Advanced age and its related comorbidity may affect both the patterns and goals of diabetes treatment. We examined the relationships of demographic variables and comorbidity with drug treatment for diabetes in the elderly. We studied both the 81,700 residents of New Jersey, aged 65-99 years, who were hospitalized between July 1, 1989 and June 30, 1991 and had prescription drug coverage either through Medicaid or the Pharmacy Assistance for the Aged and Disabled program, and a sample of 80,000 nonhospitalized elderly beneficiaries in these programs. Rates of utilization of insulin or oral hypoglycemic drugs in the 120 days before admission were substantially lower in those aged > or = 85 or in nursing homes. Among patients with previously treated and diagnosed diabetes, the likelihood of treatment after discharge declined with older age (odds ratio [OR] for treatment in those aged > or =85 relative to 65-74 years: 0.57; 95% confidence interval [CI]: 0.45-0.72), nursing home residence (OR: 0.30; CI: 0.22-0.41), and higher levels of comorbidity (OR for modified Charlson index > or = 5 relative to 0: 0.43; CI: 0.27-0.67). In patients who had a discharge diagnosis of diabetes but no prior treatment, those in nursing homes and those with greater comorbidity also had lower rates of diabetes treatment after discharge. Although the prevalence of diabetes increases with age and the risks of many consequences of diabetes remain high, the rate of drug treatment for diabetes declines with older age and greater comorbidity, perhaps because of concern about side effects or reduced treatment benefits due to competing risks of death. Absence of data from randomized clinical trials of diabetes treatment in the elderly appears to have resulted in considerable physician ambivalence on the benefits and risks of glycemic control in older diabetics.