Risk Factor Analysis And Prediction Model Construction Of Oral Mucositis During Chemotherapy In Children With Malignant Tumors

Objective: To explore and analyze the main risk factors for oral mucositis(OM)in children with malignant tumors during chemotherapy,and use machine learning algorithms to construct a risk prediction model for oral mucositis,which provides a reference for preventing and reducing the occurrence of oral mucositis,and lays a foundation for the subsequent establishment of OM intervention programs during chemotherapy for children with malignant tumors.Methods:(1)A computerized search in domestic and international databases for studies on risk factors related to OM during chemotherapy in children with malignant tumors was conducted to evaluate the quality and extract information from the included literature,and the current situation and risk factors were studied on the basis of Metaanalysis.(2)On the basis of the risk factors derived from Meta-analysis,we combined theoretical knowledge and the characteristics of electronic medical records to expand the investigation indexes of risk factors,and collected medical records of patients diagnosed with pediatric malignant tumors in a tertiary hospital in Sichuan Province from January 2019 to March 2022,and then conducted single factor analysis and multiple stepwise regression analysis by SPSS26.0 to derive risk factors.(3)Six machine learning algorithms,including Logistic Regression,Support Vector Machine,Decision Tree,Random Forest,XGBoost and Light GBM,were selected for the construction of oral mucositis prediction models based on 76 indicators,such as gender,age and body mass index,and their classification prediction performance was explored.Results:(1)Meta-analysis results: The incidence of oral mucositis at least once during chemotherapy in children with malignant tumors was 48%,and the risk factors included in the meta-analysis were age ≤ 3 years,body weight < standard body weight,methotrexate chemotherapy,chemotherapy time > two weeks,antibiotic time > 1 week,hospital stay > 2 weeks,increased AST/ALT ratio and increased creatinine.(2)Results of univariate analysis and multiple stepwise regression analysis: Among the subjects of this study,the incidence of oral mucositis after chemotherapy was 15.4%(131/852);Prolonged antibiotic use(OR=1.046,95%CI: 1.028～1.064)and increased urea/creatinine values(OR=1.003,95%CI: 1.001～1.004)were independent risk factors for oral mucositis in pediatric tumor patients,and compared with hematologic malignancies,children with solid tumors(OR=0.1,95% CI: 0.024～0.41)had a lower risk of OM.(3)Predication performance of Risk Predication Model: the accuracy of Logistic Regression,Support Vector Machine,Decision Tree,Random Forest,XGBoost,and Light GBM in diagnosing oral mucositis in children with malignant tumors was81%,83%,79%,82%,72%,83%,and AUC was 0.52,0.5,0.51,0.54,0.68 and 0.5,respectively,with the XGBoost model having the best predictive performance among the six models.Conclusion: The risk prediction model of oral mucositis in children with malignant tumors based on machine learning algorithm has high predictive value,and the classification ability of XGBoost model is better than that of traditional logistic regression model,and the results of this study will provide a reference for pediatric clinical medical staff to better carry out the assessment,prevention and intervention of oral mucositis in children with malignant tumors.