Healthcare · AI

Triage prediction for a regional clinic network

Built and clinically validated a triage-priority model, integrated directly into existing patient-intake software.

22%
Reduction in average wait
94%
Model agreement with clinicians
11
Clinics live
<200ms
Prediction latency

Illustrative case study. Written to show the page structure and the level of detail that performs well in search. Replace with a real engagement from the CMS before launch.

The challenge

Intake staff triaged patients with a paper protocol written years earlier. Priority varied by who was on shift, and the busiest clinics saw the widest variation — with the longest waits for the patients who needed care soonest.

Our approach

We began with a retrospective study on anonymised historical records, validating that a model could match clinician judgement before any software was written. Clinicians reviewed disagreements case by case. The model surfaces a suggested priority inside the existing intake screen — it never decides, and staff override freely.

The outcome

Average wait time fell 22% across eleven clinics. Model and clinician agree on 94% of cases, and every disagreement is logged for the clinical governance review.