Multicenter development and validation of a risk stratification tool for ward patients

Area of Science:

• Clinical Informatics
• Patient Safety
• Predictive Analytics

Background:

• Traditional ward risk scores often rely on subjective assessments and limited vital signs.
• Existing methods may not fully capture patient acuity using comprehensive data.

Purpose of the Study:

• To develop and validate a novel risk stratification tool utilizing readily available electronic health record (EHR) data.
• To improve the accuracy of predicting adverse events on hospital wards.

Main Methods:

• An observational cohort study included 269,999 patient admissions across five hospitals.
• Discrete-time survival analysis predicted cardiac arrest, ICU transfer, or death.
• Predictors included laboratory results, vital signs, and demographics; model developed on 60% and validated on 40% of data.

Main Results:

• The developed EHR-based risk score demonstrated superior accuracy compared to the Modified Early Warning Score (MEWS).
• Area under the receiver operating characteristic curves were significantly higher for the new score across all outcomes (e.g., 0.83 vs. 0.71 for cardiac arrest).
• The net reclassification index (NRI) indicated a substantial improvement in risk prediction accuracy.

Conclusions:

• A validated, accurate ward risk stratification tool was developed using common EHR variables from a large, multicenter dataset.
• The tool shows potential for enhancing patient safety and clinical decision-making on hospital wards.
• Further research is warranted to assess the impact of real-time implementation on patient outcomes.