|Table of Contents|
Table of Contents
List of Tables
Chapter 1. Review of literature
Chapter 2. Method
Chapter 3. Results
Chapter 4. Discussion
List of references
THE RELATIONSHIP BETWEEN HEALTH RELATED BEHAVIORS AND HEALTH
STATUS AMONG MINORITY POPULATIONS
DELIA F. OLUFOKUNBI
A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 1997
I would like to dedicate this dissertation to my mother, Storm Somers. It is
through her love, guidance, and support over the years that I have found the strength and motivation to pursue my dreams. She is truly the "wind beneath my wings."
First and foremost, I would like to express my deepest gratitude to my chair and mentor, Dr. Suzanne B. Johnson. Her wisdom, guidance, and unwavering encouragement over the years have provided me with the skills and supportive environment to achieve my dream. Through her example, I have seen rewards of personal integrity, commitment, and respect and, as a result, have grown tremendously, both professionally and personally.
Second, I would like to thank Dr. Michael Miller for his invaluable guidance on this project. He provided me with the knowledge and resources necessary to complete this project and his wonderful personality has made even the rough times seem bearable. I have learned a great deal through my interactions with him and I am very grateful.
Third, I would like to thank all my committee members for their time and
assistance in the development and completion of this project. Their collaborative insight helped to make this study the best it could be and also led to some very lively and quite interesting meetings. I would also like to thank Dan Nissen for all his assistance in helping me sort through this database and confront my most feared enemy, SAS.
Finally, I would thank my family and friends who have seen me through the best and worst of times. Their love and unconditional support have kept me sane over the years.
TABLE OF CONTENTS
ACKN OW LEDGM EN TS ................................................................................................. iii
LIST OF TABLES ............................................................................................................. vi
ABSTRA CT ..................................................................................................................... viii
I REVIEW OF LITERATURE ........................................................................................ I
Introduction .................................................................................................................... I
M ethodological Issues .................................................................................................... 3
D efinition of M inority ............................................................................................... 3
Problem s with M easurem ent ..................................................................................... 5
M inority Health Status and Health Related Behaviors .................................................. 6
Access to Care/Utilization of Health Care Services ................................................. 8
Health Related Behaviors ........................................................................................ 10
Theoretical Concepts of M inority Health Status .......................................................... 18
Health Status, Socioeconomic Status, and Health Insurance:
An interdependent relationship .................................................................................... 20
Health insurance coverage ...................................................................................... 20
Socioeconom ic status .............................................................................................. 23
N ational M edical Expenditure Survey ......................................................................... 27
Policy Im plications ...................................................................................................... 30
Purpose of Research ..................................................................................................... 31
2 M ETHOD .................................................................................................................... 34
Data and Sam ple .......................................................................................................... 34
M easures ...................................................................................................................... 36
Dependent Variables ............................................................................................... 40
M issing Data ........................................................................................................... 41
3 RESULTS ................................................................................................................... 43
Descriptive Analysis ................................................................................................... 44
M ultivariate Analysis ................................................................................................... 54
Overall Health Rating ............................................................................................. 56
Role Functioning ..................................................................................................... 58
Physical Funtioning ................................................................................................. 60
Acute Sym ptom s scale ............................................................................................ 62
Chronic Sym ptom s scale ......................................................................................... 63
M edical Conditions scale ........................................................................................ 64
Supplem ental Analysis ................................................................................................. 72
4 DISCU SSION .............................................................................................................. 84
Relationship between Race/ethnicity and Health Related Behaviors .......................... 86
Relationship between Race/ethnicity and Insurance Coverage .................................... 87
Relationship between Race/ethnicity and Dependent Variables .................................. 89
Lim itations ................................................................................................................... 94
Future Research ............................................................................................................ 97
LIST OF REFEREN CES ................................................................................................... 99
BIOGRAPHICAL SKETCH ........................................................................................... 108
LIST OF TABLES
I Leading Causes of Death by Gender and Race ........................................... 11
2. Rate ratio of age-adjusted death rates from 15 leading causes
of death, by sex and race United States, 1992 .................................... 13
3. Percentage Distribution and Test of Significance of Demographic Variables,
Insurance Status, and Health Related Behaviors by race/ethnicity .............. 48
4. Percentage Distribution and Test of Significance of Outcome Variables
by Race/ethnicity ...................................................................... 49
5. Percentage Distribution and Test of Significance of Medical Conditions
Scales by Race/ethnicity ............................................................. 50
6. Con-elation Coefficients for Demographic Variables ................................... 51
7. Correlation Coefficients for Health Related Behavior Variables ..................... 52
8. Correlation Coefficients for Health Outcome Variables ............................... 53
9. Summary of Hierarchical Regression Analysis for Variables Predicting
O verall H ealth Rating ............................................................... 66
10. Summary of Hierarchical Regression Analysis for Variables Predicting
Role Functioning .................................................................... 67
11. Summary of Hierarchical Regression Analysis for Variables Predicting
Physical Functioning ................................................................ 68
12. Logistic Regression Models Predicting Acute Symptoms ............................ 69
13. Logistic Regression Models Predicting Chronic Symptoms .......................... 70
14. Logistic Regression Models Predicting Medical Conditions ......................... 71
15. Summary of Hierarchical Regression Analysis for Variables Predicting
Health Insurance Coverage .................................................... 77
16. Summary of Hierarchical Regression Analysis for Variables Predicting
B ody M ass Index ............................................................... 78
17. Summary of Hierarchical Regression Analysis for Variables Predicting
Sm oking Index .................................................................. 79
18. Logistic Regression Models Predicting Physical Exercise .......................... 80
19. Logistic Regression Models Predicting Blood Pressure Check ..................... 81
20. Logistic Regression Models Predicting Wearing Seat-belt .......................... 82
21. Summary of Variables .................................................................... 83
Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy
THE RELATIONSHIP BETWEEN HEALTH RELATED BEHAVIORS AND HEALTH STATUS AMONG MINORITY POPULATIONS By
Delia F. Olufokunbi
Chairman: Suzanne B. Johnson, Ph.D.
Major Department: Department of Clinical and Health Psychology
Research in the area of health care has systematically found increased rates of
morbidity and mortality among various minority groups compared the White population. Specific causal explanations for these differences have included genetic factors, differences in socioeconomic status, increased exposure to environmental hazards, and differences in insurance coverage. However, one area that has received relatively little attention is the influence of health related behaviors on the health status of minorities. Using data from the 1987 National Medical Expenditure Survey, this study examines the effect of health related behaviors and insurance coverage on health status across three racial/ethnic groups (Whites, Blacks, and Hispanics). Results found significant differences across racial/ethnic groups on health status measures. Specifically, on the outcome measures of Health Rating, Role Functioning, and Physical Functioning, Blacks reported the worst overall ftinctioning while Hispanics reported the best overall
functioning. In addition, Hispanics reported the least amount of acute symptoms, chronic symptoms, or medical conditions while Blacks and Whites were not different in their report of symptoms on these three scales. Inclusion of socio-demographic variables, health related behaviors, and insurance coverage to the predictive model resulted in Blacks actually report beuer role and physical functioning compared to Whites. In addition, Blacks report significantly less acute and chronic symptoms and medical conditions compared to Whites. Results support hypothesis that differences in health related behaviors between Whites and racial/ethnic minorities influence overall health status and point to the importance of considering preventive behavioral health care as a public health priority.
The status of health care in the United States has become one of the most daunting social issues in recent years. While the nation's health care system is the most expensive in the world, the comparative health care status of American citizens ranks lower than other western, industrial countries on a number of indicators (Nickens, 1995; Vandenbos, 1993). With the spiralling costs for health care and the approximately 36 million Americans who are uninsured, "health care reform" has become one of the most emotionally charged expressions within the social and political arena (Bingaman, Frank, & Billy, 1993; Kerrey & Hofschire, 1993). The twofold policy goal of health care reform is overall cost control with increased access to quality services. In addition, health care outcome has become an important issue in assessing the relationship between cost and quality. The exact mechanism by which to achieve these goals has been an area of significant controversy. In fact, one of the most vehemently debated social issues facing the U.S. Congress presently is the restructuring of the national Medicare and Medicaid public health insurance system.
With the heightened awareness of the current state of health care in the United
States, concerns regarding the status of minority health as a public health issue will need to be addressed. It is estimated that by the year 2050, approximately 50% of the United
States population will be of minority background. While there is great diversity across and within minority groups, the overall health status of minorities compared to the white population is poor (Manton, Patrick. & Johnson, 1987; Nickens. 1991. 1995; Williams, Lavizzo-Mourey, & Warren, 1994). Research evaluating differential health status, access to care, service utilization, and health perceptions across racial/ethnic groups have found large disparities (Manton, Patrick, & Johnson, 1987; Davis, Lillie-Blanton, Lyons, Mullan, Powe. & Rowland, 1987; Yergan, Flood, LoGerfo, & Diehr. 1987; Martin, Perez-Stable. Manin, Sabogal, & Otero-Sabogal, 1990; Strogatz, 1990; Cornelius. 1991, 1993; Bernard, 1993; Lillie-Blanton, Martinez. Taylor, & Robinson, 1993; Farraro, 1993; Ford & Cooper, 1995). The risk factors attributed to this differential health status include differences in lifestyle and health behavior (e.g.. alcohol, smoking, and nutrition), health consequences associated with low socioeconomic status (e.g.. economic barriers to access to health services, a lack of health insurance due to chronic unemployment, stress), inadequate knowledge of health practices, more hazardous occupations and exposure to environmental pollution, and genetic factors (e.g., sickle cell trait) (Manton, Patrick, & Johnson, 1987; Robinson, 1984). However, specific causal explanations for these discrepancies are difficult to differentiate given the high correlation between race and other socioeconomic variables (i.e, income, education, occupation, culture, environment, health insurance coverage) (Schulman. Rubenstein, Chesley, & Eisenberg).
In this era of rapid changes in the health care industry and increased attention to health outcome in relation to health care costs, behavioral and epidemiological research can inform policy decision-makers of the special health and medical needs of minority populations and the specific variables that may mediate these needs. With this broad goal
in mind. the present study sought to expand upon and further illuminate the relationship between race/ethnicity and health status. Specifically, given the interrelationship between race and certain socio-demographic variables as well as the potential influence of health insurance coverage and differential attitudes towards health care and health insurance on health status, this study sought to examine the strength of the relationship between race/ethnicity and health status after controlling for the influence of specific socio-demographic variables, health related behaviors, health insurance coverage, and attitudes towards health care and health insurance.
Definition of Minority
There has been substantial controversy over the use of racial/ethnic classifications and the meanings associated with these classifications. Membership in particular racial/ethnic groups assumes the experience of a common heritage and life experience (Lillie-Blanton, Martinez, Taylor, & Robinson, 1993). Although there is considerable heterogeneity within the American culture, there are four generally recognized minority groups in the United States: African Americans, Asian and Pacific Islanders, Latinos, and Native Americans (Nickens, 1991). There are also significant cultural and socioeconomical variations within these minority groups. For example, although Mexican Americans and Vietnamese Americans have low mean family incomes, Cuban Americans and Japanese Americans have comparably high mean family incomes (Nickens, 1995). In addition, although persons from the Caribbean Islands and Haiti may consider themselves African Americans for the purpose of completing survey
questionnaire data, there are significant cultural differences between these groups and other African Americans. As such, evaluations involving broad racial/ethnic categories can run the risk of improper generalization and create a distorted view of actual racial/ethnic status. Unfortunately, most large scale studies observing racial/ethnic differences have focused primarily on broad categories of classification and within group data are limited.
In addition, the rationale behind racial/ethnic categorizations may prove
problematic when conceptualizing broad social issues such as health care and methods for addressing these issues. Specifically, research on racial differences in health status has historically been dominated by a genetic model that defines race as reflecting biological homogeneity and differences in health outcome across groups as largely genetically determined (Williams, Lavizzo-Mourey. & Warren, 1994). On the other hand, racial variations across groups may actually indicate differential exposure to behavioral, psycho-social, material, and environmental risk factors that are more reflective of socioeconomic rather than genetic differences across groups (Williams, Lavizzo-Mourey, & Warren, 1994). Lillie-Blanton et al. (1993) argue for the use of racial/ethnic classifications as a measure of particular sociocultural experiences rather than a biological inheritance marker. Ultimately, despite concerns regarding risks of generalization, it is important to understand the underlying mechanisms behind racial/ethnic differences on measures of health status and health outcome so remedies can be developed to specifically address these mechanisms.
Problems with Measurement
Measurement is another important methodological issue, particularly when dealing with large survey data. Problems inherent in measuring health status may exaggerate or mask actual differences between groups of interest (Andersen. Mullner, & Cornelius, 1987).
Unreliable estimates may be obtained when the sample size is small (Andersen, Mullner, & Cornelius, 1987). Even in large surveys, sampling may result in a small minority representation. In an attempt to address this problem, the National Medical Expenditure Survey oversampled specific populations of interest, including African Americans, Hispanics. the poor, and the elderly.
Some have argued that self-report measures of health status offer the least
objective assessment of health status (Andersen, Mullner, & Cornelius, 1987). Refusal rates, lack of adequate recall, misinterpretation of questions, and provision of inaccurate but socially acceptable responses may contribute to systematic but extraneous differences among groups. For example, in a review of national databases, Anderson, Muller, and Cornelius (1987) found Blacks to exhibit the greatest health deficits based on objective measures (mortality rates, clinical examinations) but better health status on self-report measures of illness conditions, symptoms, and restricted activity days.
As such, while this evaluation will explore variation in health status across
minority groups, it will be important to keep in mind the limitations in generalization given the broad variations within racial/ethnic groups and the inherent limitations of selfreport data.
Minority Health Status and Health Related Behaviors
Differences in the leading causes of death for racial/ethnic groups provide
information about the differential risk factors experienced across these groups. These types of data have important policy implications in terms of developing preventative health programs and addressing the specific needs of particular groups. Table I presents data from the Centers for Disease Control and Prevention on the leading causes of death in the United States according to gender and race status for 1992. Table 2 illustrates the ratio of age-adjusted death rates from 15 leading causes of death by sex and race. In most cases, age-adjusted death rates were higher for African Americans compared to White Americans. The greatest differences were found in the rates of death by homicide (African American death rates were 6.5 times that of White Americans) and by HIV infection (African American death rate was 3.7 times that of White Americans). Chronic obstructive pulmonary diseases and allied conditions and suicide were the only two conditions for which the death rate for African Americans was lower than that for White Americans (CDC, 1994b).
Nickens (1995) reviewed mortality data from a number of national databases. He found an exceedingly high death rate among young Native Americans which could be attributed to intentional and unintentional injuries (i.e., injuries related to alcohol use and the direct health effects of alcohol use). Among Hispanics, death rates tended to be similar to Whites. Asian/Pacific Islanders had lower death rates than Whites for all age groups. Nickens (1995) also examined racial/ethnic group differences on the six major causes of death identified by the 1985 Health and Human Services Secretary's Task Force
on Black and Minority Health: cancer, cardiovascular disease, chemical dependency, diabetes, infant mortality, and homicide. African Americans were found to have higher death rates than White Americans for all of the six categories and HIV infection.
Schwartz, Kofie, Rivo, and Tuckson (1990) evaluated differential mortality rates for 12 separate diseases (tuberculosis, cervical cancer, Hodgkin's disease, rheumatic disease, hypertensive heart disease, acute respiratory disease, pneumonia, and bronchitis). Overall, they found a 4.5-fold excess mortality among African Americans compared to White Americans. African Americans had significant elevations in the mortality rates for eight of the 12 causes evaluated.
