Accuracy of diagnostic codes to identify metastatic prostate cancer in medicare claims database

To assess the accuracy of claims-based identification of metastatic prostate cancer (mPC) compared to Surveillance, Epidemiology, and End Results (SEER) registry classification as the gold standard using the SEER-Medicare database.

This retrospective cohort study (N = 72,840) included men aged ≥ 66 years at SEER-recorded PC diagnosis between 2016 and 2019, and continuously enrolled in Medicare fee-for-service Parts A and B for at least 2 months before and after the SEER-recorded PC diagnosis date. Claims-based mPC was identified using the International Classification of Diseases (ICD)-10 diagnosis codes. Using registry-based mPC as a gold standard, accuracy measures were estimated for claims-based mPC. Multivariable logistic regression was used to identify factors associated with discordance between registry-based and claims-based mPC.

Overall, 9.2% (N = 6684) had registry-based mPC and 8.1% (N = 5899) had claims-based mPC at diagnosis. The claims-based method yielded an overall accuracy (95.4%) with specificity (98.0%), negative predictive value (96.9%), sensitivity (68.8%), and positive predictive value (78.0%). In adjusted logistic regression, older men (≥85 years), non-Hispanic Blacks, and those living in low socio-economic areas had higher odds of discordance.

ICD-10 codes in claims data demonstrated high accuracy in identifying mPC at diagnosis among elderly men with prostate cancer, with moderate sensitivity.