Charting and Predicting Risk: Artificial Intelligence/Machine Learning Pilot Model for Hospital-Acquired Pressure Injuries
Shea Polancich, Chris Hickman, Tracey Dick, Allyson Hall, Curry Bordelon, Ria HearldBackground:
Hospital-acquired pressure injuries (HAPIs) are a significant and global adverse event impacting an estimated 2.5 million patients per year, costing from $20 900 to $151 700 per pressure injury.
Purpose:
The purpose of this pilot study was 2-fold: (1) to evaluate the effectiveness of artificial intelligence and machine learning (AI/ML) models in predicting HAPIs, and (2) to compare that effectiveness with the predictive accuracy of traditional analytic methods.
Methods:
Secondary data analysis was performed for this pilot study. A training dataset was created and used for our exploratory evaluation and comparisons between AI/ML models and traditional analytic methods for pressure injury predictive accuracy.
Results:
While logistic regression provided a reasonable fit and interpretable coefficients, tree-based AI models performed notably better in predicting HAPIs.
Conclusions:
The use of AI/ML models appears to offer an increase in precision for the identification of patients at risk for developing HAPIs.