Abstract
Objectives. This study analyzed coronary heart disease (CHD) mortality trends from 1985 to 1995, by race and sex, among Black and White adults 35 years and older to determine whether adverse trends were evident in any US localities.
Methods. Log-linear regression models of annual age-adjusted death rates provided a quantitative measure of local mortality trends.
Results. Increasing trends in CHD mortality were observed in 11 of 174 labor market areas for Black women, 23 of 175 areas for Black men, 10 of 394 areas for White women, and 4 of 394 areas for White men. Nationwide, adverse trends affected 1.7% of Black women, 8.0% of Black men, 1.1% of White women, and 0.3% of White men.
Conclusions. From 1985 to 1995, moderate to strong local increases in CHD mortality were observed, predominantly in the southern United States. Black men evidenced the most unfavorable trends and were 25 times as likely as White men to be part of a local population experiencing increases in coronary heart disease mortality.
After a half century of precipitous increases, mortality from coronary heart disease (CHD) began to decline nationwide in the mid-1960s.1 Substantial declines in CHD mortality occurred among both Whites and African Americans throughout the 1970s.2–5 Between 1976 and 1985, rates of decline in CHD mortality slowed considerably among Black men, Black women, and White women but not among White men.6 Race and sex disparities in regard to declining CHD mortality were evident from 1980 to 1988, with average annual declines of 3.7% for White men, 3.1% for Black men, 2.9% for White women, and 2.2% for Black women.7 A study that examined CHD mortality trends among these 4 race–sex groups from 1980 to 1991 revealed similar disparities.8 During the period 1980 to 1993, African American men experienced the slowest nationwide declines in CHD mortality at both younger (35 to 64 years) and older (65 years and above) ages.9,10
Geographic variations in CHD death rates within the United States have been observed for decades.7,9–25 An excess of CHD mortality in the southern United States was first observed in 1950,11 and it appears to have persisted over time.9,10,12,15,22 In the Atlas of United States Mortality, maps of smoothed (estimated) heart disease mortality rates for 1988 to 1992 showed a clear southern excess among Blacks and Whites of both sexes.22 Several studies have described geographic disparities in CHD mortality based on state of residence,7,14,24 rurality–urbanicity,9–11,13,18,19,21,25 local educational achievement profiles,20 and community occupational structure.16,23,26
Concomitant with observations of geographic variation in level of CHD mortality, substantial spatial variations in CHD temporal trends have been reported.7,9,10,15,18,25,27,28 Although CHD mortality rates began declining nationally in the mid-1960s, there were substantial local variations in the timing of the onset of decline for Whites, with many local areas not experiencing declining mortality until the early to mid-1970s.27,28 Late onset of decline in CHD mortality among Whites was associated with rurality and more adverse community occupational, educational, and income profiles.28,29 A study of geographic variation in CHD mortality trends among Whites for the period 1968 to 1978 showed the slowest rates of decline in Appalachia and the mid-South for men and in parts of Texas, the Mississippi River Valley, and Appalachia for women.15
More recently, consistently slower rates of decline have been observed for both African Americans and Whites in rural areas, particularly in the South21,25 and in Appalachia.9,10 One study produced evidence of adverse (i.e., increasing) trends in CHD mortality from 1984 to 1995 among African American men of lower social class residing in North Carolina.30 In Appalachia, elderly Black men residing in rural areas experienced no decline in CHD mortality between 1980 and 1993.9,31
In this article, we report geographic patterns of CHD mortality among United States adults for the period 1985 to 1995. This study was part of a larger project focused on rural–urban disparities in heart disease, with an emphasis on identifying characteristics of the social environments of local areas that affect the prevalence of risk factors and trends in mortality. The geographic unit of analysis for this study was the labor market area (described subsequently).32 Given evidence from several studies of slowing declines among some demographic groups and adverse trends in localized populations of Black men, we were particularly interested in determining whether the much-acclaimed 25-year decline in CHD mortality was still apparent for adults in all localities of the United States.
Our study population consisted of White and African American adults 35 years or older who resided in the United States during the years 1985 to 1995. We chose to examine CHD mortality trends separately for Blacks and Whites not because we view race as a biological risk factor for CHD but, rather, because race is an important sociologic category that represents “the interaction of biological, cultural, socioeconomic, political, and legal determinants (including racism)” of health outcomes.33 Both race groups included individuals of Hispanic/Latino ethnicity. In most states before 1989, death certificates did not record data on Hispanic ethnicity; consequently, we were unable to examine CHD mortality trends for Hispanics separately. Following recent guidelines from the US Office of Management and Budget, we use the terms Black and African American interchangeably in this article.