These findings were hypothesized to reflect either a higher incidence of disease in African Americans or a higher case fatality rate due to advanced stage of disease at time of diagnosis, co-morbidity. or delays in obtaining adequate treatment. Utilizing national data sources, Andersen, Mullner, & Cornelius (1987) compared the health status of African Americans and White Americans using measures of death, disease, disability, discomfort, and dissatisfaction. They found that for the most objective measure of health status (i.e. death) and the most sub active measure (i.e., dissatisfaction), African Americans had poorer health status. However, self-report of acute conditions tended to be higher for Whites than for African Americans for all age groups. Specifically, they found the death rate among Blacks greater for all of the most common causes of death: heart disease, cancer, stroke, accidents, and homicides. Less serious acute conditions (i.e., minor injuries, colds) were more often reported by Whites than Blacks. The authors caution that the use of self-report indices of health status may mask actual differences in health status in terms of mortality rates. Observing the rate of "excess deaths" (the
difference between the actual number of deaths in a minority population and the number of deaths that would have occurred if the mortality experiences of that group were the same as among the White population). the Secretary's Task Force on Black and Minority Health Report found 42.3 percent excess deaths among Afican Americans. 14 percent for the Spanish surnamed population of Texas, 2 percent among Cuban-born persons. 7.2 percent for those Mexican-born, 25 percent for American Indians (in Williams, LavizzoMourey, & Warren, 1994).
Nickens (1991) also reported Native Americans have high excess death rates (22%) and 87% of those excess deaths occurred before age 45. High levels of alcohol abuse, suicide, and unintentional injuries, and interpersonal violence are thought to be responsible. In 1992. the ratio of infant mortality among African Americans was 2.4 times that of Whites (National Center for Health Statistics. 1995). This race differential has remained fairly constant despite an overall pattern of decline (Manton. Patrick, & Johnson, 1987). Manton, Patrick, and Johnson (1987) point to socioeconomic differences between Afican American and White American mothers that increase risk factors such as poorer prenatal care, poor nutrition, and higher rates of teenage pregnancy, However, in a review of the literature, Nickens (1995) found differences in infant mortality to persist after controlling for the effects of social class, prenatal care, and living conditions.
Access to Care/Utilization of Health Care Services.
The extent to which individuals have access to primary health care and are able to utilize that care in an effective manner can have profound effects on overall health status
and mortality rates (Franks, Clancy. Gold, & Nutting, 1993; Hadley, Steinberg, & Feder, 1991; Patrick, Madden, Diehr, Martin, Cheadle, & Skillman, 1992, Weissman, Stem, Feilding, & Epstein, 1991). Access to and utilization of health care are often determined by various factors including whether an individual has a usual source of care, the type of source of that care, and the availability and convenience of the care (Cornelius, 1991).
Cornelius (199 1 ) examined the use of ambulatory and inpatient medical care by white and African Americans under the age of 65 who experienced an episode of illness. The study found African Americans more likely to be poor, uninsured, unemployed or disabled so as to prevent them from working, and in fair or poor health. In accounting for observed racial differences in the use of medical care, the analysis found health status, age, income, insurance coverage, and usual source of care to be more significant determinants of differences than race. The author notes that differences in access to care may reflect the fact that African Americans are more likely to fall into the groups that experience disparity in access to care (i.e., low-income, uninsured, no usual source of care).
Enactment of Medicaid and Medicare in 1965 paved the way for increased access to health care services for low-income, elderly, and ethnic minorities in the United States. However, universal insurance coverage and greater access may not eliminate the racial disparities in health status. Miller and Curtis (1993) point to the Medicare program as an example of increased access not necessarily correlating with increased utilization. They cite a 1992 report to Congress in which the Physician Payment Review Commission demonstrated significant problems for African American Medicare beneficiaries in accessing health care services.
Health Related Behaviors 1
Lifestyle choice and health related behaviors often reflect sociocultural patterns and economic resources and have a profound affect on health outcome. Lillie-Blanton et al. (1993) suggest that the health profiles of racial/ethnic minorities too often include risk factors that have been established in the literature to be associated with specific states of ill-health and disease and are modifiable. For example, obesity and being overweight, excessive tobacco and alcohol consumption, drug abuse and related behaviors (i.e., needle sharing, prostitution), stress, and interpersonal violence have all been associated with decreased health status and increased mortality (Manton, Patrick, & Johnson, 1 987; Lillie-Blanton et al., 1993; Nickens. 1991; Taylor, 1990).
Obesity/overweight. A strong association has been established between obesity and the prevalence of diabetes, hypertension, and breast and uterine cancer in women (Lillie-Blanton et al., 1993; Clark & Mungai, 1997; Trentham-Dietz et al., 1997). Blacks and Hispanic are significantly more likely to be overweight compared to Whites (Myers et al., 1 995; Clark & Mungai, 1997). Analyzing, data from the National Health and Nutrition Examination Survey 11 (NHANES: 1976-1980) and the Hispanic Health and Nutrition Examination Study (HHANES 1982-1984), Lillie-Blanton et al. (1993) found twice as many African American women (44.4 percent) were overweight compared to
Table 1. LEADING CAUSES OF DEATH BY GENDER AND RACE
I Diseases of heart
2 Malignant neoplasms
3 Cerebrovascular diseases
4 Chronic obstructive pulmonary diseases
5 Unintentional injuries
6 Pneumonia and influenza
7 Diabetes Mellitus
8 HIV infection
10 Homicide and legal intervention
White Males White Females
I Diseases of heart I Diseases of heart
2 Malignant neoplasms 2 Malignant neoplasms
3 Cerebrovascular disease 3 Cerebrovascular disease
4 Unintentional injuries 4 Chronic obstructive
5 Chronic obstructive pulmonary 5 Pneumonia and influenza
6 Pneumonia and influenza 6 Unintentional injuries
7 Suicide 7 Diabetes mellitus
8 HIV infection 8 Atherosclerosis
9 Diabetes Mellitus 9 Nephritis, Nephrotic syndrome,
10 Chronic liver disease and 10 Septicemia
Black Males Black Females
I Diseases of heart I Diseases of heart
2 Malignant neoplasms 2 Malignant neoplasms
3 Homicide and legal intervention 3 Cerebrovascular diseases
4 lHV infection 4 Diabetes mellitus
5 Unintentional injuries 5 Unintentional injuries
6 Cerebrovascular diseases 6 Pneumonia and influenza
7 Pneumonia and influenza 7 Certain conditions originating
in the perinatal period
8 Chronic obstructive pulmonary 8 HIV infection
9 Certain conditions originating 9 Chronic obstructive pulmonary
in the perinatal period diseases
10 Diabetes Mellitus 10 Homicide and legal
Table 1 --continued
American Indian/Alaskan Native Males American Indian/Alaskan Native Females
I Diseases of heart I Diseases of heart
2 Unintentional injuries 2 Malignant neoplasms
3 Malignant neoplasms 3 Unintentional injuries
4 Chronic liver disease & cirrhosis 4 Diabetes mellitus
5 Suicide 5 Cerebrovascular diseases
6 Cerebrovascular diseases 6 Chronic liver disease &
7 Diabetes mellitus 7 Pneumonia and influenza
8 Homicide and legal intervention 8 Chronic obstructive pulmonary
9 Pneumonia and influenza 9 Nephritis, nephrotic syndrome,
10 Chronic obstructive pulmonary 10 Congenital anomalies
Asian/Pacific Islander Males Asian/Pacific Islander Females
I Diseases of heart I Malignant neoplasm
2 Malignant neoplasms 2 Diseases of heart
3 Cerebrovascular diseases 3 Cerebrovascular diseases
4 Unintentional injuries 4 Unintentional injuries
5 Pneumonia and influenza 5 Pneumonia and influenza
6 Chronic Obstructive pulmonary 6 Diabetes mellitus
7 Homicide and legal intervention 7 Chronic obstructive pulmonary
8 Suicide 8 Suicide
9 Diabetes Mellitus 9 Congenital anomalies
10 HIV infection 10 Nephritis, nephrotic syndrome &
Hispanic Male Hispanic Females
I Diseases of heart I Diseases of heart
2 Malignant neoplasms 2 Malignant neoplasms
3 Unintentional injuries 3 Cerebrovascular diseases
4 HIV infection 4 Diabetes mellitus
5 Homicide and legal intervention 5 Unintentional injuries
6 Cerebrovascu lar diseases 6 Pneumonia and influenza
7 Chronic liver disease and 7 Certain diseases originating in
cirrhosis the perinatal period
8 Suicide 8 Chronic obstructive pulmonary
9 Congenital anomalies 9 Diabetes mellitus
10 Nephritis, nephrotic syndrome, 10 Pneumonia and influenza
Health, United States (Sources: Center for Disease Control and Prevention, National Center for Health Statistics: Vital Statistics of the United States. Vol. 11, Mortality. Part A)
Table 2. Rate ratio of age-adjusted death rates from 15 leading causes of
death, by sex and race United States, 1992
Rank Cause of Death Male:Female Black:White
I Diseases of heart 1.9 1.5
2 Malignant neoplasms 1.5 1.4
3 Cerebrovascular disease 1.2 1.9
4 Chronic obstructive pulmonary disease 1.7 .8
5 Accidents and adverse effects 2.6 1.3
Motor vehicle accidents 2.3 1.0
all other accidents and adverse effects 3.0 1.6
6 Pneumonia and influenza 1.7 1.4
7 Diabetes mellitus 1.1 2.4
8 HIV infection 7.0 3.7
9 Suicide 4.3 .6
10 Homicide and legal intervention 4.0 6.5
11 Chronic liver disease & cirrhosis 2.4 1.5
12 Nephritis, nephrotic syndrome &
Nephrosis 1.5 2.8
13 Septicemia 1.3 2.7
14 Atherosclerosis 1.3 1.1
15 Certain conditions originating in
the perinatal period 1.2 3.2
(based on infant mortality rate)
All Causes 1.7 1.6
From Center for Disease Control (1994b). Mortality patterns United States, 1992. MMWR, 43(49), 916-920
White American women (2- ).9 percent). Latino women also had higher percentages of being overweight; however, there was some variation by subethnic group (Mexican Americans, 41.6 percent; Puerto Ricans, 40.2 percent; and Cuban Americans, 31.6 percent).
Using the 1991 and 1992 Behavioral Risk Factor Surveillance System (BRFSS), a study by the CDC attempted to describe the prevalence of certain risk factors (cigarette smoking, sedentary lifestyle, and overweight) for chronic disease among racial/ethnic groups. African American and American Indian/Alaskan Native women more frequently reported being overweight (38% and 30% respectively) compared to White American (21.7%), Asian/Pacific Islander (10. 1 %) and Hispanic (26.5%) women. The prevalence of overweight in men was highest for American Indians/Alaskan Natives (34%) and lowest for Asian/Pacific Islanders (I I %). African American and White American men had relatively equivalent rates of being overweight (CDC, 1994a).
Diabetes is the second highest cause of adult death among African Americans.
The higher prevalence of obesity among black women and differential management of the disease has been linked to the increased rate of diabetes in African Americans (Manton, Patrick, & Johnson, 1987). Diabetes is also noted to disproportionately affect the Hispanic population (Zambrana & Ellis, 1995).
Physical activity. Regular physical activity has been found to have a significant impact on health. The positive effects of physical activity range from preventing and/or managing hypertension. heart disease, diabetes, osteoporosis, and obesity to lowering rates of cancer and stroke (Healthy People, 2000, 1992). A study looking at the relationship between physical activity levels and coronary heart disease and death found
that moderate levels of leisure time physical activity were associated with 63% less cardiovascular deaths and 70% less total deaths than the low leisure time physical activity group (Leon et al., 1987). Despite the positive effects associated with regular physical activity, the literature indicates that both Black and Hispanic are less likely to engage in regular physical exercise compared to Whites (Myers et al., 1995, Lockery and Stanford, 1995).
Tobacco. According to a report by the Center for Disease Control, an estimated 89.8 million (49.8%) adults in the United States have smoked at some point in their lives. Of these, 46 million (25%) were current smokers, 20.4% were daily smokers, and 4.6% were occasional smokers (CDC, 1994c). Tobacco has been cited as one of the most preventable causes of mortality in the United States (Nickens, 1991). Data from the 1993 National Health Interview Survey (NHIS-2000) were used to determine the prevalence of smoking among adults (CDC, 1994c). American Indians/Alaskan Natives and African Americans were found to have the highest rates of smoking, 38.7 percent and 26.0% percent respectively. Prevalence of smoking was lowest among Hispanics (20.4) and AsianTacific Islanders (18.2%). Significant sex differences existed across all race/ethnic groups with men reporting a higher rate than women, except for American Indian/Native Alaskan. Interestingly, American Indian/Native Alaskan women (40%) have the highest smoking rate across race/ethnicity and sex categories. Prevalence differences were also related to income level and education level. Persons living below the poverty level had higher smoking rates than those living at or above the poverty level. Smoking prevalence was lower among persons with less than 8 years of education compared to those with 9-
15 years of education. However, prevalence varied inversely with education level among persons with 9 or more years of education (CDC, I 994c).
Data from the 1991/1992 Behavioral Risk Factor Surveillance System also found the highest percentage of cigarette smoking among American Indians/Alaskan Native men and women (CDC, 1994a). Marin, Perez-Stable, Marin, Sabogal, & Otero-Sabogal (1990) found Hispanics to report smoking significantly less than non-Hispanic whites.
Alcohol use. Cirrhosis, which is often associated with excessive alcohol
consumption, is twice as high in African Americans as White Americans (Manton, Patrick, & Johnson. 1987). In a review of the literature on alcohol consumption, Bernard (1993) found no significant differences between reported drinking patterns among African Americans and White Americans. However, he found differential age effects in how African Americans and White Americans consumed alcohol. Between 14-17 years of age, African American youths tend to drink significantly less than White American youth. However, this gap gradually decreases with age as older African Americans report higher rates of alcohol consumption than younger African Americans and older White Americans report lower rates than younger White Americans. Manton, Patrick, and Johnson (1987) note that while there are more alcohol abstainers among Blacks, there are also more heavy consumers and that it may this continued high alcohol consumption in adulthood in certain Black subpopulations that led to the high incidence of cirrhosis in Blacks.
HIV infection. Data from the Center for Disease Control present some startling figures on the effect of HIV infection on the national death rate statistics. In 1992, HIV infection was the overall leading cause of death among men aged 25-44 years and the
fourth leading cause of death for women within that age range. The rate of AIDS and HIV infection varied widely across minority populations. When stratified by race, age (25-44), and sex, HIV is the leading cause of death for African American men (25.3% of total deaths) and Hispanic men (24.1 % of total deaths), the second leading cause of death for White American men ( 18.5% of total deaths), the sixth leading cause of death among Asian/Pacific Islander men (8.8% of total deaths), and the sixth leading cause of death among American Indians/Alaskan Native men (4.5% of total deaths). For this age bracket, HIV is the second leading cause of death among African American women (16.5% of total deaths), the third leading cause of death among Hispanic women (12.4% of total deaths), the sixth leading cause of death for White American women (3.8% of total deaths), the seventh leading cause of death among American Indian/Alaskan Native women (1.9% of total deaths), and the ninth leading cause of death for Asian/Pacific Islander women (L I % of total deaths). Mortality data for Hispanic men and women are from 1991 because data from 1992 were unavailable at the time of the report (CDC, 1993b).
In 1992, the death rate from HIV infection for persons aged 25-44 was three times as high for African American men (13 6 per 100,000) compared to White American men (42.1 per 100,000). Among young women, the African American/White American differentials are even greater. Death rates among young African American women (38 per 100,000) are 12 times those of young White American women (3.3 per 100,000) (CDC, 1993b).
Violence. With increasing attention being focused on health related behaviors and its influence on overall health status, homicide and interpersonal violence is an increasing
public health concern (Nickens. 1991). The largest difference in mortality between African Americans and White Americans is in the rate of homicide (Manton. Patrick, & Johnson, 1987). The ratio of homicide rates for African Americans are currently 7 and
4.2 times those for White men and women respectively (Nickens. 1991).
