Background: Physicians routinely factor comorbidities into diagnostic and treatment decisions. Analyses of treatment patterns and outcomes using the National Cancer Data Base (NCDB) usually adjust for comorbidities; however, the completeness of comorbidity ascertainment in the NCDB has never been assessed. We compared the prevalence of comorbidities captured in the NCDB and Surveillance, Epidemiology, and End Results (SEER)-Medicare among female breast, non-small-cell lung, and colorectal cancer patients aged ≥66.
Methods: In the NCDB, ten fields were searched for comorbidities. In the SEER-Medicare dataset, Medicare claims were used to identify comorbidities for two time periods: 12 months prior to diagnosis (Prior) and Index claim alone. Chi-square tests were used to compare comorbidity prevalence using propensity score-matched subsamples from each dataset. Kaplan-Meier survival analyses by Charlson-Deyo comorbidity score and data source were conducted.
Results: Comorbidity prevalence in NCDB did not differ significantly from that identified in SEER-Medicare Index claims across all three cancer sites, except for congestive heart failure, chronic pulmonary disease, and renal disease. However, when compared to the prevalence identified through SEER-Medicare Prior claims, comorbidity prevalence in the NCDB was lower. Overall survival rates by NCDB comorbidity scores were nearly identical to those based on SEER-Medicare Index claims but were lower than those based on SEER-Medicare Prior claims, particularly in higher comorbidity score categories.
Conclusions: The study found overall similarity of comorbidity prevalence between NCDB and SEER-Medicare Index claims, but much less similarity between NCDB and SEER-Medicare Prior claims. Future researchers should understand the limitation of comorbidities ascertained in the NCDB and interpret results accordingly.