Women's Heart Disease Atlas: Appendix B

Methodological and Technical Notes

Source: Casper ML, Barnett E, Halverson JA, Elmes GA, Braham VE, Majeed ZA, Bloom AS, Stanley S. Women and Heart Disease: An Atlas of Racial and Ethnic Disparities in Mortality, Second Edition. Morgantown, WV: Office for Social Environment and Health Research, West Virginia University, 2000.
Note: Information is presented as at the time of publication. Some reference, resource and contact information may not be current.

A. County Definitions

1. Overview

Data from several different sources were used in this publication, and one of our chief methodological concerns was ensuring comparability of county definitions across datasets. We used the Federal Information Processing Standard (FIPS) codes to link county definitions across datasets, and to reconcile differences. For the majority of states, there was 100 percent comparability in county definitions among all the datasets used. Details about modifications to county definitions for specific states appear below. 

The following cities were retained as independent cities and the FIPS codes were modified to conform to the geographic database. 

2. Alaska 

In the Area Resource File (ARF), Alaska was treated as a single geographic unit. The ARF did not provide data for the Alaska county equivalents. Therefore, for each of the maps that present data from the ARF, we were unable to map data for Alaska. These maps include the following: 

Total Population per Cardiovascular Disease (CVD) Physician, 1990
Total Population per Cardiac Intensive Care Unit (CCU) Bed, 1993
Cardiac Rehabilitation Units, 1993
Local Economic Resources, 1990 

Due to differences in county coding over time, and differential coding among the various data sources, several other changes were also made to county FIPS codes. The coding changes are indicated in the following tables. 

3. Arizona

4. Hawaii

5. Virginia 

Virginia is comprised of counties and independent cities that are treated as county–equivalents in many datasets. However, not all of the datasets we used contained data for the Virginia independent cities. Many of these cities are also difficult to represent on a map because of their small land area. Therefore, the spatial geometry for most of Virginia independent cities was removed from the geographic database and data for those cities was collapsed into those counties with which they are most geographically associated. We followed the conventions of the 1996 Area Resource File. The changes made to FIPS codes to combine Virginia independent cities with their surrounding or adjacent counties are shown in the table below.

5. Yellowstone National Park

Back to top

B. Data Sources

1. Economic Resources Data

Data for the Index of Local Economic Resources were obtained from the Area Resource File (February 1996 edition) — a compilation of health–related data that have been abstracted from multiple data sources by the Bureau of Health Professions, Department of Health and Human Services. The three variables that were used to create the index were abstracted from the 1990 Census of Population and Housing, STF3A data files. The Index of Economic Resources was based on three dimensions of the local socioeconomic infrastructure: median family income, occupational structure, and unemployment rate. Occupational structure was defined as the percent of all employed persons who were engaged in white collar jobs (i.e., managerial and professional specialty occupations and technical, sales, and administrative support jobs). 

The index was calculated by ranking all counties separately for each variable. For each variable, the counties were then categorized into deciles, and each decile was assigned a score ranging from 0 to 9. Counties in the decile with the poorest economic conditions (lowest median income, lowest occupational structure, highest unemployment rate) were assigned a 0 and counties in the decile with the most advantaged economic conditions were assigned a 9. For each county, the scores from the three variables were added together to arrive at the index score. The range of the score is from 0 (counties that were in the lowest decile for all three dimensions of the Index) to 27 (counties that were in the top decile for all three dimensions of the Index). The distribution of index values across all counties was then divided into five groups with roughly equal ranges of index values. 

2. Heart Disease Mortality Data 

Death certificate data for the years 1991–1995 were obtained through the National Vital Statistics System maintained by the National Center for Health Statistics. Deaths from heart disease were defined as those for which the underlying cause of death listed on the death certificate was coded according to the International Classification of Diseases—9th Revision (ICD–9) as: 390–398, 402, 404–429. These codes comprise the category 'Diseases of the Heart' as defined by the National Center for Health Statistics.¹ For each decedent, underlying cause of death, age, race/ethnicity, gender, and county of residence at the time of death were abstracted from computerized death certificate files. Information on Hispanic ethnicity was not collected on death certificates in Oklahoma throughout the 1991–1995 study period, and prior to 1993 was not collected for New Hampshire. Consequently, we could not analyze decedents of Hispanic ethnicity for Oklahoma and New Hampshire. 

3. Medical Care Resources Data 

Data on medical care resources were obtained from the Area Resource File, (February 1996 edition) a compilation of health-related data abstracted from multiple data sources by the Bureau of Health Professions, Department of Health and Human Services. Maps were created for the following indicators of medical care resources relevant to secondary prevention of heart disease mortality: population per cardiovascular disease specialty physician, population per coronary care unit bed, and number of cardiac rehabilitation units. The primary source for the data on cardiovascular disease physicians was the American Medical Association Physician Master File. The primary source for the data on coronary care unit beds and cardiac rehabilitation units was the County Hospital File for 1993. 

