Improving the Predictive Accuracy of the National Early Warning Score 2: Protocol for Algorithm Refinement

Area of Science:

• Medical Informatics
• Clinical Decision Support
• Health Services Research

Background:

• The National Early Warning Score 2 (NEWS2) is widely used for predicting patient deterioration, but has limited accuracy over extended periods.
• Existing NEWS2 has shown reduced in-hospital mortality but struggles with predicting clinically significant outcomes beyond 24 hours.

Purpose of the Study:

• To improve the predictive accuracy of the NEWS2 scoring system, especially for predictions longer than 24 hours.
• To enhance NEWS2's predictive value in specific populations like older adults and children.
• To investigate the impact of altered data utilization and additional variables on algorithm performance.

Main Methods:

• Utilizing historical patient data from Newcastle upon Tyne Hospitals NHS Foundation Trust.
• Training and testing a predictive algorithm using observational, BMI-related, and outcome data.
• Assessing algorithm performance via accuracy, precision, F1-score, AUC, and ROC curve.

Main Results:

• The study is projected to commence in April 2025, with findings anticipated by the end of 2026.
• Results will be disseminated through symposia, conferences, and peer-reviewed publications.
• A proof of concept for a modified scoring system predicting mortality, ICU admission, sepsis, and cardiac arrest will be demonstrated.

Conclusions:

• A refined NEWS2 algorithm will overcome the original system's limitations in predicting deterioration beyond 24 hours.
• Enhanced predictive accuracy facilitates early detection and timely interventions, potentially reducing mortality and adverse events.
• The improved algorithm can be integrated into clinical decision support systems for healthcare professionals.