Skip to main content
Flock generates a cohort, not a pile of independent rows. Demographics come from US Census distributions, condition correlations from a 108-condition comorbidity matrix grounded in public epidemiology (CDC WONDER, NHANES, CDC surveillance), and temporal order is enforced across every clinical event. The result reads like a population a real facility would see, which is what makes it useful for testing.

​
Demographics

Age, sex, race, and geography are drawn from US Census distributions for the selected region. Set the geographic focus with a state abbreviation:
Age brackets, sex ratios, and the race and ethnicity mix track the census figures for that region rather than a flat random spread.

​
Correlated conditions

Conditions co-occur the way they do in real patients. The comorbidity matrix carries conditional probabilities between condition pairs, so a generated patient with type 2 diabetes is likely to also carry hypertension, obesity brings sleep apnea and fatty liver disease along with it, and a COPD patient trends toward nicotine dependence and, less often, lung cancer. This is the credibility layer, not the headline: the point of Flock is schema-aware, FK-safe, no-PHI data, and the epidemiological grounding is what keeps that data from looking synthetic.

​
Temporal coherence

Every generated timeline holds together. Admission timestamps precede discharge timestamps, lab orders precede their results, medication start dates precede end dates, and a patient’s age stays consistent with their date of birth. A downstream system that validates ordering sees data it can actually process.

​
Family linkage

Households are modeled, not scattered. Spouses share addresses, parent and child records carry age-appropriate gaps, and emergency-contact fields reference other generated people rather than dangling.

​
Reproducible runs

Pass a seed to get the identical population every time, which makes a generated dataset something you can commit to a test fixture:

​
FHIR output

Generates FHIR transaction Bundles across the 137 spec-validated resource types Post generates, with references resolved inside each Bundle.