Using Chatbot to Better Understand What Matters Most to Urogynecologic Patients

Abstract

IMPORTANCE: Urogynecology patients increasingly investigate their problems using technology-based resources, but there is a lack of evidence surrounding the way they use these resources or how patients may describe their problems. OBJECTIVE: The aim of this study was to examine how patients describe their pelvic floor symptoms to an artificial intelligence chatbot and the specific information they seek during their initial consultation. STUDY DESIGN: This was a mixed-methods secondary analysis of an Institutional Review Board-approved, single-center, randomized, nonblinded trial examining patient use of a large language model, Chat Generative Pre-trained Transformer (ChatGPT 4.0; OpenAI), at their initial urogynecologic visit. Patients who were randomized to an arm using ChatGPT were provided with a tablet and instructed to ask the program anything about their primary pelvic floor symptoms. A post hoc qualitative analysis of deidentified transcripts was performed by 2 independent reviewers with line-by-line coding and organized into themes using a predefined strategy. RESULTS: Seventy-nine conversation transcripts (41 previsit and 38 postvisit) were collected from 72 English-speaking and 9 Spanish-speaking patients. Five thematic domains were identified based on participant transcripts: (1) Language, (2) Disease-Specific, (3) Patient Experience, (4) Treatment, and (5) Chatbot Interactions. CONCLUSIONS: Medical consultations are often structured, and patients may not acknowledge when they do not fully understand the information provided. This study provides a window into understanding the patient experience that was not previously available. Pelvic floor specialists may consider the identified themes as relevant when providing patient-centered education and during expectation setting.

Ähnliche Arbeiten