Abstract
OBJECTIVE: The study aimed to set up and validate a predictive nomogram for post-infectious bronchiolitis obliterans in severe pneumonia. METHODS: We retrospectively analyzed data of 228 patients diagnosed with severe pneumonia and constructed a prediction nomogram. The least absolute shrinkage and selection operator (LASSO) regression model was utilized to optimize the selection of features for the clinical characteristics of post-infectious bronchiolitis obliterans. Individual nomograms of bronchiolitis obliterans incorporating clinical factors were developed using the multivariate logistic model. The C-index, calibration plot, and decision curve analysis were used to verify the calibration, discrimination, and clinical utility. The bootstrapping method was used for the internal validation of the model. RESULTS: Predictors in the individualized predictive nomogram included age of patients (odds ratio [OR], 0.994; 95% confidence interval; [CI], 0.990-0.998), length of stay (OR, 1.043; 95% CI: 1.015-1.073), mechanical ventilation (OR, 1.865; 95% CI: 1.236-2.817), human adenoviral infection (OR, 1.671; 95%, CI: 1.201-2.326), and the level of interleukin (IL)-2 (OR, 0.947; 95% CI: 0.901-0.955). The model discriminated reasonably well, with a C-index of 0.907 (C-index, 0.888 and 0.926) with good calibration and internal validation, which was not statistically significant by the Hosmer-Lemeshow test (P = 0.5443). Decision curve analysis showed that nomograms were useful in clinical settings. CONCLUSION: In this study, a model was developed and presented as a nomogram with relatively good accuracy to help clinicians accurately and early diagnose post-infectious bronchiolitis obliterans in children with severe pneumonia.