----------------------------------------------------------- REPORT OF THE CONFERENCE ON SOCIOECONOMIC STATUS AND CARDIOVASCULAR HEALTH AND DISEASE November 6-7, 1995 National Heart, Lung, and Blood Institute National Institutes of Health ----------------------------------------------------------- REPORT OF THE CONFERENCE ON SOCIOECONOMIC STATUS AND CARDIOVASCULAR HEALTH AND DISEASE November 6-7, 1995 For Administrative Use National Heart, Lung, and Blood Institute National Institutes of Health ----------------------------------------------------------- Foreword I am pleased to present this report of the Conference on Socioeconomic Status (SES) and Cardiovascular Health and Disease, sponsored by the National Heart, Lung, and Blood Institute (NHLBI). This important meeting addressed a topic of great timeliness and interest to those of us who are concerned with the public health of the United States. Participants at the conference were given the following charge: * Review existing knowledge of biological, behavioral, and social factors related to SES variations in CVD morbidity and mortality and their trends, particularly with respect to minorities; * Identify scientific information that is ready for transfer to health care professionals and the public to improve the cardiovascular health of the country; * Determine future opportunities and needs for research on SES factors and their relationships with cardiovascular health and disease. The Chartbook of U.S. National Data on Socioeconomic Status and Cardiovascular Health and Disease served as background to the conference. The product of more than 2 years of work by a number of distinguished nonfederal scientists and representatives of the NHLBI, the National Center for Health Statistics, and the Health Care Financing Administration, the Chartbook is unique in its wealth of U.S. national data about race-, gender-, and age-specific associations between SES and CVD. As this conference report indicates, current knowledge and understanding of the behavioral, social, psychological, and biological pathways through which SES influences CVD remains limited. In clarifying the association between SES and CVD, a number of approaches must be vigorously pursued to acquire new scientific information, validate therapeutic and preventive approaches, and transfer new knowledge into practice. It is our hope that this report and its recommendations will receive close attention by the scientific community and serve as a starting point for identifying ways to reduce CVD among all segments of society. We are grateful to conference participants for their thoughtful contribution to our efforts. Claude Lenfant, M.D. Director National Heart, Lung, and Blood Institute ----------------------------------------------------------- Table of Contents Executive Summary Jeremiah Stamler, M.D. Helen P. Hazuda, Ph.D. Session I: Setting the Stage Session I Overview Millicent Higgins, M.D. - Session Chair Biologic and Methodologic Approaches to the Association Between Socioeconomic Factors and Cardiovascular Disease George A. Kaplan, Ph.D. Socioeconomic Status and Cardiovascular Disease Mortality: National Longitudinal Mortality Study Paul D. Sorlie, Ph.D., Norman J. Johnson, Ph.D., Eric Backlund, M.S. Socioeconomic Factors in Ischemic Heart Disease Morbidity and Mortality Jacob J. Feldman, Ph.D., Diane M. Makuc, Dr.P.H. Socioeconomic Status and Biomedical, Lifestyle, and Psychosocial Risk Factors for CVD: Selected U.S. National Data and Trends Clifford L. Johnson, M.S.P.H., Christopher T. Sempos, Ph.D. SES and Medical Care Utilization Related to CVD: Selected U.S. National Data and Trends Paul Eggers, Ph.D. Differences Between Occupational Classes in Cardiovascular Disease Mortality: A Comparison of 11 European Countries Anton E. Kunst, M.A., Feikje Groenhof, M.A., Johan Mackenbach, M.D., Ph.D., and the European Union Working Group on Socioeconomic Inequalities in Health Socioeconomic Status and Cardiovascular Disease: Experience in the United Kingdom Michael G. Marmot, M.D., Ph.D. Session II: Pathways Linking SES and CVD Session II Overview Redford B. Williams, M.D. - Session Chair Black and White Populations Herman A. Tyroler, M.D., Mark Massing, M.D., Marilyn Knowles, M.P.H. Pathways Linking SES and CVD: Hispanic Populations Helen P. Hazuda, Ph.D. Socioeconomic Status and 23-Year Cardiovascular Disease and Total Mortality Among Middle-Aged Men: The Honolulu Heart Program Beatriz L. Rodriguez, M.D., Ph.D., J. David Curb, M.D., Cecil M. Burchfiel, Ph.D., Ka-On Fong, M.A., Katsuhiko Yano, M.D., Dan S. Sharp, M.D., Ph.D. Socioeconomic Status and Cardiovascular Health and Disease in American Indians: The Strong Heart Study Elisa T. Lee, Ph.D., Oscar T. Go, Ph.D., for the Strong Heart Study Investigators Socioeconomic Status and Cardiovascular Disease in Rural Populations Thomas A. Pearson, M.D., M.P.H., Ph.D. Socioeconomic Status, Cardiovascular Risk Factors, and Cardiovascular Disease: Findings on U.S.Working Populations Jeremiah Stamler, M.D., Rose Stamler, M.A., Daniel Garside, Kurt Greenlund, Ph.D., Sujata Archer, Ph.D., James D. Neaton, Ph.D., Deborah N. Wentworth, M.P.H. Psychosocial Pathways Linking SES and CVD James S. House, Ph.D., David R. Williams, Ph.D. Stress, Work, Social Supports, and Gender in Relation to Cardiovascular Disease Kristina Orth-Gomer, M.D., Ph.D. Medical Care: Access, Utilization, and Cost Daniel B. Mark, M.D., M.P.H., Nancy Clapp-Channing, R.N., M.P.H., Lai Choi Lam, M.S., John C. Barefoot, Ph.D., Ilene C. Siegler, Ph.D., M.P.H., Redford B. Williams, M.D. Session III: Experiences in Educational and Preventive Interventions Across SES Groups Session III Overview L. Julian Haywood, M.D. - Session Chair What Have We Learned About Socioeconomic Status and Cardiovascular Disease From Large Clinical Trials? Jeffrey A. Cutler, M.D, M.P.H., Greg Grandits, M.S. Smoking Cessation, Socioeconomic Status, and Ethnicity Judith K. Ockene, Ph.D., John Harrington, Ph.D. Nutrition Intervention Studies Shiriki K. Kumanyika, Ph.D., R.D., M.P.H. Physical Activity Intervention Studies and Socioeconomic Status Daniel I. Galper, M.S., Steven N. Blair, P.E.D. Worksite, Community-Based, and School-Based Intervention Russell V. Luepker, M.D. Interventions Targeted to Ethnic Minorities, Especially Women Mary Helen Deer-Smith, R.N. National Cardiovascular Disease Education Programs Gregory J. Morosco, Ph.D., M.P.H. Appendix Chairs and Speakers Contributing Authors Coordination ----------------------------------------------------------- Executive Summary ----------------------------------------------------------- Executive Summary Jeremiah Stamler, M.D. Helen P. Hazuda, Ph.D. Cochairs Disease prevention is an essential part of the mission of the National Heart, Lung, and Blood Institute (NHLBI). Over the years, the Institute has developed some remarkably effective preventive strategies based on research to identify the major risk factors for cardiovascular disease (CVD). It is no surprise that the sharp declines in CVD mortality that occurred during the past 30 years coincided with our understanding that such factors as smoking, hypertension, high blood cholesterol, obesity, and diabetes increase a person's risk of developing CVD. Gratifying as this progress is, we still have far to go. Much evidence indicates that the beneficial trends in CVD mortality have not been felt equally across all segments of society. Rather, the most striking improvements in cardiovascular health have occurred among wealthier, better- educated Americans, while progress among groups with lower socioeconomic status (SES) has lagged. The observation that the gap between high-SES and low-SES populations may be widening is particularly disturbing. To address these important public health issues, the Conference on Socioeconomic Status and Cardiovascular Health and Disease was convened November 6-7, 1995, in Bethesda, Maryland. More than 120 persons attended, representing such fields as cardiovascular and preventive medicine, epidemiology and biostatistics, behavioral and social sciences, and health policy research. The goals of the meeting were to assess the extent to which SES is related to CVD mortality, morbidity, and risk factors in men and women of various ages, races/ethnicities, and geographical locations; to assess time trends in the SES-CVD association; to explore possible biological, psychosocial, and lifestyle- related pathways by which SES may relate to CVD; to identify strategies for reducing SES-based disparities in cardiovascular health; and to recommend promising avenues for future research on this topic. Several crosscutting themes emerged from the conference presentations and discussions. First is the need to improve our understanding of the concept of SES and the ways in which it reflects the conditions of everyday life for people of various strata. This task will involve development of more sophisticated and refined measures of SES. The importance of seeking input from other related fields (e.g., sociology, demography, economics) was emphasized. The conference also highlighted the importance of rapidly transferring new scientific knowledge into practice and, most critical, of more effectively applying what we already know about risk reduction strategies (e.g., smoking cessation, blood pressure control, diet), especially to population strata that have not yet been reached. In this regard, the superb national leadership taken by the NHLBI through its education programs was repeatedly acknowledged. Irrefutable evidence that an SES gradient in CVD risk exists - that there is room for improvement at every SES level - speaks to the importance of a population-wide approach to reducing the burden of CVD. Sustained and focused efforts among all SES groups of various races/ethnicities are needed. SESSION HIGHLIGHTS Session I: Setting the Stage The first session focused on U.S. national data relating SES to CVD morbidity and mortality; to lifestyle, biomedical, and psychosocial risk factors; and to medical care utilization. Presenters gave an overview of data in the Chartbook of U.S. National Data on Socioeconomic Status and Cardiovascular Health and Disease, which had been prepared as background to the conference. The national data document that SES relates to CVD, to lifestyles, and to major lifestyle-related biomedical CVD risk factors (with variations by age, gender, ethnicity, and specific aspects of lifestyle, as well as with variations over time). The general finding is that more adverse patterns of these variables exist for lower-SES strata than for higher-SES strata in the U.S. population. For instance, a wealth of information on smoking shows strong and consistent relationships with SES (i.e., for all age groups, both genders, and all racial/ethnic groups, the lower the SES, the higher the prevalence of smoking and the lower the quit rate during recent decades). Strong and consistent inverse relationships also prevail - for adults of every age and both genders, for blacks, non- Hispanic whites, Hispanics, Asian Americans, Native Americans - between SES and high blood pressure, and between SES and obesity. Lower-SES groups have higher rates of both hypertension and obesity than higher-SES groups. Upward trends in body weight and in prevalence of overweight during the last decades have been more marked in lower- than higher-SES strata. Other SES-risk factor relationships are less strong and consistent - for example, the relationship between SES and serum cholesterol. However, trend data from successive national population surveys indicate that declines in adult average serum cholesterol levels have been smaller in lower- than in higher-SES strata. National data are limited about such relevant aspects of lifestyle as diet and physical activity and their time trends across SES strata. National data are also sparse with respect to psychosocial risk factors across SES strata. In particular, data are lacking on psychosocial variables for SES groups of various ethnic backgrounds. Furthermore, little information exists about such variables as job change, unemployment, job and income instability, work-related psychosocial variables, social support, and social isolation, and how these variables influence relationships of SES to CVD, CVD risks, and their trends. As the Chartbook documents and the presenters noted, national mortality data show clearly that for major ethnic groups in the U.S. population, lower-SES strata have higher mortality rates than higher-SES strata for coronary heart disease (CHD), all CVD, and all causes. National morbidity surveys give generally concordant findings. Discussants noted that SES is related to many aspects of daily life, experiences at work and in the community, stresses and strains, and access to and utilization of medical care, including preventive services and state-of- the-art diagnostic and therapeutic services. To understand better how SES influences CVD, an epidemiology of everyday life that encompasses these phenomena needs to be developed. Recommendations * Promote measurement of SES, in both observational and interventional research, using established, valid, reproducible, acceptable indices appropriate for a given study and its aims. Encourage data analysis approaches that include use of information on SES as both a control and a stratification variable. * Develop, validate, and incorporate more sophisticated measures of SES and SES-related aspects of everyday life into research with a major focus on SES and CVD. In studies of women, minorities, and rural residents, give explicit attention to special features that may condition the definition, meaning, and impact of SES. * Investigate the relationships between SES and CVD initiation, progression, prognosis, morbidity, disability, and death. Emphasize greater use of methods to measure preclinical disease (e.g., echocardiography, electrocardiography, sonographic measurement of carotid artery intima-media wall thickness, measurement of ankle-arm blood pressure ratio) to achieve earlier and more comprehensive assessments of the impact of SES- related exposures on CVD. Link such findings to data on nonfatal and fatal clinical CVD. Session II: Pathways Linking SES and CVD Several presenters reviewed data on SES-CVD relationships and on possible pathways of these relationships, in men and women from specific U.S. population strata (i.e., blacks, whites, Hispanics, Asians and Pacific Islanders, Native Americans, rural populations, and employed groups). Two other presenters discussed evidence from psychosocial studies and from research on stress, work, and social support. The last speaker addressed medical care access, utilization, and costs. In agreement with national data, findings from studies of specific U.S. population groups generally show an inverse relationship between SES and CVD (i.e., higher CVD rates with lower SES). Two broad pathways were noted linking SES and CVD: 1) less favorable patterns of established major lifestyle and biomedical risk factors (smoking, adverse diet, sedentary lifestyle, high serum cholesterol, high blood pressure, obesity, diabetes) in lower- compared with higher-SES strata; 2) less favorable patterns of psychosocial factors (hostility, depression, low social support, social isolation, racism, job instability- insecurity-strain- powerlessness, unemployment) in lower- compared with higher-SES strata. Data are sparse on possible SES-related biological mediators of relationships between psychosocial factors and CVD (e.g., altered sympathetic and parasympathetic nervous system function, altered hypothalamic-pituitary-adrenal axis function, altered cellular-molecular biology of key cells), and on social-environmental mechanisms whereby lower SES leads inordinately - from early childhood on - to development of more adverse behavioral