Real-time surveillance system for patient deterioration: a pragmatic cluster-randomized controlled trial.

Summary

In a multicenter pragmatic cluster-RCT of 60,893 encounters, a nursing documentation–based machine learning early warning system significantly reduced instantaneous risk of in-hospital death (HR 0.64), shortened length of stay (IRR 0.91), and reduced sepsis risk (HR 0.93), while increasing unanticipated ICU transfers (HR 1.25). No adverse events were reported.

Key Findings

• 35.6% decreased instantaneous risk of in-hospital death in intervention units (adjusted HR 0.64, 95% CI 0.53-0.78).
• 11.2% reduction in length of stay (adjusted IRR 0.91, 95% CI 0.90-0.93).
• 7.5% decreased instantaneous risk of sepsis (adjusted HR 0.93, 95% CI 0.86-0.99).
• 24.9% increased instantaneous risk of unanticipated ICU transfer (adjusted HR 1.25, 95% CI 1.09-1.43); no adverse events reported.

Clinical Implications

Health systems can consider implementing nursing documentation–driven EWS to reduce mortality and sepsis, with operational planning for increased (likely earlier) ICU transfers.

Limitations

• Generalizability may vary with documentation practices and EHR integration
• Algorithm transparency and external validation details not fully delineated in the abstract

Future Directions

Assess transportability to diverse EHRs and settings, understand mechanisms behind increased ICU transfers, and evaluate cost-effectiveness and equity impacts.