Income, income inequality, and cardiovascular disease mortality: relations among county populations of the United States, 1985 to 1994.

Objectives: Despite the major contribution of cardiovascular disease (CVD) to total mortality, and reports demonstrating strong relations between income and CVD, the joint relations of population-level income and income inequality with CVD mortality are not well described. This study was undertaken to describe relations among population-level income, income equality, and mortality due to cardiovascular disease, coronary heart disease, and stroke.

Methods: County income distributions were determined from 1990 census data, and CVD mortality rates were obtained from the Compressed Mortality File. Relations among income, income inequality, and CVD mortality were examined in stratified and Poisson regression analyses.

Results: County income was inversely related and income inequality was directly related to CVD, coronary heart disease, and stroke mortality. Relations were strongest for stroke. Relations of stroke mortality with income inequality were strongest in low-income populations.

Conclusions: The CVD mortality experiences of county populations are related to both income and income distribution in a complex, disease-dependent manner. The authors' findings are especially relevant to the Southeast, a region of high income inequality, low income, and high stroke mortality.

Key Words: cardiovascular disease, coronary heart disease, income inequality, mortality, stroke

**********

Cardiovascular disease (CVD) has been associated with the socioeconomic status of individuals and the socioenvironmental characteristics of their place of residence. (1-3) These associations vary by time, place, and the particular manifestation of CVD, such as coronary heart disease (CHD) and stroke. (4-7) In recent decades, the risks of death due to any cause and due to CVD were greater among individuals of low income than among individuals of high income in the United States. (8,9)

A number of studies comparing populations at the levels of states, counties, and metropolitan areas have revealed that mortality rates are associated not only with population income, but also with inequalities in the distribution of income among population members. (10-14) Relations between income inequality and mortality are direct, such that mortality rates are higher in populations experiencing highly unequal income distributions (high income inequality) than in populations with relatively more equal income distributions (low income inequality).

Consistent with these findings, it has been reported that populations experiencing high income inequality have greater CVD risk factor exposures than populations of low income inequality. (15) Furthermore, this direct relation between income inequality and CVD risk factor exposure may be stronger for persons of low income than for persons of high income. (15) Thus, we may expect that the relation between income inequality and CVD mortality is a direct one and that this relation may be modified by income level. Few studies have specifically investigated relations between income inequality and CVD mortality. (11,16) We are not aware of any published report comparing CHD and stroke with respect to their relations with income inequality.

We examined the joint relationships of income and income inequality with CVD, CHD, and stroke mortality from 1985 to 1994 in an aggregate-level, cross-sectional analysis of US counties and their resident populations. We also investigated how associations between income distribution and CVD-related mortality differ in high- and low-income populations. Our objective was not to identify the underlying causes of these relationships, but rather to describe the complex and disease-specific interplay of income and its distribution with CVD mortality.

Materials and Methods

Data Sources and Unit of Analysis

County-level household income data were obtained from the US 1990 decennial census public access files provided by Geolytics, Inc. (17) Mortality rates were determined using county-level death counts and population estimates obtained from the National Center for Health Statistics Compressed Mortality File. (18) The unit of analysis was the county or county equivalent.

Inclusion and Exclusion Criteria 

All counties in the 48 contiguous states were included. Counties with inconsistent geographic boundaries in the census and mortality files were excluded (n = 4), and counties with virtually nonexistent populations were excluded (n = 1). Only county resident populations aged 35 to 74 years were included. All race and ethnic groups were included.

Income and Income Inequality

The measure of population income used in these analyses was median household income. The income distribution measure was the top 50% household income share, hereafter referred to as the 50% share. This was defined as the percentage of aggregate household income earned by the highest earning 50% of households in each county. The 50% share was calculated from grouped census household income data in a manner consistent with previous studies. (10,13)

Briefly, the 50% share was determined as follows. The aggregate income earned for households in each of the 25 census income groups was estimated for each county based on the assumption that all households within each group earned the average of the group upper and lower income bounds. For the last and highest income group, open-ended at its upper bound, the aggregate income was provided in census data. Group-specific estimates of aggregate income were adjusted so that the sum of all groups was equivalent to the county aggregate household income. Aggregate income for the highest earning 50% of households was then calculated as the sum of aggregate incomes for all groups above the 50th percentile of households. Aggregate income above and below the 50th percentile within the group containing that point was estimated using linear interpolation. Finally, the 50% share was calculated for each county as the income earned by the highest earning 50% of households divided by aggregate household income of all households.

CVD-related Mortality

Mean annual mortality rates for the years 1985 through 1994 were determined as the average annual number of stratum-specific deaths during the time interval divided by the 1990 population. For stratified analyses, mortality rates for ages 35 to 74 years were age and gender adjusted by the direct method using the 1990 US population distribution as the standard. Cause-specific deaths were identified from the International Statistical Classification of Diseases, Injuries, and Causes of Death, Ninth Revision (19) rubrics 390 to 448 (CVD); 402, 410 to 414, and 429.2 (CHD); and 430 to 438 (stroke).

