AI Falls Short on Differential Dx - Summary - MDSpire
From the Journals

AI Falls Short on Differential Dx

New PrIME-LLM benchmark shows strong diagnostic accuracy but persistent gaps in clinical reasoning across 21 large language models

Share

A cross-sectional study published in JAMA Network Open evaluated the accuracy of 21 large language models (LLMs) in medical diagnoses through the Proportional Index of Medical Evaluation for LLMs (PrIME-LLM). The study found that while LLMs like GPT-5 and Grok 4 achieved high accuracy in final diagnoses, they struggled significantly with differential diagnoses, revealing a failure rate above 80%. Researchers emphasized the limitations of current LLMs in processing clinical uncertainty and called for better evaluation methods that reflect real-world decision-making processes.

Original Source(s)

AI Falls Short on Differential Dx

  • by Kathryn Wighton

  • April 13, 2026

Related Content

FDA Flags Missing Trial Results

Federal outreach seeks to improve compliance with ClinicalTrials.gov reporting requirements and make more complete clinical trial information publicly available

Can Nasal Nitric Oxide Aid PCD Diagnosis?

Evidence synthesis shows strong diagnostic performance but highlights variability across subgroups and clinical settings

Utah Just Let an Algorithm Write Prescriptions

Physician oversight waived in state sandbox; federal review status unclear