SMOOTHING APPROACHES FOR EXPLORING ECOLOGICAL BIAS

Yijie Zhou, Department of Biostatistics, Johns Hopkins Bloomberg School of Public Health
Francesca Dominici, Department of Biostatistics, The Johns Hopkins Bloomberg School of Public Health
Thomas A. Louis, Department of Biostatistics, Johns Hopkins University Bloomberg School of Public Health

Manuscript submitted to the Journal of the Royal Statistical Society, Series C

Abstract

Ecological bias arises in epidemiological studies when spatially aggregated data are used to estimate the association between an exposure and a health outcome in the absence of individual-level data. In this paper we develop spatial smoothing methods for data aggregation for evaluating ecolog- ical bias in Generalized Linear Models (GLM). Our smoothing methods for data aggregation account for both the form and degree of aggregation. We illustrate with examples that data aggregation which incorporates spatial characteristics of exposure and outcome may be less subject to ecological bias. In this paper we also provide a close form expression of first-order approximation to ecological bias from pure model specification. We apply our proposed methods to the study of racial disparities in mortality rates that includes more than 4 million Medicare enrollees residing in 2095 zip codes in the Northeast region of U.S. We find that the ecological bias in estimating the association between race and mortality rates highly depends on both the form and degree of aggregation.