and psychosocial patterns. Men with low levels of all established biomedical risk factors (blood pressure, cholesterol, smoking, diabetes, previous heart attack) have been found to experience lower CHD and CVD mortality rates than men without such a favorable profile. Such a favorable profile is found in less than 10 percent of the population. These findings indicate that controlling the established biomedical risk factors has the potential to achieve low CHD-CVD rates for all groups in the United States, including low-SES strata of various racial/ethnic backgrounds. These observations underscore the importance of primary prevention and control of the established major biological risk factors for CHD-CVD. Improved understanding of the reasons why lower-SES groups have more adverse levels of the major risk factors, the pathways producing these patterns, and the contributory role of more adverse behavioral and psychosocial patterns is key to achieving this goal. Also important is improved understanding of the lifestyle, behavioral, psychosocial, cultural, metabolic, and genetic components of maintaining low-risk status into middle age. Such knowledge will lead to development and application of better approaches for the primary (including "primordial") prevention and control of the major risk factors, with a focused emphasis on lower- as well as higher-SES strata of all racial/ethnic groups. Recommendations * Investigate relationships between SES and development/evolution of lifestyles, behaviors, and risk factors, including both adverse and favorable patterns. * Analyze existing data and data currently being collected by trials and demonstration projects to document further the effectiveness of interventions in lower- and higher- SES groups of various races/ethnicities; report and disseminate such data and their implications for programs to control CVD in lower-SES groups. * Conduct research to enhance understanding of mechanisms underlying clustering of adverse lifestyles, behaviors, and risk factors in lower-SES groups of various races/ethnicities. * Study SES and precursors of adult lifestyles and behaviors, from childhood on, including influences of the family, community, school, and workplace. * Elucidate the mechanisms whereby conditions of everyday life for lower-SES groups - exposures during prenatal life, infancy, childhood, and adulthood - contribute to the development of more adverse patterns of lifestyles, behaviors, risk factors, and psychosocial traits. Include studies of the roles of racism; sexism; deprivation; relative and absolute income levels and their trends; knowledge and attitudes; home, school, workplace, community, and mass-media exposures; peer pressures; housing conditions; relationships to organizations; early exposures to food, alcohol, and drugs; and barriers to adoption of heart-healthy behavior. * Investigate the interrelationships between psychosocial traits and lifestyles in lower-SES groups with the aim of clarifying environmental and biological mediators and pathways of these interrelationships (including neurological, endocrine, cellular, and molecular pathways) and thereby enhancing understanding of how psychosocial factors and acculturation influence CVD risk. Session III: Experience in Educational and Preventive Interventions Across SES Groups This session focused on preventive strategies that, if applied more broadly, have potential to reduce the SES gradient in CVD health and disease. Speakers reviewed evidence from clinical trials and a broad range of intervention studies to determine whether observed reductions in CVD risk factors, morbidity, and mortality extended equally to low- and high-SES participants, including those from ethnic minorities. Findings clearly indicate that, although there is more to be learned, much is already known about how to reduce CVD in low-SES groups. Interventions in several large NHLBI-supported primary prevention trials were efficacious in reducing CVD risk factors, morbidity, and mortality in multiple SES-race substrata, although only limited special efforts were made to tailor the interventions for lower-SES, nonwhite participants. Interventions consisted of antihypertensive drug treatment involving stepped care as well as multifactor behavioral interventions aimed at smoking cessation, reduction of total serum cholesterol, weight control, increased physical activity, and reduction of salt and alcohol intake. Lower SES was not a barrier to intervention success for either the drug treatment or the behavioral components. In one antihypertensive trial, the SES gradient in mortality was eliminated in the intervention group. Studies in communities, worksites, and schools confirm that public health interventions can be designed to benefit all segments of society. Favorable changes in lifestyles and lifestyle-related biomedical risk factors have been achieved across all SES groups in community interventions that have used a broad, multimedia communication approach with special efforts to target and involve low-SES, culturally diverse groups. Worksite interventions incorporating on-site classes and payroll incentives have also achieved favorable results in both blue- and white-collar employees. The success of school-based programs in achieving similar magnitudes of risk factor reduction in both low- and high- SES students is particularly noteworthy given the schools' potential for building lifelong heart-healthy habits and, thereby, contributing importantly to the key strategic goal of preventing the development of major risk factors. Preliminary findings from programs specifically designed to address the need for nutrition education materials suitable for English-speaking adults with limited literacy skills indicate that the approaches developed result in high utilization of intervention materials among ethnically diverse, low-SES persons and can lead to gradual, sustained progress toward favorable risk factor change over time. Most programs use multimedia, client-centered instructional approaches that include interactive computer technology, videotapes, audiocassettes, compact discs, and printed materials. Across all SES groups, self-help is the method of choice for smoking cessation by more than 90 percent of smokers. Programs that combine media presentations with telephone counseling hotlines or distribution of self-help materials have been particularly effective with lower-SES smokers. A community organization approach used in one study had the greatest effect in low-SES individuals. Counseling of smokers by physicians or dentists has also been shown to increase their likelihood of quitting smoking significantly. Sedentary lifestyle is particularly prevalent among lower- SES persons and ethnic minorities. Experience from community-based heart-health programs suggests that interventions promoting moderate- rather than vigorous- intensity activities are more likely to be successful, and that exercise campaign events tied to preexisting community structures or traditional community events have the greatest participation levels. Worksite interventions have demonstrated that clinically significant increases in physical activity can be achieved in both low- and high-SES employees. The National High Blood Pressure Education Program and National Cholesterol Education Program have been prime movers in the substantial progress made in achieving the Healthy People 2000 blood pressure and cholesterol objectives. Their achievements underscore the effectiveness of science-based public health strategies that rely on broad-based cooperation between government and the private sector in overcoming the barriers to reducing CVD in all sectors of the population. The overall conclusion is that more widespread application of interventions already known to work across multiple SES and ethnic groups can make a substantial contribution to eliminating the SES gradient and ending the CVD epidemic in all segments of society. Recommendations * Incorporate nonsmoking messages into educational efforts targeting all SES-ethnic groups, with a particular focus on more habituated smokers and on do-it-yourself approaches to smoking cessation. * Evaluate existing obesity intervention programs; look for successes and develop approaches based on them, with the aim of achieving national goals for obesity prevention and control for all SES strata, including reversal of the decades-long rise in obesity rates among children, youth, and adults. * Intensify efforts to increase consumption of heart- healthy foods - reduced in total fat, saturated fat, cholesterol, salt, refined sugars, and calories - in lower-SES communities. * Enhance efforts to prevent and control high alcohol intake among lower-SES groups of various races/ethnicities. * Enhance efforts to achieve daily or near-daily physical activity by lower-SES groups from childhood on and to increase the proportion of people who regularly engage in moderate activity. * Emphasize the potential contribution of stress reduction to modifying risk factors and subsequent CVD morbidity and mortality. * Encourage assessment of literacy as a relevant aspect for improving ability to intervene effectively in lower-SES groups; use group-specific programs appropriate for literacy level and sensitive to group culture. * Develop effective strategies to promote favorable behavior changes in lower-SES groups. * Conduct demonstration research in communities, workplaces, and schools with persons from lower-SES groups of various ethnicities. Include participation of community outreach workers (including workers trained and supervised by nurses and dieticians) and community representatives in all aspects of such projects (i.e., planning, intervention, data evaluation) to enhance the potential for lasting accomplishments. REPORT ORGANIZATION The report that follows is organized in three sections that parallel the structure of the conference. Each section includes an overview of the session, prepared by its chair, and summaries of the individual topical presentations, prepared by the presenters and their colleagues. Rosters of conference chairs and speakers, contributing authors, and NHLBI coordination staff are provided at the end of this report. ----------------------------------------------------------- Session I Overview Setting the Stage ----------------------------------------------------------- Session I Overview Millicent Higgins, M.D., Chair Session I presented highlights of the very extensive data gathered in the Chartbook of U.S. National Data on Socioeconomic Status and Cardiovascular Health and Disease and included discussions of measurement of socioeconomic status (SES) and the nature of its association with cardiovascular disease (CVD). These presentations, as well as comments from conference participants, are reflected in this overview. Socioeconomic Status The Chartbook used education, income, and occupation as measures of SES; each is strongly related to some aspects of CVD mortality, morbidity, and risk factors, and these measures are correlated with one another. However, each provides some unique information and has its strengths and weaknesses. Completion of formal education usually precedes the onset of CVD and is not influenced by the disease process. Because education is stable, it can be used over time and throughout the adult age range. Income, on the other hand, can vary over time and may be a less valid measure of SES at the extremes of the age range. Personal income and family income may be hard to ascertain or interpret. Measures such as the poverty-to-income ratio were reported in some studies, but little attention has been given to measuring assets such as home ownership, value of housing, and car ownership; assets as well as income need further consideration, especially for elderly populations. Occupation is predominantly useful for those who are employed and at the age where having a job is the norm. Occupations may change over time and be influenced by the presence of disease, as well as influence the development of disease. An advantage of occupation is that it has been used extensively in Europe to define social class, but there are issues of comparability and of combining categories for international comparisons. Ownership of a business and authority in the workplace are sometimes used to classify people according to their occupations. Classifying married women according to their husbands' occupation or social class may be less appropriate now than in the past, and inappropriate in some societies. Education, occupation, and income do not cover some aspects of social class and may be inadequate to describe associations or suggest mechanisms by which social class is related to health and disease. Additional social and demographic variables, such as living alone and marital status, are available from the National Longitudinal Mortality Study; deprivation was assessed in another study. Most presentations and data described and evaluated SES as an attribute of individuals, though household income and family size were assessed in a few studies and familial aggregation and cultural inheritance of social class were mentioned. SES of an individual or family should also be considered in context with respect to place, time, and the social environment of the community. The appropriateness of descriptors of SES varies with sex, age, and race; over time; and from culture to culture. Cardiovascular Health and Disease Measures of CVD included extensive national mortality data and a little information on case fatality rates. Limitations of death certificate information are well known and include inaccurate recording and coding of specific causes of death. Measures of morbidity were fewer; prevalence is often based on self-reports of disease and is influenced by ascertainment, by incidence and duration of illness, and by case fatality. Incidence data are sparse. Measures of severity of disease and of access to care were limited to information on hospital admissions, physician visits, and performance of diagnostic and therapeutic procedures for CVD. Recent information on biological risk factors including blood pressure, serum cholesterol, other lipids, and body mass index was provided by the third National Health and Nutrition Examination Survey (NHANES III). Lifestyle factors are not as easy to measure precisely, but information on diet and physical activity and their relationships to SES variables were presented. There is a wealth of information on smoking that shows very strong and consistent inverse relationships with SES. The Chartbook and presentations also included information on the frequency and distribution by SES of some psychosocial risk factors including life events, social networks, and personality characteristics. Some psychosocial variables were strongly related to SES. Although preclinical evidence of disease was not discussed in Session I, it is an important component of CVD that can be assessed using markers such as carotid artery intimal- medial wall thickness and ankle-arm blood pressure ratio. These newer measures allow studies of a broader range of cardiovascular conditions from fatal events, to type and severity of clinical presentations, preclinical disease, and risk factors. CVD-SES Relationships Some consistent patterns were apparent, particularly in recent U.S. data. Low SES was related to CVD morbidity, mortality, and some risk factors, and to less utilization of some components of medical care. Age, sex, and race, as well as socioeconomic factors, are related to access and use of medical care. Some of the data suggest conflicting patterns. Outpatient visits are as frequent or more frequent in poor people. The literature suggests that a given disease diagnosis costs more in poor people because they present later in the disease