Construction of A Prediction Model of Unplanned Reoperation After Radical Surgery for Oral Cancer Based on Machine Learning
[Objective]To investigate risk factors for unplanned reoperation following radical surgery for oral cancer and to construct a prediction model based on machine learning.[Methods]The clinical data of 684 patients with oral cancer who underwent surgical treatment from January 2017 to January 2023 were collected.Patients were divided into reoperation group(n=78)and non-reoperation group(n=606).Support vector machine,random forest method and Lasso regression were used to screen for factors associated with reoperation,and multivariate Logistic regression analysis was employed to identify independent risk factors.A Nomogram model was constructed based on these risk factors,and its predictive efficacy and clinical value were assessed.[Re-sults]The reoperation group had significantly higher proportion of patients with male,history of diabetes,smoking,surgical site infection and flap necrosis compared to the non-reoperation group.Quantitative data showed that the reoperation group also had significantly longer operation times and prothrombin time than that in the non-reoperation group.Multivariate Logistic regression analysis indicated that operation time(OR=3.294),prothrombin time(OR=1.850),platelet(OR=2.008),male(OR=0.377),history of diabetes(OR=0.368),surgical site infection(OR=0.330)and flap necrosis(OR=0.267)were independent risk factors for reoperation.The Nomogram model had an area under the curve of 0.748,and the calibration curve almost overlapped with the ideal curve.The decision curve analysis showed that the model has a benefit rate between 0-76％.[Conclu-sion]Operation time,prothrombin time,platelet,male,history of diabetes,surgical site infection and flap necrosis are independent risk factors for unplanned reoperations after radical surgery for oral cancer.The con-structed Nomogram model can effectively predict the risk of reoperation for patients,which may help clinicians to take targeted measures and reduce postoperative complications.
NETL
NSTL
万方数据
