Gates every release must pass
Compliance & privacy
Evidence from the latest release v1.0. These checks are synthetic-data controls, not a legal opinion.
Gates
29/29
Realism
100
Quality
100
Watermark
SYNTHETIC-APIARY
integrity
ATL-01 · All five database foreign keyspass
ATL-02 · Clinical orders equal RX orderspass
ATL-03 · All paid fills mapped to distribution dispensespass
ATL-04 · No negative pharmacy inventorypass
ATL-05 · Paid claim / dispensation countpass
ATL-06 · Rx patient plans consistentpass
ATL-07 · Encounter and prescription patient consistentpass
ATL-08 · Nonnegative financial splitspass
ATL-09 · Adherence range validpass
ATL-10 · No real provider identifierspass
ATL-11 · 100% formulary snapshots matchedpass
ATL-12 · No paid fills for uncovered drug-plan-monthpass
ATL-13 · All medication orders have coverage decisionpass
ATL-14 · No duplicated RX IDspass
ATL-15 · No negative monthly stockpass
ATL-16 · All 5 shared master row countspass
ATL-17 · No paid fill without paid claimpass
ATL-18 · Synthetic flags setpass
WB-01 · Workbook columns present on every generated source rowpass
licence
WB-03 · MedDRA codes withheldVOC-05 uses the workbook placeholder, not a numeric MedDRA code.
realism
FID-js_channel · Fidelity plan js_channelJS divergence on channel versus learned rx.plan. Score 0.003 vs threshold 0.12.
FID-js_coverage · Fidelity formulary js_coverageJS divergence on coverage_status versus learned rx.formulary_at_order. Score 0 vs threshold 0.08.
FID-js_specialty · Fidelity hcp js_specialtyJS divergence on specialty versus learned rx.hcp. Score 0.002 vs threshold 0.12.
FID-js_state · Fidelity hcp js_stateJS divergence on state versus learned rx.hcp. Score 0.002 vs threshold 0.12.
FID-js_ta_code · Fidelity drug js_ta_codeJS divergence on ta_code versus learned rx.drug. Score 0.032 vs threshold 0.22.
FID-ks_birth_year · Fidelity patient ks_birth_yearHistogram distance on birth year. NSCLC and psoriasis cohorts intentionally shift age, so the threshold is wider than a pure replay. Score 0.044 vs threshold 0.35.
FID-sequence_violation · Fidelity journey sequence_violationShare of generated journey steps that are absent from the learned transition matrix. Score 0 vs threshold 0.05.
REAL-01 · Realism against the TA calibration bandsfemale_pct=52.1; age_median=50; biologic_pct=66.3; us_pct=62.5; lot1_pct=64.4; msl_anchored_pct=100
regulatory
WB-02 · Workbook synthetic flagpass
Vocabulary licence tracker
| VOC | Vocabulary | Status | Mappings | Licence |
|---|
| EXT-59 | NCBI Gene | open | 1 | NCBI Gene public |
| VOC-01 | MeSH | open | 2 | NLM MeSH |
| VOC-02 | UMLS | withheld | 1 | UTS. No redistribution. |
| VOC-03 | SNOMED CT | withheld | 1 | Affiliate licence. Id not shipped. |
| VOC-04 | ICD-10-CM | open | 1 | ICD-10-CM public domain. No specific METex14 code; this is the coarse neoplasm code used in the diagnoses table. |
| VOC-05 | MedDRA | withheld | 1 | MSSO subscription. PT name only. |
| VOC-06 | ATC / WHO Drug | withheld | 1 | ATC code withheld. |
| VOC-07 | RxNorm | not_applicable | 1 | RxNorm. No code invented. |
| VOC-09 | MONDO | open | 2 | CC BY |
| VOC-10 | SPOR / IDMP | withheld | 1 | EMA terms. Not generated. |