The use of medical records, fasting plasma glucose, 2-hour post-challenge glucose and glycosylated hemoglobin is compared in a population diabetes screening survey among Indians in northern Ontario and Manitoba. As expected, screening identified additional diabetics not previously known to the health care system. Glycosylated hemoglobin was highly correlated with fasting glucose and could discriminate between diabetics and non-diabetics, although not those with impaired glucose tolerance. In multiple regression models more significant predictors were identified and a larger proportion of the overall variance accounted for when log hemoglobin instead of log fasting glucose was used as the dependent variable. In a subsample of subjects, no additional diabetics were identified when both 2-hour and fasting criteria were used, compared to fasting plasma glucose alone.