Development and validation of a nomogram to predict plastic bronchitis in children with refractory Mycoplasma pneumoniae pneumonia

Abstract

Background: Early identification of plastic bronchitis (PB) is of great importance and may aid in delivering appropriate treatment. This study aimed to develop and validate a nomogram for predicting PB in patients with refractory Mycoplasma pneumoniae pneumonia (RMPP).

Methods: A total of 547 consecutive children with RMPP who underwent fiberoptic bronchoscopy (FOB) intervention from January 2016 to June 2021 were enrolled in this study. Subsequently, 374 RMPP children (PB: 137, without PB: 237) from January 2016 to December 2019 were assigned to the development dataset to construct the nomogram to predict PB and 173 RMPP children from January 2020 to June 2021 were assigned to the validation dataset. The clinical, laboratory and radiological findings were screened using Least Absolute Shrinkage and Selection Operator (LASSO) regression and logistic regression was applied to construct a nomogram. The performance of the nomogram was evaluated by discrimination, calibration and clinical utility. Comparsion of ROC analysis and decision curve analysis (DCA) between nomogram and other models was performed to evaluate the discrimination ability and clinical utility.

Results: The development dataset included 374 patients with a mean age of 6.6 years and 185(49.5%) were men. The validation dataset included 173 patients and the mean age of the dataset was 6.7 years and 86 (49.7%) were men. From 26 potential predictors, LASSO regression identified 6 variables as significant predictive factors to construct the nomogram for predicting PB, including peak body temperature, neutrophil ratio (N%), platelet counts (PLT), interleukin-6 (IL-6), actic dehydrogenase (LDH) and pulmonary atelectasis. The nomogram showed good discrimination, calibration and clinical value. The mean AUC of the nomogram was 0.813 (95% CI 0.769-0.856) in the development dataset and 0.895 (95% CI 0.847-0.943) in the validation dataset. Through calibration plot and Hosmer-Lemeshow test, the predicted probability had a good consistency with actual probability both in the development dataset (P = 0.217) and validation dataset (P = 0.183), and DCA showed good clinical utility. ROC analysis indicated that the nomogram showed better discrimination ability compared with model of peak body temperature + pulmonary atelactsis and another model of N% + PLT + IL-6 + LDH, both in development dataset (AUC 0.813 vs 0.757 vs 0.754) and validation dataset (AUC 0.895 vs 0.789 vs 0.842).

Conclusions: In this study, a nomogram for predicting PB among RMPP patients was developed and validated. It performs well on discrimination ability, calibration ability and clinical value and may have the potential for the early identification of PB that will help physicians take timely intervention and appropriate management.

Keywords: LASSO; Nomogram; Plastic bronchitis; Refractory Mycoplasma pneumoniae pneumonia; Risk factor.

Fig. 1

Study flow. CAP community-acquired pneumonia, CHD congenital heart disease, MPP Mycoplasma pneumoniae pneumonia, PB plastic bronchitis

Fig. 2

Variable selection using least absolute shrinkage and selection operator (LASSO) logistic regression. A LASSO coefficient profiles of the 26 variables. With larger penalties, the coefficients of an increasing number of variable are compressed; finally, most of the variable coefficients are compressed to zero. B The best penalty coefficient lambda was selected using a tenfold cross-validation and minimization criterion. By verifying the optimal parameter (lambda) in the LASSO model, the binomial deviance curve was plotted versus log(lambda) and dotted vertical lines were drawn based on 1 standard error criteria. 6 variables with nonzero coefficients were selected by optimal lambda

Fig. 3

Nomogram to predict PB among RMPP children was constructed based on 6 independent predictors. Mark the value of these included factors on the corresponding axis. Draw a vertical line from the value to the top lines and get corresponding points. Then, sum the points from each variable value. Locate the sum on the total points scale and project it vertically on the bottom axis to obtain a PB risk

Fig. 4

The ROC curves of the nomogram from the development cohort (A) and the validation cohort (B). ROC: receiver operating characteristics

Fig. 5

Calibration plot of PB risk nomogram in the development cohort (A) and validation cohort (B). The ideal outcome (dashed line), the observed outcome (fine dashed line), and the bias-corrected outcome (solid line) are depicted

Fig. 6

Decision curve analysis for the PB risk nomogram. The y-axis measured the net benefit. The black solid line represented the assumption that all patients had no PB. The gray solid line represented the assumption that all patients had PB. A From the development cohort and B from the validation cohort