As mentioned, the geographic unit of analysis in this study was the labor market area. Labor market areas consist of 1 or more counties and define small regions in which participants in local labor markets both live and work; that is, they encompass the space in which labor market processes are grounded.32,34–37 Given our interest in the influence of the social environment, particularly economic development, on geographic patterns of CHD mortality, labor market areas offered a more theoretically defensible approximation of “communities” than counties or other county aggregates. Labor market areas are relatively large in land area relative to the small areas defined for other types of epidemiologic studies, such as analyses of cancer clusters. Consequently, each labor market area contains a heterogeneity of microenvironments, including neighborhoods and households. However, areas as small as neighborhoods, often defined according to census block groups, are inappropriate for studies of economic development and labor market processes, because these processes operate on a larger spatial scale.
The delineation of labor market areas used in this study was based on extensive analysis of empirical journey-to-work and residence data for counties, county equivalents, and the District of Columbia obtained from the US census.37 There are 394 labor market areas defined for the United States. These areas cross state boundaries, which is appropriate given the nature of local labor markets. Each labor market area comprised at least 100 000 inhabitants in 1990, and hence the land areas of labor market areas tend to be larger in the western United States, which is more sparsely populated. The largest city or town within a labor market area is typically used in describing the area; we follow that convention here.
We obtained death certificate computer files from the National Center for Health Statistics. Deaths from CHD were defined for this study on the basis of underlying cause of death. Specifically, we included underlying cause-of-death codes 402 (hypertensive heart disease), 410 to 414 (ischemic heart disease), and 429.2 (cardiovascular disease, unspecified) as defined in the International Classification of Diseases, Ninth Revision. This method was used in previous studies and has been recommended by several researchers.6,18,23 For each decedent, we abstracted data on age, race, sex, year of death, and county of residence. US Bureau of the Census intercensal population estimates were used to provide race-, sex-, and age-specific population counts for all counties for the years 1985 to 1995.
County-specific CHD deaths and population counts were summed to the labor market area level by year of death, sex, race, and 10-year age group. Age-specific death rates were computed and used to calculate, by race and sex, age-adjusted annual CHD mortality rates for each labor market area. The 1970 US population, the standard currently used by the Centers for Disease Control and Prevention, was used as the standard for age adjustment. Following the convention of the National Center for Health Statistics,38 we calculated race- and sex-specific CHD death rates only for those labor market areas in which the population experienced an average of at least 20 CHD deaths per year.
In the case of White men and women, rates were calculated for all 394 labor market areas in the United States. However, solely as a result of the underlying geographic distribution of the Black population in the United States, CHD death rates were calculated for only 175 labor market areas in the case of Black men and only 174 areas in the case of Black women. These labor market areas constituted more than 95% of the total African American population in the United States during the study period.
To quantify CHD mortality trends over the study period, we fit separate linear regression models to the log-transformed, age-adjusted rates for each race–sex group in each labor market area. Specifically, we used the following model: y = α + β(x) + e, where y is ln(age-adjusted rate), x is year, and e is the error term. The average annual percentage change in mortality is 100(eβ – 1). The use of a log-linear model allowed us to compare relative changes in mortality over time for populations with markedly disparate absolute CHD death rates. In addition, linear regression of the log-transformed rates assumed constant proportional change over time, as opposed to constant absolute change over time. This has been shown to be a more appropriate model for examining temporal trends in mortality.39 We used the standard errors from the regression models in calculating confidence intervals for average annual percentage change values.
Maps for each race–sex group were created on the basis of the regression estimates of average annual percentage changes in CHD death rates from 1985 to 1995. The following 5 categories of trend outcomes were mapped for each race–sex group, permitting comparisons across the maps: strong decline (5% or greater decline in mortality each year), moderate decline (1% to 5% decline in mortality each year), negligible change (less than 1% decline or less than 1% increase per year), moderate increase (1% to 5% increase in mortality each year), and strong increase (5% or greater increase in mortality each year). We took a conservative approach in categorizing local trends, considering a labor market area to have experienced an increasing trend only if death rates increased at least 1% per year. Each map shows all of the labor market areas in the United States, shaded to reflect the appropriate trend category. Overall, darker shades represent adverse trends, and lighter shades represent favorable trends. Labor market areas for which there were insufficient data to calculate CHD death rates are unshaded on the maps for Black men and Black women.
