Expert Evaluation of Clinical AI Tools on Real Point-of-Care Clinical Queries

2026-06-27

AI Summary

This study addresses the gap in clinical AI evaluation, which often relies on hypothetical questions and lacks validation by domain experts in real-world settings. The authors introduce Real-POCQi, a dataset of 620 authentic frontline clinical queries, and conduct a large-scale blinded assessment involving 149 specialist physicians who evaluated responses from three general-purpose large language models and the specialized tool OpenEvidence across five dimensions—including accuracy and clinical utility. This work presents the first large-scale expert-aligned, query-grounded blind evaluation, demonstrating that the specialized tool significantly outperforms general models across all dimensions (win-rate advantages of 25–39 percentage points, p<0.001), with results robust to multiple sensitivity analyses. The study also reveals systematic discrepancies between automated LLM scoring and expert judgments. The Real-POCQi benchmark and analysis pipeline are publicly released to establish a new paradigm for clinical AI evaluation.

Abstract

Physicians now pose millions of clinical questions to AI tools each week, yet these tools are evaluated largely on hypothetical or exam-style questions, not those actually asked in practice. We report a blinded evaluation built on 620 Real-world Point-Of-Care Queries (Real-POCQi) submitted to the OpenEvidence (OE) platform by physicians spanning 30 specialties, as well as 187 questions from HealthBench. 149 practicing physicians across 36 states made head-to-head comparisons between answers from three frontier general-purpose models (Claude Opus 4.8, Gemini 3.1 Pro, and GPT-5.5) and a specialized clinical tool (OE), with graders matched to each question's specialty. When comparing answers along five dimensions relevant to clinical decision support -- accuracy, clinical utility, source quality, verifiability, & completeness -- physicians scored the specialized tool highest on all axes; in the primary analysis on Real-POCQi, win differences (margins between win and loss rates) ranged from 25 to 39 percentage points (p<0.001). Results remained consistent in sensitivity analyses stratifying by citation display, answer length, OE-user status, and Real-POCQi versus HealthBench. In parallel, LLM judges were found to systematically differ from expert judges, though both generally agreed on the best model. These findings underscore two conclusions: (i) AI tool evaluations should reflect real-world query distributions and use expert judges that mirror the specialization defining modern medicine and (ii) the consistent advantage of the specialized tool over general-purpose models does not necessarily mean that the latter cannot serve similar purposes, but that targeted engineering and customization can yield meaningful gains in performance for its users. We release Real-POCQi as a public benchmark, as well as the prespecified statistical analysis for reproducing results of this study.

Problem

• clinical AI evaluation
• real-world queries
• point-of-care
• expert assessment
• AI benchmarking

Innovation

• real-world clinical queries
• expert evaluation
• specialized clinical AI
• benchmark dataset
• blinded head-to-head comparison