Development and application of Semiparametric Regression Models for estimation of short term health effects related to Air Pollution in Chennai

Abstract

Air pollution in the outdoor ambient environment ranks among the leading risk factors contributing to both the global and regional burden of disease Time series studies of the effects of short term exposure on morbidity and mortality from cardiovascular or respiratory diseases have provided some of the most consistent evidence of serious adverse health effects of air pollution for regulatory policies in North America Europe and Asia More recently results from a coordinated set of time series studies examining the association of natural all cause mortality with PM10 exposures in the cities of Chennai Delhi and Ludhiana have been published These initial studies in India pointed out to the need for additional validation of the methods using data over extended time periods as well as the need to develop model refinements to address unique features of exposure and health datasets available through relevant Governmental agencies in India The study is aimed at developing semiparametric methods to describe exposure response relationships between daily average ambient PM10 concentrations and short term health effects in Chennai city through time series analyses The excess risk estimates obtained across models in the study are very similar to the summary estimates obtained from the meta analyses of all Asian studies as well as European and North American Studies This study explored several approaches to improve outputs of models for estimating the exposure response relationship between PM10 and mortality or morbidity through time series analyses These were able to use data over an extended period of time compared to earlier study in Chennai to provide more refined and robust effects estimates for all cause mortality Finally the development of an autoregressive Poisson model to address serial correlation in outcomes in the face of missing exposure data together with simulation studies to validate the model has significantly enhanced the ability to address model uncertainties. Until such time when infrastructural inve