A27-08 Early Warning for Pediatric Wheeze: Developing a Predictive Health Score From Continuous Vital Sign Monitoring

Abstract Rationale Pediatric asthma exacerbations can escalate rapidly, yet current monitoring relies primarily on episodic symptom reporting rather than continuous objective data. Wheezing represents an early marker of airway narrowing and is often accompanied by subtle fluctuations in heart rate (HR) and respiratory rate (RR) that precede overt symptoms. While individual vital sign measurements provide limited predictive value, patterns of vital sign variability over time may signal physiologic instability before obvious clinical deterioration. This study evaluated whether a health score derived from continuous vital sign variability patterns could predict next-day wheeze events in pediatric asthma patients using real-world remote monitoring data. Methods We conducted a retrospective proof-of-concept analysis on six pediatric asthma patients (ages 3-17) in a remote monitoring pilot program at Cedars-Sinai Guerin Children’s. Patients wore a regulatory-approved wearable stethoscope to measure HR, RR, and detect wheeze continuously. Daily mean and variability of HR and RR were calculated using a rolling 30-day look-back window. Days were classified as “wheeze” or “no wheeze” based on recorded wheezing episodes. A model was derived using a linear discriminant approach that maximizes separability between the vital sign variability of “wheeze” and “no wheeze” days, producing a score that predicts risk of wheezing. Results 30-day look-back model achieved sensitivity of 64.0%, specificity of 70.64%, and accuracy of 70.12% for next-day wheeze prediction. Prediction performance improved with longer look-back windows: from 2 to 30 days, specificity increased from 24.1% to 70.64% and accuracy from 28.6% to 70.12% but sensitivity declined from 83.8% to 64.0%. The best trade-off was observed with a 26-day window (sensitivity 69.0%, specificity 66.1%, accuracy 66.4%), suggesting approximately one month of baseline data is needed for individualized risk prediction. Conclusion This proof-of-concept study demonstrates that a continuous vital sign variability derived score can predict next-day wheeze events, early markers of impending asthma exacerbations, in pediatric asthma patients with moderate accuracy, even with a small cohort. The need for extended observation periods (26-30 days) suggests the model captures individualized baseline physiology and deviations that precede clinical symptoms. Early wheeze detection could enable preemptive therapeutic escalation before progression to severe exacerbation requiring ED visits, systemic corticosteroids, or hospitalization. These results support the feasibility of wearable-based early warning systems for pediatric asthma. Future validation in larger cohorts could enable personalized risk stratification, timely therapeutic adjustments, and reduced asthma-related hospitalizations. Continuous vital sign monitoring may transform pediatric asthma management from reactive symptom treatment to proactive risk mitigation. This abstract is funded by: Cedars-Sinai Medical Center