肺炎支原体肺炎患儿并发大叶性肺炎的影响因素分析及列线图预测模型构建

Influencing factors analysis of children with Mycoplasma pneumonia complicated with lobar pneumonia and the construction of Nomogram prediction model

目的 分析肺炎支原体肺炎(MPP)患儿并发大叶性肺炎(LP)的影响因素并构建Nomogram预测模型。方法 回顾性分析2020年2月至2022年8月就诊于安徽省六安市人民医院的280例MPP患儿的临床资料,根据是否并发LP,分为LP组(173例)和非LP组(107例)。另收集同期就诊于安徽省六安市人民医院的132例MPP患儿的临床资料用于验证模型。绘制受试者工作特征(ROC)曲线分析各因素的最佳截断值;采用多因素Logistic回归模型分析MPP患儿并发LP的危险因素;构建预测MPP患儿并发LP的Nomogram模型,Nomogram模型的内部验证采用校正曲线验证,Nomogram模型的预测效能采用决策曲线进行评估。结果 LP组年龄、发热时间、发病至应用大环内酯类(MAs)药物时间、有心动过速的比例、有胸腔积液的比例、有肺外并发症的比例、红细胞沉降率(ESR)、淋巴细胞(LYM)百分比、中性粒细胞(NEU)百分比及乳酸脱氢酶(LDH)、C反应蛋白(CRP)水平均高于非LP组(P<0.05)。ROC曲线分析结果显示年龄、发热时间、发病至MAs药物应用时间、LYM百分比、NEU百分比、LDH、ESR、CRP判断MPP患儿并发LP的曲线下面积(AUC)分别为0.868、0.886、0.801、0.662、0.747、0.702、0.868、0.892;最佳截断值分别为4岁、7 d、3 d、29.46%、53.72%、348.1 U/L、34 mm/h、27.85 mg/L。多因素Logistic回归分析结果显示年龄>4岁、发热时间>7 d、发病至MAs药物应用时间>3 d、有胸腔积液、LDH>348.1 U/L、CRP>27.85 mg/L是MPP患儿并发LP的独立危险因素(P<0.05)。内部验证结果显示,Nomogram模型预测MPP患儿并发LP的C-index为0.884(95%CI:0.753~0.917),观测值与预测值间一致性良好。Nomogram模型预测MPP患儿并发LP的阈值>0.17,Nomogram模型提供显著的临床净收益。另以132例患儿对模型进行验证,ROC曲线分析结果显示,该模型的AUC为0.854(95%CI:0.782~0.909),校准一致性较为良好。结论 基于年龄、发热时间、发病至MAs药物应用时间、胸腔积液、LDH、CRP构建的预测MPP患儿并发LP的Nomogram模型对于MPP患儿并发LP有较好的预测价值,可用于识别并发LP的高危患儿。

Objective To analyze the influencing factors of children with mycoplasma pneumoniae pneumonia(MPP)complicated with lobar pneumonia(LP)and construct a Nomogram prediction model.Methods Clinical data of 280 MPP children attending Lu′an People′s Hospital from February 2020 to August 2022 were collected retrospectively,who were divided into LP group(n=173)and non-LP group(n=107)according to whether they were complicated with LP or not.The clinical data of another 132 children with MPP attending Lu′an People′s Hospital during the same period were collected for model validation.The best cut-off values of each factor were analyzed by subject operating characteristic(ROC)curves;multivariate Logistic regression model was used to analyze the independent risk factors influencing the concurrent LP in children with MPP;Nomogram models predicting concurrent LP in children with MPP were constructed,and the internal validation of Nomogram models was validated by calibration curves,and the predictive power of the Nomogram model was assessed with a decision curve.Results Age,duration of fever,time from onset to macrolide antibiotics(MAs)drug application,proportion of tachycardia,proportion of pleural effusion,proportion of extra-pulmonary complications,erythrocyte sedimentation rate(ESR)and LYM,NEU,lactate dehydrogenase(LDH),C-reactive protein(CRP)levels in the LP group were higher than those in the non-LP group(P<0.05).The results of ROC curve analysis showed that the AUC values of age,duration of fever,time from onset to MAs drug application,LYM,NEU,LDH,ESR and CRP were 0.868,0.886,0.801,0.662,0.747,0.702,0.868,and 0.892,respectively;the optimal cutoff values were 4 years,7 days,3 days,29.46%,53.72%,348.1 U/L,34 mm/h and 27.85 mg/L,respectively.Multivariate Logistic regression results showed that age>4 years,duration of fever>7 days,time from onset to MAs drug application>3 days,presence of pleural effusion,LDH>348.1 U/L and CRP>27.85 mg/L were independent risk factors for MPP children with LP.The internal validation results showed that the C-index of the Nomogram model for predicting concurrent LP in children with MPP was 0.884(95%CI:0.753-0.917),with good agreement between the observed and predicted values.The threshold for the Nomogram model to predict concurrent LP in MPP children was>0.17,and the Nomogram model provided significant net clinical benefit.The model was also validated with 132 children,and the ROC curve analysis results showed a relatively good calibration agreement with an AUC of 0.854(95%CI:0.782-0.909).Conclusion Nomogram model constructed based on age,duration of fever,time from onset to drug application of MAs,pleural effusion,LDH and CRP for predicting concurrent LP in MPP children has good predictive value for concurrent LP in children with MPP,and it can be used to identify children at high risk of concurrent LP.