e15148 Background: There is questionable reliability of administrative claims data to identify BM, yet historically such claims were the only source for identifying BM in SEER-Medicare (SM). Starting in 2004, SEER provided tumor registry information on incident BM. We compared frequencies of BM using registry and claims data among men diagnosed with PCa. Methods: We analyzed pts aged 66 and older from the linked SM database. Pts with stage IV (M1) PCa diagnosed between 2004 and 2007 were followed until death or censoring. Identifying pts with BM using claims required at least one inpatient or outpatient claim with a diagnosis code of 198.5 within (i.e., +/-) 1 month of the SEER diagnosis month (designated BMclaims-90) or at any time during follow up (designated BMclaims-ever). BM identified using SEER was based on an indicator for bone metastasis at diagnosis using the AJCC M1b classification (designated BMseer). We calculated the sensitivity and specificity of BM at diagnosis in claims (BMclaims-90) with respect to the incident BM data in SEER, assuming BMseer to be the gold standard. Results: Application of inclusion/exclusion criteria resulted in 3,664 stage IV (M1) PCa pts. Average age was 79 years, 13% were African American, and 61% had poorly/un-differentiated tumors. BMseer (n=2,301; 63%) was higher than BMclaims-90 (n=1,979; 54%) and lower than BMclaims-ever (n=2,878; 79%). Assuming BMseer as the standard measure of BM at diagnosis, BMclaims-90 had 58.8% sensitivity (n=1,353), 54.1% specificity (n=737), and 68.4% positive predictive value for identifying BM at diagnosis. Among men with BM identified using BMseer, 84% (1,931) had BMclaims-ever and the median (min; max; average) time to the first BM claim was 0.9 (-2.6; 57.1; 3.9) months. Among men with BM identified using BMclaims-90, the median (min; max; average) time to the first BM claim was 0.6 (-1.0; 2.0; 0.7) months. Conclusions: The incidence of BM in PCa pts differs between a commonly-used algorithm to identify BM in claims data and the AJCC staging information in tumor registry data. There was moderate concordance between the claims-based measure and the SEER-based measure of BM at diagnosis.