Effort expectancy is the most important factor for physicians adopting the assisted diagnostic systems based on large language model：Cross-Sectional Survey Study (Preprint)

BACKGROUND Background: The implementation of digital healthcare innovation was accelerating in China. Some medical large language models (LLM) were applied into different scenarios. The LLM-based assisted diagnostic systems (LLM-ADSs) were a major scenario. OBJECTIVE Objective: The purpose of this study is to evaluate the influencing factors of physicians' adoption of the LLM-ADSs in tertiary hospitals in China. METHODS Methods: We developed and evaluated a 25-item measurement scale and used the results from a web-based survey among 250 physicians to test our hypothetical model, who’s from four grade A tertiary hospitals in four provinces in China. We used partial least squares structural equation modeling (PLS-SEM) to analyze the data. RESULTS Results: The results indicated that physicians' acceptance of LLM-ADSs was moderately influenced by effort expectancy (f2=0.253), and weakly influenced by performance expectancy (f2=0.118), social influence (f2=0.042), and trust in AI (f2=0.028), with the effect size of these factors decreasing in order. Effort expectancy had a strong effect on performance expectancy (f2=0.639). Trust in AI was strongly influenced by performance expectancy (f2=0.404) and weakly influenced by effort expectancy (f2=0.115), respectively. Moreover, personal innovativeness weakly influenced effort expectancy (f2=0.037), and task-technology fit weakly influenced performance expectancy (f2=0.051). The model had explained 68% of physicians' acceptance. CONCLUSIONS The research enhances our understanding of the determinants of physicians’ acceptance of LLM-ADSs. Among these factors, effort expectancy was considered as the top priority. It highlights the importance of optimizing LLM-ADSs to be user-friendly and seamlessly integrated into clinical workflows to increase ease of use.