Artificial Neural Networks applied to assess the impact of PM2.5 on hospital admissions for cardiovascular diseases

Resumo

The high emission of atmospheric pollutants in large urban centers causes several harms to population
health. Thus, it is necessary to evaluate the negative impact of its concentration to assist in decision-making and
public policies by government agents. Several modeling techniques have been used to assess the effects of air
pollution on human health. However, due to their greater flexibility in analyzing the complex nonlinearity of
environmental data, Artificial Neural Networks (ANN) have been shown to be the most attractive approach for
solving such data modeling problems. This work aimed to compare the performance of two artificial neural
networks, Multilayer Perceptron (MLP) and Extreme Learning Machine (ELM) in estimating the number of
hospital admissions for cardiovascular diseases due to the concentration of fine particulate matter (PM2.5) in
Joinville, Brazil. Daily PM2.5 concentration and meteorological variables were considered as input variables. MLP
network was able to achieve better performance to estimate hospital attendance because of these environmental
conditions after three days of PM2.5 exposure. The results demonstrate that ANN can be used to predict hospital
admissions due to air pollution levels or adverse meteorological conditions and therefore, be used to guide
government public policies on air quality and health risk assessment.