TOP 72 - Parent Preferences for Summarized Inpatient Notes: A Qualitative Study

Publication Number: 4777.TOP 72

Jessica D. Bethel, University of Wisconsin School of Medicine and Public Health, Madison, WI, United States; Madeline Kieren, University of Wisconsin School of Medicine and Public Health, Madison, WI, United States; Gabriel Tse, Stanford University School of Medicine, Palo Alto, CA, United States; Casey A. Ohare, University of Wisconsin School of Medicine and Public Health, Madison, WI, United States; Michelle M. Kelly, University of Wisconsin School of Medicine and Public Health, Madison, WI, United States; Cris G. Ebby, UT Southwestern Medical Center, Dallas, TX, United States

Background: Federal regulations now require that patients and their caregivers have access to their medical records, including provider notes. While this access can improve patient understanding, and follow through on tests and referrals; it can also lead to confusion, especially among individuals with lower health literacy and less formal education. Plain language summaries could improve comprehension but creating them manually for each patient would be time and resource intensive. Artificial intelligence (AI), specifically large language models, offer a scalable way to generate these summaries; however, it is unclear how caregivers prefer notes to be summarized.

Objective: To explore caregiver preferences for how inpatient note content should be presented and elicit their attitudes towards the use of AI to generate plain language summaries.

Design/Methods: We will conduct 30 semi-structured interviews with English-speaking legal guardians of children aged 0-12 admitted to a pediatric hospitalist service at an urban, academic hospital. Prior to the interview, participants will complete a short demographic and health literacy survey. Prompted by sample inpatient notes, interviews will focus on experiences with reading inpatient notes, desired features in summaries, and perspectives regarding AI-generated content. All interviews will be recorded, transcribed verbatim, and analyzed using conventional content analysis, a method used to inductively describe and interpret textual data. Two researchers (JDB and MQK) will review a subset of transcripts to develop a codebook containing codes, definitions, and exemplar quotes. After refining the codebook, JDB and MQK will independently code each transcript and meet to resolve discrepancies and discuss patterns identified from the data. This study was deemed IRB exempt. Analysis of transcribed interviews will be completed in winter 2025.