Analytic Techniques

Relations among county-level populations were examined using both stratified and Poisson regression analyses. To avoid unstable mortality rates in small populations during stratified analysis, deaths were determined over a relatively wide time interval (described above) and aggregated for counties within equal intervals in a distributional range of median household income and the 50% share.

Mortality rates were calculated from aggregated death and population counts for counties grouped within three ordinal levels of income and income inequality. These three levels were defined from three equal intervals in the 5th to 95th percentile range of median household income and the 50% share. Counties with values above or below this range were included in the lowest or highest level of each ordinal variable, respectively.

We also calculated mortality rates from aggregated death and population counts for counties grouped within 40 equal intervals in the 1st to the 99th percentile range of median household income and the 50% share. Mortality rates for counties with values of income or income inequality less than the 1st percentile or greater than the 99th percentile were aggregated, resulting in a total of 42 mortality rate strata. These were plotted against the mean value of median household income or the 50% share for counties within each stratum in order to evaluate linearity of these relations.

Poisson regression was performed using PROC GENMOD of SAS Version 8 (SAS Institute, Inc, Cary, NC). We used the Poisson regression method to predict mortality rates (a death count divided by the population). Poisson regression is used when the outcome is a count variable of relatively rare events. The population is an offset variable in Poisson regression. This is a commonly used regression method in mortality studies. All models were dispersion-corrected, and the dispersion parameter for nested models was held fixed to the value determined for the full model. As indicators of model fit, we report values for deviance and log likelihood (expressed as twice the log likelihood determined from scaled deviances). We do not report probability values because the study design involves neither random sampling nor randomization. (20) All regression models included terms for age and gender.

Results

Median household income during 1989 ranged from $8,595 to $59,284 (mean, $23,858) among the 3,106 counties included in these analyses. The percentage of total household income earned by the highest earning 50% of households (ie, the 50% share) ranged from 68.6 to 87.4% (mean, 79.1%).

The county-level geographic distributions of income and income inequality in 1989 are shown in Figure 1 by quartile. Counties of relatively low median household income and high income inequality were clustered in the southeastern states, along the Gulf Coast, the southern aspect of the Mississippi River, Appalachia, and the coastal plains of the Carolinas and Georgia. This geographic pattern was generally consistent with areas of relatively high CHD and stroke mortality described elsewhere. (21,22) Also, consistent with previous reports, (23) county-level income and income inequality were inversely related (Pearson r = -0.62).

[FIGURE 1 OMITTED]

Approximately 3.5 million CVD deaths occurred from 1985 to 1994 for ages 35 to 74 years in the included counties. Of these, about 66.2% were attributable to CHD and 12.5% to stroke. The mean annual age- and gender-adjusted mortality rates per 100,000 population for these counties averaged 374, 240, and 47 for CVD, CHD, and stroke, respectively.

The CVD, CHD, and stroke mortality rates of county populations varied inversely with county median household income (Fig. 2). CVD, CHD, and stroke mortality rates varied directly with income inequality (50% share). These relations appear to be fairly linear on the log scale as demonstrated by the fitted curves (solid). Relations of income and income inequality are stronger for stroke than for CHD based on the magnitude of slopes of the fitted curves. The slopes in this figure represent the change in the natural log of the number of deaths per 100,000 population associated with each $1,000 increase in median household income or 1% increase in income inequality. Based on the slopes, each $1,000 increase in median household income was associated with CVD, CHD, and stroke mortality decreases of 1.4, 1.1, and 1.8%, respectively. Each 1% (absolute) increase in income inequality (the 50% share) was associated with mortality rate increases of 4.3, 3.6, and 5.8% for CVD, CHD, and stroke, respectively.

The joint associations among median household income, income inequality, and CVD-related mortality are illustrated using a three-level (described above) stratified approach in Table 1. Within each level of median household income, mortality increased as income inequality increased for CVD, CHD, and stroke. Cause-specific mortality rates (deaths per 100,000) were lowest in counties of high income and low inequality and were highest in counties of low income and high inequality, ranging from 315 to 457 for CVD, from 210 to 286 for CHD, and from 38 to 61 for stroke. Thus, after simultaneously controlling for income and its distribution through stratification, each factor appeared to have a separate association with cause-specific mortality.