Theoretical Concepts of Minority Health Status
Williams. Lavizzo-Mourey. and Warren (1994) present a model illustrating the
complex relationship between race and health status (Figure 1). The model illustrates the interrelationship among macrosocial factors (i.e.. historical conditions, economic structures, and political order), racism (i.e.. prejudice and discrimination), and geographic origin/biological factors (i.e., physiological, morphological, and genetic factors). Within the model, macrosocial factors are thought to create racism, placing particular emphasis on selected physical characteristics or geographic origins of specific groups. Racism
Other models look specifically at the effects of social and environental
conditions on minority health outcome (Nickens. 1995; Lillie-Blanton et al.. 1993; Zambrana. 1988). For example. Lillie-Blanton et al. (1993) point to the importance of considering social environmental conditions as a significant determinant of health. Social conditions (i.e., type and place of employment and place of residence) can have a direct effect on health status in terms of exposure to occupational and environmental hazards. Social environmental conditions can also have an indirect effect through their influence on other determinants of health such as adjusting to life stress, lifestyle behaviors, access to health care and utilization of health service. Evaluating the relationship between health and social conditions among Latino and African American women, Lillie-Blanton et al.
S-, v2 fz
.--- snq ,
(1993) note that Latino and African American women have lower median incomes than White American women. They point to the negative influence of inadequate financial resources, limited education, and the stress of living in densely populated inner cities on health status. Zambrana (1988) also argues that theoretical models that attempt to interpret differential health outcomes of minority populations take into account the interaction of socioeconomic, racial, cultural, and regional variables on health care access, health status, chronic life stress, sources of social support, and work history.
Nickens (1995) notes the fact that despite similar overall poverty rates among Hispanics and African Americans, Hispanics tend to have significantly better health status than African Americans. He proposes that the effect of intergenerational poverty, systematic opposition, and frustrations related to perceived continual discrimination may have a indirect effect on health status and mortality rates among specific minority populations such as African Americans and American Indians.
Health Status, Socioeconomic Status, and Health Insurance: An Interdependent Relationship
Health Insurance Coverage.
Approximately 50% of African Americans and Latinos under 65 are uninsured or covered by public insurance (i.e., Medicaid) compared to 20% of White Americans. In addition, of this population, 15% of White Americans, 25% of African Americans, and 35% of Latinos are uninsured (Nickens, 1991). Cornelius (1993) used data from the 1987 National Medical Expenditure Survey to characterize the various barriers to medical care for White American, African American, and Hispanic children. She found 14% of White
American children were uninsured, compared to 22% of African American children, and 33% of Hispanic children. She also found uninsured children were twice as likely as children with insurance to have no usual source of health care. Her analysis further indicated that having health insurance reduced the disparities in use of health care services across groups. However, disparities still existed by type of insurance in terms of where children went for care, likelihood of having a regular doctor, and convenience of obtaining care. Valdez et al. (1993) used data from the 1980 and 1990 Current Population Surveys to examine determinants of health insurance coverage for Latinos. They found 39% of Latinos under 65 years of age were uninsured for the entire year of 1989. They also found health insurance coverage among Latinos varied substantially. Mexican Americans and Central and South Americans experienced approximately twice the uninsured rate as Puerto Ricans and Cuban Americans. Valdez et al. (1993) hypothesize that the disparities in the rates of insurance coverage are largely a result of primary employment in lower skilled and lower paid sectors of the economy, which are less likely to provide employment benefits such as insurance.
Members of minority groups are dramatically overrepresented among those on
public insurance programs (Kasper & McMillan, 1986; Nickens, 1991). As such, a brief overview of such programs, specifically Medicaid, is important for this review. Medicare (Title XVIII) and Medicaid (Title XIX) were established by U.S. Congress in 1965 with the enactment of the Social Security Act Amendments (DeLeon et al., 1992). VAIile Medicare served to provide medical care to the elderly and disabled regardless of income, Medicaid was designed to assist states in providing health care specifically for the poor and categorically needy. The two major groups eligible for Medicaid fall into two
programs: Aid to Families with Dependent Children (AFDC) and Supplemental Security Income (SSI). Medicaid is coordinated on both the federal and state level and provides medical assistance to individuals and families with low income and resources. Operating within broad federal guidelines, states have the authority to set eligibility standards and types of services covered by medicaid as well as rates of payment for services and the duration and scope of these services.
Given the large proportion of minorities who are poor, it was believed that
Medicaid would reduce the disparities in access to health care experienced by minorities. In fact, Davis et al. (1987) indicate that racial differences in use of health services decreased substantially between 1965 and 1980. This decrease is largely attributed to the introduction of Medicaid, Medicare, and other public health programs. However, a number of problems exist with Medicaid that affect access and utilization of health care services, including limitations in the extent of coverage, low reimbursement fees resulting in limited participation of health care providers, and poor quality of service received by Medicaid beneficiaries (Davis et al., 1987). In addition, different eligibility standards and levels of coverage across states result in differential access and utilization of health care services by state. As Valdez et al. (1993; p.889) summarize, "The categorical nature of the Medicaid programs, which serve only a fraction of the nation's poor, and the inadequate medical resources available through these programs further reduce the access."
Despite the noted problems intrinsic to this system of public insurance, Medicare and Medicaid combined currently account for nearly 90% of federal health care costs. Over the years, federal expenditures for Medicare and Medicaid have grown by
exorbitant rates. For example, federal health care costs grew by 20% between 1987 and 1989 (Broskowski, 1991). A report by the Congressional Budget Office estimates that by the year 2002, 25% of all federal expenditures will be devoted solely to Medicare and Medicaid (Bingamin, Frank, & Billy, 1993). Payments for Medicaid between 1980 and 1990 grew from $14 billion to $41 billion and, by 1996, federal expenditure is projected to exceed $120 billion (DeLeon et al., 1992). As such, state Medicaid agencies are currently facing critical challenges as federal mandates for Medicaid coverage expand, the number of eligible recipients increase, and the cost of delivering health care services in this country escalates rapidly. On both the national and state level, efforts are underway to reform the Medicaid system. However, reform efforts may have potentially drastic effects for minority populations. Efforts to manage and improve the Medicaid system must take into account the specific health needs and health related problems of the populations most likely to be effected by the changes. In addition, as Medicaid funding for the delivery of health care services is curtailed, more attention will have to be focused on preventative health behaviors.
Literature evaluating the effect of socioeconomic status on differential health outcomes across racial/ethnic groups have produced somewhat conflicting results. Socioeconomic status is usually measured by combinations of occupation, income, and educational attainment (Nickens, 1991). On these three variables, minorities tend to be lower compared to Whites (Nickens, 1995). Poverty rates for African Americans and Hispanics was 33.3% and 29.3% respectively in 1992, compared to 11.6% for Whites
(National Center for Health Statistics, 1994). Native American poverty rates are noted to be similar to that of Latinos and African Americans (Nickens, 1991).
Pappas, Queen, Hadden, and Fisher (1983) examined trends in mortality across socioeconomic groups using data from the 1986 National Mortality Followback Survey and the 1986 National Health Interview Survey. They found an inverse relationship between mortality rates and socioeconomic status for both African Americans and White Americans. However, higher mortality rates were found for African Americans compared White Americans for every age group and income or educational level. Sortie, Rogot, Anderson, et al. (1992) examined mortality rates for African Americans and White Americans by family income using data from the National Longitudinal Mortality Study. They also found lower mortality rates for higher income individuals for both African Americans and White Americans. However, death rates were higher for African Americans than White Americans within each level of income.
Otten, Teutsch, Williamson, & Marks (1990) followed up participants from the National Health and Nutrition Examination Study (NHANES). They were interested in determining the effects of six known risk factors (smoking, blood pressure, body-mass index, diabetes, alcohol use, and cholesterol level) and income on the Black-White differential in mortality. They found the unadjusted mortality ratio for African Americans verses White Americans was 2.3 for adults aged 35-54. This ratio decreased to 1.9 when the six risk factors were adjusted and to 1.4 when adjusted for the risk factors and family income level. Thus, approximately 69% of the excess mortality could be accounted for by the six known risk factors and family income. These findings provide
some evidence for the association between socioeconomic status, prevalence of risk factors, and the higher mortality rate among African Americans.
Bernard (1993) conducted a review of medical literature from 1987-1991 on the health status of African Americans. He found a higher prevalence of malignancies, diabetes, hypertension, obesity, homicide, and unintentional injuries compared to White Americans. When socioeconomic status and educational level were controlled, he found racial differences decreased or disappeared for some of the conditions. However, no mention is made within the review of exactly how socioeconomic status was defined or measured or what other variables were included in the analysis of these studies. Blendon et a]. (1989) analyzed data from a 1986 national survey on access and found disparities for racial/ethnic minorities persisted even after controlling for socioeconomic status.
Bassett and Krieger (1986) analyzed the influence of race and social class on
breast cancer survival rates in a population-based sample. After adjusting for social class, age, and other medical predictors of survival, the authors found that black-white differences in breast cancer survival rates diminished greatly. However, the data the authors used in the analysis had no direct indicators of social class. As such, they measured social class by using each patient's address to determine their census block group characteristics and then calculated each block group's social class composition by determining the proportion of residents who could be classified as "working class." Using such global measures of social class presents considerable concern regarding conclusions based on this definition.
Markides and Coreil (1986) coin the term "epidemiological paradox" in
describing the relationship between SES and Hispanic health status. The overall health
status of Hispanics as a group is good and similar to that of the White population, despite the fact that the overall poverty rate among Hispanics and African Americans is similar.
In an attempt to account for the differential health status of Affican Americans and Hispanics given the similar poverty rates among the two groups, Nickens (1995) proposes considering the differential effects of perceived poverty, systematic oppression and acculturation across minority groups in the United States. He further points out that low education and low family income may be necessary but not sufficient conditions for poor health status among minority populations and suggests the need to explore other causal factors that might mediate the relationship between health status and SES and mortality rates, such as, perceived powerlessness, frustration, and negative self-image.
For example, he notes,
Despite the importance of socioeconomic status on health status, when the relative
health status of minority populations is examined, it does not simply correlate
with their socioeconomic status. Socioeconomic status may operate in minority
populations with a time component. It may be that populations that have been
poor in the United States over several generations without substantial progress up
the socioeconomic ladder, suffering continual discrimination and frustration, are likely to feel much more powerless and will have a very different perception of their lot than newly arrived immigrants who are poor but still hopeful. (Nickens,
1995, p. 152)
In addition, as mentioned earlier, significant differences exist within Hispanic
subgroups in the United States on both economic and cultural levels. The combination of these subgroups into one group may mask important differences in health status.
The National Medical Expenditure Survey
The 1987 National Medical Expenditure Survey (NMES) was designed to provide estimates of health care use and expenditures as well as information on insurance coverage and health status. The survey has been used in a number of studies thus far to study health related issues. For example, researchers have examined the characteristics of employer-sponsored health insurance (e.g., Seccombe, Clarke, & Coward, 1994; Monheit & Vistness, 1994; Coward, Clarke, & Seccombe, 1993; Cooper & Monheit, 1993), issues related to prescription medications (e.g., Hahn, 1995; Willcox, Himrnmelstein, & Woolhandler, 1994; Olfson & Pincus, 1994c), characteristics of nursing homes and nursing home residents (see Murtaugh & Freiman, 1995; Romeis, 1994; Short & Kemper, 1994), and mental health care (i.e., Olfson & Pincus, 1994a, 1994b; Shea, Streit, & Smyer, 1994; Smyer, Shea, and Streit, 1994; Shea, Smyer, & Streit, 1993). The database has also provided information on access to and utilization of health care (i.e., Cunningham & Cornelius, 1995; Himmelstein & Woolhandler, 1995; Hahn, 1994; Cornelius, 1991, 1993; McKinney & Marconi, 1992) and health care financing and expenditures (see Rasell, Bernstein, & Tang, 1994; Short & Lair, 1994; Rubin, Altman, & Mendelson, 1994).
There are a few studies that have used this database to specifically explore issues related to subjective health status. For example, Franks et al. (1993) examined the relationship between health insurance and subjective health status using the 1987 National Medical Expenditure Survey. They compared adults with private or military insurance to adults without insurance for a year on several measures of subjective health
status: health perception, mental health, physical functioning, and role ftinctioning. They controlled for medical conditions, attitude toward the value of health care and insurance, family income, education, and race/ethnicity. Their analysis found individuals without health insurance had lower perceived health status compared to individuals with health insurance even after adjusting for the above mentioned potential confounding variables. They found those without health insurance to be younger, more likely to be male, less likely to be white, more likely to have a family income below the poverty level, less likely to have graduated from high school, and to have more negative attitudes towards the value of health insurance and medical care than those with health insurance.
Short and Lair (1994) also used the 1987 NMES to compare the health status of the insured and uninsured. However, in an attempt to deten-nine potential differential expenditure patterns across health insurance subgroups, their analysis further distinguished between five different coverage groups: privately insured with employment-related insurance; privately insured with nongroup insurance; persons who qualified for public insurance on the basis of their poor health; persons who qualified for public insurance by virtue of low family income; and the uninsured. Exploratory factor analysis of health status questions from the Self-administered Health Status Questionnaire of the NMES survey produced I I scales used in the analysis. They found significant differences across the five health insurance groups on self reported health status. Adults in employer-sponsored plans were the healthiest followed by those with nongroup private insurance, the uninsured group, the low-income publicly insured group, and lastly those who qualified for Medicare or Medicaid based on poor health. Overall, their analysis found that on all health status scales but one (developmental disability),
individuals with public insurance were less healthy than either the uninsured or the privately insured. One possible explanation for the differential health status among the uninsured and those with public insurance cited in the authors review is that people who are in relatively good health may perceive the costs of purchasing health insurance outweigh the possible benefits of having health insurance. Interestingly, unlike the Frank et al. (1993) study, the authors did not explore the relationship between being uninsured and having negative attitudes towards insurance and health care.
There has been some limited use of the National Medical Expenditure Survey to examine health care issues specific to minority populations. Cornelius (1993) examined access to health care for minority children and found the African American and Hispanic children were more likely than White American children to be poor, uninsured members of single-parent households and to wait longer to see a medical provider. Specifically, she found 14% of White American children were uninsured compared to 21.6% of African American children and 32.6% of Hispanic children. She also found approximately twice as many African American and Hispanic children were reported to be in fair or poor health compared to White American children. Moy and Bartman (1995) examined the relationship between physician race and the care of minority and medically indigent patients. They found that racial/ethnic minorities and medically indigent patients were more likely to receive care from nonwhite physicians. Looking at health status outcomes, the analysis revealed that these patients also tended to be sicker (were in fair or poor health, received an emergency department service, and/or were hospitalized),
Seccombe, Clarke, and Coward (1994) examined the effects of sociodemographic and employment factors on discrepancies in employer-sponsored health insurance among
minorities. Their analysis found minorities were less likely to have medical insurance provided to them by their employers and were more likely to be uninsured compared to white Americans.
A review of the literature using the 1987 National Medical Expenditure Survey found only two studies that evaluated the relationship between health behaviors and health status. Stoddard and Miller (1995) examined the impact of parental smoking behavior on the incidence of wheezing respiratory illness in children. Cornelius (199 1 looked at the health habits of school-age children. Thus far, this database has not been used to examine the relationship between health related behaviors and health status among minority populations.
The broad social, economic, and medical circumstances that mediate the differential health status of various groups within the United States have serious implications for health care reform and related public policy initiatives. The provision of universal health insurance and increased access to health care may present an ideal avenue to closing the gap in health status across groups. However, given the escalating costs of health care and the current political atmosphere, increased funding for public insurance programs and initiatives for increasing access to health care services for the poor appears unlikely. In addition, while disparities in socioeconomic status across groups account for a portion of the difference in health status and mortality rates among minorities and white Americans, a review of the literature suggests that it does not account for all the variance. Differences in the social or physical environment between
African Americans and white Americans may be partially responsible. Variations across groups in health related behaviors may also play a significant role in overall health status.
