Healthcare professionals’ perceptions on AI-assisted decision-making in clinical practice: a qualitative meta-synthesis

Healthcare professionals’ perceptions are central to how AI-assisted decision-making (AI-ADM) is understood, accepted, and used across clinical contexts. This qualitative meta-synthesis, guided by Normalization Process Theory (NPT), synthesized healthcare professionals’ perceptions on AI-ADM and identified factors shaping its clinical use. Ten databases were searched through November 9, 2025. We conducted a framework-guided qualitative synthesis by combining the first two stages of Thomas and Harden’s thematic synthesis with NPT-based deductive framework interpretation. Analysis of 13 studies from nine countries revealed that while healthcare professionals recognized AI-ADM’s potential in accelerating diagnosis, optimizing workflows, and enabling personalized medicine, critical challenges emerged across NPT’s four constructs: Coherence (challenged by difficulties in understanding the reasoning and evidence base underlying AI-ADM recommendations), Cognitive Participation (hindered by responsibility ambiguity and autonomy concerns), Collective Action (complicated by privacy, data security, and informed consent concerns), and Reflexive Monitoring (shaped by concerns about real-world reliability, over-reliance on AI-ADM, weakening humanistic care, and equity in resource allocation). Successful AI-ADM use requires clear accountability frameworks, transparent and interpretable models with explicit performance boundaries, and validated monitoring systems addressing algorithmic bias. Addressing these interconnected challenges may support responsible AI-ADM use while upholding the fundamental values of fair, transparent, and patient-centered healthcare.