Study Finds Clinical Stigma Can Skew AI Emergency Triage

A study has found that stigmatizing language in clinical documentation can influence how artificial intelligence systems prioritize patients in emergency triage. When chart notes contain biased or judgmental phrasing, the AI may assign lower urgency to patients who later experience clinical deterioration.

The findings point to a gap in current bias testing for healthcare AI. Many evaluation frameworks focus on demographic variables such as race, gender, or age. The study indicates that the tone and word choice within clinical notes also require scrutiny, as documentation style alone can skew algorithmic decisions.

Healthcare leaders responsible for AI governance are advised to expand bias audits to include analysis of documentation language. Without this broader testing, triage systems risk embedding clinical stigma into automated workflows, potentially affecting patient outcomes.