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

Area of Science:

• Clinical Informatics
• Patient Safety
• Machine Learning in Healthcare

Background:

• Early detection of patient deterioration is crucial for improving hospital outcomes.
• Existing early warning systems (EWS) have limitations in real-time patient surveillance.
• Nursing documentation offers a rich source of data for identifying at-risk patients.

Purpose of the Study:

• To evaluate the effectiveness of the COmmunicating Narrative Concerns Entered by RNs (CONCERN) early warning system (EWS) in reducing inpatient deterioration risk.
• To compare patient outcomes between care teams utilizing the CONCERN EWS and those receiving usual care.
• To assess the impact of a machine learning-based EWS on mortality, length of stay, and other critical care events.

Main Methods:

• A 1-year, multisite pragmatic trial involving 74 clinical units across 2 health systems.
• Cluster-randomization of units into intervention (CONCERN EWS) and usual care groups.
• Analysis of 60,893 adult hospital encounters, focusing on coprimary outcomes of in-hospital mortality and length of stay.

Main Results:

• The CONCERN EWS intervention group showed a 35.6% decrease in the instantaneous risk of death (HR, 0.64; P < 0.0001).
• Length of stay was reduced by 11.2% (IRR, 0.91; P < 0.0001) in the intervention group.
• A 7.5% decrease in the risk of sepsis (HR, 0.93; P = 0.0317) was observed, alongside a 24.9% increase in unanticipated ICU transfers (HR, 1.25; P = 0.0011).

Conclusions:

• A machine learning-based EWS, informed by nursing surveillance patterns, effectively reduces inpatient deterioration risk.
• The CONCERN EWS demonstrates significant benefits in reducing mortality and length of stay.
• While improving safety, the system's impact on ICU transfers warrants further investigation.