Agreement between Surveillance, Epidemiology, End Results (SEER)- and Medicare claims-derived uterine cancer treatment data

Background: Despite prevalent use of SEER-Medicare data to study uterine cancer treatment patterns and survival, concordance between SEER and Medicare claims has not been a focus of prior research. We assessed agreement between SEER and Medicare claims, predictors of disagreement between sources, and associations between treatment (identified in SEER- vs. Medicare) and survival. Methods: Patients diagnosed with uterine cancer between 2000 and 2019 were identified using the SEER-Medicare linked database. We calculated kappa statistics to assess agreement between the two data sources for receipt of hysterectomy with or without bilateral salpingo-oophorectomy (BSO), hysterectomy with BSO, lymphadenectomy, external beam radiotherapy (EBRT), vaginal brachytherapy (VBT), and chemotherapy. For each treatment type, we examined temporal trends in the kappa and used multivariable-adjusted logistic regression to examine predictors of disagreement. Treatment hazard ratios in Cox proportional hazards regression models using treatment information from SEER vs. Medicare were compared. Results: For each treatment, we excluded patients with unknown SEER information, resulting in variable sample sizes. Agreement was lowest for hysterectomy with BSO (kappa=0.71) and highest for lymphadenectomy and chemotherapy (kappas=0.85). Temporal variation was evident, with lymphadenectomy, EBRT, and VBT agreement dropping in recent years. Black race, younger age at diagnosis, high-risk histology, and advanced stage were associated with higher disagreement odds for certain treatments. Associations between treatment identified in SEER vs. Medicare and survival outcomes were similar. Conclusions: Treatment agreement between SEER and Medicare was high. Impact: Our results support use of both data sources for uterine cancer treatment-survival analyses.