儿童慢性化脓性中耳炎鼓膜穿孔的风险预测模型构建

Construction of a risk prediction model for tympanic membrane perforation in chronic suppurative otitis media in children

目的探讨影响儿童慢性化脓性中耳炎(CSOM)发生鼓膜穿孔的因素,并以此构建风险预警模型。方法选取2021年1月—2022年12月收治在上海市东方医院吉安医院耳鼻喉科慢性化脓性中耳炎的患儿136例作为训练集,根据患儿有无发生鼓膜穿孔分为鼓膜穿孔组23例和非鼓膜穿孔组113例。于2022年5月—2023年2月纳入59例慢性化脓性中耳炎的患儿作为验证集。采用单因素及多因素Logistic回归分析慢性化脓性中耳炎患儿发生鼓膜穿孔的影响因素,应用R语言软件绘制相应的列线图,以此构建对应的风险预警模型。同时在模型组数据(内部验证)和验证组数据(外部验证)中采用受试者工作特征(ROC)曲线、校准曲线及Hosmer-Lemeshow(H-L)拟合优度检验来评估模型的预测性能;采用临床决策曲线来评估模型的临床适用性。结果多因素Logistic回归分析结果提示,病程3个月以上、急性化脓性中耳炎病史、反复上呼吸道感染、咽鼓管功能障碍、抗生素使用不合理均是影响慢性化脓性中耳炎的患儿发生鼓膜穿孔的因素。基于以上5个独立影响因素构建相应的风险预测模型。模型组的ROC曲线下面积(AUC)为0.873(95%CI=0.798~0.948,P<0.001),敏感度为95.70%、特异度为74.30%,H-L拟合优度检验结果显示χ^(2)=11.063,P=0.198,模型拟合情况良好;验证组的AUC为0.870(95%CI=0.763~0.977,P<0.001),敏感度为80.00%、特异度为85.70%,H-L拟合优度检验结果显示χ^(2)=2.048,P=0.957,模型拟合情况良好。校准曲线结果显示,模型预测概率与实际发生情况之间具有良好的一致性;临床决策曲线结果显示,模型具有较高的临床实用价值。结论慢性化脓性中耳炎患儿发生鼓膜穿孔受多种因素影响,构建的预测模型具有较好的临床应用价值。

Objective To explore the factors affecting the occurrence of tympanic membrane perforation in children with chronic suppurative otitis media(CSOM)and to construct a risk prediction model.Methods A total of 136 children with chronic suppurative otitis media admitted to the Department of Otorhinolaryngology,Ji'an Hospital of Shanghai Oriental Hospital from January 2021 to December 2022 were selected as the training set,they were divided into 23 cases in the tympanic membrane perforation group and 113 cases in the non-tympanic membrane perforation group according to the occurrence of tympanic membrane perforation or not.Fifty-nine children with chronic suppurative otitis media were included from May 2022 to February 2023 as the validation set.The influencing factors of tympanic membrane perforation in children with chronic suppurative otitis media were analysed by using unifactorial and multifactorial Logistic regression,and the corresponding column line graphs were plotted by applying the R language software,so as to construct the corresponding risk warning model.The predictive performance of the model was evaluated by using receiver operating characteristic(ROC)curves,calibration curves and Hosmer-Lemeshow(H-L)goodness-of-fit tests in the model group(internal validation)and validation group(external validation),and the clinical decision curve was used to evaluate the clinical applicability of the model.The clinical decision curve was used to assess the clinical applicability of the model.Results The results of multifactorial Logistic regression analyses suggested that disease duration of more than 3 months,history of acute suppurative otitis media,recurrent upper respiratory tract infections,Eustachian tube dysfunction,and irrational use of antibiotics were all factors influencing the occurrence of tympanic membrane perforation in children with chronic suppurative otitis media.A corresponding risk prediction model was constructed based on the above five independent influencing factors.In the model group,the area under the curve(AUC)was 0.873(95%CI=0.798~0.948,P<0.001),with a sensitivity of 95.70%and a specificity of 74.30%,and the results of the H-L goodness-of-fit test showed that the model fit was good,with χ^(2)=11.063,P=0.198.In the validation group,the area under the curve(AUC)was 0.870(95%CI=0.763~0.977,P<0.001),with a sensitivity of 80.00%and specificity of 85.70%,and the results of the H-L goodness-of-fit test showed that the model fit was good with χ^(2)=2.048,P=0.957.The results of the calibration curve showed good agreement between the predicted probability of the model and the actual occurrence,the results of the clinical decision curve showed that the model had a high clinical utility value.Conclusion The occurrence of tympanic membrane perforation in children with chronic suppurative otitis media is affected by various factors,and the constructed prediction model had good clinical application value.