Relative relations comparing cause-specific mortality for counties in the highest and lowest strata of median household income and income inequality from Table 1 are examined in Table 2. Cause-specific mortality rate ratios comparing counties of high to low income inequality were greater where median household income was low than where it was high. Cause-specific mortality rate ratios comparing counties of low to high income were greater where income inequality was high than where it was low. Rate ratios were greatest for stroke except for the ratio contrasting high- to low-income counties of low income inequality. CVD, CHD, and stroke mortality was, respectively, 1.36, 1.26, and 1.60 times higher in low-income counties of high inequality than in low-income counties of low income inequality. CVD, CHD, and stroke mortality was, respectively, 1.27, 1.15, and 1.33 times higher in high-inequality counties of low income than in high-inequality counties of high income. Examination of mortality rate ratios in these stratified analyses therefore suggest that relations between median income and CVD-related mortality may vary across levels of income inequality, relations between income inequality and mortality may vary by median income, and these relations are stronger for stroke than for CHD.

[FIGURE 2 OMITTED]

Poisson regression results (Table 3) were consistent with stratified analyses in showing that county-level CVD, CHD, and stroke mortality was inversely related to median household income (model 1) and directly related to income inequality (model 2). The regression coefficients were similar in magnitude to the slopes of the log-linear fitted curves in Figure 2. In models where both income and income inequality were considered jointly (model 3), these relations persisted but were diminished in magnitude. Based on model 3, CVD, CHD, and stroke mortality decreased by 0.8, 0.6, and 1.1%, respectively, with each $1,000 increase in median household income and increased by 3.1, 3.3, and 4.1% for each 1% (absolute) increase in the 50% share. Income and income inequality considered jointly contribute substantially more to model fit than either variable considered separately for CVD, CHD, and stroke. Models containing a multiplicative interaction term between income and income inequality (model 4) substantially enhanced fit for CVD and stroke, but much less so for CHD. These findings are consistent with stratified analyses showing that relations of CVD and stroke with income inequality were strongest where population income was low.

The relations from regression analyses are illustrated in Figure 3 with age- and gender-adjusted cause-specific mortality rates predicted from model 4 regression coefficients. Relations between predicted CHD mortality and median household income (upper graph) were inverse, with similar slopes at various levels of income inequality. Relations between predicted CHD mortality and the 50% share (lower graph) were direct and with similar slopes at various levels of median household income. Based on the slopes presented in this figure, CHD mortality decreased by 1% with each $1,000 increase in median household income at each presented level of income inequality, and CHD mortality increased by 3% for each 1% (absolute) increase in the 50% share at each presented level of median household income.

In contrast, the inverse relation between predicted mortality and median household income varied markedly by the level of income inequality for stroke (upper graph). Likewise, the direct relation between predicted stroke mortality and the 50% share varied by median income (lower graph). The association between stroke mortality and median household income was stronger where income inequality was high than where it was low, and the association between stroke mortality and the income inequality was stronger where income was low than where it was high. Based on the slopes presented in Figure 3, predicted stroke mortality decreased for each $1,000 increase in median household income by 0, 1, and 2% for county populations with 50% share values of 74, 79, and 85%, respectively. Predicted stroke mortality increased for each 1% (absolute) increase in income inequality by 1, 5, and 7% for county populations with median household incomes of $46,000, $23,000, and $13,000, respectively.

Discussion

Our finding of an inverse relation between county-level median household income and CHD or stroke mortality is consistent with earlier reports relating CVD with individual-level and population-level income. (3) An inverse relation between socioeconomic status and CHD in the United States has existed since the onset of the decline in CHD mortality (mid-1960s) and has persistently increased in magnitude. (24) Pathways linking CVD-related mortality to the socioeconomic status of individuals and to the socioenvironmental conditions of their place of residence are complex and not well understood. (3) Well-established income-related CVD risk factors only partially explain associations between income and clinical CVD. (1)

Our findings of statistically income-independent direct relations between county-level income inequality and CVD-related mortality were consistent with previous reports from similar cross-sectional analyses performed at the aggregate level. (10-13,25-27) However, reported relations between mortality and income inequality vary across studies and appear to be highly dependent on context. For example, in their cross-sectional analysis of mortality in the Canadian population, (28) Ross et al failed to identify a relation with income inequality and concluded that "the lack of a significant association between income inequality and mortality in Canada may indicate that the effects of income inequality on health are not automatic and may be blunted by the different ways in which social and economic resources are distributed in Canada and in the United States." Lynch et al. have found that relations between income inequality and mortality vary among countries and are inconsistently associated with specific causes of death. (29)

[FIGURE 3 OMITTED]

A variety of hypothetical pathways have been described that link income inequality to health. Backlund et al (8) have reported a curvilinear relation between income and mortality risk for individuals in the United States such that the health benefits of increased income diminish as income increases. Assuming that income is related causally to health, this curvilinear relation suggests that in populations where income is relatively more concentrated among fewer individuals (ie, high income inequality), the health benefit realized by the few with the highest income would be more than offset by lost health opportunities experienced by the many of lower income. Although some may gain from increasingly unequal income distributions, average population health would decline due to the greater health losses experienced by others. Thus, the shape of the individual-level curvilinear income-health relation may explain, at least partially, how the distribution of income among individuals influences health at the level of populations. This has been referred to as the absolute income hypothesis. (30)

