Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Towards conversational assistants for health applications: using ChatGPT to generate conversations about heart failure

Anuja Tayal, Devika Salunke, Barbara Di Eugenio, Paula G Allen-Meares, Eulalia P Abril, Olga Garcia-Bedoya, Carolyn A Dickens, Andrew D. Boyd

TL;DR

The paper investigates using ChatGPT (3.5-turbo and GPT-4) to generate patient–educator conversations about heart failure self-care for African-American patients, addressing a data gap in culturally relevant dialogue. It compares four prompting strategies—Domain, AAVE, SDOH, and SDOH-informed reasoning—across foods, exercise, and fluids with varying turn lengths, plus SDOH attributes. Findings show that thoughtful prompt design, including reasoning steps, can improve dialogue quality and personalization, but ChatGPT still struggles with genuine empathy and fully natural two-way engagement. The work highlights the potential and current limitations of LLM-based healthcare dialogues and points to a direction for a future task-oriented, culturally sensitive dialogue system that more effectively handles empathy and practical barriers. Overall, the study advances understanding of how SDOH-informed prompting and intermediate reasoning can shape synthetic health conversations, informing subsequent development of patient-centered digital health assistants.

Abstract

We explore the potential of ChatGPT (3.5-turbo and 4) to generate conversations focused on self-care strategies for African-American heart failure patients -- a domain with limited specialized datasets. To simulate patient-health educator dialogues, we employed four prompting strategies: domain, African American Vernacular English (AAVE), Social Determinants of Health (SDOH), and SDOH-informed reasoning. Conversations were generated across key self-care domains of food, exercise, and fluid intake, with varying turn lengths (5, 10, 15) and incorporated patient-specific SDOH attributes such as age, gender, neighborhood, and socioeconomic status. Our findings show that effective prompt design is essential. While incorporating SDOH and reasoning improves dialogue quality, ChatGPT still lacks the empathy and engagement needed for meaningful healthcare communication.

Towards conversational assistants for health applications: using ChatGPT to generate conversations about heart failure

TL;DR

The paper investigates using ChatGPT (3.5-turbo and GPT-4) to generate patient–educator conversations about heart failure self-care for African-American patients, addressing a data gap in culturally relevant dialogue. It compares four prompting strategies—Domain, AAVE, SDOH, and SDOH-informed reasoning—across foods, exercise, and fluids with varying turn lengths, plus SDOH attributes. Findings show that thoughtful prompt design, including reasoning steps, can improve dialogue quality and personalization, but ChatGPT still struggles with genuine empathy and fully natural two-way engagement. The work highlights the potential and current limitations of LLM-based healthcare dialogues and points to a direction for a future task-oriented, culturally sensitive dialogue system that more effectively handles empathy and practical barriers. Overall, the study advances understanding of how SDOH-informed prompting and intermediate reasoning can shape synthetic health conversations, informing subsequent development of patient-centered digital health assistants.

Abstract

We explore the potential of ChatGPT (3.5-turbo and 4) to generate conversations focused on self-care strategies for African-American heart failure patients -- a domain with limited specialized datasets. To simulate patient-health educator dialogues, we employed four prompting strategies: domain, African American Vernacular English (AAVE), Social Determinants of Health (SDOH), and SDOH-informed reasoning. Conversations were generated across key self-care domains of food, exercise, and fluid intake, with varying turn lengths (5, 10, 15) and incorporated patient-specific SDOH attributes such as age, gender, neighborhood, and socioeconomic status. Our findings show that effective prompt design is essential. While incorporating SDOH and reasoning improves dialogue quality, ChatGPT still lacks the empathy and engagement needed for meaningful healthcare communication.
Paper Structure (23 sections, 1 figure, 10 tables)

This paper contains 23 sections, 1 figure, 10 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Health Education.
  4. Prompting
  5. Methodology
  6. Can ChatGPT generate relevant conversations for Heart Failure Self-Care?
  7. Can ChatGPT personalize conversations based on the SDOH features of the patients?
  8. Can ChatGPT express empathy with the patients?
  9. Is having ChatGPT to generate reasoning before conversations more effective than directly generating the conversations?
  10. Approaches to Generate Simulated Conversations
  11. Approach 1- Generating Conversations According to Domain
  12. Approach 2 - African American Vernacular English
  13. Approach 3 - Integrating SDOH Features
  14. Approach 4 - Generating Reasoning as an Intermediate Step
  15. Experiment and Evaluation
  16. ...and 8 more sections

Figures (1)

  • Figure 1: Comparison of Appropriateness of SDOH Features between Approaches 3 and 4.