← All updates
Observational study

Cardiology ·

LLM Analysis Finds Frequent Nonescalation of Lipid Therapy in ASCVD

A retrospective cohort study at an integrated Northern California health system used a large language model to analyze clinical notes from patients with ASCVD and LDL-C levels of at least 70 mg/dL. Among 13,354 relevant notes, lipid-lowering therapy was not escalated in 88.7%; documented reasons included adverse effects and patient preferences.

Researchers conducted a retrospective cohort study of adults aged 18 to 90 years with at least two ASCVD-related encounters and an LDL-C level of 70 mg/dL or higher at their most recent encounter. They analyzed notes from January 2019 through May 2025 at a large integrated health system in Northern California. The LLM’s performance was compared with clinical annotations, with micro-weighted F1 scores ranging from 0.80 to 0.98 across evaluated outcomes.

Of 28,113 patients identified, 13,810 had above-goal LDL-C. In the inference set, a treat-to-target strategy was documented in 3,756 of 13,354 relevant notes (28.1%), most often with a target below 70 mg/dL. Therapy was not escalated in 11,842 notes (88.7%). At least one reason was documented in 40.4% of those notes; the most common were patient-reported adverse effects (12.6%), LDL-C close to goal (6.9%), and patient preference to avoid therapy (6.8%).

The findings describe documented care and potential targets for system-level interventions, rather than testing an intervention or establishing why therapy was not escalated in every case. The analysis was retrospective and relied on clinical-note documentation; the authors propose that LLMs may help identify care gaps in secondary prevention.

AI summary · Not yet editor-reviewed

Is this summary clinically accurate?

Help fellow clinicians: your rating sends inaccurate summaries straight to our editors.

Sign in to rate this summary →

Source

JAMA Cardiology: AI-Identified Reasons for Nonescalation of Lipid-Lowering Therapy for Secondary Prevention ↗

This is an automated AI-condensed summary that has not yet been reviewed by an editor. Always consult the full item at the original source.