A multi-pollutant model: a method suitable for studying complex relationships in environmental epidemiology

Abstract

Most of the models developed to study the effects of pollutants on the health of people are single input and single outcome while adjusting for other variables. However, the real environment is a mixture of pollutants, which affect people synergistically and varies in time and space. The aim of this work is to introduce a multiple exposures-outcomes tree regression method. An oblique tree with Weighted Oblique Decision Trees (WODT) algorithm was designed to find the share effects of pollutant(s) on health outcomes and investigate the temporal and spatial differences. Using this method, a case study was conducted on the association between O3, NO2, PM2.5 and asthma, COPD, pneumonia, and bronchitis in CA, USA. The results indicated that NO2 and O3 are responsible for asthma emergency department (ED) visits in South Coast and San Diego Air Basins during January–April and October–December for the years 2005–2015. For PM2.5, the results indicated that an increase in concentration was associated with an increase in the number of ED visits for COPD and pneumonia during January–December in the whole study area. The method introduced in this study is useful in handling multi-pollutant exposure conditions. Using this method, public health agencies and policy makers can better understand the relative effects of multiple pollutants on the health of people in temporal and spatial scales.