Sites
15
15 fictional institutions
Synthetic Data Lab
Generate useful scenarios with measurable data quality.
Fixed seed, explicit assumptions, labeled anomalies, and separate entropy definitions make the scenario repeatable and testable.
Patients
60
Pseudonymous only
Events
220
Cross-system timeline
Evidence
53
Visible provenance
Quality controls
Current deterministic seed: CTI-PANC301-v1
Referential integrity100%
All site, patient, event, evidence, and graph links resolve
Temporal consistency100%
No event precedes enrollment or follows the scenario window
Categorical coverage91%
Normalized Shannon entropy across friction and provenance
Target distribution96%
Risk mix matches 9 high / 18 medium / 33 low
Scenario coverage94%
All required rank-flip and retention narratives represented
Correlation fidelity89%
Injected operational relationships remain within target tolerance
Entropy scope
Two intentionally different uncertainty measures
Dataset distribution entropy
0.91Normalized Shannon entropy across a declared category set
Extraction entropy
0.18 avgAmbiguity among candidate semantic interpretations
Injected anomaly labels
Duplicate transport event
AN-204 high
Referral distribution shift
AN-219 medium
Source-date conflict
AN-227 medium