Examining geospatial clustering of late stage colorectal cancer

Background: Colorectal cancer (CRC) mortality is the third leading cause of cancer-related death in the US. CRC screening compliance has been historically lower in poor and minority populations. This study modeled the probability of a CRC case being diagnosed in an area at high risk of late-stage CRC using county-level CRC screening data and neighborhood SES characteristics. Methods: CRC cases diagnosed 2005-2009 and geocoded to 2010 Census boundaries were obtained from Florida Cancer Data System (n=52,152). County level screening rates were obtained from the 2010 Florida BRFSS. Census tract level SES data were obtained from the 2010 American Community Survey. SaTScan spatial scan software was used to detect late-stage clusters for Blacks, Cubans, Hispanic Whites (HW), and non-Hispanic Whites (NHW). Logistic regression was performed to describe risk factors predictive of being diagnosed in an area at high-risk for late-stage CRC diagnosis. Results: The relationship between neighborhood poverty and late-stage cluster membership was non-linear and varied across models. For example, Blacks showed a positive dose-response relationship with poverty level except those living in the most impoverished neighborhoods were significantly less likely to be diagnosed within a cluster (p=0.002). Increased segregation by ethnicity was protective for Hispanics and a risk for Blacks and NHW. In general, increased segregation by race was a risk for living in a cluster of late-stage CRC diagnoses. Screening was inconsistently protective across models. Conclusion: Unanticipated associations between late-stage/poverty and late-stage/screening may be a ‘screening effect' driven by current cancer control efforts aimed at Florida's high-risk populations.