Epidemiological and clinical characteristics of bacterial co-detection in respiratory syncytial virus-positive children in Wenzhou, China, 2021 to 2023

Abstract

Background: Respiratory syncytial virus (RSV) is a leading cause of respiratory tract infections in young children, posing significant health challenges worldwide. This study aims to analyze the etiological, clinical, and imaging characteristics of RSV infection in children from Wenzhou, China, to better understand its epidemiological profile and provide insights for early diagnosis, monitoring, and clinical interventions.

Methods: This retrospective descriptive observational study included 1,063 RSV-positive pediatric inpatients (< 14 years) from the Second Affiliated Hospital of Wenzhou Medical University (December, 2021, and April, 2023). Missing data were addressed using multiple imputation. Patients were grouped by co-infection type (viral or bacterial), and differences in demographic, clinical, and imaging features were analyzed. Univariate and multivariate logistic regression identified risk factors for bacterial co-infection in RSV-positive patients. Predictive features were selected via the Boruta algorithm and modeled using XGBoost, with SHAP applied for interpretability. Pathogen distributions were also compared across age groups in both co-infection subtypes.

Results: Among 1,063 hospitalized RSV-positive children, RSV primarily affected those under one year of age, with a substantial proportion of cases (47.3%) showing co-infections. The most frequently detected viral co-pathogen was human rhinovirus (HRV, 42.2%), while Streptococcus pneumoniae (37.3%) was the most common bacterial pathogen. Some children were found to carry multiple pathogens simultaneously. Age-stratified radiographic analysis revealed that bronchitis was the predominant imaging finding across all age groups. Compared with the viral co-infection group, children with bacterial co-infections had longer hospital stays (6.4 vs. 5.4 days), higher white blood cell and neutrophil counts, and lower IgE levels (P < 0.05). Logistic regression analysis identified length of stay, prealbumin (PA), and immunoglobulin E (IgE) as independent predictors of bacterial co-infection. Feature selection using the Boruta algorithm, combined with XGBoost modeling and SHAP interpretation, further confirmed that length of stay, WBC count, neutrophil count, and IgE were the most important predictive variables. Age-specific pathogen analysis showed that influenza A and Streptococcus pneumoniae were predominantly found in children aged 3 to < 6 years (P < 0.05).

Conclusions: RSV infections in Wenzhou predominantly occur in infants under one year of age and frequently involve co-infections with viruses and bacteria, particularly rhinovirus and Streptococcus pneumoniae. Early identification of co-infections and tailored interventions are critical to improving patient outcomes.

Trial registration: Not applicable.

Keywords: Allergic histories; Co-infections; Pathogen analysis; Respiratory syncytial virus (RSV); Risk factors.

Fig. 1

Study flowchart depicting patient inclusion and exclusion criteria

Fig. 2

Feature importance analysis of predictors for RSV-associated bacterial co-infections in RSV patients using the Boruta algorithm. Box plots represent the importance distribution of each feature, ranked from low to high. Features are categorized by color: blue for shadow features (used as controls), red for unimportant variables, yellow for tentative variables, and green for important variables

Fig. 3

SHAP summary plot of predictive features for bacterial co-infections in RSV patients. SHAP scatter plots show the impact of individual feature values on model predictions. Each point represents a patient. Features are ordered by importance (top to bottom). Positive SHAP values indicate an increased predicted risk of bacterial infection, while negative values indicate a reduced risk