A number of other hypotheses have been advanced to explain observed relations between income distribution and health. (30) The specific pathways involved are currently an area of active research and great controversy. One major area of contention is the extent to which these relations are explained by the absolute income hypothesis (ie, by individual-level income-health relations) versus competing or complementary hypotheses involving contextual factors that are properties of populations and their social environments rather than properties of individuals. (31-34) Some investigators have found income inequality to be associated with indicators of societal dysfunction. (35-38) This is consistent with the income inequality hypothesis, which suggests that income inequality and its underlying causes have adverse societal effects with detrimental health consequences for the entire population. (30) Another view of this phenomenon is that economic inequalities, far from being dysfunctional, are a key component of the functioning of modern economies. (39)

It has also been suggested that one's actual or perceived position in society, partially determined by income and income distribution, may influence health. (30) This hypothesis is consistent with animal experiments that demonstrate associations of social status and social stability with atherosclerosis. These associations presumably operate through stress-induced neurologic and endocrine pathways. (40)

Multilevel studies simultaneously evaluating individual-level and population-level characteristics may eventually determine the relative contributions of the various hypothetical pathways linking income distribution to health. (41) However, findings from recently published multilevel analyses have been mixed. Fiscella and Franks (42) found that income inequality was not independently related to mortality when controlling for individual-level income in multilevel analyses of participants from the first National Health and Nutrition Examination Survey and Epidemiologic Follow-up Study. Their findings are consistent with other multilevel studies showing that statistically independent effects of income inequality on health virtually disappear when controlling for individual-level income. (43,44) In contrast, some multilevel studies have shown that income inequality is independently related to health and risk factors after controlling for individual-level income for entire populations or specific demographic groups within populations. (15,45-47) Differences in population characteristics, geographic units of analysis, analytic methods, and income and inequality measures contribute to the difficulty in comparing these studies.

Our analyses, conducted strictly at the population level without individual-level information, cannot evaluate the relative role of individual-level versus population-level factors. Rather, we set out to determine whether and how income and its distribution are jointly associated with CVD-related mortality. Aggregate-level studies such as ours have the strength of being sensitive to both individual-level and population-level factors relating income inequality to mortality. Aggregate-level studies may provide appropriate information about combined effects when individual- and population-level phenomena are interdependent and, indeed, not conceptually separate, as is likely the case for income, income distribution, and health outcomes.

Evidence that populations of high income inequality experience relatively higher CVD risk factor exposures than populations of lower inequality suggests that relations between income inequality and CVD-related mortality are mediated, at least in part, by the well-established CVD risk factors. (15) The direct association between CVD risk factor exposure and income inequality was reported to be stronger among low-income persons than among high-income persons. (15) We speculate that high income may "protect" individuals and populations from inequality-related adverse social and environmental conditions of place of residence. This is consistent with our findings of interactions involving income and its distribution in their associations with CVD and stroke mortality. Our results suggest that entire populations on average may be protected by high median income from health adversities related to income inequality. However, high median income may not guarantee good population health if its beneficial effects are offset by high income inequality.

It is notable that relations between income inequality and CHD varied little by median income, but relations between income inequality and stroke varied markedly by median income. These different relations for stroke and CHD suggest that pathways linking income distribution to CVD are complex, disease-dependent, and may include both individual-level and population-level components.

County-level income distribution has been related to a variety of social, economic, and demographic factors that change over time. (23) The distributions of individuals defined by age, race, and education have been related to both income inequality and CVD mortality rates. In an analysis of mortality among US states, (48) Muller demonstrated that the effect of income inequality disappeared after controlling for the educational attainment of the population. McLaughlin and Stokes (27) have demonstrated that county-level income inequality was relatively high where racial minority concentration was also high, and they have shown that where blacks constitute a high percentage of the population, mortality is higher in counties of low income inequality than in counties of high income inequality.

Population age distribution, especially with respect to older age groups, may be related to population income, income inequality, and CVD mortality rates. Nielsen and Alderson (23) have suggested that the effect of the size of the elderly population was an important contributor to income inequality in 1970 and earlier, when elderly families were a relatively disadvantaged segment of the income distribution. However, the impact of Social Security programs, the indexing of Social Security benefits, and other factors have contributed to improvements in the relative income of this population. As a result, by 1980, large elderly populations were no longer a major positive contributor to income inequality as it relates to population income. Nevertheless, given the relatively high levels of CVD mortality experienced by older populations, it is possible that our findings were influenced to some extent by geographic variations in population age distributions.

