We developed and internally validated a nomogram for the early diagnosis of severe MPP associated with PB. Fever duration, NP, LDH, PCT, IL-6, and pleural effusion were included in the nomogram. The proposed nomogram is an easy-to-use tool for diagnosing PB. Internal validation showed that the nomogram had good discriminatory power and calibration. The C-index was 0.900 (0.866–0.934), indicating that the nomogram could be used in large cohorts in clinical practice. Decision curve analysis(DCA) indicated that at a threshold probability of ≥ 14%, the nomogram was more clinically useful in diagnosing PB.
P-trend analysis showed that the OR increased with the increase in PCT, indicating that high PCT increased the risk of developing PB. PCT was clinically useful, underscoring the need to incorporate this variable into this diagnostic model. The contribution of pleural effusion to the nomogram was lower; however, in clinical practice, imaging showed that pleural effusion had a higher predictive ability and clinical value for severe MPP associated with PB, suggesting that pleural effusion could be used as an independent clinical feature in the proposed model.
Fever duration was clinically useful and was longer in the PB group than in the control group (mean: 11.50 [0–32 days] vs. 8.25 [0–18 days]), suggesting that patients with severe MPP with a long fever duration (more than 11 days) are more likely to develop PB.
Bronchial casts may be caused by inflammatory cell infiltration and inflammatory mediators, resulting in congestion, edema, necrosis of the tracheal mucosa, and airway obstruction[13]. IL-6 levels are positively correlated with the severity of MPP[14]. Increased IL-6 levels in the bronchoalveolar lavage fluid in the acute phase of MPP indicate a local inflammatory response in the lungs [15–17]. IL-6 aggravates tissue inflammation by increasing the production of other immune factors [18]. In this study, IL-6 level was higher in the PB group than in the control group (mean: 68.43 [2.86–806.40] pg/ml vs. 20.89 [1.00–318.90] pg/ml), indicating that children with severe MPP and IL-6 levels higher than 68 pg/ml should be monitored for PB.
PB can be secondary to asthma, bronchitis, atelectasis, and pulmonary fibrosis[19] and may develop in healthy children; however, patients with allergies or asthma are at higher risk[20, 21]. PB is caused by influenza viruses, MP, and adenoviruses[22].
Bronchial casts are either inflammatory (type 1) or acellular (type 2)[23]. Type 1 casts are characterized by dense fibrin and the infiltration of eosinophils and neutrophils. Type 2 casts are characterized by the presence of mucin with little or no cellular infiltration. Type 2 PB is secondary to congenital cyanotic heart disease, especially after Fontan operation. Bronchial casts may be an unusual response of the pulmonary epithelium to increased pulmonary venous pressure and lymph exudation caused by surgical injury or lymphatic malformation[24]. Most reported cases of PB in China are type 1. The incidence of PB in children who underwent Fontan operation is 4–14%.
A total of 196 cases of PB were diagnosed in this study. Of these, 49 cases were excluded because the etiological agent was not identified (13, 6.60%), pneumonia was caused by pathogens other than MP (7, 3.60%), and MP was associated with other pathogens (29, 14.80%). Among the 196 patients, 147 patients (75%) were infected with MP, and 21 patients (10.71%) were infected with MP and EB virus, indicating that MP, and the mixed infections of MP and EB virus were common in our cohort, contrary to previous studies[25–26].
Using univariate logistic regression analysis and p-trend analysis, ten clinical characteristics more useful for diagnosing PB were selected. An AUC of 0.70–0.79 is considered clinically acceptable, and an AUC ≥ 0.8 is excellent[27, 28]. Therefore, an AUC ≥ 0.70 was chosen to identify diagnostic predictors of PB, and a diagnostic model with six independent clinical characteristics was developed.
The discriminative ability of the two models was assessed. Model 1 included fever duration, NP, and LDH, and model 2 included fever duration, NP, LDH, PCT, IL-6, and pleural effusion. The C-index of models 1 and model 2 was 0.854 (0.816–0.893) and 0.900 (0.866–0.934), respectively. The C-index of model 2 was significantly higher than that of model 1. Using ROC analysis, the difference in AUC between the two models was 0.046 (0.022–0.070) (p < 0.001), suggesting that PCT, IL-6, and pleural effusion were integrated into the nomogram. The net reclassification index(NRI) was 0.1803 (95% CI: 0.0039–0.3567; p = 0.0451), and the integrated discrimination improvement (IDI) was 0.1755 (95% CI: 0.1004–0.2505; p < 0.001). The accuracy of model 2 increased by 17.55%.
This study has limitations. First, the retrospective design may lead to bias. Second, only data from one hospital were analyzed. Therefore, prospective studies with larger cohorts are necessary to validate our model.