O ur research group 1 has demonstrated disparities in stage at diagnosis for melanoma based on race and ethnicity and on health care delivery system. To further document these disparities, we examine herein data for Miami-Dade County, Florida.Miami-Dade County, the most populous of Florida's 67 counties, has more than 2.5 million residents. The socioeconomic status (SES) characteristics of this county include factors that may affect health care access and utilization, such as a diverse racial and ethnic population, a large indigent population, and large percentages of foreign-born individuals. 2 We sought to determine if, within the county, differences exist in stage of melanoma diagnosis and if so, what factors influence those differences.Methods. Incidence data on melanoma as the primary site of diagnosis for the 5-year period from 1997 through 2001 were extracted from the Florida Cancer Data System (FCDS), 3 Florida's statewide, population-based cancer incidence registry. In the FCDS, stage at diagnosis is coded according to the summary staging system used by the SEER (Surveillance Epidemiology and End Results) program. 4 For the present analysis, in situ and local stage diagnoses (stages 0 and 1) were considered early, while regional and distant or systemic diagnoses (stages 2-4) were considered late. Cases with unknown stage were excluded.We used zip code of residence for each case to subdivide the county. Zip codes were used rather than census tracts because they represent larger populations and numbers of melanoma cases, although population sizes of zip code areas vary slightly. 5 Zip codes are also recognizable to the general public, planners, and policy makers, and there is precedence for using zip codes in similar evaluations. [6][7][8] Zip code population data were obtained from the US Census 2000 9 for population and housing, summary file 3, technical documentation 2002. Ten SES measures were extracted from this source for each of the 76 residential zip codes in Miami-Dade County (except for item 6, which was extracted from the Florida Agency for Health Care Administration 10 ) (Table ). We also developed zip code maps of Miami-Dade County using ArcMap, version 8.3 (Esri, Redlands, California) for each of the 10 SES measures.Using data aggregated by zip code, we computed 2-tailed Pearson correlation coefficients using SPSS, version 15.0 (IBM Corporation, Somers, New York) to examine SES factors that were correlated with the percentage of mela-noma late-stage diagnoses. Multiple logistic regressions were performed to further examine how the SES variables were related to stage at diagnosis, coded as either early or late. The analysis was performed on individual case data. The SES measures used were for the zip codes in which the individual lived, not for the individual patient. Ecologic analyses are commonly used when socioeconomic data for individual patients are not available.