This study investigates the potential for ecological studies to contribute useful information on variations in service use, both across areas and across different types of psychiatric care. The analysis uses data for the 62 counties of New York State, which include both urbanised and rural areas, with widely differing social, household and ethnic structures. We analysed data on service use by patients aged 15–64 years for several psychiatric conditions combined. The research reported here used an approach which for several reasons is innovative, compared with other ecological studies of psychiatric service use. First, the impact of population variables on both ambulatory and hospitalisation rates is considered, whereas many previous studies are confined to hospital use. Second, our method combines and weights population variables in the ‘need index’ in a way that (a) reflects geographical variations in service use rates in both hospital and ambulatory sectors (b) controls for service configuration and access as well as (c) allowing for spatial autocorrelation in the need index. To demonstrate this method, four simple indicators of poverty, social isolation, concentration of racial minorities and population density, were used in combination to define a ‘needs’ index that predicts use of psychiatric services at county level. Comparison with alternative methods of measuring need, using the same data but based on more conventional strategies, resulted in significantly different need rankings of areas. Our composite index of ‘underlying need’ showed a positive association with service use (hospital and ambulatory care for men and for women). This relationship controls for access to services, and allows for spatial correlation in the need construct. Controlling for underlying population ‘need’, a measure of spatial proximity to hospitals with psychiatric beds had an independent effect, being associated positively with hospital inpatient use, and negatively with ambulatory referral rates. For women, after allowing for population characteristics and proximity to hospital beds, there was no strong relationship between county level indicators of hospitalisation and rate of use of ambulatory care. For men, at the county population level considered in this ecological study, relatively high rates of admission in the inpatient sector were associated with higher rates of referral in the ambulatory care sector. The analysis highlights the importance of considering a set of socio-geographic factors that affect need for care (as well as simple measures of population size and age/sex structure) when making ecological comparisons of levels of mental health service use. It also demonstrates that the geographical relationship between psychiatric hospital use and population ‘need’ indicators should to be evaluated in ways that take into account a functionally complex ‘matrix’ of local conditions including population characteristics, service configuration, access and possible interrelationships between various psychiatric care sectors.