Background: Asthma is one of the most common chronic diseases worldwide, and a growing public health concern. It is characterized by chronic bronchial inflammation and a multi-factorial etiology that includes genetic, immunological and environmental factors. Our goal was to develop a patient-customizable asthma forecasting system that takes into account meteorological factors, air pollution, and pollen and viral respiratory infections, which are the most common environmental factors known to exacerbate asthma.
Methods: We analyzed the health insurance records of patients who visited the emergency departments in Seoul due to an asthma attack and who were treated with salbutamol. Our potential predictors of asthma symptoms included meteorological factors (temperature, humidity, air pressure, and amount of sunshine), air pollution factors (ozone and Asian yellow dust), environmental factors (pollen) and health-related factors (seasonal influenza viral infection). Patients were assigned to five age groups and separated according to gender and season (5 × 2 × 4 groups), which resulted in a total of 40 groups. Three models (a multiple regression model, a logistic regression model, and a decision tree) were tested for their ability to predict exacerbated asthma symptoms (referred to as “attention” symptoms). The suitability of the final model was assessed using predictability skill scores.
Results: Logistic regression and the decision tree provided the best forecasts. Using the estimated binary values for “continuous management” versus “attention” symptoms, the optimal threshold was selected.
Conclusions: Our binary forecasting model may improve the prediction of asthma exacerbation in a clinical setting.