[Development and Evaluation of Electronic Health Record Data-Driven Predictive Models for Pressure Ulcers]

Area of Science:

• Medical Informatics
• Clinical Nursing Research
• Health Data Science

Background:

• Pressure ulcers represent a significant healthcare challenge, impacting patient outcomes and increasing healthcare costs.
• Current risk assessment tools, such as the Braden Scale, have limitations in predictive accuracy.
• Electronic Health Records (EHRs) contain vast amounts of data that can potentially improve risk prediction.

Purpose of the Study:

• To develop and validate predictive models for pressure ulcer incidence utilizing EHR data.
• To compare the predictive performance of these novel models against the established Braden Scale.

Main Methods:

• A retrospective case-control study involving 202 pressure ulcer patients and 14,705 controls.
• Development of predictive models using logistic regression, Cox proportional hazards regression, and decision tree algorithms.
• Comparative analysis of model predictive validity using Area Under the Curve (AUC) metrics.

Main Results:

• The logistic regression model achieved the highest AUC (0.97), followed by decision tree (0.95) and Cox regression (0.95) models.
• All developed EHR-based models demonstrated superior predictive performance compared to the Braden Scale (AUC 0.82).
• Key predictive factors identified included decreased mobility and endotracheal tube presence.

Conclusions:

• EHR-derived predictive models offer significantly enhanced accuracy for pressure ulcer risk assessment over the Braden Scale.
• These models hold potential for integration into clinical decision support systems to aid nursing staff.
• Automated risk assessment using advanced predictive modeling can improve preventative strategies for pressure ulcers.