and need a higher level of care and more procedures. Utilization data paired with incidence, prevalence, and mortality data are needed. Some of the risk factor relationships with SES are not as strong or consistent in the three racial/ethnic groups (i.e., whites, blacks, Hispanics), in men and women, and across the age range. Researchers need to take these variations into account, as well as to assess the quality and strength of the science base in order to increase understanding of the complex interactions between SES and CVD. There were marked gradients in CVD among occupational groups in European countries, and considerable variation in the strengths of associations between SES and mortality from all CVD, ischemic heart disease, and stroke. Intriguing differences were apparent in northern Europe compared with southern Europe. United Kingdom data provided detail about the gradient of risk across the range of SES and the extent to which occupational level and income provided similar or different insights. Information on deprivation was used as another marker of SES differences in these studies. Some of the European data suggested that income was not the main determinant of the SES gradients, which were just as great in countries where income inequalities were smaller as in countries in which they were large. Time trend data were available from the United States over recent years and from the United Kingdom over a longer period. There has been a reversal in the relation between SES and CVD mortality in men, but the pattern in women has not changed. Prior to the mid-1960s, mortality rates were higher for men in the upper SES groups but then the situation reversed. In fact, the gap between extremes of SES strata has widened recently, especially among men in the United States. In some developing countries, mortality from CVD is still greater at the upper end of the SES distribution. It is clear that SES differences in CVD morbidity and mortality are mediated, in part, by the major risk factors, but there is a component of the SES-CVD relationship that is not explained by those associations. Despite the limitations of available data, and imperfect measures of SES, there is strong evidence that SES is a major determinant of cardiovascular health and disease and that further research into the nature of these relationships is warranted. RESEARCH NEEDS AND RECOMMENDATIONS * Better measures of SES are needed, especially for women and minorities. Definitions and methods should be validated and standardized. Information is lacking or sparse for several racial and ethnic groups and for rural residents who have not been studied extensively in the past. More sophisticated measures and multiple measures of SES and of aspects of everyday life are desirable, but a simple measure - such as education - would be a useful addition to observational studies and clinical trials, which cannot collect more detailed information. * Further investigations of relationships between SES and development of risk factors, severity, and course of CVD as well as diagnosis, treatment, and utilization of medical care are needed. Such information would identify groups where the need for prevention or therapy is greatest and improve understanding of the ways in which SES variables influence initiation and course of CVD. * Circumstances are changing rapidly, and relationships between SES and components of CVD need continued monitoring. National and regional data are needed for the United States and other countries to expand the range of experience and increase understanding of the ways in which SES and cardiovascular health and disease interact. * Relationships between psychosocial risk factors and biomedical risk factors are in need of further study. The biological pathways by which psychosocial factors and acculturation influence risk must be understood better to improve prediction and prevention of CVD. * More knowledge is needed about the precursors of adult health behaviors, including the influence of the family and the community on health behaviors. Many healthy and unhealthy behaviors are learned in childhood in the home and they aggregate in families. Children and families should be included in research on SES and cardiovascular health and disease. * Research is needed to identify effective strategies to promote behavior change in low-SES groups as well as to improve recruitment and retention of such people in observational studies and clinical trials. * Studies of cost should be added to our research projects. Information about how SES affects cost-effectiveness of preventive and therapeutic approaches is lacking. Policymakers need to know more about health care costs and their relationship with SES. Measures of utilization such as physician visits, hospitalizations, and use of procedures should be related to measures of disease frequency and severity, but these are not generally available and they require well-designed investigations for their collection, analysis, and evaluation. ----------------------------------------------------------- Biologic and Methodologic Approaches to the Association Between Socioeconomic Factors and Cardiovascular Disease George A. Kaplan, Ph.D. Examination of the data presented in the Chartbook of U.S. National Data on Socioeconomic Status and Cardiovascular Health and Disease, as well as numerous other sources, reveals an inverse gradient between various measures of socioeconomic status (SES) and mortality and morbidity from the major cardiovascular diseases (CVD) (1). The "better off" experience less disease, and the "worse off" more disease. This pattern of increased risk of poor health outcomes with decreasing SES is found for most, but not all, diseases across the world, and when using a variety of SES measures (2). There are some exceptions to this pattern, and it has not always been true for all groups for CVD. However, in virtually all the developed countries at the current time for which there are good data, higher SES is associated with lower CVD risk. These are compelling and remarkably consistent findings. The observation of consistent, patterned differences in risk of disease between groups or places is the starting place for epidemiological analysis. Although it is important to continue to document the nature of SES inequalities in CVD, a full understanding will require a clarification of the operative biological and social pathways. Parallel with passage through various stages of the life course, there are exposures to a variety of socioeconomic states and risk factors, all linked over time (Fig. 1). Just as risk factors may vary in importance depending on age and stage of disease progression, different aspects of SES may loom more or less important over time. Thus, researchers should not expect either a single risk factor or a single aspect of SES to be equally important at all stages in the natural history of coronary heart disease (CHD). Similarly, patterns of exposure to socioeconomic factors and risk factors are dynamic, not static. Risk factor exposures change over time, and exposures to socioeconomic conditions may also change over time. These changes, as well as the converse - the extent to which they track over time - may have important implications for understanding of the links between SES and cardiovascular health. The links between SES and risk factors are, of course, of paramount importance in understanding the association between SES and CHD. While some of the epidemiological literature on this association attempts to establish SES as an "independent" risk factor, it must perforce act through more proximal pathways. Thus, it is important to understand these links between measures of SES and the existing risk factors for CHD. Socioeconomic Factors in the Natural History of Coronary Heart Disease In understanding how SES factors might be associated with the development and manifestations of CHD, it is convenient to divide the natural history of CHD into three major stages: preclinical disease, triggers and events, and recovery. What follows, a discussion of the pathophysiological links between SES and these stages of CHD, is based, for the most part, on consistent reports in the literature. In some cases, in the absence of any information, it is based on biological plausibility. ----------------------------------------------------------- Figure Missing Figure 1. - Relationship Between Age, SES, Risk Factors, and the Natural History of Atherosclerotic CHD. ----------------------------------------------------------- The molecular and cellular events leading over the long term to atherosclerotic changes have been summarized by a number of authors (3). The preclinical disease stage refers to the early stages of alteration of the arterial walls and environment in which there are neither major obstructions to blood flow nor symptoms. In this stage, there is the gradual progression from fatty streaks to hard plaque with calcification and, ultimately, clinically significant obstruction. The focal role of endothelial injury stemming from mechanical, immunological, viral, or other forces is well recognized, with subsequent endothelial dysfunction leading to disturbances in interactions between the endothelium and macrophages, platelets, smooth muscle cells, T lymphocytes, among others, and changes in vasomotor properties of the artery. Changes in the nonthrombogenic character of the endothelium lead to platelet adhesion, accelerating lesion development via increased levels of platelet-derived growth factor and other substances and leading to smooth muscle cell proliferation and migration. Table 1 indicates a list of factors that are related to SES, variously measured, and are also related, in some studies, either to coronary atherosclerosis, found on angiography or autopsy, or, as a model of coronary atherosclerosis, to carotid atherosclerosis or progression of carotid atherosclerosis. Given the recent interest in the early origins of CVD, it is important to point out that many maternal exposures influence prenatal development and might, conceivably, have some impact on later atherosclerotic developments. While the pathways are complex, this might be seen as the intrauterine transmission of SES. Considerable evidence now indicates that the standard coronary risk factors (blood pressure, smoking, obesity, lipids, alcohol, physical activity) are related to preclinical atherosclerosis, and in many studies higher levels of these risk factors are found in lower SES groups, although the consistency of this is weakest for lipid measures. SES is also associated with fibrinogen levels, and possibly other pathways influencing both fibrinolysis and coagulation (4). Although less studied, a variety of psychosocial factors also seem to be strongly associated with coronary and extracoronary atherosclerosis and SES. In one study, 4-year progression of carotid atherosclerosis was associated with levels of hopelessness (5). Hopelessness, in turn, was inversely related to income level. The effects of psychosocial variables on atherogenesis may operate via promoting higher levels of other, more traditional risk factors, or by potentiating the effects of these other risk factors via changes in lipid peroxidation, fibrinolysis, or other mechanisms. In addition, tendencies toward chronic vascular hyperactivity to chronic psychological stress, which may be more prevalent among those who are of lower SES, may also be associated with increased atherogenesis. When we consider the second stage, that involved in the precipitation of acute events, SES may also have an important role. In this stage of the natural history of CHD, the important events are primarily those related to plaque fissuring and instability, and a cascade of additional factors and events leading to an occlusive thrombus, or other complications (Table 1). Of course, coronary spasm secondary to atherogenic changes or superimposed on these changes, and arrhythmias also are of considerable importance in this stage. Given the effects of smoking, physical activity, and alcohol consumption on coagulation and fibrinolysis, it is not surprising that factors likely to be associated with SES could potentially act as triggers of acute events. Many investigators now believe that factors such as anger, disruption of social ties, mental stress, and emotional distress, and their associated hemodynamic, neuroendocrine, and hemostatic effects may be major contributors to the timing of acute events. Importantly, existing data suggest that these events are more common among those with lower SES. An example of the magnitude is found in a study of silent myocardial ischemia, assessed by radionuclide ventriculography, in response to mental stress in patients with coronary artery disease (6). When patients were asked to give a short speech about personal faults and undesirable habits in front of others, presumably a quite stressful experience, they evidenced almost as high a frequency of wall motion abnormalities as when they underwent a graded maximal exercise test. ----------------------------------------------------------- Table 1. - SES and CHD SES and Preclinical CHD: Prenatal exposures; blood pressure; smoking; obesity; physical inactivity; diet; lipids; alcohol; hostility/anger; social instability; depression; job strain; hopelessness; vascular reactivity SES and CHD Triggers: Smoking; heavy exertion; alcohol; anger; social instability; mental stress; intense emotional distress; hyperreactivity SES and CHD Recovery: Increased severity; greater comorbidity; poorer access; poorer quality of care; poorer adherence; socioenvironmental factors + + + + + + + + Factors identified with SES are associated with CVD and events at multiple stages. The preclinical disease stage (left panel) refers to the early stages of alteration of the arterial walls and environment in which there are no symptoms or major obstructions to blood flow. In this stage there is the gradual progression from fatty streaks to hard plaque with calcification and, ultimately, clinically significant obstruction. Reflected at left are various factors related to SES, linked in some studies either to coronary atherosclerosis on angiography or autopsy, or to carotid atherosclerosis or progression of carotid atherosclerosis. Certain of these factors are likely to be associated with both SES and events that occur in the triggering and acute stage (center panel). Finally, SES could be related to poorer recovery from acute events via a number of pathways (right panel). [end Table 1] ----------------------------------------------------------- SES could be related to poorer recovery from acute events via a number of pathways (Table 1). Poorer recovery may reflect greater severity of disease and greater comorbidity, each of which would be expected to be present with lower SES. Each of these could be seen, to some extent, as a failure of primary and secondary prevention. For example, a number of studies have shown relatively limited access to preventive medical care and screening among those who are poorer or who live in poorer areas. It is of some interest that most of the mortality differentials associated with educational differences in the Hypertension Detection and Followup Program were eliminated in the stepped-care treatment group (7). Overall access to care and to health coverage varies by SES, and poorer access or absence of insurance may translate to lower quality of care. There is also some evidence that state-of-the-art interventions are available mainly for those who are higher SES. While it fits into the other stage also, there is evidence that those who live in poorer areas also have less access to healthy foods, and are targeted by the tobacco and alcohol industries. In addition, rehabilitation and supportive services may be less