OBJECTIVES: OA is a highly prevalent disease with significant economic implications. With increased aging and obesity, the incidence and prevalence of OA is expected to continually rise, resulting in higher utilization of health resources. As part of a systems dynamic modelling research programme to inform OA care planning in Alberta, we used provincial administrative data to estimate OA prevalence and incidence and examined the sensitivity of estimates to different OA case definitions. METHODS: We obtained Alberta Health and Wellness (AHW) Discharge Abstract (DAD), Physician Claims (Claims) and Ambulatory Care Classification System (ACCS) databases from 1994 to 2010with ICD-9 and ICD-10 OA diagnosis codes (715 and M15-19 codes) identified in any field. In the base case, OA incidence and prevalence was captured for patients documented with this diagnosis who had at least two physician OA visits within two years.  RESULTS: The incidence and prevalence of OA in 2008 were estimated at 7.5 cases/1000 population and 87 cases/1000 population, respectively. OA prevalence was most affected by run-in time (number of years of data), followed by the number of physician visits used to define OA, the number of years between cases, and the databases applied in the analysis. Over 15 years, prevalence approaches steady-state. Physician Claims data captured most (98%) of the OA cases. CONCLUSIONS: Administrative data have limitations but are the only routinely collected population level source for these estimates. The key factors that impact incidence and prevalence estimates for chronic diseases, like OA, is the number of years of historical data and number of visits used in the case definition. These estimates most likely underestimate OA prevalence, but likely capture clinically relevant disease for which patients seek care.  The value of these data for health authorities is to allow for better prediction of demand for planning future health services.