A representative CHD mortality trend graph for a single labor market area (nonmetropolitan Searcy City, Ark) is shown in Figure 1. On this graph, annual age-adjusted CHD death rates are displayed for each of the 4 race–sex groups. The dashed trend lines depict estimated CHD death rates derived from the log-linear regression models. In this labor market area, trends in CHD mortality from 1985 to 1995 were as follows: strong increase among Black women, moderate increase among Black men and White men, and negligible change among White women. Comparable trend data were generated and analyzed for each labor market area in the United States.
Substantial geographic variation in CHD mortality trends from 1985 to 1995 was evident for adults in each of the race–sex groups analyzed (Figure 2). Annual changes in CHD death rates for local areas ranged from –9.2% to +9.7% for Black women, from –7.9% to +9.7% for Black men, from –6.7% to +6.7% for White women, and from –8.0% to +3.2% for White men. Median average annual percentage change values in CHD mortality for labor market areas were –3.1% for White men, –2.8% for White women, –2.1% for Black men, and –2.1% for Black women. Overall, White men experienced the most favorable distribution of local CHD mortality trends. The bar along the x-axis of Figure 2 shows the 5 cutpoints used in mapping CHD mortality trends (relative to the distribution of labor market area trends for each race–sex group).
The geographic pattern of local CHD mortality trends from 1985 to 1995 for African American women is shown in Figure 3. CHD death rates strongly increased from 1985 to 1995 in 2 labor market areas. Moderate increases in CHD death rates were observed for 9 labor market areas: 2 in the Mississippi River Valley region, 1 in Alabama, 5 in Texas, and 1 in Sacramento, Calif, and surrounding areas. Negligible change in CHD mortality among Black women was found for a cluster of labor market areas in the Mississippi River Valley region, parts of rural Texas, and several large cities and their surrounding areas, including New York, San Francisco, Miami, and Buffalo. Black women in the majority of labor market areas (75%) experienced moderate or strong declines in CHD mortality.
Of the 23 labor market areas in which African American men experienced increased CHD mortality from 1985 to 1995, only 1 was located outside the South (Figure 4). Strong increases in CHD death rates among Black men were observed for 2 labor market areas. Moderate increases in CHD mortality were experienced by Black men in 21 additional labor market areas, including 10 contiguous areas in the Mississippi River Valley region of Missouri, Arkansas, Mississippi, and Louisiana. Labor market areas in which Black men experienced negligible changes in CHD mortality over the study period were also found predominantly in the South. Black men in the majority of labor market areas (72%) experienced moderate or strong declines in CHD mortality. Strong declines were experienced by Black men in several moderate- to large-sized cities and their surrounding areas, including Daytona Beach, Fla; Macon, Ga; Roanoke, Va; Muncie, Ind; and Boston.
Among White women (Figure 5), only 1 labor market area had a strong increase in CHD mortality from 1985 to 1995. Moderate increases in CHD death rates were experienced by White women in 9 additional labor market areas, all located in Texas, Oklahoma, and New Mexico, including 5 contiguous areas in western Texas and New Mexico. The majority of labor market areas in which White women experienced negligible changes in CHD mortality were located west of the Mississippi; several were clustered in Texas, Oklahoma, Arkansas, and Missouri. Moderate and strong declines in CHD mortality were experienced by White women in the majority (86%) of labor market areas.
White men in 90% of labor market areas experienced moderate or strong declines in CHD mortality from 1985 to 1995 (Figure 6). There were no labor market areas in which White men experienced strong increases in CHD mortality and only 4 in which they experienced moderate increases. There were negligible changes in CHD mortality among White men in 37 labor market areas, most clustered in the Mississippi River Valley region, Oklahoma, and Texas.
Table 1 (women) and Table 2 (men) present descriptive data for each of the labor market areas within which 1 or more race–sex groups experienced an adverse (i.e., increasing) trend in CHD mortality. In the case of several labor market areas, confidence intervals for average annual percentage change values did not include zero. Adverse trends were concentrated in the South, particularly in Texas and the Mississippi River Valley region. Among Whites, adverse trends occurred in several labor market areas that had substantial and growing Hispanic populations.
Nationwide, the proportions of adults 35 years and older who were affected by unfavorable heart disease mortality trends varied by race and sex. Moderate to strong increases in CHD mortality were experienced by 8.0% of Black men, 1.7% of Black women, 1.1% of White women, and 0.3% of White men. Numerically, adverse trends had the greatest impact among White women (n = 617 476 in 1995), followed by Black men (n = 446 545 in 1995), White men (n = 186 240 in 1995), and Black women (n = 121 008 in 1995). While the majority of the adult population in all 4 race–sex groups experienced moderate to strong declines in CHD mortality, results were most favorable for White men (96.1%), followed by White women (93.8%), Black men (82.7%), and Black women (69.0%).
