Background: Veterans Health Administration (VHA) databases are used extensively to study racial/ethnic disparities; however, these databases may not capture all care received by VHA patients. Objectives: We examined the extent to which accounting for non-VHA care changed conclusions about racial/ethnic disparities for VHA patients with diabetes. Methods: Using a cross-sectional observational study, we analyzed a national sample of noninstitutionalized Hispanic (n = 5931), black (n = 24,670), and white (n = 149,222) VHA patients with diabetes who were at least 65 years of age for receipt of annual HbA 1c testing, low-density lipoprotein (LDL) cholesterol testing, or eye examination from VHA and Medicare administrative files. Results: In VHA alone data, adjusting for patient characteristics, Hispanic and black patients were as likely as white patients to receive HbA1c testing (odds ratio 1.06 [95% confidence interval 0.99-1.13] and 1.04 [1.00-1.07], respectively), and more likely to receive eye examinations (1.31 [1.24-1.38] and 1.33 [1.29-1.37], respectively). Hispanic patients were equally likely (1.01 [0.95-1.07]) and black patients were less likely (0.81 [0.79-0.84]) to receive LDL testing versus white patients. In VHA plus Medicare data, Hispanic and black patients were less likely than white patients to receive HbA1c (0.76 [0.71-0.82] and 0.83 [0.80-0.87], respectively) and LDL testing (0.84 [0.79-0.90] and 0.70 [0.68-0.72], respectively), and equally likely to receive eye examinations (0.91 [0.86-0.96]) and 0.98 [0.95-1.01]), respectively). Accounting for VHA facility had little effect on results. Conclusions: Restricting to VHA data masks racial/ethnic disparities in care of VHA patients. VHA researchers must be aware and supplement VHA data with other sources whenever possible. Copyright © 2006 by Lippincott Williams & Wilkins.