Determinants of healthcare professionals’ acceptance of DeepSeek-assisted clinical decision support applications: a cross-sectional study

Abstract

Large language models (LLMs) have introduced new paradigms for intelligent medical decision support, with some healthcare institutions deploying private LLMs to explore efficiency improvements through workplace integration. However, systematic research on the key determinants influencing LLM adoption in healthcare settings remains scarce. This empirical study examines the acceptance mechanisms among clinical staff from cognitive, social, and risk perspectives, and proposes strategies to promote the implementation of LLMs in healthcare. A cross-sectional design was employed. A total of 392 healthcare professionals from a Chinese tertiary hospital with DeepSeek deployment participated in the study. A conceptual framework integrating the Unified Theory of Acceptance and Use of Technology (UTAUT) and Perceived Risk Theory was tested using structural equation modeling (SEM) to analyze behavioral intention. Thematic analysis was employed to analyze healthcare professionals’ core demands and suggestions for improvement regarding the clinical application of DeepSeek. The integrated UTAUT - Perceived Risk model effectively accounted for variations in clinical staff’s intention to adopt LLMs. Institutional social influence, performance expectancy, and facilitating conditions directly drove adoption, whereas effort expectancy and perceived risk showed no direct effects. Qualitative analysis identified four core needs: the establishment of training systems, optimization of user experience, enhancement of reliability, and expansion of application scenarios. The adoption of private LLMs in healthcare settings is characterized by a three-tier pattern of institutional shaping, technological mediation, and risk modulation, which informs a dual strategy of enhancing institutional governance alongside contextualized implementation. This study provides theoretical support for the deployment and application of medical LLMs and extends the explanatory scope of UTAUT and Perceived Risk Theory within the field of generative AI.

Ähnliche Arbeiten