Ultimately, attention will have to focus on health outcomes and interventions will need to extend beyond the health care system to broader social, economic, and educational systems (Nickens, 1995). Psychology could play a significant role, addressing these broader systems through preventative health care and education. However, it will be profession's responsibility to successfully educate minority populations as to the impact of behavior on health and the importance of preventative health care, recognizing the influence of cultural patterns, socioeconomic status, and health related behaviors on health status and health outcomes.
Purpose of Research
Differences in health related behaviors may help explain some of the variance
associated with the differential health status of minority populations. The purpose of this study is to examine the relationship between race/ethnicity, health related behaviors, and self-reported health status. Given the established association between health, socioeconomic status, and the probability of having health insurance, the effects of SES and insurance coverage will also be examined as they relate to race/ethnicity and health.
Specific exploratory model of interest
As illustrated in figure 2, a hierarchical causal model will examine the
relationship between health related behaviors and self-reported health status. In the model, four groups of variables will be included as direct predictors of health status: (1) race, (2) socio-demographic indices, (3) health insurance, and (4) health related
behaviors. Components of socio-demographic dimension include sex, age, education, income, and marital status. Self-reported health status will be measured through a number of components: Overall Health Rating, Subjective Health Status, Role Functioning, Physical Functioning, Acute and Chronic Symptoms, and presence of Medical Conditions.
Several indirect effects on self-reported health status are also proposed in the model. The effect of demographic variables on (a) insurance coverage and (b) health related behaviors will be examined. A secondary model will explore the effects of race and socio-demographic variables on (c) attitudes towards health care/health insurance as well as the effects of (c) attitudes towards health care/health insurance on (d) health insurance coverage.
Previous investigations have explored several pathways presented here. For
example, Seccombe, Clark, and Coward (1994) examined the influence of demographic variables on insurance coverage. Short and Lair (1994) have looked at the relationship between insurance coverage and health status as well as the health status of the uninsured mediated by family income level. Franks et al. (1993) also examined the effects of health insurance coverage on subjective health status. Cornelius (1993) investigated barriers to health care for white American, Affican American, and Hispanic children. A subset of her analysis observed the direct effects of perceived health status on racial/ethnic group membership. Based on comprehensive literature review, no study to date has used this database to explore the relationship between health related behaviors among minority groups and self-reported health status.
Data and Sample
Data used in this study were obtained from the Household Survey component of the 1987 National Medical Expenditure Survey (NMES). The National Medical Expenditure Survey (NMES) was sponsored by the Agency for Health Care Policy and Research and used a national probability sample of civilian, noninstitutionalized U.S. citizens. The survey was designed to provide nationally representative estimates of health status, health insurance coverage, use of health care services, health care expenditures, and sources of payment for the period from January 1 to December 3 1, 1987.
The NMES used a multistage stratified area probability design of approximately 15,000 households in the United States representing a total sample of 34,600 persons. To provide more accurate analysis of underrepresented populations and in an effort to address policy concerns specific to certain populations, the NMES oversampled African Americans, Hispanics, the elderly, the functionally impaired and low-income families.
The Household Survey component of the NMES was conducted over four rounds of personal and telephone interviews at 4-month intervals. A short telephone interview constituted a final fifth round. Baseline data on household composition and employment
and insurance characteristics were updated for each quarter, and information on all use of and expenditures for health care services and sources of payment was obtained. The response rate for the Household Survey across all five rounds of data collection was approximately 80% of total households identified. In addition to personal and telephone interviews, a self-administered Health Status Questionnaire for adults and children was mailed to participants between the first and second interview rounds and collected information on health habits, self-assessed health status, mental health, functional status, health attitudes, vision and hearing, and preventive care. The self-administered questionnaire included checklists of the most common chronic conditions and a checklist of symptoms that asked whether participants had experienced the symptoms in the previous 30 days and whether they had seen a physician about them. Separate questionnaires were developed for children and adults.
The present investigation used demographic and health insurance data from the second round of the Household Survey interview and information gathered from the selfadministered Health Status Questionnaire. This study is restricted to data obtained from adults, ages 18 and older, who were denoted as heads of household and who responded to the Self Administered Household Survey questionnaire portion of the survey, resulting in a total sample size of 12272. Adjustments for missing data resulted in final sample size of 11287.
The independent variables of interest in this study included race/ethnicity, sociodemographic variables (gender, age, marital status, education, poverty status), health insurance coverage (private, public, uninsured), attitudes towards health care and health insurance, and health related behaviors. Six dependent measures were constructed to reflect overall health status: Overall Health Rating, Role Functioning, Physical Functioning, Chronic Symptoms, Acute Symptoms, and Medical Conditions.
In the NMES Household Survey, participants were asked to best classify their
ethnic/racial background as American Indian, Alaskan Native, Asian or Pacific Islander, black, white, or other. In addition, participants were asked if their main national origin was among the following Hispanic subgroups: Puerto Rican, Cuban, Mexican or Mexicano, Mexican American or Chicano, other Latin American, or other Hispanic. Using similar data collection instruments and interview procedures over the same time period, NMES conducted a separate survey of American Indians and Alaska Natives living on or near reservations and eligible for services from the Indian Health Service. However, due to differences in questionnaires across the two separate surveys, the results from the Survey of American Indians and Alaska Natives were not included in this analysis. Sample size for Asian or Pacific Islanders was small and, as such, were not included in this analysis. Individuals classified as "other" were also not included in this study. In addition, sample sizes for the Hispanic subgroups were small and thus did not permit separate analysis. As a result, all Hispanic subgroups were combined into a single
group. Given the heterogeneity of national origin and the cultural differences that may exist across subgroups of the Hispanic population, results related to Hispanic outcomes should be considered cautiously. Race/ethnic classifications were dummy coded as African American (yes=l, no=O), White American (yes=l, no=O), Hispanic (yesl, no=O).
Participants were classified as having one of the following insurance coverage: 1) Uninsured; 2) Public insurance (Medicaid and other public medical assistance); 3) Private insurance. Coverage through CHAMPUS (Civilian Health and Medical Program of the Uniformed Services) was included under private insurance category since coverage benefits more closely approximate private rather than public insurance (Hahn, 1994). For analytical purposes, the insurance variable was dummy coded as Uninsured (yes= 1, no=O), Public (yes=l, no=O), and Private (yes=l, no=O).
Poverty status was used in this investigation as a measure of socioeconomic status. As part of the NMES Household survey, income data were collected on 26 separate sources of incomes. An aggregated income measure was then developed by NMES for each person by summing over each of the income sources. Family income measures were constructed for each person by summing over the income for each person in the person's family and then combining these amounts into the single family measure. The person's family income was then compared to the official poverty threshold for 1987 for the appropriate family size to create a poverty status variable. Poverty status was initially categorized by NMES into 6 categories (poor, near poor, low income, middle income, high income, negative income). However, due to skewed distribution of income data, for this present investigation, the poverty status variable was re-categorized and
dummy coded into poor income (yes= 1, no=O), middle income (yes= 1, no=O), high income (yes=1. no=O). Participant's age was coded as a continuous variable. Marital status was treated as a dichotomous variable. Individuals who reported being single, separated, divorced, or widowed were combined into one category and coded as not married (married=O). Individuals who reported being married were coded as married (married= 1). The distribution of data from the education variable was skewed. Therefore, individuals were re-categorized into three groups (less than high school, high school, more than high school).
The self-administered health status questionnaire included ten questions related to participants' attitudes towards health insurance and health care in general. Questions were rated on a 5 point likert scale, with higher scores indicating more negative attitude towards the value of health care and health insurance. For this analysis, two separate subscales were developed. The first subscale consisted of four questions measuring attitudes towards health insurance and the second subseale consist of six questions assessing attitudes about health care in general. Cronbach's alpha calculated for these two scales were .43 and .57 respectively. Following the proposed model in Figure 2, preliminary regression analysis was conducted to determine degree to which attitude towards health care and attitude toward health insurance could be predicted from demographic variables as well as degree to which the two attitude scales could predict type of insurance coverage. However, results from preliminary analysis found no significant amount of variance accounted for. Given the low alpha associated with both scales (suggesting poor reliability) as well as nonsignificant relationship with variables of
interest, both scales were dropped from further analysis, and as a consequence, the attitude component of Figure 2 was deleted.
For this investigation several items from the self-administered health status
questionnaire were used to measure various health related behaviors: Body mass index, smoking index, regular physical exercise, blood pressure checked within past year, eating breakfast, and frequency of wearing a seat belt. Two separate indices of weight were calculated for this study. Body Mass Index was calculated by dividing self-reported weight (kg) by self-reported height (meters) squared. For ease of interpretation in logistic analysis, a binary variable of Overweight was developed. Based on cutoff s obtained from Health People 2000 (Department of Health and Human Services, 1992), individuals with a body mass index equal to or greater than 28 were characterized as being overweight (yes=l). Individuals with a body mass index less than 28 were characterized as not overweight (no=O). Two separate measures of smoking behavior were also used for this study. For logistic regression, a dichotomous Ever Smoked index was developed. Individuals who reported having smoked more than 100 cigarettes in lifetime were categorized as I (yes), otherwise they were categorized as 0 (no). To allow for greater specificity in hierarchical regression analysis, an index of total cigarettes smoked in lifetime was created. For current smokers, age started smoking was subtracted from present age and multiplied by 365 to obtain total number of days smoked. This number was then multiplied by number of cigarettes smoked per day to obtain total number of cigarettes smoked in lifetime. For previous smokers, the age started smoking was subtracted from the age the participant stopped smoking and multiplied by 365 to obtain total number of days smoked. This number was similarly multiplied by number of
cigarettes smoked per day. Individuals who reported never having smoked were included as having zero number of cigarettes smoked per day. Variables for whether the participant eats breakfast and wears a seat belt were initially measured on a 4 point scale (never, seldom/rarely, almost daily, or always/everyday). However, due to skewed distribution of scores, both variables were dichotomized into yes (almost daily, everyday) and no (never, seldom/rarely). The eating breakfast variable was ultimately excluded for further regression analysis due to non-significant effect on all models. Regular physical exercise and blood pressure check were both coded as dichotomous variables (yes=l, no=O).
For this investigation, several health related measures were initially developed from questionnaire items to serve as dependent variables: Overall Health Rating, Subjective Health Scale, Mental Health Status, Role Functioning, Physical Functioning, Acute Symptoms, Chronic Symptoms, and Medical Conditions. For each of these scales, Cronbach's alpha was calculated to determine reliability of scales in relation to summed items.
To obtain participants' Overall Health Rating, they were asked to rate their health as excellent, good, fair, or poor. The Health Rating scale was normally distributed. The Subjective Health scale consisted of four separate questions rated on a four point scale (a=.87). However, due to non-normal distribution of the scale as well as it's similarity to and high correlation (.70) with the Overall Health Rating scale, the Subjective Health scale was dropped from further analysis. The Role Functioning scale consisted of two
questions related to the degree health keeps the individual from performing certain tasks (a=.81). The Physical Functioning scale consisted of five questions related to the degree health limits various physical activities ((x=.87).
The Acute Health Symptoms scale originally consisted of nine questions related to whether the participant demonstrated particular symptoms (i.e., sudden feeling of weakness, repeated indigestion or upset stomach) within the past thirty days. Three symptoms (high fever, skin rash, and bleeding) were removed from the scale due to low correlation with other variables in scale; the resulting scale had an alpha of .65. The chronic health symptoms scale included 5 questions related to more chronic symptoms (i.e., repeated backaches, frequent headaches). Similarly, due to low correlation with other variables in scale, one variable (hemorrhoids) was dropped from the scale. The resulting scale had an alpha of .52 The Household Survey Questionnaire also included a checklist of I I serious medical conditions which comprised the medical conditions scale (a=.69).
The Acute Health symptoms, Chronic Health symptoms, and Medical Conditions scales all demonstrated skewed distribution, with individuals acknowledging either no symptoms or one or more symptoms. Consequently, dummy variables were created (either having any one of the symptoms or condition or not) for further analysis by logistic regression.
Certain responses to questionnaire items were treated as missing data. These include the responses categories: 1) did not ascertain, 2) don't know, 3) refused, and 4)
inapplicable. Full regression models using the SAS system for computer statistical analysis employed listwise deletion of missing data resulting in a sample size of 11287. In order to facilitate comparison across models, the reduced sample size of 11287 was used for all analysis in this study. Elimination of missing data reduced overall sample size by 8%.
The purpose of this study was to examine the relationship between race/ethnicity and health status while controlling for various socio-demographic variables, health related behaviors, and type of insurance coverage. Data were first analyzed to determine if differences existed across racial/ethnic groups on independent and dependent variables. Chi-square and F tests were computed to determine whether differences were statistically significant. Correlations were then estimated within sets of independent variables (i.e., socio-demographic, health related behaviors) and among dependent variables to explore degree of association among measures (Statistical software automatically produces pointbiserial and phi-coefficient for dichotomous data; Cohen & Cohen, 1983)).
Multiple regression analysis was used to determine whether the effects of
race/ethnicity persist with the stepwise inclusion of sets of variables proposed to mediate the relationship between race/ethnicity and health status (stepwise inclusion followed direction proposed in the model depicted in Figure 2). Hierarchical multiple regression was used for analysis of continuous dependent variables while hierarchical logistic regression was used for dichotomous dependent variables.
Supplemental regression analysis was used to analyze the direct relationship between race/ethnicity and the variables hypothesized to mediate of the effect of race/ethnicity on health status. First, the relationship between health insurance coverage and race was analyzed, controlling for socio-demographic variables (pathway a on Figure 2). Second, the relationship between race and health related behaviors was analyzed, controlling for the effects of socio-demographic variables (pathway b on Figure 2). Descriptive Analysis
Tables of descriptive analysis are presented at the end of the section. Table 3
presents racial/ethnic comparisons among both dependent and independent variables of interest. There was a significant difference in the gender composition of the sample with women comprising a majority (55%) of the Black sample but a minority of the White (36%) and Hispanic sample (36%). Whites in the sample were relatively older with a mean age of 51 compared to Blacks (mean age=-47) and Hispanics (mean age=43). Sample differences in income level also emerged with 15% of Whites being categorized as poor compared to 26% of Hispanics and 34% of Blacks. Blacks were significantly less likely to be married (35%) compared to both Whites (56%) and Hispanics (57%). In terms of the educational distribution of the sample, approximately 50% of the Hispanics had not completed high school compared to 42% of Blacks and 29% of Whites. In addition, more Whites had post-secondary education (38%) compared to Blacks (25%) or Hispanics (23%). There were significant differences in insurance coverage across the three racial/ethnic groups. Of the Whites in the sample, 80% were privately insured compared to 57% of Blacks and 56% of Hispanics. In addition, approximately 25% of Hispanics were uninsured compared to 10% of Whites and 15% of Blacks.
There were significant differences across racial/ethnic groups on health related behaviors. Of the Whites in the sample, 54% reported engaging in regular physical activity compared to 40% of Blacks and 44% of Hispanics. Black had the highest mean body mass index and a greater proportion of overweight persons (30%) compared to Whites (19%) or Hispanics (23%). Of the three groups in the sample, Blacks were the least likely to regularly wear seat-belts while Hispanics were the least likely to report having had their blood pressure checked within the past year. More Whites reported having ever smoked (defined as smoking over 100 cigarettes in lifetime) compared to Blacks or Hispanics. In addition, Whites had a significantly higher number of cigarettes smoked in lifetime than to Blacks and Hispanics.
