Feasibility and Preliminary Efficacy of GPT-QPL: An AI-Generated, Personalized Question Prompt List Intervention for Patients With Hematologic Cancers
Summary
The goal of this study is to evaluate the feasibility and preliminary efficacy of an artificial intelligence (AI)-generated personalized question prompt list (a list of suggested questions to ask during outpatient appointments) for patients with hematologic cancers. The intervention will involve tailoring a standardized prompt to patients' individual characteristics and concerns. This prompt will then be used to ask Washington University's (WashU) HIPAA compliant ChatGPT to generate personalized question lists for outpatient appointments. Analyses will assess the impact of personalized QPLs on patients' question-asking behavior; communicative self-efficacy; and self-reported amount and satisfaction with information obtained about their disease and its treatment. Sub-analyses will explore patterns in questions generated by WashU ChatGPT. Patients will also provide feedback pertaining to the perceived helpfulness and ease-of-use of WashU-ChatGPT-generated question lists, as well as their attitudes and intentions regarding use of AI chatbots and whether they would engage in pre-appointment AI-assisted question brainstorming independently in the future.
Arms & interventions
- OtherGPT-QPL
A research team member will generate a QPL that is personalized to the patient's demographics (from EHR screening and Baseline Demographic Survey) and concerns (from Distress Thermometer Problem Checklist and Interview).
Outcome measures
Primary
Changes in communicative self-efficacy
* Communicative self-efficacy is measured by Self-efficacy survey . * Dependent samples t-tests will be used to assess within-subject changes. * The Self-Efficacy survey consists of two scales containing 10 questions each, with answers ranging from 1=not successful to 5=very successful or 1=not confident to 5=very confident. The total score on each scale ranges from 10-50 with a higher score on Scale 1 (the PEPPI-5) representing higher self-efficacy related to communicating with physicians in general and a higher score on Scale 2 (researcher-designed questions based on Social Cognitive Theory) representing higher self-perceived success at communication with the physician during the immediately preceding appointment.
Time frame: Appointment 1 (pre-intervention) and appointment 2 (post-intervention) (estimated to be 12 weeks)
Secondary
Changes in perceived amount of cancer information received as measured by an abbreviated version the EORTC QLQ - INFO25
Time frame: Appointment 1 (pre-intervention) and appointment 2 (post-intervention) (estimated to be 12 weeks)
Changes in satisfaction with cancer information received as measured by the EORTC QLQ - INFO25
Time frame: Appointment 1 (pre-intervention) and appointment 2 (post-intervention) (estimated to be 12 weeks)
Changes in the number of questions asked during appointments
Time frame: Appointment 1 (pre-intervention) and appointment 2 (post-intervention) (estimated to be 12 weeks)
Changes in self-reported degree of intention to use AI chatbots for healthcare tasks
Time frame: Appointment 1 (pre-intervention) and appointment 2 (post-intervention) (estimated to be 12 weeks)
Eligibility criteria
Study locations (1)
Washington University School of Medicine
St Louis, Missouri, 63110