Inference from ecological models: estimating the relative risk of stroke from air pollution exposure using small area data.

Maheswaran et al. (2006) analysed the effect of outdoor modelled NO(x) levels, classified into quintiles, on stroke mortality using a Poisson Bayesian hierarchical model with spatial random effects. An association was observed between higher levels of NO(x) and stroke mortality at the small area (enumeration district) level. As this model is framed in an ecological perspective, the relative risk estimates suffer from ecological bias. In this paper we use a different model specification based on Jackson et al. (2008), modelling the number of cases of mortality due to stroke as a binomial random variable where p(i) is the probability of dying from stroke in area i. The within-area variation in outdoor modelled NO(x) levels is used to determine the proportion of the population in area i falling into each of the five exposure categories in order to estimate the probability of an individual dying from stroke given the kth level of NO(x) exposure assuming a homogeneous effect across the study region. The inclusion of within-area variability in an ecological regression model has been demonstrated to help reduce the ecological bias (Jackson et al., 2006, 2008). Revised estimates of relative risk are obtained and compared with previous estimates.

[1]  S. Richardson,et al.  Hierarchical related regression for combining aggregate and individual data in studies of socio‐economic disease risk factors , 2007 .

[2]  Sebastien J-P A Haneuse,et al.  Hierarchical Models for Combining Ecological and Case–Control Data , 2007, Biometrics.

[3]  W P D LOGAN,et al.  Mortality in the London fog incident, 1952. , 1953, Lancet.

[4]  Andrew Thomas,et al.  WinBUGS - A Bayesian modelling framework: Concepts, structure, and extensibility , 2000, Stat. Comput..

[5]  Sebastien J-P A Haneuse,et al.  The Combination of Ecological and Case-Control Data. , 2006, Journal of the Royal Statistical Society. Series B, Statistical methodology.

[6]  Alan D. Lopez,et al.  Mortality from tobacco in developed countries: indirect estimation from national vital statistics , 1992, The Lancet.

[7]  Ruth Salway,et al.  A statistical framework for ecological and aggregate studies , 2001 .

[8]  Sylvia Richardson,et al.  Improving ecological inference using individual‐level data , 2006, Statistics in medicine.

[9]  Robert Haining,et al.  The effect of alternative representations of population location on the areal interpolation of air pollution exposure , 2005, Comput. Environ. Urban Syst..

[10]  Paolo Vineis,et al.  Outdoor air pollution and lung cancer: Recent epidemiologic evidence , 2004, International journal of cancer.

[11]  H. Kan,et al.  Air pollution and daily mortality in Shanghai: a time-series study. , 2003, Archives of environmental health.

[12]  J. Besag,et al.  Bayesian image restoration, with two applications in spatial statistics , 1991 .

[13]  A. Hansell,et al.  Geographic Variations in Risk: Adjusting for Unmeasured Confounders Through Joint Modeling of Multiple Diseases , 2009, Epidemiology.

[14]  L. Sheppard,et al.  Long-term exposure to air pollution and incidence of cardiovascular events in women. , 2007, The New England journal of medicine.

[15]  R. Haining,et al.  Using Modeled Outdoor Air Pollution Data for Health Surveillance , 2004 .

[16]  B. Oftedal,et al.  Urban air pollution and mortality in a cohort of Norwegian men. , 2004, Environmental health perspectives.

[17]  D. Black.  HEALTH AND DEPRIVATION: Inequality and the north , 1988 .

[18]  Ruth Salway,et al.  Sources of bias in ecological studiesof non-rare events , 2005, Environmental and Ecological Statistics.

[19]  M. Williams.  Monitoring of exposure to air pollution. , 1995, The Science of the total environment.

[20]  Majid Ezzati,et al.  Estimates of global mortality attributable to smoking in 2000 , 2003, The Lancet.

[21]  Massimo Craglia,et al.  GIS in public health practice , 2004 .

[22]  Bradley P. Carlin,et al.  Bayesian measures of model complexity and fit , 2002 .

[23]  Aki Vehtari.  Discussion to "Bayesian measures of model complexity and fit" by Spiegelhalter, D.J., Best, N.G., Carlin, B.P., and van der Linde, A. , 2002 .

[24]  S Richardson,et al.  A Bayesian approach to measurement error problems in epidemiology using conditional independence models. , 1993, American journal of epidemiology.

[25]  Jie Yang,et al.  Linking health and environmental data in geographical analysis: it's so much more than centroids. , 2009, Spatial and spatio-temporal epidemiology.

[26]  R. Burnett,et al.  Spatial Analysis of Air Pollution and Mortality in Los Angeles , 2005, Epidemiology.

[27]  A. Gelman.  Prior distributions for variance parameters in hierarchical models (comment on article by Browne and Draper) , 2004 .

[28]  D. Rubin,et al.  Inference from Iterative Simulation Using Multiple Sequences , 1992 .

[29]  Noah Nd,et al.  Letter: Meningococcal infections. , 1975 .

[30]  Andrew Gelman,et al.  General methods for monitoring convergence of iterative simulations , 1998 .

[31]  C. Gotway,et al.  Combining Incompatible Spatial Data , 2002 .

[32]  N. G. Best,et al.  Outdoor NOx and stroke mortality: adjusting for small area level smoking prevalence using a Bayesian approach , 2006, Statistical methods in medical research.