Geographically-based cancer control: methods for targeting and evaluating the impact of screening interventions on defined populations.

Successful implementation of cancer control programs depends on efficient targeting to those at highest risk of developing and dying from the disease. This study presents a methodology for targeting cancer screening on the basis of population and disease variation among small geographic areas. Techniques for quantifying the impact of targeting on the predictive value of a positive test are demonstrated, using 329 New York City health areas. Age-truncated crude incidence, late-stage incidence and mortality rates for breast, cervix, and colorectal cancer are used, using site-specific truncation points relevant to the age groups appropriate for screening. Coefficient alpha was used to determine rate stability with 2, 3, 5 and 7 years of data. The stability of most small area rates was found to reach acceptable levels only with 5 and 7 years of data. Targeting into areas where breast cancer prevalence was high increased the expected predictive value of a positive test by as much as 50% when compared with areas of average prevalence. Geographic targeting will be most useful where between-area variability in prevalence is large and within-area variability is small. The implications of these results are discussed and future studies are suggested.