Table 4 presents percentage distributions and tests of significance for health
outcome variables. On the Overall Health Rating scale, Whites and Hispanics had similar ratings of excellent and good health, while Blacks had lower ratings of excellent and good health. On the Role Functioning scale, significantly more Blacks reported not being able to work due to health problems (18%) compared to Whites (14%) and Hispanics (11%). Approximately 23% of both Blacks and Whites reported that their health limits the kinds of work they can do compared to only 14% of Hispanics. Significant differences also occurred across items on the Physical Functioning scale. Interestingly, while more Blacks indicated that their health limits moderate activity (22%) compared to Whites (18%), more Whites reported that their health limits vigorous activity (42%) compared to Blacks (37%). Hispanics reported the least limitations in moderate (13%) or vigorous (26%) activity. Finally, more Blacks reported having trouble walking a block, climbing stairs and lifting, bending, or stooping compared to Whites and Hispanics.
Differences by race/ethnicity on the health outcome scales related primarily to medical conditions are presented in Table 5. On the Acute Symptoms scales, significantly fewer Hispanics (42%) endorsed symptoms than either Whites (50%) or Blacks (50%). A similar pattern of results emerged for both the Chronic Symptoms scale and the Medical Conditions scale.
Table 6 presents correlation coefficients among the socio-demographic variables of race, gender, age, marital status, education, income, and insurance coverage. Results found significant relationships between race and socio-demographic variables. For example, comparisons between Whites and Blacks found Whites were more likely to be older, have higher incomes, higher educational levels, to be married, and to be privately insured. A similar pattern of association occurred between Whites and Hispanics, however, there was no correlation for marital status. The significant differences across racial/ethnic groups that emerged on the various socio-demographic dimensions included in this study suggest that these dimensions may influence the relationship between race/ethnicity and health status and lend credence to their inclusion in multivariate analysis as mediating variables.
Males in the sample were significantly more likely than females to be married, have higher income levels, and be either privately insured (compared to publicly) or uninsured (compared to publicly). As one would expect, higher educational levels were significantly associated with higher income levels. In addition, higher income levels was correlated with possession of private insurance (compared to public or being uninsured).
Table 7 presents correlation coefficients for the health related behaviors. There was a positive correlation between engaging in physical activity and wearing seat-belts,
indicating that those who engage in regular physical activity also tend to regularly wear their seat-belts while driving. Physical activity was negatively correlated with body mass index as well as smoking index, indicating that individuals who engage in regular physical activity tend to weigh less and smoke less than individuals who do not engage in regular exercise. There was a positive correlation between body mass index and having had blood pressure checked; individuals who weigh more were also more likely to have had their blood pressure checked within the past year.
The correlation coefficients for the health outcome variable scales are presented in Table 8. Overall, the outcome scales were moderately correlated with each other, suggesting that they are most likely tapping into a similar construct: health status. Selfreport measures of functioning (Overall Health Rating, Physical Functioning, Role functioning) tended to have somewhat higher correlations than the scales that measure actual symptoms (Acute Symptoms, Chronic Symptoms, Medical Conditions). However, the fact that all scales, expect for the association between Physical Functioning and Role Functioning, are not very highly correlated provides some evidence that they are also measuring unique components of that construct. The exception was Role Functioning and Physical Functioning, which were highly correlated (r 2 =39). Nevertheless, they were treated as separate outcome variables in this analysis to allow for greater specificity (i.e., differences in amount of variance accounted for in multivariate analysis across the two scales may provide important information about degree which there are racial/ethnic differences in the way health limits the ability to perform a role (i.e., work) verses limiting daily physical functioning (i.e., housework).
Table 3. Percentage Distribution and Test of Significance of Demographic Variables, Insurance Status, and Health Related Behaviors by Race/Ethnicity
Variable White Black Hispanic Chi-square
%Male 64.1 44.9 63.6 290.9*
%Female 35.9 55.0 36.3
Age (Mean) 51.1 47.4 43.5 113.1*
(SD) (19.1) (17.2) (15.9)
%Poor 15.3 33.8 26.4 553.1*
%Low-Mid 48.5 46.8 52.1
%High 36.2 19.4 21.5
% Married 56.0 35.0 57.6 344.7*
%>High school 37.6 24.9 22.8 192.9*
%Private 80.4 57.4 56.2 820.7*
%Public 9.2 27.2 18.5
%Uninsured 10.4 15.4 25.3
Health Related Behaviors
BodyMass Index(Mean) 24.8 26.3 25.6 83.3*
(SD) (4.8) (5.6) (4.8)
%Overweight 18.7 29.1 23.2 126.0*
%Physical Exercisea 53.9 40.3 44.9 153.8*
%Ever Smokeda 61.6 55.1 50.6 63.4*
# of Cig/Life(Mean) 105243 52969 44424 152.4*
(SD) (169880) (111911) (96432)
%Wear Seat-belta 53.7 45.6 53.1 46.5*
%Blood Pressure Checka 76.4 75.62 63.9 73.2*
a% yes, p<0.0001I
Overweight = Bodymass index >= 28, Ever Smoked = Smoked over 100 cigarettes in lifetime
Table 4. Percentage Distribution and Test of Significance of Outcome Variables by Race/Ethnicity
Variable White Black Hispanic Chi-square
Rate Health 100.7*
Mean (SD) 2.07(.81) 2.25(.81) 2.08(.79)
%Excellent 24.1 16.3 23.0
%Good 49.8 49.3 50.3
%Fair 20.5 27.6 21.8
%Poor 5.7 6.8 4.9
Role Functioning 57.9*
Mean (SD) .37(.71) .41(.75) .25(.60)
%Can't work due to health 14.0 17.7 10.9 31.0*
%Health limits kind of work 23.2 23.6 13.5 49.1 *
Physical Functioning 156.8*
Mean (SD) 1.16(1.64) 1.25(1.81) .76(1.44)
%Health limits mod. Activity 18.4 21.5 13.3 31.2*
%Health limits vig. Activity 42.4 36.6 25.9 113.0*
%Trouble walking one block 11.5 16.7 7.7 65.9*
%Trouble climbing stairs 20.7 25.2 14.7 49.4*
%Trouble lift/bend/stoop 23.2 25.2 14.7 45.1*
Means on scales: Higher scores = worse functioning
Table 5. Percentage Distribution and Test of Significance of Medical Conditions Scales by Race/Ethnicity
Variable (% yes) White Black Hispanic Chi-square
Acute Symptoms Scale 49.6 49.6 41.9 21.4*
Chronic Symptoms Scale 50.9 49.4 40.5 32.4*
Medical Conditions Scale 45.5 46.9 30.4 87.6*
Scales are coded (1=1 or more symptoms, O=no symptoms)
0 (U 0
En C> -0 Z c
11 0 C:)
0 Im 11 cc
C) C> 0 0
u C> M r
cn C; -0 it
as (U 0 E 0
E m vi : E
00 -,T r-4 CP a
Ow X v
SO > >
Cl. A v
as C> 00
> ci 00
ci on 7
1:4 A t:,
C,3 M = u 40 = =
Im 4) PZ
to > >
to < 2
u) e4. 00 4~ en 4. 0)
CO m~ 0N
0 9 0
,w to 00
~t 00 00 d oo t;C a 0 r~~
Following recommendations by Cohen and Cohen (1983), multiple and logistic regression analysis was used to test the causal model depicted in Figure 2. Table 21 provides a list of all variables included in the regression analysis along with corresponding coding. For the dependent variables of Overall Health Rating, Role Functioning, and Physical Functioning, hierarchical multiple regression was used to test the predictive model. For all three dependent variables, race/ethnicity was entered in the first step of the model. As mentioned previously, White, Black, and Hispanic were dummy coded as either having the attribute or not (White (Il=yes,O=no), Black (I =yes, O~no), Hispanic (Il=yes, O=no)). Black and Hispanic were included in the model with White left out, thereby allowing comparison of both Blacks and Hispanics to White. The second step added socio-demographic variables to the model. The three categories of income were dummy coded into poor( 1 =yes, O=no), middle income (I =yes, O--no), high income (I =yes, O=no)) and both poor and middle income were included in the model, allowing comparison to high income. Gender was coded as male (Il=yes, O-no), allowing comparison to females. Marital status was coded as married (I =yes) (O-no), comparing to those not married. Education was dummy coded into less than high school (Ilyes, O=zno), high school (Il=yes, O=no), and more than high school (Il=yes, Ono). Less than high school and high school were included in the model, allowing comparison to individuals with more than a high school education. Finally, age was treated as a continuous variable. In the third step, health related behaviors were added to the model.
Body Mass Index and Smoking Index were treated as continuous variables while Physical Exercise, Blood Pressure Check, and Wear Seat Belt were coded as I =yes and O=no. For the final ful model, health insurance coverage was added to the sets of independent variables. Public (Il=yes,O=no), private (Il=yes, O=no), and uninsured (Il=yes, O=no) were dummy coded and both public and uninsured were included in the model, thereby allowing comparison to private.
The dependent variables of Acute Symptoms, Chronic Symptoms, and Medical
Conditions consisted of binary response data (scales coded as either no symptoms (0) or I or more symptoms (1)). As such, hierarchical logistic regression was used to determine the influence of race/ethnicity, socio-demographic variables, health related behaviors, and insurance coverage on the odds of having one or more medical symptom (versus having no symptoms). The steps for inclusion in the models are identical to the multivariate regressions. However, the dichotomous variables of Overweight (yes=l, no=0), and Ever Smoked (yes= 1, no=0) were used in place of Body Mass Index and Smoking Index for ease of interpretation of odds ratios.
Separate computational analysis were done for both multiple regressions and
logistic regressions to allow for across group comparisons of variables within each step of model. Specific group comparisons made within steps included: Black vs. Hispanic, high school vs. less than high school, poor vs. Middle income, and public insurance vs. uninsured. The equation used to obtain appropriate t-values was t= B,-B2
\ var P3i+ var p2- 2(covo3.43)
Overall Health Rating
Table 9 presents a summary of the regression analysis for variables predicting Overall Health Rating. For the first step in the model, race/ethnicity was entered independently to estimate the degree to which perceived health status can be predicted by race/ethnicity. Race/ethnicity was significant with Blacks being more likely to report poorer health status compared to Whites. Separate analysis indicated that Blacks were also more likely to report poorer health status than Hispanics (t=-5.33). There was no difference in perceived health status between Hispanics and Whites.
Adding the set of socio-demographic variables (gender, age, marital status, education, and income level) to race/ethnicity as predictors in the second model accounted for an additional 21% of the variance. Age (P=.32), education (
age, have lower educational levels, and to be poor or have middle income levels compared to high income.
In the third model, the set of health related behaviors (Body mass index, smoking index, physical activity, blood pressure check within past year, wearing seat belt) were added to race/ethnicity and socio-demographic variables as predictors. Health related behaviors accounted for an additional 6% of the variance. The inclusion of health related behaviors to the model accounted for an additional reduction in the effect of being Black, although it remained statistically significant. Inclusion also eliminated the effects of being male. The reduction of effect suggest that differences in health behaviors may mediate some of the initial differences found based on race, age, and gender. In the third equation, poorer ratings of health are associated with being Black, older, having lower educational levels, low to middle income levels, weighing more, smoking more, not engaging in regular physical activity, having had blood pressure checked within past year, and not using seat belts regularly.
In the final model, insurance coverage was added to the sets of predictor variables, explaining an additional 1% of the variance in the model. Compared to model 2, adding insurance further reduced the Black effect by 25% and poor effect by 23%, although they remained significant in the model. Uninsured individuals had poorer health ratings compared to the privately insured (B=.09). Publicly insured individuals had the worst health ratings when compared to the privately insured (B=.26) or uninsured (AB=.26-.09, t--6.54).
For the full regression model, race/ethnicity, socio-demographic variables, health related behaviors, and type of insurance coverage explained a total of 28% of the variance
associated with perceived health status. Compared to the initial model with race alone, the full model resulted in a 61 % reduction in the Black effect. Overall, poor health rating was associated with being Black, older, poor or middle income, weighing more, smoking more, not engaging in physical activity, having had blood pressure checked within past year, not wearing seat belts regularly, and being uninsured or publicly insured. There were no effects for Hispanics in the model indicating that Hispanics rate their overall health in a similar manner as Whites.
Table 10 presents results of regression analysis for Role Functioning.
Race/ethnicity accounted for a small portion of the variance with Hispanics reporting significantly better role functioning than both Whites (B=-13) or Blacks (AB=-. 13-.04, t---6. 15). Blacks reported significantly poorer role functioning than Whites (B=.04). The addition of socio-demographic variables to the model increased the amount of variance accounted for to 20%. The inclusion of demographic variables eliminated the initial race effect for Blacks and reduced the effect for Hispanics by 38%, suggesting the influence of these demographic variables in accounting for some of the variance previously associated with race. Similar to the Overall Health Rating scale, both age (P3=.36) and income level (poor: 0=.21, middle: 0=.08) accounted for a large portion of the socio-demographic effect. Separate analysis indicated that low income individuals were more likely to report poorer role functioning than those having a middle income level (AB=.38-. 1, t=16.87). Poorer role functioning was significantly associated with being older, not married, having less than a high school education, and low to middle income levels.
In the third model, the inclusion of health related behaviors accounted for an additional 4% of variance. Addition of health related behaviors slightly reduced the effect of age (P=.30 vs. 0=36 in model 2), although age remained significant. Interestingly, inclusion of the health related behaviors increased the effect of being Black. This finding suggests that once certain health related behaviors are controlled for, Blacks actually report better, although not significantly, role functioning, compared to Wliites. Increased Body mass and smoking, lack of physical activity, and having had blood pressure checked within past year were all associated with poorer role functioning.
The final model included the insurance variable and accounted for an additional 2% of the variance. Both being uninsured and publicly insured were associated with significantly poorer role functioning compared to the privately insured. In separate analysis, publicly insured individuals were found to report significantly poorer role functioning compared to the uninsured (t=13.04). The inverse Black effect that emerged in model 3 becomes significant with the inclusion of insurance coverage. In addition, the significant effect for Hispanics increases further. Although still significant, the effect of being poor is reduced by 27% with the addition of the insurance variable to the model. The effect of body mass was eliminated. Age, lack of regular physical activity, and possessing public insurance had the largest effects on the model.
The full regression model measuring the impact of race, socio-demographic
variables, health related behaviors. and insurance coverage on individuals perceived role functioning accounted for a total of 27% of the variance. Overall, poorer role functioning is associated with being male, older in age, having lower income levels, more lifetime
smoking, not engaging in physical activity, having had blood pressure checked within past year, being uninsured or possessing public insurance, and being White. Physical Functioning
Results for the dependent variable of physical functioning are presented in Table I I and are similar to previous findings for Role Functioning. For the first model of race only, Hispanics demonstrated significantly better physical functioning than both Whites (B=-.40) and Blacks (ALB=-.40-.09, t=-8.16). Blacks were not significantly different from Whites in perceived physical functioning. The inclusion of demographic variables in the second equation accounted for an additional 3 1 % of the variance and reduced the race effects for Hispanics by 43%. Significant effects emerged for gender (male:p=-.04), age (0=.46), education (
Health related behaviors were added to the third model and accounted for an additional 4% of the variance. Inclusion of the health related behaviors to the model increased the effect of being Black, with Blacks reporting significantly better physical functioning compared to Whites. This finding indicates that once variations in health related characteristics such as physical activity, smoking, and body mass are controlled for, Blacks report less limitations physical functioning than Whites.
Inclusion of health related behaviors slightly increased the effect of not being married but reduced the effects of all the other demographic variables, although all
(except having a high school education) remained statistically significant in the model. Higher body mass, more lifetime smoking, not engaging in regular physical activity, and having blood pressure checked within past year were associated with poorer physical functioning. Lack of'regular physical activity had the largest effect (0=-.19) of all the health related behaviors included in the model.
