Assessment and prediction of short term hospital admissions: the case of Athens, Greece

Abstract The contribution of air pollution on hospital admissions due to respiratory and heart diseases is a major issue in the health-environmental perspective. In the present study, an attempt was made to run down the relationships between air pollution levels and meteorological indexes, and corresponding hospital admissions in Athens, Greece. The available data referred to a period of eight years (1992–2000) including the daily number of hospital admissions due to respiratory and heart diseases, hourly mean concentrations of CO, NO2, SO2, O3 and particulates in several monitoring stations, as well as, meteorological data (temperature, relative humidity, wind speed/direction). The relations among the above data were studied through widely used statistical techniques (multivariate stepwise analyses) and Artificial Neural Networks (ANNs). Both techniques revealed that elevated particulate concentrations are the dominant parameter related to hospital admissions (an increase of 10 μg m−3 leads to an increase of 10.2% in the number of admissions), followed by O3 and the rest of the pollutants (CO, NO2 and SO2). Meteorological parameters also play a decisive role in the formation of air pollutant levels affecting public health. Consequently, increased/decreased daily hospital admissions are related to specific types of meteorological conditions that favor/do not favor the accumulation of pollutants in an urban complex. In general, the role of meteorological factors seems to be underestimated by stepwise analyses, while ANNs attribute to them a more important role. Comparison of the two models revealed that ANN adaptation in complicate environmental issues presents improved modeling results compared to a regression technique. Furthermore, the ANN technique provides a reliable model for the prediction of the daily hospital admissions based on air quality data and meteorological indices, undoubtedly useful for regulatory purposes.