OBJECTIVES: To assess treatment patterns among patients with shoulder osteoarthritis (OA). METHODS: Data from Thomson MarketScan, a large national managed care population, was used to identify patients with a shoulder OA diagnosis in the first 6 months of 2005 (i.e., the index date). The 360 days post index (identification period) was used to establish baseline treatments (i.e., conservative management (pharmaceutical and physical therapy), steroid injections and shoulder surgery). Patients were required to be continuously eligible for 54 months post-index and were excluded if they had a shoulder surgery claim in the identification period.  Four cohorts were followed based on the baseline treatments:  C1- conservative treatment; C2- conservative treatment and at least one steroid injection; C3- at least one steroid injection; C4- no treatment claims. Progression to additional treatments was evaluated descriptively from day 361 to 1260 in 180 increments.  Logistic regression was used to model the odds or having a claim for a treatment. RESULTS: A total of 3646 patients met the analysis criteria (C1, n=2,815(77.2%); C2, n=171(4.7%); C3, n=27(0.7%); C4, n=633(17.4%)). The distribution was split evenly between males (50.2%) and females (49.8%). Patients who received steroid injections in the identification period had the greatest likelihood of having a steroid injection in the observation period (C1-19.2%;C2-43.9%;C3-44.4%;C4-14.1%).  The percentage with shoulder surgery was 6.4%, 15.2%, 11.1% and 6.5% for C1 to C4, respectively.  Patients with steroids in the observation period (C2 and C3) were more likely to have surgery in the first year of observation.  Logistic regression showed that females who had steroid injections (C2 and C3 combined) had odds of surgery that were 2.9 times greater than females with no treatments (C1). CONCLUSIONS: The most significant predictor of surgery was presence of steroid injections.  Rates of steroid injections and surgery differed based on presence of pre-existing treatments.