available to those who are of lower SES. Finally, more difficult working conditions and neighborhoods, greater family problems, and greater demands in other domains may all contribute to worse recovery among those in lower SES groups. All of these factors would translate into poorer survival and recovery after acute events. Conceptualization and Measurement of Socioeconomic Factors SES factors that might influence cardiovascular health (Table 2) can be measured in many ways. Terms that describe such measurements are often used within epidemiological analyses without proper recognition of their intellectual roots in sociology (8). While it is probably not necessary for epidemiologists to concern themselves with the arcane world of sociological theory, a general reading helps the epidemiologist to be aware of differences in meaning and emphasis that may be of some importance in understanding both the measurement of socioeconomic factors and their impact on health. Not all measures of SES are equivalent. Each carries with it different problems in measurement, interpretation, and implications with respect to underlying causal pathways. For example, education is taken by some as the best measure because it is easy to measure with high response rates, valued by the researcher, and not subject to reverse causation - health problems in adulthood cannot logically cause differing levels of educational achievement at an earlier time. However, educational achievement is often socially patterned, so highest level of education completed may simply be a proxy for whatever forces are important in determining completion of school. Because there are still large discrepancies in educational opportunities between groups that vary by gender, race or ethnicity, and class, level of education should not necessarily be seen as reflecting forces that are solely located within the individual. There are also large variations by birth cohort in level of education completed - with a general trend toward more and more education in more recent birth cohorts. Thus, the behavioral, social, and psychological factors associated with a given level of education may vary by age and birth cohort. In such a case, asking if the association between education and CHD varies by age may raise numerous interpretive problems. Finally, the consequences of more or less education (e.g., income, housing, job security) may vary by age, race or ethnicity, and gender. Income is less often measured in epidemiological studies because of somewhat higher nonresponse rates and the possibility that poor health might lead to declines in income. However, income is critically important, as it provides access to other beneficial things (e.g., goods and services, high-quality education, medical care, good housing). To not measure it is to miss the potential links between material conditions and health. However, measuring income is not simple: individual or family income can be measured, it can be adjusted for family size or not, and noncash benefits such as Medicare or food stamps can be included or not. ----------------------------------------------------------- Table 2. - Socioeconomic Factors Influencing Cardiovascular Health Socioeconomic status; social class; education; income; occupation; employment status; occupational grade; living conditions; wealth; housing tenure; car ownership; deprivation; poverty; income inequality [end Table 2] ----------------------------------------------------------- Economic resources as represented by wealth may also be important to measure, although this has not been done very much in epidemiological analyses. Measures of wealth based on such assets as bank accounts, stocks and bonds, pensions, and home equity may be critical as they represent resources that may potentially buffer the impact of stressors on health, or may make certain health care resources more available after acute events. More commonly, measures of assets such as car ownership and home ownership have been used, although in the latter case there has seldom been information on size of mortgage. Occupation is often measured, although the focus is generally not on specific agents that might be associated with increased risk. Instead, categorizations are used that involve status, roles, power, prestige, lifestyle, job characteristics, income, education, traditions, beliefs, and values. For example, a reasonably consistent set of 12 categories has been used since 1910 by the U.S. Bureau of the Census, and the Registrar General of Great Britain has used a categorization since 1911 (currently six categories). It is also possible to categorize people within a broad occupation, such as the classification of civil service workers in Great Britain as administrative, professional- executive, clerical, or other (9). General problems with occupational classification include a usual lack of occupational histories so that downward mobility due to health problems cannot be ruled out, substantial heterogeneity within occupational classifications, and large secular trends in the jobs that fall in specific categories. Most important, such broad classifications do not go very far in pinpointing what characteristics, in particular, are associated with increased risk. Some progress in that regard has been made in the recent work relating job strain and CHD, which suggests that people in jobs characterized by high psychological demands and low decision authority may be at increased risk (10). It is also important to consider employment status because that, too, has been linked to disease and health outcomes. Many studies use measures of SES that are based on characteristics of the areas in which people live. Characteristics of areas [e.g., census tracts, block groups, zip codes, Standard Metropolitan Statistical Areas (SMSAs)] that are used include income, education, proportion in poverty, occupational distribution, crowding, housing conditions, relative equity of income distribution, and many other indicators. Often, the choice of an area-based measure of SES reflects the lack of availability of individual measures. However, area-based measures can capture characteristics of the locations in which people live that are different from individual characteristics, and these area or community characteristics may be of great significance for health. For example, one study showed that residence in a poverty area was associated with increased risk of death over and above a variety of individual characteristics (11). In many analyses in which information about SES is not available, classifications of race or ethnicity are used instead. While SES levels do vary by race and ethnicity, substituting race and ethnicity for social class information leaves much to be desired. For example, examination of the distribution of whites, blacks, and Hispanics by quintiles of income for the United States shows that there is substantial overlap of the three populations in the middle of the income distribution. This overlap means that there will be substantial misclassification if race or ethnicity and SES are treated as identical. In addition, race, ethnicity, and SES represent different systems of stratification with implications for health which cannot be held equivalent. Indeed, associations between SES indicators and cardiovascular outcomes may vary by race and ethnicity. For these reasons, and others, race and ethnicity should not be used as proxies for SES information. Understanding the Connection between Social Class and Cardiovascular Disease The most common way of addressing our understanding of the association between SES and CVD has been to use multivariate models to examine the extent to which particular risk factors account for the association and the extent to which the association is "independent" of these risk factors. Probably the most exhaustive attempt at this examined the association between income quintiles and risk of death over approximately 6 years, with adjustment for age and 22 additional risk factors (plasma fibrinogen, serum HDL, serum apolipoprotein B, blood leukocytes, serum copper, mercury in hair, serum ferritin, blood hemoglobin, serum triglycerides, systolic blood pressure, body mass index, height, cardiorespiratory fitness, cigarette smoking, alcohol consumption, leisure-time conditioning physical activity, depression, hopelessness, cynical hostility, organizational participation, quality of social support, and marital status) (12). With adjustment for these 22 risk factors, there was no longer an association between income and risk for either all-cause or cardiovascular death. This pattern of results (i.e., a decrease in an association with addition of potential confounder) is commonly taken as indicating that the association between SES and disease has been explained. ----------------------------------------------------------- Figure Missing Figure 2. - Critical Components of the Relationship Between SES and CVD. ----------------------------------------------------------- However such an approach takes on only part of the task of understanding (12). For example, many of these risk factors show trajectories that reflect the influence of SES (2). Thus, children from poorer families are more likely to begin smoking. Because SES is causally antecedent to smoking, smoking cannot be said to explain the association between SES and CVD. What this type of analysis does is to help us understand how SES manifests itself in increased risk of disease. Much more is still to be learned by such analyses, as we dissect out the critical set of pathophysiological pathways that result in the inverse association between SES and cardiovascular outcomes. From a public health point of view, it is unlikely that the identification of these pathways will tell us much about how to lower the excess cardiovascular mortality among those who are less "well-off." The problem is not likely to be solved by pharmacological means or other physiological interventions. Instead, we will have to emphasize study of why particular risk factors are differentially distributed by SES (13). Such an approach will require examination of multiple factors (Fig. 2) and include an appreciation of life-course trajectories of SES and risk factors. Such a perspective exposes a variety of loci in which the connections between SES and CVD can be broken. Many of these loci are not the usual focus of CVD prevention efforts. However, as Geoffrey Rose (1992) pointed out so eloquently in his final book, The Strategy of Preventive Medicine, "The primary determinants of disease are mainly economic and social, and therefore its remedies must also be economic and social." REFERENCES 1. Kaplan GA, Keil JE: Socioeconomic factors and cardiovascular disease: a review of the literature. Circulation 1993;88:1973-1998. 2. Haan MN, Kaplan GA, Syme SL: Socioeconomic status and health: old observations and new thoughts. In Pathways to Health: The Role of Social Factors, edited by JP Bunker, DS Gomby, BH Kehrero. Menlo Park, CA: Henry J. Kaiser Family Foundation, 1989, pp. 76-135. 3. Ross R: The pathogenesis of atherosclerosis: a perspective for the 1990s. Nature 1993;362: 801-809. 4. Wilson TW, Kaplan GA, Kauhanen J, Cohen RD, Wu M, Salonen R, Salonen JT: Association between plasma fibrinogen concentration and five socioeconomic indices in the Kuopio Ischemic Heart Disease Risk Factor Study. Am J Epidemiol 1993;137:292-300. 5. Everson SA, Kaplan GA, Goldberg DE, Salonen R, Salonen JT: Hopelessness predicts 4-year progression of carotid atherosclerosis among eastern Finnish men. 1995, submitted for publication. 6. Rozanski A, Bairey CN, Krantz DS, Friedman J, Resser KJ, Morell M, Hilton-Chalfen S, Hestrin L, Bietendorf J, Berman DS: Mental stress and the induction of silent myocardial ischemia in patients with coronary artery disease. New Engl J Med 1988;318:1005-1012. 7. Tyroler HA: Socioeconomic status in the epidemiology and treatment of hypertension. Hypertension 1989; 13 (Suppl):I94-97. 8. Susser M, Watson W, Hopper K: Sociology in Medicine, third edition. New York: Oxford University Press, 1985. 9. Rose G, Marmot MG: Social class and coronary heart disease. Br Heart J 1981;45:13-19. 10. Schnall PL, Landsbergis PA, Baker D: Job strain and cardiovascular disease. Ann Rev Public Health 1994;15:381-411. 11. Haan M, Kaplan GA, Camacho T: Poverty and health: prospective evidence from the Alameda County Study. Am J Epidemiol 1987;125:989-998. 12. Lynch JW, Kaplan GA, Cohen RD, Tuomilehto J, Salonen JT: Do known risk factors explain the relationship between socioeconomic status, risk of all-cause mortality, cardiovascular mortality, and acute myocardial infarction? 1995, submitted for publication. 13. Kaplan GA: Where do shared pathways lead? Some reflections on a research agenda. Psychosom Med 1995;57:208-212. 14. Rose G: The Strategy of Preventive Medicine. Oxford: Oxford University Press, 1992. ----------------------------------------------------------- Socioeconomic Status and Cardiovascular Disease Mortality: National Longitudinal Mortality Study Paul D. Sorlie, Ph.D. Norman J. Johnson, Ph.D. Eric Backlund, M.S. The National Longitudinal Mortality Study (NLMS) is a prospective study of mortality in persons who were interviewed during various Current Population Surveys conducted monthly by the Bureau of the Census (1). Detailed social, economic, and demographic information on a sample of the U.S. population is obtained by household interview. These surveys provide, among other information, the monthly unemployment rate in the United States. This analysis uses information from approximately 600,000 persons age 25 years or more. The National Death Index was used to identify approximately 65,000 persons from these surveys who died during the years 1979 through 1989. Underlying cause of death was obtained from the death certificate. A person- years approach for calculating annual death rates and a proportional hazards model for calculating relative risks were utilized to describe the relationship between education, income, and mortality from cardiovascular disease (CVD). Our analyses focus on the following: * the strength of the income and education associations with mortality; * the independence of the associations when income, education, and other socioeconomic status (SES) characteristics are considered together; * the consistency of the associations in ethnic and gender groups; and * the differences in the strength of associations in the NLMS relative to the Kitagawa-Hauser study in 1960 (2). Figure 1 presents the relative mortality from CVD for categories of family income for men and women 25 to 64 years of age. When adjusted for age and race, there is a steadily declining mortality by increasing income so that the mortality at $50,000 or more is one-third that of the lowest income categories. After further adjustment for SES (i.e., education, marital status, employment status, and household size), the relationship is not as strong, as would be expected, but there is still a substantial effect of income independent of these other variables. Figure 2 presents the relative mortality from CVD by education. Again, CVD mortality decreases with increasing education; as seen for income, the risk at the highest education level is about one-third that at the lowest. This relationship remains evident after adjustment for SES (i.e., income, marital status, employment status, and household size). It is noteworthy that the steepest declines in mortality risk occur at the highest levels of education, namely, beyond high school. ----------------------------------------------------------- Figure Missing Figure 1. - Relative Cardiovascular Mortality by Income. Estimates were obtained from the proportional hazards model using indicator variables for each income category with the lowest income as the