Statewide surveillance of colorectal cancer screening using geographic and insurance claims data.

162 Background: Available estimates of colorectal cancer (CRC) screening across small geographic areas (SGAs) in the United States are limited in their breadth and reliability due to dependence on resource-intensive survey methodology. We developed a novel method to estimate CRC screening utilization at SGA level using multiple-payer insurance claims combined with geographic and other public contextual data. Methods: From statewide multi-payer insurance data in the Cancer Information and Population Health Resource (CIPHR) of the University of North Carolina Lineberger Comprehensive Cancer Center, we assembled nine annual cohorts (2012-2020) age >50 with near-continuous insurance coverage (>96 of 120 months with no coverage gap > 90 days) for the prior 10 years. Individuals were assigned to an SGA defined by the intersection of home address ZIP code tabulation area (ZCTA) and county boundary. For each year in each SA, we defined the proportion of eligible individuals with up-to-date CRC screening status defined by recency and modality (e.g. fecal test past 2 years; colonoscopy past 10 years) per national guidelines. Annual estimates were rolled up to a single estimate of the proportion of person-years up to date for each SGA over the whole study period (2012-2020). We estimated mean odds ratios (OR) and 95% credible intervals (CI) from a Bayesian spatial binomial regression model to examine the association between SGA % Black residents (2010 Census) and SGA screening utilization. Stratification was used to adjust for sex and age (i.e. 50-55, 56-60, 61-65, 66-70, and 71-75). Results: Among 1,545,696 age-eligible individuals without prior CRC diagnosis enrolled in a participating payer in 2020, 23% (n = 349,289) were evaluable based on coverage continuity; the proportion evaluable was similar in other years. Among 1180 small geographic areas across North Carolina, the median area-level estimated proportion of up-to-date person time for insured residents was 78.4% (range: 43%-93%). We observed negative associations between higher SGA Black racial composition and the proportion of person years up to date across all age groups, with the strongest associations among the oldest individuals: OR male,71-75 = 0.34 (95%CI: 0.26, 0.45), OR female,71-75 = 0.40 (95%CI: 0.31, 0.51). Ongoing work will use supplemental data sources to transform insured-population estimates of screening adherence to whole-population estimates, and to examine area-level health services determinants of screening uptake. Conclusions: Use of multi-payer insurance claims data to produce granular area-level estimates of CRC screening utilization is feasible at scale in a large and geographically diverse state. Higher Black racial composition was strongly associated with lower estimated CRC screening utilization, even with continuous insurance coverage, and may be an appropriate focus for community-level intervention.