Use of Dasymetric Areal Interpolation to Estimate the Burden of Diabetes Hospitalization in Chicago Neighborhoods

Tuesday, June 11, 2013: 4:30 PM
106 (Pasadena Convention Center)
Roderick C. Jones , Chicago Department of Public Health, Chicago, IL
Chieko Maene , University of Chicago, Chicago, IL
Monica Peek , University of Chicago, Chicago, IL
Elbert S. Huang , University of Chicago, Chicago, IL
BACKGROUND:  In Chicago, the geographic unit used most frequently to delineate neighborhoods for the purposes of monitoring health status and identifying and solving localized health problems is community area. The Illinois hospital discharge dataset is a rich source of data on the burden of chronic disease, but patient records are geocoded to residential ZIP code. These two geographies overlap and do not have common boundaries.

METHODS:  To develop the procedure we analyzed hospital discharges in 2010 for residents of Chicago ZIP codes with a first-listed diagnosis of diabetes. US Census 2010 files were used for block-level sex-by-age population counts, and block-to-ZIP and block-to-community area relationships were determined through geographic information systems. Using SAS, a dasymetric areal interpolation technique was implemented to calculate age and sex-specific discharge rates, apply the rate to block-level counts, sum imputed counts to the community area level, and calculate crude and age-adjusted rates.

RESULTS:  There were 58 residential ZIP codes, with areas ranging from <0.1 to 17.6 square miles (median, 3.8) and populations ranging from 493-113,916 (median, 45,962). The 77 community areas had areas ranging from 0.6-13.5 square miles (median, 2.8) and populations ranging from 2,876-98,514 (median, 31,028). As many as 7 ZIP codes contributed to the imputed for a single community area. Three ZIP codes contain substantial population that resides outside the city border. Choropleth maps of crude and age-adjusted rates using both the ZIP code and imputed community area data illustrate spatial interdependence, with similar patterns evident despite different geographic unit boundaries. By ZIP code, discharge counts ranged from 0-393 (median, 109), crude rates per 10,000 ranged from 0-57 (median, 18) and age-adjusted rates ranged from 0-57 (median, 19). By community area, imputed discharge counts ranged from 12-462 (median, 76), imputed crude rates per 10,000 ranged from 7-58 (median, 26), and imputed age-adjusted rates ranged from 9-56 (median, 29).

CONCLUSIONS:  We created a reproducible method to impute hospital discharge rates for Chicago community areas. This allows for comparison and discussion of disparities according to the geographic unit most-frequently used in our city. Due to frequent changes in ZIP code boundaries, additional methods must be developed to analyze data pertaining to intercensal years. Current work is on the development of an alternate approach, which incorporates race and ethnicity into the imputation procedure.