Trust in artificial intelligence at the bedside: a qualitative interview study with neonatologists

Abstract Background Artificial Intelligence (AI) promises to improve patient outcomes, increase efficiency, and reduce healthcare costs. While AI has demonstrated success in imaging fields like radiology, its application is expanding into other data-rich medical domains, including neonatology. Nonetheless, a disparity persists between technical advancements of AI and its practical use in clinical care. A significant obstacle for integrating these systems into bedside decision-making is physicians’ difficulty in determining when AI can reasonably be trusted. Methods We explored the perspectives of neonatologists at a Dutch academic hospital on trust in AI via in-depth, semi-structured interviews ( n = 17). Thematic analysis was used to identify key themes. Results Three overarching themes emerged: (1) Trusting AI in practice, (2) Maintaining control over trusted AI systems, and (3) Limits of trust. Overall, AI in neonatal clinical settings was welcomed. The concept of trust was delineated as being distinct from that of confidence and reliance. For trusting AI, it was deemed important that AI’s role would be supplemental to the role of the physician. Also, AI validation and adequate oversight were considered essential for physicians’ trust. Preserving human expertise while working with AI in clinical settings was deemed important. Concerns about overreliance and incompatibility of clinical intuition with AI in practice were frequently mentioned as reasons to lose trust in AI. Conclusion This study offers practical insights into how trust in medical AI is conceptualized by neonatologists and highlights key considerations for its acceptance. Clarifying what is expected of physicians in terms of reliance on AI at the bedside and formulating concrete guidance for trust in AI is recommended. Trial registration Not applicable. The research protocol was reviewed by the research quality control committee of the University Medical Center Utrecht and was deemed exempt from the Medical Research Involving Humans Act (research proposal no. 23U0364).