Comparison of Early Warning Scores Utilising Patient Trends

Area of Science:

• Clinical Medicine
• Healthcare Informatics
• Patient Safety

Background:

• Early Warning Scores (EWS) are crucial for identifying in-hospital patient deterioration.
• Current EWS often lack trend analysis of physiological data, despite its known importance.
• This limitation can hinder timely intervention and potentially lead to adverse outcomes.

Purpose of the Study:

• To compare the performance of trend-based EWS against a static, state-of-the-art EWS.
• To evaluate two novel trend-based EWS models against an existing trend-based model and a static reference.
• To demonstrate the predictive performance benefits of incorporating physiological trends into EWS.

Main Methods:

• Developed and evaluated two new trend-based Early Warning Scores using logistic regression.
• Compared these new scores, an existing trend-based score, and the National Early Warning Score (a static model).
• Assessed performance using the area under the receiver operating characteristic curve (AUC).

Main Results:

• All evaluated trend-based Early Warning Scores outperformed the static National Early Warning Score (reference model).
• The study provides evidence supporting the advantages of trend-based EWS in predictive performance.
• High predictive accuracy was achieved using a minimal set of predictors in the developed models.

Conclusions:

• Trend-based Early Warning Scores offer superior performance in predicting patient deterioration compared to static scores.
• Incorporating physiological trends into EWS models is a valuable strategy for improving patient safety.
• Effective EWS can be developed with a limited number of predictive variables.