Rather than map the number of physicians per county, we chose to map the ratio of county population size to each cardiovascular specialty physician. This approach provides a better comparative measure of the availability of physicians when examining counties with large populations vs. counties with small populations. Similarly, we chose to map the ratio of county population size to each coronary care unit bed. Because cardiac rehabilitation units are intended to serve more than one individual at a time, we mapped the total number of cardiac rehabilitation units in each county. 

4. Population Data 

Population count data for all counties in the U.S. were obtained from the Bureau of the Census for the years 1991–1995. These intercensal estimates were calculated by the Bureau of the Census through extrapolation of linear trends in population growth and inter-county migration patterns between census years 1980 and 1990.

5. Social Isolation of Women Data 

Data on several dimensions of women’s social isolation were obtained from the 1990 Special Tabulation on Aging compiled by the Bureau of the Census. This dataset contains a variety of variables, abstracted from the 1990 Census of Population and Housing, for elderly women and men. We examined data for women aged 60 years and older. The majority of heart disease deaths for women aged 35 years and older actually occur to women aged 60 years and older, because of the strong association between increasing age and heart disease risk. 

Two indicators of women’s social isolation were mapped for this atlas: percent of women living alone, and the percent of women with either mobility or self-care limitations. Living alone was defined as an individual living in a household without a spouse or other family members or friends. A mobility limitation was defined as a health condition, either physical or mental that lasted for 6 or more months, which made it difficult to go outside the home alone. A self-care limitation was defined as a health condition, either physical or mental that lasted for 6 or more months, which made it difficult to take care of personal needs, such as dressing, bathing, or getting around inside the home.

To produce the maps of women’s social isolation, we first excluded 32 counties with fewer than 100 women over the age of 60 years old in 1990. For each of the two measures of women’s social isolation, the range between the 1st and 99th percentiles of the distribution of the percentage values was divided into 5 equal categories. Counties below the first and above the 99th percentiles of the distribution were included in the lowest and highest categories respectively. These five categories provided the cutpoints for mapping.

Back to top

C. Map Projections

1. National Maps

To facilitate the presentation of information for all U.S. counties, several different map projections were used. For the coterminous United States an Albers-Conic Equal Area projection was used. Alaska was projected to the Miller Cylindrical projection and Hawaii is presented using geographic coordinates (latitude and longitude). Neither Alaska nor Hawaii is to proper geographic scale relative to the continental United States. The combinations of projections and scales allowed the presentation of a relatively familiar orientation of these geographic features. 

The coordinate information for the contiguous United States was projected using the Albers Equal-Area projection with the following parameters:

The coordinate information for Alaska has been projected using the Miller Cylindrical project with the following parameters:

2. State Maps 

All state maps were projected using the State–Plane projection systems of each state. The state maps are presented to maximize the reader's ability to interpret results for each state and are therefore not to proper geographic scale relative to one another. However, State–Plane coordinate systems are commonly used by state agencies and therefore their use here maximizes the reader's ability to compare these maps with other information. 

Many states did not have significant populations of women of particular racial and ethnic groups. In many cases racial and ethnic specific rates could not be calculated for any of the counties within the state. Rather than present blank maps for these states, we elected only to generate race and ethnicity–specific state maps if there were at least two counties with heart disease mortality rates for any given racial and ethnic group.

Back to top

D. Race and Ethnicity Definitions 

The race and ethnicity categories used in Women and Heart Disease were defined according to Office of Management and Budget, Directive 15,2 and are not based upon biological or anthropological concepts. The categories were developed in response to needs for collecting standardized data to be used by federal agencies for record keeping, collection and presentation of data (i.e., Federal surveys, the decennial census and monitoring various civil rights laws). 

According to the Office of Management and Budget, the federal agency that defines standards for government publications, there are six minimum categories for race and ethnicity classification (listed below). Hispanic or Latino is considered a designation of ethnicity, not race, and people of Hispanic or Latino origin may be of any race.

Back to top

E. Spatial Geometry 

The geographic database, which includes spatial geometry and attribute information for all U.S. counties, was obtained from Environmental Systems Research Institute's (ESRI) ArcUSA database. ESRI has modified source data from the 1973 Digital Line Graph (DLG) data produced by the U.S. Geological Survey to improve the currency of the county boundary information to 1988. The geographic scale of the spatial geometry (linework) is 1:2,000,000, and is sufficient to identify major county features. Mortality, population, socioeconomic, and medical resource data were linked to county geography using the Federal Information Processing Standards (FIPS) codes. 

