The use of a neural network for studying the relationship between air pollution and asthma-related emergency room visits

Abstract

To establish the relationship between air pollution levels and bronchial asthma-associated emergency room (ER) visits, we adapted artificial network technology to conduct this study which focused on three different pollutants, sulphur dioxide, nitrogen oxide and ozone. The study population was comprised of adults presenting to the emergency room of a large metropolitan hospital in Israel during a 3-month period with acute exacerbation of bronchial asthma and who had a past history of intermittent airway disease compatible with bronchial asthma. The range of mean daily pollutants levels for the whole period were: O 3 = 15–26 μg m −3, NO x = 36–108 μg m −3, NO = 16–70 μg m −3, and SO 2 = 11–32 μg m −3. The data sets were composed of input air pollution levels and output ER visits. The first 126 data sets used for the training phase showed that maximal ER visits were mainly associated with the highest cumulative values of air pollution and mostly with nitrogen oxide. In phase two, an attempt was made to predict ER visits based on air pollution level in 49 data sets. The study findings demonstrated that ordinary network technology can be used for learning the effect of air pollution ER visits and, although limited in accuracy, to also predict future ER visits.