Recent studies of trends in CHD mortality, both national trends7,8,10,40 and trends in the local catchment areas of large cohort studies,41 have reported average declines of at least 1% per year for all race–sex groups. Clearly, national studies can mask unfavorable trends for particular localities; similarly, caution should be exercised in generalizing the results of cohort studies to larger populations. For example, in the present study, all 4 race–sex groups experienced strong declines in CHD mortality of more than 5% per year in the labor market area that contains Framingham, Mass, the site of the longest-running cohort study of CHD in the nation. However, only 3% to 5% of adults nationwide resided in labor market areas that experienced similarly strong declines in CHD mortality from 1985 to 1995.
Our results are consistent with previous reports of widening geographic inequalities in CHD mortality nationwide,17 unfavorable patterns of CHD mortality in the South,10–12,15,22,25,28 and recent adverse trends in CHD mortality among African American men.9,30 Whereas Black women were about 1.5 times more likely than White women to experience adverse CHD mortality trends, Black men were more than 25 times as likely as White men to experience adverse CHD mortality trends from 1985 to 1995. In addition, a substantial minority of African American adults (9.3% of men and 29.3% of women) experienced negligible changes in CHD mortality over the study period.
Geographic disparities in health outcomes may be attributed to either compositional or contextual effects.42,43 Composition refers to the individual characteristics of people who live in a local area (e.g., socioeconomic status, prevalence of behavioral or physiologic risk factors), while context refers to characteristics of local social environments that are not reducible to individual traits (e.g., labor market structure, tax base, income inequality). Several recent studies have shown that community context affects health outcomes above and beyond the impact of compositional effects.42–47 Ameliorating the adverse trends reported here for localities across the country will first require an understanding of the relative effects of individual traits and social environmental contexts in producing population patterns of CHD.
aConfidence interval (CI) did not include zero, indicating a statistically significant increase in CHD mortality from 1985 to 1995. aConfidence interval (CI) did not include zero, indicating a statistically significant increase in CHD mortality from 1985 to 1995. FIGURE 1— Coronary heart disease mortality trends: Labor Market 43, nonmetropolitan Searcy City, Ark, 1985–1995. FIGURE 2— Distribution of coronary heart disease (CHD) mortality trend outcomes across labor market areas among adults 35 years or older, by race and sex: United States, 1985–1995. FIGURE 3— Local trends in coronary heart disease (CHD) among Black women: United States, 1985–1995. FIGURE 4— Local trends in coronary heart disease (CHD) among Black men: United States, 1985–1995. FIGURE 5— Local trends in coronary heart disease (CHD) among White women: United States, 1985–1995. FIGURE 6— Local trends in coronary heart disease (CHD) among White men: United States, 1985–1995.
