A machine learning algorithm-based predictive model for pressure injury risk in emergency patients: A prospective...

Area of Science:

• Emergency Medicine
• Data Science
• Clinical Informatics

Background:

• Pressure injuries are a significant concern in emergency departments.
• Effective risk prediction tools are needed for timely intervention.

Purpose of the Study:

• To develop and optimize machine learning models for predicting pressure injury risk in emergency patients.
• To identify the most effective model for clinical application.

Main Methods:

• A cohort of 312 emergency patients was divided into modeling and validation groups.
• Logistic regression, decision tree, and neural network models were constructed and compared.
• Model performance was evaluated using ROC curve, sensitivity, specificity, and Yoden index.

Main Results:

• The incidence of pressure injuries was 8.97%, with sacrococcygeal region and stage 1 being most common.
• Serum albumin, incontinence, perception, and mobility were identified as independent risk factors.
• The decision tree model demonstrated superior predictive efficacy (AUC 0.866 in validation).

Conclusions:

• The decision tree model is highly effective for individualized pressure injury risk prediction in emergency settings.
• This model serves as a valuable tool for preventing and managing pressure injuries in emergency patients.