基于血小板参数构建肺炎支原体肺炎患儿发生塑型性支气管炎的风险预测列线图模型

Nomogram Model for Predicting the Risk of Plastic Bronchitis in Children with Mycoplasma Pneumoniae Pneumonia Based on Platelet Parameter

目的基于血小板参数构建肺炎支原体肺炎(MPP)患儿发生塑型性支气管炎(PB)的风险预测列线图模型。方法选取2020年3月—2023年6月南通市海门区人民医院收治的188例MPP患儿为研究对象。查阅患儿电子病历,收集其年龄、性别、发热持续时间、异常呼吸音发生情况及入院时实验室检查指标。根据患儿是否发生PB将其分为PB组(n=21)和无PB组(n=167)。通过最小绝对收缩和选择算子(LASSO)回归、支持向量机(SVM)算法和随机森林算法筛选MPP患儿发生PB的风险因素,通过多因素Logistic回归分析探讨MPP患儿发生PB的独立影响因素。基于多因素Logistic回归分析结果构建MPP患儿发生PB的风险预测列线图模型,绘制ROC曲线、校准曲线、决策曲线以分别评估该列线图模型的区分度、准确性、临床有效性;使用DynNom包开发MPP患儿发生PB的风险预测列线图模型的网络计算器。结果PB组发热持续时间长于非PB组,中性粒细胞与淋巴细胞比值(NLR)、C反应蛋白(CRP)、D-二聚体(D-D)、纤维蛋白原(FIB)、丙氨酸氨基转移酶(ALT)、天冬氨酸氨基转移酶(AST)、血小板指数评分(PIS)高于非PB组,乳酸脱氢酶(LDH)、血红蛋白(Hb)、血小板计数(PLT)低于非PB组,平均血小板体积(MPV)、血小板体积分布宽度(PDW)大于非PB组(P<0.05)。LASSO回归、SVM算法、随机森林算法分析结果均显示,发热持续时间、NLR、CRP、FIB、ALT、PIS是MPP患儿发生PB的风险因素。多因素Logistic回归分析结果显示,发热持续时间、NLR、CRP、FIB、PIS是MPP患儿发生PB的独立影响因素(P<0.05)。基于多因素Logistic回归分析结果构建MPP患儿发生PB的风险预测列线图模型。ROC曲线分析结果显示,该列线图模型预测MPP患儿发生PB的AUC为0.971〔95%CI(0.936~0.990)〕;校准曲线分析结果显示,该列线图模型预测MPP患儿发生PB的概率与其实际概率之间具有良好的一致性;决策曲线分析结果显示,当阈值概率为0.021~0.996时,该列线图模型的净获益率>0。基于MPP患儿发生PB的风险预测列线图模型开发网络计算器(https://bpo-nomogram.shinyapps.io/DynNomapp/)。结论基于PIS及发热持续时间、NLR、CRP、FIB构建的MPP患儿发生PB的风险预测列线图模型的区分度、准确性较高,且具有一定临床有效性;此外,基于上述列线图模型开发的网络计算器简单实用。

Objective To construct the nomogram model for predicting the risk of plastic bronchitis(PB)in children with mycoplasma pneumoniae pneumonia(MPP)based on platelet parameter.Methods A total of 188 children with MPP admitted to Nantong Haimen People's Hospital from March 2020 to June 2023 were selected as the research objects.The electronic medical records of the children were reviewed,and their age,gender,duration of fever,occurrence of abnormal respiratory sounds and laboratory examination indexes at admission were collected.The children were divided into PB group(n=21)and non-PB group(n=167)according to whether PB occurred.The risk variables of PB in children with MPP were screened by least absolute shrinkage and selection operator(LASSO)regression,support vector machine(SVM)algorithm and random forest algorithms,and the independent influencing factors of PB in children with MPP were screened by multivariate Logistic regression analysis.Based on the results of multivariate Logistic regression analysis,the nomogram model for predicting the risk of PB in children with MPP was constructed.The ROC curve,calibration curve and decision curve were drawn to evaluate the discrimination,accuracy and clinical effectiveness of the nomogram model.The DynNom package was used to develop the network calculator of nomogram model for predicting the risk of PB in children with MPP.Results The duration of fever in the PB group was longer than that in the non-PB group,and the neutrophil-to-lymphocyte ratio(NLR),C-reactive protein(CRP),D-dimer(D-D),fibrinogen(FIB),alanine aminotransferase(ALT),aspartate aminotransferase(AST)and platelet index score(PIS)were higher than those in non-PB group,lactate dehydrogenase(LDH),hemoglobin(Hb)and platelet count(PLT)were lower than those in non-PB group,while mean platelet volume(MPV),platelet volume distribution width(PDW)were greater than those in the non-PB group(P<0.05).The results of LASSO regression,SVM algorithm and random forest algorithm analysis showed that duration of fever,NLR,CRP,FIB,ALT and PIS were risk factors for PB in children with MPP.Multivariate Logistic regression analysis showed that duration of fever,NLR,CRP,FIB and PIS were independent influencing factors of PB in children with MPP(P<0.05).The nomogram model for predicting the risk of PB in children with MPP was constructed based on the results of multivariate Logistic regression analysis.ROC curve analysis showed that the AUC of the nomogram model in predicting PB in children with MPP was 0.971[95%CI(0.936-0.990)].The calibration curve analysis results showed that there was a good consistency between the probability of PB in children with MPP predicted by the nomogram model and the actual probability of them.The decision curve analysis results showed that when the threshold probability was 0.021-0.996,the net benefit rate of the nomogram model was>0.The network calculator(https://bpo-nomogram.shinyapps.io/DynNomapp/)was developed based on the nomogram model for predicting the risk of PB in children with MPP.Conclusion The nomogram model for predicting the risk of PB in children with MPP which was constructed based on PIS,duration of fever,NLR,CRP and FIB has high discrimination and accuracy,and certain clinical effectiveness.In addition,the network calculator developed based on the above nomogram model is simple and practical.