Background: A prior assessment of concordance between the diagnosis month in SEER records and Medicare claims found reasonable agreement; however, no assessment of the impact of discordance on cancer treatment ascertainment was conducted.
Objectives: The aim of this study was to assess the concordance between the SEER diagnosis date (Sdx) and Medicare claim–derived diagnosis date and the impact of discordance on identification of treatment received.
Methods: The first Medicare claim date with a cancer diagnosis (Mdx) was compared with the Sdx among patients diagnosed with breast, colorectal, or lung cancer. The Mdx was considered concordant with the Sdx if the Mdx was within 16 days. Claims within 4 months after both the Mdx and Sdx were examined to collect treatment information. Treatment rate agreement was measured by κ-statistics.
Results: Among 50,731 breast, 51,025 colorectal, and 61,384 lung cancer patients, the Sdx and Mdx were concordant in 79%, 86%, and 73% of cases, respectively. Most discordant Mdx cases were identified in the month after the SEER diagnosis month. A small proportion of cases (7%–12%) preceded the SEER diagnosis month. Agreement for receipt of surgery was very good across all 3 cancer sites (κ>0.88) and was excellent for radiation therapy (κ>0.96).
Conclusions: Although most cases were concordant for both diagnosis date and treatment ascertainment, there was still a small proportion of cases discordant for both diagnosis date and treatment identification. This study underscores the importance of examining claims in the months preceding diagnosis in the SEER-Medicare dataset to ensure patients are appropriately selected for analysis.