AI in Medical Societies (Concurrent)-Neurosymbolic AI for Automated Clinical Quality Measure Abstraction in Pathology

Neurosymbolic AI for Automated Clinical Quality Measure Abstraction in Pathology

Video Transcription

Video Summary

The College of American Pathologists, in collaboration with AI startup Pharoah, is advancing automated clinical quality reporting using neurosymbolic AI to address challenges in extracting clinically relevant data from complex pathology reports. Traditional methods—structured fields, manual abstraction, keyword or rule-based approaches—fall short due to high clinician burden, cost, or inaccuracies, especially in interpreting nuanced narrative data with context. Pharoah’s neurosymbolic AI breaks down quality measure abstraction into many smaller sub-tasks, combining AI-driven analysis with logical, auditable reasoning trees. This approach avoids hallucination, ensures precise adherence to measure definitions, and produces consistent, explainable results. Evaluated against human abstractors on 2,000 reports, the system showed no significant difference in accuracy. The technology promises reduced reporting burden, accelerated measure development, and the potential to identify clinical practice gaps and safety events. Ongoing efforts focus on expanding measures, improving validation, and deploying AI tools to better capture meaningful clinical quality data with minimal clinician disruption.

Keywords

automated clinical quality reporting

neurosymbolic AI

pathology report data extraction

Pharoah AI startup

clinical measure abstraction

AI-driven clinical data analysis