We did not focus on the population distributions of education, race, age, and other factors that may explain relations among income, income distribution, and CVD mortality. Rather, our approach was to examine and describe these relations at the aggregate level as a necessary prerequisite to more complex hypothesis-driven studies. It is likely that relations among income, income inequality, and CVD mortality are influenced by population-level distributions of multiple factors not considered in our analyses.

Our findings also suggest that national relations among income, income inequality, and CVD mortality might be driven by region-specific conditions. Low income, high income inequality, and high CVD-related mortality are prominent characteristics in the southeastern region of the United States. This region is also notable for its low levels of educational attainment and its high black population densities. Historic patterns of socioeconomic and health disparities in black and white populations of the Southeast may play a role in explaining our national level findings.

Prominent among the many reasons to better understand relations between income inequality and health are the potentially substantial population exposures and the increase in these exposures over time. The income inequality hypothesis suggests that all members of society, regardless of individual income, could be adversely affected by impairment of health-related societal functions in populations of high income inequality. (30) If this is true, then exposures may be large indeed, essentially including all population members. Not only are these exposures potentially universal, but they are also growing. Rapid income growth in the United States from World War II through the early 1970s was accompanied by little change in income inequality; however, in later years, national income growth benefited primarily higher income earners, whereas for the lowest income earners, real earnings increased little or declined. (49) In 1973, the top-earning 5% of US households earned 16.6% of the national aggregate household income and the bottom-earning 20% earned 4.2% of the aggregate. By 1996, the earnings share for the top 5% increased to 21.4% and the earnings share for the bottom 20% declined to 3.7%. (49) The US income gap has continued to widen through the late 1990s. According to a report from the Center on Budget and Policy Priorities citing Congressional Budget Office data, the income gap in the year 2000 was the widest ever recorded during the period beginning in 1979. According to this report, the inflation-adjusted (year-2000 dollars) average after-tax income gain from the year 1979 through the year 2000 was $1,100 for the lowest earning fifth of the US population, compared with an average gain of $576,400 for the top earning 1%. (50)

It is important to consider the causal implications of this cross-sectional analysis of county data. Although we and others have shown that income inequality is associated with poor health at the population level, there is little evidence that income inequality actually "causes" adverse health effects in populations or individuals. Indeed, simplistic notions of causation may be conceptually inadequate when applied to the complex phenomena considered in this study. Causation in epidemiology is usually founded on principles of experimental design and analysis in which the goal is to isolate the effect of a specific exposure while controlling for all other influences. In contrast, income and income inequality may impact health simultaneously through toxicologic, nutritional, behavioral, psychosocial, medical care, economic, and political mechanisms. Furthermore, increased disease in populations adversely affected by the complex influences of income level and inequality could itself limit income growth, especially among low-income segments of the population, and lead to increased inequality. Causal relationships in this case are more appropriately conceptualized as complex, interdependent, and historical phenomena rather than as universal relationships that can be subjected to experimental and quasiexperimental analyses. Our cross-sectional study should be considered in this context rather than as an attempt to provide causal evidence regarding an independent exposure-disease effect.

In addition to the changes in income inequality noted above, the geographic distribution of CVD mortality has also changed in the US population in recent years. (51) CHD mortality continues to be relatively higher in the southeastern United States, especially in rural and low-income populations. The geographic distribution of stroke has changed over time such that some southeastern states are no longer among those with the highest stroke mortality rates, whereas other states in the northwestern United States are beginning to exhibit relatively high stroke mortality rates. Thus, our findings may not reflect current conditions. To minimize the effect of low numbers in counties with small populations, our analyses aggregated death counts over a 10-year period centered on the 1990 Census. Compressed mortality data required to repeat this analysis centered on the 2000 Census were not available as of the date of this report. If changes in the geographic distribution of CVD mortality are associated with changes in population income and income inequality, findings consistent with ours may be expected in future studies.

Ongoing and future research may more clearly explain relations between income distribution and health. The evaluation of temporal relations among population-level changes in income, income inequality, and CVD is one area of research that could be especially valuable in examining the shifting geographic distribution of CVD. Recent research has focused on establishing whether and how population income distribution, per se, affects the health of individuals. Related research specific to stroke and CHD is needed. Future investigations focusing on the Southeast may be of particular interest, given the unique characteristics of this region, such as its racial composition and its relatively high disease burden.

Conclusion

We have shown that the CVD mortality experiences of populations are related to population income and its distribution in a joint and complex manner. Therefore, health scientists may want to consider both dimensions of income together to better understand geographic disparities of CVD and to explore changes in these disparities over time. Our finding that relations with income inequality differ for CHD and stroke suggests that disease-specific pathways are involved, and these need to be explored in future research. Finally, our finding that stroke mortality is most strongly associated with income inequality in low-income populations suggests that factors linking stroke mortality to income distribution may differ between impoverished and affluent populations. Economic and health policy planners may be especially concerned with policies that increase income inequality in more vulnerable populations of low income. In an era of increasing income inequality, these findings are especially relevant to the southeastern United States, a region of high inequality, low income, and high stroke mortality.

