Inferring Unknown Race in Central Cancer Registries

Area of Science:

• Public Health
• Biostatistics
• Cancer Epidemiology

Background:

• Completeness of race information is crucial for cancer registry data certification in the US.
• Missing race data can lead to inaccuracies in cancer incidence rates and disparities analysis.

Purpose of the Study:

• To present and evaluate the Bayesian Improved Surname Geocoding (BISG) method for reducing unknown race information in cancer registries.
• To assess the accuracy and effectiveness of BISG in improving race-specific cancer incidence data.

Main Methods:

• Utilized the Bayesian Improved Surname Geocoding (BISG) method, a technique with over 15 years of use in social sciences and public health.
• Employed North Carolina voter rolls as a proxy for cancer patients, sampling to mirror national race/ethnicity distributions.
• Integrated BISG into freely available computer code for widespread application.

Main Results:

• Achieved a reduction in unknown race information by up to 75%.
• Demonstrated high accuracy (95%), with race-specific sensitivity ranging from 81-99% and positive predictive value from 88-97%.
• Showed potential to increase accuracy of race-specific cancer incidence rates by up to 3% in registries with high missingness.

Conclusions:

• The BISG method offers a validated approach to significantly improve the completeness and accuracy of race information in cancer registries.
• Accurate race data is essential for reliable cancer surveillance, research, and targeted public health interventions.
• The availability of BISG in computer code facilitates its adoption and enhances the quality of cancer data nationwide.