The Challenges of AI Applications in Health Care: Qualitative Study

The application of AI in clinical settings is becoming increasingly widespread, exerting a profound impact on health care professionals' work patterns, decision-making efficiency, and clinical practice. However, there is currently a lack of systematic research on health care professionals' actual experiences, barriers to use, and attitudes toward the clinical application of AI. AI has shown great potential to streamline clinical workflows and improve service quality, yet its real-world adoption faces notable obstacles.

This study aimed to comprehensively explore the multifaceted challenges encountered during the implementation of AI in routine health care work. The findings are intended to offer empirical evidence and practical references for improving the deployment, management, and popularization of AI tools in clinical settings.

We adopted a qualitative design with inductive thematic analysis. Through purposive sampling, 23 health care workers with diverse job roles, professional titles, and work experience were recruited for one-on-one semistructured interviews. Two researchers independently performed data coding and cross-checked the results to ensure intercoder reliability. NVivo (version 15.0; Lumivero) software was used to organize the qualitative data and support thematic analysis.

All data were fully transcribed. The interviews identified 5 themes and 13 subthemes related to the challenges of applying AI in the health care sector. The five themes are as follows: (1) construction of AI trust mechanisms and shaping of rational cognition, (2) AI-assisted decision-making models and health care collaboration, (3) AI user experience, (4) barriers to AI use and risk concerns, and (5) attitudes toward AI substitution and perception of core health care value. Participants generally recognized the value of AI in clinical work, yet they highlighted multiple practical barriers and potential risks.

The application of AI in health care is shaped by the interplay of complex factors spanning technology, organization, cognition, and regulation. A balanced perspective is essential to ensure that AI integration enhances rather than undermines clinical practice. While AI has great potential to boost efficiency and support clinical decision-making, it also highlights risks such as overreliance, cognitive decline, and the long-term deterioration of professional skills and independent clinical reasoning. Accordingly, efforts should be made to optimize the design and deployment of AI tools, along with a fundamental rethinking of professional training and governance frameworks. Future work should prioritize developing intelligence-augmenting rather than intelligence-replacing systems. Critical evaluation of AI outputs needs to be incorporated into clinical education, and relevant policies should be formulated to prevent cognitive complacency. The ultimate goal is to achieve a sound, sustainable, and people-centered integration of AI into routine clinical practice.