Using census and mortality data to target small areas for breast, colorectal, and cervical cancer screening.

Abstract

The goal of this study was to develop and validate quantitative models for estimating cancer incidence in small areas. The outcome for each cancer site was the incidence of disease that had reached a late stage at the time of diagnosis. Two sets of predictors were used: (1) census-based demographic variables and (2) census-based demographic variables together with the cancer-specific mortality rate. The best models accounted for a substantial percentage of between area variability in late-stage rates for cancer of the breast (46%), cervix (61%), and colon/rectum (58%). A validation procedure indicated that correct identification of small areas with high rates of late-stage disease was two to three times more likely when model-based estimates were used than when areas were selected at random. Additional testing is needed to establish the generality of the geographic targeting methodology developed in this paper. However, there are strong indications that small-area estimation models will be useful in many regions where planners wish to target cancer screening programs on a geographic basis.