Age and sex adjusted inpatient episode ratios were calculated from hospital records over a two year period for 555 census wards in Cambridgeshire, Norfolk and Suffolk, UK. Hospital episodes were divided into acute, psychiatric and geriatric specialties, and elective and emergency acute admissions were distinguished. Variations in inpatient episode ratios between wards were compared with census indicators of the 'needs' of local populations for inpatient services, measures of the local provision of hospital and other related services, and measures of the distance to the nearest general practitioner surgery and the distance to the nearest hospital. Hospital episodes were found to be strongly related to both distance measures, but the associations were partially explained by a tendency for the health status of local populations (measured by the needs indicators) to be worst in urban areas close to health services. Including needs and provision variables together with the distance variables explained most of the variations in age and sex adjusted inpatient ratios for small areas. Needs were the most important determinants of emergency acute and psychiatric inpatient rates, but service provision was more important for elective acute and geriatric inpatients. Controlling for needs and provision, distance to hospital produced 17% reduction in acute episodes, 37% reduction in psychiatric episodes and 23% reduction in geriatric episodes over the range of distances observed. Distance to nearest GP surgery reduced elective acute episodes by up to 15% after controlling for confounders. These results demonstrate that the NHS is falling short of its aim to provide fair access to services irrespective of geography, and new policies will be required if geographical inequities are to be reduced.