Integrated Machine Learning Approach for the Early Prediction of Pressure Ulcers in Spinal Cord Injury Patients

Area of Science:

• Biomedical Informatics
• Clinical Research
• Machine Learning in Healthcare

Background:

• Pressure ulcers (PUs) significantly impair quality of life for spinal cord injury (SCI) patients.
• Prompt intervention is crucial for managing PUs in SCI populations.
• Developing advanced predictive models for PUs in SCI patients is a critical clinical need.

Purpose of the Study:

• To develop and validate machine learning (ML) models for predicting the occurrence of pressure ulcers (PUs) in patients with spinal cord injury (SCI).
• To identify key predictors of PU development using a multidimensional data approach.

Main Methods:

• Analysis of medical records from 539 SCI patients, including 139 variables (demographics, neurological status, functional ability, laboratory data).
• Application of various ML algorithms: GNN, DNN, SVM (linear and RBF), KNN, RF, and LR.
• Integrative analysis combining laboratory, neurological, and functional data for model development.

Main Results:

• The linear support vector machine (SVM_linear) model achieved superior predictive performance (AUC = 0.904, accuracy = 0.944) via 5-fold cross-validation.
• Key predictors identified included limb functional status and inflammatory laboratory markers.
• External validation indicated challenges in model generalization, guiding future research.

Conclusions:

• A comprehensive, multidimensional data approach is essential for effective PU prediction in SCI patients, particularly in acute and subacute phases.
• The developed ML models demonstrate potential for early PU detection and prevention.
• These models can significantly enhance patient care in clinical settings by facilitating proactive interventions.