Application of Machine Learning to Support Self-management of Asthma with mHealth

While there have been several efforts to usemHealth technologies to support asthma management, noneso far offer personalised algorithms that can provide real-time feedback and tailored advice to patients based on theirmonitoring. This work employed a publicly available mHealthdataset, the Asthma Mobile Health Study (AMHS), and ap-plied machine learning techniques to develop early warningalgorithms to enhance asthma self-management. The AMHSconsisted of longitudinal data from 5,875 patients, including13,614 weekly surveys and 75,795 daily surveys. We applied sev-eral well-known supervised learning algorithms (classification)to differentiate stable and unstable periods and found that bothlogistic regression and na ̈ıve Bayes-based classifiers providedhigh accuracy (AUC>0.87). We found features related to theuse of quick-relief puffs, night symptoms, frequency of dataentry, and day symptoms (in descending order of importance)as the most useful features to detect early evidence of lossof control. We found no additional value of using peak flowreadings to improve population level early warning algorithm