reference point. The y-axis displays the relative risk on a logarithmic scale. Source: NLMS. ----------------------------------------------------------- Figure Missing Figure 2. - Relative Cardiovascular Mortality by Education. High school graduate is used as the reference group. Source: NLMS. ----------------------------------------------------------- Figure 3 reflects CVD mortality rates for three levels of education - less than high school, high school, and greater than high school. The annual mortality rates are shown for blacks, whites of Hispanic background, and non- Hispanic whites. Each group shows strong and steady decreases in CVD mortality with increasing education. The last question to be addressed is how these data compare to similar data collected in 1960 from the Kitigawa-Hauser Study (2). These results (Fig. 4) are courtesy of Drs. Preston and Elo of the University of Pennsylvania, who analyzed the NLMS data available in a public use data set (3). The 1960 data (left) indicate that the difference in the death rate from the lowest education to the highest was 3.9 deaths per 1,000 for men 25 to 64 years of age. Dividing this difference by the annual death rate (8.0 per 1,000) to obtain a ratio of the death rate reveals that the difference is 0.49 of the death rate for that group as a whole. NLMS data for the 1980s (right) show a difference of 4.1 deaths per 1,000 (about the same as in 1960), but the ratio of 0.80 is much higher since the average death rate is lower. Because this graph is plotted on a log scale, the slope for the 1980s is steeper than that for 1960. Thus, for this age group of men, the impact of education on mortality is stronger now than in 1960. ----------------------------------------------------------- Figure Missing Figure 3. - Cardiovascular Mortality by Education, Sex, and Race: Age 45 to 64 years. Source: NLMS. ----------------------------------------------------------- Figure Missing Figure 4. - Estimation of Slope Index of Inequality: White Men, 25 to 64 Years. The slope index of inequality represents the decrease in the death rate from the lowest education to the highest education, with education scaled as a percentile. The bars indicate the death rates for each level of education. The width of each bar represents the percent of population at each education category. The line shows the slope of a regression of the death rates on education scaled as cumulative percentiles of the actual education levels. "Difference" is the difference in the death rate from the lowest to the highest education level. "Average death rate" is the death rate for the group as a whole. Source: Preston and Elo (3). ----------------------------------------------------------- Table Missing Table 1. - Magnitude of the Education Versus Mortality Gradient: Ratio of the Decrease in the Mortality Rate to the Average Mortality Rate* *Slope index of inequality / Mortality rate Source: Preston and Elo (3). ----------------------------------------------------------- Table 1 shows the difference in mortality across the education scale (slope index of inequality) as a ratio of the death rate for each sex and age group. This is the same statistic that was described in the Figure 4 (note the 0.49 and the 0.80). These results show that the mortality differentials for men in the current NLMS data are larger than the differentials found in 1960. For women, the mortality differentials do not appear that much different from the earlier study. Simply put, this means that for men the mortality gap between those with high and low education as a percent of the level of mortality has widened. In conclusion, this summary of results from the NLMS shows that the CVD mortality differentials by education and income are very strong, that income and education each contribute independently to the differentials in mortality, that this finding is evident and consistent in men and women and in blacks, Hispanic whites and in non-Hispanic whites. There is further evidence that for men, the mortality gradient by education is stronger now than it was in 1960. REFERENCES 1. Sorlie PD, Backlund E, Keller JB: U.S. mortality by economic, demographic, and social characteristics: The National Longitudinal Mortality Study. Am J Public Health 1995;85:949-956. 2. Kitagawa EM, Hauser PM: Differential Mortality in the United States: A Study in Socioeconomic Epidemiology. Vital and Health Statistics monographs. American Public Health Association. Cambridge, MA: Harvard University Press, 1973. 3. Preston SH, Elo IT: Are educational differentials in adult mortality increasing in the United States? J Aging and Health 1995;7:476-496. ----------------------------------------------------------- Socioeconomic Factors in Ischemic Heart Disease Morbidity and Mortality Jacob J. Feldman, Ph.D. Diane M. Makuc, Dr.P.H. This report describes data on educational differentials in morbidity from the first National Health and Nutrition Examination Survey (NHANES I) Epidemiologic Followup Study (NHEFS). Point prevalence estimates are the only morbidity data appearing in the pre-conference Chartbook on U.S. National Data on Socioeconomic Status and Cardiovascular Health and Disease, but incidence and case fatality are more informative measures of morbidity. Prevalence and death rates can be viewed as a function of incidence and survival (1,2). In designing interventions to lessen the disease burden among the disadvantaged, we need to know at what stages in the process disparities develop. NHEFS is the only nationally representative database that provides information on socioeconomic status (SES) differentials in heart disease incidence and survival for the United States. The study is being conducted by the National Center for Health Statistics (NCHS) with funding from some 12 other Federal agencies, primarily components of the National Institutes of Health. The basic design of this longitudinal cohort study is depicted in Figure 1 and in various NCHS publications (3,4). ----------------------------------------------------------- Figure Missing Figure 1. - Basic Design of the NHEFS. ----------------------------------------------------------- Table 1. - Definition of Incident IHD Cases * No history of heart disease at baseline * Never had a heart attack or heart failure * No heart medication during past 6 months but during followup period either: * Death certificate with any mention of IHD (ICD-9 codes 410-414) or * Hospital record with an IHD diagnosis [end Table 1] ----------------------------------------------------------- The analysis is based on observations for the entire period from 1971 through 1993. As this is the first analysis that includes the only recently collected data for the 1987-1993 period, the results are preliminary. More intensive analysis of these data may be expected to result in the identification of some data errors and, therefore, a moderate number of edits. However, it is certain that the basic findings will be sustained. There have been more than 150 publications and presentations based on the earlier waves of NHEFS; many of them have been concerned with cardiovascular disease (CVD) and have treated SES as a covariate in analyses focusing on other risk factors (5). The SES differentials themselves were the major focus of two earlier papers of particular relevance. Trends over the past several decades in the differentials were examined in a paper by Feldman, Makuc, Kleinman, and Cornoni-Huntley (6). (Excerpts from that paper were reprinted in the 1994 Report of the NHLBI Task Force on Research in Epidemiology and Prevention of Cardiovascular Diseases.) For older white males, earlier research had shown no relationship between educational attainment and heart disease mortality in 1960 (7). On the other hand, the NHEFS observations covering the period from 1971 to 1984 showed a steep heart disease mortality gradient in relation to educational attainment. The decline in heart disease death rates in the 1970s and 1980s appeared, for males, to have occurred only among the more educated. The rates for the less educated older males changed very little over that period. It was also found that, among older white females, the more highly educated were already experiencing lower heart disease mortality rates in 1960 than the less educated. This gradient did not appear to grow steeper over time. Cox proportional hazards regression models for age- gender strata of the NHEFS data - assessing the educational attainment-heart disease mortality relationship controlling for smoking, overweight, hypertension, and cholesterol - were also presented. Educational attainment remained a highly significant risk factor, the estimate of relative risk being reduced rather little in the full model. The decrease in the risk of heart disease mortality with increased education did not appear to be the result of the covariates. A number of alternative explanations of the trend over time in the education-heart disease relationship were discussed. In a second paper, Makuc, Feldman, and Gillum separated into incidence and case fatality components the educational differentials in mortality from ischemic heart disease (IHD) and total heart disease, again based on observations from the 1971-1984 period of NHEFS (8). Among both men and women, the less educated experienced higher incidence rates for IHD and total heart disease than did more educated individuals. In Cox proportional hazards regression models, a relatively strong SES differential remained after controlling for smoking, body mass index, hypertension, cholesterol, and diabetes. The less educated also experienced appreciably shorter survival times after onset than did the more educated. Certain of the confidence intervals were rather broad because of the relatively small number of incident cases in some age-gender strata. Extending the database to include an additional 10 years of heart-related events facilitates analyses of survival differentials. ----------------------------------------------------------- Figure Missing Figure 2. - Risk Ratios for IHD Incidence by Education for White Men and Women Ages 45 to 64 and 65 to 74 at Baseline in NHEFS, 1971-1993. High school graduation or above is used as the reference group. Age at baseline, in single years of age, has been entered as a covariate in these proportional hazard models. The vertical lines indicate 95 percent confidence intervals surrounding the estimates of relative risk. ----------------------------------------------------------- Incidence, NHEFS, 1971-1993 We require either hospital record or death certificate evidence in ascertaining cases of IHD (Table 1). There are some problems of underascertainment and of dating of the onset due to failure or delay in diagnosis, delay in hospitalization, reporting errors, and the inability to gain access to certain hospital records. The likely extent of such errors, and their potential impact on inferences based on this database, will be considered in a later publication. One might assume, however, that the less educated, because of inferior access to care, would be likely to be diagnosed and hospitalized later in the course of illness, a phenomenon that would underestimate their relative risk. Figure 2 presents estimates of the relative risks of IHD incidence for white males and females in two age groups according to educational attainment, with high school graduation or above as the reference group. Without exception, estimates indicate that those who had not graduated from high school were more likely to have developed IHD by any given age than those who had graduated from high school. The risk ratios for the less educated remain high at older ages. The estimates presented in Figure 2 are derived from a model parametrized in terms of age at baseline, but we have also performed analyses based on age at the time of risk that confirm the persistence of the SES gradient in incidence at quite advanced ages. These findings are particularly noteworthy because it has been previously observed that the impacts of many CVD risk factors tend to diminish with increasing age (9). Notwithstanding the fact that the SES differentiation in atherosclerosis and other preclinical cardiovascular pathology probably begins decades earlier in the lifespan, the incidence differentials remain large at advanced ages. It is conceivable that the relative weights of the various etiological factors responsible for the incidence gradient change over the lifespan. ----------------------------------------------------------- Figure Missing Figure 3. - Risk Ratios for IHD Deaths among IHD Cases by Education for White Men and Women in NHEFS, 1971-1993. Age at onset and gender are stratifying variables; a separate proportional hazards model was estimated for each of the six groups. Age at onset, in single years of age, was also entered into the model as a covariate. ----------------------------------------------------------- Figure Missing Figure 4. - Risk Ratios for IHD Deaths by Education for White Men and Women, Ages 45 to 64 and 65 to 74 at Baseline, in NHEFS, 1971-1993. Age at baseline, in single years of age, has been entered as a covariate in these proportional hazards models. ----------------------------------------------------------- Case Fatality We define survival time as the interval between the date of onset and the date of death with IHD coded as the underlying cause. Deaths from causes other than IHD were treated as observations censored at the date of death. Differentials in survival time for IHD are shown in Figure 3. The more educated had longer survival (in terms of a trajectory of lower ischemic heart disease mortality rates) after onset in five of the six age-gender strata. The survival advantage of the more educated appears, in the present dataset, to be greatest at the youngest age at onset but the trend is not statistically significant. It is not possible to determine from the present dataset the extent to which the differentials in survival primarily reflect differences in disease severity or whether the outcome differences reflect other factors related to SES. Mortality and Prevalence Proportional hazards models for IHD mortality, following the same structure as the incidence models in Figure 2, have also been estimated for the 1971-1993 database (Fig. 4). Substantial differentials by educational attainment were found in all four age-gender strata. Both incidence rates and the trajectory of the mortality hazard from IHD after disease onset were considerably less favorable among the less educated. These two tendencies compound to produce the large observed age-specific mortality differentials. The prevalence of IHD is a function of incidence and survival time (1,2). The educational differentials in prevalence tend to be smaller than for incidence and mortality, as the longer survival time of the more educated increases the number of IHD cases alive at any given point in time. (Survival models shown in Figure 3 predict death from IHD while prevalence is a function of the all-cause mortality hazard subsequent to onset. The estimates of the SES risk ratios for all-cause models are consistent with those for IHD.) Estimates of the average point prevalence of IHD cases among white males aged 65 or older for NHEFS during the 1992-1993 period exhibit practically no gradient at all for educational attainment. Averaging over the 6- year period 1988-1993, the National Health Interview Survey (NHIS) shows a positive self-reported IHD point prevalence gradient with education for non-Hispanic white males aged 65 or older (i.e., the more highly educated have somewhat higher point prevalence rates than the less educated). On the other hand, in younger age groups, the more highly educated report lower point prevalence than the less educated. The Medicare Current Beneficiary