Elderly populations are susceptible to many non-communicable diseases, including diabetes. Lack of awareness regarding disease status and risk factors increase complications and mortality. We conducted a cross-sectional community-based study of 1633 randomly selected participants aged 60 years and above in urban and rural areas in the Kathmandu Valley of Nepal. Study goals were: (i) to determine the prevalence of diabetes in elderly subjects as diagnosed prior to and during the study, and (ii) to identify and compare the determinants of diabetes as diagnosed prior to and during the study. A structured questionnaire was used to collect information regarding prevalence and potential determinants of diabetes diagnosed during and before the survey. Anthropometric measurements, blood pressure measurements and fasting blood sugar tests were also measured. Risk factor analysis was done using multinomial logistic regression; subjects with no diabetes constituted the reference group. We detected an overall diabetes prevalence of 25.9%, 17.3% diagnosed during the survey and 8.6% previously diagnosed. Age, disturbed sleep, and family history of hypertension were marginally significantly (p < 0.10) associated with diabetes diagnosed in the survey. In contrast, education, exercise, health perception, family history of hypertension, having a caretaker at night, receiving help from friends when seeking health care and mean waist circumference were significantly (p pounds 0.05) or marginally associated with diabetes detected before the survey. The high prevalence of diabetes in the elderly population studied, and the low level of prior diagnosis, identify an important public health problem. There is a need for improved screening of diabetes and improved health care and education in the elderly. There is also need for better understanding of risk factors associated with previously undiagnosed diabetes.