Continuous Pulse Oximetry Analysis to Predict Outcomes in Young Infants With Bronchiolitis Presenting to the Emergency Department.

BACKGROUND

Predictive analytics using continuously measured vital signs from bedside monitors has improved outcomes in critically ill children. We investigated whether continuous pulse oximetry data collected during emergency department (ED) evaluation can identify infants ≤90 days old with bronchiolitis at low risk of respiratory deterioration.

METHODS

We conducted a single-center retrospective cohort study of infants ≤90 days old with bronchiolitis who presented to a tertiary-care pediatric ED from November 2021 to November 2025. Infants were identified using the ICD-10 code with manual eligibility confirmation. Heart rate and oxygen saturation features were generated from continuous pulse oximetry during the first 2 hours of ED evaluation. The primary outcome was respiratory deterioration within 24 hours of ED arrival, defined as initiation of high-flow nasal cannula, continuous/bilevel positive airway pressure, or invasive mechanical ventilation. Data were split into derivation and validation cohorts (70/30, stratified by outcome). An L2-regularized logistic regression model estimated the probability of respiratory deterioration, with a classification threshold selected to prioritize sensitivity.

RESULTS

Of 290 infants meeting clinical eligibility, 117 had sufficient continuous pulse oximetry data; 57 (49%) had respiratory deterioration. The primary model included 3 features: the median and interquartile range (IQR) of heart rate, and oxygen saturation IQR. In the validation cohort, the model showed moderate sensitivity (0.78, 95% CI: 0.58-0.95) and fair discrimination (AUROC=0.78, 95% CI: 0.62-0.93).

CONCLUSION

We derived and validated a model using data from continuous pulse oximetry to identify low-risk infants with bronchiolitis. With further refinement and integration into monitor design systems, this approach may enhance clinical utility of continuous pulse oximetry.