitivity analyses used a fixed effects model that dummy-coded state-county instead of using random intercept terms. Results were virtually the same to those reported using the random intercepts; those results are shown in the Supplement (Tables S1 and S2) and not discussed further.Figure 1 shows store-level characteristics for 2001–2011: cigarette prices, taxes and hospitality bans, and smoking prevalence (all prices/taxes were indexed to 2010 dollars). Over the study period, average cigarette PPP rose from $4.46 to $6.15 (a $1.69 or 38% increase). During this time federal taxes rose from $0.43 to $1.01 with the largest increase in April 2009 [38]. Minus federal- and state tobacco taxes, store-level prices rose from $3.80 to $4.77 and accounted for most of the total price of a pack, although on average it represented a slightly declining share of the price (decreased from 85% to 78% of the total PPP). The percent of stores in states-counties with hospitality smoking bans rose from 1% to 46%. Total smoking prevalence declined from 23.5% to 19.4% (4.1 points or 17% lower) but the decline was only evident in the prevalence of daily smokers. (By 2011, counties included in our dataset matched the prevalence of current smoking in the U.S. as a whole: 20% [39]). Table 1 shows descriptive data for store-level cigarette price, policy exposures and smoking rates stratified by region, urban class and then by area-level attributes. Averaged across years, stores in the West and Northeast had the highest cigarette prices. Stores in less urbanized areas and in the southern U.S. had the lowest cigarette prices and state taxes accounted for a much lower proportion of the total price, state/county hospitality bans were less prevalent, and daily smoking more prevalent. For example, in the south: state taxes were only 11% of the price of a pack, only 5% of stores were in areas that had hospitality bans, and 17.1% of adult residents were daily smokers. In areas with higher SES, cigarette prices were higher, daily smoking was lower, and state/county bans more prevalent (by the end of the study period).After adjustment, the magnitude of cigarette price differences by region and urbanicity was diminished, but was still apparent (Table 2, adjusted for year, cigarette taxes, state/county smoking bans, and age distribution, state tobacco control funding, store type, SES, and proportion non-Hispanic white). Western areas had the most expensive cigarette prices and Midwest and South the cheapest: relative to the West, per-pack prices in these regions were $0.42 and $0.28 cheaper, respectively (p < 0.0001, Table 2 model 1.4). In adjusted models, micropolitan and non-core areas had the lowest cigarette prices than large metros ($0.15 and $0.19 cheaper, respectively).State excise taxes were only partially passed onto the consumer and the presence of bans was associated with a slightly higher PPP. After adjusting for year (including federal tax), region, urbanicity, state/county smoking bans, and age distribution, state tobacco control funding, store type, SES, and proportion non-Hispanic white, a $1 increase in state tax was associated with $0.90 increase in PPP and the presence of bans was associated with a $0.09 higher PPP, p < 0.0001 (Table 2 Model 1.4). In the presence of bans the state tax pass-through was slightly higher (0.852 + 0.096 = 0.95, p for interaction >0.0001). Price of cigarettes was higher with neighborhood SES index ($0.02, per 10% increase in SES index) but was not associated with proportion non-Hispanic white and there was no evidence of an interaction between state tax and either SES or race (p > 0.2).Supplement Figure S1A illustrates state and store-level differences in prices. Table 2 variance decomposition analyses confirmed that full adjustment for taxes, bans, and area-level characteristics accounted for much of this variability: before and after full adjustment intraclass correlation decreased from 30% to 15%.Midwest, Northeast and Southern regions had daily smoking prevalence at least 2 percentage points higher than the Western region (Table 1). Supplement Figure S1B, illustrates that relative to other states, daily smoking was lowest in Utah and California and highest in Oklahoma, Tennessee and West Virginia. Comparing adjusted models for daily and non-daily smoking, state tobacco tax appeared to have a stronger inverse association with daily smoking. Hospitality bans had an inverse association with both daily and non-daily smoking (Table 3). For example, at an equivalent level of cigarette taxes and prices, the presence of bans was associated with −0.12 and −0.18 percentage point reduction in non-daily and daily smoking, respectively. Interaction results suggest that the presence of bans strengthened the inverse association between cigarette taxes and daily smoking but did not impact the association for non-daily smoking (interaction p < 0.0001 and p = 0.9, respectively). For example, for each $1 increase in cigarette tax, the expected prevalence of daily smokers decreased by −0.28 percentage points in areas without smoking bans and by −0.53 percentage points in areas with bans (−0.282 − 0.247 = −0.53), after adjusting for geographic factors and covariates (Statistical note: in this calculation, the main effect of ban is omitted because it does not describe the effect of changing taxes on price).Using a large dataset of chain supermarkets and drug stores between 2001 and 2011, we found wide variation in store-level cigarette prices that was in part due to differences in state taxes. Cigarette excise taxes were only partially passed onto consumers (a $1 tax was associated with a $0.9 increase in PPP) and pass-through rates did not differ based on area-level socio-demographics but was slightly higher in areas with hospitality bans. Hospitality bans were associated with a slight increase in cigarette price ($0.17 PPP adjusted for state tax, year, area-level age, SES index, and race). Exploratory work analyzing county-level daily smoking found higher taxes and bans inversely associated with daily smoking and impacts were stronger in the presence of the other (indicating a synergistic association). Smoking prevalence has declined over time [1,40] yet remains a significant public health problem and varies by locale. One potential reason for local heterogeneity is due to differences in how much tobacco taxes are passed onto consumers. Our finding, that taxes were only partially passed on, was consistent with two studies that used disaggregated store-level UPC-level data. Harding (2012) used a shorter time period and found a pass-through of approximately 0.85 among a household panel and Chiou (2014) used data from supermarkets in a single municipality (the Chicago area) and found an average pass-through of 0.80 [5,28]. Other studies that reported taxes were fully passed or over-shifted onto consumers while using data from supermarkets only or fewer municipalities or price data from individual self-reports [4,8,9,10] Advantages of the dataset we used over other studies were: use of prices of specific goods rather than average price paid which may reflect substitution; included many more years of data relative [4,7,8] and offered better spatial resolution [9,10]. Our findings suggest that cigarette price varied by SES but the variation was small and in general, state cigarette taxes were equitably passed to consumers as there was no evidence that associations between state excise tax and price were different based on area-level SES or minority composition of residents near the chain stores. However, more work on this topic is needed as Harding (2012), using 2 years of household panel data, found that the pass though was less than full throughout the sample but increased with panelists’ household income (with increases in income, pass through rose from 0.811 to 0.897).Prior studies have estimated that, in the U.S., a 10% increase in price would reduce per capita tobacco consumption between 1% and 6% (price elasticities ranging from −0.1 to −0.6) [2,41,42]. Our exploratory analyses confirmed that after adjusting for area-level characteristics and temporal trend and tobacco prices independent of state tax, a $1 increase in the state cigarette tax was associated with average 0.44 lower daily smoking prevalence (−0.44, p < 0.0001), although there was minimal effect on non-daily smoking. If we assume that 90 cents out of $1 state cigarette tax is shifted to consumers, average price elasticity is −0.1 (or if fixed to 2011 prices and daily smoking prevalence it would be −0.2), thus within the range found in prior literature. Reasons that our estimates are slightly lower are: our data are from the 2000s and elasticity estimates were strongest in the U.S. in the 1990s [42]; most studies did not use prices from stores located in a large area that was primarily urbanized. In addition, most studies did not distinguish between daily and non-daily smoking but they benefited from having individual-level data whereas we only reported county-level estimates. Our results may suggest that in the presence of smoking bans, state cigarette taxes could be more effective in reducing daily smoking (the presence of hospitality bans strengthened the inverse association between cigarette taxes and daily smoking prevalence, see Section 3.2.2). The effect of hospitality bans itself on smoking prevalence was quite small although statistically significant for county-level non-daily smokers who are sometimes referred to as “social smokers” [43,44] (as opposed to daily smokers). As hospitality bans become more prevalent, more research will be needed on their effects across diverse geographies and on diverse smoking outcomes measured at the individual level.On average, state taxes were <25% of the total price of cigarettes and <50% of counties in this study had hospitality bans by the end of the study period. Our data primarily represented urbanized areas nevertheless results suggested that policies varied regionally and by urbanicity, with counties in less urbanized areas and in the southern United States having relatively low state taxes and low prevalence of county hospitality bans. Variations in regulations have been shown to give rise to tax avoidance via internet purchases and purchases from bordering states with lower tax rates [5,8,45,46], limiting the efficacy of state-level cigarette taxes. A more effective strategy might be to implement a higher federal cigarette tax rate, as a federal tax would apply uniformly across the United States.This study used a very detailed novel price dataset that included many items, specified brand, and prices were temporally and spatially resolved. However, the dataset only included chain grocery and drug store prices; and stores were almost exclusively in urbanized areas (and excluded New York City and Chicago stores, see methods) thus, generalizability of prices are limited to these venues/contexts. Cigarette carton price was not included thus generalizability of results is limited to cigarette packs. Some work has found higher pass through for cartons [28] while others have found lower pass through for cartons [8,11]. We carefully constructed an average price per store using methods developed by others [7] and that minimized biases specific to our research question (see details in Section 2.1.5); nevertheless, we note that other aggregation methods may generate different results. We did not control or stratify for discount cigarettes (where pass through may be higher [28], see methods for rationale) or store proximity to a jurisdiction with low-taxes/no ban (where pass through may be lower [28] and may have generated non-differential measurement error in the smoking analyses may have weakened the estimates). We were unable to include bans enacted by cities within state-counties that did not have any ban. This omission would not have impacted county-level analyses (Table 3 smoking outcome). Note that among the bans enacted in our study area and time period, bans were commonly enacted at the state level (61% of states in our IRI database enacted ordinances prior to or coincident with our study period) thus obviating the need to assess city ordinances [24]. Nevertheless, possible that this omission could have impacted ban estimates on price outcomes (Table 2 price outcome). Finally, this study did not have access to individual-level data on purchases and smoking status nor number of cigarettes smoked per day and number of quit attempts, smoking outcomes that others have found are responsive to price increases [47,48,49]. Despite its limitations, our study makes a clear contribution to the literature due to its ability to characterize local variation, adjust for local context, and include a long study period and large geographic coverage.Cigarette prices at chain stores varied widely within and between states and counties and were only partially explained by differences in state taxes. Wide variation may weaken the effectiveness of these policies at reducing smoking prevalence. Consistent rigorous policies across county and state borders—such as higher federal taxes and/or a federal hospitality smoking ban—may be needed in order to further reduce average smoking prevalence. The following are available online at www.mdpi.com/1660-4601/14/3/318/s1, Figure S1: Variation by state in (A) average cigarette price per pack by state (B) daily smoking prevalence; with 95% confidence intervals. Y-axis is the mean residual and X-axis displays rankings low to high. Y-axis residuals means that 0 is the overall average, −1.0 on the y-axis means the value is −1.0 than average, Table S1: Sensitivity to using fixed effects model. Mean differences in store-level cigarette prices per pack; data from 2001 to 2011, n = 2973 chain supermarkets and convenience stores, Table S2: Sensitivity to using fixed effects model. Adjusted mean difference in county-level smoking prevalence according to cigarette price, state tax, hospitality ban and interactions, data from 2001 to 2011.The authors thank Mark Stehr for his comments on the paper. The authors take sole responsibility for data analyses, interpretation, and views expressed in this paper. Any errors in the manuscript are the sole responsibility of the authors, not IRI (who supplied the pricing dataset). Mention of trade names, commercial practices, or organizations does not imply endorsement by the authors, the institutions where the authors work, nor by the funding entities. This work was partially supported by National Institute on Minority Health and Health Disparities at the National Institutes of Health (USA) (grant number P60 MD002249).Amy H. Auchincloss obtained the data, designed the study and contributed to data analyses. Lance S. Ballester compiled the database and lead the data analysis. Lance S. Ballester, Amy H. Auchincloss, Stephanie L. Mayne drafted and revised the manuscript. Lucy F. Robinson contributed to study design, analyses, and critically reviewed the manuscript. All authors edited, reviewed and approved the final version of the submitted manuscript. The authors declare no conflict of interest.Per pack cigarette price and taxes at chain supermarkets and drug stores, county hospitality indoor smoking bans, and county smoking prevalence, 2001–2011. All prices and taxes were adjusted to 2010 dollars.Characteristics of stores in the dataset: cigarette prices (per pack), taxes (per pack), hospitality bans, and smoking prevalence by region, urbanicity, and area-level socio-demographic characteristics.* Area-level refers to block group cluster; † Hispanic and non-white were collapsed in order to obtain sufficient population numbers across block groups. “non-Hispanic White” can also be interpreted as percent that are Hispanic or non-white (1-non-Hispanic white), thus the tertile distribution is 29%–97%, 13.4%–29%, %0–13.4%.Mean differences in store-level cigarette prices per pack; data from 2001 to 2011, n = 2973 chain supermarkets and convenience stores; estimates are derived from a random intercept model *.Est. = Estimate. Ref = Referent value; * Random intercepts were included for state and county. Intraclass correlation is the proportion of variation in price that is accounted for by price differences at the state and county levels; † Base adjustment. Time was a dummy variable in the model (year 2001–2011) and accounts for federal cigarette tax. Additional base adjustment variables are: region, urbanicity, state tobacco control funding, store type, and area-level age (percent of population aged 10–19, 20–39, 40–64, 65+); ‡ Urbanicity is a county-level variable (see methods); § Area-level refers to block group cluster. Area-level socio-economic index units are displayed in 10 percentile increments (cigarette price increases $0.02 per 10% increase in socio-economic index). Adjusted mean difference in county-level smoking prevalence according to cigarette price, state tax, hospitality ban and interactions, data from 2001 to 2011, n = 477 U.S. counties. Estimates are derived from a random intercept model *.Est. = Estimate. Ref = Referent value; * Random intercepts were included for counties (within-states); † Base adjustment. All models include time which was entered as a linear term for year + year squared + dummy variable to indicate before or after year 2009 (the year when the federal tax increased across all U.S. states). Additional base adjustment variables are: region, urbanicity, cigarette price, state tobacco control funding, store type, and area-level age (percent of population aged 10–19, 20–39, 40–64, 65+).

Med-MDPI/ijerph_2/ijerph-14-03-00319.txt

@@ -0,0 +1 @@
+Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).A hand-foot-mouth disease outbreak occurred in 2014 around Guangdong. The purpose of this study was investigating the status and susceptibility of infectious neutralizing antibodies to enterovirus 71 among children so as to provide scientific evidence for the population immunity level of hand-foot-mouth disease and prepare for enterovirus 71 vaccination implementation. Serum specimens were collected from children in communities from January 2014 to March 2015 in Guangzhou. A total of 197 serum samples from children 1–5 years old were collected for this cross-sectional study via non-probabilistic sampling from the database of Chinese National Science and Technique Major Project. Neutralization activity was measured via micro neutralization test in vitro. The positive rate of enterovirus 71 neutralizing antibodies was 59.4%, whereas the geometric mean titre was 1:12.7. A statistically significant difference in true positive rates was found between different age groups but not between different genders. Being the most susceptible population of hand–foot–mouth disease, children under 3 years of age are more likely to be infected with enterovirus 71, and the immunity of children increases with increasing age. Further cohort studies should be conducted, and measures for prevention and vaccination should be taken.According to the updated surveillance summary of World Health Organization [1], an outbreak of hand-foot-mouth disease (HFMD) was recorded in China in 2014. In 2015, 2,014,999 cases of HFMD were reported in China. Zhang [2] reported that the incidence of HFMD in Guangdong increased, peaking twice each year from 2009 to 2012 (incidence range was from 9.75 × 10−5 to 32.08 × 10−5). Enterovirus 71 (EV71, species Enterovirus A, genus Enterovirus, family Picornaviridae, order Picornavirales) is a major causative agent of HFMD. Jin [3] conducted an etiological study and found that 51% of HFMD cases in China in 2011 were caused by EV71. EV71 infection causes fever, skin eruptions on hands and feet, as well as vesicles in the mouth. Meningitis, encephalomyelitis, and neurogenic pulmonary oedema [4] may be involved in rare cases, which may cause serious sequelae or death [5]. EV71 seroepidemiological studies were conducted in Germany [6], Brazil [7], Singapore [8], Taiwan [9,10], and several cities in mainland China, such as Shanghai, Shenzhen [11], Henan [12], and Handan. Guangzhou, the capital of Guangdong, presents high population mobility in Southern China; Guangzhou residents, especially children, may be highly susceptible to EV71. Hence, few seroepidemiological studies focused on this region during the 2014 outbreak. Estimation of the seroprevalence of EV71 neutralizing antibodies and the level of herd immunity is important to provide the supporting principle for HFMD prevention and control strategies.An individual who is susceptible to EV71 can be infected by asymptomatic infection. Sun [13] reported a high asymptomatic infection rate of EV71 (24%) during the epidemic season in Wenzhou in 2012. Given that HFMD is an acute self-limited infectious disease, humans can produce neutralizing antibodies by inducing immune responses, either with symptomatic infection or asymptomatic infection. EV71 neutralizing antibodies can be detected 1 or 2 weeks after infection and may persist for at least a year [14]. Xu [15] conducted a seroepidemiological research on 254 children in Henan between 2010 and 2012, and found a positive correlation (correlation coefficient r = 0.80) between the positive rate of EV71 neutralizing antibodies and morbidity within the same year. The status of EV71 neutralizing antibodies should be explored to determine HFMD morbidity and population immunity. By contrast, the seroepidemiological study of Rong [16] in Guangzhou revealed significant differences in seropositivity between the younger age group and the eldest group. The samples were obtained not only from children but also from adults aged 24–35 years. Therefore, investigation of the positive rate of EV71 neutralizing antibodies and the susceptibility to EV71 among children aged 1–5 years in Guangzhou is essential to provide the seroepidemiological information necessary for region disease control implementation and EV71 vaccination.All subjects provided their signed informed consent to inclusion prior to participation in the study. The study was conducted in accordance with the 1964 Declaration of Helsinki, and the protocol was approved by the Ethics Committee of Institutional Review Board of School of Public Health, Sun Yat-sen University. The ethical code is L2016[021]. Our study was conducted as a cross-sectional study via non-probabilistic sampling. We calculated the sample size as 119 with an estimated neutralizing antibody positive rate (abbreviation as “p”) of 59%, significance level of 0.05, and permissible error of 0.15 “p”. Considering the small sample number for each age group, we collected 197 samples between 2014 and 2015. All serum specimens and demographic characteristics of children in Guangzhou were collected from the blood sample database set up by Chinese National Science and Technique Major Project (2012ZX10004912). Child participants were divided into four groups: under 2 years old, between 2 and 3 years old, between 3 and 4 years old, and between 4 and 5 years old. As positive control, the EV71 positive serum was collected from child patients identified by Yuebei Hospital in Shaoguan City. Human rhabdomyoma (RD) cells were provided by Guangdong Centre for Diseases Control (CDC) and used for virus sub-cultivation and antibody neutralization test. EV71 (virus strain number 2014XN37281) was collected from HFMD cases, isolated, cultured, and sequencing identified by Guangzhou CDC.Virus titration was conducted via endpoint dilution assay. This assay was selected because of its higher sensitivity and cost-effectiveness compared with plaque assay [17,18]. Virus liquid was sequentially diluted by minimum essential medium (MEM) in gradient dilutions from 10−1 to 10−10 and then cultured in 96-well culture plates (8 × 12 wells). Eight wells per vertical row in each culture plate were for one dilution level of virus liquid inoculated (50 µL was added to each well). Each culture plate included one cultivating liquid group and one cell suspension group as controls. To prevent the infection from being trailed from non-neutralized virus wells to the cell culture suspension, all tips were changed for each new row. A 50% Tissue Culture Infective Dose (TCID50) was generated to measure the EV71 antibody titre, and mathematically calculated using the Reed–Muench method [19]. Observation and recording lasted for 3 days.EV71 neutralizing antibody titre was determined by the absence of cytopathic effects (CPE) in the micro neutralization test [20,21]. Serum samples were diluted at a sample-treatment-liquid ratio of 1:4 and processed under homogeneous vibration. After being placed at a constant temperature of 4 °C overnight, the samples were inactivated the next day at 56 °C for 30 min. Then, the samples were diluted sequentially twice from 2−1 to 2−9 and cultured in 96-well culture plates within two wells for one dilution level (50 µL per well). Then, diluted 100 TCID50 virus liquid (50 µL per well) was added. To prevent cross infection, all tips were changed for each new row. After homogeneous vibration, these culture plates were placed in a carbon dioxide incubator for 2 h neutralization. RD cell liquid was added (100 µL per well), and the plates were placed back into the carbon dioxide incubator at 35 °C. RD cells were used for virus subcultivation. Occurrence of CPE in RD cell culture indicates the presence of virus activities. When EV71 neutralizing antibody is present in RD cell culture, the infectivity of EV71 is reduced, leading to the absence of CPE in the cell culture. Therefore, the absence of CPE proved the positive result of EV71 neutralizing antibodies. CPE of these 96-well culture plates was observed under an inverted microscope daily for 3 days. The serum control group, virus control group, and blank control group were added simultaneously.Standard prediction was performed after the micro neutralization test. The multiplicative inverse of the highest dilution level under CPE absence was the determined point of neutralizing antibody titre. The antibody titre value determined the positive or negative results of EV71 neutralizing antibodies. If the neutralizing antibody titre values were less than or equal to 1:4, 1:4 was recorded as negative. The value was classified as positive if the neutralizing antibody titre values were greater than or equal to 1:8. All neutralizing antibody titre values that were greater than 1:32 were recorded as ≥1:32. Calculations of the geometric mean titre (GMT) of EV71 neutralizing antibodies and its 95% confidence interval (95% CI) were generated.Statistical analysis of data was performed by SPSS 19.0 (IBM SPSS Statistics, Armonk, NY, USA). Chi-square test was used to compare the positive rates of EV71 neutralizing antibodies within different ages, genders, and years. Analysis of variance (ANOVA) was used to compare the logarithm values of neutralizing antibodies GMT within different ages, genders, and years. Cochran-Armitage Trend Test was performed to verify the trend of positive rates along with age. The significance level was set to 5% (α = 0.05). Pairwise comparison of the neutralizing antibody GMT between two age groups was processed in Fisher’s least significant difference (LSD) with significant level correction.In total, 197 children aged 1–5 years participated in this study. The participants comprised 112 males and 85 females (male-to-female ratio was 1.3:1). Groups of children aged <2, 2–3, 3–4, and 4–5 years included 17, 60, 60, and 60 samples, respectively.RD cells can support the replications of EV71; therefore, CPE occurred in the RD cell culture as shown in Figure 1a, indicating the presence of EV71 activity. The absence of CPE shown in Figure 1b indicates a positive EV71 neutralizing antibodies result.As shown in Figure 2, 117 specimens tested positive among the 197 serum specimens of children (positive rate was 59.4%). The GMT of EV71 neutralizing antibody was 1:12.7 (95% CI: 1:10.8, 1:14.6). The positive rates of EV71 neutralizing antibodies in the groups of children aged <2, 2–3, 3–4, and 4–5 years were 35.3% (6/17), 43.3% (26/60), 65.0% (39/60), and 76.7% (46/60), respectively. The positive rates significantly increased with increasing age (z = 3.361, p < 0.001). Statistically significant differences in positive rates were observed among the four groups (Pearson χ2 = 18.715, p < 0.05). The total GMT value of EV71 neutralizing antibodies was 1:12.7 (95% CI: 1:10.8, 1:14.6). Furthermore, the GMT values were 1:8.0 (95% CI: 1:1.8, 1:14.2), 1:9.1 (95% CI: 1:5.8, 1:12.4), 1:13.9 (95% CI: 1:10.6, 1:17.2), and 1:18.4 (95% CI: 1:15.3, 1:21.5) in the four groups (Figure 3). More neutralizing antibodies were detected in the serum of the elder-age group than in the sera of the other age groups of children.Statistically significant differences in the logarithm values of GMT were observed among the four groups (F = 7.092, p < 0.05). However, statistically significant differences in GMT only were found between the groups of children aged <2 and 4–5 years (p = 0.002), as well as between the groups of children aged 2–3 and 4–5 years (p = 0.000077).The positive rates of EV71 neutralizing antibodies were 59.8% for boys and 58.8% for girls. Among the boys’ titre values, 40.2% (45 samples out of 112) were less than 1:4, 8.0% (9/112) were 1:16, and 51.8% (58/112) were equal or greater than 1:32. Among the girls’ titre values, 41.2% (35/85) were less than 1:4, 4.2% (4/85) were 1:8, 7.1% (6/85) were 1:16, and 47.1% (40/85) were equal or greater than 1:32. No statistically significant difference in positive rates of EV71 neutralizing antibodies were observed between different genders (p = 0.888). The GMT values of EV71 neutralizing antibodies were 1:13.1 (95% CI: 1:10.6, 1:15.6) for boys and 1:12.1 (95% CI: 1:9.3, 1:15.0; Figure 4) for girls. No statistically significant difference in GMT was found between different genders (p = 0.579).Enterovirus 71 is the major pathogen that causes HFMD, especially among children. In Shenzhen, China, where the Pearl River delta is located along with Guangzhou, the positive rate of EV 71 Immunoglobulin G (IgG) antibodies was 45.64% in 2012 among children below 5 years of age [22]. According to Kuang [23], this value was 30.83% in Guangzhou in 2010. These findings show that the positive rate of antibodies in Guangzhou was higher in this study (59.4%) than before, indicating that the immunity of children increased during the past 5 years. Zeng [24] observed an inverse correlation between specific age and EV71-infected HFMD cases in Shanghai. Therefore, individual susceptibility to EV71 could be determined by the level of EV71 neutralizing antibodies. In the present study, statistically significant differences in positive rates, as well as in the logarithm values of GMT, were observed among the four age groups. Therefore, we can conclude that during 2014 and 2015, different groups of children under 5 years of age in Guangzhou presented various levels of susceptibility to EV71. Moreover, the positive rate for elder children was higher than that for younger children, implicating that children aged less than 3 years were the most susceptible population. Zhou [11] conducted a seroepidemiological study of EV71 in 2007 in Shenzhen and found that the positive rate of antibodies (30.8%) in the elder group (2–5 years old) was higher than that in the younger groups (1–2 years old and <1 years old, 19.3% and 19.0% respectively). The results of the present study agree with those found by Zhou. Given that children under this age are either under home care (under 2 years old) or going to nursery school (between 2 and 3 years old), preventive measures at home and in school should be reinforced, including avoidance of direct contact with infected persons and proper hand-washing procedures.Focusing on the details for each group, we found the lowest positive rate of EV71 neutralizing antibodies in children under 2 years old (35.3%), and the GMT of this group was 1:8.0. Similar results were obtained by Li [25] in Guangdong, China and by Linsuwanon [26] in Thailand, indicating that children under 2 years of age presented the lowest seropositive rate of EV71 neutralizing antibodies. Children between 4 and 5 years old in this study showed the highest positive rate (76.7%) among all the groups, presenting more than two times the positive rate for children under 2 years of age. Younger children with incomplete immune system may show more than two times the probability of being secondary infected. In Jin’s [3] HFMD epidemic features study in 2011, morbidity rapidly increased with age under 1 year of age and reached the peak of morbidity. Among children older than 1 year, morbidity decreased rapidly after the peak but remained on a higher level with increasing age. The high morbidity of HFMD at this age is due to the fluctuation in the amount of maternal antibodies, as well as the unstable, incompletely developed immune system function of children. In Ooi’s [9] study from Singapore, 44.0% of pregnant mothers possessed EV71 neutralizing antibodies, which waned rapidly. After a month, none of the new-born babies who were tested possessed maternal antibodies to EV71. With low immunity level and rapidly decreasing number of maternal antibodies during child growth, the risk of infection would be higher if the children did not possess the EV71 neutralizing antibody. Younger children are less aware of HFMD risk and good hygienic preventive actions, and also lack supervision because of the insufficient numbers of nurses or teachers in nursery schools and kindergartens. These children are more easily infected by EV71 via respiratory droplets from close contact. Meanwhile, children aged beyond 4 years show decreasing morbidity of HFMD, which may be associated with recessive infection or higher positive rate of EV71 neutralizing antibodies. Raising public awareness of HFMD in children under 3 years of age, broadcasting HFMD general information and its prevention measures, improving the situation of health care in kindergartens and nursery schools, and implementing vaccines against EV71 are needed. Proper hand hygienic preventions, especially for children under 3 years of age, should be supervised by the guardians and tutors. Currently in Mainland China, inactivated vaccines against EV71 passed phase III clinical trial [27,28], and these inactivated vaccines as Class 1 preventive biological products received the approval of production registration from China Food and Drug Administration (CFDA) in 2015. Children under 3 years of age should be primarily considered as the targeted population for vaccination. The seroprevalence of EV71 neutralizing antibodies based on age groups would be useful to establish immunization protocols.The positive rates of EV71 neutralizing antibodies for boys and girls were 59.8% and 58.8%, showing no statistical significance between genders. In a German study on 696 individuals, Rabenau [29] found that the positive rate of EV71 neutralizing antibodies in males is 41.3%, which is higher than that of females (44.4%), but the difference is not statistically significant. Similarly, Horwood [30] found that the seroepidemiology of EV71 neutralizing antibodies relative to genders in Cambodia does not show a significant difference (89.8% for females and 87.7% for males, p = 0.18). Consistent with the results of these studies, the transmission of EV71 can be inferred to be nonselective toward gender in the Guangzhou population. By contrast, no significant differences in positive rate or GMT were found among children in Guangzhou between 2014 and 2015. This outcome differed from the findings of Li [25] in 2013 that the seropositivity rate of EV71 neutralizing antibodies in Guangdong significantly increased after the 2010 epidemic. In the comparison of the two findings from the same region, the Guangzhou 2014 HFMD outbreak did not affect the antibody immunity level among children. However, these results could also be due to the closeness of the collection years 2014 and 2015 for observing a significant difference. The surveillance of EV71 should continue even after the HFMD outbreak, and inactivated vaccines against EV71 are emergently needed, particularly for children (both boys and girls) under 3 years of age, after the 2014 HFMD outbreak to improve the immunity of the susceptible population.The limitation of our study is the different number of samples for each age group. The sample size of children under 2 years of age was markedly smaller than those of the three other groups. Chang [31] indicated that the highest mortality rate (15.6/105) can be found in children between 6 and 11 months old. Therefore, further seroepidemiological study on EV71 should be continued, including sample collection from under children below 1 year of age. As an important factor of the detection prevalence validation of EV71 neutralizing antibodies, the precise duration of EV71 neutralizing antibodies should be investigated. Given that the susceptible population involves children under 3 years old, specific medical treatments and vaccination to EV71 for children should be under the national and regional implementations.Children under 3 years of age are highly susceptible to HFMD. During 2014 and 2015, the positive rate of EV71 neutralizing antibodies among 197 serum specimens from children aged 1–5 was 59.4%, whereas the GMT was 1:12.7. Statistically significant differences in positive rates were observed between different age groups but not between different genders. Given the HFMD transmission routes and the characteristics of susceptible population, prevention and control measures should be taken. Further epidemiological studies and vaccination to EV71 should be continued under the national and regional protocols of implementation.This study was funded by the Guangzhou City Science and Technology Plan—The Pearl River New Star special project (grant number 201506010072); the Natural Science Foundation of China (grant number 81473064); and the National Science and Technique Major Project of China (grant number 2012ZX10004912, 2012ZX10004213). We owe special thanks to Qinlong Jing and Jinmei Geng from Guangzhou Centre for Diseases Control and Prevention for providing the enterovirus 71 virus strain, Zhanzhong Ma from Yuebei Hospital in Shaoguan City for providing EV71 positive serum, and Huanying Zheng from Guangdong Centre for Diseases Control and Prevention for providing human rhabdomyoma cells (RD cells).Dingmei Zhang designed the study and drafted the manuscript. Yan Chen was responsible for data collection and analysis. Xiashi Chen analysed the data and reviewed the work and the manuscript. Zhenjian He was responsible for the laboratory experiments. Xun Zhu was responsible for cell culture. Yuantao Hao, the corresponding author, approved and reviewed the work, as well as supported the study with serum sources and funding sources. All authors read and approved the final manuscript.The authors declare no conflict of interest. The founding sponsors performed no role in the design of the study; the collection, analyses, or interpretation of data; writing of the manuscript; and the decision to publish the results.(a) RD cells with CPE; (b) normal RD cells; absence of CPE testified the positive result of EV71 neutralizing antibodies; multiplicative inverse of the highest dilution level without the presence of CPE was the determined point of EV71 neutralizing antibody titre.Antibody titre and positive rates of EV71 neutralizing antibodies in different age groups of children. Negative results were determined by titres of antibodies (≤1:4), and positive results were determined by titres of antibodies (1:8, 1:16, and ≥1:32). All neutralizing antibody titre values that exceed 1:32 were recorded as ≥1:32. Positive rates were generated from all positive results.Geometric mean titre (GMT) and GMT 95% confidence intervals of EV71 neutralizing antibodies in different age groups of children.GMT and GMT 95% confidence intervals of EV71 neutralizing antibodies in different genders of children.