Survey (MCBS) shows a flat relationship between educational attainment and self-reported IHD point prevalence for white males aged 65 or older (10). The NHANES III for 1988-1991 shows the highest point prevalence of cases with either self-reported myocardial infarction or angina (Rose Chest Pain Questionnaire) among the more highly educated for white men aged 65 or older. As in the NHIS, there is a higher-order interaction between age, educational attainment, and point prevalence; in the age groups below 65 years, the more highly educated have the lowest point prevalence rates. On the basis of the incidence and survival differentials, one can deduce that heart disease patients among the more educated white males age 65 or older have, on average, been living with diagnosed IHD for a longer period of time than those with less education. While the mean survival times differ, it is uncertain how similar the shapes of the distributions of durations since onset are for the populations with differing educational attainment. It could be informative to investigate the shapes of those distributions. Analysis of the NHEFS data covering the 1971- 1984 period for the white population indicated a higher rate of heart-related surgery among the more highly educated (11). Analyses of 1993 Medicare data based on SES characteristics of the zip code of residence suggest a more or less flat relationship between the rates for various atherosclerosis-related surgical procedures and SES for white beneficiaries (12). To the extent that the size of the pool of prevalent cases of IHD influences the surgical procedure rate, the relatively small SES differentials in surgical rates for the older white population are consistent with the weak SES gradient in point prevalence. Further Research This is the first presentation based on the NHEFS database extended to 1993. A number of more detailed analyses are under way. Earlier analyses focused primarily on the white population because of sample size considerations. An analysis performed on the 1971-1987 data showed that educational differentials in mortality were similar for black and white males (11). With the heart disease events occurring in the 1987-1993 period added to the database, more definitive investigations of educational differentials in heart disease among blacks have become feasible and will be conducted. In the analyses of the NHEFS database thus far, the events of the entire period from 1971 to 1993 are aggregated. Although great changes have taken place since 1960, we have not examined individual-level data for changes in differentials in more recent decades. The 1979-1989 National Longitudinal Mortality Study data presented by Dr. Sorlie also have been treated as deriving from a single time period (13). As the SES gradient shifted so markedly between 1960 and the early 1970s, it seems probable that the gradient has continued to shift since then in one direction or the other. Modeling by specifying time period stratification and the use of time as a covariate may help clarify this question. Medicare Part A benefit records have, since 1984, been linked to the NHEFS database. These will provide the basis for the methodological research regarding the completeness of ascertainment and problems with the accuracy of the dating of onset. In addition, these linked benefit records facilitate research into SES differentials in the cost of services related to heart disease. In examining various risk factors as covariates in our proportional hazards models, we have relied exclusively on measurements from the 1971-1975 baseline examination. Data on potential risk factors were collected in the 1982-1984, 1986, and 1987 waves. The inclusion of measures from these waves in our models may well elucidate the mechanisms through which the SES differentials operate. In addition, following, through successive waves, cases with onset early in the study period should increase our understanding of SES differences in the natural history of IHD. While primary prevention may be the ideal, intervention at later stages in the disease process may also be effective in reducing SES differentials. Conclusions Our analysis of the NHEFS database, extended to 1993, has shown a strong SES gradient for the incidence of IHD in the white population for both males and females extending throughout the age span. Analyses of data from the first decade of the study were not able to account for the SES differentials in terms of the risk factors for which data were available. We also found a strong SES gradient for case fatality in the white population for both males and females, particularly at the younger ages. On the other hand, there does not currently appear to be a marked SES gradient in the prevalence of IHD among older white males. This presentation was the first based on the extended NHEFS database. More detailed analysis of these data holds promise of elucidating some of the mechanisms underlying the SES gradients. References 1. Seigel DG, Krueger DE: Consistency of incidence, survival, and prevalence data on myocardial infarction from various studies. J Chron Dis 1967;20:603-614. 2. Freeman J, Hutchison GB: Prevalence, incidence and duration. Am J Epidemiol 1980;112:707-23. 3. Cohen BB, Barbano BE, Cox CS, et al: Plan and operation of the NHANES I Epidemiologic Followup Study: 1982-84. Vital Health Stat 1987;1(22). 4. Cox CS, Rothwell ST, NWans JK, et al: Plan and operation of the NHANES I Epidemiologic Followup Study, 1987. Vital Health Stat 1992;1(27). 5. Leaverton PE, Havlik RJ, Ingster-Moore LK, et al: Coronary heart disease and hypertension. In Health Status and Well-Being of the Elderly: National Health and Nutrition Examination Survey-I Epidemiologic Followup Study, edited by JC Cornoni-Huntley, RR Huntley, JJ Feldman. New York: Oxford University Press, 1990, pp. 53-70. 6. Feldman JJ, Makuc DM, Kleinman JC, Cornoni-Huntley JC: National trends in educational differentials in mortality. Am J Epidemiol 1989;129:919-933. 7. Kitagawa EK, Hauser PM: Differential Mortality in the United States: A Study in Socioeconomic Epidemiology. Cambridge, Massachusetts: Harvard University Press, 1973. 8. Makuc DM, Feldman JJ, Gillum RF: Educational differentials in heart disease incidence and case fatality in a national sample. Presentation at the Annual Conference of the Society for Epidemiological Research, 1988. 9. Kannel WB, Gordon T: Cardiovascular risk factors in the aged: The Framingham Study. In Second Conference on the Epidemiology of Aging, edited by SG Haynes, M Feinleib. NIH Publication No. 80-969, 1980, pp.65-89. 10. National Heart, Lung, and Blood Institute: Chartbook of U.S. National Data on Socioeconomic Status and Cardiovascular Health and Disease, 1995. 11. Makuc DM, Feldman JJ: Socioeconomic differentials in all-cause and heart disease mortality; individual level data. Presentation at the Annual Conference of the American Public Health Association, 1989. 12. Health Care Financing Administration: Monitoring the Impact of Medicare Physician Payment Reform on Utilization and Access. HCFA Publication No. 03379, 1995. 13. Sorlie P: Presentation at this meeting. ----------------------------------------------------------- Socioeconomic Status and Biomedical, Lifestyle, and Psychosocial Risk Factors for CVD: Selected U.S. National Data and Trends Clifford L. Johnson, M.S.P.H. Christopher T. Sempos, Ph.D. During the past 30 years, there has been a marked decline in U.S. death rates from cardiovascular disease (CVD), including coronary heart disease (CHD) and stroke. Although every segment of the population has benefited from this decline, not all population subgroups have benefited to the same extent. This observation is almost certainly a result of the differential effect of the numerous biomedical, lifestyle, and psychosocial risk factors on the prevalence of CVD. The relationship between socioeconomic status (SES) and these risk factors is important. As noted in the National Heart, Lung, and Blood Institute (NHLBI) Report of the Task Force on Research in Epidemiology and Prevention of Cardiovascular Diseases, "Membership in less affluent and less educated groups is associated with higher levels of cigarette smoking, high blood pressure, obesity, and physical inactivity. These groups also tend to consume more adverse diets and to lack access to health care services, particularly prevention services"(1). The major risk factors for CVD are well known and include: * Abnormal blood lipids and lipoproteins * Elevated blood pressure * Smoking * Obesity * Physical inactivity * Diabetes mellitus * Adverse dietary patterns * Psychosocial factors While there are other potential CVD risk factors, both old and emerging, the list describes those risk factors with the largest potential for reducing or preventing CVD with improved status in the general population. The focus of this analysis is a review of national data and national trends using two surveys conducted by the National Center for Health Statistics (NCHS) of the Centers for Disease Control and Prevention. The National Health and Nutrition Examination Survey (NHANES) is a complex, multistage, national probability sample of the civilian noninstitutionalized population of the United States. Conducted periodically by NCHS and unique in that it includes both an interview component and a standardized physical examination as part of the survey (2), each NHANES produces information on many of the risk factors for CVD in its examination component. The National Health Interview Survey (NHIS), an interview survey conducted annually by NCHS (3), obtains data on the incidence of acute illness and injuries, the prevalence of chronic conditions and impairments, the extent of disability, and utilization of health care services. Both surveys have the capability to assess the relationship between SES and various CVD risk factors at a given point in time, as well as to assess trends in various CVD risk factors over time. The ability to assess trends is possible since many of the questions or examination methods have remained comparable over time. The findings are primarily based on information found in the Chartbook of U.S. National Data on Socioeconomic Status and Cardiovascular Health and Disease (4). Although the Chartbook contains information from a number of surveys and surveillance systems, only findings from the NHIS and NHANES surveys are used as examples in this analysis. Findings for the other studies shown in the Chartbook were generally similar to those shown for the NHIS or NHANES. A large number of biomedical, lifestyle, and psychosocial risk factor findings are presented in the Chartbook. All of these variables have been implicated as risk factors for CVD based on previous studies, and many were discussed extensively in the NHLBI Report of the Task Force on Research in Epidemiology and Prevention of Cardiovascular Disease. It was logical to believe that similar conclusions would be found when more recent data from NHANES III and the 1993 NHIS were analyzed. However, this was not always the case. Table 1 categorizes certain CVD risk factors into two groups based on an overall assessment of their relationship to the three SES variables used in the Chartbook analyses (i.e. education, income, and occupation). These SES variables were not related in any consistent manner to at least half of the previously defined risk factors in these two national surveys. Although the categorization into two groups is possibly an oversimplification, the fact remains that the relationship between many of these risk factors and SES is uncertain. One of the CVD risk factors that did show a strong relationship with SES was smoking. Higher levels of cigarette smoking were associated with persons who are less educated and less affluent. Figure 1 shows the prevalence of cigarette smoking among men and women ages 45 to 64 years by education, sex, and race/ethnicity group from the 1993 NHIS. For all six race/ethnicity-gender groups, the highest prevalence of smoking was observed in the group with less than high school education. This pattern was similar to that observed in the 25 to 44 and 65+ years age groups (not shown). Similar results were found when income and occupation were used as the surrogate for SES. The lowest income group had the highest prevalance of cigarette use, and the blue collar and service groups had the highest smoking prevalences among the occupation categories. However, the relationship between income, education, or occupation and cigarette usage was different for the Hispanic subgroup. Whether this inconsistent pattern for the Hispanic population resulted from smaller sample sizes that did not allow for consistent patterns to be observed, an inadequate definition of SES for this population subgroup, or a truly different relationship within this population group remains unresolved. With respect to trends over time, the overall conclusion is that for most sex-race/ethnicity-SES subgroups there has been a decrease in the prevalence of cigarette smoking. For example, the prevalence of smoking among 25 to 44 year old black and white males decreased from 1976 to 1993 in all educational subgroups (Fig. 2). However, the decrease is less obvious in some of the lowest education groups. In contrast, among 25 to 44 year old black and white women, the percent who smoked cigarettes actually stayed the same or increased in the lowest educational subgroup. Thus, while a generally decreasing pattern in cigarette smoking was observed when comparing NHIS data for 1976 and 1993, the decrease was much more significant in the highest SES subgroups. Similar patterns were observed when income was used as the SES variable. This finding is likely to be related to the increasing gap in CVD death rates between various race/ethnicity groups. ----------------------------------------------------------- Table 1. - Strength of Association Between CVD Risk Factors and SES Measures Weak: Blood pressure; cholesterol; diet; psychosocial factors Strong: Overweight; cigarette smoking; diabetes; physical activity [end Table 1] ----------------------------------------------------------- Figure Missing Figure 1. - Current Cigarette Smoking by Race/Ethnicity and Education: Men and Women Ages 45 to 64. Source: NHIS, 1993. ----------------------------------------------------------- Figure Missing Figure 2. - Trends in Cigarette Smoking by Education and Race/Ethnicity: Men and Women Ages 45 to 64. NHB = non- Hispanic black; NHW = non-Hispanic white. Source: NHIS, 1976, 1993. ----------------------------------------------------------- The relationship between the SES variables used in the Chartbook and many other CVD risk factors was much more complicated than that observed for cigarette smoking and was affected by the way the analyses were conducted. The observed relationship between overweight and income in the NHANES III (1988-91) data is such an example. For these analyses, overweight was defined as a value greater than the age-sex specific 85th percentile cut points for body mass index for 20 to 29 year old males and females from NHANES II (5). The table in the Chartbook used data from the first phase of NHANES III, and family income was classified into four different groups. In general, differences were observed in the prevalence of overweight by income status. However, the pattern was not consistent among men and women and among the various racial/ethnic groups. Moreover, for many subgroups the survey sample sizes were too small to make meaningful comparisons. Figure 3 presents similar analyses using another measure of income status and fewer categories. The alternative definition of income status is the poverty income ratio (PIR). PIR is a standardized variable calculated from annual census tables that takes into account family size, family