The predictive value of column chart model and random forest model based on preoperative systemic inflammatory factors for postoperative infection in patients with open tibiofibular fractures

SONG Zhi-hui; Zulipikar Yasen; AN Wei

doi:10.3969/j.issn.2095-252X.2025.09.007

Abstract：Objective To explore the predictive value of column chart model and random forest model based on preoperative systemic inflammatory factors for postoperative infection in patients with open tibiofibular fractures.Methods 196 open tibiofibular fracture patients admitted to our Orthopedic Center from June 2023 to October 2024 were selected as the research subjects.They were randomly divided into training set(137 cases)and testing set(59 cases)in 7:3 ratio.Patients in the testing set were divided into infected group(39 cases)and uninfected group(98 cases)according to whether postoperative infection occurred.The general clinical data and the expression level of serum systemic inflammatory factors of patients between the two groups were compared.The independent predictors of postoperative infection in patients with open tibiofibular fractures were obtained by multivariate logistic regression analysis.A column chart model and random forest model was established.The predictive performance and net return of column chart model and random forest model were evaluated through ROC analysis and decision curve analysis,respectively.Results Multivariate logistic regression analysis showed that CRP,PCT,TNF-α and IL-6 before operation were independent predictors of postoperative infection in patients with open tibiofibular fracture(P<0.05).A column chart model and random forest model was constructed to predict the postoperative infection risk of open tibiofibular fractures patients based on CRP,PCT,TNF-α and IL-6.The ROC analysis results showed that the area under the curve(AUC)for predicting postoperative infection in open tibiofibular fracture patients in testing set of column chart model and random forest model were 0.807(95%CI:0.732-0.879)and 0.862(95%CI:0.778-0.925),respectively.The AUC of the random forest model was significantly higher than that of the column chart model,and the differences were statistically significant(Z=5.645,P<0.05);The decision curve analysis results showed that within the threshold probability range of 0.28-0.81,the predictive performance and net return of random forest model predicting postoperative infection in open tibiofibular fracture patients were higher than the column chart model in the testing set.Conclusions The nomogram model and random forest model constructed based on CRP,PCT,TNF-α,and IL-6 both demonstrate good predictive value for postoperative infections in patients with open tibiofibular fractures,with the random forest model showing higher predictive performance.