Med-MDPI/ijerph_2/ijerph-14-03-00320.txt

@@ -0,0 +1 @@
+Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).In the absence of pipe-borne water, many people in Africa, especially in rural communities, depend on alternative water sources such as wells, boreholes and rivers for household and personal hygiene. Poor maintenance and nearby pit latrines, however, lead to microbial pollution of these sources. We evaluated the abundance of Escherichia coli and the prevalence of pathogenic E. coli virulence genes in water from wells, boreholes and a river in a South African peri-urban community. Monthly samples were collected between August 2015 and November 2016. In all, 144 water samples were analysed for E. coli using the Colilert 18 system. Virulence genes (eagg, eaeA, stx1, stx2, flichH7, ST, ipaH, ibeA) were investigated using real-time polymerase chain reaction. Mean E. coli counts ranged between 0 and 443.1 Most Probable Number (MPN)/100 mL of water sample. Overall, 99.3% of samples were positive for at least one virulence gene studied, with flicH7 being the most detected gene (81/140; 57.6%) and the stx2 gene the least detected gene (8/140; 5.7%). Both intestinal and extraintestinal pathogenic E. coli genes were detected. The detection of virulence genes in these water sources suggests the presence of potentially pathogenic E. coli strains and is a public health concern.Although the Millennium Development Goals’ (MDGs) target for drinking water supply was globally exceeded, this was not the case in the African region [1]. Many people in Africa, especially in rural communities, still depend on alternative water sources such as rivers [2] and boreholes [3] for household and personal hygiene. Unlike the exceeded target for access to safe drinking water, access to improved sanitation was not met globally, and in the African region, only 7% of the targeted 50% reduction in people without access to improved sanitation was achieved [1]. The lack of sanitation facilities results in the uncontrolled disposal of household and human waste into surrounding water bodies, leading to pollution [4]. In some instances, poorly constructed pit latrines have been used for human waste disposal. However, the construction of these pit latrines uphill from, and in close proximity to, water sources also leads to the pollution of these water sources through leaching [5]. These pollutants could include chemicals such as pharmaceuticals [6] and pathogenic microorganisms [7]. These microorganisms which include bacteria, viruses and parasites, have been found to cause numerous waterborne diseases in humans that use untreated water from such polluted sources [8].Escherichia coli has been extensively studied as an indicator of faecal pollution in water resources, owing to the fact that the organism has been found to colonise the gut of many warm-blooded animals [9]. Though originally considered a commensal, the organism has evolved to include pathogenic strains, causing both intestinal and extraintestinal disease in humans [10,11]. Strains that have been found to cause gastrointestinal illnesses include the enterohaemorrhagic E. coli (EHEC), enterotoxigenic E. coli (ETEC), enteropathogenic E. coli (EPEC), enteroinvasive E. coli (EIEC) enteroaggresive E. coli (EAEC) and diffusely adherent E. coli (DAEC) [12] and they are collectively known as diarrhoeagenic E. coli. Those that have been reported to cause extraintestinal illnesses include neonatal meningitis E. coli (NMEC) and uropathogenic E. coli (UPEC) [13]. Their ability to cause infections in humans lies in a number of virulence traits which the pathogenic forms possess. The virulence genes associated with these intestinal and extraintestinal E. coli pathotypes and how they cause disease, have been reviewed previously [13,14].Groundwater has been recognised as the most abundant and most important source of portable water for human uses around the globe [3]. In Africa in particular, and the world in general, groundwater is considered a relatively safe and cost-effective water source compared to other water sources, especially in those regions where surface water is not readily available [15]. As such, this natural resource is facing constant decrease in quantity and deterioration in quality as a result of both natural and anthropogenic activities, a challenge that is anticipated to get even worse in the 21st century and beyond [16]. Although the limited access to clean portable water has forced many households in South Africa (and many African countries as a whole) to consider alternative sources such as rivers and streams, these sources are usually of poor microbial quality [17,18,19,20]. As such, many other communities have resorted to boreholes and wells to meet their daily water needs. The World Health Organization (WHO) recommends that to ensure good protection of such groundwater sources, they should not be constructed near to or downhill of possible contamination sources such as pit latrines and runoff from animal manure in agricultural areas [8]. However, these recommendations have often been neglected in most communities using groundwater sources, resulting in the pollution of boreholes with chemical and/or pathogenic microorganisms [5]. These pollutants could have adverse health impacts on the lives of people using such polluted water for personal and household hygiene. Currently, there are no clear polices and legislature also governing the use and quality of groundwater in South Africa. While the South African Department of Water and Sanitation is currently drafting a national groundwater strategy [21], there is a need to understand the quality of these water sources so as to ensure protection of the users of the water. Therefore, the current study was carried out to evaluate the microbial quality of wells and boreholes in Stinkwater, a peri-urban community of South Africa, using E. coli as an indicator organism. More importantly, the study also sought to determine the prevalence of pathogenic E. coli virulence-associated genes in these water sources so as to infer any possibility of infection from the consumption of untreated water from these water sources.The Stinkwater community lies in the northern part of the Gauteng Province of South Africa, close to the border with the Northwest Province. Presently, the Stinkwater community consist of eight extensions comprising a total of 8250 households spread over an estimated surface area of 27.92 km2 [22]. Although the main portions of the Stinkwater community consist of subsidised housing, there also exist informal settlements in the area, especially in the outer boundaries of the community [23]. A remarkable characteristic of the Stinkwater community is the lack of access to piped water [24]. As such, the inhabitants are dependent on water delivery from roaming municipal water tankers. As municipal water deliveries are often sporadic, many residents habitually rely on water obtained from hand-dug wells or boreholes in their yards or other communal areas.A total of 14 wells (W1–W14), four boreholes (BH1–BH4) and two river sites (R1 and R2) were chosen for the purpose of the current study (Figure 1).Water samples were collected once a month for a period of 12 months between August 2015 and November 2016 (samples were not collected during the months of December 2015, and January, March and April of 2016). Samples were collected following the South African sampling guide for groundwater [25], transported to the laboratory in cooler boxes containing ice packs and then analysed within 2 h upon arrival at the laboratory. Samples were analysed using the Colilert 18/Quanti-Tray 2000 system (IDEXX Laboratories Inc., Pretoria, South Africa) as previously described [26]. E. coli counts were recorded as the Most Probable Number (MPN) per 100 mL (MPN/100 mL) of water sample.Following 18–24 h incubation of the Quanti-Tray 2000 plates as recommended by the manufacturer, 1 mL of the content of a fluorescent cell on the Quanti-Tray 2000 plates was transferred into 1.5 mL microcentrifuge tubes and DNA was extracted as previously described [27]. The extracted DNA was then used in separate real-time PCR assays for the identification of virulence genes pertaining to different E. coli pathotypes. The primers used in each of the PCR assays are given in Table 1.Sets 2, 3 and 5 PCR were run in multiplex assays while sets 1 and 4 were run in singleplex real-time PCR assays. The multiplex assays were run in total volumes of 20 µL each, while the singleplex assays were run in total volumes of 10 µL each. Assay 2 consisted of 10 µL of 2× SensiFAST High-Resolution Melt (HRM) mix (SF) (Bioline GmbH, Luckenwalde, Germany), at a final concentration of 1×, 0.75 μL each of forward and reverse primers (eaeA and eagg) at a final concentration of 0.75 μM, 0.2 μL of ipaH forward and reverse (final concentration 0.75 μM), 1.6 μL of nuclease free water (NFW) and finally 1 μL of deoxynucleotide (dNTP) Mix (Thermo Fisher Scientific, Edenvale, South Africa) was added to the final overall mixture at a final concentration of 400 μM. Assay 3 consisted of 10 µL SF, 1.5 μL stx1 (final concentration of 1.5 μM), 0.2 µL stx2 (final concentration of 0.2 μM), 1.6 μL NFW and 1 μL dNTP Mix. The reaction mixture for Assay 5 was similar to that of Assay 3 with volumes and concentrations of ST and ibeA being similar to those of stx1 and stx2 respectively. Unlike Assays 2, 3 and 5, Assays 1 and 4 were run in a total volume of 10 μL. Assay 1 consisted of 5 µL SF, 0.5 μL mdh (forward and reverse; final concentration of 0.5 μM) and 1 μL NFW. The reaction mixture for Assay 4 included 5 µL SF, 0.75 μL stx2 (forward and reverse; final concentration of 0.75 μM) and 1 μL NFW. For Assays 2, 3 and 5, 5 μL of extracted sample DNA was added while 3 μL of sample DNA was added for Assays 1 and 4. A positive control containing DNA from known E. coli strains was included in each of the respective PCR runs. No-template controls, consisting of the respective reaction mixes with NFW (and void of any DNA) added to make up the desired volume, were also included in each assay. All positive controls and PCR cycling conditions were as previously described [27] with a modification of the number of cycles to 45 for all assays.The Spearman’s rank correlation was computed to check for any correlation between the abundance of E. coli and the overall prevalence of the virulence genes. Data analysis was performed using SPSS 20 (Statistical Package for the Social Sciences; IBM Corporation, Armonk, NY, USA). Before analysis, E. coli data was log transformed and statistical tests were considered significant at a 95% confidence limit.A total of 144 samples was collected between August 2015 and November 2016 from the selected 20 sampling sites. Not all the sites were sampled equally as they could not always be accessed routinely due to the unavailability of the owners. The mean MPN of E. coli per 100 mL of sample at each site is shown in Table 2. To obtain the mean, all values of <1 were considered as zero while all values of >2419.6 were converted to the nearest whole number (2420).Due to the vast diversity of microorganisms that can be found in a polluted water source at any given time, it would be extremely laborious and financially demanding to do a complete assessment of the microbial quality of such water. As such, the WHO recommends the use of indicator organisms, of which E. coli is the most widely used, for the evaluation of the microbial quality of water [8]. Based on this, the WHO recommends that no E. coli be found in any water meant for human consumption [8]. This same limit is adopted by the South African water management authorities [32].Considering these limits, only one site (BH1) met the WHO and South African guidelines for drinking water quality. However, this cannot be conclusive given that only a single sampling round was analysed during the entire study because of the inaccessibility of the site. Similarly, it cannot be concluded that site W9, BH2 and BH3 were unsafe for consumption given that they were all sampled only once. The high E. coli counts observed in samples from these sites could have been the result of a single pollution event and as such, only further sampling of these sites could give a conclusive quality of their water. Nevertheless, consumption of water from these sites on the sampling day could still represent a health risk for household members, especially children and immunocompromised individuals.The river samples recorded the highest E. coli counts during all the sampling days for the entire sampling period (Table 2). This is an indication that the water is constantly polluted. Pollution of this water source could arise from the informal settlements around the neighbouring Soshanguve community through which the river flows, or from the small cattle farms that are present along the banks of the river on the Stinkwater side. Most informal settlements are characterised by the lack of basic sanitation facilities and the inhabitants of such settlements mostly resort to rivers and nearby bushes as their main points of waste disposal. The impact of animal farming [33,34] and informal settlements on the microbial quality of aquatic ecosystems has been previously reported [4,35,36,37,38].All the wells sampled in this study recorded mean E. coli counts above the WHO and the South African recommended limits for drinking water. Most of these wells are shallow, and poorly protected. Rudimentary materials such as zinc or aluminium sheets are typically used to cover the wells and water is extracted with a rope or chain attached to a bucket. These buckets are usually kept under unhygienic conditions on the ground near the well, especially in the communal wells. For example, W1 which recorded the highest mean E. coli count of 443.1 MPN/100 mL during the entire study is a communal well located in a public space. Although not used for drinking as indicated by the inhabitants, water from this well is used to wash other household items and motor vehicles. Activities such as motor vehicle washing near water sources has been reported to have negative impacts on water quality [39]. As such, seepage from the nearby pit latrines and the car washing activities around the well could have accounted for the high E. coli counts recorded at this site. Similarly, W3 with a mean E. coli count of 415.0 MPN/100 mL is located at close proximity to a pit latrine and is used for drinking and other household purposes. Therefore, the poor maintenance of the well, poor hygienic conditions associated with the water extraction process and the close proximity of the well to a pit latrine are factors that could account for the high mean E. coli count observed at this site. Such poor hygienic practices have been reported to lead to contamination of groundwater sources [40].A total of 141 fluorescent Quanti-Tray 2000 cells were selected for the identification of the E. coli virulence genes. The melt curve analyses of the optimised PCR assays are shown in Figure S1 (Supplementary Materials). Prior to analysing for the virulence genes, all samples were first tested for the presence of the malate dehydrogenase (mdh) gene which is a house-keeping gene common to all E. coli strains [30]. This was to ensure that all the fluorescence observed was due to the presence of E. coli. Of the 141 fluorescent cells analysed, only one cell was negative for the mdh gene. This cell was subsequently negative for all other virulence genes tested. Although false positive results have previously been reported with the use of the Colilert system for the identification of E. coli, only a single sample (1/141; 0.7%) analysed in the current study was falsely positive compared to the 7.4% to 36.4% reported in other studies [20].The various virulence genes investigated in the current study were not evenly distributed between the sites (Table 3). As with the abundance of E. coli, the Stinkwater River sites recorded the highest number of samples positive for the virulence genes. These sites (R1 and R2) were also the only sites that harboured all the virulence genes (as studied). Virulence genes were not identified in the BH1 sample. Apart from R1 and R2, only W5 was positive for the stx2 gene of EHEC.The overall prevalence of each virulence gene investigated in the current study is shown in Figure 2. Almost all samples (139/140; 99.3%) were positive for at least one of the pathogenic E. coli virulence genes investigated in the current study.There was a positive correlation between the abundance of E. coli and the prevalence of the virulence genes (p = 0.000; p < 0.05) with a correlation coefficient (rs) of 0.889. This indicates that more virulence genes were identified as the number of E. coli increased. The results of the current study contradict the findings of Shelton et al. [41]. In their study, however, Shelton et al., investigated the correlation between the abundance of E. coli in a watershed and the presence of only two virulence genes (the intimin (eae) and shiga toxin production (stx) genes). Like the findings of Shelton et al., but contrary to our findings, Sidhu et al. also reported a lack of correlation between the abundance of E. coli and the presence of virulence genes in some sub-tropical surface waters in Brisbane, Australia [42]. Contrary to the approach used in our study, Sidhu et al. worked on pure isolates and this could account for the difference in the correlation between the two studies.Members of the EHEC group have been globally implicated in numerous bloody diarrhoea outbreaks and haemolytic uremic syndrome [43]. Virulence in these strains is characterised by the presence of the shiga toxin producing genes (stx1 and stx2), the intimin (eaeA) gene and other factors such as the flicH7 gene which codes for the structural flagella antigen in EHEC O157:H7 [44,45]. Strain O157:H7 has been reported to be the most common member of the EHEC group and has been involved in several diarrhoeal disease outbreaks in many developed countries such as the UK, USA, Ireland and Canada [46] as well as many developing countries, including South Africa [47]. In the United States for example, the organism causes an estimated 73,000 cases of illnesses resulting in about 60 deaths every year [48]. The eaeA gene is also found in EPEC strains which are mostly responsible for infantile diarrhoea-associated mortality in developing countries and were the first E. coli to be associated with human infections [49]. Although diarrhoea caused by EPEC members are usually self-limiting and can easily be managed through oral-rehydration, the onset of the disease is very short, usually 2.9 h following ingestion [50]; complications may arise if the disease is poorly handled, especially in immunocompromised individuals and children. In the current study, all the above genes were detected, with the flicH7 gene being the most isolated gene (81/140; 57.6%) while the stx2 gene was the least detected gene (8/140; 5.7%). This indicates that the various water sources in the current study are not only polluted, but also harbour pathogenic forms of E. coli with the possibility of adverse health effects to those using the water for drinking and other household purposes without prior treatment. Also, the stx2 has been isolated in both healthy and diseased cows in South Africa [29], indicating that the high presence of this gene in the current study could have originated from both animal and human sources. The genes responsible for virulence in EIEC (ipaH; 18.6%), EAEC (eagg; 40.7%) and ETEC (ST; 24.3%) were also detected. A study conducted on different surface water bodies in Iran reported that 97% of E. coli that was isolated carried the ipaH gene [51]. The presence of the invasive plasmid antigen H (ipaH) gene in any water body meant for human consumption represents a health risk. Given that the ipaH gene (which is also present in Shigella spp.) is carried in a plasmid, the gene could easily be transmitted to non-pathogenic E. coli strains and other bacterial species, thus transferring the invasive characteristics to the new organisms [51,52]. In a study conducted by Sansonetti et al. [53], the authors reported that loss of the invasive plasmid resulted in the loss of the virulent invasive potential in Shigella spp., while the reintroduction of the plasmid into avirulent strains led to the reestablishment of the invasive potentials.Apart from the genes coding for virulence in the diarrhoeagenic E. coli, the gene responsible for virulence in NMEC, an extraintestinal E. coli, was also detected. The ibeA (invasive brain epithelium A) gene codes for the ability of NMEC strains to invade the meninges in infants, thereby causing meningitis, 10%–30% of which may lead to death [54]. Although these strains are mostly transmitted from mother to child during birth, environmental transmission has also been reported whereby after faecal–oral acquisition, the organism succeeds in crossing the mucosal barrier and then spreads through blood to other organs such as the brain [55,56]. As such, the presence of the ibeA gene (51/140; 36.4) in the water from the wells studied indicates that the water contains NMEC strains and is not suitable for consumption, especially in infants. Furthermore, this study adds to the growing evidence that there is a need for clear policies, which can be filtered down to communities, on best practices to protect groundwater sources and to treat and use groundwater. It should be noted, however, that the results presented in the current study are for the prevalence of virulence genes in whole samples and should not be regarded as the abundance of specific E. coli pathotypes, given that pure isolates were not obtained prior to detection of the virulence genes.The water from the wells and boreholes in the Stinkwater community is not microbiologically safe for human consumption. The high number of samples positive for pathogenic E. coli virulence genes indicates that the groundwater is not only faecally polluted, but could also be harbouring pathogenic E. coli strains with the potential to cause infection, especially in children and immunocompromised individuals. It is therefore highly recommended that in the absence of clean pipe-borne water, Stinkwater community members be advised to treat the water from these sources before consumption or use in the house.The following is available online at www.mdpi.com/1660-4601/14/3/320/s1, Figure S1: High-resolution melt curves (HRM) for each PCR assay.The study was funded by Parliamentary Grant No.: 0000006070 from the Council for Scientific and Industrial Research (CSIR), Pretoria, South Africa.Akebe Luther King Abia designed the study, performed the molecular experiment and drafted the manuscript. Lisa Schaefer performed field work and bacteriological analysis. Eunice Ubomba-Jaswa supervised the molecular work and contributed in drafting the manuscript. Wouter Le Roux was the project leader, performed the sample collection, and supervised the study. All authors read and approved the final manuscript.The authors declare no conflict of interest.Map of the study area showing sampling points (Source: Google earth). W: well; BH: borehole; R: river (sample collection).Overall prevalence of the various E. coli virulence genes detected during the entire study period.Primer sequences used for the identification of E. coli virulence-associated genes.Mean E. coli count (Most Probable Number per 100 mL; MPN/100 mL) per sampling site.* Values represent results of a single sampling round and should not be regarded as means.Number of samples positive for each gene per sample site.

Med-MDPI/ijerph_2/ijerph-14-03-00321.txt

@@ -0,0 +1 @@
+Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).This study systematically reviewed forest therapy programs designed to decrease the level of depression among adults and assessed the methodological rigor and scientific evidence quality of existing research studies to guide future studies. This systematic review was conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines. The authors independently screened full-text articles from various databases using the following criteria: (1) intervention studies assessing the effects of forest therapy on depressive symptoms in adults aged 18 years and older; (2) studies including at least one control group or condition; (3) peer-reviewed studies; and (4) been published either in English or Korean before July 2016. The Scottish Intercollegiate Guideline Network measurement tool was used to assess the risk of bias in each trial. In the final sample, 28 articles (English: 13, Korean: 15) were included in the systematic review. We concluded that forest therapy is an emerging and effective intervention for decreasing adults’ depression levels. However, the included studies lacked methodological rigor. Future studies assessing the long-term effect of forest therapy on depression using rigorous study designs are needed.Forest therapy or “forest bathing” refers to visiting a forest or engaging in various therapeutic activities in a forest environment to improve one’s health and wellbeing [1,2]. Societies have been urbanizing rapidly and more people reside in an urban environment with limited access to nature; therefore, diverse efforts including political and landscaping efforts have been made to make nature more accessible [3]. With an increasing awareness of health benefits of forest therapy, it has been implemented on diverse population [1]. Particularly, the psychological benefits of forest therapy have received special attention as people residing in urban environments have been reported to be at an increased risk of prolonged exposure to stressful situations and mental health problems [4,5,6]. Compared to control groups, forest therapy significantly improves adults’ mental health by decreasing stress, depression, anxiety, and anger levels [7].A systematic review summarizes the results of the available research studies and provides synthesized evidence on the effectiveness of those studies [8]. It enables researchers to identify the current state of the science, areas for future researchers to improve upon, and provides strong evidence for up-to-date practices and policy developments [9]. It is also beneficial for emerging topics that require systematic evaluation and synthesis of the evidence quality (e.g., feasibility and effectiveness of intervention) as well as well-established areas of research with accumulated scientific evidence that need be updated regularly.Despite the increased attention to the various health benefits of forest therapy, until now, systematic reviews of the body of evidence for the effectiveness of forest therapy on mental health have not been conducted. A clearer and comprehensive understanding of the effectiveness of forest therapy on mental health is important for further refinement of forest therapy programs. Among the several mental health outcomes included in the forest therapy research, our paper will focus on depression. Depression is the leading cause of disability; approximately 350 million (5% of the world’s population) suffer from this debilitating disorder [10]. The specific aims of this study were to: (1) provide a broad overview and synthesize the evidence on the usefulness of forest therapy to improve the level of depressive symptoms in adults; and (2) assess the methodological rigor and scientific evidence quality of existing research studies to guide future studies evaluating the effects of forest therapy on adults’ experiencing depressive symptoms. In the present review, forest therapy was defined as visiting a forest or engaging in various therapeutic activities in a forest environment to improve one’s health and wellbeing [1,2].This systematic review was conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines. We searched PubMed, EMBASE, Cumulative Index to Nursing and Allied Health, PsycARTICLES, Korean Studies Information Service System, Research Information Sharing Service, and DBpia to identify relevant studies published until July 2016. The search terms were chosen from the USNLM Institutes of Health list of Medical Subject Headings for 2015. Search terms included “trees”, “forests”, “wood”, “affect”, “depression”, “emotions” and “depressive disorder”. Search terms used to identify relevant studies for the review are listed in Appendix A, Table A1.The initial eligibility assessment was conducted by one author by reviewing the title and abstracts. Then, two authors (MinKyung Song and Buhyun Lee) independently screened the full text versions of 66 articles using the following criteria: (1) intervention studies assessing the effects of forest therapy on depression among adults aged 18 years and older; (2) studies including at least one control group or condition; (3) peer-reviewed studies; and (4) been published either in English or Korean.The four authors (Heeseung Choi, Kyung-Sook Bang, MinKyung Song, and Buhyun Lee) independently performed the data extraction. The following data were extracted from each study: first author, date and place of publication, study design, sample size, setting, ethical consideration, participants’ characteristics, number of participants enrolled, summary of the intervention and control conditions, measures, reported outcomes, and risk of bias. The extracted data were input into standardized MS word (Microsoft Corporation, Seattle, WA, USA) files. Any disagreements were resolved by discussion between the authors.The Scottish Intercollegiate Guideline Network (SIGN) measurement tool (Healthcare Improvement Scotland, Edinburgh, Scotland) was used to assess the risk of bias in each study included in this review. The SIGN was developed in 1993 to improve the quality of health care for patients in Scotland by reducing the variation in practice and outcome, through the development and dissemination of national clinical guidelines containing recommendations for effective practice based on current evidence [11]. Using the SIGN, we evaluated the internal validity and risk of bias of the study and assigned values of “high quality (++),” “acceptable (+),” “low quality (−),” or “unacceptable—reject (0)” to each study. The risk of bias was evaluated independently by four reviewers (Heeseung Choi, Kyung-Sook Bang, MinKyung Song, and Buhyun Lee) and any disagreements were resolved through a consensus process.Records (N = 8355), including 4399 records published in English and 3956 records published in Korean, were retrieved from the initial database searches. These search results were imported using EndNote X7 and 1516 duplicates (1356 English articles and 160 Korean articles) were removed. A detailed flow diagram of the screening process is shown in Figure 1. After excluding an additional 6773 records based on the review of the study titles and abstracts, the remaining 21 articles published in English and 45 articles published in Korean were assessed for eligibility. Many articles were excluded because those studies addressed topics that were not relevant to forest therapy, such as tree analysis (i.e., classification and regression tree analysis), biliary/bronchial tree, forest modeling, and forest fragmentation. Finally, 28 articles (English: 13, Korean: 15) were included in the present systematic review.The general characteristics of the studies included in this review are summarized in Table 1. And summary of the studies included in this review are presented in Table 2 and Table 3. The selected studies were published between 1996 and 2016, and 24 of the 28 studies were published within the last five years. All the studies were conducted in Asian countries (Korea, Japan, and China) except one, which was conducted in the United Kingdom. Sixteen studies were conducted on healthy adults and the rest of the studies (n = 12) were conducted on adults with various health problems such as hypertension, cancer, and mental disorders. Among the 12 studies conducted for adults with health problems, six studies targeted psychiatric patients [12,13,14,15,16,17]; however, only one study [17] was conducted with patients with major depressive disorder. Two studies [12,15] were conducted for psychiatric patients with various diagnoses such as substance use disroders, schizophrenia, other psychotic disorders, mood disorders, and anxiety disorders. These two studies, however, did not specify what percentage of their samples were patients with major depressive disorder. Other studies targeted specific diagnoses such as Hwa-Byung [13], neurocognitive disorders [14], and alcoholism [16].Regarding the study design, 11 studies used a crossover trial design and only six of the studies [16,18,19,20,21,22] used a randomized controlled trial (RCT) design. Four out of six RCT design studies were conducted with adults with health problems. The most common types of control condition used in the non-equivalent control group design studies were “normal daily routines” for healthy adults (five out of six studies) and “treatment-as-usual” for adults with health problems (three out of five studies).The sample size ranged from 11 [23] to 92 [16]; for almost 43% of the studies, the sample size was less than 20. Furthermore, about one third of the studies (28.5%) did not follow the ethical protocol such as being reviewed and approved by the Institutional Review Board (IRB).The forest therapy programs tested in these 28 studies varied in terms of format and content of the programs. The length of time that the interventions were undertaken ranged from one day to twelve weeks. The duration of the forest therapy ranged from twelve minutes [29] to three hours [33]. About one third of the studies [23,26,27,28,29,32,34,39] offered forest therapy programs and control conditions (such as activities in downtown) every other day during the two-day periods. Three studies [12,21,25] used a one-time intervention that lasted a few hours to half a day. One study did not report duration details of the intervention [36].Regarding the content of forest therapy, walking in the forest was the key component of the forest therapy that was included in most studies except one [28]. Other therapeutic activities included in forest therapy programs were experiencing forest through the five senses (seeing, hearing, touching, smelling, and tasting), forest viewing, forest meditation, Qi-Qong, aromatherapy, herbal tea therapy, and craftwork using natural materials.The most commonly used self-report measure for depression in these 28 studies was the Profile of Mood States (POMS). For articles published in English, nine studies [12,19,21,22,25,28,29,34,39] used POMS to assess the level of depression and three studies [28,29,39] used Semantic differential (SD) method. Other scales used by the studies included the Hamilton Rating Scales for Depression [20], Beck Depression Inventory (BDI) [16,20,24,31], positive and negative affect schedule (PANAS) [34]. For studies published in Korean, the POMS [15,23,26,27,32,38] and BDI [13,14,15,17,18,35] were the most commonly used scale. Other scales used to measure depression were the Hospital Anxiety and Depression Scale (HADS) [38], Hamilton Rating Scales for Depression [17], Montgomery-Asberg Depression Rating Scales [17], Symptom Check List (SCL-90-R) [32], and Zung Self-Rating Depression Scale (SDS) [36]. Fourteen studies [13,15,17,18,19,21,22,23,24,25,26,28,29,39] used both self-report, paper-and-pencil questionnaires, and physiological measures, while 14 studies [12,14,16,20,27,30,31,32,33,34,35,36,37,38] used only self-report, paper-and-pencil questionnaires. Physiological or objective measures included heart rate variability (HRV), blood pressure, heart rate, and amylase concentration. Detailed information of the measures included in these studies is summarized in Table 2 and Table 3.All 28 studies assessed the level of depression before and after the intervention; however, no study conducted additional long-term follow-up assessments. Regarding the changes in the level of depression, 21 studies showed significant improvement in depression, whereas seven studies reported no significant changes in depression compared to the control group [12,15,18,27,28,29,33]. The studies that failed to demonstrate a significant improvement in the level of depression were the ones that targeted only healthy adults and the ones that conducted “viewing or walking in the forest” activities only for the intervention group.Regarding the differential pattern of findings associated with research design, while 8 out of 11 crossover trials and 8 out of 11 non-equivalent control group design studies reported significant improvement in depression scores, five out of six RCTs reported significant results. However, the differential patterns of findings could be partly attributable to the sample characteristics of the studies; four out of six RCTs were conducted with adults with health problems.Based on the SIGN checklist, ten out of the twenty-eight studies [16,18,19,20,21,23,25,30,34,37] met an “acceptable quality” rating and the rest eighteen studies were rated as low quality [12,13,14,15,17,22,24,26,27,28,29,31,32,33,35,36,38,39]. Six studies [16,18,19,20,21,22] used random allocations; however, no detailed description of the procedure was provided except two studies [18,20]. Five out of six RCT studies included in this review met the criteria for the “acceptable quality” [16,18,19,20,21], meeting all the items in the checklist except the criterion of blinding the treatment allocation to participants/researchers. One of the RCTs was rated “low” in terms of quality because the homogeneity between the experimental group and the control group at baseline was not ensured and the significant differences between the two groups had not been adequately addressed [22]. Three out of eleven crossover trials [23,25,34] were rated “acceptable” in terms of quality because they had low drop-out rates and the only difference between the experimental and control groups was the treatment under investigation. The main reasons for the “low quality” ratings were inadequate random allocation or method of concealment used. Among 11 categories in the SIGN checklist, studies were rated to have “high risk of bias” particularly for three categories: “the assignment of participants to treatment groups is randomized,” “an adequate concealment method is used,” and “the design keeps participants and investigators ‘blind’ about treatment allocation.” Please see Figure 2 for risk of bias graph.As a result of the extensive literature review, we could identify 28 studies meeting the criteria for the present review. All the studies were data-based, intervention studies with at least one comparison group. Moreover, most studies (24 out of 28 studies) were published within the last five years, confirming that forest therapy is one of the emerging therapeutic approaches and it has been gaining popularity. An analysis of the 227 regional healthcare program plans proposed in Korea between 2011 and 2014 also revealed that 35 healthcare programs were utilizing forest resources [40]. These findings demonstrated that forest therapy is a fast-growing treatment approach used in the community.All 28 studies varied in terms of their sample characteristics and intervention types such as format, content, and study settings. Regardless of the wide variations, in general, the studies demonstrated that forest therapy is effective in improving depression, particularly for adults with health problems. However, programs that targeted only healthy adults and the ones that used “viewing or walking in the forest” activities as the only main intervention were not effective in improving depression. Per Stigsdotter et al. forest therapy is classified into three different levels of contact: “viewing nature,” “being in the presence of nearby nature,” and “active participation and involvement with nature [41].” All the activities had a certain amount of health benefits [3]; however, this review revealed that “viewing nature” or “being present near nature” may not be enough to have a significant impact on the level of depression. Therefore, future studies testing the effects of forest therapy need to include a higher level or dosage of therapeutic component of forest therapy. Another possible reason for the not-significant effect of forest therapy on improving depression in healthy adults is ceiling effect [42]. For future studies, well-thought-out intervention contents and careful selection of the outcome measures and target population are recommended. It is also important to develop structured and theory-based forest therapy programs based on the scientific evidence for the specific health benefits of forests.Despite the increasing number of studies testing the effects of the forest therapy, these published studies are still lacking methodological rigor, mainly due to a small sample size and not having an RCT design. The majority of the studies used either non-equivalent control group design or crossover design. Crossover design, or within-subjects design, increases statistical power and enables researchers to test the effect of the intervention with relatively small samples compared to studies using between-subject design. In addition, the internal validity of crossover designs is not influenced by random assignment or between-subject variation [43,44]. On the other hand, crossover design has several limitations. For crossover design, carryover effects, the treatment effect that is carried over from one experimental period to the next experimental period, needs to be carefully examined. In addition, dropout or missing data could be the significant problem because each participant serves as both the intervention and control group; therefore, the amount of contribution made by one participant is relatively large [43,44,45,46]. However, carryover effects inherited in the crossover study design were not properly addressed in the reviewed studies. Only two crossover trials conducted by the same investigator mentioned washout periods. In addition, only 4 out of 28 studies mentioned dropout rates. Overall, issues associated with dropout or missing data were not discussed in the reviewed studies.A second issue related to methodological rigor is the inadequacy of the control group/condition employed in the non-equivalent control group design studies. The majority of the non-equivalent control group studies used “usual care” for the control group and did not properly address factors that may threaten the findings’ validity such as the Rosenthal effect. In the future, well-designed studies with structurally equivalent control groups are needed to improve the quality.Another shortcoming of these studies was the lack of reliable measures for assessing the level of depression. About half of these studies used self-report, paper-and-pencil-based questionnaires that only assessed the level of depression. Since the significant correlations between the physiological findings (e.g., electroencephalogram asymmetry) and the level of perceived depression has received attention [47,48], scientists have begun to use various physiological measures to assess depression in addition to self-report questionnaires. Heart rate variability (HRV) was one of the commonly used measures in the studies included in the present paper. HRV is a physiological marker that reflects the functioning of the sympathetic and parasympathetic nervous system and is also a well-established indicator of stress and depression [49,50]. A significantly reduction in HRV has been observed among patients with depression compared to the healthy adults [51,52].Other physiological measures that have been used to assess the level of stress and psychological conditions, including depression, natural killer cell activity [53], salivary amylase activity [54,55], salivary and serum cortisol, immunoglobulin A (IgA) concentrations [56], and urinary adrenaline levels [57]. In addition, electroencephalogram-based biomarkers (i.e., rACC theta, LDAEP, iAPF, P300, frontal theta activity) were found to predict the prognosis of the course of mental illness and treatment response [58]. Therefore, future studies examining the effects of forest therapy on depression need to use well-established and reliable physiological measures in addition to self-reported questionnaires to capture the full picture of the therapeutic effects of forest therapy.Lastly, sample characteristics of the reviewed studies deserve mention. The majority of the reviewed studies targeted healthy adult participants; only three studies tested the effects of forest therapy on adults diagnosed with major depressive disorder. Therefore, the extent that the results are applicable to clinical depression is still uncertain. More studies with clinical samples are needed to establish evidence of the therapeutic value of forest therapy. Furthermore, longitudinal studies testing the long-term effects of forest therapy on depression and the changes in depressive symptoms over a span of time are needed.A limitation of this systematic review is language bias since we only included studies that were published in English and Korean. Studies published in other languages, such as Chinese and Japanese, were not included in this review. Despite this limitation, this study increased the understanding of the therapeutic benefits of forest therapy and identified gaps in the literature.This review demonstrated that forest therapy is an emerging and effective intervention for decreasing adults’ depressive symptoms. However, the studies included in this review lacked methodological rigor. Future studies assessing the long-term effects of forest therapy on depression using rigorous study designs are needed.This work has been supported by a research grant from the Korea Forest Service (No. S211214L010140).Insook Lee, Sungjae Kim, Kyung-Sook Bang, and Heeseung Choi conceived and designed the study; Heeseung Choi, Kyung-Sook Bang, MinKyung Song and Buhyun Lee conducted the systematic review and analyzed the data; Heeseung Choi, MinKyung Song, and Buhyun Lee wrote the paper; and Insook Lee, Heeseung Choi, Kyung-Sook Bang, and Sungjae Kim reviewed and refined the paper.The authors declare no conflict of interest.Search Terms Used to Identify Relevant Studies for the Review.Trees */Tree/Forests/Forest/Forest Areas/Area, Forested/Areas, Forested/Forested Area/Woodland/Woodlands/Forestlands/Forestland/Wood/Woods/Shinrinyoku/Green exercise/1 OR 2 OR 3 OR 4 OR 5 OR 6 OR 7 OR 8 OR 9 OR 10 OR 11 OR 12 OR 13 OR 14 OR 15 OR 16 OR 17Affect/Affects/Mood/Moods/Depression/Depressions/Depressive Symptoms/Depressive Symptom/Symptom, Depressive/Symptoms, Depressive/Emotional Depression/Depression, Emotional/Depressions, Emotional/Emotional Depressions/Emotions/Emotion/Regret/Regrets/Feelings/Feeling/Depressive Disorder/Depressive Disorders/Disorder, Depressive/Disorders, Depressive/Neurosis, Depressive/Depressive Neuroses/Depressive Neurosis/Neuroses, Depressive/Depression, Endogenous/Depressions, Endogenous/Endogenous Depression/Endogenous Depressions/Depressive Syndrome/Depressive Syndromes/Syndrome, Depressive/Syndromes, Depressive/Depression, Neurotic/Depressions, Neurotic/Neurotic Depression/Neurotic Depressions/Melancholia/Melancholias/Unipolar Depression/Depression, Unipolar/Depressions, Unipolar/Unipolar Depressions/Sadness18 OR 19 OR 20 OR 21 OR 22 OR 23 OR 24 OR 25 OR 26 OR 27 OR 28 OR 29 OR 30 OR 31 OR 32 OR 33 OR 34 OR 35 OR 36 OR 37 OR 38 OR 39 OR 40 OR 41 OR 42 OR 43 OR 44 OR 45 OR 46 OR 47 OR 48 OR 49 OR 50 OR 51 OR 52 OR 53 OR 54 OR 55 OR 56 OR 57 OR 58 OR 59 OR 60 OR 61 OR 62 OR 63 OR 6417 AND 65Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) Flow Diagram of the Screening Process.Risk of Bias Graph. Note: Authors’ judgments about each risk of bias item presented as percentages across all included studies. 1.1 The study addresses an appropriate and clearly focused question; 1.2 The assignment of participants to treatment groups is randomized; 1.3 An adequate concealment method is used; 1.4 The design keeps participants and investigators “blind” about treatment allocation; 1.5 The treatment and control groups are similar at the start of the trial; 1.6 The only difference between groups is the treatment under investigation; 1.7 All relevant outcomes are measured in a standard, valid, and reliable way; 1.8 What percentage of the individuals or clusters recruited into each treatment arm of the study; dropped out before the study was completed? (<20% = low risk of bias); 1.9 All the participants are analyzed in the groups that they were randomly allocated (often referred to as intention to treat analysis); 1.10 Where the study is carried out at more than one site, results are comparable for all sites; 2.1 How well was the study done to minimize bias?General Characteristics of Included Studies (N = 28).Summary of Included Studies for Healthy Adults (N = 16).Note: Exp.: Experimental group, Cont.: Control group; * Significant finding; 2-day forest therapy camp; † consisted of walking, therapeutic activities, psychoeducation for coping with pain and stress, bodily exercises and mindfulness-based meditation in the forest and indoor music therapy.Summary of Studies for Adults with Health Problems (N = 12).Note. Exp.: Experimental group, Cont.: Control group; TAU: Treatment-as-usual; * Significant finding; † three-day forest-experience-integration intervention consisted of preparation phase (30 min), physical intervention (20 min), psychological intervention (20 min), physical intervention (20 min), and completion phase (30 min).; ‡ three-day forests healing program conducted in the healing forest area and consisted of forest healing activities and oriental medicine treatments. Forest healing activities included various activities in the forest, such as exercise, Qi-Qong program, and experiencing forest using five senses. Oriental medicine treatments included natural herbal diet, herbal footbath therapy, aroma therapy, herbal tea therapy, and oriental medicine music; § nine-day forest therapy camp consisted of three sessions and each session lasted for three days. Each session included various therapeutic activities including nature games and nature interpretation (1st session); mountain-climbing, trekking, and orienteering (2nd session); nature-meditation and counseling in forest environment (3rd session).

