Journal of Asthma and Allergy | 2021
Predicting Environmental Allergies from Real World Data Through a Mobile Study Platform
Abstract
1Sharecare Inc., Atlanta, GA, USA; 2Stanford Medicine, Palo Alto, CA, USA; 3Clinical Excellence Research Center, Stanford, CA, USA Introduction Environmental allergies are extremely disruptive to the daily life of many globally. Nearly 15 million clinic visits, 3.5 million days of missed work, and $24.8 billion in direct costs are incurred annually due to allergic rhinitis in the United States alone. In treatment protocols, allergen avoidance is the primary clinical recommendation; however, few interventions have been broadly effective. If patients and clinicians can better predict the risk of symptom flares, preventative steps can be taken to mitigate downstream consequences. The heterogeneity of triggers and symptom patterns across individuals and geographies, however, has historically precluded collective learning and predictive efforts have been only modestly successful. Advances in machine learning and smartphone-based data collection can help clarify these relationships and provide personalized actionable intelligence. We launched a mobile research platform designed to gather real world subjective symptom data and objective sensor data, linked to select external datasets. Smartphone sensor data including physical activity (steps/day) and geolocation (linked to external ambient pollen data) were collected, and participants logged their allergy symptoms in an e-diary using the app (Figure 1). These data were used to develop and train a machine learning algorithm to predict the emergence and severity of symptoms related to allergic rhinitis.