AI Chart Review Boosts SEP-1 Performance

Artificial intelligence-enabled medical record abstraction and near-real-time feedback improved performance on a sepsis quality measure in emergency departments, according to results from a cluster randomized trial published in JAMA Network Open.

The trial evaluated whether large language model (LLM)-driven assessment of sepsis care could improve compliance with the Centers for Medicare & Medicaid Services (CMS) Severe Sepsis and Septic Shock Management Bundle (SEP-1). The authors reported that hospital quality reporting is often manual, costly, and delayed, limiting its usefulness for improving care processes.

The single-blind, unstratified trial was conducted from December 13, 2024, to July 8, 2025, at 2 academic emergency departments within the University of California, San Diego health system. A total of 66 attending physicians were randomized to receive either targeted feedback based on LLM-determined SEP-1 compliance at the time of patient discharge or the standard feedback process. The analysis included 301 patients who met CMS inclusion criteria for SEP-1, including 121 in the control group and 180 in the intervention group.

Physicians in the intervention group had an SEP-1 compliance rate of 82.9%, compared with 70.1% in the control group. Assignment to the intervention group was associated with a 13.0% absolute improvement in SEP-1 compliance (95% CI, 2.5%-23.4%; odds ratio, 2.10; 95% CI, 1.15-3.81; P=0.02). The largest difference between groups was in noncompletion of the 30mL/kg crystalloid fluid bolus component, which occurred in 1.7% of patients in the intervention group and 13.2% of patients in the control group. Agreement between the LLM determination and expert review was 92%.

The authors reported no significant differences in intensive care unit admissions or 30-day mortality. They concluded that AI-enabled medical record abstraction and timely physician feedback may help address limitations in current quality measurement, while noting that whether this approach improves clinical management or patient-centered outcomes remains uncertain.

Source: Boussina A, Allison C, Quintero K, et al. Medical record abstraction for quality improvement in sepsis care using artificial intelligence: a cluster randomized trial. JAMA Netw Open. 2026;9(6):e2611885. doi:10.1001/jamanetworkopen.2026.11885