Med-MDPI/ijerph_2/ijerph-14-03-00322.txt

@@ -0,0 +1 @@
+Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).Environmental noise is known to cause noise annoyance. Since noise annoyance is a subjective indicator, other mediators—such as noise sensitivity—may influence its perception. However, few studies have thus far been conducted on noise annoyance in South Korea that consider noise sensitivity and noise level simultaneously. The aim of this study was to evaluate the correlations between noise sensitivity or noise level and noise annoyance on a large scale in South Korea. This study estimated the level of noise exposure based on a noise map created in 2014; identified and surveyed 1836 subjects using a questionnaire; and assessed the impact of transportation noise and noise sensitivity on noise annoyance. The result showed that noise exposure level and noise sensitivity simultaneously affect noise annoyance, and noise sensitivity has a relatively larger impact on noise annoyance. In conclusion, when study subjects were exposed to a similar level of noise, the level of noise annoyance differed depending on the noise sensitivity of the individual.Environmental noise is defined as “unwanted or harmful outdoor sound created by human activities” [1]. This includes transportation noise caused by airplanes, automobiles, or trains; neighbourhood noise; and leisure noise [2]. Environmental noise is known to cause a wide array of health problems. The World Health Organization (WHO) reported on such health problems as tinnitus, cardiovascular disease, child cognitive disabilities, sleep disorder, and noise annoyance in 2011 [3].Among those health problems, noise annoyance is defined as “a feeling of displeasure caused by noise” [4]. As noise annoyance has a recommended threshold and shows a dose-response relationship, it has been widely used in assessing the health effects of environmental noise [5,6]. Since noise annoyance is a subjective indicator, it is affected not only by the level of noise exposure, but also by other mediators, including fear of danger from the noise source [7,8], noise preventability [7], attitude towards the noisy situation [9], and noise sensitivity [7,8,9,10]. Among these mediators, noise sensitivity is defined as “a factor involving underlying attitudes towards noise in general” and is also known to affect noise annoyance; many studies have suggested considering noise sensitivity together when analysing noise annoyance [2,11,12,13,14,15,16]. Few studies have thus far been conducted on noise annoyance in South Korea that consider noise sensitivity and noise level simultaneously, and there have been few investigating a large-scale population.This study aims to examine correlations between noise sensitivity or noise level and noise annoyance by using data about transportation noise exposure from a community-dwelling setting of a local population on a large scale.The study site selected was Yangcheon-gu in Seoul and Nam-gu in Ulsan, areas for which we completed noise maps in 2014. Based on the noise maps, we stratified the buildings in those selected districts into four levels based on noise level (below 50 dBA, 50–59.9 dBA, 60–69.9 dBA, and above 70 dBA), then grouped them into similar areas. We determined the sample size for each level based on the size of the population. In order to extract households at the same probability, the study used a local sampling method and recruited 1000 subjects in Seoul and Ulsan, respectively. Until the required sample size was achieved, we contacted 2341 subjects in Seoul and 1965 subjects in Ulsan by using available methods (mainly home visiting, e-mail, and phone calls). When we contacted subjects and accounted for this study, if subjects refused to participate then we excluded those subjects. We visited subjects’ houses, described the object of this study, provided guidance, and obtained written consent. We conducted face-to-face interviews using computer-assisted personal interviewing (CAPI) to reduce the missing rate. We administered the questionnaire survey from July 2015 to January 2016. Out of 2000 possible subjects, 1836 were included in the study after excluding 164 whose questionnaire answers were missing (Figure 1).The questionnaire included questions on social and demographic information as well as questions on noise sensitivity and noise annoyance. Social and demographic variables included age, sex, education level, marital status, monthly income, smoking status, alcohol drinking, exercise, and length of time at the present residence. Education level was divided into high school graduation or lower and two-year college graduation or higher; marital status was divided into married and single; and the monthly income was divided into less than 3 million KRW and at least 3 million KRW. Smoking status was divided into current smoker and current non-smoker (including past smoker and non-smoker); current smokers were defined as those who smoked currently; past smokers were defined as those who smoked more than 100 cigarettes over their lifetime and did not smoke currently; non-smokers were defined as those who smoked less than 100 cigarettes over their lifetime and did not smoke currently [17]. Alcohol drinking was categorized into current drinker and current non-drinker, and exercise status was categorized into regular exerciser and non-regular exerciser.To assess noise sensitivity and noise annoyance, we used an 11-point visual analog scale (VAS) ranging from 0 to 10 that we created based on the International Organization for Standardization Technical Specification (ISO/TS) 15666 (2003) [18]. For noise sensitivity, subjects exceeding the average scale value of the total subjects were classified as “high sensitivity (6–10 points)” group, and others classified as the “low sensitivity (0–5 points)” group. For noise annoyance, subjects exceeding 72% of the point scale (8–10 points) were classified as the “highly annoyed” group, while subjects exceeding 50% of the point scale (6–10 points) were classified as the “annoyed” group [6].In order to estimate the noise level of the residential districts in which the subjects resided, this study used a noise map that we created in 2014. With data from the noise map, we used noise prediction software (Cadna A, DataKustik, Gilching, Germany) to calculate the transportation noise levels at the exterior wall of the residential buildings of the study subjects, based on addresses confirmed during the questionnaire survey. The number of passing vehicles per hour and the percentage of heavy vehicles per hour are the main input variables for the calculation of road traffic noise. Those values are measured for each time interval on the real road. Additionally, road shape, road surface, barriers by the roads, and the speed limit of the road are included for the input values. Furthermore, the geographical and meteorological inputs are used, such as three-dimensional building polygons, contour lines, and annual temperature and pressure values. The verification of the noise map was conducted by measuring twenty points of the study area and comparing the calculated values to the measured values. If the difference between calculated values and measured values was less than 3 dB, then the noise map was considered reliable for use in the study.This study used the day–night average sound level (Ldn) as a noise indicator. The low-noise group and the high-noise group were divided based on the threshold at which environmental noise could pose a risk to health [19]; 55 dBA—the average transportation noise in the roads of the study districts—was set as the threshold in this study. Subjects were classified as the low noise group when their noise exposure level was less than 55 dBA, while subjects were classified as the high noise group when their level of exposure was 55 dBA or higher.We performed a Pearson’s correlation analysis to look into correlations between noise level or noise sensitivity and noise annoyance. We performed multiple linear regression analysis in order to check multi-collinearity between noise level and noise sensitivity. To compare the ratio of those “highly annoyed” and “annoyed” based on noise level and noise sensitivity, we conducted a chi-square test. Based on noise level and noise sensitivity, we categorized the subjects into four combinations: “low sensitivity/low noise”, “low sensitivity/high noise”, “high sensitivity/low noise”, and “high sensitivity/high noise”. To compare age and length of time at the present residence according to the four combinations, this study used analysis of variance (ANOVA) and Tukey’s method for post-hoc verification. To compare sex, education level, marital status, monthly income, smoking status, alcohol drinking, exercise, highly annoyed, and annoyed, we performed a chi-square test.We conducted logistic regression to show an interaction by modelling interaction variables (noise sensitivity × noise exposure). We also conducted multiple logistic regressions to calculate the adjusted odds ratio (aOR) to adjust for confounders that could affect annoyance (age, residential period, sex, education level, marital status, monthly income, smoking status, alcohol drinking, and exercise).We used SPSS 21.0 (IBM SPSS Inc., Chicago, IL, USA) to analyse all data. The significance level was set at 0.05, and we considered a p-value of less than 0.05 to be significant.This study was approved by the Institutional Review Board of Ulsan University Hospital (IRB No. 2014-08-008). This study has been conducted from 2014 to the present. From 2015, we examined the health effects of environmental noise on humans. All participants took part in this study voluntarily and written consent was obtained from participants.The results of the correlation analysis between noise sensitivity, noise level, and noise annoyance to transportation noise showed that the correlation coefficient between noise sensitivity and noise annoyance was 0.39 (p < 0.001) while the correlation coefficient between noise level and noise annoyance was 0.20 (p < 0.001), each of which showed a positive correlation. There was no multi-collinearity between noise level and noise sensitivity in the results of multiple linear regression. We have not presented the results of regression as a table.The general characteristics of all subjects and the four combination groups are presented in Table 1. The average age of subjects was 47.0 ± 16.1 years; the average residence period was 9.1 ± 8.5 years; and the average noise exposure level was 55.2 ± 10.4 dBA. After the characteristics of the four combination groups were analysed, we found that the average age was higher in the two high noise sensitivity groups than the two low noise sensitivity groups (p = 0.019). The average residence period was also longer in the two high sensitivity groups than the two low sensitivity groups (p = 0.009). The proportion of women was higher in the two high sensitivity groups than the two low sensitivity groups (p < 0.001), while the education level was lower in the two high sensitivity groups than the two low sensitivity groups (p = 0.001). The monthly income was higher in the two high noise groups than the two low noise groups (p < 0.001). The proportion of both smoking status and regular exercise was higher in the two low noise groups than the two high noise groups (p < 0.001, Table 1).The proportion of “highly annoyed” and “annoyed” by noise exposure showed an increasing trend as noise exposure increased. This trend also appeared in noise sensitivity, but the proportion was higher in the noise sensitivity group than the noise exposure group. When the four combinations of “low sensitivity/low noise”, “low sensitivity/high noise”, “high sensitivity/low noise”, and “high sensitivity/high noise” were categorized in consideration of noise exposure level and noise sensitivity together, and the proportion of “highly annoyed” and “annoyed” for those four groups were analysed, it was found that the proportion of the “highly annoyed” for each group was 4.2%, 6.6%, 15.3%, and 23.0% (p < 0.001), respectively, while the proportion of “annoyed” was 13.8%, 22.0%, 41.7%, and 55.2% (p < 0.001, Table 2), respectively.A model that considered the interaction variables (noise sensitivity × noise exposure) showed statistical significance in “highly annoyed” (OR 3.37; 95% CI 2.51–4.53) and “annoyed” (OR 3.81; 95% CI 3.03–4.78) groups. After analysing the risk of annoyance in consideration of both noise level and noise sensitivity, the aOR of being “highly annoyed” in “low sensitivity/high noise”, “high sensitivity/low noise”, and “high sensitivity/high noise” was 1.72 (95% CI 0.98–3.02), 4.14 (95% CI 2.46–6.99), and 7.08 (95% CI 4.28–11.73), respectively, compared to the “low sensitivity/low noise” group. The aOR of being “annoyed” for those groups was 1.74 (95% CI 1.25–2.42), 4.30 (95% CI 3.10–5.97), and 7.38 (95% CI 5.33–10.21), respectively (Figure 2).To assess the noise annoyance of transportation noise, this study analysed data from 1836 residents in Yangcheon-gu, Seoul, and Nam-gu, Ulsan which were located on a developed noise map, and compared noise annoyance depending on the noise level. The average noise level estimates based on residential districts on the noise map were 55.2 ± 10.4 dBA (ranging from 46.0 ± 5.7 to 64.0 ± 5.7 dBA). This noise level was not as high as occupational noise that reaches about 90 dBA [2], so noise sensitivity would have a greater impact on noise annoyance [10]. In this respect, this study stratified the subjects according to noise level, noise sensitivity, and noise level and noise sensitivity together, and analysed their impact on noise annoyance, respectively.Initially, we performed a correlation analysis to verify correlations between noise level or noise sensitivity and noise annoyance. The results showed that there were significant correlations between the two variables and noise annoyance, and that the correlation coefficient between noise sensitivity and noise annoyance (0.39) was higher than that between noise level and noise annoyance (0.20). When the proportion of “highly annoyed” and “annoyed”—depending on noise level or noise sensitivity—was analysed, the higher noise level group and the higher noise sensitivity group showed a higher proportion of “highly annoyed” and “annoyed”, but the difference was much larger for noise sensitivity. Past studies reported that differences in noise annoyance depending on noise level were not distinctive when there was a low level of noise exposure, but noise sensitivity had a larger impact on noise annoyance when there was a low level of noise exposure [10,20,21]. The reason behind such results could be that noise annoyance—the indicator we used in this study—is a subjective indicator, and it could be affected by noise sensitivity—a subjective characteristic [13,16,22].Based on our initial findings, we re-classified the subjects into four groups in consideration of the noise level and the noise sensitivity together, and analysed the proportion of “highly annoyed” and “annoyed”. The results showed that the proportion of “highly annoyed” and “annoyed” increased in the order of “low sensitivity/low noise”, “low sensitivity/high noise”, “high sensitivity/low noise”, and “high sensitivity/high noise”. Furthermore, the results of multiple logistic regression analysis showed that the aOR of being “highly annoyed” and “annoyed” tended to gradually increase in the order of “low sensitivity/high noise”, “high sensitivity/low noise”, and “high sensitivity/high noise” compared to the “low sensitivity/low noise” group. Although many previous studies found that noise level and noise sensitivity affected noise annoyance, most of those studies presented results regarding analysis of correlations only [2,10,22]. Unlike the methods of the past studies, this study stratified subjects according to noise level and noise sensitivity, re-classified subjects into four groups, and analysed the impact on noise annoyance. We found that noise level and noise sensitivity simultaneously affect noise annoyance, and when we analysed with four groups, the impact of noise sensitivity on noise annoyance was more prominent than that of the noise level. In addition, although exposed to a similar level of transportation noise on the road, reactions to noise annoyance differs depending on noise sensitivity. Therefore, if noise level were considered alone when assessing the impact of transportation noise on noise annoyance, there could be a possibility of underestimating the impact of noise.Moreover, in the results of the general subject characteristics, it was found that higher noise sensitivity was correlated with relatively higher age, lower education level, and female sex. Even though there have been few studies looking into factors that impact noise sensitivity, we could find that a previous study found similar results [23,24]. In summary, noise sensitivity was higher among those who could be considered a relatively vulnerable group, and noise annoyance was higher, although they were exposed to a similar level of transportation noise as those who reported low noise annoyance.Thus, we found that noise level and noise sensitivity simultaneously affect annoyance, and noise sensitivity has a relatively larger impact on noise. Furthermore, as seen in this study’s results, when noise sensitivity was considered together with noise level, the impact on annoyance could be assessed in more detail for similar levels of noise exposure. In addition, unlike the industrial workplace population comprising mostly physically healthy workers, environmental noise—including transportation noise—could impact a vulnerable group who could have relatively higher noise sensitivity [23,24,25]. Therefore, if noise sensitivity is considered with noise together when assessing not just noise annoyance, but also other health impacts of environmental noise, it would ensure a more appropriate health assessment.This study has some limitations. First, it was a cross-sectional study that could only evaluate correlations between noise level or noise sensitivity and noise annoyance, and was not able to verify causal relationships or assess long-term exposure. Second, we used an 11-point VAS scale based on ISO/TS 15666 (2003) to assess noise sensitivity because there is no universally used simple noise sensitivity scale. Thus, an absolute cut-off value would be inaccurate, nevertheless we used the average value of the subjects as a cut-off value because we thought it is reasonable. Third, assessment of noise annoyance—one of the most widely used indicators to evaluate the health impact arising from environmental noise exposure—is generally conducted based on a questionnaire (a subjective indictor), and an objective assessment method to support the questionnaire has thus far not been established. Therefore, self-report bias could have occurred on the questionnaire survey in this study. Fourth, this study was funded by the Korean Ministry of Environment (MOE), and the MOE did not want to cover extreme noise levels. Thus, the noise level ranged from 46 to 64 dBA, and we could not assess annoyance at higher noise levels.Nevertheless, this study has several important implications. First, it is the first study in South Korea that has assessed the health impact of environmental noise on noise annoyance in a large-scale study of a population exposed to transportation noise in their daily lives. Second, while previous studies mostly focused on assessing noise annoyance depending on noise level, this study considered noise sensitivity as well. Thus far, there have been few studies on the health impact of environmental noise in South Korea that assessed noise sensitivity and environmental noise levels together. Therefore, this study could be meaningful in that it is the first large-scale study in South Korea that considers noise level and noise sensitivity in assessing noise annoyance.In conclusion, we could see that when a population is exposed to a similar level of noise, the level of noise annoyance varies depending on noise sensitivity—especially with relatively low noise levels, such as environmental noise. When other variables that could affect the subjective assessment are controlled, the results are identical. Therefore, a future study on the health impact of environmental noise needs to consider not only the physical effects of noise, but also individuals’ noise sensitivity.This study was supported by the Korea Ministry of Environment (MOE) as “the Environmental Health Action Program (grant number 2014001350001)” and Ulsan University Hospital (Biomedical Research Center Promotion Fund, grant number 2013804). The authors would like to thank the Occupational and Environmental Medical Center, Ulsan University Hospital and Korea Ministry of Environment (MOE). The authors also are grateful to the participants of the survey.Joo Hyun Sung, Jiho Lee, Kyoung Sook Jeong, Min-Woo Jo, and Chang Sun Sim developed the conception and design of the study; Soogab Lee and Changmyung Lee created the noise map and calculated the noise levels; Joo Hyun Sung analyzed the data and wrote the manuscript; Joo Hyun Sung and Chang Sun Sim interpreted the result and revised the manuscript; and all authors critically reviewed and approved the final manuscript.The authors declare no conflict of interest.Flowchart of subject selection criteria.Adjusted odds ratio of being (a) “highly annoyed” and (b) “annoyed” according to noise sensitivity and noise exposure.General subject characteristics.Unit: mean ± standard deviation, number (percentage); a Low noise sensitivity and low noise exposure; b Low noise sensitivity and high noise exposure; c High noise sensitivity and low noise exposure; d High noise sensitivity and high noise exposure; * post hoc comparison using Tukey’s method: a,b < c,d; † post hoc comparison using Tukey’s method: a,b < c; ** post hoc comparison using Tukey’s method: a,c < b,d.Proportion of highly annoyed and annoyed according to noise exposure, noise sensitivity, and a complex of noise sensitivity and exposure.Unit: number (percentage); a NS, noise sensitivity; b NE, noise exposure; * p < 0.001; † p = 0.001.

Med-MDPI/ijerph_2/ijerph-14-03-00323.txt

@@ -0,0 +1 @@
+These authors equally contributed as corresponding author in this study.Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).The objective of this study was to investigate the aquatic-toxic effects of glyoxal-containing cellulose ether with four different glyoxal concentrations (0%, 1.4%, 2.3%, and 6.3%) in response to global chemical regulations, e.g., European Union Classification, Labeling and Packaging (EU CLP). Toxicity tests of glyoxal-containing cellulose ether on 11 different microbial strains, Microcystis aeruginosa, Daphnia magna, and zebrafish embryos were designed as an initial stage of toxicity screening and performed in accordance with standardized toxicity test guidelines. Glyoxal-containing cellulose ether showed no significant toxic effects in the toxicity tests of the 11 freeze-dried microbial strains, Daphnia magna, and zebrafish embryos. Alternatively, 6.3% glyoxal-containing cellulose ether led to a more than 60% reduction in Microcystis aeruginosa growth after 7 days of exposure. Approximately 10% of the developmental abnormalities (e.g., bent spine) in zebrafish embryos were also observed in the group exposed to 6.3% glyoxal-containing cellulose ether after 6 days of exposure. These results show that 6.3% less glyoxal-containing cellulose ether has no acute toxic effects on aquatic organisms. However, 6.3% less glyoxal-containing cellulose ether may affect the health of aquatic organisms with long-term exposure. In order to better evaluate the eco-safety of cellulosic products containing glyoxal, further studies regarding the toxic effects of glyoxal-containing cellulose ether with long-term exposure are required. The results from this study allow us to evaluate the aquatic-toxic effects of glyoxal-containing cellulosic products, under EU chemical regulations, on the health of aquatic organisms.Cellulose ethers are water-soluble polymeric substances derived by the etherification of cellulose, which is one of the most widespread natural organic compounds. They are known as environmentally friendly polymers as a result of their properties (e.g., water solubility, pH stability, and biodegradation) [1,2,3,4,5]. Chemical modification of cellulose ethers has been performed to improve their physical and chemical properties (e.g., organic solubility, viscosity stability, water retention, and non-ionic charges), and cellulose ethers in mixed with organic or inorganic chemicals have also been produced to enhance the cellulosic product quality by physical blending of functionalized additives [2,5,6,7]. These derivatives can be used in a wide range of industrial applications, including construction products, ceramics, paints, personal care products, and pharmaceuticals [2,5,8,9]. With the increasing use of chemicals in industrial applications, the Globally Harmonized System of Classification and Labeling of Chemicals (GHS) proposed harmonized hazard communication elements, including safety data sheets (SDSs) for chemicals. It was adopted in the chemical regulations of the European Union (EU) Registration, Evaluation, Authorization, and Restriction of Chemicals regulation (REACH) and Classification, Labeling and Packaging (CLP) regulation [10,11]. Since most industrial applications of chemicals are based on mixtures containing more than a single chemical substance, the classification and labeling of chemical substances have been extended to a whole mixture of various substances from 2015 [11,12,13]. Therefore, SDSs for cellulosic products contained in functionalized additives under the EU CLP regulation are required.Glyoxal is one of the most extensively used cross-linking agents in cellulosic products due to the advantage that its chemical reaction can enhance the solubility and dispersion of the polymer without chemical modification [6,7]. However, the glyoxal toxicity for human health is an issue and may cause sensitization for eye and skin contact, and its properties have led to the regulation of glyoxal content in functionalized products with glyoxal [6,14,15,16]. For example, the use of glyoxal in paper and textile wall coverings is prohibited [14]. The European Commission (EC) defined manufactured cosmetic products with a content of up to 100 mg/L glyoxal as safe [15,16]. When more than 1% of glyoxal is used for the reaction process with cellulose ether, companies are obliged to describe the cellulose ether products under EU CLP registration [6,15]. These regulations imply that the use and production of glyoxal-containing products may result in glyoxal being released into the environment and may thus arouse concern with respect to the potential impact on human and environmental health [15,16,17].The predominant target compartments for glyoxal-containing cellulosic products in the environment are hydrospheres rather than sediment, when considering production to final emission [16,17]. Nevertheless, the aquatic hazards for glyoxal-containing cellulosic products are currently unclear, and the toxicity tests for these cellulosic products have been not performed for aquatic organisms. Cellulose is not classified as a hazardous polymer under the chemical regulations of the EU due to its being a natural organic compound [1,2,3,4,5,10,11]. Glyoxal is classified as a non-hazardous substance in the environment due to high half-effective concentrations of more than 100 mg/L in aquatic organisms, i.e., microorganisms, algae, Daphnia, and fish [16,17,18,19,20]. However, the eco-safety assessment of glyoxal-containing cellulosic products is urgently needed in response to the EU CLP regulations.Therefore, the aim of this study was to provide information on aquatic toxicology available for glyoxal-containing cellulose ether based on a reliable evaluation of its hazardous properties. Surface-treated cellulose ethers with glyoxal were chosen in this study, since these glyoxal-containing cellulose ethers were specially developed to prevent lumping effects or to improve the rheological properties in wet blending applications, such as paints and emulsion. These cellulose ethers with four different concentrations of glyoxal were classified hazardous substances for human health by EU CLP regulations (Table 1). We evaluated the aquatic-toxic effects of glyoxal-containing cellulose ethers with four different concentrations by analysis of the toxicity tests, using four different aquatic organisms [21,22,23,24].Glyoxal-containing cellulose ethers with four different glyoxal concentrations (0%, 1.4%, 2.3%, and 6.3%) were obtained from Lotte Fine Chemical Co., Ltd., Ulsan, Korea (cellulose, CAS 9004-65-3) (Table 2), and test solutions were prepared in accordance with the manufacturer’s manual (Lotte Fine Chemical Ltd., Ulsan, Korea) [25]. Briefly, surface-treated test samples in powder form were dissolved using a magnetic stirrer in purified cold water for 1 h at room temperature to ensure the complete removal of any undissolved powder samples before use in the toxicity tests with the different aquatic organisms.A schematic design for the aquatic-toxic effect tests is shown Figure 1. Releases to environment of the glyoxal-containing cellulose ethers are primary emissions to water due to high solubility [15,16,25]. The ecological effects on aquatic organisms may depend on ingestions or biodegradation of glyoxal-containing cellulose ethers [15,16]. However, the aquatic toxicity of glyoxal-containing cellulose ethers is unknown. To understand the aquatic toxicity on different biological levels in this study, the aquatic toxicity tests were conducted with four different aquatic organisms. Aquatic toxicity testing of glyoxal-containing cellulose ethers on 11 freeze-dried microbial strains, Microcystis aeruginosa, Daphnia magna, and zebrafish embryos have been covered by standardized methods [22,23,24,26]. The following points were investigated: the growth inhibition of microbial strains and cyanobacteria, Daphnia acute toxicity based on mortality and immobility, and zebrafish embryo toxicity based on mortality, hatchability, and abnormalities.A MARA system involving 11 freeze-dried microbial strains was purchased from NCIMB Ltd., Aberdeen, UK (Table 3). The MARA for test solutions (i.e., glyoxal-containing cellulose ether with four different glyoxal concentrations) was performed according to the manufacturer instructions [26]. Briefly, 11 freeze-dried microbial strains were pre-incubated in a 96-well MARA plate with 150 µL of aqueous nutrient peptone (2% phytone peptone) for 4 h at 30 °C. After pre-incubation, 200 µL of the test solution was transferred to the 96-well MARA plate with 100 µL of medium containing 0.01% redox indicator. A volume of 15 µL of each microbial strain was added to each well and incubated for 18 h at 30 °C. After 18 h of incubation, the MARA plates were scanned by a scanner (HP Scanjet G4050, Hewlett Packard, Palo Alto, CA, USA) using transmitted light with a resolution of 100. The average growth rate of the 11 freeze-dried microbial strains exposed to test solutions was determined with the reduction of tetrazolium red using MARA software in quadruplicate.Microcystis aeruginosa was purchased from the Culture Collection of Algae and Protozoa, Cambria UK strain (CCAP 1450/1), and cultivated at 25 ± 1 °C with 16 h of light and 8 h of darkness (16L/8D) in Blue-Green medium (BG11, Sigma-Aldrich, Darmstadt, Germany) until use for the algal growth inhibition test (a total volume of ≥1 × 105 cell/mL). Based on the finding in the MARA system that there was no influence between 11 freeze-dried microbial strains exposed to test solutions, except on 6.3% glyoxal-containing cellulose ethers, algal growth inhibition tests for 6.3% glyoxal concentrations only were conducted based on the protocol described in the OECD Guideline 201 [23,27].Daphnia magna ephippia was obtained from Daphtoxkit FTM (MicroBioTests Inc., Gent, Belgium) and was cultured according to the Daphtoxkit FTM manual. Toxicity tests for four different glyoxal concentrations (0%, 1.4%, 2.3%, and 6.3%) of Daphnia magna neonates within 8 h after hatching were conducted in six-well cell culture plates (Cellstar®, greiner bio-one, Frickenhausen, Germany) filled with 10 mL of each test solution [28]. During the experimental period, Daphnia magna neonates were maintained at 23 ± 0.5 °C under a light cycle of 16 h light and 8 h darkness (16L/8D). The toxicity to Daphnia magna was investigated after an exposure period of 48 h to glyoxal-containing cellulose ether [22]. During the experimental period, the dead and immobilized individuals for each cellulose ether solution were recorded for calculating the immobilization (immobilization = (numbers of dead and immobilized individuals /number of initial individuals) × 100) after exposure to each cellulose ether solution. Tap water filtered through a Millipore® 0.22 µm of nitrocellulose membrane (GSWP) filter (Merck KGaA, Darmstadt, Germany) after sterilization at 130 °C for 1 h was used as a control medium. Toxicity for each cellulose ether was determined in quadruplicate (10 daphnids per replicate).Wild-type zebrafish (AB strain) breeding and maintenance was performed under a long photoperiod (16L/8D light/dark cycle) at a 26.5 ± 0.5 °C water temperature [25,28]. Tap water filtered through a Millipore® 0.22 µm GSWP filter (Merck KGaA, Darmstadt, Germany) after sterilization at 130 °C for 1 h was used as the control group. The zebrafish embryo toxicity tests with four different glyoxal concentrations (0%, 1.4%, 2.3%, and 6.3%) were conducted in six-well cell culture plates (Cellstar®, Greiner bio-one, Frickenhausen, Germany) filled with 10 mL of each test solution. Embryos at 8 h post-fertilization were treated with glyoxal-containing cellulose ether for 6 days which is commonly referred to as the embryonic stage. The mortality, hatchability, and developmental abnormalities in each test solution were recorded to assess the embryo toxicity of each glyoxal-containing cellulose ether [25,28]. The embryo toxicity for each cellulose ether was determined in quadruplicate (10 embryos per replicate).All errors are expressed as mean ± standard error of mean (SEM). Comparison between the toxicities for each glyoxal-containing cellulose ether was carried out using a post hoc Student–Newman–Keuls test in the one-way ANOVA (SigmaPlot version 12.5, Systat Software, Inc., San Jose, CA, USA). Statistical significance was set at p < 0.05.Tests for ecotoxicological effects of chemicals on algae, waterflea (Daphnia magna), and fish (Danio rerio) are covered by the OECD test guidelines for the testing of chemicals [22,23,24]. Toxicity tests with the zebrafish embryos were conducted in accordance with the protocol as prescribed by the OECD [24].The toxicity of each glyoxal-containing cellulose ether for 11 freeze-dried microbial strains and cyanobacteria are shown in Table 4 and Figure 2. There was no remarkable difference between the growth rates of the 11 freeze-dried microbial strains exposed to glyoxal-containing cellulosic products of ≤2.4%, due to the growth rates being more than 75% in all strains (Table 4). However, when exposed to glyoxal-containing cellulosic products of 6.3%, the growth rate of Microbial Strain #2 (Brevundimonas diminuta, NCIMB 30256) was less than 75% and tended to decrease in a dose-dependent manner (Table 4, Figure 2A). In addition, the growth inhibition rate of M. aeruginosa in the 6.3% glyoxal-containing cellulose ether-exposed group for 7 days was significantly higher (i.e., 60.6%) when compared to that of the group exposed for 3 days (a growth inhibition rate of 32.9%), and the decrease in growth was dependent on exposure time (Figure 2B). The difference between the microbial strains and M. aeruginosa in terms of the growth inhibition after exposure to glyoxal-containing cellulose ethers may be a result of exposure time rather than glyoxal concentration. These results imply that 6.3% glyoxal-containing cellulose ether may have bactericidal properties or may not be totally biodegradable over the short term, unlike in previous reports [1,2,3,4,5,15,16,17]. Therefore, the lack of effects on the aquatic microbial or bacterial growth may be due to the biodegradation of glyoxal-containing cellulose ether with exposure time. In order to examine the safety on aquatic microbial or bacterial growth, further study for examining the effect of biodegradation activity on aquatic microbial or bacterial growth is needed.After 48 h of exposure to the glyoxal-containing cellulose ethers with four different concentrations, the mortality and immobility of Daphnia magna were rarely observed in the glyoxal-containing cellulose ether-exposed group, due to the results that normal behavior was observed in more than 85% of cases in all groups (Figure 3A). The zebrafish embryo toxicity was also not observed in the glyoxal-containing cellulose ether-exposed group throughout the experimental period. The survival rate and hatching rate of zebrafish embryos exposed to glyoxal-containing cellulose ethers with four different concentrations were higher than 90% and 85%, respectively (Figure 3B). It appears that glyoxal-containing cellulose ethers had no effect on Daphnia magna acute toxicity or zebrafish embryo toxicity [16,17], even if 6.3% glyoxal-containing cellulose ether inhibited M. aeruginosa growth. The results indicate that the toxicity of glyoxal-containing cellulose ethers are sensitively reflected to M. aeruginosa, rather than waterflea (Daphnia magna) or zebrafish embryos [17,29].Alternatively, while there was no acute toxicity of glyoxal-containing cellulose ether for the zebrafish embryos, developmental abnormalities, such as bent spines, were observed in the glyoxal-containing cellulose ether-exposed group, in particular 2.3% and 6.3% glyoxal-containing cellulose ethers (Figure 4). However, there were no significant differences between embryo developmental abnormalities among any of the groups (p > 0.05). It is possible that glyoxal-containing cellulose ethers do not result in acute toxicity in mortality and hatchability on zebrafish embryos, but may cause developmental abnormalities after 6 days of exposure. Unfortunately, studies examining the development toxicity of glyoxal-containing cellulose ethers on zebrafish embryos are rare and their toxicity mechanisms are also unknown. Further studies on the effects induced by glyoxal-containing concentration during zebrafish embryo development are therefore required.Consequently, cellulose ethers with four different glyoxal concentrations have a lack of toxicity on different aquatic environments within the short term after exposure, suggesting that aquatic organisms are not taken in in the short term. However, in order to better understand the ecological safety of glyoxal-containing cellulose ethers, an evaluation of toxicity by long-term exposure on aquatic organisms is required.In this study, toxicity tests using different aquatic organisms were conducted to investigate the effects of cellulosic products containing four different glyoxal concentrations (i.e., 0%, 1.4%, 2.3%, and 6.3%) as functional additives to aquatic environments under EU chemical regulations. A major finding from this study is the non-toxic effect of ≤6.3% glyoxal-containing cellulose ethers: even ≥1.4% glyoxal-containing cellulose ethers were classified as hazardous substances for human health by the EU CLP. Therefore, our systemic toxicity assessments might be valid for the risk assessment of glyoxal-containing cellulosic products under EU chemical regulations.This work was funded by the KIST Europe Institutional Program (Project No. 31601). We would like to thank the Korea Institute of Science and Technology (KIST) Europe and Lotte Fine Chemical (LFC) for helpful comments that greatly improved earlier drafts of the manuscript.Chang-Beom Park and Youngjun Kim had the original idea for the study and performed the aquatic toxicity test of glyoxal-containing cellulose ethers with different aquatic organisms. Min Ju Song, Nak Woon Choi, and Sunghoon Kim was responsible for sample preparation and analysis including characterization of glyoxal-containing cellulose ethers. Hyun Pyo Jeon and Sanghun Kim were involved in data processing. All authors contributed to the conception and design of the experiments and have given their approval to the final version of the manuscript.The authors declare no conflict of interest.The sequential assessing of the aquatic-toxic effects of glyoxal-containing cellulose ethers on the various aquatic organisms according to the Organization for Economic Co-operation and Development (OECD) guidelines.Growth inhibition rate of microbial strain #2 (A) and M. aeruginosa (B) after exposure to 6.3% glyoxal-containing cellulose ether. The inhibition rates (IR) of M. aeruginosa growth at 3 and 7 days after exposure to 6.3% glyoxal-containing cellulose ether were 32.9% and 60.6%, respectively. * denotes significant differences between algal cell density (p < 0.05).Acute toxicity of Daphnia magna after 48 h (A) and zebrafish embryo toxicity after 6 days (B) to glyoxal-containing cellulose ether with four different glyoxal concentrations after exposure.Zebrafish embryo toxicity to glyoxal-containing cellulose ether with four different glyoxal concentrations after 6 days of exposure. All scale bars are 500 µm.Hazardous classification of glyoxal-containing cellulose ethers with four different glyoxal concentrations calculated by the EU CLP.“-“ = Not calculable; “Skin sens. 1” = May cause an allergic skin reaction; “Muta. 2” = Suspected of causing genetic defects.Physicochemical characteristics of surface treated cellulose ethers with four different concentrations of glyoxal.Freeze-dried microbial strains for microbial assay for toxicity risk assessment (MARA, NCIMB Ltd., Aberdeen, UK).Average growth rate of freeze-dried microbial strains exposed to surface treated cellulose ethers with four different concentrations of glyoxal.† Growth rate is lower than 75%, which was the lowest growth rate in this study. 

