One hundred and nine residents with known diabetes mellitus were randomly selected from 19 nursing, 16 residential, six elderly mentally infirm, and three dual registered homes in the catchment area of one hospital in north west England. Controls were sex and nearest age matched non-diabetic residents in the same homes. All residents were interviewed by a research nurse, who completed a comprehensive questionnaire and examined each subject during six months from January 1995. Hospital and general practice records were subsequently reviewed.