National Trends in Active Surveillance for Prostate Cancer: Validation of Medicare Claims-based Algorithms

Objective: To better describe the real-world use of active surveillance. Active surveillance is a preferred management option for low-risk prostate cancer, yet its use outside of high-volume institutions is poorly understood. We created multiple claims-based algorithms, validated them using a robust clinical registry, and applied them to Medicare claims to describe national utilization.

Materials And Methods: We identified men with prostate cancer from 2012-2014 in a 100% sample of Michigan Medicare data and linked them with the Michigan Urologic Surgery Improvement Collaborative (MUSIC) registry. Using MUSIC treatment assignment as the standard, we determined the performance of 8 claims-based algorithms to identify men on active surveillance. We selected 3 algorithms (the most sensitive, the most specific, and a balanced algorithm incorporating age and comorbidity) and applied them to a 20% national Medicare sample to describe national trends.

Results: We identified 1186 men with incident prostate cancer and completely linked data. Eight algorithms were tested with sensitivity ranging from 23.5% to 88.2% and specificity ranging from 93.5% to 99.1%. We found that the use of surveillance for men with incident prostate cancer increased from 2007 to 2014, nationally. However, among all men in the population, there was a large decrease in the rate of prostate cancer diagnosis and an increased or stable rate in the use of active surveillance, depending on the algorithm used. Less than 25% of men on active surveillance underwent a confirmatory prostate biopsy.

Conclusion: We describe the performance of claims-based algorithms to identify active surveillance.