| Sumario: | Objective: To establish a predictive model of pressure injury (PI) in intensive care unit (ICU) patients by combining limb muscle strength and transcutaneous oxygen pressure (TcPO2) with clinical data. Methods: 429 patients admitted to the ICU of our hospital from April 2021 to January 2023 were retrospectively selected as the research objects. The patients were divided into a training set (n= 299) and a test set (n=130) according to the ratio of 7:3 by random sampling. Logistic regression model was used to analyze the influencing factors of pressure injury in patients, and random forest algorithm was used to construct a random forest model. The area under the receiver operating characteristic (ROC) curve, sensitivity, and specificity were used to evaluate the predictive efficacy of the two models. Results: 71 cases occurred pressure injury among 299 patients in the training set, with an incidence of 23.75%. Multivariate analysis showed that diabetes mellitus, APACHE-II score, Braden score, TcPO2, limb muscle strength, mechanical ventilation, and use of vasoactive drugs were the influencing factors of PI in ICU patients (P<C0.05). The order of importance of predictors in the random forest model was TcPO2, APACHE- II score, limb muscle strength, Braden score, mechanical ventilation, use of vasoactive drugs, and combination of diabetes. The test set was introduced to verify the prediction efficiency of the two models. It was found that the area under the ROC curve of the logistic regression model was 0.871, the sensitivity was 84.5%, and the specificity was 81.4%. The area under the ROC curve of the random forest model was 0.912, the sensitivity was 88.5%, and the specificity was 84.2%. Conclusions: TcPO2, APACHE-II score, limb muscle strength, Braden score, mechanical ventilation, use of vasoactive drugs, and combination of diabetes were the influencing factors of PI risk in ICU patients. Furthermore, the predictive performance of the random forest model for PI risk in ICU patients was better than that of the Logistic regression model.
|
|---|