Med-MDPI/ijerph_2/ijerph-14-03-00324.txt

@@ -0,0 +1,12 @@
+Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).Objective: Childhood obesity is a serious concern for developed and developing countries. This study aimed to assess the level of support in Australia for regulation and to assess whether systematic differences occur between individuals who support increased regulation and individuals who oppose it. Methods: An online survey (n = 563) was used to assess parental/caregiver preferences for taxation policy options and nutrition labelling designed to address the incidence of childhood obesity. Participants were parents or caregivers of young children (3 to 7 years) who were actively enrolled in an existing birth cohort study in South-East Queensland, Australia. Results: The majority of the parents (over 80%) strongly agreed or agreed with labelling food and drink with traffic light or teaspoon labelling. Support for taxation was more variable with around one third strongly supporting and a further 40% of participants equivocal about using taxation; however, a quarter strongly rejected this policy. Cluster analysis did not detect any socio-demographic differences between those who strongly supported taxation and those who did not. Conclusions: Better food labelling would be welcomed by parents to enhance food choices for their children. Taxation for health reasons would not be opposed by most parents. Implications for Public Health: Governments should consider taxation of unhealthy drinks and improved labelling to encourage healthy food purchasing.Prevalence of childhood obesity worldwide is at an all-time high with dramatic rises in both high and low-income countries over the last 40 years [1]. In Australia, over the last two to three decades, the problem has risen from a trivial level to a major health problem, with one in four children now overweight or obese [2]. Children from low socio-economic areas are 1.7 times more likely to be overweight or obese [3]. Sugary drinks are of particular concern with modelling showing that over 180,000 deaths per year worldwide are attributable to the consumption of sugar-sweetened drinks [4]. Australian health survey data indicates Australian children are consuming high quantities of sugar-sweetened drinks despite previous public health efforts, such as restrictions on advertising and limitations on sales in schools [5]. Australian public health advocates have therefore called for taxation and improved labelling to address the obesity epidemic [6,7,8].For young children incapable of making or purchasing their own food, parents act as an agent to provide food for them and it is likely that food choices by parents are not always driven by health concerns. First, many parents may have poor information about the future health consequences of a calorie-rich diet or, more likely, the relative risks associated with different food items. Second, even if parents have full information about all the consequences from their actions, present bias means they may assign little importance to events that occur in the distant future [9,10]. Parents may also value convenience, taste or other aspects over the health value of the food. Consequently, parents with these biases (acting as agents for their children) may be inadvertently putting their children at increased risk of harm.In addition to potential failures regarding information and personal preference, on a per-calorie basis, energy-dense foods (those containing fats and sugars) are cheap, whereas foods low in energy density (like fresh fruit and vegetables) are more expensive [11]. Simply put, there are prevailing market biases for cheap energy-dense foods. The parent who is seeking to maximise utility but constrained by both time and budget, is forced to trade-off between buying: (i) cheap high-fat/high-sugar food and more of all other goods; or (ii) expensive low-fat/low-sugar food and fewer of all other goods.Food labelling and pricing measures were two methods of improving food choices with respect to tackling childhood obesity explicitly considered by a panel of experts and by a Citizens’ Jury in Australia [12]. Taxation is a direct measure that can be employed to overcome the prevailing market discrepancy to price energy-dense (high-fat/high-sugar) foods lower than those that are less energy dense but with greater nutrient value. Recent evidence from countries that have introduced taxation on sugar-sweetened beverages is promising. In Mexico, a tax of around 10% has led to a decline in the purchase of taxed beverages of up to 12%, with a complementary increase in sales of bottled water [13]. A systematic review of nine studies found that a tax rise of 10% leads to a reduction in consumption of 5 to 39 kJ per day [14]. Consequently, food taxation polices may provide both a mechanism for individuals to move towards their ideal health state as well a direct measure to reduce consumption of energy dense (high-fat/high-sugar) foods.The provision of improved information at point of sale is an immediate remedy to the current labelling standing in Australia: mandatory back-of-pack labels; voluntary star rating system; and front-of-pack daily intake guides. Voluntary measures have been criticised as being small, indistinct and seriously flawed, as well as too confusing for the average consumer to understand [15]. Additionally, due to the voluntary nature of the scheme, they are often absent from energy dense (high-fat/high-sugar) foods [16]. On the other hand, the use of mandatory, distinct front-of-pack labelling has been shown to help consumers make better food choices [16,17].The emergence of childhood obesity as a national public health crisis is not a recent phenomenon, nor is the call for government action in response. However, few countries have demonstrated willingness to implement substantive regulations, as most have relied on industry self-regulation to improve food quality [18]. It is therefore paramount to understand whether the inactivity of government is a reflection of the will of its constituents or a case of business interests, not citizens, exerting substantial influence on government policy [19]. This study aims to investigate the level of support for taxation and improved nutrition labelling of food and drinks to prevent childhood obesity. Specifically, the study examines whether systematic differences occur between individuals who support increased regulation and individuals who oppose it, and explores the preferences regarding current front-of-pack, traffic light and teaspoon nutrition labels among parents of young children aged 3 to 7 years old.We developed an online survey using the Qualtrics® software (Qualtrics, Provo, UT, USA) to measure the preferences of parents/caregivers of young children (3 to 7 years old) for taxation policy options and nutrition labelling designed to reduce the incidence of childhood obesity. We collected demographic information from the parent/caregiver including: age; gender; type of caregiver; relationship status; education level; employment status; frequency of grocery shopping; household and child consumption of soft drink; household grocery spend per week; and household take-away spend per week.This study was conducted according to the guidelines laid down in the Declaration of Helsinki and all procedures involving human subjects were approved by the Griffith University HREC (2012/828). Written informed consent was obtained from all subjects.We recruited parents/caregivers from those who were actively enrolled in the existing Environments for Health Living (EFHL) study, a longitudinal birth cohort study of children who were born at one of three public hospitals in South-East Queensland and Northern New South Wales, Australia [20]. This geographically defined area has a high proportion of families from lower socio-economic groups who have a higher risk of childhood obesity and, therefore, the consequences of policy directives in this area are relevant. The population recruited in the study is broadly representative of all births in the region [21]. We invited a total of 1332 parents/caregivers of young children who were active in the 2006 (n = 457), 2007 (n = 368) and 2009 (n = 507) cohort groups to participate in the current study. We collected data between October 2013 and February 2014. To maximise response, we entered respondents into a draw to win a $200 debit card for the completion of the online questionnaire. We provided an information sheet at the commencement of the online survey and obtained participant consent through the completion of the survey (in full or in part).Questions on taxation of food and drinks for the current study were selected from questions developed for a Citizens’ Jury conducted in Queensland, Australia [12]. A Citizen’s Jury is a method of public consultation used in participatory action research that was developed by the Jefferson Center and draws on the symbolism and methods of a legal trial by jury [22]. Participants (jurors) were selected from a random sample of the electoral roll to represent the diversity of the Australian population. The questions put to the jurors were based on a literature review of current patterns of consumption in Australian children and taxation measures on foods and drinks, as well as the deliberations of a panel of Australian experts on nutrition and obesity [23].Following a presentation of the evidence by various experts and the subsequent deliberative discussions during the Citizens’ Jury, the jurors unanimously supported taxation on sugar-sweetened drinks but generally did not support taxation on the other types of foods presented. However, the jurors were supportive of taxation on snack foods in conjunction with traffic light nutrition labelling on the packaging. Based on these findings, we asked the participants of the current study to respond to the following questions on taxation:
+          
+            
+              In your opinion, is taxing unhealthy food and drink an appropriate strategy for reducing childhood obesity amongst 0–5 year old children?
+            
+            
+              In your opinion, is it appropriate to tax sugar-sweetened drinks as a strategy for reducing childhood obesity?
+            
+            
+              In your opinion, is it appropriate to tax snack foods as a strategy for reducing childhood obesity?
+            
+          In your opinion, is taxing unhealthy food and drink an appropriate strategy for reducing childhood obesity amongst 0–5 year old children?In your opinion, is it appropriate to tax sugar-sweetened drinks as a strategy for reducing childhood obesity?In your opinion, is it appropriate to tax snack foods as a strategy for reducing childhood obesity?A horizontal middle-marked visual analogue scale (VAS) was displayed with a slider below each of the three questions. The scale was anchored at each end and ranged from 0 (strongly disagree) to 100 (strongly agree). We asked participants to move the slider along the scale to represent their level of agreement with each question.Questions on nutrition labelling of food and drinks for the current study were developed based on the findings from the same Citizens’ Jury as described above [12]. Jurors recommended the introduction of a traffic light labelling system and more graphical representations of the sugar content in products. The current star system in place in Australia was not recommended as an option by the Jury. In light of these results, we asked participants of the present study questions regarding three types of package labelling: current front-of-pack; traffic light; and teaspoon labelling.Participants were shown an example of the current front-of-pack daily intake guide labelling in Australia showing energy plus four key nutrients (fat, saturated fat, sugars, and sodium) [24] (see Supplementary Materials Figure S1). Participants were asked whether they had seen the label before (yes/no), whether the label was considered useful (five point Likert scale), and whether the label was used to make purchasing decisions (VAS scale). Full questions are provided in the supplementary materials.Participants were then shown examples of front-of-pack nutrition labelling using the traffic light system (Figure S2) and a teaspoon label system (Figure S3). The traffic light example was taken from the UK’s Food Standards Agency [25,26]. The teaspoon label example was adapted from The Nutrition Source, Harvard School of Public Health [27], with the nutritional profiling based on that of the UK’s Food Standards Agency [26]. Participants were asked whether these labels would be considered useful in relation to purchasing food for their children and whether they favoured implementing these labels as standard.We sent the questionnaire to a random sample of 50 parents identified from the EFHL study to ensure that the questionnaire software and administration procedures were working. Following successful piloting, we emailed a link to the online questionnaire to all remaining participants who had previously provided their email address and mailed a letter with a web link to the remaining participants who did not have an email address. In order to maximise the participation rate, after two weeks, a reminder email or letter was sent to participants who had not completed the questionnaire. After a further 2 weeks, a reminder SMS text message was sent. Following this, we attempted to personally contact all participants by telephone who had not yet completed the questionnaire.Of the 1332 participants invited to participate in the study, nine participants were non-contactable. Of the remaining 1323 participants, a total of 563 (42.5%) participants (2006 group, n = 75; 2007 group, n = 274; 2009 group, n = 214) agreed to participate in the survey. Of these, 101 participants had incomplete survey data. We used pairwise analyses which excluded participants with incomplete data for each of the respective analyses.Statistical analyses were conducted in STATA13® (Stata IC 13.1; StataCorp, College Station, TX, USA). Comparisons between categorical variables were analysed using chi-square statistics. Continuous variables were analysed for normality using histograms and skewness and kurtosis tests. Comparisons between the continuous variables were analysed using parametric (one-way analysis of variance (ANOVA)) and/or non-parametric (Kruskal–Wallis equality-of-populations rank test) tests, as appropriate.We performed an exploratory cluster analysis using the k-median cluster algorithm [28] to identify clusters in participants’ responses regarding taxation. The primary outcome measures of acceptability of taxation of unhealthy foods and drinks, sugar-sweetened drinks, and snack foods were used to determine to which clusters the participants belonged. The optimal number of clusters was determined using the Calinski–Harabasz pseudo-F index [29]. We used univariate analyses to compare socio-demographic characteristics between the identified clusters.Table 1 presents the demographic characteristics of the sample. Participants ranged in age from 22 to 68 years with a mean age (standard deviation, SD) of 35.6 (5.6) years. This was similar to the mean (SD) age for all participants (n = 1173) in relevant cohort groups recruited in the EFHL study of 34.5 (6.1) years. Almost all participants (99.6%) were mothers and most participants (97.5%) identified as the primary caregiver. Most participants (84%) were in a relationship (married or de facto) and the majority (92%) of participants identified as the main person in the household that did the grocery shopping. Two-thirds of participants had completed higher education and 62% of participants were employed either full-time or part-time. More than 25% of participants reported that their child enrolled in the EFHL study consumed soft drink more frequently than once per week. Grocery spending per week averaged around $220 and ranged up to $700. Spending on takeaway foods and drinks (excluding alcohol and groceries) per week averaged just under $35 with a maximum reported value of $350.Figure 1 presents the support for the different taxation strategies: taxing unhealthy food and drinks (Figure 1a); taxing sugar-sweetened drinks (Figure 1b); and taxing snack foods (Figure 1c). For all three taxes, the results had a trimodal distribution with most people clustered around either of the polar ends (0% or 100% support) and another group clustered around 50%, or ambivalence to taxation.Cluster analysis identified three distinct clusters representing those strongly in support, strongly against and ambivalent towards taxation. These aggregated clusters were based on the participants’ responses to three questions of support for different taxation strategies. The three-cluster solution was identified as the optimal cluster solution using the largest Calinski–Harabasz pseudo-F index as an indicator of distinct clusters, compared with the two- and four-cluster solutions. There were 124 (24%) participants in cluster one, 221 (43%) participants in cluster two and 167 (33%) participants in cluster three. Table 2 presents the clusters by average support for taxation on the three taxation strategies. The average level of support for taxation within each cluster was similar across the three taxation strategies, with somewhat weaker support shown for taxing snack foods as opposed to sugar-sweetened drinks or unhealthy food and drinks.The key caregiver, household, and child characteristics are presented in Table 2 with further details in the online appendix. As there was no difference in results between the parametric and non-parametric tests, and due to the large sample size, only the results from the ANOVA tests are reported. Comparison of various demographic and household characteristics showed that the majority of factors did not differ between the clusters. Only three characteristics were significantly different between groups. Those participants who purchased more soft drink and whose children had more frequent soft drink consumption were more likely to reject the use of taxation while those who were more frequent users of current labels were more likely to be supportive of taxation.Over eighty percent of participants thought teaspoon labelling or traffic light labelling systems would be useful or very useful compared to just over 50% for the current labelling system (Figure 2a). Chi-square goodness-of-fit comparing the new nutrition labelling (i.e., teaspoon or traffic light labelling) to the current nutrition labelling system found that a statistically significantly larger proportion of participants found the new nutrition labelling systems to be useful or very useful compared to the current labelling system.The majority of participants (84%) either strongly agreed or agreed with having traffic light labelling on the front of food and drink packs compared to current front-of-pack labels, while only a very small proportion (1.3%) of participants were not in favour of this new type of label (Figure 2b). Similarly, over 85% of participants either strongly agreed or agreed with introducing front-of-pack teaspoon labelling on drinks, with less than 3% of participants strongly disagreeing or disagreeing with the new labelling system.Taxation caused a polarised response with a little over a third of people being strongly supportive, one quarter strongly opposing and around 40% being indifferent. Analysis of the identified clusters revealed no strong differences between these groups with factors that might be considered to be important such as education, income and work status. The clustering of opinion around indifference to taxation may be due to measurement bias of a middle-marked VAS instrument.The response to the taxation question was similar to initial voting recorded at the commencement of the Citizens’ Jury [12]. This is despite the fact that the sample of people in the Citizen’s Jury was very different to the present study. The Citizens’ Jury contained a broad selection of the public, whereas the present study focused on parents of young children. By the end of the two-day period, the Citizen’s Jury showed unanimous support for taxation on sugar-sweetened drinks, indicating that the educative nature of a Citizens’ Jury can modify preferences. Advocates wishing to introduce taxation may wish to engage in educating the public on the costs and benefits of this strategy in order to win broad support for reform.With respect to labelling, the results of our study demonstrate that parents are likely to welcome the introduction of traffic light and teaspoon labelling on foods and drinks. Improving children’s diet through improved information and price signals could result in large health gains. A large proportion of the parents surveyed reported that their young children are consuming soft drink regularly despite Australian guidelines recommending consumption occur rarely or not at all in this age group [30]. Sugar-sweetened drinks are an easily modifiable component of the diet that can improve diet quality and health.There are several limitations to this study. It is important to note that the study had a restricted sample: almost all participants were mothers of young children recruited from a geographical area in South-East Queensland and Northern New South Wales. This may limit the generalisability of the results; however, results were consistent with a previous Citizens’ Jury with a broadly representative population [12]. The survey had a response rate of 42.5%. While that is relatively high for similar web based surveys [31], it is unknown if the non-responders would have different preferences to those in this sample. The survey was based on self-report and may be prone to issues of under-reporting child soft drink consumption where it is likely that participants considered this socially unacceptable. Previous research has found that people systematically under-report consumption of food and drinks [32]. Parents’ opinions on the usefulness of new front-of-pack nutrition labels via a survey may not reflect their use in real-life purchasing decisions in shop environments where competing information and stimuli can overwhelm consumers.Given the grave consequences of lifelong obesity on health for individuals and society, government should be considering regulatory action to introduce price signals and improved information to address current market failures.Government may be reluctant to introduce taxation on junk foods, believing that most people oppose additional taxes. This study suggests opposition is likely to only come from around one quarter of the community. In addition, the Australian Government has also failed to regulate for compulsory clear labelling, despite indications from this study that this change would be broadly supported. The public and elected representatives may be divided in their opinions as the result of being inadequately educated on the pros and cons of regulation and the consequences of failure to act. To help address this stalemate, more pressure should be placed on government to provide informed debate regarding regulatory strategies. Furthermore, increased effort is required from researchers to provide information to the public at large. Without this process, policy change is unlikely to happen.The Supplementary Materials are available online at www.mdpi.com/1660-4601/14/3/324/s1. Figure S1: Example of current front-of-pack daily intake guide label, Figure S2: Examples of “traffic light” food labels for front of food packs, Figure S3: Example of a teaspoon label for sugar contained in the drink. Table S1: Full sample characteristics, Table S2: Full characteristics of the three identified clusters with respect to approval of taxation.The research reported in this publication is part of the Griffith Study of Population Health: Environments for Healthy Living (EFHL) (Australian and New Zealand Clinical Trials Registry: ACTRN12610000931077). Core funding to support EFHL is provided by Griffith University. The EFHL project was conceived by Rod McClure, Cate Cameron, Judy Searle, and Ronan Lyons. We are thankful for the contributions of the Project Manager, Rani Scott, and the current and past Database Managers. We gratefully acknowledge the administrative staff, research staff, and the hospital antenatal and birth suite midwives of the participating hospitals for their valuable contributions to the study, in addition to the expert advice provided by Research Investigators throughout the project. The authors wish to thank the parents that participated in this study and the Environments for Health Living Project Staff for their assistance. We also thank Angela Simons and Erin Pitt for support in gaining ethical clearance and Erin Pitt for providing assistance in the development of the online questionnaire. Funding for this study was provided through a grant from the previous Australian National Preventive Health Agency (ANPHA) (17COM2011). The ANPHA had no role in the design, analysis or writing of this article.Tracy Comans and Joshua Byrnes conceived the overall study and drafted the manuscript. Nicole Moretto performed data analysis and help draft the manuscript. All authors read and approved the final manuscript.The authors declare no conflict of interest.Support for the different taxation strategies: (a) taxing unhealthy food and drinks; (b) taxing sugar-sweetened drinks; (c) taxing snack foods.Usefulness of and support for different nutrition labelling systems.Key sample characteristics.Notes: SD, standard deviation.Characteristics of the three identified clusters with respect to approval of taxation.Notes: a associated with the one-way analysis of variance (ANOVA); b associated with the Chi-square test; IQR, interquartile range; SD, standard deviation. Three identified clusters with respect to approval of taxation in which participants were aggregated based on three questions of support for different taxation strategies. Small amount of missing data from some of the chi-square and one-way ANOVA analyses (<5%).

Med-MDPI/ijerph_2/ijerph-14-03-00325.txt

@@ -0,0 +1 @@
+Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).International recommendations for mental health care have advocated for a reduction in the length of stay (LOS) in full-time hospitalization and the development of alternatives to full-time hospitalizations (AFTH) could facilitate alignment with those recommendations. Our objective was therefore to assess whether the development of AFTH in French psychiatric sectors was associated with a reduction in the LOS in full-time hospitalization. Using data from the French national discharge database of psychiatric care, we computed the LOS of patients admitted for full-time hospitalization. The level of development of AFTH was estimated by the share of human resources allocated to those alternatives in the hospital enrolling the staff of each sector. Multi-level modelling was carried out to adjust the analysis on other factors potentially associated with the LOS (patients’, psychiatric sectors’ and environmental characteristics). We observed considerable variations in the LOS between sectors. Although the majority of these variations resulted from patients’ characteristics, a significant negative association was found between the LOS and the development of AFTH, after adjusting for other factors. Our results provide first evidence of the impact of the development of AFTH on mental health care and will provide a lever for policy makers to further develop these alternatives.The burden of mental disorders worldwide is high. They will affect one in three individuals over the course of their lifetime [1], and are anticipated to become the leading cause of disability-adjusted life years by 2020 [2,3]. In France, national prevalence data are scarce, but it is estimated that mental disorders contribute to 14% of the overall disease burden, with mental illness constituting the leading cause of disability [4,5], and the suicide rate is among the highest in Europe [6]. Moreover, the costs associated with mental disorders are considerable. They account for 8% of the total national health spending and represent the first item of expenditures for Statutory Health Insurance [7,8].A major challenge of the mental health care system in France is providing optimal care to confront this epidemiological and economic burden. This system is characterized by a territorial organization into geo-demographic areas (sectors) where multidisciplinary teams enrolled and paid by a hospital coordinate and supply inpatient and outpatient services, including ambulatory care and community-based care, to cover the mental health needs of their population. Sectors are the cornerstone of the organization of public mental health care delivery which represents nearly 70% of the costs of psychiatric care in France [8]. Historically, they have relied mostly on inpatient care for the treatment of mental disorders, but this model has been questioned, in particular because of its costs and patients’ dissatisfaction [9,10]. Following recent international recommendations [11,12,13], several countries have extensively developed alternatives to full-time hospitalizations for inpatients (AFTH) [14,15,16,17,18]. AFTH encompass ambulatory care, part-time hospitalizations (day or night care, part-time therapy centers and therapeutic workshops) as well as full-time care outside of inpatient settings integrated in the community, i.e., hospitalizations at home, stays in therapeutic apartments, stays in specially trained families, crisis centers and rehabilitation centers. As a result, several different kinds of staff can work in those alternatives. They include psychiatrists and other medical doctors, nurses, nursing auxiliaries, psychologists, physiotherapists, social and educational staff as well as administrative staff. In France, the development of AFTH is still limited [19], despite support from policy makers [20]. An assessment by the French National Court of Auditors has shown that the implementation of AFTH was slowed down by the resistance of health professionals [21]. This is possibly due to a lack of consensus regarding the benefits of AFTH among the different schools of thought in the mental health field [21].In parallel, international recommendations for mental health care have advocated for a reduction in the length of stay (LOS) in full-time hospitalization [22,23] as prolonged hospitalizations can result in isolation and loss of autonomy, and are unpopular among patients [13,24,25,26]. The development of AFTH could facilitate alignment with those international recommendations through two main mechanisms. First, previous international research has shown the benefit of AFTH, in particular in terms of increased quality of life, clinical outcomes, adherence to treatment, accessibility and continuity of care [27,28,29,30,31,32,33]. This suggests that AFTH have the potential to decrease patients’ severity of illness through increased quality of care. As a consequence, when their development is satisfactory, patients will only require full-time hospitalizations for a limited period of time. Second, it has been widely demonstrated that health care supply influences practice [34,35]. The lack of AFTH could therefore result in an increased length of stay (LOS) in full-time hospitalization when no satisfactory option is available at the end of a patient’s full-time hospitalization. Physicians in hospitals with more AFTH may be more inclined to discharge inpatients earlier because they know AFTH are available and can provide alternatives to patients in full-time hospitalization well enough to be discharged but not well enough to be sent home without further care [36].There is currently a dearth of research to assess if the development of AFTH does result in reduced LOS and previous work has advocated for more research in that field worldwide [37]. However, when studying the impact of AFTH on LOS, a wide range of factors, which may also be associated with LOS, should be considered. They include patient, health care provider and environment characteristics. Some of them can influence LOS through similar mechanisms as AFTH provided within psychiatric sectors. Indeed, some factors can impact the patient’s health status and readiness for discharge, in particular clinical factors such as diagnosis, symptoms severity and comorbidities [38,39,40,41] while some factors can be associated with early discharge such as the implementation of discharge planning or the availability of medical and social care in the community [38,41,42,43]. In addition, associations were shown between the LOS and patients’ demographics and socio-economic characteristics [38,40,41] as well as either institutional characteristics (such as specialization or teaching status) or organizational characteristics (such as number of beds) of the health care provider [38,44].In this context, the objective of our study was to evaluate whether the development of AFTH in French psychiatric sectors was associated with a reduction in the LOS in full-time hospitalization, taking into account the other factors potentially associated with the LOS.A retrospective study was carried out using the French national discharge database (Recueil d’informations médicalisé en psychiatrie, RIM-P) [45], which records all hospital stays and outpatient care contacts in psychiatric hospitals, the annual national survey on health care providers (Statistique annuelle des établissements, SAE) where hospitals report their activity in a declarative manner for the past year [46], and other databases non-specific to psychiatry.Psychiatric care in public and private non-profit hospitals performing public service duties in France is delivered by sectors. They are multidisciplinary teams enrolled and paid by a hospital in charge of providing care to the population of a given geo-demographic area, either through ambulatory care, part-time hospitalization, full-time hospitalization or full-time care outside of inpatient settings. The catchment areas of psychiatric sectors are relatively homogenous in size across France, except in overseas French territories where they are larger. Sectors therefore represent an optimal unit of analysis of variations in psychiatric care as they remain the cornerstone of the organization of public mental health care delivery in France both for inpatient and outpatient care. There are specific sectors—with differing organizations—for adult, child and adolescent, and forensic psychiatry. Given these elements and to ensure comparability, our study focused on psychiatric sectors at public and private non-profit hospitals that perform public services in mainland France and provide care for adult patients outside of forensic settings.We included patients in full-time hospitalization whose care was reported in the RIM-P for the year 2012 (most recent year available at the start of the study) and who were diagnosed with a mental disorder from Chapter V of the International Classification of Diseases, tenth revision (ICD-10) [47], excluding organic mental disorders (F00-F09), mental retardation (F70-F79) and psychological development disorders (apart from pervasive developmental disorders) (F80, F81-F83, F88-F89). This diagnosis scope corresponds to the scope of psychiatrists’ expertise in France and has been used in previous studies on mental health [48,49]. Patients with at least one diagnosis of mental disorder outside of this scope were excluded from the analysis.As the databases used for the study were not totally exhaustive for the year 2012 and in order to ensure data quality, we further excluded patients seen in sectors belonging to a hospital which: (i) did not report consistently its annual full-time inpatient activity and/or its number of psychiatric sectors in the RIM-P and the SAE databases; (ii) did not report its outpatient activity; or (iii) did not report data requested to assess the development of its AFTH as described below.There is no direct measure of the development of AFTH. One way to estimate it is to determine the share of human resources allocated to those alternatives out of the total human resources allocated to psychiatry in a given hospital. Human resources indeed represent 70% of the hospital budget for the treatment of somatic illnesses [50,51] and it is estimated that this percentage is even higher for the care of psychiatric disorders [52,53,54]. The development of AFTH was therefore estimated by the ratio of the number of full-time equivalents (FTEs) of staff working in departments providing alternatives to full-time hospitalizations over the total number of FTEs in the hospital to which each sector belonged, i.e., (total number of FTEs—total number of FTEs in full-time hospitalization)/total number of FTEs. All kind of AFTH provided by psychiatric sectors (ambulatory care, part-time hospitalizations and full-time care outside of inpatient settings) and all kind of staff (psychiatrists and other medical doctors, nurses, nursing auxiliaries, psychologists, physiotherapists, social, educational and administrative staff) were considered. The total number of FTEs and the total number of FTEs allocated to full-time hospitalization were extracted from the SAE database. On a given territory, where both inpatient and outpatient mental health care is coordinated by a single hospital, the FTEs employed in psychiatry and reported in the database are either allocated to full-time hospitalization or to AFTH. FTEs which are not employed in mental health services providing full-time hospitalization are de facto employed in services providing AFTH and even FTEs of administrative staff are reported based on the type of care provided by the service they belong to. In addition, considering the overall proportion of FTEs allocated to AFTH allows comparability of data across sectors by adjusting on their overall capacity.Our variable of interest, the LOS for each full-time hospitalization, was computed in number of days until discharge using the RIM-P 2012. This was done after obtaining the authorization to access this database from the French data protection authority (CNIL) in July 2013 (Decision DE-2013-077). No informed consent was required from patients as data from the RIM-P is entirely anonymized.In addition to the development of AFTH, three types of factors, potentially associated with LOS, were considered as adjustment factors: patients’ characteristics, health care providers’ characteristics and environmental factors.The patients’ demographic (age and sex) and clinical characteristics (diagnosis) were extracted from the RIM-P database. In accordance with previous research [55], ICD-10 codes were grouped together into broader diagnostic groups (see Table 1). As it is difficult to establish a diagnosis during a single care contact in psychiatry and as comorbidities are frequent [56,57], we considered all diagnoses present in the database for a given patient over the course of the year. To overcome the lack of data on the patients’ socio-economic characteristics in the RIM-P, we created a proxy deprivation index based on the patients’ residential zip codes. We used a validated composite index specifically developed for the French context, called the FDep. This index takes into account the median household income, the percentage of high school graduates in the population aged 15 years and older, the percentage of blue-collar workers in the active population and the unemployment rate, and does not include any health indicator that could lead to circularity [58,59].Characteristics of sectors and their related hospital were extracted from the SAE database and included legal status (public vs. private non-profit), specialization and participation in teaching activities, as well as organizational factors, such as number of full-time inpatient beds as an indicator of the size of the hospital [44,55,60]. Moreover, the mean value by sector of the patients’ characteristics described above (case-mix) can also have an influence on practice. For example, if a sector treats on average older patients than another sector, this might cause variations in practice. We therefore also considered sectors’ case-mix characteristics.Finally, environmental factors were extracted from administrative databases and census data [61,62,63,64,65] and calculated for the catchment area of each sector. Those catchment areas, defined as the geographic zone where the sector’s patients originate, were built for each sector after excluding zip codes corresponding to fewer than five patients to avoid bias resulting from a few isolated patients coming from long distances on an occasional basis. The catchment areas were constructed using a geographic information system (Geoconcept® software, Bagneux, France) to convert patients’ text-based zip codes found in the RIM-P into spatial data. Environmental factors included characteristics of the overall health status of the population and the level of urbanization as well as variables related to the supply of additional medical and social care outside of the scope of public psychiatry. Such variables were the availability of inpatient and outpatient psychiatric care provided by the private sector (private for-profit hospitals, self-employed community-based psychiatrists or psychologists and general practitioners) and through social care institutes (residential care or services for disabled individuals) in sectors’ catchment area.The characteristics of the study population were described either by the mean and standard deviation (SD) or by number (%).Variations in the LOS in full-time hospitalization and in the development of AFTH between psychiatric sectors were studied by calculating the mean, SD, median, interquartile range, and range of the LOS for each sector and of the development of AFTH for each hospital. A coefficient of variation (CV) [66], which measures the dispersion around the national mean, was computed together with the ratio between the 90th and the 10th percentiles of the distribution, which is less sensitive to outlier values [67].The association between the variations in the LOS and in the development of AFTH was then assessed through the calculation of the Spearman correlation coefficient.To study this association while adjusting for other factors potentially associated with the LOS, we conducted a multivariate analysis with the LOS in full-time hospitalization as a dependent variable. We carried out a natural logarithmic transformation to achieve a more normal distribution given the data skewness. To account for the nested structure of the data, we ran a multi-level model. It was possible to divide variability into four levels: stay, patient, psychiatric sector and hospital. The reliability of hierarchical models however depends on the number of groups [68] and the number of observations per group [69,70,71,72,73,74]. As the mean number of stays per patient and the mean number of psychiatric sectors per hospital were low, two levels were considered: stay/patient level (level 1) and sector/hospital level (level 2).The development of AFTH was introduced in the model as an explanatory variable as well as the patient, psychiatric sector and environmental characteristics associated with the LOS in the bivariate analyses at a significance level of 0.20 or for which there were strong hypotheses on their association with the LOS. When variables were highly correlated, only one of them was kept in the model based on the strength of association with the LOS and clinicians’ advice. Characteristics of patients were attributed to each of their stays and characteristics of hospitals (in particular the level of development of AFTH) were attributed to each of their sectors, according to the approach used by previous research [75].To confirm the existence of a random effect at the sector level, we first ran a null model without any explanatory variables (model 1). Second, we introduced the patients’ characteristics (model 2). Third, we added the variables calculated at the sector level (characteristics of the sectors and their environment) (model 3). For each model, we calculated the intraclass correlation coefficient (ICC), which is the proportion of variance that is accounted for by the centre level (i.e., psychiatric sectors), and the proportional change in variance (PCV) to determine the proportion of variance explained by each type of explanatory variable [76]. Finally, we interpreted the value of the estimated regression coefficient associated to the level of development of AFTH (β1) after retransforming the coefficient based on the logarithmic transformation of the dependent variable: %∆ LOS = 100*(e^β1 − 1) [77].We used a statistical significance level of 0.05 and the analyses were performed using SAS software version 9.3 (SAS Institute Inc., Cary, NC, USA).Of the 248 public and private non-profit hospitals participating in public services in mainland France, discharge data from 122 hospitals (49.2%) were included in the analysis based on data quality. These hospitals consisted of 413 sectors of adult psychiatry (see Figure 1) representing 51.4% of all sectors of adult psychiatry in mainland France. Included hospitals did not present any statistically significant differences with excluded ones in terms of main organizational and institutional characteristics or case-mix.107,668 patients, matching our diagnostic criteria, were treated in full-time hospitalization in the selected sectors. They accounted for 182,230 stays in full-time hospitalization over the study period and represented 52% of all patients within our diagnosis scope seen in adult psychiatric sectors. The mean age of patients was 46 years (±16) and 54% were female. The two most common diagnoses were mood disorders not including bipolar disorders (27%) and schizophrenia (21%). The mean LOS in full-time hospitalization was 37 days (±72).Considerable variations between psychiatric sectors were observed both for the LOS in full-time hospitalization and the development of AFTH. The overall mean full-time hospitalization LOS by sector was 36 days and it ranged from 11 to 247.9 days between sectors with a coefficient of variation reaching 60%. These variations were not only a result of sectors with extreme LOS as the ratio between the 90th and the 10th percentiles of the distribution was superior to three (Table 2). The mean value of the ratio of FTEs allocated to AFTH out of the total number of FTEs by hospital amounted to 0.34 (±0.11) and varied between 0.08 and 0.66 among hospitals with a coefficient of variation reaching 33% and a ratio between the 90th and the 10th percentiles of the distribution close to 3 (Table 2).In the bivariate analysis, a decrease in full-time hospitalization LOS was observed when the level of development of AFTH increased. However, this association was not statistically significant (ϱ = −0.07; p = 0.08).In the multivariate analysis, we introduced ten individual patient characteristics at level 1. At level 2 we introduced three case-mix characteristics, five institutional or organizational characteristics (in addition to the development of AFTH) of psychiatric sectors, six characteristics of the overall health status of the population in the psychiatric sectors catchment area, as well as eight variables related to the availability of medical and social care in the catchment area (Table 3).The null model confirmed the existence of a significant centre effect and the necessity to take into account the nested structure of the data (variance = 0.22, p-value < 0.0001). Thirteen percent of the total variation in the LOS was related to practice differences between sectors (inter-sector variations) while 87% resulted from differences within sectors linked to case-mix (intra-sector variations) (Table 4).Patients’ individual characteristics explained 33% of the variations between sectors while sectors’ characteristics explained less than 20% of those variations. The level of development of AFTH was significantly and negatively associated with the LOS (p = 0.0493) (Table 5). For each 10% increase of the level of development of AFTH, the LOS in full-time hospitalization decreased by 3.4% (Table 5) when all other patient, health care provider and environmental characteristics were held constant.Some adjustment factors were also significantly associated with the LOS in full-time hospitalization, in particular characteristics of patients and of the available supply of health and social care outside of public psychiatry. Patient’s age and a female gender were positively associated with the LOS in full-time hospitalization. The presence of a diagnosis of schizophrenia and other psychotic disorders as well as a diagnosis of bipolar or other mood disorders was also associated positively with the LOS. In addition, even if there was not a linear association between patients’ deprivation and LOS, patients in the first and third quintiles of the deprivation index had significantly longer LOS than the most deprived patients (patients in the fifth quintile of the deprivation index). Similarly, some characteristics of patients’ case-mix at the sector level (mean age and percentage of patients suffering from anxiety disorders) were associated with LOS. A positive association was also found for the number of inpatient beds of private psychiatry in the catchment area while the capacity of housing and social rehabilitation centres in the catchment area was negatively associated with the LOS. Among sectors institutional and organizational characteristics, only the level of development of AFTH was significantly associated with the LOS. Finally, the inpatient psychiatric admission rate in the catchment area was negatively associated with LOS while an overall poor somatic health status of the population, approximated by the acute admission rate for somatic disorders for the population in a sector’s catchment area, was positively associated with the LOS (Table 5).Considerable variations were observed in full-time hospitalization LOS between psychiatric sectors in France. While the majority of those variations resulted from different patients’ characteristics, our results show a significant negative association between the LOS in full-time hospitalization and the development of AFTH after adjusting for a broad range of factors. Our findings are consistent with studies carried out in other local contexts which found that the development of AFTH was associated with a reduced use of inpatient services [28,32]. The considerable variations in the LOS between psychiatric sectors observed in France are also of the same order of magnitude as those underscored by a study carried out on depressed patients in 107 medical centres in the US, which showed that there was a fourfold difference in the mean LOS between the medical centres with the shortest and the longest LOS [78].Our results provide the first evidence of the benefits of developing AFTH in the French setting as the reduction of the LOS in full-time hospitalization is in alignment with international recommendations for mental health care [13,22,23,24]. They suggest that the development of AFTH can significantly benefit the quality of mental health care in France, taking into account the strong influence of patient and environmental characteristics on LOS. Variables related to the availability of private medical care on sectors’ catchment area (general practitioners community-based private psychiatrists and psychologists or private psychiatric hospitals), which complement the supply of public care provided by psychiatric sectors, were not associated with the LOS in our multivariate analysis. Our main hypothesis is that patients suffering from mental disorders which are seen in private care are not the same patients as those treated in sectorized public psychiatry. They might for instance be wealthier populations (as out-of-pocket costs are higher for private care). This could explain why the availability of private medical care would not influence the LOS of patients seen in psychiatric sectors while the development of AFTH within psychiatric sectors would.One of the main strengths of our study is the cross-referencing of different data sources, which allowed us to adjust our analysis on a broad range of patient, psychiatric sector and environmental characteristics, using multi-level modelling that has been employed increasingly for the study of factors associated with the LOS in psychiatry [42,55,79,80]. In addition, we did not focus on a limited geographic area but carried out an analysis at the national level which facilitates the generalizability of our results. Additional studies using multiple consensual indicators of quality of care and/or a longitudinal design can now be carried out. Those first findings would in particular be usefully complemented by research aimed at disentangling the mechanisms underlying the impact of AFTH on LOS in the French context. In addition, further work focusing on other variables such as readmission and involuntary admission rates whose reductions are also supported by international recommendations for mental health care [13] will be developed.The results of our study should however be interpreted in consideration of several limitations linked to the retrospective use of administrative databases, which are considered less accurate than prospective studies, but are the most cost-effective way to gather data on a national scale and allow non-invasive data collection for patients [81]. The RIM-P database was first implemented in 2006 and its limitations have been mentioned previously [21], but it is estimated that since 2010, data quality is sufficient for research purposes [82]. Moreover, whenever possible, we compared data from the RIM-P with data from other databases, such as the SAE to ensure data quality.However, there are limitations linked to data comprehensiveness as some psychiatric sectors belonging to hospitals where FTEs data were not available could not be included in the analysis, which might limit the external validity of our results. Furthermore, there are some limitations linked to data precision: the share of FTEs allocated to AFTH was calculated on all types of staff while they might not be completely equivalent. In addition, no information was available in the SAE database regarding the distribution of FTEs between the different types of alternatives to full-time hospitalization, while some studies have shown that the impact on the use of inpatient services could vary with the type of AFTH considered [32]. Similarly, there was no information available on the distribution of FTEs between the different forms of care at the psychiatric sector level, while there could be differences between the sectors belonging to the same hospital even if they follow the same general policy.We were also limited by data availability for the inclusion of adjustment factors potentially associated with the LOS, which could account for some of the unexplained variations in LOS remaining in our multivariate analysis. First, regarding patients’ characteristics, there was no information in the RIM-P database on either symptom or illness severity nor on the socio-economic situation of individual patients. To address this latest limit, we used a deprivation index at the level of patient zip code of residence, following approaches adopted by previous research [83,84,85,86,87,88,89]. Such proxies present the advantage of being mobilizable and operational quickly at a large scale (the national scale in our case) and their use was recently recommended by the French High Council for Public Health whenever individual data are lacking in available information systems [90]. Those limitations result from the use of administrative databases, which however allow the conduction of analyses on a large number of patients (more than 100,000 in our study).Second, regarding organizational sectors’ characteristics, data was not available on implementation of discharge planning in psychiatric sectors.Finally, regarding environmental factors related to the availability of medical and social care on the catchment areas of psychiatric sectors, we used data directly available in administrative databases which only provide information relating to a limited number of structures. Detailed mapping of all relevant structures at the local level, similarly to what has been developed in other countries [91,92], should be added to future research to describe with more precision services availability as there are strong hypotheses on its association with the LOS [38,42].This study provides the first evidence of the benefits of the development of AFTH at the national level taking into account the particularities of the French system and will provide a lever for policy makers to further its development. One of its main strengths is the cross-referencing of different data sources which allowed us to adjust our analysis on a broad set of patient, psychiatric sector and environmental characteristics. These first results should be supplemented with a focus on other endpoints that may also be impacted by the development of alternatives to full-time hospitalization and are frequently associated with the quality of care, such as unplanned readmission and involuntary admission rates, as well as by studies focusing on other aspects, such as treatment adherence or reduction of caregiver burden. In addition, qualitative research involving the different actors of mental health could usefully add to those first findings.We are grateful for the strong support of Morgane Michel and for her proof-reading of the manuscript. We would also like to acknowledge two anonymous reviewers for their helpful comments. Finally, we would like to thank the policy makers, health professionals, database managers and other researchers working in mental health services research who participated in the monitoring committee of the call for proposals, which funded this research, and who provided on-going advice on the project. This study was indeed funded by the French Ministry of Health in the framework of the 2013 Mental health and psychiatry call for proposals of the direction of research, study, evaluation and statistics (DREES). The funding source had no role in the conception and design of the study, in the acquisition, analysis and interpretation of data or in the writing of the manuscript.C.G. participated in the conception and design of the study, in the acquisition, analysis and interpretation of data and wrote the manuscript; J.G. participated in the acquisition, analysis and interpretation of data; J.T. participated in the acquisition and analysis of data; J.-M.M. participated in the conception and design of the study and in the acquisition of data; J.-L.R. participated in the conception and design of the study and in the interpretation of data; K.C. participated in the conception and design of the study, in the interpretation of data and in the drafting of the manuscript. All authors give final approval of the manuscript submitted.The authors declare no conflict of interest.Flowchart for the selection of included sectors based on data quality.Diagnostic groups included in the study.Variations of the LOS in full-time hospitalization and of the development of AFTH between psychiatric sectors.CV: coefficient of variation; AFTH: alternatives to full-time hospitalization; FTE: full-time equivalent.Explanatory variables introduced in the multivariate analysis in addition to the level of development of AFTH.* The number of full-time inpatient beds per 1000 inhabitants of the catchment area was highly correlated with the total number of sectors per hospital (ρ = 0.90; p < 0.0001) and we therefore only introduced the number of beds in the model.Estimation of random effects.ICC: intraclass correlation coefficient.Estimation of fixed effects in the final model (model 3).

