Differentiating Pressure Ulcer Risk Levels through Interpretable Classification Models Based on Readily Measurable...

Area of Science:

• Clinical Nursing
• Machine Learning in Healthcare
• Patient Safety

Background:

• Pressure ulcers pose a significant clinical risk, necessitating accurate risk assessment tools.
• Existing nursing datasets often suffer from imbalanced learning, hindering model performance.
• Clinical decision-making for pressure ulcer prevention requires interpretable and practical tools.

Purpose of the Study:

• To develop and validate an interpretable decision tree model for estimating pressure ulcer risk.
• To address data imbalance issues in nursing datasets through oversampling techniques.
• To introduce a novel visualization method for enhancing clinical usability of risk models.

Main Methods:

• Utilized decision tree modeling for pressure ulcer risk prediction.
• Applied various oversampling configurations to mitigate imbalanced learning in nursing data.
• Focused on three key indicators: patient mobility, activity level, and skin moisture.
• Developed a novel tabular visualization for decision tree interpretation.

Main Results:

• Decision tree models demonstrated interpretable and effective estimation of pressure ulcer risk levels.
• Oversampling techniques improved data quality for more robust modeling.
• The proposed visualization method enhances the practical application of the models in clinical settings.
• Stratified cross-validation confirmed the models' performance.

Conclusions:

• The developed decision tree approach provides a practical and interpretable tool for nursing professionals to assess pressure ulcer risk.
• The models support informed clinical decision-making for timely and tailored preventive interventions.
• This research offers a valuable enhancement to nursing care by improving pressure ulcer risk estimation.