The significance of a very elevated erythrocyte sedimentation rate (ESR) in elderly patients is debated. In a retrospective study, we searched the records of a laboratory providing the sole service to a health district for ESR measurement and identified all non-surgical and non-psychiatric patients over the age of 65 who had had an ESR above 50 mm/h. Diagnoses and mortality in a 1-year follow-up were determined from case notes. Four hundred and nine subjects (median age 75; range 65-99) were identified and data on 401 of these (155 male, 246 female; median ESR 80 mm/h, range 50-148) were adequate for 1 year follow-up. Forty-eight percent had a persistently raised ESR (two values > 50 mm/h separated by at least 14 days; group 1); 39% had a single ESR measurement only (group 2), and 13% had a transiently raised ESR (group 3). The commonest diagnosis in group 1 patients was rheumatological disease (51.8%), followed by infection (31.9%) and non-haematological malignancy (11%). Infection was the commonest diagnosis in groups 2 (47.4%) and 3 (43.7%), followed by non-haematological malignancy (19.9%) in group 2 and rheumatological disease (20.4%) in group 3. In only 1 in 20 cases was no diagnosis apparent at 1 year. The standardised mortality ratio (SMR) of the combined groups 1 and 2 (482; CI: 421-544) was strikingly raised, and even more so if patients with rheumatoid arthritis were excluded (542; CI 458-625). Where there were sufficient numbers of deaths to make SMR estimations valid, a gradient of mortality against the level of the ESR could be observed.