Inconsistency of race/skin color data in hospital-based cancer registry: a cancer center cross-sectional study

Keywords

race/skin color - cancer registries - racial disparity

Introduction: Cancer registries are essential for epidemiological analyses, public health planning and quality of care evaluation. Most of its data come from hospital-based cancer registries. In Brazil, following National Cancer Institute and World Health Organization standards, basic sociodemographic characteristics on these databases originate from administrative records. Brazil has a regulation for filling out the race/skin color question in health information system forms since 2017. Racial disparity in cancer care and outcomes is one of the greatest challenges that Brazil faces. A great amount of information and analyses to understand and deal with this disparity comes from secondary data, such as cancer registries.

Objective: To compare self-reported skin color to race/skin color reported in the same patient's hospital cancer registry.

Methods: This is a cross-sectional study of breast cancer patients. A review of electronic medical records was performed including all patients whose treatment for breast cancer began in 2022/23. Skin color self-reported data was obtained from a cancer center retrospective cohort study that analyzed time between diagnostic and first treatment.

Results: We retrieved 176 cases where women self-reported skin color. Mean age was 61, ranging between 27 and 87 years old. Concordant data was seen in 86,9% of the records. In this sample, 40 patients (22,7%) self-reported as black (the aggregation definition of the IBGE Brazilian categories for “preto” and “pardo”), even though only 15,9% was registered as non-white patients. Among the 23 discordant color identification, 56,5% was a misclassification of black patients to white. There was just one patient that was classified the other way around, she self-reported to be white and it was registered as “Pardo”. The discordant data left was inside the black category (“preto” to “pardo” and vice-versa).

Discussion: We found almost 7% of relevant discordance rate. This is different from American and English studies, whose records and self-reported data did not agree on around 5% of cases. The precision of this data is crucial to study disparity in cancer care. Although this is a small, selected sample, probably, it represents the race/skin color information quality on cancer registries. Understanding the source of inconsistency is extremely relevant for improving sociodemographic information in medical records. More Brazilian studies are necessary to confirm these results.

Corresponding author: Alice Zelmanowicz (e-mail: [e-mail]).