Med-MDPI/ijerph_2/ijerph-14-03-00327.txt

@@ -0,0 +1 @@
+Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).Given that public transportation networks are less developed in rural than in urban areas, a lack of accessibility to dental care facilities could be a barrier to routine dental checkups. Thus, we hypothesized that the distance to the dental care facilities is a risk factor for tooth loss. The aim of this study was to test whether there is an association between the distance to dental care facilities, estimated by geographic information systems, and number of teeth, assessed by an oral examination, among elderly residents of a rural area in Japan. Data were collected in 2016 from a cross-sectional study conducted in Shimane prefecture, Japan. After excluding participants with missing data (n = 21), we analyzed data from 710 participants. Of them, 40.6% were male and the mean (standard deviation) age was 67.4 (7.4) years. Further, 68.0% (n = 483) had at least 20 teeth. We found that the distance to dental care facilities was significantly associated with the number of teeth (less than 20) (odds ratio = 1.07, 95% confidence interval = 1.01–1.12) after adjustment for potential confounders. This result suggested that individuals without easy access to dental care facilities may be important targets for dental care.Previous studies have indicated that tooth loss in elderly populations is associated with mortality [1,2,3]. Japan is an aging society and the percentage of the elderly in 2060 will reach 39.9% [4]. Maintaining dental health in an aging population is a major public health concern in Japan.Several risk factors affecting tooth loss have been suggested, including sociodemographic factors (e.g., age and educational attainment) [5,6], lifestyle and health factors (e.g., nutritional status, smoking habits, and current history of disease) [5,6,7,8,9], and oral health conditions and behaviors (e.g., complaint of oral condition and dental care visits) [10,11]. Distance to dental care facilities is also a potential factor for tooth loss, especially in rural areas [12]. Given that public transportation networks are often less developed in rural compared to urban areas, a lack of accessibility to dental care facilities could be a barrier to routine dental checkups. We hypothesized that the distance to dental care facilities is a risk factor for tooth loss, assessed by an oral examination. To the best of our knowledge, no previous studies have examined this association in a rural area of Japan.In Japan, a list of targets for Health Japan 21 (the second term) was published in 2012 [13]. This is a 10-year plan that began in 2013 and the government will work toward increasing the number of people with 20 teeth at the age of 80 years (8020 campaign) [13]. The aim of this cross-sectional study was to test whether there is an association between the distance to dental care facilities, estimated by geographic information systems (GIS), and the number of teeth (less than 20) among elderly residents in a rural area.Data was gathered from participants in the Shimane Center for Community-Based Health Research and Education (CoHRE) Study. The Shimane CoHRE study is a cohort study to examine the determinants of lifestyle-related diseases, including oral health, in rural areas in the southern part of Shimane prefecture, Japan [14,15,16,17,18]. The present study was conducted as a cross-sectional study of Shimane CoHRE study for which participants were recruited in 2016. 731 of the residents who were covered by the National Health Insurance and aged between 40 and 74 years of age in the town of Ohnan participated in the 2016 survey. After excluding participants with missing data (21 participants, 2.9%), we analyzed data from 710 participants. The Ethics Committee of the Shimane University School of Medicine approved the study protocol in 2016 (number 2227, 5/12). Written informed consent was obtained from all participants.Dental examination was conducted by a trained dental hygienist, with both the dental hygienist and participants in a seated position. The number of teeth was counted in the examination, and participants were divided into the following two categories: those with less than 20 teeth and with 20 or more teeth [13].The Geographic Information Systems software (ArcGIS, version 10.0, Environmental Systems Research Institute, Redlands, CA, USA) was employed for database queries and used to estimate distance to dental care facilities from the individuals’ addresses. Network analysis, which determined the shortest path between the participant locations and the dental care facilities, was performed on road networks.Age (years, analyzed as a continuous variable), gender (male vs. female), body mass index (BMI) (analyzed as a continuous variable), current smoker (yes vs. no), current alcohol drinker (yes vs. no), with regular physical activity (engaged in regular physical activity = yes vs. not engaged in regular physical activity = no), medication for disease treatment (medication against hypertension, diabetes mellitus and hyperlipidemia, yes vs. no), receiving a dental health check within the past one year (yes vs. no), having any oral health problems (yes vs. no), having enough sleep (yes vs. no), elevation estimated by the GIS (median value, ≤258 m vs. >258 m), and accessible transportation (driver = yes vs. non-driver = no). Accessible transportation was assessed by the following question: “Do you have a valid driving license and regularly drive a car?” (if yes = driver, if no = non-driver).The χ2 and t-tests were used to compare characteristics according to the number of teeth (less than 20 vs. 20 or more). Multivariable logistic regression model was performed to derive odds ratios (ORs), 95% confidence intervals (95% CIs), and p-values. p-values less than 0.05 were considered statistically significant. All statistical analyses were performed using IBM SPSS Statistics 20 (IBM Corporation, Tokyo, Japan).The characteristics of the study participants are shown in Table 1 by number of teeth. There were statistically significant differences between the two groups (less than 20 teeth vs. 20 teeth or more) in the distance to dental care facilities, age, gender, medication (hypertension), and car driver. On the other hand, there were no statistically significant differences in current smoker, current alcohol drinker, regular physical activity, medication (diabetes mellitus and hyperlipidemia), BMI, elevation, having enough sleep, use of dental health checks, and having any oral health problems.Table 2 shows the results of the multivariable logistic regression analysis. The distance to dental care facilities was significantly associated with the number of teeth (less than 20) (OR = 1.07 per km, 95% CI = 1.01–1.12). The factors of age, non-smoker, and non-driver were also significantly associated with the number of teeth (less than 20) (OR = 1.19, 95% CI = 1.14–1.25, OR = 0.47, 95% CI = 0.25–0.90, and OR = 1.88, 95% CI = 1.12–3.13, respectively).Although a previous study conducted in a rural region of the United States examined the association between the distance to care facilities and care visits [12], no studies have been performed on the potential effects of distance to dental care facilities, estimated by GIS, on the number of teeth, assessed by an oral examination, in a rural area of Japan. Our results showed that the distance to dental care facilities increased the OR of the number of teeth (less than 20) (OR = 1.07, 95% CI = 1.01–1.12), independently of sociodemographic factors, lifestyle, and oral health conditions. The present result is consistent with our previous study, indicating that residential location influences health conditions; a cross-sectional study conducted in a rural area found that distance from a city center affected the incidence of hypertension [17]. Generally, health care facilities in a rural area are clustered at the center of the town. Although the average distance to dental care facilities was 3.8 km for the participants in this study, this may be too far for the elderly residents. Thus, further studies are required to examine the burden of accessibility to dental care facilities by distance.A previous study pointed out that transportation available for residents should be considered when discussing the association between distance to health care facilities and residents’ health [12]. Although information about modes of transportation to dentists (e.g., use of public transportation or of a ride offered by neighbors or family members) was not available in this study, we included the driving status of participants which might be an important determinant of a moving range in a rural area [12]. In our analysis, the distance to dental care facilities was associated with the number of teeth, independently of the driving status. We also tested the logistic regression model without a driver. As a result, the OR for the distance to dental care facilities was a similar value (OR = 1.06, 95% CI = 1.01–1.06, p = 0.016). Although more research is needed to examine reasons for why similar associations were shown, these results implicated that another explanatory factor (e.g., socio-economic status) should be considered to explain the association between accessibility to dental care facilities and the number of teeth. For example, previous studies revealed that lower socio-economic status is associated with the utilization of dental care and oral health behavior [19,20]. In addition, locational differences in attitude about dental care may account for the residential variation in the use of dental care [21].The present study has several strengths. To our knowledge, this is the first study to examine the potential association between the distance to dental care facilities, estimated by GIS, and the number of teeth, assessed by an oral examination, in a rural area of Japan. GIS is a computer-based system that integrates and analyzes spatial data, including latitude and longitude, and its application to epidemiological research has been increasing in recent years [22]. This approach contains less measurement noise than the information evaluated by respondents’ responses to questions as to the distances, because GIS can estimate the distance between population locations and health care facilities based on road network information. Furthermore, the number of teeth was counted by dental hygienists rather than relying on self-reported data. On the other hand, there are also a number of potential limitations in the current study. First, due to the cross-sectional study design, it is difficult to argue the causal relationship between independent and dependent parameters. Second, our results could be explained by other unmeasured risk factors for tooth loss (e.g., socio-economic status and dietary intake). This is particularly evident in socioeconomic factors. Previous studies revealed that a low education level was associated with tooth loss [6,23]. The reason is that education level is associated with the utilization of dental care and oral health behavior [19,20]. Our data could not evaluate differences in socio-economic status between the center of the town and remote areas, so further studies are required to examine this issue. Third, our data did not include accurate information on years of smoking and volume of alcohol intake. Fourth, our analyses could not consider the effect of residence year. Finally, misclassification may have occurred in the self-reported data as a consequence of recall errors. The distance to dental care facilities was a significant risk factor for tooth loss, independently of sociodemographic, lifestyle and behavioral factors. Those who reside in a location far from dental care facilities may be important targets for dental care.This work was supported by the Taiyo Life Welfare Foundation and MEXT KAKENHI (Grant Number 15H05365).Tsuyoshi Hamano, Miwako Takeda, Kazumichi Tominaga, and Kristina Sundquist conceived and designed the experiments; Tsuyoshi Hamano and Miwako Takeda analyzed the data; Kazumichi Tominaga and Toru Nabika contributed reagents/materials/analysis tools; Tsuyoshi Hamano and Miwako Takeda wrote the paper. Tsuyoshi Hamano and Miwako Takeda contributed equally.The authors declare no conflict of interest. The funding sponsors had no role in the design of the study, in the collection, analyses, or interpretation of data, in the writing of the manuscript, and in the decision to publish the results.Characteristics of study participants.SD, standard deviation.Multivariable logistic regression analysis with the number of teeth as the dependent variable.Independent variables were coded as follows: gender (0 = male, 1 = female), current smoker, current alcohol drinker, medication for disease treatment, regular physical activity, car driver, having enough sleep (0 = yes, 1 = no), use of dental health checks and having any oral health problems (0 = no, 1 = yes), and elevation (0 = ≤258 m, 1 = >258 m). Note that 0 as the reference category. OR: odds ratio; 95% CI: 95% confidence interval.

Med-MDPI/ijerph_2/ijerph-14-03-00328.txt

@@ -0,0 +1 @@
+Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).The aim of the present study was to obtain the prevalence of malocclusions in preschool children in Shanghai, China. A cross-sectional survey was conducted among 2335 children aged 3–5 years from kindergartens. Several occlusal parameters were clinically assessed, including second deciduous molar terminal plane, canine relationship, degree of overjet and overbite, anterior and posterior crossbite, and the presence or absence of physiologic spaces and crowding. All parents of subjects were asked to fill in the oral health knowledge questionnaires. The prevalence of malocclusion in primary dentition in Shanghai was 83.9%, and no significant differences were found in genders. Data showed that the prevalence of deep overbite (63.7%) was the highest in children with malocclusion, followed by deep overjet (33.9%), midline deviation (26.6%), anterior crossbite (8.0%) and anterior crowding (6.5%). The results revealed a high prevalence of malocclusion in primary dentition in children aged 3–5 years old of Shanghai, especially in vertical anomalies. The need for preventive orthodontic therapy is extremely desired and oral health education about malocclusion should be strengthened.As an improving quality of living, people are hoping to own an aesthetically pleasing appearance. This evolution has driven many industries to satisfy their clients’ aesthetic needs, including dentistry. The rapid development of China’s economy has resulted changes in diet, making people consume more refined food. However, this dietary change results in insufficient jaw growth [1]. Malocclusion is a disorder of the craniofacial complex that affects the development of dental maxillofacial region and masticatory function [2]. Serious malocclusion may cause both psychological and physiological conditions. Therefore, it is important to find out the incidence of various malocclusions and corresponding methods to prevent or correct them.Early intervention for children in, or before, the peak of growth and development can reduce not only the prevalence of malocclusion or the severity in permanent dentition, but also the psychological impact. A number of studies had investigated the prevalence of malocclusions in the primary dentition in different countries and populations, with prevalence values ranging from 21.0% to 88.1% [3,4,5,6,7,8,9,10]. A study about Chinese people from 1956 to 1960 showed the prevalence ranging from 29.33% to 48.87%. A national survey conducted by Chinese Stomatological Association (CSA) in 2000 concluded the prevalence of malocclusion as 51.84% in Chinese children [11]. Some studies suggested that malocclusions were also related with bad oral habits, such as mouth-breathing and non-nutritive sucking habits [1,12].The aim of the present study was to evaluate the prevalence of malocclusion in the primary dentition and bad oral habits of preschoolers in the city of Shanghai, in order to provide an epidemiological reference for the development of early intervention and prevention of the occurrence of malocclusion. A multistage, stratified sampling method was applied to obtain a representative sample of preschoolers, and we selected four districts (Hongkou District, Putuo District, Pudong District, and Minhang District) by probability proportional to size sampling (PPS) (Figure 1). The sample was composed of 2335 children (1247 boys and 1088 girls) aged 3 to 5 years from 12 kindergartens. The included children were studied in the kindergartens which were sampling surveyed and we also obtained their parents’ or guardians’ informed consent before examination was initiated. The exclusion criteria were the presence of permanent teeth, loss of any primary teeth, dental caries that affected the judgment, orthodontic treatment history, tooth agenesis, and other congenital malformation (such as cleft lip/palate) or severe illness and children unable to cooperation. The survey was conducted from January to June, 2016. This study was approved by the Ethics Committee of Shanghai Stomatological Hospital (2015-0012).The investigation was composed by an anamnestic questionnaire and oral examination measurements without radiograms, which were mostly based on the WHO basic methods for conducing oral health surveys [13].The questionnaires were completed by parents under the dentists’ instruction. The first section was about general information, such as age and gender of the child. The second section contained ten questions about the children’s oral habits and parents’ awareness of oral health.The oral examination was carried out by five calibrated trained orthodontic dentists. A pilot study on 50 children was conducted before beginning the present investigation to ensure the accuracy of diagnosis and to standardize the procedures, and substantial inter-examiner reliability was found (Kappa agreement value >0.9). The children were examined in schools’ infirmaries. Each child was checked with a pair of disposable latex gloves and a disposable mouth mirror.Following items were included in the oral examination:Deciduous canine relationship: Equal to Angel’s classification. The canine relationship was recorded as class II or class III, if it was class I on one side and class II or class III on the other. Children with class II canine relation on one side and class III on the other side were recorded as mixed.Terminal plane relationship of the second primary molars: The relationship of the distal surface between the upper and lower second deciduous molar including three types (flush type, mesial type and distal type). The relationship of molars and canines were recorded on the basis of bilateral occlusion.Maxillary overjet: This was measured from the palatal surface of the mesial corner of the most protruded maxillary incisor to the labial surface of the corresponding mandibular incisor. (0 mm: edge-to-edge; >3 mm, ≤5 mm: mild; >5 mm, ≤8 mm: moderate; >8 mm: severe).Mandibular overjet (anterior crossbite): This was recorded when one or more of the maxillary incisors or canine occluded lingual to the mandibular incisors.Overbite: This was graded according to coverage of the mandibular incisor by the most protruded fully erupted maxillary incisor. (<1/2: normal; >1/2, ≤3/4: mild; >3/4, <1: moderate; all cover: severe).Open bite, anterior (<3 mm: mild; >3 mm, ≤5 mm: moderate; >5 mm: severe).Posterior crossbite: This was recorded when one or more of the maxillary primary molars occluded the lingual to the buccal cusps of the opposing mandibular teeth.Scissors bite: This was recorded when one or more maxillary primary molars occluded the buccal to the buccal surfaces or the lingual to the lingual surfaces of the corresponding mandibular teeth.Midline displacement.Crowding (anterior, posterior): >0, ≤2 mm: mild; >2 mm, ≤4 mm: moderate; >4 mm: severeSpacing: >0, ≤2 mm: mild; >2 mm, ≤4 mm: moderate; >4 mm: severeDental arch shape: triangular; U-shape; square-shapeTonsil: normal; antiadoncus I°; antiadoncus II°; antiadoncus III°Temporomandibular joint disorderNasal ventilationMandibular plane angleAnterior crossbite, posterior crossbite, deep overbite (>1/2), deep overjet (>3 mm), anterior open bite, anterior edge-to-edge, posterior scissor bite, and crowding (>2 mm) all indicated malocclusion. The preschool children who exhibited at least one of these conditions were classified with malocclusion.Data were recorded in a spreadsheet computer program (Microsoft Excel 2010, Microsoft Corp., Redmond, WA, USA). SPSS 22.0 software (SPSS, Chicago, IL, USA) was used for analyses. The results of intra‑examiner reliability were tested using the kappa agreement statistic method.The prevalence of malocclusion was reported by age and gender, and in total. The chi-square test was applied to determine the statistical associations between the independent variables and the malocclusion variable. For all tests, significant difference was assumed when the p value is < 0.05. The clinical registrations were based on the method evolved by the Angle’s classification, which has been used in many studies [14].The present study showed that 16.1% of children had dentitions without any irregularity and 83.9% had different degrees of anomaly (Table 1). There was no significant difference found in genders. Data showed that prevalence of deep overbite (63.7%) was the highest in children with malocclusion, followed by deep overjet (33.9%), midline deviation (26.6%), anterior crossbite (8.0%), and anterior crowding (6.5%) (Table 2, Table 3, Table 4 and Table 5).The study revealed that the most common molar relationship at the 3–5 years of age was the flush terminal plane (38.7%), followed by mesial step (38.5%) (Table 2). With respect to the canine relationship, the normal type was observed as 57.0%, and the distal type as 32.4% (Table 2).There were 63.4% children with bad oral habits, 32.7% of them had sucking habits, and 48.5% parents had no awareness about orthodontic treatments (Figure 2).The results showed that prevalence of malocclusion from 3 to 5 years old was 83.9%, which is much higher than 51.84% for children with primary dentition in China [11]. The prevalence was also different from that reported in studies which were carried out in different countries, such as 26.0% reported in India and 42.0% to 74.7% in Germany [6,7,15]. These differences may be due to the different methodology used by the authors, or different subjects and decades. Race, living environment, and eating habits were different in the various regions, which may affect the incidence of malocclusion.Our study on the Shanghai population showed that distribution of flush terminal molar relation was 38.7% on both side. A study by Infante pointed out that the distal step molar relationship decreased with the increase of age [16]. Other studies by Nanda et al. and Ravn indicated that the distal step molar relationship was invariably maintained throughout the primary dentition stage and always transferred unchanged to the permanent dentition [3,17]. The research done by Ravn was a longitudinal study, to ensure the result was more reliable. With regard to the flush and mesial terminal plane, Onyeasoet et al. found out most of them developed into Angle class I in the permanent dentition [18]. The present study was a cross-sectional study which inevitably imposed limitations on the estimation. Further longitudinal studies are needed to obtain the changes in occlusal pattern from the deciduous dentition to permanent dentition.The prevalence of the class I canine relationship in our study was 57.0%. Children with a class II canine relationship reached 32.4%, which was much lower than 45% in British children [19], but was similar to 31.6% in the Danish children [17]. The difference could be caused by small sample size in the former study, which may enlarge the sampling error to misunderstand the actual situation.Table 2 suggested that the two more prevalent types of anomalies were dental space and deep overbite, which is consistent with previous studies [14,20]. Primate space and leeway space are normal in deciduous dentition. A study published by Center of Human Development at the University of Michigan showed the sum of mesiodistal diameters of primary teeth was 6, shorter than that of permanent teeth in the maxillary [21]. The permanent dentition may be crowded if there are no spaces in the deciduous dentition. However, this theory is questionable since Baume showed that nine of 16 individuals with no interdental spaces in the primary dentition did not exhibit crowding in the permanent dentition [22]. This indicates that leeway space does not necessarily solve the problem of crowding. More longitudinal studies should be conducted to determine this in the future. Hence, dental space was temporarily not classified as malocclusion in our study. The prevalence of crowding (6.5%) was much lower than in Colombia (52.1%) [14]. This may be due to the cutoff value of the latter article being more than 0 mm.As interdental space was removed from malocclusion in primary dentition, the most prevalent type of malocclusion became deep overbite (63.7%), followed by deep overjet (33.9%). These results were similar to those reported in previous studies [7,23,24,25]. The prevalence of deep overbite was high in deciduous dentition and increased to the late mixed dentition, which may be explained by the common use of extraction of deciduous molars, a procedure that will usually result in collapsed dentition. Full eruption of the premolars and second molars could stabilize the occlusion, and the prevalence of deep overbite may decrease in the permanent dentition. During craniofacial growth, the mandible will rotate in a backward direction [26], while the overjet will decrease.In the present study, 33.9% of the children showed deep overjet, which was higher than 29.7% in Brazil [27] and 26.0% in Finnish subjects [23]. These differences may be caused due to the use of different methodologies by the authors, who considered an accentuated overjet to be greater than 3 mm, in comparison to the 2 mm used in the present study for the determination of this condition.The prevalence of anterior crossbite was 8.0% in the present study, which was more than the Saudi (1.7%) and the British (1.0%). However, it was similar to the prevalence in the Finnish (8%) [28] and African-Americans (5%) [29]. Previous studies on Americans and Europeans indicated the incidence of posterior crossbite in the primary dentition ranging from 7.2% to 20.81% [5,7]. Another study showed that it was one of the most prevalent malocclusions in the primary and early mixed dentitions [30]. However, in the present study, the prevalence of posterior crossbite was much lower, which meant only 6 in 2334 children had posterior crossbite. Our result was similar to the result found in India (0.4%) [31]. It observed that Caucasians generally showed higher incidence rate of posterior crossbite than Africans and Asians [17,32,33]. The different prevalence of posterior crossbite between different regions may be caused by the difference in prevalence of sucking habits. Three studies on posterior crossbite associated this alteration to finger-/dummy-sucking habits which, in the present study, was 32.7% [5,25,34]. The children who adopted such habits tend to have a greater chance of exhibiting posterior crossbite than those that did not. However, research shows the scientific evidence could not confirm what type of malocclusion is associated with non-nutritive sucking habits [35].Malocclusion not only destroys aesthetics, but also creates functional problems. Studies suggest that deep overjet and anterior open bite were predisposing factors of dental trauma [36,37,38]. Young children start to crawl, walk, run, and fall, and take up high-risk activities when they grow up, such that dental injuries and dislocated teeth are common [39]. Cross-bite is unlikely to lead to the development of oral disease, but dysfunction can arise from the resulting impairment of mastication [40]. The definition of early treatment it a treatment which is started in the primary or mixed dentition to enhance the dental and skeletal development before the eruption of permanent dentition [41]. These early therapeutic methods are usually brief and simple, which elicits little cooperation from patients and their parents. Early treatments could prevent the malocclusion from worsening and greatly simplify subsequent orthodontic treatment. Malocclusion caused by bad oral habits, such as open bite caused by tongue protrusion, can be corrected by tongue crib appliance [42]. It seems that some early interventions are needed in order to prevent the malocclusion from worsening and obtain a well-balanced dental and skeletal development.The present study indicated a high prevalence of malocclusion for 3 to 5-year-old children in Shanghai, which should be taken seriously. However, it is also important to construct a more precise definition of primary dentition malocclusion such that common standards are defined for further studies. The change of malocclusion with increasing age could not be indicated due to our cross-sectional research, and a further longitudinal study is needed to determine this.The present study offers evidence that malocclusion is a remarkable problem in Shanghai’s preschool children. The prevalence of malocclusion was high among the children analyzed and increased significantly in recent 17 years, suggesting that malocclusion is a public health problem worthy of note in the Chinese population. This work was supported by grants from the National Natural Science Foundation of China, 81271192 (2012). The authors acknowledge all the doctors involved in this investigation and 12 kindergartens for their cooperation. Yuehua Liu and Ying Zhang conceived and designed the experiments; Xinhua Zhou, Yan Wang, and Li Chen performed the experiments; Xinhua Zhou completed data statistics; Hao Zhang analyzed the data; Xinhua Zhou wrote the paper.The authors declare no conflict of interest.The locations where the study took place.Composition ratio of inadequate oral habits at each age.Descriptive analyses of demographic characteristics of sample.Chi-square test: p > 0.05.The composition and prevalence of sagittal occlusal characteristic.Mix 1: Child with class II canine relation on one side and class III on the other side was recorded as mixed.The composition and prevalence of vertical anomalies.The composition and prevalence of transversal anomalies.The composition and prevalence of space discrepancies.

