Use of homogeneous social areas for ecological analyses - A study of accident rates in pre-school children

Background: The results of ecological analyses are sensitive to the geograp hical areas used and census areas are not necessarily the most suitable uni ts. This study compares the effects of using specially constructed areas wi th similar social characteristics rather than standard census areas as basi c units in a small area ecological study. Methods: The study used ecologica l regression analyses of accident rates in pre-school children on social an d demographic predictors using census enumeration districts, wards and spec ially constructed social areas as the units. The setting was the city of No rwich, UK and adjacent suburbs, consisting of 349 census enumeration distri cts nested within 30 wards or 21 social areas. Results: Analyses at differe nt geographical scales produced similar estimates of the effects of materia l deprivation, presence of lone parent households and young population age structures on child accident rates, but the r(2) values varied considerably , the weakest relationships being found for enumeration districts and the b est fitting for social areas. Adjusted r(2) values between log accident Fat es and material deprivation scores, for example, were 10.5% for enumeration districts, 52.7% for wards and 63.3% for social areas. Conclusions: Specia lly constructed areas were preferable to enumeration districts and wards as units for identifying ecological relationships between accident rates in p re-school children and social conditions. Homogeneous social areas are pote ntially useful units of analysis for ecological studies in epidemiology.