Abstract
BACKGROUND: Human rhinovirus (HRV) is a leading cause of acute respiratory infections and hospitalization in children. However, risk factors for progression to bronchitis vs. pneumonia remain incompletely characterized. OBJECTIVE: This study aimed to identify and distinguish independent predictors for these outcomes using a machine learning approach. METHODS: A retrospective cohort study was conducted among hospitalized children with HRV infection at the First Affiliated Hospital of Ningbo University. A two-stage feature selection method (Random Forest and LASSO) was used, followed by multinomial logistic regression to quantify associations with progression to bronchitis or pneumonia. RESULTS: Of 1,125 children, 29% progressed to bronchitis and 59% to pneumonia. Multinomial logistic regression revealed distinct risk profiles. Progression to bronchitis was strongly associated with lower airway obstruction, including dyspnea (OR: 4.35, 95% CI: 2.35-8.06), wheezing on auscultation (OR: 2.98, 95% CI: 1.89-4.69), and cough (OR: 2.61, 95% CI: 1.52-4.49). Progression to pneumonia was uniquely linked to prior antibiotic use (OR: 2.09, 95% CI: 1.34-3.25), viral co-infection (OR: 1.97, 95% CI: 1.25-3.10), and coagulation abnormalities (OR: 1.04, 95% CI: 1.00-1.08). Elevated IgM was a common risk factor for both, while older age was protective against both (OR: 0.89, 95% CI: 0.83-0.96). CONCLUSION: Bronchitis progression is primarily associated with airway obstruction, whereas pneumonia is linked to more complex clinical scenarios, including prior medication, co-infections, and coagulopathy. These findings can improve early risk stratification and guide targeted interventions.