OpenEvidence errs on the safe side in a structured test of triage recommendations.
TL;DR
Under this benchmark, OpenEvidence produced fewer missed emergencies than the historical ChatGPT Health comparison, while errors concentrated in over-triage and evidence-seeking refusal.
Abstract
Background
Large language model (LLM) chatbots are increasingly consulted for triage decisions. A structured benchmark reported that ChatGPT Health, a consumer-facing health assistant, under-triaged 51.6 % of true emergencies and was susceptible to social anchoring. Whether a physician-facing clinical decision support platform fails similarly is unknown.
Objective
To characterize the triage safety profile of OpenEvidence, a retrieval-augmented, physician-facing platform, using the identical benchmark previously applied to ChatGPT Health.
Methods
We evaluated 60 clinician-authored vignettes from 30 clinical scenarios across 21 domains. Each scenario was written with and without objective clinical data and crossed with demographic and contextual modifiers in a 2 × 2 × 2 × 2 factorial design, yielding 960 prompts (480 clear-case, 480 edge-case). Responses were classified against a clinician gold standard as correct triage, under-triage, over-triage, or evidence-seeking refusal. Analyses used cluster bootstrap resampling, mixed-effects logistic regression, and Holm-Bonferroni correction.
Results
Among 449 clear-case responses that returned a recommendation, accuracy was 71.3 %. OpenEvidence under-triaged 12.5 % of emergency presentations versus 51.6 % in the previously reported ChatGPT Health benchmark, and over-triaged 68.0 % of nonurgent Home presentations (ChatGPT Health, 64.8 %). Anchoring statements did not alter recommendations (OR = 1.08, 95 % CI 0.62-1.88; Holm-adjusted p = 1.0). Objective clinical data eliminated emergency under-triage (25 % to 0 %; p = 0.005) and reduced nonurgent over-triage (78.7 % to 57.8 %; p = 0.014). In 65 of 960 responses (6.8 %), the platform declined to assign a triage level, exclusively in symptom-only Home or Routine prompts.
Conclusions
Under this benchmark, OpenEvidence produced fewer missed emergencies than the historical ChatGPT Health comparison, while errors concentrated in over-triage and evidence-seeking refusal. These findings support evaluating health AI within its deployment context and treating refusal as a distinct output category whose clinical implications require separate assessment.