Med-MDPI/ijerph_2/ijerph-14-03-00329.txt

@@ -0,0 +1 @@
+Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).In this study, a chlorine dioxide solution (UC-1) composed of chlorine dioxide was produced using an electrolytic method and subsequently purified using a membrane. UC-1 was determined to contain 2000 ppm of gaseous chlorine dioxide in water. The efficacy and safety of UC-1 were evaluated. The antimicrobial activity was more than 98.2% reduction when UC-1 concentrations were 5 and 20 ppm for bacteria and fungi, respectively. The half maximal inhibitory concentrations (IC50) of H1N1, influenza virus B/TW/71718/04, and EV71 were 84.65 ± 0.64, 95.91 ± 11.61, and 46.39 ± 1.97 ppm, respectively. A 3-(4,5-Dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) test revealed that the cell viability of mouse lung fibroblast L929 cells was 93.7% at a 200 ppm UC-1 concentration that is over that anticipated in routine use. Moreover, 50 ppm UC-1 showed no significant symptoms in a rabbit ocular irritation test. In an inhalation toxicity test, treatment with 20 ppm UC-1 for 24 h showed no abnormality and no mortality in clinical symptoms and normal functioning of the lung and other organs. A ClO2 concentration of up to 40 ppm in drinking water did not show any toxicity in a subchronic oral toxicity test. Herein, UC-1 showed favorable disinfection activity and a higher safety profile tendency than in previous reports.Chlorine dioxide, a strong oxidant, can inhibit or destroy microbes [1,2,3,4,5]. Studies have investigated the application of chlorine dioxide in numerous fields such as water or wastewater treatment, environment and food disinfection, and medicine [6,7,8,9,10,11,12,13]. Typically, chlorine dioxide is produced using either an acid-based or an electrolytic method [7,8,10,12]. In the acid-based method, chlorine dioxide is produced by mixing starting materials, such as sodium chlorite and hydrochloric acid, sodium chlorite and ferric trichloride, or sodium chlorite and chlorine gas. In the electrolytic method, the reactants are aqueous sodium chloride or saturated saline and sodium hypochlorite. According to the disinfectants and disinfection byproducts rule (DBPR) of the United States Environmental Protection Agency Microbial and Disinfection Byproduct Rules Simultaneous Compliance Guidance Manual [14], the maximum residual disinfectant level goals (MRDLG) and maximum residual disinfectant levels (MRDL) of chlorine dioxide are 0.8 mg/L [14]. The permissible exposure limits (PELs) for chlorine dioxide defined by the Occupational Safety and Health Administration are as follows: (a) General industry: 0.1 ppm and 0.3 mg/m3; (b) Construction industry: 0.1 ppm and 0.3 mg/m3 time weighted average (TWA); (c) American Conference of Governmental Industrial Hygienists threshold limit value: 0.1 ppm and 0.28 mg/m3 TWA; 0.3 ppm and 0.83 mg/m3 short term exposure limit (STEL); (d) National Institute for Occupational Safety and Health recommended exposure limit: 0.1 ppm TWA; 0.3 ppm STEL.The application of chlorine dioxide products or their contact with food or the human body is a serious issue if the products contain high levels of impurities. Impurities are mainly caused by impure reactants such as 10% H2SO4 and 15% NaClO2 or reaction byproducts such as Cl2 and chloroxy anion. For example, 10% H2SO4 and 15% NaClO2 contain 90% and 85% unknown impurities, respectively. The chlorine dioxide product obtained from a mixture of 10% H2SO4 and 15% NaClO2 is highly impure. The Cl2 product can react with organic matter to form trihalomethane, which is a carcinogen. Chloroxy anions, such as ClO2− or ClO3−, can be harmful to human health [15]. The domestic and industrial use of chlorine dioxide should be assessed according to product purity, for which the preparation method is an essential step. Low purity starting materials (e.g., 5% HCl and 10% NaClO2) have a high impurity content. If the product of these reactions is not further purified, then the chlorine dioxide products produced, which also contain high levels of impurities, are useful only for wastewater treatment and are unsuitable for contact with humans or food because of the harmful impurities. Therefore, a higher percentage of chlorine dioxide gas molecules must be obtained through further chlorine dioxide gas molecule purification.To increase the safety of chlorine dioxide solution, eliminating or reducing the impurities and increasing the gas chlorine dioxide concentration in a solution is a reasonable approach. Herein, a clean and concentrated process for chlorine dioxide gas production was designed and implemented. We produced a chlorine dioxide solution (UC-1) containing 2000 ppm chlorine dioxide gas in water through the electrolytic method. The solution was further purified with a film membrane, and subsequently dissolved in reverse osmosis (RO) water. UC-1 was investigated to determine its efficacy, and safety issues such as the antimicrobial activity, in vitro cytotoxicity, in vivo rabbit ocular irritation, in vivo inhalation toxicity, and in vivo subchronic oral toxicity were assessed.The UC-1 solution is produced in an apparatus, the technical details of which will be published later in the form of a patent application (PCT applied PCT/CN2016/080198; PCT applied PCT/CN2016/080199; PCT applied PCT/CN2015/099515; DE202016103175) by an electrochemical method. Briefly, sodium chloride solution was made from 99% (food grade) sodium chloride and RO water and pumped into the electrobath equipment. The electrolysis was operated by 6–12 V and 40–80 A current. After electrolysis, the ClO2 gas was mixed with water using a water-ClO2 mixer which was designed based on the Venturi effect. Mixing of water with ClO2 gas was continued by the cycle till the concentration of ClO2 in water was over 2000 ppm (Figure 1) and pH value was 2.2. The chlorine dioxide solution produced by this process is named as UC-1.The chemical composition of the UC1 solution was determined according to a standard method [16]. The following data were obtained: ClO2: 2120 ppm, free chlorine (Cl2): 882 ppm, and total chlorine (Cl2 + HOCl + OCl−): 900 ppm. The total chlorine concentration is somewhat higher than in the case of other ClO2 generators because the electrolyte applied by us does not contain any NaClO2. The UC-1 solution was produced by using only 25% NaCl solution, with no other additive, which is an obvious advantage. At the same time, despite the higher total chlorine content (which is present in the diluted UC-1 solutions mostly as HOCl), no detectable adverse effects were observed on the test animals or animal tissues.The test was performed following U.S. Pharmacopeia 34 NF29 Microbiological Tests/<51> [17]. Antimicrobial Effectiveness Testing. The test organisms were as follows: Escherichia coli (BCRC 11634/ATCC 8739), Staphylococcus aureus (BCRC 10451/ATCC 6538P), Pseudomonas aeruginosa (BCRC 11633/ATCC 9027), S. aureus subsp. aureus (BCRC 15211/ATCC 33591), Bacillus subtilis subspecies. (BCRC 10447/ATCC 6633), Listeria monocytogenes (BCRC 14848/ATCC 19114), Acinetobacter baumannii (BCRC 10591/ATCC 19606), Salmonella enterica subspecies. (BCRC 12947/ATCC 13311), Klebsiella pneumoniae (BCRC 16082/ATCC 4352), Penicillium funiculosum (BCRC 30438/ATCC 11797), and Candida albicans (BCRC 21538/ATCC10231).Viruses were amplified in MDCK/RD cells. MDCK/RD cells were cultured in 10% fetal bovine serum Dulbecco’s modified Eagle’s medium (FBS DMEM). When the cells reached 90% confluence, they were washed with phosphate-buffered saline (PBS) and infected at a multiplicity of infection of 0.01. Following the infection, 0% FBS DMEM was added, and the cells were incubated at 35 °C in a 5% CO2 incubator for 48 h.A 1-mL cell suspension (6 × 105 cells) was loaded into each well of a 6-well plate, which was incubated at 37 °C for 18–24 h. PBS was used to dilute UC-1 to final concentrations of 0, 25, 50, 100, and 200 ppm in wells reacted with cells and viruses for 2 min at 37 °C. Following the reaction, the total reaction mixture was diluted to 10−8. Subsequently, the 10−8 dilution mixture was incubated at 37 °C for 48–64 h. The cells were fixed with 10% formalin for 1 h and stained with 0.1% crystal violet for 5 min. The virus-formed plaque number was counted and compared between the test and control groups. The antiviral activity is shown as the percentage of virus control = plaques in the test group/plaques in the control group × 100. The virus control is defined as infected virus with cells without the testing agent and is considered as 100%.Mouse lung fibroblast L929 cells were cultured in complete Eagle minimum essential medium (MEM) and incubated at 37 °C ± 1 °C in 5% ± 1% CO2. Furthermore, 100 μL of L929 cell suspension (1 × 105 cells/mL) was transferred into each well of a 96-well cell culture plate. The cells were subsequently incubated at 37 °C ± 1 °C for 24 h ± 2 h. The culture medium was replaced with 100 μL of the test solution or blank, positive, or negative control. The test solutions contained 0 (control), 200, 400, 600, and 800 ppm UC-1 in MEM. The blank control medium contained 10% horse serum. The cells were incubated for another 24 h. The cells were treated with the solutions in triplicate. After the MTT solution was added to each well, the plate was incubated for 2 h ± 10 min at 37 °C ± 1 °C. The MTT solution was replaced with 100 μL of dimethyl sulfoxide and subsequently subjected to a microplate reader equipped with a 570-nm filter for colorimetric measurement (reference, 650 nm). The triplicate results of the MTT assay are presented as mean ± standard deviation (SD). Cell viability (%) = optical density of the test group/optical density of the control group × 100.Six 2–3-kg female New Zealand white rabbits were purchased from the Taiwan Livestock Research Institute (Xinhua, Tainan, Taiwan); the rabbits were quarantined and acclimatized before treatment. The animals were fed ad libitum and maintained at 20–26 °C under 30%–70% humidity. Furthermore, 0.1 mL of 50 ppm UC-1 (test solution) was administered to the left eye of the rabbits, and 0.1 mL of 0.9% normal saline (control solution) was administered to the right eye. Subsequently, the eyelids were held together for 1 s for instillation. Each treatment was repeated three times. Ocular irritations were observed for at the 1st, 24th, 48th, and 72nd hour using an ophthalmoscope (Welch Allyn, Skaneateles Falls, NY, USA). Extended observation was necessary in case of persistent lesions to determine the progression or reversal of the lesions. Ocular irritation scores were based on the system for grading ocular lesions (ISO 10993-10). When more than one animal in the test group showed a positive result at any stage of the observations, the test component was considered an eye irritant and further testing was not required or performed. When only one of the test groups showed a mild or moderate reaction that was equivocal, the procedure was conducted on three additional animals. When more than half of the eyes showed a positive result at any stage of the observation, the test component was considered an eye irritant. A severe reaction in only one animal was considered sufficient to label the test component as an eye irritant.Fifteen 4-week-old BALB/c male mice were purchased from the National Laboratory Animal Center (Taipei, Taiwan); they were quarantined and acclimatized before treatment in an animal room at China Medical University, Taiwan. The animals were fed ad libitum and maintained at 20–25 °C and 65%–80% humidity. Five mice were housed in one cage and fed with 0 (PBS) and 10 or 20 ppm UC-1 (test solution), which was administered as mist by using a humidifier in an airtight box for 24 h. The clinical symptoms and body weight of the animals were observed; they were subsequently sacrificed to examine their lung sections and organ weight. The experimental animals were observed, and their clinical symptoms were recorded as abnormality (%), defined as the animals behaving abnormally compared with normal animals, and mortality (%), defined as animal death.During the experiment, the animals were immediately dissected on death, and a record was made. All surviving animals were sacrificed and autopsied to observe their appearance and all organs in the mouth, chest, and cranial and abdominal cavities. Subsequently, the organs, including the liver, adrenal glands, kidneys, and gonads, were removed, weighed, and recorded.Tissue sections frozen in the optimal cutting temperature compound were fixed in acetone and chloroform; the sections were immersed in filtered Harris hematoxylin (Leica Biosystems Richmond, Inc., Richmond, IL, USA) for 1 min. The slides were rewashed with Tris-buffered saline and Tween 20 (Biokit Biotechnology Inc., Miaoli, Taiwan), and the sections were counterstained with eosin (Leica Biosystems Richmond, Inc., Richmond, IL, USA) for 1–2 min. The sections were dehydrated in ascending alcohol solutions and cleared with xylene. The prepared slides were examined through light microscopy.Twenty-five 4-week-old BALB/c male mice were purchased from the National Laboratory Animal Center; they were quarantined and acclimatized before treatment in an animal room at China Medical University. The animals were fed ad libitum and maintained at 20–25 °C under 65%–80% humidity. Five mice were housed in one cage and fed 0 (PBS; control), 5, 10, 20, and 40 ppm UC-1 (test solutions) continuously for 90 days. PBS or test solutions fed as drinking water were freshly prepared daily before treatments.The experimental animals were observed, and their clinical symptoms were recorded as abnormality (%), defined as the animals behaving abnormally compared with normal animals, and mortality (%), defined as animal death.The body weight of the experimental animals was recorded at treatment initiation and once per week during the experimental period using an electronic balance (AND, FX-2000i, Tokyo, Japan).During the experiment, the animals were dissected immediately on death, and a record was made. All surviving animals were sacrificed and autopsied to observe their appearance, and all organs in the mouth, chest, and cranial and abdominal cavities were analyzed. Subsequently, the organs, including the liver, adrenal glands, kidneys, and gonads, were removed, weighed, and recorded.The results were analyzed using SPSS Version 20.0 (IBM Corp., Armonk, NY, USA) with one-way analysis of variance, F-test, and Duncan’s new multiple range test for comparing more than two mean values; results with p < 0.05 indicated significant differences. The results represent at least 3 independent experiments and are shown as the mean ± SD.This research was approved by the China Medical University Laboratory Animal Service Center. Program Number: 10442699 (for the white rabbit ocular irritation test) and 10442686 (for the inhalation toxicity and subchronic oral toxicity tests).In this study, a UC-1 containing 2000 ppm chlorine dioxide in water was produced through the electrolytic method with food-grade salt (99% NaCl) and RO water as the starting reactants. Subsequently, the chlorine dioxide was purified through a film and dissolved in RO water. Because a chlorine dioxide solution can be directly applied to food or human hygiene or preventative health measures, its safety and efficacy were investigated.The in vitro antimicrobial activity of UC-1 was examined. The in vitro antimicrobial activity was more than 98.2% reduction for bacteria and fungi (Table 1); excellent antimicrobial activity was observed at low concentrations of 5 and 20 ppm UC-1 for bacteria and fungi, respectively.The antiviral activity of 0, 25, 50, 100, and 200 ppm UC-1 after 2 min of reaction is shown in Figure 2. For H1N1 and influenza virus B/TW/71718/04, 200 ppm UC-1 had the most significant effect in inhibiting viral plaque formation. The half maximal inhibitory concentration (IC50) of H1N1 was 84.65 ± 0.64 ppm and that of influenza virus B/TW/71718/04 was 95.91 ± 11.61 ppm. For EV71, 50 ppm UC-1 showed significant inhibition activity, with an IC50 of 46.39 ± 1.97 ppm at 2 min. The results showing statistical significance (p < 0.05) are presented. Bars are plotted as means ± SD.The cytotoxic effect of 0, 200, 400, 600, and 800 ppm UC-1 against L929 lung fibroblast cells was analyzed. The cell viability was 74.0%–100.0% at UC-1 concentrations below 600 ppm. L929 cell viability was reduced to 40.3% at 800 ppm UC-1 (Figure 3).The cornea, iris, and conjunctivae were evaluated in a rabbit ocular irritation test. The 50 ppm UC-1 solution induced neither significant clinical signs nor ocular gross changes in the rabbits at each time point (Table 2). Therefore, single ocular applications with 0.1 mL of 50 ppm UC-1 did not cause ocular irritation in rabbits.In an inhalation toxicity test, we used 0, 10, and 20 ppm UC-1, which was administered as mist by using a humidifier in an airtight box containing five mice. The test showed no abnormality and no mortality for the control and test components within 24 h (Table 3). The weights of the heart, liver, spleen, and kidney of the test group did not differ significantly from those in the control group (Table 4). Hematoxylin and eosin staining of the mice lung sections (Figure 4) showed that 10 and 20 ppm UC-1 did not induce significant clinical signs of changes in the mouse lung cells at 24 h. Therefore, inhalation of 10 and 20 ppm UC-1 did not cause irritation in the mice.In the subchronic oral toxicity test, 0, 5, 10, 20, and 40 ppm UC-1 was prepared to feed the mice. Clinical observations of the mice showed no abnormality and no mortality after 90 days for the control and test groups (Table 5). The mouse weight was not influenced (Figure 5). Moreover, necropsy and gross examination did not show any pathological symptoms (Figure 6). The weights of the heart, liver, spleen, and kidney of the test groups did not differ significantly compared with those in the control group (Table 6). Therefore, administration of up to 40 ppm UC-1 to mice for 90 days is nontoxic.UC-1 with gas chlorine dioxide and fewer impurities was prepared using a patented green process design, and the efficacy and safety of UC-1 were evaluated. Many studies have reported the potent oxidant and antimicrobial activity of chlorine dioxide in vitro. Recent reports have addressed concerns related to microbial decontamination of food by chlorine dioxide [6,9,18,19,20,21]. In these studies, chlorine dioxide was produced using various methods (e.g., 2% NaClO2 with H3PO4, 4% Cl2 with 80% NaClO2, and by using an electrogenerator); the antimicrobial activity was more than 2% for 5–75 mg/L ClO2 within 5–30 min. Here, the antimicrobial activity was more than 98.2% reduction at UC-1 concentrations of 5 and 20 ppm for bacteria and fungi, respectively. In the MTT test, the viability of L929 cells was 93.7% at 200 ppm UC-1 that a concentration is over routine use.No significant symptoms were observed with 50 ppm UC-1 in the ocular irritation test. No abnormality or mortality was observed in clinical symptoms, lungs, and other organs at 10 ppm or 20 ppm UC-1 in the inhalation toxicity test. Paulet and Desbrousses [22] administered 2.5, 5, and 10 ppm chlorine dioxide to rats and rabbits in 1970; they reported that 2.5 ppm chlorine dioxide had the lowest-observed-adverse-effect level (LOAEL), causing thoracic effects in rats at 7 h/day for 30 days and pulmonary effects in rabbits at 4 h/day for 45 days. Paulet and Desbrousses [23] increased the test concentration to 5, 10, and 15 ppm chlorine dioxide and reduced the dose time to 15 min per dose, 2–4 times per day, for 4 weeks in rats. The results showed a no-observed-averse-effect level (NOAEL) of 5 ppm and an LOAEL of 10 ppm for lung damage.The mice were fed drinking water containing up to 40 ppm UC-1 for 90 days; the concentration showed no toxicity in the sub-chronic orally toxicity test. Daniel et al. [24] reported the oral exposed toxicity of chlorine dioxide in drinking water administered to Sprague–Dawley rats for 90 days; they used different concentrations of chlorine dioxide (0, 25, 50, 100, and 200 mg/L corresponding to doses of 0, 2, 5, 8, and 15 mg/kg·day). The spleen and liver weight decreased significantly at 25 and 50 mg/L, respectively. They showed nasal lesions caused by 25 mg/L chlorine dioxide vapors in drinking water. In that study, the LOAEL was 25 mg/L. Bercz et al. [25] conducted a similar test in African green monkeys (Cercopithecus aethiops) by using 0, 30, 100, and 200 mg/L chlorine dioxide for 4–6 weeks. Furthermore, 200 mg/L chlorine dioxide caused erythema and ulceration of the oral mucosa after 1 week, and 100 mg/L chlorine dioxide reduced the serum thyroxine (T4) levels after 6 weeks; in that study, the NOAEL was 30 mg/L and LOAEL was 100 mg/L for the oral exposure of the monkeys. UC-1 was produced through a green process with clean starting materials and procedures. UC-1 solution demonstrated satisfactory antibacterial, antifungal, and antiviral activity. Low toxicity was demonstrated through an in vitro cytoxicity test (high IC50 765 ± 18 ppm), 50 ppm ClO2 did not cause eye irradiation in an ocular irritation test, mice did not exhibit abnormality and mortality in a 20 ppm ClO2 inhalation toxicity test, and concentrations of UC-1 up to 40 ppm were nontoxic to mice for 90 days in subchronic oral toxicity test. Therefore, a higher safety profile for UC-1 than those yielded in previous studies was demonstrated.We thank Unique Biotech Co. Ltd. for providing UC-1 solution support. This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.Hao-Chang Yin and Shan-Shue Wang were involved in research design, discussion, analysis, decision, revision, and funds. Jui-Wen Ma and Bin-Syuan Huang processed the data, implemented most experiments, and wrote the manuscript. Chu-Wei Hsu, Chun-Wei Peng, and Ming-Long Cheng helped execute detailed experiments as well as collect and manage data. Jung-Yie Kao and Tzong-Der Way were technical advisors and study coordinators and provided helpful suggestions for experimental design, novel opinions for the study, and collaboration with other laboratories.The authors declare no conflict of interest.Flowchart of chlorine dioxide solution production.Antiviral efficacy against influenza virus A/WSN/33, influenza virus B/TW/71718/04, and enterovirus 71. Bars are plotted as means ± standard deviation (SD). Means with the same letter did not differ significantly at p < 0.05 according to the ANOVA (Analysis of Variance) F-test and Duncan’s new multiple range test.Cytotoxic effects of various UC-1 concentrations on L929 cells.Hematoxylin and eosin staining of mouse lung sections in the inhalation toxicity test. The Scale bar labeled in this figure was 100 μm.Mouse weight trend chart in the subchronic oral toxicity test. CTL: control.Observation of mouse lungs and organs in the subchronic oral toxicity test.Antimicrobial efficacy of UC-1.a The contact time was 10 min. b Percent reductions of <1% represent no significant bacteriostasis or fungistasis. c UC-1 concentrations were 5 and 20 ppm for bacteria and fungi, respectively. * presented as bacteria; Δ presented as fungi. CFU: colony-forming unit.Grades in the clinical observation of individual rabbits for the ocular irritation test.Evaluation of clinical symptoms in the inhalation toxicity test.PBS: phosphate-buffered saline.Evaluation of organ weight for the inhalation toxicity test.Statistical analyses of the presented data were performed at the 95% significance level (p < 0.05).Evaluation of clinical symptoms in the subchronic oral toxicity test.Evaluation of organ weight in the subchronic oral toxicity test.Statistical analyses of the presented data were performed at the 95% significance level (p < 0.05).

Med-MDPI/ijerph_2/ijerph-14-03-00330.txt

@@ -0,0 +1 @@
+Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).Background: The majority of older adults prefer to remain in their homes, or to “age-in-place.” To accomplish this goal, many older adults will rely upon home- and community-based services (HCBS) for support. However, the availability and accessibility of HCBS may differ based on whether the older adult lives in the community or in a senior housing apartment facility. Methods: This paper reports findings from the Pathways to Life Quality study of residential change and stability among seniors in upstate New York. Data were analyzed from 663 older adults living in one of three housing types: service-rich facilities, service-poor facilities, and community-dwelling in single-family homes. A multinomial logistic regression model was used to examine factors associated with residence type. A linear regression model was fitted to examine factors associated with HCBS utilization. Results: When compared to community-dwelling older adults, those residing in service-rich and service-poor facilities were more likely to be older, report more activity limitations, and provide less instrumental assistance to others. Those in service-poor facilities were more likely to have poorer mental health and lower perceived purpose in life. The three leading HCBS utilized were senior centers (20%), homemaker services (19%), and transportation services (18%). More HCBS utilization was associated with participants who resided in service-poor housing, were older, were female, and had more activity limitations. More HCBS utilization was also associated with those who received instrumental support, had higher perceived purpose in life, and poorer mental health. Conclusions: Findings suggest that older adults’ residential environment is associated with their health status and HCBS utilization. Building upon the Person–Environment Fit theories, dedicated efforts are needed to introduce and expand upon existing HCBS available to facility residents to address physical and mental health needs as well as facilitate aging-in-place.Older adults prefer to remain in their homes, or to “age-in-place”, because doing so allows them to maintain independence [1,2]. Housing type preferences among older adults in the United States are diverse; they often vary based on financial or healthcare-related needs, and affect aging-in-place. The majority of older adults express desire to remain in their homes [3,4], a smaller proportion with fewer needs prefer to reside in independent living apartments or active living communities, and those with additional needs reside in assisted living facilities or affordable housing.Home- and community-based services (HCBS) can facilitate the ability of older adults to remain in their own homes by providing various forms of assistance to accomplish activities of daily living [5,6]. Typical HCBS include assistance with bathing, meal provisions, homemakers, respite care, transportation, in-home health care (e.g., nurse visits, physical therapy), and legal advice [7]. However, older adults often encounter barriers to HCBS utilization [8,9], which include affordability and cost, lack of awareness, and unavailability of services. These impediments suggest the presence of an unmet need among older adults, and researchers are beginning to examine HCBS utilization patterns to increase service access and uptake among older adults who could benefit from such resources [8,9].Barriers to utilization are problematic, because the increased burden of chronic conditions experienced by older adults further contributes to the need for HCBS. Approximately two-thirds of older adults have two or more chronic conditions [10]. The most commonly-occurring conditions include arthritis, heart disease, cancer, diabetes, and hypertension [11,12], all of which increase the risk for functional decline, impairment [13], and the need for assistance with activities of daily living (ADLs (Activities of Daily Living); e.g., feeding, dressing, bathing). This ADL assistance often becomes the responsibility of family members, most often the spouse or adult children [14,15]. In the absence of supportive services, such caregiving tasks become particularly burdensome on the older adult and care provider alike [16,17,18,19].The physical (built) environment has been identified as an important contributor to HCBS use [20,21,22]. For example, HCBS use varies substantially across housing types based on service availability, accessibility, and geospatial proximity. While facilities such as Continuing Care Retirement Communities (CCRCs) enmesh services and housing [3,23,24]. HCBS are rarely offered accommodations in government subsidized senior housing (i.e., Section 202 housing) that provide independent living apartments to older adults of limited means [25,26].Less is known about the influence of the social environment on HCBS use. Social isolation has adverse effects on the health and well-being of older adults, including a greater risk of disability and mortality in comparison to other age groups [27,28,29]. HCBS use can be of particular benefit to address the needs of older adults at increased risk of social isolation who have small social networks to assist them (e.g., physically, socially, financially) [30]. While substantial arguments have been made to link research about physical and social environments, research that adequately combines these aspects of aging has not proliferated [31,32].Therefore, the purposes of the study were to (1) identify the demographics, health status, psychosocial factors, and HCBS utilization among older adults based on the type of housing in which they live; and (2) examine how demographics, health status, psychosocial factors, and residence type are associated with HCBS utilization. These purposes are achieved through the examination of three older adult subgroups who live in single-family homes (e.g., community-dwelling), service-rich facilities (e.g., CCRC), and service-poor senior housing (e.g., apartments with few to no service provisions).The study of social and physical environments on the functional and psychosocial well-being of older adults is often difficult given the division of theories by discipline [20,31,33,34]. It has been proposed that the differential concepts of social and physical aspects of environment entail issues of meso- and micro-levels of analyses, thus limiting the inclusion of these aspects as covariates in the same study [31]. However, these two aspects are components within the person–environment fit processes, which considers the physical and social environments as transactional [35]. Both cross-sectional and longitudinal studies report the influence of social networks on subjective well-being, including exchanges of support [36], social contact improving mood [35,37], and adaptation to losses associated with aging [38].This study draws upon the Lawton’s ecological model [20,34,39] and Person-Environment (P-E) framework [21]. P-E fit conceptualizes the relationships between older adults and their surrounding systems. The “person” component is comprised of individual-level competencies, such as motor skills, cognitive function, and biological health [40]. The environment is often conceptualized as the physical surroundings such as the home or community. The ecological model provides context for P-E fit through the inclusion of physical, personal, small-group (e.g., family and friends), supra-personal (e.g., proximal family/staff) and mega-social environments (i.e., culture, society) [20]. Activities of Daily Living (ADLs) limitations are an index of the relationship between the person (ability) and environment (management of tasks) [41,42]. It is hypothesized that older adults experiencing difficulties in maintaining ADL tasks may particularly benefit from assistance provided by HCBS. We use these theoretical frameworks to guide our investigation about the associations of the physical and social environments with the use of HCBS.This study examined data from the Pathways to Life Quality Study, a longitudinal study of residential change and stability in an upstate New York community [43,44,45,46,47,48,49]. Samples were recruited through two pathways. Survey Sampling, Inc. and voter registration records provided the initial list of 55,000 potential participants over the age of 60 years. Using a random number generator (www.randomizer.org), names were selected at random and information about the study was mailed to them. Follow-up telephone calls were conducted a week later, and interviews were scheduled by the researchers. Letters returned with no forwarding address were flagged in the database along with any returns indicating that the contacts were deceased. The final response rate for the random sample was 43%. Convenience samples were recruited from the senior housing communities through fliers, mailings, social events and presentations. Based on these procedures, the resulting analytic sample consisted of randomly-selected community-dwelling county residents aged 60 and older (n = 343), residents of service-rich facilities (n = 184), and residents of service-poor facilities (n = 136). Service-rich facilities were those in which onsite services were available and included in the monthly fees. Within this study, there were two service-rich facilities. One was a continuing care retirement community (CCRC) that provided assisted and skilled-nursing levels of care, meal provisions; a rehabilitation and fitness center staffed with physical, occupational, and speech therapists; a care clinic with pharmacy; and transportation. The second service-rich facility had independent and assisted living options, a recreation suite and a partnership with a nearby college that allowed allied health students experience and credit by providing therapies to residents; a swimming pool, fitness center, and exercise classes; and transportation. Service-poor facilities included low-income HUD-subsidized apartment buildings and affordable apartments for senior living. Service provisions were not part of the housing package, but were available through Medicare/Medicaid waivers and local home health care agencies. The Pathways to Life Quality Study was approved by the Institutional Review Board at Cornell University and Ithaca College during data collection. Institutional Review Board (00003614) approval was granted for this secondary data analyses at the University of Georgia.As the guiding theoretical framework of this study, the Ecological Model was used to inform variable selection for this study [20]. In particular, the psychosocial variables represent the small group and supra-personal aspects of the social environment. Descriptions of study variables are provided below.Our main variable of interest to examine across housing types was HCBS service use. The survey directly asked participants if they used HCBS. A list of services was presented, dichotomously coded as “uses” and “does not use.” HCBS included in the list were home health care, senior centers, transportation, home-delivered meals, legal assistance, and homemaker. Within this upstate, New York county home health care included visits from nurses, nurse aids, and physical or occupational therapists from area health care agencies. Senior centers provide activities, recreation, and community meals. Senior centers also assist with coordination and delivery of home delivered meals for those unable to visit the center. Legal assistance was a local service that provided assistance with wills and end-of-life documents for older adults. Homemaker services included assistance with companionship, cleaning, and errands. These variables were used in descriptive statistics and then summed into a variable of total number of services used.Subjective assessments of health were ascertained through self-report of health on a ten-point ladder where 10 represented best possible health [50], a four-point scale comparing one’s own health to others one’s own age, self-report levels of pep or energy on a ten-point ladder, and summed score of ADL limitations using the Medical Outcomes Survey (MOS) scale [51]. Objective measures of health included (a) months since last physician visit; (b) incidence of hospitalization in the preceding two years; (c) the number of hospitalizations; and (d) days per hospitalization.Embedded within the interviews were scales assessing psychosocial well-being, including social integration and social support, positive and negative affect, purpose in life, instrumental support, and a one-item life satisfaction measure. The Cutrona Provisions of social relationships subscales were included as indices of psychosocial well-being, to measure social integration and social support [52]. The Positive and Negative Affect Scales (PANAS) provided summed scores for the affect measures [53]. Also included were subscales on the Purpose in Life from the Ryff Scales of Psychosocial Well-Being [54] and the instrumental support from the Piedmont Health Survey [55]. The instrumental support subscale measures both receiving help from and providing help to others. Generativity, the concern for guiding future generations, was measured through the Loyola Generativity Scale [56]. Life satisfaction was rated on a ten-point scale, modeled after the self-rated health ladder, with higher numbers representing greater satisfaction. A 4-item Likert-response format scale was used to capture home satisfaction, an index of P-E fit. We controlled for demographic variables, including age (continuous), sex, and marital status (non-married versus married/partnered).All analyses were performed using SPSS version 24 [57]. Basic descriptive statistics were run on measures of interest across the three housing types. One-way ANOVA tests were performed to assess mean differences for continuous and count variables. Post-hoc tests were used to identify significant mean differences between groups. Chi-square and Fisher’s exact tests were performed on categorical variables. A multinomial logistic regression model with backwards entry was fitted to examine factors associated with residence type. In this analysis, those living in the community-based housing served as the referent group. Resident profile typologies emerge from the multinomial analyses. Then, a multivariate linear regression analysis with backwards entry was used to identify demographic, health, and psychosocial factors associated with increased HCBS use. Given the large number of independent variables included in this study, backwards entry was used in multivariate analyses to generate more parsimonious and interpretable models.Table 1 provides sample characteristics by housing type. Overall, the average age of participants was 76.35 (±7.9) years. The majority of participants was female (67%) and 28% were married. Within the overall sample, self-assessed health averaged 7.40 on a ten-point scale. Health compared to others averaged 3.13 on a four-point scale, both indicative of good health. Participants averaged two limitations in activities of daily living, and had last visited their physician on average 2.20 months prior. Nearly one-quarter (24%) had been hospitalized in the two years preceding the interview for a mean average of 1.65 times for an average of ten days per stay. The three leading HCBS used by participants were senior centers (20%), homemaker services (19%), and transportation services (18%). On average, participants reported using less than one HCBS.When comparing study characteristics by housing type, participants residing in service facilities were significantly older than community-dwelling participants. A larger proportion of those residing in service-poor facilities were female, and a significantly smaller proportion was married relative to other housing types. On average, community-dwelling older adults higher self-rated health, better ratings of health compared to others, and higher energy levels compared to the other two groups. On average, community-dwelling participants provided more instrumental support relative to those residing in facilities.On average, community-dwelling participants reported higher life satisfaction relative to those residing in service-rich and service-poor housing. On average, feelings of generativity and having a purpose in life were lowest among service-poor participants relative to the other participant groups. On average, service-poor residents used significantly more HCBS relative to other participant groups. More specifically, a significantly larger proportion of those in service-poor facilities used home health care, transportation, home-delivered meals, and homemaker services than other participant groups. A significantly smaller proportion of service-rich residents used senior centers relative to other participant groups. A significantly smaller proportion of service-poor residents used legal services relative to other participant groups.Table 2 presents findings from a multinomial logistic regression model that examined factors associated with residence type to generate emergent resident profiles for each housing type. Compared to community-dwelling individuals, participants who were older and married were more likely to reside in service-rich facilities. Participants with more ADLs were more likely to reside in service-rich facilities, whereas those who provided and received instrumental support were less likely to reside in service-rich facilities. Participants with higher perceived life satisfaction were less likely to reside in service-rich facilities.Compared to community-dwelling individuals, older participants were more likely to reside in service-poor facilities. Participants who were married were less likely to reside in service-poor facilities. Participants with more ADLs were more likely to reside in service-poor facilities, whereas those who provided instrumental support were less likely to reside in service-poor facilities. Participants with higher positive affect and negative affect scores were more likely to reside in service-poor facilities; whereas those with higher perceived life satisfaction were less likely to reside in service-poor facilities.Table 3 presents findings from a linear regression model that examined factors associated with HCBS use. Participants who resided in service-poor housing, those who were older, and females used more HCBS. Participants who had more ADLs and received more instrumental support used more HCBS. Those who provided less instrumental support used more HCBS. Participants with worse positive affect and those with higher perceived purpose in life used more HCBS.This study provides a glimpse into the interplay of environment, health, and service utilization among a sample of older adults. When comparing demographics, health status, and psychosocial factors by housing type, differences were observed. These findings suggest that an older adult’s environment can influence their health status, or that their health circumstance can influence where they reside (choice or by force).Findings in this study revealed that participants in service-poor housing had higher risk in terms of health and psychosocial factors (i.e., ADLs, negative affect, life purpose). This is supported by previous research suggesting residents in senior housing have higher rates of unmanaged health needs and depression [58]. While these poorer health outcomes may hinder aging-in-place, additional research is needed to explore the underlying causes of these health indicators and the benefits of HCBS utilization. Overall utilization of HCBS was low among the study sample, despite reported poor health and psychosocial factors that can be addressed/improved by such services. While HCBS use was highest among those residing in service-poor housing, it is unclear whether low reported service utilization was attributed to the lack of knowledge about services, low perceived benefits from accessing services, absence of services in their local area, or service ineligibility. Findings highlight the need for additional awareness raising and recruitment efforts to promote HCBS to housing facility residents.Providing and receiving instrumental support were associated with service use and varied across housing types. Community-dwelling older adults engaged in more instrumental support compared to facility residents. Providing less support was associated with HCBS use, while receiving more support was associated with HCBS use. Given ADLs were also associated with HCBS use, findings suggest that individuals in worse physical health may be utilizing services and resources required to meet their needs (e.g., home health care, transportation, home-delivered meals, homemaker services). Recognizing these services can be instrumental in managing health conditions and physical limitations among at-risk older adults; HCBS can be beneficial for all older adults and prevent negative health consequences. For example, because older adults’ mental health and social well-being can decline alongside growing physical limitations, and given mental health disorders are largely untreated among older adults [59,60], opportunities exist to increase mental health screening, resources, and service utilization among housing facility residents.Based on P-E fit, study findings suggest the need to increase service coordination and build community partnerships with agencies and providers to improve fit and promote aging-in-place. For example, to combat poorer health among residents of service-poor housing, one strategy to improve health outcomes is to improve the integration of primary care and behavioral health services within housing facility communities [61]. Another strategy to improve health among housing facility residents could be to employ and work with a Health and Aging Residential Service Coordinator (HARSC), who can assess the health status of residents, determine their eligibility for services, link them to such services, and follow-up with them to ensure their needs are met [62].In this study, the highest utilized resource was senior centers, primarily among community-dwelling and service-poor residents. Senior centers are community hubs for community-based services, especially in their offering of evidence-based programs that address health topics including chronic disease, fall prevention, and physical activity [63,64,65,66,67]. However, senior centers use and locale may limit utilization. For instance, senior centers are not widely used by diverse older adults [68], or are often located in more affluent areas. Given that transportation is among the highest reported needs for American older adults [9,69], the location of senior centers may indicate the need for transportation services among facility residents to ensure that they can access programs and resources offered at such entities. Facilities are encouraged to create partnerships with non-emergency medical transportation brokers as a strategy to increase mobility among older adults with limited travel options [70,71].A limitation of this study is its cross-sectional design, thus limiting the ability to determine the causal relationships among the variables. For instance, it would be interesting to determine if living in service-poor communities actually contributes to poorer health among older adults. Second, the list of HCBS in this study may not have been comprehensive. Other services should be examined in future studies such as durable medical equipment, home safety assessments, and financial services [72]. This study is limited by the inability to delineate the actual rate of HCBS use among participants or changes in health status as a result of service utilization. Finally, the senior housing samples were convenience samples and therefore not generalizable to the greater population. Future studies should be replicated to purposively include marital dyads and use multi-level models to examine household-level service use across housing types. Study findings did not fully elucidate the factors supporting aging-in-place; future studies should empirically test the influence of the P-E framework within the context of the ecological model to include expanded environmental influences (psychosocial). Further investigation is also warranted to examine the longitudinal impacts of environment on health status among the aging population. Additionally, further investigations should examine the health-related impact on migration and relocation, which are indicators of adults’ ability to age-in-place.To meet the needs of a growing aging society, improving P-E fit between older adults and their respective environments should be a priority of service providers and policy makers. When the fit between an older adult and his or her environment is insufficient and leaves health and psychosocial needs unfulfilled (e.g., needs for meal provisions, homemakers, respite care, transportation, in-home health care), it is incumbent upon service providers and policy makers to work together to improve fit to increase aging-in-place opportunities. One way to achieve this goal is by implementing interventions at multiple systems levels (e.g., individual, family, and community) that create new resources, sustain existing services, and promote health and aging-in-place.The Pathways to Life Quality study was supported through funding from Atlantic Philanthropies. Heidi H. Ewen conceived the study, performed statistical analyses, and drafted the manuscript. Tiffany R. Washington assisted with statistical interpretations and drafted the manuscript. Kerstin G. Emerson and Andrew T. Carswell drafted and reviewed the manuscript. Matthew Lee Smith conceived the study, assisted with statistical interpretations, and drafted and reviewed the manuscript.The authors declare no conflict of interest.Sample characteristics by housing type.Percentages reported for categorical variables. Means and standard deviations (SD) reported for continous and count variables. * Signifies significant mean differences determined by post-hoc analyses. Multinomial logistic regression examining factors associated with housing types.Referent Group: community-dwelling individuals. Model fit statistics: Nagelkerke R2 = 0.49; −2Log = 951.90; χ2 = 348.62; df = 28; p < 0.001.Linear regression examining factors associated with service utilization.Adjusted R2 = 0.22; F (8628) = 22.84; p < 0.001.