Back to top

F. Spatial Smoothing of Heart Disease Death Rates 

1. Spatial Smoothing Methods

Heart disease death rates were calculated for women 35 years and older for the period 1991–1995. Separate rates were calculated for the following population groups: all women, American Indian and Alaska Native women, Asian and Pacific Islander women, black women, Hispanic women, and white women. For each population group, a smoothed death rate for heart disease, based on a spatial moving average, was calculated for each county. 

For each county, heart disease deaths (numerators) and population counts (denominators) for ten–year age groups (e.g., 35–44 years old, 45–54 years old, etc.) were summed for the five–year study period 1991–1995. County numerators and denominators were then summed together with death count numerators and population count denominators of all neighboring counties, and then divided by the number of neighbors plus one to produce an average rate. "Neighbors" were defined based solely on contiguity (as opposed to distance). This process produced spatially smoothed age–specific (by 10–year age group) heart disease death rates. The spatially smoothed age–specific heart disease death rates were then directly age–adjusted to the 1970 United States population, for the age range 35 years and older. 

Two constraints were applied to the calculation of county–level heart disease death rates for each race and ethnicity group. For a particular population group (e.g., Latina women aged 35 years and older), a heart disease death rate was not calculated for any county for which the total number of deaths in that county plus its neighbors was fewer than 20 during 1991–1995. To avoid calculating rates for counties that had no population themselves but whose neighbors had significant populations, rates were calculated only for counties that had a population count of 5 or greater for 1991–1995 (i.e., had 5 or greater person–years). 

Information on Hispanic ethnicity was not collected on death certificates in Oklahoma throughout the 1991–1995 study period, and prior to 1993 was not collected for New Hampshire. Consequently, we removed all counties in Oklahoma and New Hampshire from the contiguity matrix when the rates for Latinas were spatially smoothed, and no rates for Hispanics in Oklahoma and New Hampshire were calculated.

2. Standard Population Weights 

Age–specific heart disease death rates were directly age–adjusted using the 1970 U.S. population as the standard. The 1970 standard weights were based on the total resident population in the United States as of April 1, 1970. 

Because we generated heart disease death rates only for women ages 35 and over, and weights used in the age–adjustment of mortality rates are required to sum to 1, the weights for 10–year age groups for ages 35 and over were recalculated from the 1970 standard weights. The 1970 standard weights were summed for age groups 35–44 through ages 85+. New weights for each of these age groups were calculated by dividing the original weight by the sum of the weights for ages 35 and older (i.e., .418101). The new weights were rounded to two decimal places for subsequent calculation of age–adjusted heart disease death rates.

3. Hispanic Population in New York City 

During 1991–1993, information on Hispanic origin was not reported on approximately 22% of heart disease death certificates for women aged 35 years and older residing in New York City. During 1994–1995, the percent of death certificates for women that were missing information on Hispanic origin dropped to less than 3%. Based on a detailed examination of the New York City death certificate data for our five–year study period, we concluded that the majority of the deaths with "unknown" Hispanic origin occurred among non–Hispanic women. As evident in the table below, the percent of heart disease deaths for Hispanic women rose only slightly between 1991–1993 and 1994–1995, while the percent of heart disease deaths for non–Hispanic women rose markedly after reporting improved in 1994. From 1991–1993 to 1994–1995, the average annual number of heart disease deaths increased 7% for Hispanic women and 22% for non–Hispanic women, while the number of deaths with unknown Hispanic origin declined 96%. 

However, since a small proportion of the deaths with missing Hispanic origin data did occur among Hispanic women, it is almost certain that the heart disease death rates reported here for Hispanic women are modestly (but not severely) underestimated. In addition, the extent of underestimation may have varied among the five city boroughs; therefore prudence should be exercised in comparing individual county rates.

4. Contiguity Matrix for Alaska

A contiguity matrix for all U.S. counties was obtained from the 1996 Area Resource File (ARF). The matrix identifies maximum of fourteen contiguous neighbors for every U.S. county. Because Alaska was treated as a single geographic unit in the ARF, we created our own contiguity matrix for Alaska (shown below). Columns n1–n9 identify contiguous neighbors to each county. Counties are identified by FIPS code.

Back to top

G. References

1. National Center for Health Statistics. Public Use Data Tape Documentation, Mortality Detail, 1992. Rockland, MD: National Center for Health Statistics, 1992.
2. Wallman KK, Hodgdon J. Race and ethnic standards for federal statistics and administrative reporting. Statistical Reporter, July 1977 (no. 77–10):450–54.

Back to top