Despite the promise of geocoding and use of area-based socioeconomic measures to overcome
the paucity of socioeconomic data in US public health surveillance systems, no consensus
exists as to which measures should be used or at which level of geography. The authors
generated diverse single-variable and composite area-based socioeconomic measures
at the census tract, block group, and zip code level for Massachusetts (1990 population:
6,016,425) and Rhode Island (1990 population: 1,003,464) to investigate their associations
with mortality rates (1989-1991: 156,366 resident deaths in Massachusetts and 27,291
in Rhode Island) and incidence of primary invasive cancer (1988-1992: 140,610 resident
cases in Massachusetts; 1989-1992: 19,808 resident cases in Rhode Island). Analyses
of all-cause and cause-specific mortality rates and all-cause and site-specific cancer
incidence rates indicated that: 1) block group and tract socioeconomic measures performed
comparably within and across both states, but zip code measures for several outcomes
detected no gradients or gradients contrary to those observed with tract and block
group measures; 2) similar gradients were detected with categories generated by quintiles
and by a priori categorical cutpoints; and 3) measures including data on economic
poverty were most robust and detected gradients that were unobserved using measures
of only education and wealth.