个体化预测儿童肺炎支原体肺炎支气管镜下灌洗次数的模型建立与验证

Model establishment and validation of individualized prediction of the number of bronchoalveolar lavage for Mycoplasma pneumoniae pneumonia in children

目的:建立预测儿童肺炎支原体肺炎(MPP)行支气管肺泡灌洗术(BAL)次数的Nomogram模型,并分析模型的临床价值。方法:对2023年1月至2024年1月就诊于我院的240例行BAL治疗的MPP患儿的临床资料进行回顾性收集。根据随机数字表法,将所有患儿分为训练集和验证集(比例7∶3)。根据患儿行BAL治疗的次数将训练集患儿分为单次组和多次组,基于单因素分析和多因素Logistic回归分析筛选出独立危险因素,以此为基础构建Nomogram模型。模型的准确度采用校准曲线和受试者工作特征(ROC)曲线评价,临床适用范围和安全性采用临床决策曲线分析评估。结果:单因素分析显示,单次组和多次组的BAL前热程、合并胸腔积液、肺外并发症、乳酸脱氢酶(LDH)及C反应蛋白(CRP)相比较,差异有统计学意义(P<0.05)。多因素分析显示,BAL前热程≥10 d、合并胸腔积液、有肺外并发症、LDH≥462 U·L^(-1)及CRP≥31.4 mg·L^(-1)是MPP患儿行多次BAL的独立危险因素(P<0.05)。基于上述危险因素构建的Nomogram模型的ROC曲线分析显示,训练集和验证集的曲线下面积(AUC)分别为0.820、0.843(P<0.05)。校正曲线分析显示,模型在训练集和验证集中拟合度较高,一致性较好。临床决策曲线分析显示,该模型在预测MPP患儿行BAL次数的方面,具有广泛的临床使用范围和较高的安全性。结论:MPP患儿多次行BAL治疗与BAL前热程、合并胸腔积液、肺外并发症及LDH、CRP水平有关。基于上述因素构建的Nomogram预测模型准确度高且临床适用范围广,为临床医生提前识别出MPP患儿多次行BAL的风险提供理论支持。

Objective:To develop a Nomogram model to predict the number of bronchoalveolar lavage(BAL)performed for Mycoplasma pneumoniae pneumoniae(MPP)in children and to analyze the clinical value of the model.Methods:Clinical data of 240 children with MPP who attended our hospital for BAL treatment from January 2023 to January 2024 were collected retrospectively.All the children were divided into training and validation sets(ratio 7∶3)according to the random number table method.The children in training set were divided into single and multiple groups according to the number of times they underwent BAL treatment.Independent risk factors were screened based on univariate and multivariate Logistic regression analyses,and Nomogram model was constructed on this basis.The accuracy of the model was evaluated using calibration curve and receiver operating characteristic(ROC)curve,and the clinical applicability and safety were assessed using clinical decision curve analysis.Results:Univariate analysis showed a statistically significant difference when comparing the pre-BAL thermal course,combined pleural effusion,extrapulmonary complications,lactate dehydrogenase(LDH)and C-reactive protein(CRP)between the single and multiple groups(P<0.05).Multivariate analysis showed that pre-BAL thermal course≥10 d,combined pleural effusion,presence of extrapulmonary complications,LDH≥462 U·L^(-1)and CRP≥31.4 mg·L^(-1)were independent risk factors for multiple BALs in children with MPP(P<0.05).The ROC curve analysis of the Nomogram model constructed based on the above risk factors showed that the area under the curve(AUC)in the training set and validation set were 0.820 and 0.843(P<0.05),respectively.Calibration curve analysis showed that the model fitted well in the training and validation sets with good agreement.Clinical decision curve analysis showed that the model had a wide range of clinical use and high safety in predicting the number of BALs performed in children with MPP.Conclusion:Multiple BALs in children with MPP were associated with pre-BAL thermal course,combined pleural effusion,extrapulmonary complications,LDH and CRP levels.The Nomogram prediction model constructed based on the above factors is highly accurate and has a wide range of clinical applicability,providing theoretical support for clinicians to identify the risk of multiple BALs in children with MPP in advance.