composition, and total family income. It also allows comparison of findings from various surveys over time, which would not be possible with income data alone. For this alternative analysis, the sample sizes were adequate to divide PIR into three categories. The lowest category (less than or equal to 1.30) uses a cut point that determines eligibility for selected government sponsored food assistance programs. The upper category is a calculated PIR greater than 3.50, and the middle category is between these two cut points. The relationship between income status and overweight is different for men and women. In addition, the relationship differs between race/ethnicity groups. For non-Hispanic white men, the prevalence of overweight is lower in the upper income group, whereas, for both non-Hispanic black men and Mexican-American men the opposite is true: the prevalence of overweight is lowest in the poorest income group (Fig. 3). For non-Hispanic white women, the pattern is similar to that for white men, that is, the lowest prevalence is found in the upper income category. While this pattern is also observed in Mexican-American women, non-Hispanic black women show very little change in the prevalence of overweight regardless of income status. Thus, the "effect" of low income on the prevalence of overweight is not consistent among the major U.S. population groups, and the "effect" is not the same in men as in women. This leads one to conclude that the relationship of SES to some of these CVD risk factors is complex and not easily definable. Many more examples could be shown using other biomedical, lifestyle, or psychosocial risk factor data found in the Chartbook. The relationship between these risk factors, SES, and health outcomes such as cardiovascular health is complex and reflects the complex manner in which social and economic status and race interact in the United States. This topic has been discussed recently in a number of articles. Dr. Gregory Pappas noted in an editorial in the American Journal of Public Health that "the complex ways in which social and economic class and race create disadvantages and produce disparities in health must be more fully investigated" (6). Based on these findings, more investigation is needed. It is extremely difficult to summarize all the findings on SES and CVD. It is possibly due to the lack of a consistent and agreed-upon definition of SES or "social class." Although "class" is a difficult and sometimes contentious term, Dr. Pappas contended that some measure of it is necessary in order to examine disparities in health among various racial and ethnic populations. Agreement on a standard measure of "class" or SES is critical to the process as well. ----------------------------------------------------------- Figure Missing Figure 3. - Age-adjusted Prevalence of Overweight by Poverty Status and Race/Ethnicity: Men and Women Ages 20 and Over. Source: NHANES III, 1988-91. ----------------------------------------------------------- Similar issues and conclusions were also noted in a Public Health Reports article by Moss and Krieger (7). In an introduction, Dr. Phillip R. Lee, the Assistant Secretary for Health, acknowledged the limitations of SES measures (especially occupation) in many of the Federal, state, and local health data and indicated that many of the recommendations from that conference would be implemented in the coming months and years. With the currently available national survey data and with agreed-upon indicators of "social class" or SES, some evaluation of disparities in health among population subgroups is possible. The recently completed second phase of NHANES III will be available in the near future. This will allow analyses of the full 6-year study with increased sample sizes that will clearly resolve some of the limitations that were encountered by the Chartbook developers when creating tables based on the first phase of data from that survey. In addition, NHANES III has a number of SES variables (including occupation) collected in a standardized manner, which will allow numerous analyses of SES and cardiovascular health based on the wealth of information collected in that survey. Many of these variables were consistently collected across previous NHANES surveys and will provide the basis for assessing trends over time. These data, in conjunction with information from the NHIS and other non-national or subpopulation-specific national data sets, will greatly increase our knowledge of the relationship between SES (or social class) and CVD. We have the beginnings of that information data base today in the Chartbook. The rest remains for us to analyze, interpret, and implement in the future. References 1. National Heart, Lung, and Blood Institute: Report of the Task Force on Research in Epidemiology and Prevention of Cardiovascular Diseases, 1994. 2. National Center for Health Statistics: Plan and operation of the Third National Health and Nutrition Examination Survey, 1988-1994. Vital Health Stat 1994;1(32). 3. National Center for Health Statistics: Current estimates from the National Health Interview Survey, 1993. Vital Health Stat 1995; 10(190). 4. National Heart, Lung, and Blood Institute: Chartbook on U.S. National Data on Socioeconomic Status and Cardiovascular Health and Disease, 1995. 5. Kuczmarski RJ, Flegal KM, Campbell SM, Johnson CL: Increasing prevalence of overweight among US adults: the National Health and Nutrition Examination Surveys, 1960 to 1991. JAMA 1994;272:205-211. 6. Pappas G: Elucidating the relationships between race, socioeconomic status, and health. Am J Public Health 1994;84:892-893. 7. Moss N, Krieger N: Measuring social inequalities in health: report on the conference of the National Institutes of Health. Public Health Reports 1995;110:302-305. ----------------------------------------------------------- SES and Medical Care Utilization Related to CVD: Selected U.S. National Data and Trends Paul Eggers, Ph.D. Although there is a great deal of research on the effects of socioeconomic status (SES) on global utilization rates, such as physician visits and hospitalizations, there is very little on cardiovascular disease (CVD) specific use rates. This is due to at least two reasons. The first is sample size. Surveys are usually designed to obtain health care use and expenditure data on overall medical services. Rarely are diagnosis-specific use rates sufficiently sampled for analysis, even for large categories such as CVD. Second, many surveys rely on patient recall for the identification of utilization. Nonmedically trained respondents are very poor in identifying cause-specific utilization. They can usually recall important medical events, i.e., hospitalization, but the underlying diagnosis related to the event is usually poorly understood or remembered. A good review of the literature on race and CVD utilization was recently published by Earl Ford and Richard Cooper (1). Oberman and Cutter found that blacks who received coronary arteriography were only 40 percent as likely to receive a coronary artery bypass graft (CABG) as were whites (2). Maynard, et al, found that the use of CABG was 30 percent lower for black males and 10 percent lower for black females than for their white counterparts (3). Gillum used the National Hospital Discharge Survey (NHDS) to show that blacks received angiography only one-half as frequently as did whites and were 70 percent less likely to receive a CABG (4). Ford, et al, also examined NHDS data, but controlled for the rate of CVD by using discharge rates for myocardial infarction (MI) (5). They estimated a two-fold racial difference in CABG use, in contrast to the three-fold difference shown by Gillum. Wenneker and Epstein examined Massachusetts hospitalizations for persons admitted with CVD or chest pain (6). Blacks had 20 to 30 percent fewer angiographies and about one-half the number of revascularizations. This analysis included a control for estimated income. Maynard, using data from Seattle, showed that, although blacks were equal to whites in use of angiography and thrombolytic therapy, their rates of revascularization were still only about one-half those of whites (7). Hannon, et al, attempted to control for potential hospital technology bias by limiting their analysis to hospitals that actually provide these cardiovascular procedures (8). Still, they found similar racial differences. Two recent studies have used the Medicare data base to look at the race issue. Goldberg, et al, found that, in 1986, aged black males were only 20 percent as likely as aged white males to receive a CABG (9). For black women, the relative use rate was 40 percent. Goldberg did not attempt to control for the rate of catheterization. Franks, et al, also using Medicare data, did examine catheterization rates (10). They found aged black males hospitalized with an MI to be 50 percent less likely than aged white males to receive a catheterization. For females the relative use rate was 70 percent. Contingent upon catheterization, both black males and females were only 60 percent as likely as their white counterparts to undergo any revascularization (either percutaneous transluminal coronary angioplasty [PTCA] or CABG). Ford and Cooper reached several conclusions. They include: * The question of whether race or SES accounts for many of the observed differences in health and health care utilization is controversial and remains unanswered. * The demonstration of racial differences in the Medicare data, where one could argue that economic factors should constitute less of a barrier, shows that factors other than economic ones play an important role. * During the 1980s, after arteriography, PTCA, and CABG were routinely available, blacks were far less likely to undergo these procedures. * A surveillance system, preferably one that is nationally based, should be established to monitor various aspects of health care utilization, including the use of cardiac procedures. Very few of the national studies on race and cardiovascular utilization directly address the issue of SES. The remaining part of this paper presents data from two national data bases with SES variables included. The National Medical Expenditure Survey (NMES) The NMES was a set of surveys that collected information about use of and expenses for health services for the year 1987. This survey included about 14,000 households representing the civilian, noninstitutionalized population of the United States. Data were collected through three personal surveys and a single telephone survey. Respondents were asked to record medical events in a calendar or diary. Physician visits were identified from "provider probes" for visits to medical doctors, hospital emergency rooms, and hospital outpatient departments. Hospital admissions were similarly identified by a probe for either an overnight stay or an admission and discharge the same day. Demographic information was obtained from the primary respondent from each family. Figures 1 and 2 depict NMES-determined physician visit and hospitalization rates, respectively. For whites, there is a fairly clear gradient for use by education level. Persons with more than high school education see a physician for a cardiovascular problem less than one-half as often as those with less than a high school education (264 and 574 visits, respectively). The effect of education on use rates among blacks is less dramatic. However, both high school graduates and those with more education have about one-third fewer visits than those with less. The gradient is more consistent for both racial groups with respect to income. For both groups, the highest income category has less than one-half the use rate of the lowest income group, with intermediate levels for the other income groups. Hospitalizations are shown in Figure 2 for all conditions, all cardiovascular conditions, and for ischemic heart disease by educational groupings. The first thing that is evident is the very low rate of cardiovascular hospitalization - thus, the rates have fairly high errors of estimation. Nevertheless, educational gradients seem evident. For blacks, those with less than high school education have approximately twice the overall hospitalization and up to 5 times the cardiovascular hospitalization of those with high education levels. Among whites the gradient is not so pronounced. However, for all three measures of hospitalization, decreases are associated with increased income levels. ----------------------------------------------------------- Figure missing Figure 1. - Physician Visits for Any Cardiovascular Condition by Education, Income, and Race. Rates shown are for persons ages 25 to 64, both sexes. Source: Adapted from 1987 NMES. ----------------------------------------------------------- Figure Missing Figure 2. - Hospitalizations by Education and Race for Patients Ages 25 to 64. Hospitalization rates are shown for all conditions, all cardiovascular conditions, and ischemic heart disease (IHD) in both sexes by educational groupings. Source: Adapted from 1987 NMES. ----------------------------------------------------------- Medicare Link With Census Data In order to explore the issue of the effect of SES in the Medicare population, U.S. Census data were used as a proxy measure for SES, specifically income. Briefly, utilization rates were calculated for Medicare beneficiaries by the zip code of their residence. The median income for each zip code obtained from the 1990 U.S. Census was used to estimate household income for all aged persons living in a zip code. Zip codes were then aggregated into four income levels, and utilization rates were developed. The population was limited to aged Medicare beneficiaries who were not members of health maintenance organizations (HMOs). The counts were based on total months of non-HMO enrollment. These data were taken from the Medicare denominator file for 1993. For each zip code, person-year equivalents (total months of non-HMO enrollment divided by 12) were calculated for each of three age groups (65 to 74, 75 to 84, 85 and over) for males and females separately, and for white, black, and other race categories. Medicare Trends in Cardiovascular Utilization (1986 to 1993) The Health Care Financing Administration (HCFA) is the agency responsible for the Medicare program. Medicare covers almost the entire U.S. population over the age of 65 (about 31 million), as well as three million disabled persons and 200,000 persons with end-stage renal disease. About 90 percent of the Medicare population receives care in a fee-for-service environment. Each of the fee-for-service encounters generates a bill that can be linked to the individual beneficiary. The fact that Medicare is almost universally available to persons over age 65 means that population-based use rates can be calculated for fairly specific kinds of services. Medicare beneficiaries account for approximately 11 million hospitalizations annually. For each of these hospitalizations, information is available for both diagnoses and procedures in ICD9-CM format. As many as five diagnostic conditions and up to three procedures are reported. All procedures were considered but only the principal condition was examined. Data were arrayed for the years 1986 through 1993. About one-fifth of all hospitalizations among aged Medicare beneficiaries are for CVD. Although hospitalization rates for all heart disease do not differ greatly between black and white beneficiaries, as seen in Figure 3, black beneficiaries are much less likely to be hospitalized for ischemic heart disease (Fig. 4). ----------------------------------------------------------- Figure Missing Figure 3. - All Heart Disease Hospitalizations of Medicare Beneficiaries by Race. Between 1986 and 1993, the hospitalization rate for whites increased by 10 percent