Why Nurses Intend to Override AI Alerts: How Alert Fatigue, Moral Distress, and Team Psychological Safety Shape Self-Reported Trust Calibration Toward Clinical Decision Support

Background and Objectives: Hospitals increasingly use AI tools that give nurses on-screen alerts and recommendations (AI-supported clinical decision support, AI-DSS). When nurses override these alerts too often, useful guidance can be lost; when they trust them blindly, errors can slip through. We examined which work and wellbeing factors are associated with nurses’ self-reported intention to override AI alerts, rather than observed override behavior. Methods: We surveyed 239 registered nurses (76.6% female; mean age 33.7 years) at a large hospital in Timișoara, Romania, from January to March 2025. Questionnaires measured alert fatigue, moral distress, mental workload, sleep problems, resilience, team psychological safety, and how strongly nurses intended to override AI alerts. Results: Nurses fell into three groups: those who tended to over-trust AI (26.8%), those with balanced trust (41.0%), and those who resisted it (32.2%). The resistant group had the strongest intention to override alerts and the weakest sense of psychological safety. Alert fatigue was the factor most strongly associated with override intention, and this association was partly accounted for by moral distress. The indirect association was weaker among nurses reporting higher team psychological safety. An exploratory model using these factors distinguished nurses with high self-reported override intention with acceptable accuracy. Because all variables were measured at a single time point, findings are associative and hypothesis-generating rather than causal. Conclusions: How nurses respond to AI alerts depends less on the technology than on their workload, ethical strain, and team climate. Cutting unnecessary alerts, easing moral distress, and building psychological safety may help nurses use AI more safely.