Clinical data is scarcer still. Same forge, a new kind of record.
Real diagnostic workup data almost never leaves the hospital that generated it \u2014 not for students, not for methods research. Health is Dataset Foundry's third line family, deliberately clinical rather than insurance/finance-shaped: no policy or loan window, no name field, and each line's own severity structure (single-stage classification vs. two-stage classification-plus-staging) is a per-generation choice instead of a fixed shape. Ghana is today's default, not a restriction \u2014 the generator takes a country parameter, chosen per dataset rather than fixed by the line.
seed 42 — this exact seal, every time
Two clinical lines today, more staged behind them
Breast Cancer and Chronic Kidney Disease are both live. Structurally realistic diagnostic-workup data, deterministic and seeded, mimicking the shape of real scarce clinical data rather than generating noise.
Breast Cancer
Available nowA single diagnostic workup's malignancy probability from clinical/imaging features, optionally extended with a stage-at-diagnosis (I–IV) severity measure for malignant findings — the first line where the stage count itself (single vs. two-stage) is a runtime choice rather than a fixed shape. Deterministic, seeded, ready to forge.
Generate a datasetChronic Kidney Disease
Available nowA single renal risk assessment's CKD-positive probability from cardiometabolic risk factors, optionally extended with KDIGO GFR/albuminuria staging and a primary-cause call for CKD-positive records — same runtime-choice stage-count shape Breast Cancer established. Deterministic, seeded, ready to forge.
Generate a datasetRelationships tested and confirmed in every batch
These aren't hidden in the noise — they're the reason the line exists. Two hypotheses drive malignancy probability and hold no matter which stage mode you generate in. Two more, verified only in two-stage mode and scoped to malignant findings, drive stage severity.
Malignancy probability — both stage modes
Patient age
Higher malignancy probability
The single most robust, most consistently replicated breast-cancer risk factor in the epidemiological literature — carries the base-rate anchor itself rather than a loading on top of one.
Family history (first-degree relative)
Higher malignancy probability
A first-degree relative's history roughly doubles risk in pooled hormonal-factors literature — the largest single loading in the malignancy table after mass margin.
Stage severity — two-stage mode only, malignant findings
Tumor size at diagnosis
Higher stage severity
Directly underlies the T-component of AJCC TNM staging — a larger tumor is mechanically closer to a higher T-stage by the staging system's own definition, not just correlated with one.
Detection pathway (symptomatic vs. screening-detected)
Higher stage severity (symptomatic)
Screening exists specifically to catch disease before symptoms appear — this line's heavy symptomatic skew is what actually drives its stage distribution toward III/IV, reflecting Ghana's low screening uptake.
Relationships tested and confirmed in every batch
Same structure as Breast Cancer's own hypotheses above. Two hypotheses drive CKD-diagnosis probability and hold no matter which stage mode you generate in. Two more, verified only in two-stage mode and scoped to CKD-positive records, drive GFR-stage severity.
CKD-diagnosis probability — both stage modes
Hypertension status
Higher CKD-diagnosis probability
Documented as the single most commonly attributed cause of CKD in Ghana — roughly 30% of pooled cases — and often poorly controlled regionally; the larger of this line's two core diagnosis loadings.
Diabetes status
Higher CKD-diagnosis probability
A real but secondary driver in Ghana's documented etiology mix (under a fifth of pooled cases, versus hypertension's roughly 30%) — unlike many Western case series where diabetic nephropathy dominates. Still a core hypothesis, just a smaller loading than hypertension's.
GFR-stage severity — two-stage mode only, CKD-positive records
Patient age
Higher GFR-stage severity
eGFR mechanically declines with age through nephron loss, independent of any specific disease process — the more mechanistic of this line's two core stage-severity hypotheses.
Screening pathway (routine vs. incidental/symptomatic)
Higher GFR-stage severity (incidental/symptomatic)
CKD screening isn't routinely performed in many diabetes/hypertension clinics across sub-Saharan Africa — this line's heavy incidental-detection skew is what actually drives its GFR-stage distribution toward G3a/G3b.
Configurable length, configurable shape
Pick a record count, a seed, and a stage mode (single-stage classification, or two-stage with severity fields \u2014 two-stage is the default), and the deterministic model runs server-side and hands you a CSV, manifest, and citation block. Same seed-and-cite reproducibility as every insurance and finance line.
Both lines are live — more staged behind them
Breast Cancer and Chronic Kidney Disease are both shipped now, each with its own hypothesis tables and methodology page. Nothing else is scheduled in this family while the freeze holds.
Other countries' distributions are on the roadmap too, opt-in per generation — Ghana stays the default, not a ceiling.