Management of Patients With Comorbid Asthma and Obesity: A Large Language Model Evaluation of Clinical Documentation.

Abstract

Obesity is a well-known asthma risk factor, and weight loss benefits asthma outcomes. Asthma guidelines recommend weight reduction as part of care; however, provider practices are not known. To evaluate whether weight management is integrated into routine asthma care for patients with comorbid obesity. We analyzed outpatient encounter notes from patients with both asthma and obesity seen by primary care, allergy/immunology, or pulmonary providers at a large health system between January 1, 2020, and September 30, 2023. Notes were extracted from electronic health records for visits with a primary asthma diagnosis. Using Generative Pretrained Transformer-4 Omni (GPT-4o), a large language model (LLM), we assessed documentation of (1) asthma management, (2) obesity management, (3) integration of obesity management into asthma care, and (4) specific weight management strategies. Inclusion rates were compared across specialties, and encounter-level predictors were identified. Chart review evaluated the performance of GPT-4o. Of 17,658 encounters (N = 8992 patients), only 12.6% included obesity management as part of asthma care, more frequently in subspecialty (11.0%) than in primary care (1.6%) settings. In adjusted models, male sex, middle age, higher body mass index, higher education, and pulmonology care increased odds of an encounter with obesity management linked to asthma care; oral steroid use decreased the odds. Obesity management strategies differed by specialty, though exercise and general weight counseling was common. GPT-4o demonstrated robust performance. LLM evaluation of >17,000 encounters demonstrate that, in contrast to guidelines, weight management is infrequently addressed in asthma care. These results highlight actionable opportunities to improve asthma outcomes.