Synthetic patient-level dataset spanning 40 rare pediatric disorders across 10 clinical classes — lysosomal storage, organic acidemias, urea-cycle defects, mitochondrial, neuromuscular, leukodystrophy, skeletal dysplasia, immune deficiency, dermatologic, and syndromic disorders. The dataset models the diagnostic odyssey, multi-system phenotype, genetic testing cascade, orphan-drug access and health economics, clinical outcomes, caregiver burden, and social determinants of health.

This is a 500-patient sample of the full HC-GEN-005 product. It is synthetic — generated by a calibrated simulation engine. It contains no real patient data.

Not for clinical use. This dataset is for ML development, benchmarking, schema prototyping, and education only. It must not be used to inform real patient care, diagnosis, or treatment decisions.

Calibration anchors

Sample-level observed values (seed 42, n=500):

Metric	Observed	Target	Anchor
Diagnostic delay (median, non-NBS)	3.53 yr	2.8–4.6	NORD Diagnostic Delay Survey (4.8y)
Specialists seen (median)	7.0	6.0–9.0	Shire Rare Disease Impact Report (~7.3)
Pathogenic variant confirmed	0.78	0.72–0.84	IRDiRC molecular-confirmation yield
Orphan cohort on DMT	0.785	0.68–0.88	orphan-drug-eligible treatment uptake
Overall by-18 mortality	0.316	0.24–0.36	severity-mixed rare-disease cohort
Newborn-screening capture	0.302	0.22–0.36	NBS-eligible capture fraction
PedsQL total (mean)	52.5	48–58	severity-mixed pediatric QoL
Caregiver Zarit burden (mean)	46.9	42–52	rare pediatric caregiver burden
NBS delay reduction	4.84 yr	≥3.0 (floor)	NBS collapses diagnostic delay
Severity → cognitive deficit	34.8 pts	≥15 (floor)	severity-graded impairment
MPS urine-GAG separation	0.983	≥0.80 (floor)	GAG elevation specific to MPS
OA/UCD amino-acid separation	0.814	≥0.55 (floor)	metabolic class abnormality
Respiratory FVC deficit	34.2 pp	≥20 (floor)	FVC reduction with respiratory involvement
Supportive-only DMT-cost violations	0	=0 (floor)	cost-integrity check

Validation: Grade A+ (10.00/10) across all six canonical seeds (42, 7, 123, 2024, 99, 1), deterministic. The engine additionally passes its own internal 4/4 Grade A+ benchmark suite on every seed.

Schema highlights by module

149 columns, single table. Module groups:

Demographics & access — sex (inheritance-aware X-linked skew), race/ethnicity, region, residence, SES, language, insurance.
Disease taxonomy — primary_disease_category (40 disorders), disease_class (10 classes), Orpha & OMIM codes, inheritance pattern, disease_severity_index, untreated by-18 mortality prior.
Diagnostic odyssey — symptom onset, age at diagnosis, diagnostic delay, prior misdiagnoses, specialists/PCP/ER visits, newborn-screening eligibility & capture, genetic-test modality/cost/denial, pathogenic confirmation, diagnosis source, odyssey cost, parent advocacy.
Clinical presentation — developmental milestones, motor/language/cognitive quotients, autism/ADHD, seizures, cardiac (LVH severity), respiratory (FVC, NIV, trach), GI/G-tube, hepatic/renal/ophtho/hearing/skeletal involvement, growth z-scores.
Metabolic & biomarkers — amino acids, organic acids, acylcarnitine, urine GAG, chitotriosidase, lyso-Gb1/Gb3, GAA/GLA activity, ammonia, lactate/pyruvate, CK, LDH, AST, ferritin, sIL-2R.
Treatment & health economics — therapy class (ERT/SRT/chaperone/gene-therapy/ASO/small-molecule/HSCT), specific products, treatment start/delay/cost, payer coverage, appeals, PAP enrollment, copay, ICER value, QALY, adherence, infusion/ADA reactions, efficacy, HSCT details.
Outcomes & function at 18 — progression velocity, hospitalizations, metabolic crises, ICU, respiratory failure, ambulation/communication/G-tube/vent status, schooling, mortality (flag/age/cause), hospice.
Caregiver & QoL — Zarit, PHQ-9, GAD-7, employment disruption, lost income, sleep, OOP cost, PedsQL subscales, sibling impact, marital strain, advocacy/registry/trial engagement.
Social determinants — distance to specialty center, travel burden, telehealth, language barrier, immigration concern, food/housing insecurity, transport barrier, missed appointments, SDOH composite.

