Statewide surveillance of colorectal cancer screening using geospatial and Multi-Payer claims data
Katherine E Reeder-Hayes, Bradford E Jackson, Jennifer Lund, Tzy-Mey Kuo, Arrianna Marie Planey, Anisha P Ganguly, Brian J Reich, Daniel S Reuland, Stephanie B Wheeler, Christopher D BaggettAbstract
Background
Screening is a key tool to improve colorectal cancer (CRC) outcomes and mitigate disparities, and reliable estimates of screening variation across communities can help prioritize screening interventions. However, existing screening surveillance infrastructure relies on labor-intensive telephone survey methodology. We aimed to develop novel methods of estimating CRC utilization for small geographic areas using North Carolina multi-payer claims data and to identify patterns and determinants of geographic variation in screening adherence.
Methods
Using data from the Cancer Information and Population Health Resource, a multi-payer claims database, we generated annual screen-eligible cohorts and identified screening procedures. We then used Bayesian spatial models to estimate screening adherence for 1180 zip-code-based small areal units. We used geographic contextual data to re-weight estimates from the insured to the total (uninsured + insured) population. We examined associations between area-level racial composition and screening adherence stratified by age and sex.
Results
Across all years, the observed proportion of person-years up-to-date among insured age-eligible adults statewide was 80.3%; among all age-eligible adults, 75.8% were estimated to be up-to-date. Model-estimated screening adherence in small areas ranged between 43.1% and 93.1%. Increasing Black racial composition was associated with lower screening adherence across all sexes and age strata. Older age and female sex were associated with higher screening adherence.
Conclusions
Model-based estimation of CRC adherence from insurance claims and geographic data is feasible at scale and yields comprehensive estimates at small-area levels. There is wide variation in community-level screening adherence, with many areas falling below public health targets.