One engine,
every therapeutic area.
The method is therapeutic-area agnostic. We benchmarked it first in psychiatry, the hardest place, where outcomes live in unstructured notes. Select an area to see how the same pipeline ingests its data, derives its endpoints, and assembles a matched cohort. A maturity tag on every area keeps the picture honest: we do not imply live validated cohorts everywhere.
Try PANSS, ctDNA, or eGFR
Psychiatry and Mental Health
Psychiatric outcomes live in therapy notes and clinical narratives, not structured fields, and the endpoints are subjective. Deriving validated, traceable scales from that unstructured text is the hardest version of this problem, and it is what we built first. Solving it is what makes the engine credible everywhere else.
- Psychiatric EHRs
- Therapy and progress notes
- Clinician assessments
- Ambient / scribe transcripts
- Clinical and behavioural measures
- Digital phenotyping
- Actigraphy
- Speech and language features
- PHQ-9
- HAM-D (HDRS)
- MADRS
- GAD-7
- HAM-A
- PANSS
- BPRS
- CGI
- YMRS
- Response and remission rates
Derive each scale item from unstructured narratives with item-level traceability to the source sentence, then validate against assessed ground truth with clinical partners, and assemble longitudinal, protocol-aligned cohorts.
Fewer external-control approvals than oncology because subjective endpoints are harder, which is precisely the gap Sapien addresses. Aligns to FDA external-control and AI guidance.