Diabetes is recognized as a growing national and international epidemic as prevalence rates for other major chronic diseases such as stroke and heart disease have decreased (¹,²). The challenges of addressing the epidemic are exacerbated by the disparities in the prevalence of the disease. These disparities are complicated by quality-of-care issues and socioeconomic determinants (^3-7), which may include local geographically clustered factors such as availability and access to health care, education and employment opportunity, and social capital and social cohesion. Furthermore, variations in the prevalence of diabetes at the local level are largely unreported, making understanding the disparities associated with the disease more difficult.

Duval County is a consolidated city/county government located on the northeast coast of Florida. It is a large (more than 840 square miles) and diverse area that has a population of more than 900,000 (⁸). The city/county contains areas that reflect urban, suburban, and rural areas. In 2007, Duval County, which encompasses Jacksonville, had an age-adjusted diabetes death rate of 32 per 100,000, compared with the 10 other largest counties in Florida (range, 14-29; median = 20) (⁹). In 2007, the total hospitalization costs for adult diabetes-related treatment in Duval County exceeded $714,000,000, and the cost for emergency department (ED) visits due to diabetes-related treatment was more than $57,000,000 (¹⁰). The growing disparity in diabetes deaths between Duval County and the state of Florida as a whole has been an alarming trend during the past 15 years (Figure 1).

We used several methods to study the local prevalence of diabetes, including the use of administrative data for the number of hospital and physician visits for diabetes (¹¹,¹²) and Behavioral Risk Factor Surveillance System (BRFSS) survey self-reported data (¹³). Both methods rely on health system diagnosis, either documented through administrative records or communicated to patients. We used a method of weighting prevalence rates to adjust for undiagnosed cases (¹⁴). Diabetes death rates and various measures of diabetes prevalence capture different forms of observable characteristics or effects of the disease. We assessed the comparable sensitivity of these measures, particularly as the measures relate to geographic distribution of ethnicity and social determinants, and analyzed diabetes-related disparities at the local level by using different sources of data to provide implications for public health and health care policy.

We used a secondary data analysis research design that included multiple sources to assess the prevalence and effect of diabetes in Duval County. Data sources were ED and hospital discharge data for the year 2007 reported to the Florida Agency for Health Care Administration (AHCA), vital statistics data for the year 2007 reported to the Florida Department of Health, 2007 BRFSS data collected by the Florida Department of Health, population data collected by the US Census Bureau and census estimates generated by the Florida Office of Economic and Demographic Research and Nielsen Claritas, and previously created geographically defined areas, identified as health zones.

Diabetes rates vary extensively in Duval County, depending on the source of data and the type of measure. Figure 3 shows the relationship of these measures in descending order, ranging from an estimated diabetes prevalence of 12,371, per 100,000 population to a death rate of 40 per 100,000.

Overall, residents from health zone 1 are less educated and poorer than residents from the other health zones, and health zone 1 has a higher African American population than the other health zones (Table 1).The extensive local variations for these different measures are illustrated in Table 2, which shows major disparities in the county. Health zone 1, the urban core, had an age-adjusted death rate of 93.5 compared with the lowest rates in the county, health zones 2, 3, and 6 with rates of 30.5, 31.0, and 31.6, respectively. These 3 rates were lower than the county rate of 39.9 and the state rate of 34.9. The other health zones (4 and 5) had rates that were less than half of the urban core rate. The rate of age-adjusted diabetes death for adults varied dramatically by health zone. Health zone 1 had more than double the rate of health zone 5, which had the next highest rate, and more than triple the 2 lowest rates (health zones 2 and 3).

Rates for hospitalization and ED visits revealed even more profound disparities in terms of location. The hospitalization rate for the urban core (health zone 1) was 747 compared with 148 for the health zone with the lowest rate (health zone 3). The urban core hospitalization rate was more than double that of all other health zones but 1. The distribution of diabetes rates for ED use was similar in that the rate of health zone 1 far exceeded those of the other health zones. Health zone 1 had an unusually high ED visit rate (692) compared with health zone 3, which had the lowest rate (105) and had more than twice the rate of the other health zones. Health zone 1, which had a rate of self-reported diagnosed diabetes of 14,250, exceeded the county rate, but this is the only measure for which another health zone (health zone 5, rate of 15,446) exceeded the urban core (ie, health zone 1) (Table 2).

The ratios of prevalence and relative risk in health zone 1 compared with the rest of the county for each diabetes measure complemented the GIS analysis, providing markers of significance for the disparities in the county (Table 3). Significant differences between health zone 1 and the rest of the county (health zones 2-6) were established for each of the diabetes measures. The largest difference in ratios was for diabetes ED use, followed by diabetes-related ED use. The health zone 1 ratio for hospitalization for diabetes and diabetes-related illness were also high compared with the other zones. The ratios for diabetes deaths and diabetes-related deaths were also comparatively high for health zone 1. The ratio of prevalence of diabetes for health zone 1 compared with the other health zones was the lowest ratio.

