In the current analysis, data were drawn from the 1984 National Health Interview Survey's Supplement on Aging (1984 NHIS-SOA), the 1986 Area Resource File System (ARF), and 1986 Medicare Enrollment Statistics.
The 1984 NHIS-SOA is a multistage area probability sample which provides self-reported characteristics for 11,497 civilian noninstitutionalized elderly (age 65 and over). It includes information on their family structure, living arrangement, social support, conditions and impairments, functional abilities (ADL and IADL), and other health-related and social information.
To develop the regression models, contextual variables from the ARF were attached to individuals on the NHIS-SOA using geographic markers. The ARF is a compilation of county and other geographic area statistics concerning a wide range of health planning related variables drawn from a multitude of survey sources. Using the geographic identifiers available on the 1984 NHIS-SOA, corresponding community data were attached at the Standard Metropolitan Statistical Area (SMSA) for individuals residing in one of 31 large self-representing SMSAs. Individuals on the data set who resided outside these 31 areas were assigned the corresponding regional (northeast, north central, south or west) and urbanity (SMSA or nonSMSA) average for their type of residence. The result was 39 distinct geographic areas: 31 self-representing SMSAs, and 4 urban and 4 nonurban regional areas.
To generate regression-adjusted synthetic estimates of the functionally dependent elderly population in an area, rates of dependency produced by the model on national data must be multiplied by population data from small areas. Any explanatory variable included in the national model must also be available in the small area population data. As intercensal age, sex and race specific population data for the elderly are not readily available in small age increments at the small area level, we used Medicare Enrollment data for our estimates. Necessary adjustments to the Medicare data to account for nonenrollment among the elderly, and for the proportion of the elderly residing in nursing homes are discussed later in the report.