SHM Converge Scientific Abstract ePoster Gallery-Mobility Functional Status Identification Leveraging Large Language Models in Hospitalized Adults

Pdf Summary

This Mayo Clinic study explored whether a large language model (LLM) could identify mobility-related functional status from hospitalized adults’ clinical notes and whether these findings were associated with discharge disposition.<br /><br />The researchers reviewed 1,104 clinical notes from 152 hospitalized patients aged 65 and older, including both medical and surgical admissions. Notes from the week before admission and throughout hospitalization were analyzed. The LLM classified mobility function into five International Classification of Functioning, Disability, and Health–based categories: changing and maintaining body position, carrying/moving/handling objects, walking and moving, moving around using transportation, and mobility unspecified. Each category was labeled as impaired, possible impairment, unimpaired, or no information.<br /><br />Results showed that functional information was present in most notes, with only 5.6% lacking any functional content. The most commonly documented areas were changing and maintaining body position (83.8%) and walking and moving (74.8%), while carrying, moving, and handling objects was rarely mentioned (6.5%). Overall, the LLM identified significant levels of mobility impairment across notes.<br /><br />Importantly, the study found a statistically significant association between LLM-derived functional status and hospital discharge disposition (p < 0.0001). Patients discharged to skilled nursing facilities were much more likely to have multiple functional impairments: 76.5% had two or more impairments. In contrast, among patients discharged home, most (58.2%) had one or fewer impairments.<br /><br />The authors concluded that the functional status LLM can effectively extract mobility-related information from clinical notes and that its findings correlate strongly with discharge outcomes. They suggest that LLMs could help advance healthcare delivery by improving assessment of functional status.

Asset Subtitle

Sandeep R. Pagali

Meta Tag

Author List Heling Jia, Jennifer St.Sauver, Muskan Garg, Sandeep R. Pagali, Sunghwan Sohn, Xingyi Liu

Category Research

Concept Mobility

Concept Functional Status

Concept Large Language Model

Concept Clinical Note

Concept Mobility Functional Status Class

Distinguished Non-Finalist

Presenter Organization Division of Hospital Medicine, Department of Medicine, Mayo Clinic, Rochester, MN

Presenting Author Sandeep R. Pagali

Track Technology in Hospital Medicine

Keywords

large language model

clinical notes

mobility impairment

functional status

hospitalized older adults

discharge disposition

skilled nursing facility

walking and moving

ICF categories

healthcare delivery

Mobility

Functional Status

Large Language Model

Clinical Note

Mobility Functional Status Class