For the final full model, insurance was added and explained another 2% of the variance. As with the Health Rating scale and the Role Functioning scale, publicly insured individuals reported significantly poorer physical functioning compared to privately insured individuals (0=.14). Publicly insured individuals' physical functioning was also significantly poorer than uninsured individuals (AB=.71-.09, t--12.2). There was no difference between being privately insured and being uninsured for the Physical Functioning scale. Although still significant, the effect of being poor and having middle income is further reduced (26% and 12%, respectively) in the full model, suggesting that insurance coverage may help explain the income effect. As with Role Functioning, inclusion of the insurance variable in the model increased the race effects with both Blacks and Hispanics having significantly better physical functioning compared to whites. Overall, individuals with poorer physical functioning are more likely to be white, female, older, not married, low to middle income levels, and to weigh more and smoked more cigarettes in lifetime, have had their blood pressure checked within past year, and to be publicly insured.
Acute Symptoms Scale
Results for the logistic regression models predicting Acute symptoms are presented in Table 12. For the first model of race/ethnicity alone, Blacks were not different from whites in their report of acute symptoms. On the other hand, Hispanics were 27% less likely to report acute symptoms compared to Whites. Addition of sociodemographic variables to the second model increased the effect for Blacks. Thus, controlling for socio-demographic variables resulted in Blacks reporting significantly less acute symptoms than Whites. Interestingly, poor individuals were 67% more likely and middle income individuals were 25% more likely to report acute symptoms compared to high income individuals. Overall, being female, older, poor or middle income, and having less education were associated with an increased likelihood of reporting one or more acute symptom. Addition of health related behaviors to the model eliminated the effect for education and reduced the effect of all other socio-demographic variables except for the Black effect. The effect for Blacks increased by 38% from the previous model suggesting that when racial differences in health related behaviors and characteristics are controlled for, Blacks actually have a significantly lower experience of acute symptoms than Whites. Of the health related behaviors, being overweight, having a history of smoking (verse no smoking history), not engaging in regular physical activity, having blood pressure checked within past year, and not wearing car seat belts were associated with an increased likelihood of reporting one or more acute symptoms. Addition of health insurance to the final model further increased the Black effect; Blacks reported 20% fewer symptoms than Whites while Hispanics reported 24% fewer
symptoms. Compared to being privately insured, possession of public health insurance increased the odds of reporting acute symptoms by 44%. Individuals possessing public insurance were also significantly more likely to report acute symptoms compared to uninsured individuals (AB=.36-. 14. t=2.72). Although the effect of income was reduced with the addition of health insurance, being poor was still significantly associated with increased odds of reporting acute symptom (odds ratio: 1.39). Chronic Symptoms Scale
On the Chronic Symptoms scale, Hispanics were 3 1% less likely to report
experiencing any chronic symptoms than Whites. Hispanics were also significantly less likely to report chronic symptoms than Blacks (AB=-.36-.06, t=-5.3 8). There were no initial differences between Whites and Blacks. As with the Acute Symptoms scale, addition of socio-demographic variables to the model significantly increased the effect of being Black (P3=-. 10) with Blacks being less likely to report chronic symptoms compared to Whites. Males were 42% less likely to report chronic symptoms than females. In addition, there was a significant effect for education as well as income. Individuals with less than a high school education were 26% more likely to report chronic symptoms compared those with more than a high school education and poor individuals were 59% more likely to report chronic symptoms compared to high income. Poor individuals were also significantly more likely to report chronic symptoms compared to middle income individuals (AB=.46-. 17, t=5.58). Interestingly, individuals who were married were 23% more likely to report chronic symptoms.
Adding health related behaviors to the third model increased the Black effect to significance but slightly reduced the effect of being Hispanic compared to White. The effect of age was eliminated from the model. With health related behaviors included in the model, males were 39% less likely to report chronic symptoms while married individuals were 17% more likely to report chronic symptoms. Having a history of smoking, not engaging in regular physical activity, having blood pressure checked within past year, and not wearing seat belts were all associated with increased odds of reporting chronic symptoms.
Similar to the Acute Symptoms scale, addition of insurance coverage to the model further increased the Black effect. Thus, while initially similar in their report of chronic symptoms, when the effects of socio-demographics, health related behaviors, and insurance coverage are controlled for, Blacks report significantly less chronic symptoms than Whites. Being publicly insured increased the odds of reporting chronic symptoms by 27%. There was no difference between being privately insured and being uninsured in the reporting of chronic symptoms. However, publicly insured individuals were more likely to report chronic symptoms compared to the uninsured (AB=.24-.03, t-=2.66). Medical Conditions Scale
Similar to the Acute Symptoms and Chronic Symptoms scale, compared to Whites, Hispanics were significantly less likely to report experiencing any medical conditions (odds:.52). Blacks were not different from Whites in their reporting of medical conditions but were significantly more likely to report medical conditions compared to Hispanics (AB=.06-(-.64), t=8.75). Adding socio-demographic variables to
the model increased the odds of reporting medical conditions to significance for Blacks (odds ratio: 1.06 in step I to 1.28 in step 2, compared to Whites). In addition, while Hispanics were 48% less likely to report medical conditions in step 1, addition of sociodemographics reduced that effect by )0%. Older age was significantly associated with a higher likelihood of reporting medical conditions. Income had a very strong effect on the model; Poor individuals were approximately 70% more likely to report experiencing one or more medical condition compared to individuals with high incomes. Poor individual were also significantly more likely to report medical conditions than middle income individuals (AB=.52-.07, t=7.50). There was no difference between individuals with high and middle incomes.
For the third step in the model, health related behaviors were added and resulted in the elimination of the Black effect and a reduction in the Hispanic effect as well as the age and poor income effect, although these remained significant. Among the health related variables, being overweight increased the odds of reporting medical conditions by 147%! Having had blood pressure checked within past year was also significantly associated with probability of reporting medical conditions (odds ratio:3.05). Having a smoking history and not engaging in regular physical activity also significantly increased the odds of reporting one or more medical conditions. The final step added insurance to the model predicting medical conditions. While there was no difference between being privately insured and being uninsured, possessing public insurance was associated with a 35% increase in the odds of reported medical conditions compared to being privately insured. Individuals possessing public insurance were also more likely to report medical conditions compared to uninsured individuals (AB=.30-.02, t--2.98).
Table 9. Summary of Hierarchical Regression Analysis for Variables Predicting Overall Health Rating
Variables B SE B p T Value F Value R2
Step 1 42.7*** .007
Intercept 2.07 .01 .00 241.2***
Black .17 .02 .08 9.29***
Hispanic .01 .03 .00 .44
Step 2 398.2*** .22
Intercept 1.09 .03 .00 42.3***
Black .11 .02 .06 6.63***
Hispanic .02 .02 .01 .65
Male -.07 .02 -.04 -4.19***
Age .01 .00 .32 38.00***
Married .03 .02 .02 2.27
Poor .39 .02 .19 18.28***
Middle Income .17 .02 .11 11.04***
Step 3 312.55*** .28
Intercept 1.10 .04 .00 24.72***
Black .08 .02 .04 4.81***
Hispanic .02 .03 .01 .81
Male -.02 .02 -.01 -1.27
Age .01 .00 .25 27.98***
Married .01 .02 .00 .71
Poor .34 .02 .17 16.23***
Middle Income .16 .02 .09 10.05***
Body Mass Index .00 .00 .05 6.29***
Smoking Index .00 .00 .08 9.73***
Physical Activity -.34 .01 -.20 -24.49***
Check Blood Pressure .17 .02 .09 10.59***
Wear Seat Belt -.07 .01 -.04 -5.14**
Step 4 285.56*** .29
Intercept 1.08 .04 .00 24.3***
Black .06 .02 .03 3.41**
Hispanic -.00 .03 -.00 -.09
Male -.01 .02 -.01 -.82
Age .01 .00 .25 27.60***
Married .03 .02 .01 1.64
Poor .26 .02 .13 11.51***
Middle Income .14 .02 .09 8.97***
Body Mass Index .01 .00 .05 5.87***
Smoking Index .00 .00 .08 9.36***
Physical Activity -.33 .01 -.20 -23.80***
Check Blood Pressure .17 .02 .09 10.74***
Wear Seat Belt -.06 .01 -.04 -4.64***
Uninsured .09 .02 .04 4.31***
Public Insurance .26 .02 .12 11.73***
*p<.01, **p<.001,***p<.0001, R2=Adjusted R-Square Race, Education, Income, and Insurance coverage are dummy coded (l=variable, 0-=no variable). Gender(Male=1, Female=O). Age, Body Mass Index, Smoking Index are treated as continuous variables. Physical Exercise, Check Blood Pressure, and Wear Seat belt are codes as l=yes, 0-no. See Table 21 for addition details on variable codes.
Table 10. Summary of Hierarchical Regression Analysis for Variables Predicting Role Functioning
Variables B SE B 3 T Value F Value R2
Step I 19.8*** .003
Intercept .37 .01 .00 49.72***
Black .04 .02 .02 2.47*
Hispanic -.,13 .02 -.05 -5.37***
Step 2 362.9*** .21
Intercept -.46 .02 .00 -20.34***
Black .00 .02 .00 .16
Hispanic -.08 .02 -.03 -3.94***
Male -.01 .01 -.01 -.99
Age .01 .00 .36 42.41***
Married -.04 .01 -.03 -2.69*
Poor .38 .02 .21 20.29***
Middle Income .11 .01 .08 8.23***
Step 3 267.5*** .25
Intercept -.45 .04 .00 -1 1.45***
Black -.02 .02 -.02 -1.76
Hispanic -.07 .02 -.02 -2.91*
Male .03 .01 .02 1.98
Age .01 .00 .30 35.75***
Married -.05 .01 -.04 -3.67**
Poor .36 .02 .20 18.98***
Middle Income .11 .01 .07 7.67***
Body Mass Index .00 .00 .02 2.72*
Smoking Index .00 .00 .05 5.72***
Physical Activity -.24 .01 -.17 -19.49***
Check Blood Pressure .18 .01 .10 12.36***
Wear Seat Belt -.00 .01 -.00 -.35
Step 4 360.9*** .27
Intercept -.45 .04 .00 -11.37***
Black -.06 .02 -.03 -3.92***
Hispanic -.09 .02 -.03 4.06***
Male .04 .01 .03 2.86*
Age .01 .00 .29 31.67***
Married -.03 .01 -.02 -2.43*
Poor .26 .02 .14 13.10***
Middle Income .09 .01 .06 6.56***
Body Mass Index .00 .00 .02 1.98
Smoking Index .00 .00 .04 5.23***
Physical Activity -.22 .01 -.16 -18.41***
Check Blood Pressure .17 .01 .10 12.27***
Wear Seat Belt .00 .01 .00 .29
Uninsured .04 .02 .02 2.37*
Public Insurance .35 .02 .17 17.84***
*p<.01, **p<.001,***p<.0001, R2=Adjusted R-Square Race, Education, Income, and Insurance coverage are dummy coded (1=variable, 0=no variable). Male (yes=l, no=0). Married (yes=i, no=0) Age, Body Mass Index, Smoking Index are treated as continuous variables. Physical Exercise, Check Blood Pressure, and Wear Seat belt are coded as I=yes, 0=no. See Table 21 for addition details on variable codes.
Table 11. Summary of Hierarchical Regression Analysis for Variables Predicting Physical Functioning
Variables B SE B 3 T Value F Value R2
Step I 31.2*** .004
Intercept 1.16 .02 .00 65.90***
Black .09 .04 .02 2.31
Hispanic -.40 .06 -.06 -7.14***
Step 2 627.45*** .32
Intercept -1.16 .05 .00 -23.43***
Black .01 .03 .00 .34
Hispanic -.23 .05 -.04 -4.79***
Male -.21 .03 -.06 -6.88***
Age .04 .00 .46 58.37***
Married -.08 .03 -.02 -2.58*
Poor .82 .04 .19 20.05***
Middle Income .26 .03 .08 8.75***
Step 3 462.43*** .36
Intercept -1.41 .09 .00 -16.54***
Black -.09 .03 -.02 -2.70*
Hispanic -.20 .04 -.03 -4.19***
Male -.10 .03 -.03 -3.35***
Age .04 .00 .39 46.49***
Married -.13 .03 -.04 -4.17**
Poor .77 .04 .18 18.84***
Middle Income .25 .03 .07 8.34***
Body Mass Index .02 .00 .07 8.94***
Smoking Index .00 .00 .06 7.61***
Physical Activity -.63 .03 -.19 -24.10***
Check Blood Pressure .41 .03 .10 13.11**
Wear Seat Belt -.02 .03 -.01 -.85
Step 4 432.3** .38
Intercept -1.41 .09 .00 -16.46***
Black -.16 .03 -.04 -4.72***
Hispanic -.26 .05 -.04 -5.28***
Male -.08 .03 -.02 -2.59*
Age .03 .00 .39 45.25***
Married -.09 .03 -.03 -3.00*
Poor .57 .04 .14 13.24***
Middle Income .22 .03 .07 7.29***
Body Mass Index .02 .00 .06 8.32***
Smoking Index .00 .00 .06 7.16***
Physical Activity -.61 .03 -.18 -23.12***
Check Blood Pressure .40 .03 .09 13.03**
Wear Seat Belt -.00 .03 -.00 -.25
Uninsured .09 .04 .02 2.23
Public .71 .04 .14 16.69***
*p<.01, **p<.001,***p<.000l, R2=Adjusted R-Square. Race, Education, Income, and Insurance coverage are dummy coded (l=variable, 0=no variable). Male (yes=l, no=0). Married (yes=l, no=0) Age, Body Mass Index, Smoking Index are treated as continuous variables. Physical Exercise, Check Blood Pressure, and Wear Seat belt are coded as l=yes, 0=no. See Table 21 for addition details on variable codes.
Table 12. Logistic Regression Models Predicting Acute Symptoms
Variable B SE B Chi-Square Odds Ratio 95%CI
Step 1 21.54***
Intercept -.02 .02 .50
Black .00 .04 .00 1.00 (.91-1.09)
Hispanic -.31 .07 20.84*** .73 (.64-.84)
Step 2 625.19***
Intercept -.80 .07 119.34***
Black -.13 .05 6.65*** .88 (.79-.97)
Hispanic -.31 .07 18.94*** .73 (.64-.84)
Male -.48 .05 101.46*** .62 (.56-.68)
Age .01 .00 189.08***
Married .09 .04 4.45 1.10 (1.00-1.21)
Poor .50 .06 68.98*** 1.67 (1.47-1.87)
Middle Income .23 .04 25.76*** 1.25 (1.14-1.37)
Step 3 880.78***
Intercept -.79 .09 70.98
Black -.18 .05 12.39*** .83 (.75-.92)
Hispanic -.23 .08 10.05** .79 (.68-.91)
Male -.41 .05 67.78*** .66 (.60-.73)
Age .01 .00 96.79***
Married .04 .05 .82 1.05 (.95-1.15)
Poor .45 .06 50.75*** 1.57 (1.38-1.78)
Middle Income .20 .05 19.09*** 1.22 (1.12-1.34)
Overweight .29 .05 36.86*** 1.34 (1.22-1.47)
Ever Smoked .32 .04 62.56*** 1.37 (1.27-1.49)
Physical Activity -.40 .04 98.82*** .67 (.62-.72)
Check Blood Pressure .41 .05 72.54*** 1.51 (1.38-1.66)
Wear Seat Belt -.19 .04 23.34*** .82 (1.38-1.66)
Step 4 910.79
Intercept -.84 .09 75.26
Black -.22 .05 17.16*** .80 (.73-.90)
Hispanic -.27 .08 12.87** .76 (.66-.88)
Male -.40 .05 64.49*** .67 (.60-.74)
Age .01 .00 93.39***
Not married .06 .05 1.83 1.07 (.97-1.77)
Poor .33 .07 24.31*** 1.39 (1.22-1.59)
Middle Income .18 .05 14.59*** 1.19 (1.09-1.31)
Overweight .28 .05 34.67*** 1.33 (1.21-1.46)
Ever Smoked .31 .04 58.54*** 1.36 (1.26-1.47)
Physical Activity -.39 .04 91.35*** .68 (.63-.74)
Check Blood Pressure .42 .05 73.30*** 1.52 (1.38-1.67)
Wear Seat Belt -.18 .04 20.92*** .83 (.77-.90)
Uninsured .14 .06 5.03 1.16 (1.02-1.31)
Public .36 .07 28.83*** 1.44 (1.26-1.64)
*p<.01, **p<.001,***p<.0001, df =13, Percent concordant pairs for full model=65.7%, c=.66 Race, Education, Income, and Insurance coverage are dummy coded (l=variable, 0=no variable). Gender(Male=l, Female=0). Age, Body Mass Index, Smoking Index are treated as continuous variables. Physical Exercise, Check Blood Pressure, and Wear Seat belt are codes as 1=yes, 0=no.
