| Sumario: | Objective: To analyze the risk factors of nosocomial multi-drug resistant bacterias (MDRO) infection in trauma patients undergoing surgery by Lasso "Logistic regression analysis and classification tree(CHAID) algorithm, build a risk prediction model and compare the results. Methods: The clinical data of trauma inpatients in Zhengzhou University Affiliated Zhengzhou Central Hospital from January 2019 to January 2022 were retrospectively analyzed. The risk prediction models were established by CHAID algorithm and Lasso-Logistic regression, respectively. The goodness of fit test was used to evaluate the effect of the model, and the area under the receiver operating characteristic curve (ROC) curve (AUC) was used to compare the advantages and disadvantages of the two prediction models. Results: A total of 821 trauma patients were included as the modeling group, including 191 trauma patients with MDRO 23.26%; Classification tree model and logistic regression showed that APACHE II score ≥ 20 scores, fever days ≥ 3 days, hospitalization days ≥ 10 days, PCT level ≥ 0.5 ng/L on admission were independent risk factors for postoperative MDRO infection in trauma patients. The risk prediction accuracy of classification tree model was 79.2%, and the model fit effect was good. The Hosmer-Lemeshow goodness of fit test for Lasso-Logistic regression showed that the fitting effect of the model was relatively good (P=0.146). And the Bootstrap internal validation showed that the prediction ability of the model was good. The AUC of classification tree model was 0. 792 (95% CI 0.763-0.819),and the AUC of Lasso-Logistic regression model was 0. 862 (95%CI 0.836-0.885),the predictive value of the two models were medium. The difference between the two models was statistically significant (P<0.001). Net Reclassification Index(NRI) evaluation indicated that the Lasso-Logistic regression model was superior to the classification tree model (NRI=0.153 6). Conclusion: Both models could provide a more intuitive form of presentation. The complementary combination of the two models could early identify the risk factors of postoperative MDRO infection in trauma patients from different perspectives. We should take effective prevention and control measures to reduce the incidence of MDRO nosocomial infection.
|
|---|