ARE YOU INTERESTED IN REVIEWING BOOKS FOR THE SOUTHERN MEDICAL JOURNAL?

The Southern Medical Journal frequently receives books from publishing companies requesting published reviews. If you are interested in reviewing books for the Journal, please email your name and degree, address, phone and fax numbers, professional affiliation, email address, and area(s) of specialty to [e-mail], or call the editorial office at [phone].

All books may be kept by the reviewer.

Table 1. Mean annual age- and gender-adjusted mortality rates among
counties aggregated within three levels of median household income and
income inequality, 1985-1994 (a)

                                CVD      CHD  Stroke

Low median household income
  Low income inequality         337 (b)  227  38
  Medium income inequality      390      254  47
  High income inequality        457      286  61
Medium median household income
  Low income inequality         347      233  41
  Medium income inequality      369      242  46
  High income inequality        428      282  54
High median household income
  Low income inequality         315      210  38
  Medium income inequality      344      235  41
  High income inequality        360      249  46

(a) CVD, cardiovascular disease; CHD, coronary heart disease.
(b) Deaths per 100,000.

Table 2. Mortality rate ratios comparing county groups of high and low
median household income and income inequality (a)

                                         CVD   CHD   Stroke

High to low income inequality mortality
  rate ratio
  Low median household income            1.36  1.26  1.60
  High median household income           1.14  1.19  1.21
Low to high median household income
  mortality rate ratio
  High income inequality                 1.27  1.15  1.33
  Low income inequality                  1.07  1.08  1.01

(a) CVD, cardiovascular disease; CHD, coronary heart disease.

Table 3. Poisson regression parameter estimates and fit statistics for
models controlling for age and gender (not shown), median household
income, income inequality, and the multiplicative interaction between
income and inequality (a)

                         MHI     II
           Intercept  (/$1,000)  (%)     MHI X II  Deviance  df

CVD
  Model 1   -5.5342   -0.0126                      170,632   24,842
  Model 2   -9.4161              0.0441            170,990   24,842
  Model 3   -8.1055   -0.0088    0.0310            156,919   24,841
  Model 4  -10.4887    0.0714    0.0611  -0.0010   155,382   24,840
CHD
  Model 1   -6.3286   -0.0100                      150,913   24,842
  Model 2   -9.9252              0.0415            145,585   24,842
  Model 3   -9.0095   -0.0061    0.0323            141,017   24,841
  Model 4   -9.2464    0.0018    0.0353  -0.0001   141,007   24,840
Stroke
  Model 1   -6.9854   -0.0156                       53,897   24,842
  Model 2  -11.8669              0.0556             53,606   24,842
  Model 3  -10.2960   -0.0107    0.0399             51,047   24,841
  Model 4  -14.3110    0.1255    0.0905  -0.0017    50,485   24,840

                Log
           likelihood (b)

CVD
  Model 1  5,546
  Model 2  5,489
  Model 3  7,739
  Model 4  7,984
CHD
  Model 1  2,596
  Model 2  3,535
  Model 3  4,340
  Model 4  4,341
Stroke
  Model 1  3,227
  Model 2  3,370
  Model 3  4,629
  Model 4  4,906

(a) MHI, median household income; II, income inequality; df, degrees of
freedom; CVD, cardiovascular disease; CHD, coronary heart disease.
(b) Twice the log likelihood determined from scaled deviances in
dispersion-corrected models.

Acknowledgments

The authors thank the staff of the Institute for Research in the Social Sciences at the University of North Carolina at Chapel Hill for their assistance with and provision of US Census data and the National Center for Health Statistics Compressed Mortality File.

Accepted December 8, 2003.

Copyright [c] 2004 by The Southern Medical Association

0038-4348/04/9705-0475

References

1. Diez-Roux AV, Nieto FJ, Tyroler HA, Crum LD, Szklo M. Social inequalities and atherosclerosis: The atherosclerosis risk in communities study. Am J Epidemiol 1995;141:960-972.

2. Diez Roux AV, Merkin SS, Arnett D, et al. Neighborhood of residence and incidence of coronary heart disease. N Engl J Med 2001;345:99-106.

3. Kaplan GA, Keil JE. Socioeconomic factors and cardiovascular disease: A review of the literature. Circulation 1993;88(4 pt 1):1973-1998.

4. Marmot MG, Adelstein AM, Robinson N, Rose GA. Changing socialclass distribution of heart disease. BMJ 1978;2:1109-1112.

