An AI framework developed by Nashville, Tenn.-based Vanderbilt University identified statin intolerance, contraindications and patient deferrals documented in clinical notes for 10% of adult patients.
Researchers processed 197,761 clinical notes from 47,192 adult patients using a three-part framework that included a rule-based natural language processing filter with 100% sensitivity, a large language model refinement filter with 97.3% specificity and an LLM-based multicategory classifier, according to a Feb. 9 university news release.
The tool flagged statin intolerance in 6.4% of patients, contraindications in 0.7% and deferrals in 2.9%, according to the study published Jan. 1 in the International Journal of Medical Informatics. The AI tool was highly accurate in identifying statin intolerance and performed well in spotting contraindications and cases in which patients declined treatment.
