Bias in spatial analyses that overlook compositional and contextual factors of communities can be substantial. We first examined spatial patterns among 11,728 prostate cancer cases across Connecticut, 1994-98. A spatial scan statistic (SatScan™) identified two locations where average annual incidence rates significantly exceeded the statewide level and two locations with significantly lower disease rates. Extending the analysis to adjust rates for age and race/ethnicity greatly minimized, but did not eliminate, geographic variation. Adjustment for age and poverty level of communities eliminated significant variability across locales. Similarly, analysis adjusted for age and covariation of colorectal cancer incidence rates across the state accounted for all significant variation previously observed. These results suggest that accounting for a "detection effect" due to clinical patterns of another screenable condition may be as useful as adjusting spatial data for variability of socio-economic conditions.
Keywords: BRFSS; Behavioral Risk Factor Surveillance System; CTR; Connecticut Tumor Registry; EXP; Geographic analysis; NCI; National Cancer Institute; OBS; PSA; Prostate cancer incidence; Prostate-Specific Antigen; RR; SEER; Spatial co-variation; Surveillance Epidemiology and End Results; expected; observed; relative rate/risk.
Copyright © 2013 Elsevier Ltd. All rights reserved.