Med-MDPI/ijerph_2/ijerph-14-03-00331.txt

@@ -0,0 +1 @@
+Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).Adverse childhood experiences are associated with an array of health, psychiatric, and behavioral problems including antisocial behavior. Criminologists have recently utilized adverse childhood experiences as an organizing research framework and shown that adverse childhood experiences are associated with delinquency, violence, and more chronic/severe criminal careers. However, much less is known about adverse childhood experiences vis-à-vis specific forms of crime and whether the effects vary across race and ethnicity. Using a sample of 2520 male confined juvenile delinquents, the current study used epidemiological tables of odds (both unadjusted and adjusted for onset, total adjudications, and total out of home placements) to evaluate the significance of the number of adverse childhood experiences on commitment for homicide, sexual assault, and serious persons/property offending. The effects of adverse childhood experiences vary considerably across racial and ethnic groups and across offense types. Adverse childhood experiences are strongly and positively associated with sexual offending, but negatively associated with homicide and serious person/property offending. Differential effects of adverse childhood experiences were also seen among African Americans, Hispanics, and whites. Suggestions for future research to clarify the mechanisms by which adverse childhood experiences manifest in specific forms of criminal behavior are offered.In the seminal survey of 9508 adults in the Adverse Childhood Experiences (ACE) study, Felitti and colleagues [1] introduced the concept of adverse childhood experiences to account for the negative health and behavioral consequences of various forms of childhood abuse, neglect, and exposure to household dysfunction. The seminal ACE study and subsequent studies [2,3,4,5] supported the notion that individuals who endured more adverse childhood experiences tended to suffer the most throughout the life course and evinced the greatest number of health problems, maladaptive behaviors, and comorbid psychiatric conditions.One of the maladaptive behaviors that results from adverse childhood experiences is antisocial behavior. In recent years, criminologists have utilized the adverse childhood experiences conceptual framework to examine associations with delinquency, crime, and serious violence. It has become clear that diverse adverse childhood experiences are fairly pervasive among clinical and correctional samples relative to those in the general population. For example, Baglivio and colleagues [6] examined a sample of more than 60,000 juvenile offenders from the United States and found that delinquents were significantly more likely than those in the general population to have pervasive adverse childhood experiences and significantly less likely to never experience adverse childhood experiences. In other words, clinical samples of youth evince a high preponderance of abusive experiences and deprivation. These findings were consistent with studies of youth residing in detention, correctional, and confinement facilities [7,8,9,10,11] where adverse childhood experiences are endemic.There is also compelling evidence that adverse childhood experiences are particularly serious risk factors for more pathological forms of offending. For instance, Fox and colleagues [12] reported that each additional adverse childhood experience increased the likelihood of serious, violent, and chronic juvenile offending by 35%. Others have similarly found that adverse childhood experiences were linked to more severe offending trajectories, earlier onset of antisocial conduct, and shorter time to recidivism post-juvenile justice services [13,14]. For instance, Boduszek and colleagues [15] found that male prisoners in Poland who had been exposed to family violence were approximately six times more likely to perpetrate a homicide than offenders who lacked violence exposure. In terms of sexual violence, a host of investigators have found that assorted adverse childhood experiences, particularly childhood sexual abuse victimization, were associated with an increased likelihood of perpetrating sexual crimes during adolescence or adulthood [16,17,18,19,20,21]. The evidence is clear that adverse childhood experiences increase the liability for externalizing symptoms and diverse forms of antisocial conduct.Despite these general findings, there remain important knowledge gaps. For instance, in the United States there is extraordinary heterogeneity of background experiences among delinquents in addition to sharply divergent local life circumstances and socioeconomic backgrounds by race and ethnicity [22,23,24,25,26], which have been demonstrated to affect adverse childhood experience wherein juvenile offenders residing in more disadvantaged communities evidence greater exposures. Prior criminological research has shown that disaggregated analyses by race and ethnicity are a fruitful way to understand different pathways of offending [27,28,29]. Given these varying background experiences, it is unclear to what extent adverse childhood experiences are associated with crime in the same way for different racial and ethnic groups, and to what degree the adverse childhood experience effects are potentially crime-specific. For example, Duke and colleagues [30] examined 136,549 youth who responded to the 2007 Minnesota Student Survey and found differential associations between adverse childhood experiences by race and gender in addition to differential effects on specific forms of antisocial behavior, including delinquency, bullying, physical fighting, dating violence, and weapons carrying. Adverse childhood experiences were negatively associated with delinquency, but positively associated with various forms of violence and weapons carrying. To our knowledge, no prior study has examined the adverse childhood experiences-crime link using models that are disaggregated by race, ethnicity, and offense type. The current study was conducted to examine three research questions: (1) whether there are differential effects for the number of adverse childhood experiences that a youth experienced; (2) whether these effects vary across race and ethnic groups, and (3) whether the adverse childhood experiences-antisocial behavior association depends on the commitment offense type.Cross-sectional data are based on a sample of (n = 2520) adjudicated male delinquents committed to confinement facilities in a large southern state in 2009. Information on each juvenile offender was compiled by the juvenile correctional system at a statewide intake unit upon the youths’ commitment and during their institutionalization. All state-committed youth were housed at the intake facility for approximately two months and then were transferred to specific facilities around the state to complete their commitment period. Additional offender data were collected during the youth’s confinement from numerous sources including state-level and county-level official records, on-site diagnostic procedures at intake, observations from professional and correctional staff, self-report information from youth, or a combination of these sources. Delinquency history and additional file information for all state-committed delinquents were collected with standardized instruments at each juvenile facility in the state and maintained at a centralized location.A count-measure of adverse childhood experience exposures (Mean = 2.08, Standard Deviation = 1.51, range = 0–7) was constructed from seven dichotomous variables indicating (1) whether the youth had been physically abused; (2) whether the youth had been sexually abused; (3) whether the youth had been emotionally abused; (4) whether the youth was reared in poverty; (5) whether the youth was reared in a chaotic home characterized by residential instability and multiple family members and acquaintances living in the home; (6) whether the youth had family members who were in gangs; and (7) whether the youth had been violent toward members of his immediate family. The use of multiple and diverse indicators of adverse childhood experiences is consistent with prior research [31,32]. In terms of prevalence of adverse childhood experiences, 17.3% had zero, 18% had one, 30.9% had two, 17.1% had three, 8.7% had four, 5.8% had five, 1.82% had six, and 0.3% had all seven.Analyses were performed separately for African Americans (n = 889, 35.3%), Hispanics (n = 966, 38.3%), and whites (n = 625, 24.8%).Commitment offense type is the most serious charge for which the ward was committed to the confinement facility. Homicide offenses (n = 681, 27.02%) included capital murder, attempted capital murder, murder, attempted murder, criminally negligent homicide, and voluntary manslaughter. Sexual offenses (n = 930, 36.9%) included aggravated sexual assault, attempted aggravated sexual assault, attempted sexual assault, and sexual assault). Serious person/property offenses (n = 498, 19.76%) included aggravated robbery and attempted aggravated robbery.Three delinquent career parameters were used as statistical controls in the adjusted tables of odds based on their empirical associations with serious delinquency [33,34,35,36,37]. These were age at first commitment to a confinement facility which captures onset (Mean = 15.29, Standard Deviation = 1.14, range = 10–18), total previous delinquent adjudications (Mean = 1.60, Standard Deviation = 0.95, range = 0–10), and total prior out of home placements (Mean = 0.45, Standard Deviation = 1.12, range = 0–15).Epidemiological tables of odds were used to examine the associations between adverse childhood experiences and the three commitment offense types for the full sample, African Americans, Hispanics, and whites. Epidemiological tables of odds are used for case-control and cross-sectional data to evaluate the odds of a binary outcome (e.g., commitment for homicide, sexual, or serious person/property offense) by score on a predictor variable (e.g., adverse childhood experiences). Two additional tests were conducted. The test of homogeneity evaluates if the odds of failure associated with adverse childhood experiences the youth experienced are equal. The score test for trend of odds indicates the positive or negative trend in the association between adverse childhood experiences and the commitment offense types.Two sets of models were performed. First, an unadjusted model showed the associations of adverse childhood experiences on commitment offense type. Second, adjusted models controlling for the three delinquent career confounders were used with Mantel-Haenszel odds ratios.As shown in Table 1, adverse childhood experiences were variously associated with homicide as a commitment offense. For the total sample in the unadjusted model, wards with five adverse childhood experiences were 49% less likely and wards with six adverse childhood experiences were 72% less likely to be committed for homicide. Indeed, none of the adverse childhood experiences had an odds ratio >1 for the total sample. The test of homogeneity was significant, indicating there were not equal odds across scores on adverse childhood experiences, and the score test for the trend of odds was significant and revealed a negative association between adverse childhood experiences and homicide. In the adjusted model, the only significant association was for wards with six adverse childhood experiences who were 68% less likely to be committed for homicide.The effects of adverse childhood experiences were inconsistently related to homicide across racial and ethnic groups. In the unadjusted model, African Americans with one adverse childhood experience were 39% less likely to be committed for homicide. In the adjusted model, Hispanics with four adverse childhood experiences were two times more likely to be committed for homicide. In the unadjusted model, whites with five (68% less likely) and six (85% less likely) adverse childhood experiences evinced a reduced likelihood of homicide offending.As shown in Table 2, there were dramatic associations between adverse childhood experiences and being committed for a sexual offense; however, the effects varied across models. For the total sample in both the unadjusted and adjusted models, wards that had two or more adverse childhood experiences were more likely to be committed for a sexual offense. Indeed, those with seven adverse childhood experiences had an odds ratio of 6.3 for sexual offending. The adjusted models showed disparate effects. African Americans with three (Odds Ratio = 1.76) and five (Odds Ratio = 4.67) adverse childhood experiences were more likely to be committed for a sexual offense. Hispanics with two to six adverse childhood experiences were more likely to be committed for a sexual offense and the trend was positive. For whites, only those with five adverse childhood experiences were significantly likely to be committed for a sexual offense and the trend was positive and linear.As shown in Table 3, adverse childhood experiences mostly showed a negative association with serious person/property offending. For the total sample in both the unadjusted and adjusted models, wards with three, four, five, and six adverse childhood experiences were negatively associated with serious person/property offending. For instance, those with three adverse childhood experiences were 44% less likely to be committed for serious person/property offending. The likelihood was 49% less likely for four adverse childhood experiences, 60% less likely for five adverse childhood experiences, and 84% less likely for six adverse childhood experiences. No significant effects were found in the adjusted model for African Americans. However, Hispanics with three adverse childhood experiences were 60% less likely to be committed for serious person/property offending, those with four experiences were 73% less likely, and those with five experiences were 78% less likely. For whites, different sorts of effects were found. Those with one adverse childhood experience were 387% more likely and those with two adverse childhood experiences were 329% more likely to be committed for serious person/property offending.Adverse childhood experiences have been shown to have long-term effects on behavioral functioning and have recently been utilized by criminologists as a conceptual framework to understand antisocial development [6,12,13,14,15,16,17,18,19,20]. Using a large sample of 2520 serious male juvenile delinquents, the current study examined whether there were differential effects on crime based on the number of adverse childhood experiences that a ward experienced, whether these effects varied across race and ethnic groups, and whether the adverse childhood experiences’ association with offending depended on the commitment offense type. The results revealed that adverse childhood experiences do not operate on criminal outcomes in a directly linear manner and have considerably differential effects.First, it is not the case that a youth who has been exposed to the greatest number of adverse childhood experiences will necessarily be at the greatest risk of offending. Youth who had six adverse childhood experiences were 72% less likely to be committed for homicide, 413% more likely to be committed for a sexual offense, and 50% less likely to be committed for a serious person/property offense. Across models, these inconsistent effects were common. In the unadjusted model, African Americans with one adverse childhood experience were 39% less likely to be committed for homicide, whereas in the adjusted model for whites, those who had six adverse childhood experiences were 85% less likely to be committed for homicide. Still, there was also evidence of a gradient effect for crimes other than homicide. There was a clear positive trend where more adverse childhood experiences were associated with a greater likelihood of sexual offending, and in the total sample and among Hispanics, a negative trend where more adverse childhood experiences were associated with less risk for serious person/property offending.Second, adverse childhood experiences had differential associations with the three offenses by racial and ethnic group. African Americans with one adverse childhood experience were 39% less likely to be committed for homicide in the unadjusted model. Hispanics with four adverse childhood experiences were 100% more likely to be committed for homicide. Whites with five or six adverse childhood experiences were 68% and 85% less likely, respectively, to be committed for homicide in the unadjusted model. There was more consistency in the effects for sexual offense where adverse childhood experiences were associated with elevated risk for sexual offending across racial and ethnic groups. However, even within this consistency, there were differences. Among wards with five adverse childhood experiences, the likelihood of being committed for sexual offense was 367% for African Americans, 314% for Hispanics, and 257% for whites. Whites with fewer adverse childhood experiences had extremely high likelihood of commitment of a serious person/property offense (Odds Ratio = 4.87 for one adverse childhood experience and Odds Ratio = 4.29 for two adverse childhood experiences), yet African Americans and Hispanics with between three and five adverse childhood experiences had reduced likelihood of serious person/property offending. Although non-white offenders experience generally more disadvantaged backgrounds [24,25,26,38] than whites, that was not the case in these current data. Whites had the highest mean adverse childhood experiences (Mean = 2.43, Standard Deviation = 1.71), followed by Hispanics (Mean = 2.01, Standard Deviation = 1.42) and African Americans (Mean = 1.94, Standard Deviation = 1.41), and these group differences were significant (F(3, 2520) = 17.00, p < 0.0001). This finding is consistent with prior work which indicated a higher proportion of white juvenile offenders at the extreme upper end of the range of adverse childhood experiences (22% compared to 15% of African American youth and 11% of Hispanic youth) [22].Third, adverse childhood experiences were not associated with homicide, sexual offending, and serious person/property offending in the same ways. The effects were mostly negative for homicide, were sharply positive for sexual offending, and were negative for serious person/property offending, with the exception of whites where the effects were positive. Other criminologists have also recently found evidence of adverse childhood experiences on specific forms of crime and negative behaviors. For instance, Perez and colleagues [39,40] found that adverse childhood experiences were directly associated with substance abuse and with serious, violent, and chronic delinquency in their study of more than 64,000 delinquent youth in Florida. However, the effects were rather small, with adverse childhood experiences increasing the likelihood of substance abuse by 14% and of serious, violent, and chronic delinquency by 8%. Additional research is needed to examine how and to what degree adverse childhood experiences are associated with variance in specific forms of serious crime, including homicide and aggravated robbery. Moreover, while there is considerable evidence that sexual abuse victimization can be a distal predictor of subsequent sexual offending [17,18,19,20,21], much less is known about how poverty, physical abuse, emotional abuse, and other deprivations are linked to other forms of violence.There are clear policy applications from our findings. Any program that attempts to reduce various forms of childhood abuse and neglect, and to generally reduce the incidence of adverse childhood experiences is certainly a worthy endeavor. Even one type of adverse childhood experience can be the driving force in propelling a youth toward delinquent and other antisocial conduct. Central among these programs would be parenting education and training that provides adaptive, instructive ways for parents to interact with and discipline their children. These healthy alternatives engender a cascade of healthy development compared to the pernicious and enduring effects of abuse and neglect.There are key limitations to the present study that should be considered to not only contextualize the findings, but also to guide future research. There are numerous important criminological variables, such as religiosity, family support, and various social bonds that we lacked measures of and thus were not able to specify them in the multivariate models. Although the adjusted models controlled for three important parameters of the delinquent career, there were omitted variables that would be helpful to specify the mechanisms by which adverse childhood experiences translate into elevated or attenuated risks for various types of serious delinquency. It is likely that the inconsistent ways that adverse childhood experiences manifest in crime in the present study are reflective of the mechanism advanced by differential susceptibility theory [41,42]. Some youth suffer considerably from even one adverse childhood experience, such as the case of whites in the current data and their risk for serious person/property offending, whereas others are able to avoid specific forms of crime despite many adverse childhood experiences, such as the case of whites in the current data and their risk for homicide. In their landmark study, Caspi and colleagues [43] revealed how specific genetic polymorphisms moderate abuse experiences and explain differential antisocial outcomes. Biosocial scholars should model adverse childhood experiences relative to well-known genetic risks for antisocial traits and behaviors, such as dopaminergic and serotonergic genes, to more clearly articulate for whom adverse childhood experiences are most damaging.Another fruitful research endeavor is to examine adverse childhood experiences vis-à-vis psychopathic traits and serious offending. Anda and colleagues [44] found that persons with a score of five or more (maximum of eight) adverse childhood experiences had nearly a three-fold increase in psychopathic personality features. Given the strong association between psychopathy and the most severe forms of juvenile delinquency [45,46,47,48] and the association between psychopathy and instrumental forms of offending, such as murder, rape, and armed robbery, it is likely that some of the youth in the current study exhibited these traits, but we were unable to measure them. Moreover, a bevy of studies have revealed that psychopathic offenders often experience severely abusive and neglectful childhoods [49,50,51,52,53,54], which contributes to their pernicious antisocial development. However, no studies to our knowledge have explicitly examined the interplay between psychopathy and adverse childhood experiences relative to crime.The current study examined only three types of offending; thus, future research should extend this by examining other types of offenses. There could be interesting developmental sequela associated with specific forms of offending. For instance, Dube and colleagues’ [55] study of the original ACE data found that each additional adverse childhood experience increased the likelihood of early initiation of drug use two- to four-fold, and those with the greatest number of adverse childhood experiences were seven to 10 times more likely to be early-onset drug users. However, it is likely that some adverse childhood experiences, such as parents who openly use drugs in front of their children or furnish drugs to their children, would produce a stronger association with drug crimes than poverty or emotional abuse, for instance. In addition, the coding of adverse childhood experiences might obscure variation. Counts used here reflect the diversity of experiences but not the frequency or severity of those experiences, nor the relationship of the youth to the perpetrator. It is possible that frequency and severity vary by subgroup and the resultant error from this omission contributes to the discrepant findings. Additional study of adverse childhood experiences and specific forms of crime could also inform the understanding of criminal specialization to the degree that a specific form of abuse is associated with subsequent offending, as in the case of sexual abuse/sexual assault [17,18,19,20].Adverse childhood experiences are variously associated with commitment offenses and work in similar and dissimilar ways across racial and ethnic groups. Youth with more adverse childhood experiences were generally less likely to be committed for homicide or serious person/property offending but more likely to be committed for sexual offenses, particularly when models were adjusted for onset of first commitment, previous delinquent adjudications, and prior out of home placements. The effects of adverse childhood experiences on sexual offending are robust and work similarly for African Americans, Hispanics, and whites; however, there remained differential effect sizes. The study of adverse childhood experiences is an invaluable framework for criminology, and one where greater specificity of the mechanisms and ultimate effects of these experiences is needed.The data that support the findings of this study are available from Chad Trulson, but restrictions apply to the availability of these data, which were used under license for the current study, and so are not publicly available. Data are however available from the corresponding author upon reasonable request and with permission of Chad Trulson.Matt DeLisi, Justin Alcala, Chad Trulson, and Michael Baglivio conceived and designed the study, Matt DeLisi conducted the data analysis, Matt DeLisi, Justin Alcala, Abdi Kusow, Andy Hochstetler, Mark Heirigs, Jonathan Caudill, Chad Trulson, and Michael Baglivio all contributed to the writing of this article.The authors declare no conflicts of interest.Table of odds for homicide as commitment offense. ACE, Adverse Childhood Experiences. OR, Odds Ratio.* p < 0.05; ** p < 0.01; *** p < 0.001.Table of odds for sexual offense as commitment offense.* p < 0.05; ** p <0.01; *** p < 0.001.Table of odds for serious person/property offense as commitment offense.* p < 0.05; ** p < 0.01; *** p < 0.001.

Med-MDPI/ijerph_2/ijerph-14-03-00332.txt

@@ -0,0 +1,6 @@
+Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).While DeKalb County, Georgia, offers free radon screening for all eligible residents, portions of the county remain relatively under-sampled. This pilot study focused on 10% of the census tracts in the county with the lowest proportion of radon testing; most were in southern DeKalb County. In total, 217 households were recruited and homes were tested for indoor radon concentrations on the lowest livable floor over an eight-week period from March–May 2015. Tract-level characteristics were examined to understand the differences in socio-demographic and economic factors between the pilot study area and the rest of the county. The pilot study tracts had a higher proportion of African Americans compared to the rest of DeKalb County (82% versus 47%). Radon was detected above 11.1 Bq/m3 (0.3 pCi/L) in 73% of the indoor samples and 4% of samples were above 148 Bq/m3 (4 pCi/L). Having a basement was the strongest predictive factor for detectable and hazardous levels of radon. Radon screening can identify problems and spur homeowners to remediate but more research should be done to identify why screening rates vary across the county and how that varies with radon levels in homes to reduce radon exposure.Radon has been classified as a known human lung carcinogen [1]. In pooled case-control studies in both Europe and North America, researchers found direct evidence that residential exposure to radon was associated with lung cancer risk [2,3]. In the United States, radon is currently the second leading cause of lung cancer, behind only smoking cigarettes; it is the number one cause of lung cancer in non-smokers [1]. From 1988 to 2013, approximately 25 million tests for radon were completed in the U.S. although most of these were the result of real estate transactions [4]. However, the results of these screening tests are often not shared widely as a lack of resources constrains the information about the proportion of homes above hazardous levels.The U.S. Environmental Protection Agency (EPA) has provided radon risk maps for all counties in the U.S. and categorized them into three tiers according to potential radon exposure: zone 1 counties have predicted indoor radon screening levels >148 Bq/m3 (4 pCi/L), zone 2 are between 74 and 148 Bq/m3 (2 and 4 pCi/L), and zone 3 are estimated to be <74 Bq/m3) (2 pCi/L) [5]. There are four zone 1 counties in Georgia: DeKalb, Fulton, Gwinnett, and Cobb. The DeKalb County Board of Health (DBH), through the Division of Environmental Health, offers free radon screening to all residents [6]. Although the county—which consists of 144 census tracts—offers radon screening through homeowner initiated tests, screening prevalence is overall relatively low and some census tracts are estimated to have less than 0.5% of the eligible households screened for radon. Since the screening program is voluntary and provided to those who request the service, there is a need to understand exposure in areas of the county which have relatively low screening prevalence. Examining exposure in these areas of the county can determine whether indoor radon exposure may constitute a significant health risk.The objectives of this study were:
+		
+        To describe a pilot study of randomly selected households for recruitment for in-home radon measurements in 14 census tracts in DeKalb County, GA, USA
+        To analyze tract-level socio-economic and demographic characteristics to understand differences between the pilot study tracts and the remaining tracts in the county
+        To analyze radon levels in homes and identify housing characteristics associated with radon in homes.
+        To describe a pilot study of randomly selected households for recruitment for in-home radon measurements in 14 census tracts in DeKalb County, GA, USATo analyze tract-level socio-economic and demographic characteristics to understand differences between the pilot study tracts and the remaining tracts in the countyTo analyze radon levels in homes and identify housing characteristics associated with radon in homes.Georgia State University graduate students from the Department of Geosciences and the School of Public Health recruited additional student volunteers from both departments. All students and faculty members who were involved in the field work completed human subjects’ research training via the Collaborative Institutional Training Initiative (CITI) collaborative (https://www.citiprogram.org/). In addition, each member of the team received additional training with details for household recruitment, surveying, use of the radon test kit, and information regarding radon resources to be shared with participants. Training took place in March 2015, and household recruitment took place in March through May 2015. Because this pilot study involved interaction with human subjects, institutional review board (IRB) approval was required and obtained from the Georgia State University IRB (IRB No. H14542).To examine radon in under-sampled census tracts, the county was evaluated for percentage of single family homes screened for radon. Data was provided by the DBH and from Air Chek (Air Chek Inc., Mills River, NC, USA). The screening prevalence (see Supplemental Materials Figure S1) was calculated as follows: the numerator was the number of radon tests and the denominator was the number of residential parcels. We ranked the county by the radon screening prevalence and then chose the bottom 10% of the tracts (n = 14) for our sampling. The 10% was used as a pilot to understand the residential exposure in the low-screening areas in order to prepare a study focused on screening-disparity research. The recruitment goal for the project was to collect indoor air samples on the bottom livable floor from 200 houses. A list of all study-eligible, single-unit land parcels within the selected 14 census tracts was compiled. From this list, households were randomly selected with allocation across census tracts that was proportional to the total number of single-unit parcels within the census tract. As a result, selection probabilities were approximately equal. For each census tract, we generated a list of randomly selected households and addresses were drawn from this list for visits. Starting in March 2015, groups of two traveled to the census tracts weekly, typically on Saturdays. Randomly selected addresses were provided for initial approach. If the household did not respond (because no one came to the door), a household on the same street within three to five households was contacted. If no houses were available, the next address on the list was visited.Homeowners were approached and provided information about the project and invited to participate. Informed consent was obtained and documented, a survey was conducted, and a radon test kit was installed in the home which was retrieved within 3–5 days. The survey was conducted to collect housing characteristic data by self-report or observation including: age of home, foundation type, building type, and housing type. In addition, the primary respondent answered questions about the presence of any children under 18 years of age, smoking and any prior knowledge of radon. The test kits (Air Chek Inc., Mills River, NC, USA) were hung at eye-level on any interior wall of the lowest livable floor. For every 20 homes sampled, a duplicate test kit was placed on the same floor to measure the reproducibility of the test kit and laboratory analysis. For successful completion of the radon test (aka retrieval 3–5 days later), the resident received a $15 Walmart gift card. Upon retrieval from the home, each kit was sealed and immediately dropped in the mail for analysis by Air Chek laboratories. Results were shared with the principal investigator and provided via mail or email to the homeowner as well.Radon test results and questionnaire data were manually entered into Microsoft Excel. The data were then analyzed using descriptive statistics and bivariate and multivariate logistic regression. For analytical purposes, all non-detects were assigned a value of 5.5 Bq/m3 (0.15 pCi/L) or one half of the detection limit of the test. To evaluate initial differences, radon data were plotted using box and whisker plots. However, because the radon data were not normally distributed, non-parametric comparisons were performed using Kruskal-Wallis tests. In addition, two binary categories were used to assess statistically significant associations between housing characteristics and radon: one variable that was developed based the detection of radon above the sample kit detection level (11.1 Bq/m3 or 0.3 pCi/L) and one that assessed hazardous levels of exposure to radon (>148 Bq/m3 or >4 pCi/L). All statistical tests were conducted using Stata version 13.0 (Stata, College Station, TX, USA).In addition to the data collected in the pilot study, data from the 2011–2015 American Community Survey (ACS) database were also examined. The following variables were extracted from the ACS database for all tracts in DeKalb County: educational attainment, income, race and poverty status. Mean percentages were generated for each variable of interest for the pilot tracts and the remaining 130 tracts and compared using Wilcoxon rank-sum tests.As shown in Table 1, the most prominent difference between our 14 sampled tracts and the remaining 130 census tracts in DeKalb was a higher proportion of African Americans present in the 14 pilot study tracts (82% versus 47% respectively); a difference that was statistically significant via Wilcoxon rank-sum test. Comparison of educational attainment for the pilot study tracts and the rest of the county suggested a lower proportion with a bachelor’s degree (or higher) compared to the rest of the county. Mean and median income were higher in the rest of the county and the proportion in poverty was lower for the rest of the county, but these were not statistically significant differences.Recruitment logs of houses visited were kept in an attempt to assess response rate. Based on these recruitment logs, we estimated a response rate of 33%. A total of 217 homes were recruited during the March to May 2015 pilot study; however, 16 radon test results were not valid and therefore only 201 households could be included in the analysis of household characteristics. To assess the geographic distribution of our sampled households, we performed a nearest-neighbor analysis (α = 0.01; one-tailed) on the sampled locations at each census tract, respectively, based on their minimum bounding rectangles. The results suggest that they represent either random or dispersed distribution. The analysis, however, may be limited in three or four tracts because of their sparse locations.As shown in Table 2, the homes were built over a broad time range as construction year of homes ranged from 1950 through 2014; the median year of home built was 1997. Approximately half of homes tested were multi-story homes (n = 113, 52%), 28% were split-level homes, and 20% were ranch-style homes. Homes were identified (by observation) to be mostly frame construction (64%), with the remaining homes either brick/block construction (20%) or some combination of frame, brick and/or block (15%). Household participants reported that the majority of foundations were slab (61%), with 27% with basements and 11% with crawl spaces. In terms of risk factor knowledge assessment, we found the following, that 49 respondents (23%) reported that someone smoked cigarettes in the house; more than half reported they had not heard of radon before (115 participants), and almost as many reported that children under 18 years of age resided in the home.Of the 201 radon test kits that came back with a valid test result, 154 (78%) were collected from the 1st floor with the remaining collected from the basement (Table 2). The radon concentrations were positively skewed (Figure 1). Even after log transformation, the radon test results did not appear to be normally distributed (data not shown). Radon concentrations ranged from <5.5 Bq/m3 (<0.3 pCi/L) to 400 Bq/m3 (10.8 pCi/L), with a mean of 45.6 Bq/m3 (1.2 pCi/L) (Std. Dev. 53.3 Bq/m3) and a median of 29.6 Bq/m3 (0.8 pCi/L). Approximately 1/4th of the samples with a valid result found radon concentration below detectable levels of the test kit. In Figure 2, the results are provided via a map of DeKalb County and, as shown, the analysis does not suggest spatial trend for the 14 sampled census tracts. The northern-most census tract was on average higher (with fewer samples) but this was not statistically significant when examined with the Kruskal-Wallis test (p = 0.17).Housing and building type had little influence on radon concentrations, while foundation type had a substantial influence (Figure 3, Figure 4 and Figure 5, respectively). Average radon concentrations were the highest for ranch homes at 56.6 Bq/m3 (1.53 pCi/L), with multi-story and split-level homes at 43.2 Bq/m3 (1.17 pCi/L) and 42.6 Bq/m3 (1.15 pCi/L), respectively (not statistically significant differences). Building type, Figure 4, was also not found to be significantly associated with radon concentrations (p = 0.52 via the Kruskal-Wallis test). With respect to foundation type, basement had the highest interquartile range and there was a statistically significant difference between the three foundation types, p-value = 0.0001 (via Kruskal-Wallis). Houses with basements as their foundation type had the highest average radon concentration of 71.2 Bq/m3 (1.92 pCi/L) with slab and crawl space average concentrations substantially lower at 36.5 Bq/m3 (0.98 pCi/L) and 29.8 Bq/m3 (0.81 pCi/L), respectively.When examining the location of the radon test kit, there was a statistically significant difference between the basement and first floor as shown in the plot below and evidenced by the Kruskal-Wallis test (p value = 0.001), Figure 6.The results from bivariate and multivariate logistic regression are shown in Table 3. We examined housing characteristics associated with detecting radon above the test kit limit and found that foundation type was statistically significantly associated with increased odds of detectable levels of radon in both bivariate and multivariate models. Households with a basement were found to have increased odds of detecting radon compared with slab foundations and crawl spaces (which had the lowest odds of detectable radon). We also analyzed factors associated with odds of detecting radon above EPA actionable levels (148 Bq/m3 or 4 pCi/L). In our analysis of housing characteristics that were associated with detection of hazardous levels of radon, both foundation type and location of sample were statistically associated with increased odds of detecting radon >148 Bq/m3 in bivariate analysis. However, when all three housing characteristics were included in the model, both remained positively associated with increased odds of detecting higher levels of radon, but neither remained statistically significant.In our pilot study in DeKalb County, GA, USA we successfully recruited 217 geographically random households in 10% of the under-sampled census tracts of the county to participate in radon screening. Based on our spatial analysis, we argue that the sampled locations were geographically representative in their tracts given their random or dispersed distribution. Our estimated response rate of 33% is very similar to Duckworth et al. (2002), which had a 29% response rate for randomly selected households they interviewed and screened for radon in DeKalb County, Illinois, USA [7].In the pilot samples, we found that 74% of households had radon above 11.1 Bq/m3, with 18% above 74 Bq/m3 (2 pCi/L (EPA moderate risk level)) and 4% above 148 Bq/m3 (4 pCi/L (EPA high risk level)). Aggregate reports from DeKalb County, GA, suggest that approximately 26% of the county samples were found to have moderate levels (between 74 Bq/m3 and 144 Bq/m3) and 17% were found to have hazardous levels [8]. In a recent analysis of more than 3900 indoor samples from DBH and real estate transaction data from DeKalb County, GA over a 20-year period, Neal (2016) [9] found mean indoor concentrations of 70.3 Bq/m3 (1.9 pCi/L) (compared to 45.6 Bq/m3 (1.2 pCi/L) for our samples). In comparison to estimates from DeKalb County, our mean concentration and proportion of samples in the moderate-to-high range may be lower. However, it is difficult to compare our sample populations because of the difference in sample selection procedures. The DBH program is free to households that select the service. However, there is some evidence that awareness and radon tests may be affected by socio-demographic factors [10]. It is also possible that those who select the radon test do so because they are aware of high radon tests in the neighborhood, which might result in a selection of samples that are elevated on average. In our study, approximately half of our participants had not heard of radon before and this was not associated with radon detection in homes or with hazardous levels of radon (data not shown).Whether or not the lower concentration of radon we found is more typical of the county when randomly sampled would be difficult to say without additional evidence. However, it is possible that these regions were not heavily sampled because they are not prone to high indoor radon measurements and the localized radon potential is low. A recent study conducted in DeKalb County, GA that measured gamma radiation as a potential radon predictor found that, from our pilot study areas, only areas of southeastern DeKalb County had higher radon potential [11]. While our samples suggested lower indoor concentrations (on average), there is still a potential for increased risk of cancer as a result of radon exposure even below EPA action levels. Darby et al., in their case control study on radon and lung cancer, found a linear dose-response relationship between radon concentration and lung cancer risk for households <200 Bq/m3 [3].In terms of other differences between our sample and the rest of DeKalb County, GA when comparing our 14 census tracts to the rest of the county, we found a statistically significant difference in the proportion of African Americans compared to the rest of the county, 82% versus 46% respectively. Educational attainment, specifically attaining a Bachelor’s degree or higher, was statistically significantly lower in the pilot census tracts (via one-sided test). The other factors, including the proportion in poverty and income, were somewhat lower in our pilot study tracts but were not statistically significant in our analysis. Whether or not socio-economic and demographic differences can fully account for differences in radon screening in the pilot study census tracts is difficult to determine. In their analysis of radon risk perception and socio-demographic correlates using data from the National Health Interview Survey, Halpern and Warner found that minority respondents were less likely to be aware of the harmful effects of radon, yet apparently willing to test once made aware, suggesting that educational outreach programs were not effectively communicating [12]. A smaller, qualitative study in Michigan explored radon perceptions in African Americans and found some inaccurate beliefs regarding radon and concern about ability to avert the exposure [10]. Although we can compare ACS data regarding the census tracts in our study, we did not collect socio-economic or demographic data from our participants and cannot determine whether or not the participants in our random sample would be representative of these census tracts, although the geographic dispersion of the households does suggest a spatially representative sample. As one potential indicator of similarities with health indicators, approximately 20% of the homes in our study had at least one smoker in them, which is similar to data for DeKalb County adults (average 17% adult smokers) [13]. Ultimately, our results suggest that further investigation into socio-demographic and economic indicators for radon awareness and screening in DeKalb County need to be explored to understand how and why people participate in the program and how to maximize its impact as almost half of our participants had not heard of radon but were willing to participate.In our study, we found that foundation type (i.e., having a basement) was a significant predictor of increased odds of detecting radon. We found that household foundation type was also predictive of odds of radon concentration above EPA action levels (148 Bq/m3 (≥4 pCi/L)) but not when controlling for other housing characteristics. However, it is likely that we were limited in analyzing factors associated with hazardous levels of radon as only 4% of samples (8/201) were found to have this level of radon in the sample. In a study from under-sampled areas in New York, researchers found that a higher proportion of samples from basements exceeded 148 Bq/m3 compared to measurements on other floors of the home [14]. Similarly, researchers in Minnesota also found that concentrations in basements were about twice as high as measurements made on other floors of the home [15]. Our study results are also supported by recent analysis of data from the county from the DBH screening program and real estate transactions in which foundation type was also statistically significantly predictive of hazardous levels of radon, but few other housing characteristics were associated with increased odds of hazardous indoor radon levels [10]. Overall, these data suggest that in DeKalb County, GA, household foundation type should be considered in new construction as a potential way to reduce exposure to radon.Limitations. This pilot study aimed to understand the spatial distribution of radon within DeKalb County, GA in census tracts with low screening rates. While our sample was randomly assigned, we had difficulty reaching the initially identified household. We estimate our acceptance rate to be approximately 33% (based on recorded visitor logs and willingness to participate once approached). While this is low, it is similar to other studies that attempted to randomly recruit participants [7,14]. The difficulty to reach the desired pilot number of homes may have somewhat impacted the random selection of the households. However, it did not impact the random geographic distribution of the samples. For future work, we propose sharing the information about the study at community and neighborhood meetings prior to study recruitment and also contacting via mailers to further enhance outreach and enhance the ability to generalize results. Seasonal variation was also a limitation to this study. The screening and recruitment process was conducted during the spring months in Georgia in 2015. A requirement for accurate testing using the active-carbon detectors was that all windows and doors remain shut with no fans or air conditioning blowing on the tests, allowing maximum concentrations of radon to be observed. It is possible that households did not observe these conditions and that this reduced detection and diminished concentrations of radon in the home during screening tests. However, the agreement between short-term and long-term radon tests has been estimated to be approximately 90% [4]. Lastly, while our duplicate samples did not suggest any differences in average concentrations taken at the same time (data not shown), if time permitted, repetitive testing of participating homes as well as long-term testing should be examined to understand and identify seasonal variations of radon concentrations in the home as well as more thoroughly estimated exposure.Radon is the second leading cause of lung cancer nationally and its concentration in the indoor environment remains unknown unless tested. Our pilot study of 201 households in 10% of the lowest screened census tracts in DeKalb County, GA, was designed to avoid the potential bias of self-selecting free screening programs. We found that radon exceeded EPA moderate risk levels in 18% of households and exceeded high risk in 4% of the homes tested. While this proportion is lower than other aggregate results from the county, it does suggest that households in under-sampled areas may also be at risk for potential health outcomes from exposure to residential radon. Foundation type—specifically owning a home with a basement—was the factor most likely to be associated with increased radon. Awareness of the dangers of radon is still a barrier that needs to be addressed as half of our study participants had not previously heard of radon but were willing to participate in our study, which suggests that more effective outreach and communication strategies may be successful at increasing screening in under-sampled areas. As the entire county is considered a high-risk zone by EPA, more needs to be done to understand the socio-economic and demographic factors that may be influencing screening prevalence in order to capitalize on the free radon screening resources available to all eligible county residents.The following are available online at www.mdpi.com/1660-4601/14/3/332/s1, Figure S1: Radon screening prevalence in DeKalb County, GA.The study was supported in part by the National Institute of Minority Health and Health Disparities 1P20MD009572-01 grant. We would like to thank all of the participants who agreed to be in the study. We would also like to thank the two dozen students from Georgia State University School of Public Health and Department of Geosciences (graduate and undergraduate) who worked on the study. We also thank the reviewers for their helpful comments.Dajun Dai, Richard Rothenberg, Christine E. Stauber and Scott R. Weaver conceived and designed the experiments; Sydney Chan, Christine Stauber, Dajun Dai and Jeremy E. Diem. performed the experiments; Sydney R. Chan and Christine Stauber analyzed the data; Sydney R. Chan, Christine E. Stauber, Dajun Dai, Scott R. Weaver, Richard Rothenberg and Jeremy E. Diem wrote and edited the paper.The authors declare no conflict of interest. The founding sponsors had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, and in the decision to publish the results.Histogram of radon results from homes tested in the DeKalb County pilot study in 2015.Graduated colors representing radon concentrations for each sample location within the 14 sampled tracts of DeKalb County, GA, in 2015.Association with housing type and radon concentration in the pilot study of homes tested in DeKalb County in 2015.Association with building type and radon concentration in the pilot study of homes tested in DeKalb County in 2015.Association with foundation type and radon concentration in the pilot study of homes tested in DeKalb County in 2015.Association with sample location and radon concentration in the pilot study of homes tested in DeKalb County in 2015.American Community Survey data comparing the 14 selected tracts to the rest of DeKalb County.* Comparisons indicated a statistically significant difference (p < 0.05) via a two-sided Wilcoxon rank-sum test; ** Comparison indicated a statistically significant difference via a one-sided Wilcoxon rank-sum test.Descriptive results of the homes tested in the DeKalb County pilot study in 2015.Bivariate and multivariate logistic regression assessing housing characteristics and radon in homes in DeKalb County, 2015.

