Development and validation of a nomogram for predicting necrotizing pneumonia in children with refractory Mycoplasma pneumoniae pneumonia

Background The early prediction of pulmonary necrosis in children with severe pneumonia improves patient prognosis and prevents complications. The aim of this study was to establish a linear model for predicting necrotizing pneumonia (NP) caused by Mycoplasma pneumoniae (MP) infection and to investigate the risk factors for lung necrosis in children with refractory Mycoplasma pneumoniae pneumonia (RMPP). Methods A total of 536 children with RMPP were enrolled, including 95 with NP and 441 with nonnecrotizing pneumonia (NNP). A prediction model was built on 375 cases and validated on 161 cases, which were divided by random sampling in R software. Multivariate logistic regression was performed to determine optimal predictors and to establish a nomogram for predicting NP. The performance of the nomogram was evaluated by the area under the characteristic curve (AUC), calibration ability and decision curve analysis (DCA). Results There were 315 (84.0%) NNP patients and 60 (16.0%) NP patients in the training group (n = 375) and 126 (78.3%) NNP patients and 35 NP patients (21.7%) in the validation group (n = 161). Multivariate logistic regression analysis identified 4 independent predictors that were used to construct a nomogram for predicting NP in children with RMPP, namely, fever duration (OR = 1.475; 95% CI 1.296–1.678; P  < 0.001), WBC count (OR = 1.149; 95% CI 1.073–1.231; P  < 0.001), IL-6 concentration (OR = 1.007; 95% CI 1.002–1.013; P  = 0.007) and D-dimer concentration (OR = 1.361; 95% CI 1.121–1.652; P =  0.002). The area under the curve (AUC) of the nomogram was 0.899 (95% CI, 0.850–0.947) in the training set and 0.920 (95% CI, 0.874–0.966) in the validation set, indicating a good fit. The calibration plot and Hosmer‒Lemeshow test indicated that the predicted probability had good consistency with the actual probability in the training ( P  = 0.439) and validation ( P  = 0.526) groups. The DCA curve demonstrated a significantly better net fit in the model. Conclusions We developed and validated a nomogram model for predicting RMPP-associated NP in its early clinical stages using four risk factors. This four-risk factor model may assist physicians in predicting NP induced by RMPP.