5. Tyroler HA, Wing S, Knowles MG. Increasing inequality in coronary heart disease mortality in relation to educational achievement: Profile of places of residence, United States, 1962 to 1987. Ann Epidemiol 1993; 3(suppl):S51-S54.

6. Carter LR, Walton SE, Knowles MK, Wing S, Tyroler HA. Social inequality of stroke mortality among US black populations, 1968 to 1987. Ethn Dis 1992;2:343-351.

7. Wing S, Barnett E, Casper M, Tyroler HA. Geographic and socioeconomic variation in the onset of decline of coronary heart disease mortality in white women. Am J Public Health 1992;82:204-209.

8. Backlund E, Sorlie PD, Johnson NJ. The shape of the relation between income and mortality in the United States: Evidence from the National Longitudinal Mortality Study. Ann Epidemiol 1996;6:12-20.

9. Sterling T, Rosenbaum W, Weinkam J. Income, race, and mortality. J Natl Med Assoc 1993;85:906-911.

10. Kaplan GA, Pamuk ER, Lynch JW, Cohen RD, Balfour JL. Inequality in income and mortality in the United States: Analysis of mortality and potential pathways. BMJ 1996;312:999-1003.

11. Kennedy BP, Kawachi I, Prothrow-Stith D. Income distribution and mortality: Cross sectional ecological study of the Robin Hood index in the United States. BMJ 1996;312:1004-1007.

12. Lynch JW, Kaplan GA, Pamuk ER, et al. Income inequality and mortality in metropolitan areas of the United States. Am J Public Health 1998;88:1074-1080.

13. Brodish PH, Massing M, Tyroler HA. Income inequality and all-cause mortality in the 100 counties of North Carolina. South Med J 2000;93:386-391.

14. Cooper RS, Kennelly JF, Durazo-Arvizu R, Oh HJ, Kaplan G, Lynch J. Relation between premature mortality and socioeconomic factors in black and white populations of US metropolitan areas. Public Health Rep 2001;116:464-473.

15. Diez-Roux AV, Link BG, Northridge ME. A multilevel analysis of income inequality and cardiovascular disease risk factors. Soc Sci Med 2000;50:673-687.

16. Kawachi I, Kennedy BP, Lochner K, Prothrow-Stith D. Social capital, income inequality, and mortality. Am J Public Health 1997;87:1491-1498.

17. Census CD++Maps [machine readable 1990 US Census Files and software]. Prepared by Geolytics, Inc, East Brunswick, NJ, 1998.

18. National Center for Health Statistics, Public Health Service. Public Use Data Tape Documentation: Compressed Mortality File 1968-95. Washington, DC, US Government Printing Office, February 1998.

19. World Health Organization. International Statistical Classification of Diseases, Injuries, and Causes of Death, Ninth Revision. Based on the Recommendations of the Ninth Revision Conference, 1975. Geneva, World Health Organization, 1977.

20. Greenland S. Randomization, statistics, and causal inference. Epidemiology 1990;1:421-429.

21. Pickle LW, Mungiole M, Gillum RF. Geographic variation in stroke mortality in blacks and whites in the United States. Stroke 1997;28:1639-1647.

22. Feinleib M. Trends in heart disease in the United States. Am J Med Sci 1995;310(suppl 1):S8-S14.

23. Nielsen F, Alderson AS. The Kuznets curve and the great U-turn: Income inequality in US counties, 1970 to 1990. Am Sociol Rev 1997;62:12-33.

24. Feldman JJ, Makuc DM, Kleinman JC, et al. National trends in educational differentials in mortality. Am J Epid 1989;129:919-933.

25. Rodgers GB. Income and inequality as determinants of mortality: An international cross-sectional analysis. Popul Stud (Camb) 1979;33:343-351.

26. Franzini L, Ribble J, Spears W. The effects of income inequality and income level on mortality vary by population size in Texas counties. J Health Soc Behav 2001;42:373-387.

27. McLaughlin DK, Stokes CS. Income inequality and mortality in US counties: Does minority racial concentration matter? Am J Public Health 2002;92:99-104.

28. Ross NA, Wolfson MC, Dunn JR, Berthelot JM, Kaplan GA, Lynch JW. Relation between income inequality and mortality in Canada and in the United States: Cross sectional assessment using census data and vital statistics. BMJ 2000;320:898-902.

29. Lynch J, Smith GD, Hillemeier M, Shaw M, Raghunathan T, Kaplan G. Income inequality, the psychosocial environment, and health: Comparisons of wealthy nations. Lancet 2001;358:194-200.

30. Wagstaff A, van Doorslaer E. Income inequality and health: What does the literature tell us? Annu Rev Public Health 2000;21:543-567.

31. Stanistreet D, Scott-Samuel A, Bellis MA. Income inequality and mortality in England. J Public Health Med 1999;21:205-207.

