Abstract
INTRODUCTION: Bronchiolitis, a viral lower respiratory tract infection, is the leading cause of hospitalisation for infants, with healthcare utilisation highest among young infants (aged ≤90 days). Clinical models to predict respiratory deterioration in infants with bronchiolitis have been developed for a broad age group that includes children up to 2 years old, not focusing specifically on young infants. These models have also been limited by exclusion of viral aetiology and by use of vital signs measured at a single time point during clinical evaluation, overlooking the variable and dynamic course of bronchiolitis. This study aims to combine clinical history and examination factors with viral aetiology (primary aim) and continuous physiological data from bedside monitors (secondary aim) to develop accurate prediction models to identify young infants at low risk of respiratory deterioration and enable safe discharge to home. METHODS AND ANALYSIS: We are conducting a single-centre, prospective cohort study of young infants with bronchiolitis presenting to the paediatric emergency department (ED) of a tertiary care children's hospital. Enrolment began in November 2021 and will end in April 2027. Young infants with a clinical diagnosis of bronchiolitis are included. Infants with hospitalisation within 1 week of the index ED visit or baseline (home) use of supplemental oxygen or respiratory support are excluded. We are collecting clinical, laboratory, imaging and treatment data from the ED and hospitalisation, if admitted. All patients have a standardised initial examination as well as a repeat examination at 2-4 hours if still physically located in the ED. We are also obtaining swabs from the anterior nares and the nasopharynx. Continuous physiological data are collected through pulse oximetry and 5-lead ECG. We will contact parents or guardians at 7, 14 and 21 days for follow-up of symptom resolution and unplanned visits to a healthcare provider. The primary outcome is the development of respiratory deterioration within 24 hours of ED presentation. Respiratory deterioration is defined as new use of non-invasive or invasive respiratory support within 24 hours of ED arrival. For the primary aim, regression and recursive partitioning techniques will create the low-risk model. For the secondary aim, we will use machine learning models such as Lasso regression and support vector machines. ETHICS AND DISSEMINATION: Ethics approval was obtained through the Columbia University Institutional Review Board (IRB-AAAT8528). Written informed consent will be obtained for all patients. Results will be disseminated through academic conferences, peer-reviewed publications and appropriate free open-access medical education.