AI-CDSS adoption among healthcare professionals: the influence of endowment-based autonomy threats and configuration preferences

Objective The rapid deployment of artificial intelligence-based clinical decision support systems (AI-CDSS) has outpaced current understanding of factors shaping healthcare professionals' adoption and resistance. This study examines how functional and endowment-based autonomy threats relate to resistance and intention to use AI-CDSS, and whether attitudes toward assistive versus replacement-oriented autonomous configurations differ across clinical scenarios. Methods A cross-sectional survey was conducted among healthcare professionals at six hospitals across five Chinese cities that had deployed AI-CDSS, yielding 551 valid responses. The structural model integrated the Unified Theory of Acceptance and Use of Technology, Technology Threat Avoidance Theory, and endowment effect theory, and was analyzed using partial least squares structural equation modeling and exploratory neural network analysis. A scenario-based repeated-measures analysis compared attitudes toward assistive versus replacement-oriented autonomous configurations across five clinical tasks. Results Performance expectancy and social influence positively predicted behavioral intention, with social influence showing a stronger effect. Functional threats (to the diagnostic process and outcomes) and endowment-based threats (attachment to expertise and ownership loss) were significant predictors of resistance, which was associated with lower behavioral intention. Neural network analysis identified ownership loss as the most salient predictor of resistance. Healthcare professionals reported consistently more favorable attitudes toward assistive than toward replacement-oriented autonomous configurations across all tasks. Conclusions Resistance to AI-CDSS reflects not only functional concerns about diagnostic processes and outcomes but also endowment-based threats related to expertise and ownership over clinical judgment. The consistent preference for assistive configurations suggests that preserving professional control is important for acceptance. Institutions and technology providers should prioritize autonomy-preserving design and clarify organizational expectations to support sustainable adoption and inform digital health policy. Public interest summary Healthcare institutions are increasingly using artificial intelligence-based clinical decision support systems (AI-CDSS) to support clinical decision-making. This study surveyed 551 healthcare professionals across six hospitals in China to understand how they respond to AI-CDSS. Professionals expressed stronger intention to use AI-CDSS when they found it useful and when colleagues, supervisors, or institutional expectations supported its use. However, concerns about losing control over clinical decisions and feeling that professional expertise was being devalued increased resistance. Across five clinical situations, professionals consistently favored AI-CDSS that assisted their decisions over systems designed to perform tasks with minimal human involvement. These findings suggest that successful AI-CDSS implementation depends not only on technical usefulness, but also on designing systems that support, rather than replace, healthcare professionals' clinical authority and expertise.