LMA No. Average Annual Change in CHD Mortality, % (95% CI) LMA Name US Census Region Size of Largest Locale Hispanic, %, 1990 Hispanic, %, 1995 Black 318 1.4 (–5.7, 9.0) Bryan–College Station, Tex South Small metropolitan 8.8 10.3 328 1.4 (–1.2, 4.1) Waco, Tex South Small metropolitan 5.6 6.5 98 2.1 (–2.1, 6.4) Nonmetropolitan Auburn City, Ala South Larger urban 0.4 0.4 334 2.1 (–1.6, 5.9) Longview–Marshall, Tex South Small metropolitan 1.2 1.4 53 2.1 (–0.4, 4.7) West Memphis City, Ark South Larger urban 0.5 0.7 335 2.4 (–2.3, 7.3) Texarkana, Tex–Texarkana, Ark South Small metropolitan 1.0 1.2 258 2.7 (–2.0, 7.6) Nonmetropolitan Blytheville City, Ark Midwest Larger urban 0.5 0.6 374 3.0 (0.0, 6.0) Sacramento, Calif West Major metropolitan 8.8 10.6 319 4.0 (0.6, 7.6)a Brazoria, Tex South Major metropolitan 13.0 14.7 332 5.1 (–1.6, 12.2) Paris City, Tex South Larger urban 1.4 1.6 43 6.8 (1.5, 12.3)a Nonmetropolitan Searcy City, Ark South Small urban 0.2 0.3 White 309 1.1 (–0.7, 2.8) Amarillo, Tex South Small metropolitan 11.7 13.1 324 1.1 (–2.4, 4.7) Nonmetropolitan Big Spring City, Tex South Larger urban 17.3 19.7 301 1.2 (–0.4, 2.7) Fort Smith, Ark–Fort Smith, Okla South Small metropolitan 0.6 0.9 326 1.5 (–0.9, 4.0) Wichita Falls, Tex South Small metropolitan 4.4 5.1 308 1.6 (–0.8, 4.0) Lubbock, Tex South Small metropolitan 17.4 19.4 319 1.6 (0.1, 3.1)a Brazoria, TX South Major metropolitan 13.0 14.7 339 1.7 (–0.4, 3.9) Sherman–Denison, Tex South Small metropolitan 0.9 1.1 307 2.0 (0.7, 3.3)a Nonmetropolitan Roswell City, NM West Larger urban 26.2 27.2 314 4.1 (1.9, 6.3)a Odessa–Midland, Tex South Small metropolitan 22.8 25.1 329 6.7 (2.2, 11.5)a Killeen–Temple, Tex South Medium metropolitan 8.4 9.7 LMA No. Average Annual Change in CHD Mortality, % (95% CI) LMA Name US Census Region Size of Largest Locale Hispanic, %, 1990 Hispanic, %, 1995 Black 42 1.0 (–1.1, 3.2) Little Rock–North Little Rock, Ark South Medium metropolitan 0.5 0.9 53 1.0 (–0.6, 2.7) West Memphis City, Ark South Larger urban 0.7 1.0 35 1.0 (–1.5, 3.7) Baton Rouge, La South Medium metropolitan 1.4 1.5 66 1.1 (–2.2, 4.4) Rome City, Ga South Larger urban 0.8 1.2 197 1.1 (–0.3, 2.5) Philadelphia, Pa–NJ Northeast Major metropolitan 2.7 3.2 28 1.2 (–2.1, 4.6) Nonmetropolitan Laurel City, Miss South Small urban 0.3 0.4 311 1.5 (–2.6, 5.8) Victoria, Tex South Small metropolitan 19.4 22.6 48 1.7 (–3.6, 7.3) Nonmetropolitan Greenville City, Miss South Larger urban 0.8 0.8 321 1.7 (–1.1, 4.7) Beaumont–Port Arthur, Tex South Medium metropolitan 3.1 3.9 322 1.7 (–2.8, 6.5) Nonmetropolitan Nacogdoches City, Tex South Larger urban 5.2 6.5 333 1.8 (–1.8, 5.4) Tyler, Tex South Small metropolitan 3.3 4.1 39 1.9 (–1.7, 5.6) Monroe, La South Small metropolitan 0.6 0.7 3 1.9 (–4.9, 9.2) Knoxville, Tenn South Medium metropolitan 0.4 0.5 334 3.3 (–1.7, 5.6) Longview–Marshall, Tex South Small metropolitan 1.7 2.1 32 3.3 (0.2, 6.5)a Nonmetropolitan Vicksburg City, Miss South Larger urban 0.4 0.4 98 3.7 (–1.6, 9.3) Nonmetropolitan Auburn City, Ala South Larger urban 0.4 0.5 41 3.7 (0.8, 6.7)a Pine Bluff, Ark South Small metropolitan 0.4 0.7 319 3.9 (–1.2, 8.1) Brazoria, Tex South Major metropolitan 14.6 16.8 340 4.0 (0.6, 7.5)a Nonmetropolitan Hot Springs City, Ark South Larger urban 0.7 1.0 43 4.0 (1.7, 6.5)a Nonmetropolitan Searcy City, Ark South Small urban 0.3 0.4 258 4.8 (–2.0, 12.0) Nonmetropolitan Blytheville City, Ark Midwest Larger urban 0.6 0.9 332 6.3 (0.9, 12.0)a Paris City, Tex South Larger urban 1.9 2.3 314 9.7 (2.9, 17.0)a Odessa–Midland, Tex South Small metropolitan 24.3 27.3 White 43 1.2 (–1.3, 3.7) Nonmetropolitan Searcy City, Ark South Small urban 0.3 0.4 326 1.4 (–0.1, 2.9) Wichita Falls, Tex South Small metropolitan 5.4 6.5 314 2.7 (0.7, 4.7)a Odessa–Midland, Tex South Small metropolitan 24.3 27.3 329 3.2 (–0.4, 6.9) Killeen–Temple, Tex South Medium metropolitan 9.5 11.2 
This study was funded by an American Heart Association Scientist Development Grant (9630294N) to Elizabeth Barnett.
