Data-driven approach to Early Warning Score-based alert management.

Area of Science:

• Clinical Informatics
• Patient Safety
• Health Systems Engineering

Background:

• Electronic health records (EHRs) with integrated alerting systems are crucial for patient safety.
• Designing effective EHR-driven alerts that manage alert burden and ensure timely notifications remains challenging.
• Proactive development and evaluation of alert-generating approaches are needed for successful implementation.

Purpose of the Study:

• To proactively develop and evaluate a systematic alert-generating approach for Early Warning Score (EWS) implementation.
• To assess the impact of different alert algorithms on alert quantity, frequency, and early recognition of clinical deterioration.
• To inform strategies for optimizing alert deployment in clinical practice.

Main Methods:

• Quantified alert system impact using three distinct alert algorithms (triggering/muting criteria).
• Utilized retrospective EHR data from December 2015 to July 2016 across three hospital units.
• Included general medical, acute care for the elderly, and heart failure patient populations.

Main Results:

• Alert generation quantity and frequency varied significantly based on care location, patient severity, and characteristics.
• Compared algorithms for early recognition of clinical deterioration against estimated alert burden.
• Demonstrated a clear dependency between alert metrics and patient/unit-specific factors.

Conclusions:

• EWS-based alert algorithms show potential for effective alert management before clinical integration.
• Tailoring alert systems to specific patient and care location needs is essential for enhancing patient safety.
• Findings support alternative EWS-based alert deployment strategies to optimize early detection and manage alert burden.