Racioethnic differences in missing data in the national cancer database: A window into inequities in cancer care.

Abstract

e18502 Background: While the National Cancer Database (NCDB) is frequently used for observational studies about cancer care including disparities, study populations are often overwhelmingly White. The aim of this study was to assess racioethnic inequities in the NCDB by characterizing the completeness of data in the registry. We hypothesized that Black and Hispanic patients are more likely, compared to White patients, to have missing data about staging, treatment, and survival. Methods: We used the 2004-2016 NCDB files for breast, colon, and lung cancer focusing on non-Hispanic White, Black, and Hispanic patients. The primary outcome was the comparative missingness in data between the three racioethnic groups, measured using multivariable logistic regression models. A stepwise study progression from diagnosis of cancer, histology, clinical staging, treatment documentation, survival, and guideline-concordant care was used. Results: Compared to the U.S. Census data from 2010, Black and especially Hispanic patients were much less likely to be registered in the NCDB compared to Whites. Black and Hispanic patients were also more likely to have missing or non-usable histology data, incomplete or missing clinical staging, and treatment documentation (Table). Survival and follow-up documentation was similar between the three groups. Black and Hispanic patients were less likely to receive surgery for stage I-III disease compared to White patients. Conclusions: Our study suggests that the NCDB may be a limited resource for study of cancer in Black and Hispanic patients, and that inequities in missing data in its own structure may reflect racioethnic inequities in cancer care in the United States. Summary of fraction of missing treatment status, by malignancy, stratified by race/ethnicity in the NCDB. Data are represented as count (percent) unless otherwise noted. The p values are obtained by the Pearson’s chi squared test. Results from a multivariable logistic regression examining the odds of missing treatment status are provided for each malignancy as well. [Table: see text]