Big data, internet-of-things and aRtificial intelligence to study the impact of personal Exposure to Air pollution on asTHma Exacerbations (BREATHE): Beta Test

Introduction: Asthma is a heterogeneous inflammatory disease, characterized by exacerbations, which minimization is one of the main treatment goals. Exacerbations could be triggered by many factors, including infections, smoke exposure, allergens and air pollution. Prognostic models combining risk factors have been proposed to predict the risk of exacerbations; however data on the individual exposure to air pollution in these models are limited. BREATHE aims at studying the effect of air pollution on asthma exacerbations. Patients will be provided with an air quality sensor, a sensor of activity and sleep, and a portable sensor of respiratory function. Clinical and functional data will be integrated in a dedicated mobile app for patients and a web application for clinicians. Data-driven predictive models empowered by advanced AI techniques will be used to identify factors predicting exacerbations. Before extending the assessment to the whole BREATHE cohort, we conducted a beta test aiming to evaluate platform, patients and clinicians needs, satisfaction and adherence to the program. Methods: For the beta testing, we recruited 8 patients with mild/moderate asthma and 8 pulmonologists working in the U.O.C. of Pneumologia of Padua University Hospital. Each clinician enrolled one patient using the dedicated application. Patients have been asked to use personal devices and insert in the mobile app: spirometry data, medication use, hours spent outdoors daily, Asthma Control Test (ACT) biweekly, number and characteristics of exacerbations. Both patients and clinician underwent systematic usability tests (tasks) during the enrollment visit using a scale from 1 (difficult) to 7 (easy) to describe the difficulty of the proposed steps and skills. After 28 days of testing, System Usability Scale (SUS) Score was administered to obtain feedback regarding usability, along with a questionnaire evaluating the amount of time spent completing tasks, user’s satisfaction and possible system bugs. Results: Among clinicians, the easiest task scored a median value of 7/7,the most difficult scored 6/7. Conversely, among patients, the easiest task scored a median value of 7/7, while the most difficult scored 5.5/7. Clinicians required a median enrollment time of 18 minutes for patients (range 13’-23’) and fully agreed on the user-friendliness of the platform. The time spent by the patients for app download, set-up and device synchronization was 4 minutes and 20 seconds (range 2’01-6’).During the 28 days patients recorded 62.05% of expected daily data (spirometry, medication use, hours spent outdoors), 6 out of the 8 ACT required. Two over 8 patients experienced an exacerbation, promptly recorded on the app. SUS Score for clinicians was 85 (77.5-100) while for patients was 58.75 (52.5-97), with air quality sensor being the most impacting factor. The time of use of FitBit on the total observation period was 92.85 (76.92–100)%.Conclusions: The beta testing demonstrates that the system is time-efficient, well-fitting within the constraints of clinical practice. Clinicians recorded positively usability and satisfaction, while patients brought to light some device issues which can be solved, thus increasing the adherence to data recording.