Predictive model for estimating risk of crush syndrome: a data mining approach

Area of Science:

• Disaster Medicine
• Emergency Medicine
• Public Health

Background:

• Lack of standardized triage for earthquake victims with crush injuries.
• Scarcity of epidemiologic and quantitative data for crush injury management.
• Need for predictive tools to guide medical response in mass casualty events.

Purpose of the Study:

• To develop and validate predictive triage models for crush injury victims.
• To identify key clinical factors associated with severe crush syndrome.
• To improve resource allocation and patient management in earthquake scenarios.

Main Methods:

• Retrospective cohort study of 372 Kobe earthquake victims with crush injuries.
• Logistic regression analysis to assess 21 risk factors for crush syndrome outcomes.
• Development of two predictive models: initial field triage and secondary hospital assessment.

Main Results:

• Initial triage model (pulse rate, delayed rescue, abnormal urine color) achieved an AUC of 0.73.
• Secondary hospital triage model (WBC, tachycardia, abnormal urine color, hyperkalemia) achieved an AUC of 0.76.
• Both models demonstrated utility in classifying patient risk for severe crush syndrome.

Conclusions:

• Developed triage models can assist non-experts in identifying high-risk crush injury patients.
• Models facilitate efficient utilization of limited medical and transportation resources post-disaster.
• Enhanced triage can improve outcomes for earthquake victims with crush injuries.