Geographical variation in the utilisation of in-patient medical services for elderly people in a health district in England was examined in relation to supply of in-patient geriatric medical care and indicators of need. An ecological study design based on electoral wards was used. The health district had a resident population of 67,919 aged 65 y or more and was divided into three localities, each with a different supply of in-patient medical services for elderly people. Locality A had a traditional model of geriatric medical care, Locality B an integrated model and Locality C an age-related model. Localities A and C also had a high provision of general practice hospital beds. The main outcome measure was the age and sex standardised hospital admission ratio for people aged 65 y or more admitted under geriatric medicine, general medicine or general practice in April 1991-March 1992. There were 8829 admissions in 1991/2, 48% in general medicine, 40% in geriatric medicine and 12% in general practice, giving an overall unadjusted admission rate of 130 per 1000 population aged 65 y or more for the three specialties combined. Locality A had the highest, and Locality B the lowest, unadjusted admission rate for the three specialties combined. This rate remained highest in Locality A if second and subsequent admissions in the same period were excluded. Lengths of stay in geriatrics were longest in this locality but lengths of stay for the three specialties combined were similar in the three localities. Multiple regression was used to examine the effect of three indicators of need, the Jarman score, standardised mortality ratio and prevalence of limiting long-term illness, on standardised admission ratios at the electoral ward level. Jarman score had a significant independent association with the standardised admission ratio but adjustment for this factor did not alter the ranking of the three localities, with the standardised admission ratio remaining highest in Locality A. Subject to the limitations of the study, the results suggest that factors related to the supply of in-patient medical services may be associated with geographical variation in medical admissions for elderly people.