Utility of the SEER-Medicare data to identify medical androgen deprivation therapy

Publisher Website
Google Scholar
Abstract
6113 Background: Researchers are increasingly using SEER-Medicare data to measure the use of various cancer therapies. However, the validity and completeness of Medicare data as a source of information for identifying medical androgen deprivation has not been established. In this study, we compared the classification of ADT use based on SEER-Medicare linked data and medical record review. Methods: Among patients who were included in the Prostate Cancer Outcomes Study (PCOS) aged 65 or older and have Medicare part A and B coverage as their primary health insurance, their use of medical androgen deprivation therapy (ADT) identified from the SEER-Medicare data was compared with the results of medical record review. The linkage of the datasets was accomplished by using encrypted IDs. Two algorithms were evaluated. Algorithm A defines ADT use liberally, if at least one of the following HCPCS codes is present: J1950, J9202, J9217, J9218, or J9219. Algorithm B, developed by the authors to identify consistent androgen deprivation, defines ADT only if there is evidence of treatment lasting for 3 months or longer, indicated by either the 12-month treatment J9219 or at least 3 months of treatment with J9202 or J9217. Results: For the 673 patients eligible for this validation study, algorithm A has a sensitivity of 92.5% and a specificity of 85.2% in identifying medical ADT compared to medical record review (see table ), while algorithm B has a sensitivity of 81.8% and a specificity of 88.9% ( Table 1 ). When the study was further restricted to patients with Medicare insurance only (N=142), both algorithms improved: 96.2% sensitivity and 85.6% specificity for algorithm A, and 88.5% sensitivity and 91.1% specificity for algorithm B. Conclusions: Both algorithms have reasonable sensitivity and specificity. Medicare claims can be used to track ADT use for the treatment of prostate cancer in the Medicare population. [Table: see text] No significant financial relationships to disclose.

This publication has 0 references indexed in Scilit: