儿童耐多药结核病发生的相关影响因素及其预测模型的建立

目的 探究儿童耐多药结核病（MDR-TB）发生的相关影响因素及其预测模型的建立。方法 回顾性选取2010年3月1日至2022年2月28日昆明市第三人民医院结核病病儿，统计MDR-TB发生情况，比较MDR-TB病儿与非MDR-TB病儿临床资料，通过logistic多因素回归模型分析儿童MDR-TB发生的相关影响因素，采用R语言构建儿童MDR-TB发生的列线图预测模型，并进行一致性检验。结果 共纳入850例结核病病儿，其中初治MDR-TB病儿占5.65%，复治MDR-TB病儿占24.88%，总MDR-TB病儿占10.47%;logistic多因素回归模型分析显示，年龄、吸烟史、治疗情况、感染播散性、与结核病病人接触、合并人类免疫缺陷病毒（HIV）感染、合并乙型肝炎病毒（HBV）感染是儿童MDR-TB发生的独立危险因素，规范用药是独立保护因素（P<0.05）；根据影响因素构建儿童MDR-TB发生的列线图预测模型，通过Bootstrap自抽样法验证显示该预测模型预测值与实际观测值基本一致，一致性指数（C-index）为0.960，具有良好的区分度；预测模型ROC曲线的曲线下面积（AUC）...

Establishment of related influencing factors and prediction model of multidrug-resistant tuberculosis in children

Objective To explore the related influencing factors of multidrug-resistant tuberculosis (MDR-TB) in children and the es. tablishment of a predictive model.Methods Children with tuberculosis in Kunming Third People''s Hospital from March 1, 2010 toFebruary 28, 2022 were retrospectively selected, the incidence of MDR-TB was counted, and the clinical data of children with MDR-TB and those without MDR-TB were compared, the relevant influencing factors of the occurrence of MDR-TB in children by logistic multifactorial regression model were analyzed, a column-line graph prediction model of the occurrence of MDR-TB in children was con. structed by using the R language, and the consistency test was performed.Results A total of 850 children with tuberculosis were in. cluded in this study, of which 5.65% were initially treated with MDR-TB, 24.88% were retreated with MDR-TB, and 10.47% were in to. tal MDR-TB. Logistic multivariate regression model analysis showed that age, smoking history, treatment status, dissemination of infec.tion, contact with tuberculosis patients, HIV infection, and hepatitis B virus (HBV) infection were independent risk factors for the occur.rence of MDR-TB in children, and standardized medication was an independent protective factor (P<0.05); According to the influenc. ing factors, a nomogram prediction model for the occurrence of MDR-TB in children was constructed. The Bootstrap self-sampling meth.od showed that the prediction value of the prediction model was basically consistent with the actual observation value, and the C-index was 0.960, which had a good degree of discrimination; the prediction model ROC The area under the curve (AUC) of the curve was 0.9695%CI:(0.94,0.98)], indicating high predictive power.Conclusions Children''s age, smoking history, treatment, infection spread, con.tact with tuberculosis patients, combined HIV infection, combined HBV infection, etc. are all influencing factors of MDR-TB. The no. mogram prediction model based on the influencing factors can better predict the risk of MDR-TB in children with tuberculosis.