Predicting Lung Healthiness Risk Scores to Identify Probability of an Asthma Attack

For asthma, monitoring of symptoms and progression of the disease while avoiding triggers and minimizing the frequency of attacks are the main objectives of care. Tracking and analyzing lung healthiness, symptoms, medications, level of interference with daily life activities provides a hint of the level of severity of asthma, and that is indeed a strong predictor of subsequent and future exacerbations. In a prior effort, we employed Text Classification method to classify asthma severity. However, asthma attack prediction depends on many other risk factors such as physiological measurements, patient behaviors and characteristics, environmental triggers, and personal risk scores. In this paper, we propose a method to forecast future asthma attacks using historical data linking the asthma severity level and the personalized risk scores of triggers. Specifically, we used the Additive Interaction Analysis of Exposures technique combined with the Long Short-Term Memory method to predict risk scores of respiratory and oxygen saturation. These risk scores are then used to forecast the probability of an asthma attack.