Med-MDPI/ijerph_2/ijerph-14-03-00333.txt

@@ -0,0 +1,155 @@
+Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).The aim was to study objectively assessed walkability of the environment and participant perceived environmental facilitators for outdoor mobility as predictors of physical activity in older adults with and without physical limitations. 75–90-year-old adults living independently in Central Finland were interviewed (n = 839) and reassessed for self-reported physical activity one or two years later (n = 787). Lower-extremity physical limitations were defined as Short Physical Performance Battery score ≤9. Number of perceived environmental facilitators was calculated from a 16-item checklist. Walkability index (land use mix, street connectivity, population density) of the home environment was calculated from geographic information and categorized into tertiles. Accelerometer-based step counts were registered for one week (n = 174). Better walkability was associated with higher numbers of perceived environmental facilitators (p < 0.001) and higher physical activity (self-reported p = 0.021, step count p = 0.010). Especially among those with physical limitations, reporting more environmental facilitators was associated with higher odds for reporting at least moderate physical activity (p < 0.001), but not step counts. Perceived environmental facilitators only predicted self-reported physical activity at follow-up. To conclude, high walkability of the living environment provides opportunities for physical activity in old age, but among those with physical limitations especially, awareness of environmental facilitators may be needed to promote physical activity.Physical activity plays an important role in maintaining health and function in old age. Older adults are mostly physically active in the near vicinity of the home [1]. In accordance with the environmental press or person-environment fit model, an individual may limit his or her physical activity when the demands of the environment exceed the capabilities of the individual [2,3]. Low physical activity predisposes adults to loss of muscle strength, balance and endurance, which in turn may cause avoidance behavior and a vicious circle of declining physical activity and further declining capacity [4,5]. Lower physical capacity may lower older adults’ threshold to overcome environmental barriers [6,7], but attractive environments may still motivate older adults to go outdoors and be physically active. Currently, it is unclear whether attractive environmental features affect physical activity differently in older adults with and without physical limitations.Geographical information such as land use mix, population density, and connectivity have been frequently combined into one walkability index reflecting possibilities to walk to different destinations [8]. A higher walkability index, indicating better walkability, has been associated with objectively assessed and self-reported physical activity, but not consistently [9,10]. Such objective environmental measures may not always correlate well with perceptions of the environment [11,12,13]. Subjectively perceived environmental factors reflect the capacity of the individual and the environment used by the individual [12]. Few studies have shown that reporting higher numbers of perceived environmental facilitators for outdoor mobility were associated with higher mobility function in older adults [14,15].Person-environment relationships may be region-specific as environments and physical activity behaviors vary between countries [16,17,18]. Associations between walkability and physical activity have primarily been studied in cities in the USA, Australia and central Europe [9,17]. Urban areas in Nordic countries typically include more blue and green spaces. According to our knowledge, walkability was previously studied in one project among adults in Sweden [19]. Perceived environmental factors have been studied previously in Finland [14,15], but combined with objective environmental features only in one study on the immediate home environment [20]. Including both objectively assessed and subjective perceptions of the environment is important to provide a comprehensive picture of person-environment associations related to physical activity behavior in older adults.Our aim was to study in older adults: (1) associations between objectively assessed walkability of the environment and participant perceived environmental facilitators for outdoor mobility, (2) their association with objective (step count) and self-reported physical activity at baseline, and (3) their association with self-reported physical activity over time. Based on the person-environment fit model, we hypothesized that associations between the environment and physical activity behavior in old age may differ by the presence or absence of physical limitations in the lower extremities; thus we stratified the analyses accordingly.These cross-sectional and longitudinal analyses are part of the “Geographic characteristics, outdoor mobility and physical activity of older people” (GEOage) project. In this project, freely available geographic information characterizing the environment is linked to participant data of the “Life-space mobility in old age” (LISPE) cohort comprising 75–90-years-old community-dwelling people living in Jyväskylä and Muurame in Central Finland [21]. Briefly, a random sample of 2550 was drawn from the population register. Those living independently, able to communicate, residing in the recruitment area, and willing to participate were eligible to participate. In spring 2012, baseline data (n = 848; 62% female) were collected in a home interview [21]. One (n = 816; 62% female) and two (n = 761; 63% female) years later participants were re-interviewed over the phone [22]. At baseline, a subsample of 190 wore a tri-axial accelerometer for seven days following the baseline assessments. Participants were included based on the availability of accelerometers and willingness to participate. Valid days included ≥10 hours of accelerometer wear time. Technical problems (n = 4), <4 valid accelerometer days (n = 11) and >1 days in-between consecutive measurement days (n = 1) led to exclusion of data, thus leaving 174 participants (64% female) for the analyses [23]. Participants signed an informed consent prior to the data collection. The GEOage and LISPE study were approved by the University of Jyväskylä Ethical Committee.Lower extremity performance was assessed with the Short Physical Performance Battery (SPPB), comprising of three tests that assess standing balance (narrow stance, semi-tandem stance, tandem stance), walking speed over 2.44 m, and five timed chair rises [24,25]. Each task was rated from 1 to 4 according to Finnish age- and gender-specific cut-off points [25]. Participants unable to complete the task due to balance- or mobility-related difficulties were assigned a score of 0, those unable to complete the task due to temporary medical conditions, wheel chair use, severe visual impairment, lack of a suitable chair or unwillingness to cooperate were not assigned a score (missing). By summing the task scores, a sum score (range 0–12) was calculated. For those with one task score missing, the final score was multiplied by 1.5. Higher scores indicate better performance. In order to stratify the sample for the analyses, participants were categorized according to the median; those with SPPB score ≤9 were labeled as having lower-extremity physical limitations.Self-reported physical activity was assessed using a seven-point scale combining frequency and intensity of common physical activities [26]. Participants were categorized into at most light physical activity (at most light housework or gardening and short walks once or twice a week), moderate physical activity (at least moderate physical activity <3 h/week), and regular physical activity (moderate physical activity ≥4 h/week or strenuous physical activity). The self-reported physical activity scale and its categorization have shown to have concurrent validity with accelerometer assessed step counts and different measures of mobility [26].In the substudy, average daily step counts were derived from an accelerometer (Hookie, tri-axial AM20 Activity Meter, Hookie Technologies Ltd., Espoo, Finland; sampling frequency 100 Hz and measurement range ±15 g0 (gravity of the Earth)) that was worn on the right hip for 7 days following the face-to-face interview at baseline [23]. Participants were instructed to wear the accelerometer daily from waking up to going to sleep, removing it for water activities only, and registered the wear times of the device in a diary.The number of perceived environmental facilitators was calculated from the checklist for perceived environmental facilitators for outdoor mobility (PENFOM; 16 items), designed to identify the presence of environmental features that people perceive as facilitating their possibilities for outdoor mobility [15]. The PENFOM comprise parks, walking routes, nature, appealing landscape, familiar surroundings, good lighting, own yard, other people outdoors motivate, services or shops near, even sidewalks, walkways without steep hills, resting places by the walking route, peaceful and good quality pedestrian routes, no car traffic, no cyclist on walkways, and safe crossings.Participants’ homes were located on the map by geocoding their home addresses in a geographic information system (GIS) [27]. The GIS-based walkability index was modified from Frank et al. [8]. In line with previous research [19], retail floor area ratio was omitted from the walkability index; thus leaving three components, land use mix, street connectivity, and population density, for which z-scores were calculated and summed to obtain the walkability index. Higher index scores indicate better walkability. The walkability index was calculated within 1 km from the participants’ home. Older adults frequently travel a 1 km distance on foot or by bike [28] and the respective distance has been used to define areas around participants homes for studying environmental factors [17,29].Land use mix was based on the following land use types considered relevant for physical activity [30,31,32]: (1) Residential areas, (2) Services, (3) Sport and leisure facilities, and (4) Forest and semi-natural areas (build and natural green spaces) [33]; Land use mix represents the distribution of land use types in proportion to the dry land area within the 1 km circular buffer area around the participant’s home. We excluded water areas from the total surface area, which are abundantly present in the study area, but inaccessible and at the same time attract people to go outdoors [15]. Agricultural and industrial areas, which are not likely accessible or relevant for physical activity, were accounted for by including them in the total surface area [8]. Land use mix, ranging from 0 (homogeneity; single land use type) to 1 (heterogeneity; even distribution of land use types), was calculated as entropy value using Equation (1) [8]:
+        
+          (1)
+          
+            
+			  
+              
+                Land
+                 
+                use
+                 
+                mix
+                =
+                −
+                 
+                
+                  
+                    
+                      (
+                      
+                        P
+                        1
+                      
+                      )
+                    
+                    ×
+                     
+                    ln
+                    
+                      (
+                      
+                        P
+                        1
+                      
+                      )
+                    
+                    +
+                    
+                      (
+                      
+                        P
+                        2
+                      
+                      )
+                    
+                    ×
+                    ln
+                    
+                      (
+                      
+                        P
+                        2
+                      
+                      )
+                    
+                    +
+                    
+                      (
+                      
+                        P
+                        3
+                      
+                      )
+                    
+                    ×
+                    ln
+                    
+                      (
+                      
+                        P
+                        3
+                      
+                      )
+                    
+                    +
+                    
+                      (
+                      
+                        P
+                        4
+                      
+                      )
+                    
+                    ×
+                    ln
+                    
+                      (
+                      
+                        P
+                        4
+                      
+                      )
+                    
+                  
+                  
+                    ln
+                    
+                      (
+                      4
+                      )
+                    
+                  
+                
+              
+			  
+              
+                P
+                =
+                
+                  
+                    surface
+                     
+                    area
+                     
+                    of
+                     
+                    respective
+                     
+                    land
+                     
+                    use
+                     
+                    type
+                     
+                  
+                  
+                    total
+                     
+                    surface
+                     
+                    area
+                     
+                    of
+                     
+                    buffer
+                     
+                    without
+                     
+                    water
+                     
+                    areas
+                  
+                
+              
+			  
+			  
+			  ln
+			  =
+			  natural
+			  logarithm
+			  
+              
+			
+          
+        Street connectivity was the number of street intersections of walkable ways in all possible directions along the road network up to 1 km distance from the home. Only at least 3-way intersections were included [8] and street intersections within 10 m from each other were merged for the calculations. To obtain walkable ways, motorways, trails, winter roads, railroads and ferry lines over water were omitted from the road network [34]. Absolute resident numbers in the 1 km × 1 km squares of the study area [35] where participants resided was used as an indicator of population density [11].Age, sex and time lived in the current home, calculated based on the date of the latest address change (missing values were imputed with the average (n = 25)), were derived from the national register. The number of self-reported chronic conditions were calculated from a list of 22 physician diagnosed chronic diseases and an additional open-ended question about any other physician diagnosed chronic conditions. For those participating in the accelerometer substudy, daily wear time of the accelerometer was calculated from the self-report diary data. To retain as many participants in the analyses as possible, missing accelerometer wear time values were imputed with the average of the respective individual (if missing 1 or 2 days; n = 15) or the group (if missing for all days; n = 1) after visual inspection of the data to assure ≥10 h of wear time [23]. The assessment month (January–June; 1–6) was used to account for climatological circumstances.Nine participants were excluded from all analyses due to two or three missing task scores of the lower extremity performance test. Thus, baseline analyses included 839 participants and analyses on step counts 174 participants. Longitudinal analyses were conducted excluding participants who moved prior to any study follow-ups (n = 18), and those without any follow-up assessment of self-reported physical activity (n = 34). Consequently, 787 participants were included in the longitudinal analyses and self-reported physical activity was based on the latest available assessment. All analyses were conducted in two ways; including all eligible participants (non-stratified) and stratified by lower-extremity physical limitations (SPPB score 0–9 versus 10–12).Z-scores of the walkability index were divided into tertiles. The number of perceived environmental facilitators was used as such as a dependent variable or divided into tertiles when used as an independent variable. Group differences were tested with independent T-tests (continuous variables) and Chi-square tests (categorical variables). Spearman correlation coefficients were calculated between objective and perceived environmental variables. In all analyses, we consistently found similar associations for “moderate” and “regular” self-reported physical activity when compared to at most light self-reported physical activity. Consequently, we merged these two categories into one category “at least moderate physical activity”. Walkability index and perceived environmental facilitators were included in separate regression models. All regression analyses were first adjusted for age and sex, and then in addition adjusted for number of chronic conditions, years lived in the current home, years of education, and climatological circumstances one at a time and jointly (fully adjusted). In the non-stratified models, lower-extremity physical limitations was added as an independent variable. Regression coefficients (B and standard error) derived from Generalized Linear Models (GLM) are reported to describe cross-sectional associations between walkability variables and number of perceived environmental facilitators (identity link) at baseline (study aim 1). In addition, GLM were used to study associations between walkability variables and step counts (log link transformation) at baseline (study aim 2). Models studying step counts were additionally adjusted for accelerometer wear time. Logistic regression models were used to calculate odds ratios (OR and 95% confidence intervals) for reporting at least moderate self-reported physical activity at baseline (study aim 2). Logistic regression analyses stratified by baseline physical activity were conducted to predict at least moderate self-reported physical activity at the latest follow-up (study aim 3).SPSS Statistics 22 (IBM Inc., Armonk, NY, USA) was used for all statistical analyses, and statistical significance was set at p < 0.05. ArcMap 10.3.1 (ESRI, Redlands, CA, USA) was used to join map layers and create GIS variables.In the full baseline sample, participants with lower-extremity physical limitations were older, and they had more chronic conditions, lower levels of education and lower levels of physical activity than those without physical limitations (Table 1). In the accelerometer subsample, those with lower-extremity physical limitations had lower levels of physical activity and a higher number of chronic conditions. The Spearman correlation coefficient between the walkability index and perceived environmental facilitators was 0.2 (p < 0.001).Living in an area characterized by a higher walkability index was associated with reporting higher numbers of environmental facilitators (Table 2). Each tertile of higher walkability progressively increased the number of environmental facilitators reported by participants with lower-extremity physical limitations, while among those without physical limitations the regression coefficients were similar for the middle and highest tertile of walkability.Table 3 shows that participants living in areas with the highest walkability index had higher step counts than those living in an area with the lowest walkability index. However, in participants with lower-extremity physical limitations, the association was statistically significant only in the fully adjusted model (p = 0.019). Perceived environmental facilitators were not associated with step counts irrespective of lower-extremity physical limitations.The odds for reporting at least moderate physical activity at baseline were higher for participants perceiving higher numbers of environmental facilitators than for those in the lowest tertile (Table 4), and the association was especially strong among those with lower-extremity physical limitations. Living in an area with the highest walkability index was associated with higher odds for reporting at least moderate physical activity, but statistical significance was reached in the non-stratified sample only.Walkability index was not associated with self-reported physical activity at the latest follow-up (Table 5). In participants with at least moderate physical activity at baseline, higher numbers of perceived environmental facilitators for outdoor mobility were associated with reduced odds for maintaining at least moderate self-reported physical activity over time in those with lower-extremity physical limitations only, but the association attenuated after full adjustment (p = 0.070). In participants with low physical activity at baseline, however, reporting the highest number of perceived environmental facilitators more than doubled the odds for reporting at least moderate physical activity at the latest follow-up, but the associations reached statistical significance in the non-stratified sample only.This study provides a comprehensive picture of associations between objectively assessed walkability of the environment and perceived environmental facilitators for outdoor mobility, and how these factors relate to objective and subjective measures of physical activity in older adults with and without lower-extremity physical limitations. The results show that higher numbers of perceived environmental facilitators and better walkability were associated with higher physical activity levels, but not consistently across physical activity measures and time points. Furthermore, the current results show that better walkability was associated with higher numbers of perceived environmental facilitators, although rather weakly. Finally, the presence of lower-extremity physical limitations affected the strength of some person-environment relationships.A major part of older adults’ physical activity is carried out in the home environment [1], but moving through larger areas may add to the accumulation of physical activity [36]. The area, where an individual with physical limitations moves, shrinks and ultimately becomes restricted to the neighborhood [37,38]. Consequently, people with physical limitations may consider a smaller area when reporting perceived facilitators for outdoor mobility than those without physical limitations who move through larger areas in their everyday life [28]. In addition, those without physical limitations may be more motivated and better able to reach attractive destinations for outdoor mobility and physical activity beyond their own neighborhood. This may explain why the associations between perceived environmental facilitators and self-reported physical activity were somewhat weaker for participants without lower-extremity physical limitations.The walkability index is an objectively assessed measure characterizing the environment. In line with previous research, a higher walkability index did not fully translate into higher numbers of environmental factors facilitating outdoor mobility reported by older adults [11,12]. Theoretically, more road intersections and more people living in a designated area implies better availability of or access to services such as shops [8,39]. Having services within walking distance from home has been found to motivate older adults to go outdoors and be physically active [39,40]. However, other factors such as nature and water areas may also motivate people to go outdoors and be physically active [15,30]. The land use mix component of walkability to some extent covers such facilitators, but it does not capture specific preferences of individuals [32]. Because of the larger effort required to move outdoors and be physically active in those with physical limitations [41], these individuals are more likely to convey strategies to conserve their energy [38,42]. Consequently they may be more specific about which features in the neighborhood environment they consider to be attractive. The walkability index is a rather generic measure [43]. In future research, it may be useful to focus on more detailed environmental features including availability, accessibility and use of specific destinations for physical activity such as services or natural areas [30,39,40].The relatively weak association between walkability assessed based on map data and perceived environmental facilitators for outdoor mobility may be related to the measurement dimension. Perceived environmental facilitators reflect the environment that an individual uses, the ability and preferences of the individual [12,13]. The question on perceived environmental facilitators used in this study asked specifically about the presence of environmental features that motivate individuals to go outdoors in their home environment [15]. Information on use of such environments was not directly available, but use is likely. Objectively assessed GIS-based variables, such as the walkability index, do not take into account individual preferences or abilities of the individual, and, in addition, require assumptions to be made regarding neighborhood boundaries. Previous research showed that definitions of neighborhood boundaries and the area where an individual actually moves (activity-spaces) vary in size and orientation between individuals [29,44]. Individuals may demonstrate preferences for a certain direction due to attractive routes and destinations or to avoid difficult terrain or heavy traffic. Consequently, the area used to define the neighborhood in the GIS-based walkability index may not be an accurate reflection of the environment used by the individual [28,29,44]. By incorporating behavioral patterns in area definitions reflecting actual use of the environment, e.g., by using GPS or interactive mapping tools [29,45], it may be possible to assess more accurately walkability with objective GIS-based measures.This is one of the first studies looking at longitudinal associations between objective and perceived environmental variables and self-reported physical activity. In the current study, walkability was not associated with maintenance of or increase in self-reported physical activity over the two-year follow-up. Previously, proximity of popular destinations was associated with maintenance of physical activity [40], and higher walkability was associated with smaller declines in walking for transport, but not leisure walking over time in older adults [31]. In addition, it has been observed that adults changed their physical activity behavior in response to relocation and subsequential changes in destinations [46]. In all, earlier studies point toward highly walkable environments having a positive, even though modest, influence on maintaining physical activity at a higher level. The lack of association in the current study may be partly explained by the high age of the participants in the current study, predisposing them to a decline in health and functioning and, consequently, also a decline in physical activity.The longitudinal findings of perceived environmental facilitators and physical activity were equivocal. Higher numbers of perceived environmental facilitators were associated with an increased likelihood to become at least moderately physically active over the follow-up, but at the same time, also with reduced likelihood of maintaining at least moderate self-reported physical activity for those with lower-extremity physical limitations. The finding that perceived environmental facilitators increased the likelihood of becoming at least moderately physically active over the follow-up may be real or an artifact due to temporary restriction of habitual physical activity at baseline, as awareness of environmental facilitators may indicate larger exposure to the environment [13]. Either way, this stresses the importance of older adults’ awareness and recognition of environmental facilitators for outdoor mobility in their home environment as a means to overcome restrictions in physical activities.The finding that perceived environmental facilitators reduced the likelihood of maintaining higher levels of physical activity was surprising. Previous research showed higher numbers of perceived environmental facilitators decreasing the risk for development of walking difficulty [14]. It is possible that those perceiving few environmental facilitators for outdoor mobility at baseline (the reference group), may not move outdoors as frequently, but they may be more adapted to the situation (e.g., by using compensatory strategies) [13,42]. Consequently, they may be better able to maintain their physical activity over time than those with lower-extremity physical limitations who reported more environmental facilitators. However, it is also possible that over time participants modified their perceptions of the environment due to declining physical function and subsequent changes in goals and valuation of meaningful activities [42,47]. Unfortunately, we could not take this into account in the analyses due to lacking data on perceived environmental facilitators at the follow-ups.Even in longitudinal studies, it is difficult to disentangle whether the environment affects physical activity or whether the environment was chosen partly because of favorable characteristics for walking [46,48]. In our study, the average duration that an individual had lived in the same home was well over 20 years. In the past decades the city of Jyväskylä has experienced a marked increase in population and infrastructure. It is likely that both the environment and the preferences of individuals may have changed and thus the effect of neighborhood selection reduced over time. However, when physical function starts to decline, older adults may accommodate by moving to an environment better suited to their needs [49]. These are likely areas with better access to services [49], which likely coincides with better walkability. The current data suggest that participants with lower-extremity physical limitations living in the least walkable areas had lived on average 5 years longer in the current home than those living in more walkable areas. As a result, for participants with physical limitations only, adjustment for the duration lived in the current home markedly strengthened the association between the walkability index and step counts. Strengths of this study include a population-based sample with large sample size and few missing values. This is one of few studies assessing environmental features both objectively and as perceived by the individual, and physical activity behavior of older adults. Lower-extremity physical limitations were assessed objectively. The walkability measures were categorized, in line with previous studies [8,44], which may have reduced variability, potentially leading to underestimation of associations. On the other hand, a continuous measure of the walkability index may not capture variation relevant for physical activity [44], especially considering non-linearity of certain associations. Furthermore, the walkability index has been widely used, but not standardized. There is no consensus about calculation strategies of its components [17]. GIS data were available for the calculation of a walkability index in Finland, although not fully in accordance with the originally proposed variables by Frank et al. [8]. Furthermore, both self-reported and accelerometer-based physical activity measures may be prone to bias [50]. Accelerometers may underestimate physical activity especially among those walking slowly (in the current study those with lower-extremity physical limitations) and those engaging in low-impact physical activities (e.g., cycling, swimming or skiing) [50]. Finally, the analyses using objectively assessed step counts were based on a relatively small sample size, warranting caution when interpreting the results.Considering the trend of aging in community homes and the preventive effect of physical activity on maintaining health and function into old age, knowledge of attractive environmental features motivating older adults with physical limitations in the lower extremities to go outdoors is essential. This study provides a comprehensive picture of associations between objectively assessed walkability, perceived environmental facilitators and physical activity in community-dwelling older adults in Finland. Findings suggest that walk-friendly environmental design may provide opportunities for physical activity in old age. Associations between perceived environmental facilitators and physical activity suggest that creating awareness of attractive environmental factors in the home environment may be needed to promote physical activity, especially when physical function starts to decline. However, also motivation may play a role in this respect. Intervention studies are needed to confirm reported associations.We thank the participants for their time and effort to participate in our study. This work was financially supported by grants of the Finnish Ministry of Education and Culture (to E.P. and M.R.); and the Academy of Finland [grant no 255403 (to T.R.), no 285747 (to M.R.)]. Gerontology Research Center is a joint effort between the University of Jyväskylä and the University of Tampere.Erja Portegijs, Taina Rantanen and Merja Rantakokko conceived and designed the study; Erja Portegijs, Li-Tang Tsai, Taina Rantanen and Merja Rantakokko contributed to the participant data collection; Erja Portegijs and Kirsi Keskinen contributed to the spatial data collection and analyses; Erja Portegijs conducted the statistical analyses; Erja Portegijs wrote the paper; Kirsi Keskinen, Li-Tang Tsai, Taina Rantanen and Merja Rantakokko critically revised the paper.The authors declare no conflict of interest. The funding sponsors had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, and in the decision to publish the results. Due to ethical and legal restriction, data are available on request from the LISPE research group. To ensure the protection of privacy and compliance with relevant Finnish laws, researchers interested in using the data must obtain approval for data usage. Additional, restrictions and conditions may apply. To request the data please contact Professor Taina Rantanen (taina.rantanen@jyu.fi).Participant and environmental characteristics in the full baseline sample (n = 839) and the physical activity subsample (n = 174).a Independent t-test, b Chi-square test.Regression coefficients for reporting higher numbers of perceived environmental facilitators according to tertiles of the walkability index (n = 848).B = Regression coefficient, SE = Standard error. Generalized linear models adjusted for a age and sex, b age, sex, number of chronic conditions, years lived in the home, years of education, and climatologic circumstances, and c additionally for lower-extremity physical limitations.Regression coefficients for higher step counts according to tertiles of perceived environmental facilitators or tertiles of the walkability index (n = 174).B = Regression coefficient, SE = Standard error. Generalized linear models adjusted for a age, sex, and accelerometer wear time, b age, sex, accelerometer wear time, number of chronic conditions, years lived in the home, years of education, and climatologic circumstances, and c additionally for lower-extremity physical limitations. ^ Model could not be computed as the Hessian matrix is singular. ~ The maximum number of step-halvings was reached but the log-likelihood value cannot be further improved.The odds for reporting at least moderate physical activity at baseline according to tertiles of perceived environmental facilitators or tertiles of the walkability index (n = 848).Bold values indicate p < 0.050, OR = Odds ratio, 95% CI = 95% confidence interval. Logistic regression models adjusted for a age and sex, b age, sex, number of chronic conditions, years lived in the home, years of education, and climatologic circumstances, and c additionally for lower-extremity physical limitations.The odds for reporting at least moderate physical activity (PA) by the latest follow-up according to tertiles of perceived environmental facilitators or tertiles of the walkability index stratified by physical activity level at baseline (BL).Bold values indicate p < 0.050, OR = Odds ratio, 95% CI = 95% confidence interval. Logistic regression models adjusted for a age and sex, b age, sex, number of chronic conditions, years lived in the home, years of education, and climatologic circumstances, and c additionally for lower-extremity physical limitations.