OBJECTIVES: CRPC is an advanced form of prostate cancer with poor prognosis. The changing treatment landscape calls for use of real-world data to provide current information on disease epidemiology, treatment patterns, comparative effectiveness, and treatment cost. Administrative claims databases are a rich source of real-world data; however, they lack clinical values needed to identify CRPC patients. As currently there is no standard method to identify CRPC patients when clinical variables are unavailable, we developed and validated an algorithm using claims data to reliably identify CRPC patients. METHODS: Male patients with an ICD-9 code for PC between January 1, 2004 and November 30, 2012 were identified from IMS pharmacy and medical claims data. Metastatic PC was identified by additional tumors on or after the PC diagnosis. Surgical castration was determined using CPT and ICD-9 procedure codes for orchiectomy; medical castration was identified using codes for hormone therapies. The algorithm identified patients as having CRPC if they had any of the following: 1) switch in hormone therapy; 2) new metastases while on hormone therapy; 3) switch in treatment from any hormone therapy to chemotherapy. The algorithm was validated against a PSA-based definition of CRPC (at least two PSA increases following castration or new metastases while on hormone therapy). RESULTS: A total of 269 patients met inclusion criteria and had 3 or more PSA measurements. The two methods agreed in 88% of patients; 220 (81.8%) were identified as CRPC by both methods and 17 (6.3%) were identified as not castration resistant by both methods. A statistical comparison of the two methods yielded a Cohen’s kappa of 0.4491, indicating moderate agreement. CONCLUSIONS: The algorithm was concordant with a PSA-based definition of CRPC and serves as a new tool to identify CRPC patients using claims data. Future validation against a different data source is needed.