Predictive Modeling of Pressure Injury Risk in Patients Admitted to an Intensive Care Unit

Area of Science:

• Medical informatics
• Clinical decision support systems
• Predictive modeling in healthcare

Background:

• Pressure injuries are a significant hospital care challenge.
• Existing risk assessment tools like Norton and Braden scales have limitations.
• Data mining and machine learning offer potential solutions for comprehensive risk factor analysis.

Purpose of the Study:

• Develop a predictive model for pressure injury risk in intensive care unit (ICU) patients.
• Implement and validate the model in a real-world clinical setting.

Main Methods:

• Utilized data mining to extract variables from electronic medical records.
• Employed machine learning techniques, specifically logistic regression, to build the predictive model.
• Conducted retrospective training and testing, followed by prospective validation in a clinical environment.

Main Results:

• The final logistic regression model incorporated 23 variables.
• Achieved high performance with 0.90 sensitivity, 0.74 specificity, and 0.89 area under the curve in initial testing.
• Demonstrated sustained effectiveness when tested in a real-world environment one year post-implementation, outperforming the Norton scale.

Conclusions:

• The developed model accurately predicts pressure injury risk.
• Enables targeted nursing interventions for high-risk patients, optimizing resource allocation and reducing workload.