while the black rate increased by 23 percent, increasing the black-to-white use ratio from 0.92 to 1.03. Source: HCFA. ----------------------------------------------------------- Revascularization rates are depicted in Figures 5 and 6. Cardiac catheterization (not shown) increased rapidly for both white and black beneficiaries during these years. The white rate increased by 75 percent, from 9.5 procedures per 1,000 to 16.4 procedures per 1,000. The black rate more than doubled, from 5.0 procedures per 1,000 to 11.5 procedures per 1,000. The black-to-white use ratio increased from 0.52 in 1986 to 0.70 in 1993. CABG rates (Fig. 5) increased at comparable rates (to catheterization) during these years. As with the other use measures, the increase was greater for black beneficiaries than for white beneficiaries. In 1986, the CABG black-to-white use ratio was 0.28. By 1993 this had increased to 0.40. The largest increase in utilization during these years was for PTCA (Fig. 6). PTCA was performed only one-half as frequently as CABG in 1986 but by 1993 the PTCA rate exceeded the CABG rate for both black and white beneficiaries. The black-to- white differential was greater for both CABG and PTCA than for catheterization. This is consistent with previous studies showing lower revascularization rates for black than for white beneficiaries (11). ----------------------------------------------------------- Figure Missing Figure 4. - Ischemic Heart Disease Hospitalizations of Medicare Beneficiaries by Race. Although black Medicare beneficiaries are much less likely than whites to be hospitalized for IHD, their hospitalization rates increased 14 percent between 1986 and 1993, whereas those for whites increased only 6 percent. The result was an increase in the relative black-to-white use ratio from 0.69 to 0.74. Source: HCFA. ----------------------------------------------------------- Figure Missing Figure 5. - CABG Hospitalizations of Medicare Beneficiaries by Race. Between 1986 and 1993, hospitalizations for coronary artery bypass graft (CABG) procedures rose 67 percent for white Medicare beneficiaries and 140 percent for blacks. The black-to-white use ratio increased from 0.28 to 0.40. Source: HCFA. ----------------------------------------------------------- Figure Missing Figure 6. - PTCA Hospitalizations of Medicare Beneficiaries by Race. The greatest increase in hospitalization of Medicare beneficiaries for revascularizations from 1986 to 1993 occurred among those undergoing percutaneous transluminal coronary angioplasty (PTCA). Utilization increased 315 percent for whites and 503 percent for blacks; the black-to-white use ratio increased from 0.32 to 0.46. Source: HCFA. ----------------------------------------------------------- Income and Racial Differences in Cardiovascular Utilization Figure 7 shows four measures of cardiovascular utilization by income groupings for both white and black Medicare beneficiaries. The income groups are based on median income levels of aged persons by zip code of residence. White beneficiaries are arrayed according to median white income and black beneficiaries are arrayed according to median black income. For white persons, there appears to be an income effect for ischemic heart disease, with rising hospitalization associated with decreasing income. The 37.7 hospitalizations per 1,000 persons in the lowest income quartile is 28 percent greater than the rate for persons living in higher income zip code areas. Little evidence exists of an income differential for cardiovascular procedures. To the extent that there is an income differential among black beneficiaries for ischemic heart disease, it runs in the opposite direction from that of white persons. The rate in the lowest income quartile (24.3 per 1,000) is 13 percent below the rate (28.0 per 1,000) in the highest income quartile. As with white patients, there are no consistent effects of income on the procedure rates. Summary Racial differentials between black and white persons in use of cardiovascular health care services have been well documented by a number of studies. Revascularization among black persons is about one-half that of white persons. At least part of this difference is due to lower rates of catheterization among black persons. However, even among those who receive a catheterization, subsequent revascularization is considerably lower among black persons. Most of the studies on these racial differences were performed on data from the late 1970s through the late 1980s and were cross-sectional in nature. There is virtually no evidence in the literature of potential trend effects. However, data from the Medicare program suggest that, at least among the elderly population, the racial disparity in use rates may be narrowing somewhat. ----------------------------------------------------------- Figure Missing Figure 7. - Cardiovascular Hospitalization by Race and Income for Patients Ages 65 and Above, 1993. Source: HCFA. ----------------------------------------------------------- The role of SES in racial differentials has not been delineated and will be difficult to measure, short of indirect aggregate methods such as linkages with census data. Initial data from the Medicare-Census data link show different income patterns. Among whites, decreasing income is associated with an apparent increased need (as measured by hospitalization for ischemic heart disease) for revascularization, but a steady, if not decreasing, use of revascularization. Among black persons, neither the measure of need nor level of revascularization seems to be related to income levels. REFERENCES 1. Ford ES, Cooper RS: Racial/ethnic differences in health care utilization of cardiovascular procedures: a review of the evidence. Health Services Res 1995;30(1):237-252. 2. Oberman A, Cutter G: Issues in the natural history and treatment of coronary heart disease in black populations: surgical treatment. Am Heart J 1984;108(3, part II):688-694. 3. Maynard C, Fisher LD, et al: Blacks in the Coronary Artery Surgery Study (CASS): race and clinical decision making. Am J Public Health 1986;76(12):1446-1448. 4. Gillum RF: Coronary artery bypass surgery and coronary angiography in the United States, 1979-1983. Am Heart J 1987;113: 1255-1260. 5. Ford ER, Cooper RS, et al: Coronary arteriography and coronary bypass surgery among whites and other racial groups relative to hospital based incidence rates for coronary artery disease: findings from NHDS. Am J Public Health 1989;79(4):437-440. 6. Wenneker MB, Epstein AM: Racial inequities in the use of procedures for patients with ischemic heart disease in Massachusetts. JAMA 1989;261(2):253-257. 7. Maynard C, Litwin PE, et al: Characteristics of black patients admitted to coronary care units in metropolitan Seattle: results from the Myocardial Infarction Triage and Intervention Registry (MITI). Am J Cardiol 1991;67(1):18-23. 8. Hannon EL, Kilburn H, et al: Interracial access to selected cardiac procedures for patients hospitalized with coronary artery disease in New York State. Medical Care 1991;29(5):430-441. 9. Goldberg KC, Hartz AJ, et al: Racial and community factors influencing coronary artery bypass graft surgery rates for all Medicare patients. JAMA 1992;267(1):1473-1477. 10. Franks AL, May DS, et al: Racial differences in the use of invasive coronary procedures after acute myocardial infarction in Medicare beneficiaries. Ethnicity and Disease 1993;3(3):213-220. 11. Ayanian AZ, Udvarhelyi IS, Gatsonis CA, et al: Racial differences in the use of revascularization procedures after coronary angioplasty. JAMA 1993;269(20):2642-2646. ----------------------------------------------------------- Differences Between Occupational Classes in Cardiovascular Disease Mortality: A Comparison of 11 European Countries Anton E. Kunst, M.A., Feikje Groenhof, M.A., Johan Mackenbach, M.D., Ph.D., and the European Union Working Group on Socioeconomic Inequalities in Health Socioeconomic status (SES) differences in health exist in Europe as well as in the United States (1-3). The number of studies on health differences in Europe strongly increased during the 1980s. There has been a parallel increase in the concern with SES differences in health by policy makers in some European countries. Much of the research on health differences in Europe has given special attention to SES differences in cardiovascular diseases (CVD), an interest that was stimulated by the observation that social class differences in CVD mortality in England and Wales have widened since World War II (4). Parallel trends were observed in more recent studies from several other European countries (5-7). This variability over time raises the question of whether a similar regional variability exists between countries. If large variations are observed for place as well as time, that would highlight the changeable nature of SES differences in CVD. In addition, a closer look at the different countries may reveal which circumstances are associated with smaller or larger inequalities in CVD. European countries differ widely in standard of living, socioeconomic policies, health care systems, and culture and history. This diversity has resulted in large differences in the overall level of mortality from CVD and in trends over time. For example, the epidemic rise and decline of ischemic heart disease (IHD) mortality occurred earlier and was more pronounced in northern Europe than in France and more southern countries (8). An obvious - but still unanswered - question is whether European countries also differ with respect to the SES distribution of mortality from CVD. In this paper, 11 countries from northern and southern Europe are compared for SES differences in CVD mortality in men about 45 to 59 years old. The data, from the 1980s, come from an international project sponsored by the European Union. Information on mortality by SES was collected centrally and reanalyzed using a standardized methodology. Most European countries have nationally representative data on CVD mortality by occupation, but surprisingly few (about five) have similar data with education or income as the SES indicator. Therefore, mortality differences between occupational classes are compared in these countries. Materials and Methods Nationally representative data were available on mortality by age, sex, occupation, and cause of death. For England/Wales, Finland, Sweden, Norway, Denmark, and Italy, cause-specific mortality data were available from national longitudinal studies with approximately 10 years of followup. For Ireland, France, Switzerland, Spain, and Portugal, cause-specific mortality data were obtained from so-called unlinked cross-sectional mortality studies. The study period was ca. 1981-90 for most longitudinal studies and ca. 1980-82 for the study from Italy and the cross- sectional studies. The data presented in this paper refer to men of working age. Men older than 60 years and all women were excluded due to problems with reliability and crossnational comparability of the measurement of occupational class. Data from different countries refer to the same age group in terms of age at death. For studies where men were classified according to their age at death, the age group 45 to 59 years was distinguished. For longitudinal studies in which birth cohorts were followed 10 years, we distinguished the birth cohort aged 40 to 54 years at the start of followup. Men were classified on the basis of occupation according to the Erikson-Goldthorpe (E-G) scheme (9). This is a social class scheme that is gaining increased acceptance in social sciences in Europe and elsewhere. A commonly used version of the E-G scheme distinguishes the following seven social classes (the percentage of the male working population of northern and western European countries is given in parentheses): * Employers, administrators, managers, and professionals (20 to 30 percent) * Routine nonmanual employees (5 to 10 percent) * Self-employed workers, except professionals and farmers (5 to 10 percent) * Foremen and skilled manual workers (20 to 30 percent) * Semiskilled and unskilled manual workers (20 to 25 percent) * Farmers (5 to 10 percent) * Farm laborers (0 to 5 percent) A more condensed version of the E-G scheme collapses these seven classes into three broad groups: all nonmanual and self-employed workers, all wage-earning manual workers, and farmers and farm laborers. The seven-class version of the E-G scheme could not be constructed for Denmark, Italy, Spain, or Portugal. However, the occupational classifications that were available for these countries could be used to approximate the broad groups of the three-class version. Ideally, economically inactive men are assigned to an occupational class on the basis of a previous occupation. In most data sources, however, information on previous occupations was lacking for most of the economically inactive men, and these men therefore had to be excluded from the analysis of the association between mortality and occupational class. As a result, however, the magnitude of mortality differences between occupational classes may be substantially underestimated (10). However, this problem can be remedied in part. We have developed adjustment factors for estimating the size of mortality differences among the entire male population from the mortality differences observed among men with known occupation only. These adjustment factors have performed well in a number of tests (11). The relative mortality level of men in specific occupational classes was measured by means of standardized mortality ratios (SMRs), using national age-specific mortality rates as the standard. Several summary indices were calculated to express the magnitude of mortality differences by occupational class (10). In this paper, we present an often-used summary index with a straightforward interpretation: the (age-standardized) mortality rate ratio (RR) that compares all manual workers to all nonmanual workers. Comparative research on inequalities in health is treacherous if no extensive attention is paid to potential data problems (10). Three major problems with reliability and comparability of data on mortality by occupation are exclusion of economically inactive men from most data sets, use of social class schemes other than the E-G scheme, and biases inherent to "unlinked" cross-sectional studies. Each of these problems has been evaluated extensively for its potential effect on the RR estimates for specific countries. Details are given in the final report of our project (11). Data sets were classified according to their level of comparability: * Data from England/Wales and Sweden were highly comparable. If manual vs. nonmanual RRs for these two countries differed by more than 15 percent, it is unlikely that systematic errors could explain that difference. These two countries were comparable on the basis of both the three-class and the seven-class versions of the E-G scheme. * Data for most other countries were reasonably comparable. If manual vs. nonmanual RRs for these countries differed by more than about 30 percent, it is unlikely that systematic errors could explain that difference. These countries were comparable only on the basis of the three-class E-G scheme. * Data for Ireland, Spain and Portugal were poorly comparable. Only tentative comparisons can be made on the basis of the three-class E-G scheme. Results Table 1 presents the results for mortality from all CVD. The SMRs for three broad occupational classes are given together with the RRs that represent the magnitude of the mortality difference between nonmanual and manual classes. Countries are ordered from high to low RRs. The SMRs for the nonmanual class were between 0.90 and 1.00 for most countries, but smaller for England/Wales, Finland, Norway, and Sweden. In most countries, the S