Cancer
Measuring completeness of reporting of US cancer registries Lindsay J Collin* Lindsay Collin Richard F. MacLehose Thomas P. Ahern Rebecca Nash James Baurley Kevin C. Ward Timothy L. Lash
Background: Cancer surveillance is crucial for cancer control and research. Cancer registry completeness is a measure of the proportion of true cases captured by a registry. Gold star certification of a registry requires a completeness estimate of ≥95%. When completeness is less than 100%, the incident cancer rates are underestimated. If completeness is differential by demographic groups, then information on cancer health disparities will be biased. Current methods for measuring completeness do not include data from the most common cancers, all racial and ethnic groups, or account for regional variability. We propose a new, comprehensive method to assess completeness.
Methods: We obtained county-level cancer incidence and mortality data by sex, age, and race from the California Cancer Registry. We fit Bayesian joint hierarchical Poisson models for county-level incidence and mortality rates with a shared conditional autoregressive error structure, which allowed information borrowing across counties. Predicted incidence and mortality rates were estimated for each county to generate county-specific Incidence-to-Mortality Rate Ratios (IMRRs). We decomposed the overall registry completeness estimate into geographic and demographic completeness estimates.
Results and Future Directions: Completeness estimates were assessed by region (Northern, Central, and Southern California), county, race (Black, White) and natal sex. Region-specific completeness IMRRs were 99%, 103%, and 94% for Southern, Central, and Northern regions, respectively. Figure 1 highlights county-specific estimates below 95% completeness. In both the Southern and Central regions, Black women had substantially lower levels of completeness than any other group, falling below the gold certification threshold. Our method highlights differential completeness, which allows registries to target efforts for improvement. Future work will assess completeness for the entire US and across demographic groups.