Table 13. Logistic Regression Models Predicting Chronic Symptoms
Variable B SE B Chi-Square Odds Ratio 95%CI
Step 1 32.56***
Intercept -.02 .02 1.38
Black .06 .05 1.75 1.06 (.97-1.16)
Hispanic -.36 .07 27.69*** .69 (.61-.79)
Step 2 379.59***
Intercept -.17 .07 5.35***
Black -.10 .05 4.44*** .90 (.82-.99)
Hispanic -.46 .07 41.18*** .63 (.55-.73)
Male -.55 .05 138.05*** .58 (.53-.63)
Age .00 .00 6.87***
Married .21 .05 19.88 1.23 (1.12-1.35)
Poor .46 .06 58.68*** 1.59 (1.41-1.78)
Middle Income .17 .04 14.89*** 1.19 (1.09-1.29)
Step 3 529.53***
Intercept -.24 .09 6.58***
Black -.14 .05 7.69*** .87 (.78-.96)
Hispanic -.42 .07 32.07** .65 (.57-.76)
Male -.49 .05 98.19*** .61 (.56-.66)
Age .00 .00 .07***
Married .16 .05 10.73 1.17 (1.07-1.29)
Poor .46 .06 50.69*** 1.56 (1.38-1.77)
Middle Income .15 .05 11.56*** 1.17 (1.07-1.28)
Overweight .14 .05 8.39*** 1.15 (1.05-1.26)
Ever Smoked .19 .04 23.08*** 1.21 (1.12-1.31)
Physical Activity -.27 .04 44.51*** .77 (.71-.83)
Check Blood Pressure .43 .05 78.73*** 1.53 (1.39-1.68)
Wear Seat Belt -.09 .04 6.17*** .91 (.84-.98)
Step 4 542.73***
Intercept -.25 .09 6.78
Black -.17 .05 10.23*** .85 (.77-.94)
Hispanic -.44 .08 34.39** .64 (.55-.75)
Male -.48 .05 94.89*** .62 (.56-.68)
Age .00 .00 .21**
Married .17 .05 12.49 1.19 (1.08-1.31)
Poor .38 .07 32.52*** 1.46 (1.28-1.67)
Middle Income .14 .05 9.82*** 1.15 (1.06-1.26)
Overweight .13 .05 7.61*** 1.14 (1.04-1.25)
Ever Smoked .18 .04 21.76*** 1.20 (1.11-1.29)
Physical Activity -.26 .04 40.73*** .77 (.72-.84)
Check Blood Pressure .42 .05 77.34*** 1.53 (1.39-1.69)
Wear Seat Belt -.09 .04 5.48*** .91 (.84-.99)
Uninsured .03 .06 .26 1.03 (.91-1.17)
Public .24 .07 12.99** 1.27 (1.11-1.44)
*p<.01, **p<.001,***p<.0001. df =13. Percent concordant pairs for full model=65.7%/, c=.66. Race, Education, Income, and Insurance coverage are dummy coded (I-=variable, 0=no variable). Male (yes=, no=0). Married (yes= i, no=0) Age, Body Mass Index, Smoking Index are treated as continuous variables. Physical Exercise, Check Blood Pressure, and Wear Seat belt are coded as l =yes, 0=no. See Table 21 for addition details on variable codes.
Table 14. Logistic Regression Models Predicting Medical Conditions
Variable B S E B Chi-Square Odds Ratio 95%/oCI
Step 1 90.49***
Intercept -.18 .02 71.56***
Black .06 .05 1.57 1.06 (.97-1.16)
Hispanic -.64 .07 79.04*** .52 (.45-.60)
Step 2 3469.71***
Intercept -3.59 .09 1579.05***
Black .24 .06 19.40*** 1.28 (1.15-1.42)
Hispanic -.35 .08 17.07*** .70 (.60-83)
Male -.18 .06 10.80** .83 (..75-.93)
Age .06 .00 2203.88***
Married .10 .05 3.41 1.10 (.99-1.23)
Poor .52 .07 56.75*** 1.69 (1.47-1.93)
Middle Income .07 .05 2.17 1.07 (.98-1.19)
Step 3 4000.82***
Intercept -4.58 .12 1399.49***
Black .13 .06 4.51 1.14 (1.01-1.28)
Hispanic -.33 .09 12.68** .72 (.60-.86)
Male -.05 .06 .74 .95 (.84-1.07)
Age .06 .00 1779.26***
Married .01 .06 .04 1.01 (.90-1.14)
Poor .50 .07 45.30*** 1.65 (1.43-1.91)
Middle Income .08 .05 2.42 1.09 (.97-1.21)
Overweight .90 .05 273.05*** 2.47 (2.22-2.75)
Ever Smoked .27 .05 32.48*** 1.31 (1.19-1.44)
Physical Activity -.28 .05 35.23*** .76 (.69-.83)
Check Blood Pressure 1.12 .06 338.12*** 3.06 (2.71-3.45)
Wear Seat Belt .00 .05 .00 1.00 (.91-1.09)
Step 4 4016.57***
Intercept -4.59 .12 1363.82***
Black .10 .06 2.85 1.11 (.98-1.24)
Hispanic -.34 .09 14.04** .71 (.59-.84)
Male -.04 .06 .49 .96 (..85-1.08)
Age .06 .00 1696.77***
Married .03 .06 .25 1.03 (.92-1.15)
Poor .41 .08 27.32*** 1.52 (1.29-1.78)
Middle Income .07 .05 1.66 1.07 (.96-1.19)
Overweight .89 .05 267.94*** 2.45 (2.19-2.73)
Ever Smoked .26 .05 30.57*** 1.29 (1.18-1.43)
Physical Activity -.26 .05 31.47*** .77 (.70-.84)
Check Blood Pressure 1.12 .06 333.02*** 3.05 (2.71-3.44)
Wear Seat Belt .00 .05 .03 1.01 (..92-1.10)
Uninsured .02 .08 .05 1.02 (.87-1.19)
Public .30 .08 14.94*** 1.35 (1.16-1.58)
*p<.01, **p<.001,***p<.0001, df =13, Percent concordant pairs for full model=65.7%, c=.66. Race, Education, Income, and Insurance coverage are dummy coded (l=variable, 0=no variable). Male (yes=l, no=0). Married (yes=l, no=0) Age. Body Mass Index, Smoking Index are treated as continuous variables. Physical Exercise, Check Blood Pressure, and Wear Seat belt are coded as I =yes, 0=no. See Appendix A for addition details on variable codes.
Overall, being White, older, poor, weighing and smoking more, having had blood pressure checked in past year, and being publicly insured were associated with increased odds of reporting medical conditions.
Multiple and logistic regression analysis was used to examine the relationship
between race/ethnicity and health insurance coverage as well as health related behaviors while controlling for socio-demographic variables. Insurance Coverage
Table 15 presents secondary logistic regression analysis looking at the
relationship between insurance coverage and demographic variables as depicted in Figure
2. Race/ethnicity was entered in the first step of the model predicting private insurance coverage. Blacks and Hispanics were significantly less likely to possess private insurance compared to Whites (odds ratio: .3 3 and .3 1, respectively). Addition of sociodemographic variables reduced the Black and White effect, although they remained significant in the model (odds ratio:.57 and .45, respectively). Overall, possession of private insurance was associated with being white, older, married, having greater than a high school education, and having high income.
For the model predicting public insurance coverage, Blacks (266%) and Hispanics (124%) were significantly more likely to possess public insurance compared to Whites. Addition of socio-demographic variables to the model reduced the effect for both Blacks and Whites although they remained significant in the model (odds ratio: 2.48 and 1.95, respectively). Possession of public insurance was significantly associated with being
Black or Hispanic, female, older, not married, having a high school or less than high school education, and being poor or having middle income.
Hispanics were 192% more likely to be uninsured compared to Whites. Blacks were also significantly more likely to be uninsured compared to Whites (odds ratio: 1.56). Addition of socio-demographic variables to the model eliminated the effect for Blacks and reduced the effect for Hispanics (odds ratio: 99 and 1.78, respectively). Being uninsured was significantly associated with being male, not married, having a high school or less than high school education, and having lower income. Body Mass Index
Table 16 presents summary of regression analysis for Body Mass Index.
Race/ethnicity was entered in the first step of the model and accounted for I% of the variance. Blacks and Hispanics were significantly more likely to report a higher body mass compared to Whites. Blacks also had a significantly higher body mass index compared to Hispanics (AB=1I.49-.76, t=7. 14). Addition of socio-demographic variables to the model increased the amount of variance explained by 2% and slightly increased the both the Hispanic and Black effect. Being older, married, and having a lower education were associated with an increased body mass index. Interestingly, neither gender nor income had significant effects on the model. Smoking Index
Table 17 presents regression results for Smoking Index. In the first step of the
model with race/ethnicity alone. Blacks and Hispanics reported smoking significantly less cigarettes compared to Whites. Controlling for socio-demographic variables in the
second step of the model reduced the effects of being Black by 27% and Hispanic by 18%, suggesting that at least some of the differences across racial/ethnic groups in smoking behavior may be mediated by differences in across groups in various sociodemographic characteristics. There was a significant effect for both gender and age with males and older individuals smoking more than females and younger individuals. In addition, individuals with less education (less than high school: P=.07, high school: P=.08) were significantly more likely to have a higher smoking index compared to individuals more education (greater than high school). There was no significant effects for marital status and income.
Results from logistic model predicting physical activity are presented in Table 18. For the first step, Blacks and Hispanics were significantly less likely to engage in regular physical activity (odds ratio: .57 and .69, respectively) compared to Whites. Sociodemographic variables were added in the second step of the model and resulted in a slight reduction in the Black effect, although it remained highly significant. Males were approximately 56% more likely to engage in regular physical activity than females. Likewise, younger individuals were more likely to engage in physical activity than older individuals. While there was no significant difference between individuals with a high school education and those with more than a high school education, individual with less than a high school education were almost 40% less likely to engage in regular physical exercise than those with more than a high school education. Compared to high income
individuals, poor individuals were 35% less likely to exercise while middle income individuals were 13% less likely to exercise. Blood Pressure Check
Table 19 presents results for logistic model predicting Blood Pressure Check. For the first step of the model, Hispanics were 45% less likely to have had their blood pressure checked within the past year compared to Whites. Interestingly, there was no significant difference between Whites and Blacks, The significant effect found for Hispanics might be due to that fact that Hispanics were overall the healthiest group in the sample and therefore would be less likely to have visited a health professional and have blood pressure checked. On the other hand, Hispanics in the sample had the largest proportion of uninsured individuals.
In Step 2 of the model, socio-demographic variables were added and resulted in a reduction in the Hispanic effect by 34%, although it remained highly significant. Males were 53% less likely than females to have had their blood pressure checked while married individuals were 28% more likely to have had their blood pressure checked. Compared to individuals with high incomes, poor and middle income individuals were 29% and 19% less likely to have their blood pressure checked, respectively. Wear Seat-Belts
Table 20 presents results for logistic regression predicting the use of Seat-belts. In the first step with race/ethnicity alone, Blacks were 32% less likely than Whites to wear their seat-belts. There were no differences between Hispanics and White. Addition of socio-demographic variables to second step in the model reduced the Black effect by
54%, although it remained significant in the model. There was a significant effect for both gender and age with males being 28% less likely to wear seat-belts than females and younger individuals also being less likely to wear seat-belts. Interestingly, individuals with less than a high school education were 64% less likely to wear seat belts compared to individuals with more than a high school education. Likewise, compared to those with post secondary education, high school graduates were 48% less likely to wear seat belts. Poor individuals were 34% less likely to wear seat-belt compared to high income individuals while middle income individuals were 19% less likely to wear seat-belts compared to high income individuals.
