Background: The individualized PREdiction of DIsease Control using digital sensor Technology (iPREDICT) program was developed for asthma management using digital technology. Devices were integrated into daily lives of patients to establish a predictive model of asthma control by measuring changes from baseline health status with minimal device burden.
Objectives: To establish baseline disease characteristics of the study participants, detect changes from baseline associated with asthma events, and evaluate algorithms capable of identifying triggers and predicting asthma control changes from baseline data. Patient experience and compliance with the devices were also explored.
Design: This was a multicenter, observational, 24-week, proof-of-concept study conducted in the United States.
Methods: Patients (⩾12 years) with severe, uncontrolled asthma engaged with a spirometer, vital sign monitor, sleep monitor, connected inhaler devices, and two mobile applications with embedded patient-reported outcome (PRO) questionnaires. Prospective data were linked to data from electronic health records and transmitted to a secure platform to develop predictive algorithms. The primary endpoint was an asthma event: symptom worsening logged by patients (PRO); peak expiratory flow (PEF) < 65% or forced expiratory volume in 1 s < 80%; increased short-acting β2-agonist (SABA) use (>8 puffs/24 h or >4 puffs/day/48 h). For each endpoint, predictive models were constructed at population, subgroup, and individual levels.
Results: Overall, 108 patients were selected: 66 (61.1%) completed and 42 (38.9%) were excluded for failure to respond/missing data. Predictive accuracy depended on endpoint selection. Population-level models achieved low accuracy in predicting endpoints such as PEF < 65%. Subgroups related to specific allergies, asthma triggers, asthma types, and exacerbation treatments demonstrated high accuracy, with the most accurate, predictive endpoint being >4 SABA puffs/day/48 h. Individual models, constructed for patients with high endpoint overlap, exhibited significant predictive accuracy, especially for PEF < 65% and >4 SABA puffs/day/48 h.
Conclusion: This multidimensional dataset enabled population-, subgroup-, and individual-level analyses, providing proof-of-concept evidence for development of predictive models of fluctuating asthma control.
Keywords: algorithm; asthma control; digital tool; mobile application; precision medicine; self-management; severe asthma.