OBJECTIVE  To analyze the risk factors for multidrug-resistant organism (MDRO) infections in patients in the intensive care unit (ICU), this study constructed prediction models for patients with MDRO infections based on 6 machine learning algorithms, and selected the optimal model by evaluating relevant model indicators, thus providing references for clinicians to identify high-risk patients early and implement corresponding preventive measures in a timely manner.

METHODS  A total of 946 patients admitted to the ICU of the Affiliated Hospital of Xuzhou Medical University from Jun. 2019 to Jun. 2023 were included (473 with MDRO infections and 473 without infections). Binary logistic regression analysis was used to screen risk factors as characteristic variables for constructing prediction models. Six models were constructed, including logistic regression model, artificial neural network model, decision tree model, random forest model, support vector machine model and extreme gradient boosting model.RESULTS Risk factors for MDRO infections in ICU patients included admission from another hospital or the emergency department (OR=2.635), ICU stay duration ≥7 days (OR=1.291), surgery (OR=3.089), chronic pulmonary disease (OR=3.664), peripheral venous catheterization (OR=2.111), indwelling abdominal drainage tube (OR=3.382), use of ≥3 types of antimicrobial agents (OR=1.001) and antimicrobial agent use duration ≥1 week (OR=2.323) (P ＜0.05). Among the prediction models for MDRO infections in ICU patients constructed through machine learning algorithms, the logistic regression model demonstrated superior performance in terms of the area under the receiver operating characteristic curve, sensitivity, specificity, positive predictive value, negative predictive value and F1 score, making it the optimal model.

CONCLUSION  Clinical healthcare workers should pay attention to the risk factors for MDRO infections in patients and provide targeted interventions as early as possible to reduce the risk of MDRO infections in ICU patients.