Abstract： Objective Establishing a predictive model for adverse behavior changes after tonsillectomy in children.Methods A total of 2 368 patients who visited the Department of Otorhinolaryngology Head and Neck Surgery at the First People's Hospital of Tianmen City from July 2013 to December 2021 were selected as research sample.They were divided into a modeling group and a validation group in chronological order of hospitalization.All patients underwent tonsillectomy,and their behavior questionnaire was evaluated after hospitalization.The patients were divided into two groups:a group with adverse behavior changes and a group without adverse behavior changes.Lasso regression and random forest screening were used to select independent variables for the modeling group to construct different models A and B.The minimum Akashi information criterion was selected as the optimal model to establish a column chart through the constructed model.The model from three aspects of discrimination,calibration,and clinical applicability were evaluated.At the same time,data from the verification group were used to evaluate the prediction efficiency of the model.Results 2 368 patients underwent a one-time tonsillectomy and the incidence of adverse personality changes after surgery was approximately 17.06％(404/2 368).Model A had a small Akashi information criterion(AIC=1 257.736),and Model A was ultimately selected as the final model,with residence(OR=3.875),maintenance(OR=0.003),preoperative concomitant secretory otitis media(OR=22.483),preoperative white blood cells(OR=1.898),personality type(OR=1.985)were the risk factors for adverse behavior changes after tonsillectomy,Y=-1.946+1.355xResidence-5.766xMaintenance+3.113xPreoperative concomitant secretory otitis media+1.376xPreoperative white blood cells+0.685xPersonality type.When the specificity was 75.6％and the sensitivity was 56.7％,the maximum Youden index was 0.323,corresponding to a P value of 0.173.Based on P=1/(1+e-y),Y=-1.56 can be calculated.Based on a multivariate Logistic regression model,a column chart was constructed.In the modeling group:the AUC was 0.73(95％CI:0.69-0.76),and the Hosmer Limeshow goodness of fit test showed that x2=8.24,P=0.51;In the validation group,the AUC was 0.71(95％CI:0.66-0.77),and the Hosmer Limeshow goodness of fit test showed x2=5.65,P=0.77.The modeling and validation groups had good clinical applicability.Conclusions This model can provide a reference for clinician to evaluate the occurrence of adverse behavior after tonsillectomy in children before surgery.