32. Wildman J. The impact of income inequality on individual and societal health: Absolute income, relative income and statistical artifacts. Health Econ 2001;10:357-361.

33. Wolfson M, Kaplan G, Lynch J, Ross N, Backlund E. Relation between income inequality and mortality: Empirical demonstration. BMJ 1999; 319:953-955.

34. Mellor JM, Milyo J. Reexamining the evidence of an ecological association between income inequality and health. J Health Polit Policy Law 2001;26:487-522.

35. Coburn D. Income inequality, social cohesion and the health status of populations: The role of neo-liberalism. Soc Sci Med 2000;51:135-146.

36. Kawachi I, Kennedy BP. Health and social cohesion: Why care about income inequality? BMJ 1997;314:1037-1040.

37. Kawachi I, Kennedy BP. Income inequality and health: Pathways and mechanisms. Health Serv Res 1999;34(1 pt 2):215-227.

38. Kennedy BP, Kawachi I, Prothrow-Stith D, Lochner K, Gupta V. Social capital, income inequality, and firearm violent crime. Soc Sci Med 1998; 47:7-17.

39. Persson T, Tabellini G. Is inequality harmful for growth? Theory and evidence. Am Econ Rev 1994;84:600-621.

40. Manuck SB, Marsland AL, Kaplan JR, Williams JK. The pathogenicity of behavior and its neuroendocrine mediation: An example from coronary artery disease. Psychosom Med 1995;57:275-283.

41. Duncan C, Jones K, Moon G. Health-related behaviour in context: A multilevel modeling approach. Soc Sci Med 1996;42:817-830.

42. Fiscella K, Franks P. Individual income, income inequality, health, and mortality: What are the relations? Health Serv Res 2000;35(1 pt 2):307-318.

43. Blakely TA, Lochner K, Kawachi I. Metropolitan area income inequality and self-rated health: A multi-level study. Soc Sci Med 2002;54:65-77.

44. Osler M, Prescott E, Gronbaek M, Christensen U, Due P, Engholm G. Income inequality, individual income, and mortality in Danish adults: Analysis of pooled data from two cohort studies. BMJ 2002;324:13-16.

45. Lochner K, Pamuk E, Makuc D, Kennedy BP, Kawachi I. State-level income inequality and individual mortality risk: A prospective, multilevel study. Am J Public Health 2001;91:385-391.

46. LeClere FB, Soobader MJ. The effect of income inequality on the health of selected US demographic groups. Am J Public Health 2000;90:1892-1897.

47. Weich S, Lewis G, Jenkins SP. Income inequality and self rated health in Britain. J Epidemiol Community Health 2002;56:436-441.

48. Muller A. Education, income inequality, and mortality: A multiple regression analysis. BMJ 2002;324:23-25.

49. Ryscavage P. Inequality in the post-World War II era. Income Inequality in America: An Analysis of Trends. Armonk, NY, ME Sharpe, 1999, pp 45-80.

50. Greenstein R, Shapiro I. The new, definitive CBO data on income and tax trends. Center on Budget and Policy Priorities. Available at: http://www.centeronbudget.org/9-23-03tax.htm. Accessed October 26, 2003.

51. Cooper R, Cutler J, Desvigne-Nickens P, et al. Trends and disparities in coronary heart disease, stroke, and other cardiovascular diseases in the United States: Findings of the national conference on cardiovascular disease prevention. Circulation 2000;102:3137-3147.

RELATED ARTICLE: Key Points

* Inequality in the distribution of income within county populations is associated with high mortality rates for cardiovascular disease, coronary heart disease, and stroke.

* Stroke mortality is more strongly associated with income inequality than is coronary heart disease mortality, and this association is stronger where income is low than where it is high.

* In an era of persistently increasing income inequality, these findings are especially relevant to the Southeast, a region of high inequality, low income, and high stroke mortality.

Mark W. Massing, MD, PHD, Wayne D. Rosamond, PHD, Steven B. Wing, PHD, Chirayath M. Suchindran, PHD, Berton H. Kaplan, PHD, and Herman A. Tyroler, MD

From the Departments of Epidemiology and Biostatistics, School of Public Health, University of North Carolina at Chapel Hill, Chapel Hill, NC.

This research was funded in part by a Cardiovascular Disease Epidemiology training grant from the National Institutes of Health, National Heart, Lung, and Blood Institute, and a National Research Service Award (grant 5-T32-HL007055). Contributions from Dr. Wing were supported in part by National Institute of Environmental Health Sciences grant R25-ES08206-04.

Reprint requests to Mark W. Massing, MD, PhD, Department of Epidemiology, Bank of America Center, Suite 306, 137 East Franklin Street, Chapel Hill, NC 27514. Email: [e-mail]