Does choice of spatial unit matter for estimating small-area disparities in health and place effects in the Vancouver Census Metropolitan Area?

BACKGROUND The purposes of this study were to determine (i) the extent to which small-area estimates of self-rated health are dependent upon the choice of areal unit and measure of socio-economic (SES) status, and (ii) the extent to which place effects on self-rated health are dependent upon the choice of areal unit and measure of SES. METHODS The data were obtained from a subset of respondents in the Canadian Community Health Survey 2.1 (2003) aged 18 to 74 residing in the Vancouver Census Metropolitan Area. General health status was estimated using an item assessing respondents' self-rated health. Small-area data were obtained from the Statistics Canada 2001 Census at two spatial levels: larger Census Tract (CT) (average population 2,500-8,000) and smaller Dissemination Area (DA) (average population 400-700). SES quintiles were constructed using median family income and two indices. Hierarchical non-linear modelling was used to test for place effects. RESULTS A gradient was found of increasing prevalence of "fair or poor" self-rated health by decreasing SES quintile at both the DA and CT level. With age category, sex, family income and education controlled for, hierarchical analysis showed that compared with living in a high SES CT or DA the odds of reporting fair or poor self-rated health increased for respondents living in the lowest quintile CT or DA. INTERPRETATION Aggregation using DAs or CTs produces only small differences in estimates of fair or poor self-rated health by quintiles of SES. Gradients are somewhat stronger for DAs. Place effects are somewhat stronger for deprivation indices than the measure of median income.