Files

hc_gen_005_sample.csv — 500 synthetic patients, 149 columns.
README.md — this file.

Loading

import pandas as pd
df = pd.read_csv("hc_gen_005_sample.csv")
print(df.shape)  # (500, 149)

# Diagnostic delay by newborn-screening capture
print(df.groupby("newborn_screening_caught")["diagnostic_delay_years"].median())

from datasets import load_dataset
ds = load_dataset("xpertsystems/hc-gen-005-sample")

Use cases

Diagnostic-odyssey and time-to-diagnosis modeling across heterogeneous rare disorders.
Orphan-drug access, payer-coverage, and health-economic (ICER/QALY) analysis.
Newborn-screening impact simulation (delay collapse, mortality reduction).
Multi-class rare-disease classification and severity stratification.
Caregiver-burden and social-determinant modeling.
Synthetic-data methodology benchmarking and schema/ETL prototyping (FHIR R4 export available in the engine).

Limitations (honestly disclosed)

Synthetic, cross-sectional snapshot. Each row is one patient with onset/diagnosis ages, status-at-18 projections, and annualized rates; there is no visit-level longitudinal table.
Diagnostic-delay median runs slightly below the NORD 4.8y anchor. The engine's observed non-NBS median is ~3.5y, within its own ±1.5y tolerance of the NORD figure; the scorecard ranges the metric on the observed value and notes the anchor. Real-world delay varies widely by disorder.
Relative, not absolute, disease prevalence. Category weights are relative within the represented rare-pediatric cohort, not population incidence; the 40-disorder mix is illustrative.
Marginal calibration, not full joint fidelity. Univariate prevalences and the engineered structural separations (NBS delay collapse, severity-cognitive gradient, MPS-GAG and OA/UCD metabolic specificity, respiratory FVC deficit) are anchored; higher-order correlations beyond those engineered are not independently validated.
Therapy/economic parameters are illustrative. Treatment products, costs, ICER/QALY values, and access pathways represent plausible ranges rather than payer-specific real-world data.
No engine defects identified. The HC-GEN-005 engine passes its own internal 4/4 Grade A+ benchmark suite across all six canonical seeds with a stable 149-column schema; all sampling uses the seeded generator (no global-RNG issue); the Weibull sampler uses the accepted bare lam*rng.weibull(k,n) form. No defects were observed or suppressed.
Small-sample variance. At n=500 some rates carry sampling variance; scorecard ranges accommodate this without masking real misses, and structural floors are weighted to dominate.

Commercial / full version

	Sample (this)	Full (commercial)
Patients	500	20,000+ (configurable)
Columns	149	149
Disorders	40 categories / 10 classes	40 categories / 10 classes
Formats	CSV	CSV / JSON / Parquet / FHIR R4 bundle
Seeds / reproducibility	6 canonical	Unlimited
License	CC-BY-NC-4.0	Commercial
Support	—	SLA, custom calibration, vertical extensions

Contact pradeep@xpertsystems.ai · https://xpertsystems.ai

Citation

@dataset{xpertsystems_hcgen005_2026,
  title  = {HC-GEN-005: Synthetic Rare Pediatric Disorders Dataset ---
            Diagnostic Odyssey, Multi-System Phenotype, Orphan-Drug Access
            & Outcomes (Sample)},
  author = {XpertSystems.ai},
  year   = {2026},
  publisher = {Hugging Face},
  note   = {Synthetic data. Not for clinical use. Calibration anchors:
            NORD Diagnostic Delay Survey; IRDiRC diagnostic-odyssey data;
            Shire Rare Disease Impact Report; EURORDIS Rare Barometer;
            Orphanet prevalence; disease-specific trials (ENDEAR, STR1VE,
            CHERISH, ENGAGE) and ICER value assessments.},
  url    = {https://huggingface.co/datasets/xpertsystems/hc-gen-005-sample}
}

Downloads last month: 16

Total file size:

353 kB