Predicting the risk of hospital-acquired pressure ulcers in intensive care unit patients based on decision tree

Abstract

Objective To explore and evaluate the feasibility and accuracy of applying decision tree methods to predict the risk of hospital-acquired pressure ulcers (HAPUs) in intensive care unit (ICU) patients, and provide a theoretical basis for the prevention of HAPUs in clinical practice. Methods A retrospective design was used to collect 468 patients' records, and all of the patients were hospitalized in ICUs including medical intensive care unit and coronary care unit between October 2011 and October 2013 in a 3 A grade hospital in Guangzhou. The CART algorithm was used to construct the decision tree risk prediction model. The area under the receiver operating curve (AUC), sensitivity and specificity were used to evaluate the predictive validity of the decision tree model compared with the Braden Scale. Results The decision tree model had four stratums and eleven nodes. Six classification rules and three styles of high-risk populations were screened out: (1)age >81; (2)age ≤81 combined with fecal incontinence; (3)age ≤81 combined with total Braden score ≤13 and diastolic blood pressure < 60mmHg (1 mmHg= 0.133 kPa). The AUC of the decision tree model was significantly higher than the Braden Score (Z=2.31,P < 0.05). The sensitivity and specificity of the decision tree model (0.809 and 0.703) were higher than the Braden Score (0.777 and 0.587). Conclusions The decision tree model is an easy and feasible tool to predict the risk of HAPUs in ICU patients, and it can be used to screen high-risk populations. Key words: Intensive care units; Pressure ulcer; Decision tree; Prediction