Integrating Large Language Models in Healthcare Decision-Making

Authors

Anh L. Huynh, University of HoustonFollow
M. Mahbub Hossain, University of HoustonFollow
Jyoti Das, Colorado State UniversityFollow

Abstract

Large language models (LLMs) are increasingly adopted across industries, leaving healthcare as no exception. Despite a growing societal interest in LLMs, there is a scarcity of research synthesis on how to use LLMs in healthcare decision-making, factors influencing such applications, and associated outcomes. This study conducts a systematic review using a grounded theory approach (Walsh and Rowe 2023), and evaluates the empirical evidence on integrating LLMs in healthcare decision-making. A total of 176 articles were identified from five scholarly databases that reported the integration of LLMs in different healthcare contexts supporting decision-making in patients and providers. Notable applications of LLMs in healthcare decision-making were categorized into four types. First, LLMs have demonstrated high efficacy in analyzing clinical texts and images and recommending diagnostic options in different clinical scenarios. Second, LLMs provided comparable performance in designing treatment plans for people with chronic diseases. Third, LLMs were used in clinical boards and assisted compliance with guidelines, enhancing the accuracy of clinical evaluations. Lastly, LLMs facilitated informed decision-making by providing explanations and supportive information within decision aids. Despite several benefits of using LLMs in healthcare decision-making, there are unique challenges that affect the broader adoption and operationalization of LLMs in healthcare settings. Limited use cases in clinical contexts do not provide meaningful insights into how LLMs can support decision-making in community health organizations or large healthcare systems with diverse information management practices. Moreover, the digital transformation of healthcare emphasizes value creation through the use of IT (Matt et al 2015), whereas early-stage applications of LLMs do not provide clarity on how they complement or promote digital transformation alongside other technological innovations in healthcare. This study provides a grounded theoretical model for analyzing healthcare-specific implementation challenges for decision-making using LLMs that may facilitate meaningful and sustainable use of current and future LLM models. Also, this study offers empirical contributions to IS scholarship by synthesizing existing evidence on the applications of LLMs in healthcare decision-making. Furthermore, this study contributes to IS theories by establishing theoretical constructs and relationships for adopting LLMs in healthcare, which is consistent with emerging theories of artificial intelligence (AI) adoption in practice settings. References Matt, C., Hess, T., & Benlian, A. (2015). Digital transformation strategies. Business & Information Systems Engineering, (57), pp. 339-343. Walsh, I., & Rowe, F. (2023). BIBGT: combining bibliometrics and grounded theory to conduct a literature review. European Journal of Information Systems, (32:4), pp. 653-674.

Paper Number

tpp1337

Recommended Citation

Huynh, Anh L.; Hossain, M. Mahbub; and Das, Jyoti, "Integrating Large Language Models in Healthcare Decision-Making" (2024). AMCIS 2024 TREOs. 123.
https://aisel.aisnet.org/treos_amcis2024/123