Small area variation in hospital admission: random or systematic?

For some conditions hospital admission is mandatory. This should lead to low variability in admission rates and no effect on admission rate of distance from hospital. If admission is discretionary, we would expect high variability in small area admission rates, and a decline in admission rate as travel time to hospital increases. We wanted to see if non-random variability of admission rates, as measured by the systematic coefficient of variation (SCV), and distance decay, as estimated in regression models, were related. We examined variability and travel time dependence of hospital admission for seven conditions in 62 small (mean population 9900) areas of Surrey, England. Age and sex standardized admission ratios (SAR) were calculated, and their dependence on travel time, adjusting for deprivation, were estimated by linear multiple regression adjusted for spatial correlation. Deprivation was measured by Jarman's score, and time by computerized estimates of drive time to the nearest acute hospital. We found an inverse relationship between time to hospital and admission ratio for ischaemic heart disease, bronchopneumonia and chronic bronchitis. Admission ratios for diabetes mellitus and stroke were related to neither deprivation nor time. For these seven conditions there was no simple relationship between SCV and travel time dependence.