Table 15. Summary of Hierarchical Regression Analysis for Variables Predicting Health Insurance Coverage
Variables B SE B Chi-square Odds Ratio 95%CI
Intercept 1.41 .03 2786
Black -1.11 .05 504.7 .33 (.30-.36)
Hispanic -1.16 .07 276.8 .31 (.27-.36)
Intercept 1.90 .11 321.5
Black -.57 .06 95.7 .57 (.51-.64)
Hispanic -.79 .09 94.8 .45 (.37-.53)
Male -.05 .06 .76 .95 (.84-1.07)
Age .02 .00 203.3
Married .58 .06 93.7 1.79 (1.59-2.02)
Poor -2.75 .08 1049.7 .06 (.05-.08)
Middle Income -1.19 .08 250.8 .30 (.26-.35)
Intercept -2.28 .04 3894.6
Black 1.29 .06 486.2 3.66 (3.26-4.11)
Hispanic .80 .09 79.8 2.24 (1.87-2.67)
Intercept -4.58 .16 827.2
Black .91 .07 174.6 2.48 (2.17-2.84)
Hispanic .67 .10 41.71 1.95 (1.59-2.39)
Male -.36 .07 21.33 .70 (.60-.81)
Age .02 .00 101.14
Married -.58 .08 48.5 .56 (.48-.66)
Poor 2.45 .12 388.5 11.57 (9.07-14.76)
Middle Income 1.17 .12 93.75 3.22 (2.53-4.08)
Intercept -2.15 .03 3844.5
Black .45 .07 45.2 1.56 (1.37-1.78)
Hispanic 1.07 .08 172.9 2.92 (2.49-3.42)
Intercept -1.37 .13 112.7
Black -.01 .07 .02 .99 (.86-1.14)
Hispanic .58 .09 41.25 1.78 (1.49-2.13)
Male .44 .07 37.6 1.56 (1.35-1.79)
Age -.05 .00 646.28
Married -.28 .07 15.35 .76 (.66-87)
Poor 2.03 .11 371.9 7.59 (6.18-9.33)
Middle Income 1.12 .09 142.52 3.06 (2.54-3.67)
*p<.01, **p<.00I,***p<.00I, R2=Adjusted R-Square Race, Education, and Income are dummy coded (I-variable, 0=no variable). Male (yes= 1, no-0). Insurance coverage private (yes=l, no=0), public (yes=l, no=o), uninsured (yes=l, no=0)
Table 16. Summary of Hierarchical Regression Analysis for Variables Predicting Body Mass Index
Variables B SE B (3 T Value F Value R2
Step 1 83.25*** .01
Intercept 24.84 .05 .00 463.43***
Black 1.49 .12 .12 12.64***
Hispanic .74 .17 .04 4.33***
Step 2 39.38*** .03
Intercept 23.13 .18 .00 128.54***
Black 1.67 .12 .13 13.67***
Hispanic .79 .18 .04 4.50***
Male .24 .12 .02 2.10
Age .02 .00 .07 6.94***
Married .69 .11 .07 5.99***
Poor .03 .14 .00 .25
Middle Income -.19 .11 -.02 -1.83
*p<.01, **p<.001,***p<.0001, R2=Adjusted R-Square Race, Education, and Income are dummy coded (1 =variable, 0=no variable). Male (yes= 1, no=0), Marital status (1- =married, 0=not married) Body Mass Index Continuous variable
Table 17. Summary of Hierarchical Regression Analysis for Variables Predicting Smoking Index
Variables B SE B [3 T Value F Value R2
Step 1 142.36*** .02
Intercept 106699 1685.49 .00 63.30***
Black -52167 3720.79 -.13 -14.02***
Hispanic -61301 5442.74 -.10 -11.26***
Step 2 124.20*** .09
Intercept -22279 5512.60 .00 -4.04***
Black -38130 3741.03 -.08 -10.19***
Hispanic -50295 5390.38 -.08 -9.33***
Male 48874 3560.96 .15 13.73***
Age 1587.94 80.17 .18 19.81***
Married 2624.42 3516.86 .01 .75
Poor -2371.29 4571.72 -.01 -.52
Middle Income -2795.37 3338.08 -.01 -.84
*p<.01, **p<.001,***p<.0001, R2=Adjusted R-Square Race, Education, and Income dummy coded : (1 =variable, 0-=no variable) Male (yes= 1, no=0)
Marital status (1=married, 0=not married) Smoking Index Continuous variable
Table 18. Logistic Regression Models Predicting Physical Activity
Variable B SE B Chi-Square Odds Ratio 95%CI
Step 1 154.5***
Intercept .16 .02 56.38***
Black -.55 .05 138.66*** .57 (.53-.63)
Hispanic -.36 .07 28.41*** .69 (.61-79)
Step 2 1264.6***
Intercept 1.23 .08 266.86***
Black -.44 .05 77.17*** .64 (.58-.71)
Hispanic -.39 .07 30.43*** .67 (.58-77)
Male .51 .05 110.98*** 1.66 (1.51-1.83)
Age -.03 .00 380.25***
Married -.01 .05 .03 .99 (.90-1.09)
Poor -.42 .06 45.60*** .66 (.58-.74)
Middle -.15 .05 10.29*** .87 (.79-.94)
Race, Education, and Income dummy coded: (1=variable, 0=no variable). Male (yes= 1, no=0)
Marital status (1- =married, 0=not married) Physical Activity (yes= 1, no=0)
Table 19. Logistic Regression Models Predicting Blood Pressure Check
Variable B S E B Chi-Square Odds Ratio 95%CI
Step 1 67.84***
Intercept 1.17 .03 2208.39***
Black -.04 .05 .04 .96 (.86-1.07)
Hispanic -.60 .07 71.09*** .55 (.86-1.06)
Step 2 581.66***
Intercept .73 .08 73.02***
Black .02 .06 .21 1.02 (.92-1.14)
Hispanic -.40 .07 28.85*** .67 (.57-.77)
Male -.75 .06 184.02*** .47 (.42-.53)
Age .02 .00 263.11 ***
Married .30 .05 33.18*** 1.35 (1.22-1.50)
Poor -.18 .07 6.88* .82 (.72-.96)
Middle -.11 .05 4.32*** .89 (.81-.99)
Race, Education, and Income are dummy coded: (l=variable, 0=no variable) Male (yes= 1, no=0)
Marital status (1-=married, 0-=not married) Blood Pressure Check (yes= 1, no=0)
Table 20. Logistic Regression Models Predicting Wearing Seat-belt
Variable B S E B Chi-Square Odds Ratio 95%CI
Step 1 65.49***
Intercept .06 .02 9.13***
Black -.37 .05 64.77*** .68 (.62-.75)
Hispanic -.10 .07 2.24 .90 (.79-1.02)
Step 2 709.39***
Intercept .30 .07 18.58***
Black -.17 .05 11.92*** .84 (.76-.93)
Hispanic .20 .07 8.19* 1.22 (1.07-1.41)
Male -.33 .05 47.83*** .72 (.65-.79)
Age .00 .00 83.08**
Married -.13 .04 9.97* .88 (.82-.95)
Poor -.42 .06 47.24*** .66 (.59-.74)
Middle -.20 .04 21.59*** .81 (.74-.89)
Race, Education, and are dummy coded: (l=variable, 0-=no variable) Male (yes=1, no=0)
Marital status (1 =married, 0-=not married) Wear Seat Belt (yes=l, no=0)
Table 21. Summary of Variables
Independent Variables Measurement
Black I =yes, O=no
White I =yes, 0=no
Hispanic I =yes, 0=no
Gender I =Male, O=Female
Age Actual age continuous
Marital Status 1 =married, 0=not married
High I =yes, O=no
Middle I =yes, 0=no
Low I =yes, 0=no
Health Related Behaviors
Body Mass Index Range (8-114) continuous
Smoking Index Range (0-1569500) continuous
Overweight I =yes, 0=no
Ever Smoked I =yes, O=no
Physical Exercise I =yes, O=no
Blood Pressure Check I =yes, O=no
Wear Seat Belt I =yes, 0=no
Private I =yes, O=no
Public I =yes, O=no
Uninsured I =yes, 0=no
Dependent Variables Measurement
Overall Health Rating l=Excellent
Role Functioning Range (0-2) Lower score poorer functioning
Physical Functioning Range (0-5) Lower score poorer functioning
Acute Symptoms Scale 1=1 or more symptoms, O=no symptoms
Chronic Symptoms Scale 1=1 or more symptoms, O=no symptoms
Medical Conditions Scale != or more symptoms, 0=no symptoms
Epidemiological and health care research have long established the differential health status and increased mortality rates among minority groups in this country. However, the exact determinants of these differences remain unclear. Most of the research exploring this issue has focused primarily on the influence of SES as an explanatory variable. Numerous studies have confirmed the relationship between lower income levels and increased mortality and morbidity (i.e., Sorlie et al., 1992; Marmot & Smith, 1997; House et al., 1990; Ettner, 1996). However, as noted earlier, racial effects often persist even after controlling for SES. In addition, using such global indices as SES to explain differential health outcome across racial/ethnic groups may mask more direct and specific causes. Also, from a public policy perspective, while the elimination of poverty and economic disadvantage in this country presents an altruistic yet improbable task, the delineation of specific environmental and cultural factors that may contribute to the differential health outcome across racial and ethnic groups, over and above the effects of SES, may prove both tangible and worthwhile. For example, certain lifestyle behaviors that affect health (smoking, excessive alcohol consumption, being overweight) may reflect cultural styles and dietary patterns as well as socio-economic status (LillieBlanton et al., 1993). Research in the area of health, and particularly minority health, will
need to focus on identifying the potential markers underlying racial differences in health outcome that extend beyond SES and those that are policy malleable. As Williams (1997) notes "when researchers identify social status differences in the distribution of a disease they should initiate a detailed examination of the contribution of environmental and genetic factors to observed differences. A broad range of factors intervene between race and health. These intervening mechanisms include health behavior; stress in family, occupational, and residential environments; social ties; psychological factors, including personality characteristics; culture; religious beliefs and behavior; and medical care."
This present study sought to examine the influence of some of these intervening mechanisms in hopes of ftirther enhancing what we currently know about the relationship between health status and race. Specifically, this study investigated the influence of socio-demographic variables, various health related behaviors, and health insurance coverage on health outcome across three racial/ethnic groups.
Overall, the demographic characteristics of the present sample approximate
findings from previous studies regarding differences in income and insurance coverage across racial/ethnic groups (Nickens, 1991; National Center for Health Statistics, 1994; Lillie-Blanton et al., 1993; Flack et al.,1995). Blacks were more likely to be poor and publicly insured while Whites were more likely to have high incomes and be privately insured. Hispanics had the highest rate of no insurance compared to Blacks or Whites. The effect of income on perceived health status found in this study is also consistent with earlier studies that reported individuals with higher SES to have better ratings of perceived health (i.e., Ettner, 1996; Lillie-Blanton et al., 1993; Ziff et al., 1995).
Relationship Between Race/Ethnicity and Health Related Behaviors.
Descriptive analysis revealed varied results with regard to the relationship
between race/ethnicity and health related behavior variables. For example, while Whites were more likely to engage in regular physical activity, they were also more likely to report having smoked and reported the highest number of cigarettes smoked a in lifetime. Likewise, while Blacks were more likely to report wearing seatbelts regularly, they were also more likely to weigh more and least likely to engage in physical activity. Results related to weight and physical exercise are relatively consistent with previous studies in this area. Both Black and Hispanic women have been found to have higher percentages of being overweight and to be less likely to report engaging in regular physical exercise compared to White women (CDC, 1 994a; Durazo-Arvizu et al., 1997; Health People 2000, 199 1; Lillie-Blanton et al., 1993; Kumanyika, 1993; Myers et al., 1995).
As the model in Figure 2 depicts, socio-demographic variables may mediate the relationship between race/ethnicity and health related behaviors. The significant differences found across the three racial/ethnic groups on various socio-demographic dimensions provide some rationale for this contention. However, multivariate analysis found that Blacks and Hispanics were still significantly more likely to weigh more and be less physically active even after controlling for gender, age, marital status, education, and income level. Likewise, Whites remained significantly more likely to have smoked more cigarettes than Blacks or Hispanics after controlling socio-demographic variables. Previous research examining smoking prevalence across racial/ethnic group have found similar low rates of smoking among Hispanics (U.S. Department of Health and Human
Services, 1991). However, contrary to present results, Blacks have been found to have a higher prevalence of smoking compared to Whites (Myers et al., 1995; U.S. Department of Health and Human Services, 1991). Interestingly, while income did not emerge as a significant predictor of increased weight or smoking, having a high school education or less was significantly associated with increased weight and smoking compared to having more than a high school education. This finding has important policy implications in terms of which population may benefit more from health promotion campaigns aimed at reducing smoking behavior.
Relationship Between Race/Ethnicily and Insurance Coverage
Hispanics in this sample were more likely to be uninsured compared to Whites. This relationship persisted even after controlling for socio-demographic variables in multivariate analysis. This pattern is consistent with the literature on racial/ethnic differences in health insurance coverage (Nickens, 1991; Cornelius, 1993; Valdez et al, 1993; Johnson et al., 1995). However, this finding must be considered cautiously due to possible subgroup differences in insurance coverage among Hispanics. For example, using the 1989 Current Population Survey, Trevino et al. (1992) found 16% of Puerto Ricans and 20% of Cuban Americans were uninsured compared to 37% of Mexican Americans. Cuban Americans were more likely to be privately insured while Puerto Ricans were more likely to be publicly insured. Amey et al. (1995) found similar patterns of insurance coverage for Puerto Rican, Cuban American, and Black families. However, they found Mexican Americans to be significantly less likely to have their entire family insured and more likely to have none of their family members insured. The greater
proportion of uninsured Mexican Americans compared to other Hispanic subgroups is presumably due to the fact that they are more likely to be undocumented and seasonal workers and therefore less likely to possess public or employer provided private insurance. In addition, Puerto Ricans are more likely to have households headed by single women and thus more likely to be covered by categorical public insurance programs such as Medicaid (Vega &Amaro, 1994).
Consistent with previous research, Blacks in this sample were more likely to be publicly insured compared to Whites. Inclusion of insurance to the model resulted in improved self-reported health status for Blacks on all dependent measures except Rate Health. Interestingly, possession of public insurance was significantly associated with poorer health status on all outcome measures included in this analysis. For this study, all types of public insurance were combined (low-income publicly insured (i.e., AFDC), Medicare, and those who qualify for Medicaid due to poor health). As such, the categorical nature of the program may have self-selected individuals with poor health. Using the same database, Short and Lair (1994) examined the health status of adults across five different insurance coverage options : privately insured with employmentrelated insurance; privately insured with non-group insurance; individuals who qualified for public insurance due to poor health; individuals who qualified for public insurance due of low income, and the uninsured. They found that individuals who were enrolled in the employer-sponsored plans were the healthiest, followed by individuals with private non-group insurance, the uninsured, the low-income publicly insured, and the publicly insured by virtue of poor health, respectively. Thus, while analyzing particular insurance groups separately allows for greater specificity, overall, the publicly insured (both low-
income and poor health eligible) appear to have the worst health status. It may be that public insurance represents a sort of catch net when all other options have been exhausted. For example, individuals whose health care costs may have exceeded their private insurance benefits or who have preexisting conditions that limit participation in private insurance programs may have to rely on public insurance when personal resources have been depleted. Likewise, individuals who are uninsured report overall better health status than those publicly insured. Given their health status, the uninsured may not see the value of purchasing health insurance. However, once faced with a medical illness and having depleted personal resources, these individuals ultimately may fall back on public insurance for coverage.
Relationship Between Race/Ethnicity and Dependent Variables
As noted earlier, the primary relationship of interest in this study is the effect of race/ethnicity on overall health status. On the outcome variables of Health Rating, Role Functioning, and Physical Functioning, Blacks reported the worst overall ftmctioning while Hispanics reported the best. In addition, Hispanics acknowledged fewer Acute Symptoms, Chronic Symptoms, and Medical Conditions. Interestingly, Blacks and Whites were relatively similar in their report of symptoms on these three scales.
One of the more fascinating findings of this study relates to the health status of
Blacks. Descriptive analysis of the health status variables for Blacks indicate poorer selfratings of overall health, physical functioning, and role functioning compared to both Whites and Hispanics. However, controlling for socio-demographic variables, health related behaviors, and insurance coverage reversed the Black effect incrementally to the
point where Blacks actually reported better role and physical functioning than Whites. Thus, as hypothesized in proposed model, the indirect effect of these variables served to "distort" the true relationship between race/ethnicity and health status (i.e., converted a negative relationship into a positive relationship)(Rosenberg, 1968). To further illustrate this distorter effect, the relationship between race/ethnicity, insurance coverage, and the outcome variable Role Functioning will be briefly explored. Observing the relationship between race and Role Functioning alone, significantly more Blacks report that health limits their role functioning compared to Whites. However, when this relationship is broken down by insurance coverage a different relationship emerges. Specifically, significantly less Blacks report limitations in Role Functioning compared to Whites for those possessing private insurance (15% verse 22%); there is no significant difference for those possessing public insurance (5 1% verses 5 3%); and slightly more Whites report limits in Role Functioning compared to Blacks for those who are uninsured (15% verse 20%). However, since a larger proportion of Whites possess private insurance (8 0%) compared to Blacks (57%), their higher endorsement of role functioning limitation artificially elavates the sum total level of role functioning limitation. Once insurance coveraged is controlled for, the true relationship between race/ethnicity and role functioning emerges.
Similarly, while there were no significant differences between Blacks and Whites on the measures of acute symptoms, chronic symptoms, and medical conditions, controlling for socio-demographic variables, health related behaviors, and insurance coverage resulted in Blacks reporting significantly less symptoms on all three scales compared to Whites. In this case, indirect model pathways of socio-demographic
variables, health related behaviors, and insurance coverage served to "suppress" or neutralize the true relationship between the independent and dependent variables (Rosenberg, 1968). For example, looking solely at the relationship between race and acute symptoms, an equal percentage of Blacks and Whites report experiencing acute symptoms (49.64% verses 49.62%, respectively). However, when broken down across each educational level, Whites consistently report more acute symptoms than Blacks. Analysis of frequency distributions indicate that individuals with less than a high school education are more likely to report acute symptoms and Blacks are more likely to have less than a high school education. Thus, the greater proportion of Blacks with less than a high school education suppresses the true relationship between race and the report of acute symptoms. Once education is controlled for (along with other control variables demonstrating similar relationships), Blacks are actually less likely to report acute symptoms than Whites. The Rate Health scale was the only outcome measure where the effect of being Black persisted after controlling for all other variables.
These findings provide evidence supporting the model proposed in Figure 2.
Socio-demographic variables accounted for the largest amount of variance in the model. For all six outcome variables, income and age (except for Chronic Symptom scale in the case of age) had the largest effect of socio-demographic variables. However, the addition of health related behaviors and insurance coverage to the models resulted in a reduction in the negative effect of being Black or an increase in the positive effect of being Black. This suggests that, as hypothesized, differences in health related behaviors and insurance coverage do in fact influence differential health outcome.