Evaluating healthcare inequalities using ChatGPT4 – The case of potentially inappropriate medication

Abstract Issue/problem Regular and standardized monitoring of healthcare is necessary to inform decision makers and clinicians about healthcare quality. Equity is a fundamental dimension in such evaluations. We aimed to investigate if artificial intelligence (AI) large language models (LLM), such as ChatGPT4, may facilitate routine monitoring of healthcare inequalities using an established framework for healthcare quality and equity evaluation, analysis of individual heterogeneity and discriminatory accuracy (AIHDA). Description of the problem We have previously demonstrated that compared to traditional methods, AIHDA improves the evaluation of healthcare inequity. We asked ourselves if ChatGPT4 could facilitate the standardized application of the AIHDA approach in routine monitoring of healthcare. Results Using strict guidelines, we instructed a GPT model, based on ChatGPT 4, to utilize the AIHDA framework for evaluating healthcare inequalities. To demonstrate the analysis, we used the quality indicator potentially inappropriate medication among elderly and analyzed simulated data of individuals >75-year-old belonging to 36 socioeconomic strata and residing in 21 Swedish regions. The GPT performed an accurate analysis, correctly interpretated the results and formulated an informative short rapport with illustrative tables and figures. Lessons AI appears to facilitate routine monitoring of health care inequity using the AIHDA approach. AI also allows a constructive interaction with the user but, so far, it needs be closely supervised. Key messages • ChatGPT 4 can improve the standardized evaluation of healthcare inequity. • Although the technology is under development and must be used with caution.