The results of our study show that the diabetes prevalence ratios within the high-minority, low-socioeconomic area of Duval County were statistically different when compared with the other parts of the county. Understanding the effect of the disease and the distribution of that disease in the community has implications for policy development and resource allocation. The costs associated with hospitalization and ED use are much higher in the high-minority, low-socioeconomic part of the county. The cost per capita of diabetes-related hospitalizations in health zone 1 in 2007 was $2,010, which was nearly double the cost per capita for the county ($1,059). The charity cost per capita of diabetes-related hospitalizations in health zone 1 in 2007 was $1,053, which was more than double the cost per capita for the county ($465). The reasons for the acute disparities identified by this study deserve considerably more discussion than is feasible here, but they include a range of socioeconomic and health care disparities (^17-29).

The results of this study provide insights about the distribution of diabetes in specific areas of the county, insights that get lost in data aggregated at the metropolitan level. An unexpected result of our study was the low rate of diagnosed cases of diabetes, which were inferred from BRFSS data. This could be due to a lack of access to prevention and primary care for people in health zone 1, resulting in poorer outcomes related to delayed care, which are reflected in the other measures such as higher rates of hospitalization and death, as previously discussed. However, it may also reflect flaws in BRFSS methods related to low participation of African Americans in the BRFSS telephone surveys, which is exacerbated by declining land-line use. Although weighting is used to compensate for underrepresentation, it may not adequately address disproportionate underrepresentation of the highest-risk patients among African Americans.

Our study has limitations that are associated with most efforts to measure disease, illness, and death. The accuracy of the data is dependent on the people observing and recording the data and may be affected by the data collection process. Another limitation is that BRFSS data use sampling frames and telephone interviews that have inherent issues with sampling bias, particularly when refusals and land-line issues are considered. However, examining multiple sources of data is beneficial because together these sources provide a more accurate picture of disease effect, similar to the concept of triangulation found with qualitative research.

Currently allocated resources may be insufficient or inappropriate to adequately deal with diabetes and its complications in the areas of highest need. Health zone 1 is the urban core, which has the lowest socioeconomic levels. Areas with the highest prevalence of diabetes contain the patients who have the fewest resources to deal effectively with the disease. This disparity may account for the disproportionate number of hospital and ED visits, which drive up the cost of health care for the poorest because of a lack of adequate preventive resources.

Our analyses revealed that diabetes disproportionately affects the geographic part of the community that has the highest minority population and the lowest socioeconomic status. The most sensitive measures of the effects of diabetes at the local level were hospitalization data and ED use, and the least sensitive measure was prevalence, determined from BRFSS data. Our analyses show that diabetes-related disparities affect not only people and their families, but also the community that absorbs the costs associated with the disproportionate health care use that results from these disparities. Analyzing data at the subcounty level has implications for health care planning and public health policy development at the local level.

The opinions expressed by authors contributing to this journal do not necessarily reflect the opinions of the US Department of Health and Human Services, the Public Health Service, the Centers for Disease Control and Prevention, or the authors' affiliated institutions. Use of trade names is for identification only and does not imply endorsement by any of the groups named above. URLs for nonfederal organizations are provided solely as a service to our users. URLs do not constitute an endorsement of any organization by CDC or the federal government, and none should be inferred. CDC is not responsible for the content of Web pages found at these URLs.

Suggested citation for this article: Livingood WC, Razaila L, Reuter E, Filipowicz R, Butterfield RC, Lukens-Bull K, et al. Using multiple sources of data to assess the prevalence of diabetes at the subcounty level, Duval County, Florida, 2007. Prev Chronic Dis 2010;7(5) http://www.cdc.gov/pcd/issues/2010/sep/09_0197.htm. Accessed [date].

The weighting formula for the data was adapted from the Behavioral Risk Factor Surveillance System method for calculating "FINALWT" (www.cdc.gov/BRfss/technical_infodata/weighting.htm). A variable, "HZFINALWT," was created, which was the final weight assigned to each respondent. It was obtained by replacing "POSTSTRAT" with "ZONEPOSTRAT" in the formula of "FINALWT." It was calculated by multiplying 2 variables, "WT2" and "ZONEPOSTSTR." The variable "WT2" (equals STRWT × 1/NPH × NAD) is precomputed by Centers for Disease Control and Prevention and takes into account the number of adults in the respondent's household, the inverse of the number of residential telephone numbers in the respondent's household, and the differences in the basic probability of selection among strata. The variable "ZONEPOSTSTR" was created to account for zone, age, race, and sex of the respondent and is equal to the number of people in a health zone-by-age-by-race-by-sex category divided by the sum of the products of the preceding weights for the respondents in that same health zone-by-age-by-race-by-sex category (³⁰).