The objective of this study was to compare self-reported measures of diabetes care with measures derived from medical records in a well-defined population. Diabetes measures were collected through a 1997 Behavioral Risk Factor Surveillance System telephone survey of American Indians living on or near 7 Montana reservations (N = 398) and were compared with data collected from charts of a systematic sample of American Indians with diabetes seen in 1997 at Indian Health Service (IHS) facilities. Survey respondents were more likely to report a duration of diabetes > or = 10 years (44 vs 31%), annual dilated retinal exam (75 vs 59%), and an influenza immunization in the past year (73 vs 57%) compared with estimates from the chart audit. Estimates of pneumococcal immunization (88 vs 42%), annual cholesterol screening (86 vs 69%), and overweight, based on body mass index (67 vs 50%), were significantly higher from the chart audit. No significant differences were found between the survey respondents and the chart audit data for annual foot exams (65 vs 61%), annual blood pressure checks (98 vs 93%), high cholesterol (35 vs 41%), and high blood pressure (54 vs 64%). These findings suggest that self-reported data may over and underestimate specific measures of diabetes care.