Objective: To examine black-white and Hispanic-white differences in total knee arthroplasty from 2001 to 2013 in a large cohort of patients diagnosed with osteoarthritis (OA) in the Veterans Affairs (VA) health care system. Methods: Data were from the VA Musculoskeletal Disorders cohort, which includes data from electronic health records of more than 5.4 million veterans with musculoskeletal disorders diagnoses. We included white (non-Hispanic), black (non-Hispanic), and Hispanic (any race) veterans, age ≥50 years, with an OA diagnosis from 2001–2011 (n = 539,841). Veterans were followed from their first OA diagnosis until September 30, 2013. As a proxy for increased clinical severity, analyses were also conducted for a subsample restricted to those who saw an orthopedic or rheumatology specialist (n = 148,844). We used Cox proportional hazards regression to examine racial and ethnic differences in total knee arthroplasty by year of OA diagnosis, adjusting for age, sex, body mass index, physical and mental diagnoses, and pain intensity scores. Results: We identified 12,087 total knee arthroplasty procedures in a sample of 473,170 white, 50,172 black, and 16,499 Hispanic veterans. In adjusted models examining black-white and Hispanic-white differences by year of OA diagnosis, total knee arthroplasty rates were lower for black than for white veterans diagnosed in all but 2 years. There were no Hispanic-white differences regardless of when diagnosis occurred. These patterns held in the specialty clinic subsample. Conclusion: Black-white differences in total knee arthroplasty appear to be persistent in the VA, even after controlling for potential clinical confounders.