Background: Clinical deterioration in hospitalized patients often precedes adverse events such as cardiac arrest, unplanned intensive care unit (ICU) admission, and mortality. While traditional early warning systems (EWS) based on vital signs have improved detection, their predictive accuracy and integration into nursing workflow remain suboptimal. Artificial intelligence (AI)-driven systems leveraging nursing surveillance data may enhance early detection and response.

Objective: To evaluate the effectiveness of an AI-driven early warning system integrated into nursing practice on early detection of clinical deterioration, nursing response time, and patient safety outcomes in adult medical-surgical wards.

Methods: A pragmatic randomized controlled trial conducted across 12 medical-surgical units in two tertiary hospitals. Adult patients (N = 4,286) were randomized to either the AI-driven EWS (intervention, n = 2,143) or standard EWS care (control, n = 2,143). The AI system analyzed nursing documentation patterns and vital signs in real time, generating risk scores displayed within the electronic health record. Primary outcomes included time to clinical deterioration detection, nursing response time to alerts, and composite patient safety outcomes (in-hospital mortality, unplanned ICU transfer, and cardiac arrest). Data were analyzed using intention-to-treat principles with mixed-effects models.

Results: The AI-driven EWS significantly reduced time to deterioration detection (median 2.1 vs. 4.8 hours; p < 0.001) and nursing response time (median 18 vs. 42 minutes; p < 0.001). The composite patient safety outcome occurred in 11.2% of intervention patients versus 16.8% of controls (adjusted odds ratio 0.64, 95% CI 0.53–0.78; p < 0.001). Unplanned ICU transfers were higher in the intervention group (24.9% increase; p = 0.001), suggesting appropriate escalation.

Conclusion: AI-driven EWS integrated into nursing practice substantially improved early detection, response time, and patient safety outcomes. These findings support implementation of AI-enhanced surveillance systems in acute care settings.