Clinical decision support · demonstration
Cohort Risk & Review Lab
Which patients should a care-management team call first? This app shows a responsible way to answer that: the model decides who is flagged, an AI agent explains why, and a human makes the call — with fairness measured for every group, not just on average.
How it works
Four steps, each with a clear owner. The AI assistant only ever does step 3.
- 1
A statistical model scores every patient
A calibrated machine-learning model — not a chatbot — estimates each patient's chance of a high illness burden over the next six months. It never sees race.
- 2
Fairness is checked per group
Accuracy is reported separately for every age, sex, race and insurance group, with uncertainty ranges — so a model that fails one group can't hide behind a good average.
- 3
An AI agent gathers the evidence
For each flagged patient, an AI assistant pulls encounters, labs and care gaps into a short case brief. It cannot change the score or recommend enrolment.
- 4
A clinician decides
A person reads the brief and chooses to enrol, decline or defer. Nothing happens to a patient without that decision.
Explore
Want the full story?
The documentation covers the architecture, every design decision, the evidence behind each claim, and how to run it yourself.