Abstract
The present study examined how self-rated health was influenced by sociodemographic characteristics, physical health indicators, and sociocultural resources among four racial/ethnic groups of older adults. The data source was the Survey of Older Floridians, a statewide sample of Whites (n = 503), African Americans (n = 360), Cubans (n = 328), and non-Cuban Hispanics (n = 241) who were age 65 and older. Hierarchical regression models of self-rated health were estimated to explore the direct effects of the predictor variables as well as their interactive roles in each racial/ethnic group. Compared to Whites, racial/ethnic minority older adults rated their health more poorly. Although physical health indicators were significant predictors of self-rated health across all groups, the authors found group-specific predictors and interactions. Findings show similarities and differences in predictors of self-rated health across diverse racial/ethnic groups and suggest the importance of understanding group-specific factors in efforts to improve older adults’ perceived and actual health.
The single self-rated health question “How would you rate your overall health?” has been widely used in health outcome studies, and its significance in predicting general health and well-being and even mortality has been well documented (Idler and Benyamini 1997; Lee et al. 2007; Read and Gorman 2006; Ware 2004). Given the difficulty of obtaining objective data on health, self-reported health is often used to compare the health status across different racial/ethnic groups. The evidence suggests that racial/ethnic disparities exist, with African Americans and Hispanics being more likely to report poorer health than any other groups even after adjusting sociodemographic characteristics and the disparities remaining throughout late adulthood (August and Sorkin 2010; Borrell and Dallo 2008; Cummings and Jackson 2008). As yet, however, little is known about factors that affect within-group variations in self-rated health. The present study explored within-group and cross-group factors predicting self-reported health, based on older adult samples of Whites, African Americans, Cubans, and non-Cuban Hispanics.
Health research has shown that self-rated health is an important construct with compelling evidence that perceptions of health predict health outcomes as well as or better than objective measures of physical health including diagnoses (Idler and Benyamini 1997; Jang et al. 2009; Wang and Satariano 2007). According to holistic theories of health, health includes not only the objective form of health indicators such as physical symptoms but also the individual’s phenomenological view of his or her general state of health, including quality of life (Nordenfelt 2007). Study of self-rated health is thus in line with the premise that health is not a value-free concept whose measurement rests on the presence or absence of disease.
Self-rated health has been suggested to be linked to various individual-level factors, such as sociodemographic characteristics, physical health indicators, and social and cultural characteristics. Of sociodemographic characteristics, older age (Amstadter et al. 2010; Liang et al. 2010), female gender (Cummings and Jackson 2008; Read and Gorman 2006), and low socioeconomic status (Gallo, Smith, and Cox 2006; Kawachi, Kennedy, and Glass 1999) have been associated with poor ratings of health. Studies also report racial/ethnic differences in the relationship between sociodemographic characteristics and self-rated health. Over the past three decades, for example, the gender gap in health assessments has decreased significantly (Schnittker 2007), but minority women—especially African American women—tend to have unfavorable health ratings (Cummings and Jackson 2008; Read and Gorman 2006). Individual socioeconomic indicators such as educational attainment and economic resources have proved to be robust determinants of perceived health across race/ethnicity (Cummings and Jackson 2008), at least in part because these resources facilitate access to and use of health care and preventive services (Sambamoorthi and McAlpine 2003). Racial/ethnic groups may have different opportunities for attaining such resources, and there may be racial/ethnic variations in the extent to which the attained sociodemographic characteristics exert an impact on self-rated health.
Chronic health conditions and functional disabilities are strong predictors of poor self-rated health in all populations (August and Sorkin 2010; Salomon et al. 2009; Seeman et al. 2010). Among disadvantaged populations, however, the impact of physical health strains may be greater. In part because of persistent social and economic disadvantages and limited access to health care over the course of life, minority older adults may experience a heightened burden of physical health challenges (Louie and Ward 2011; Mezuk et al. 2010; Schoeni et al. 2005). Although disparities in health, compared to Whites, appear to apply to older members of many minority populations (August and Sorkin 2010; Dunlop et al. 2007), the majority of studies have focused on disparities between Whites and African Americans or Blacks (Cummings and Jackson 2008; Kelley-Moore and Ferraro 2004; Lee et al. 2007).
Expanding health comparisons beyond those involving White–Black groups provides an opportunity to understand the broader implications for health of factors such as sociocultural resources including social support, religious attendance, and English proficiency. These are examples of nonhealth factors whose influence on perceived as well as actual health may vary widely across diverse populations (Dickinson et al. 2011; Koenig, McCullough, and Larson 2001; Krause 1987; Krause and Bastida 2011). Studies have found that various cultural groups have a different repertoire of social resources, which can affect experiences with physical health strains differentially (Jang et al. 2003; Krause 2006; Litwin 2006; Umberson and Montez 2010). For example, among Hispanics—a group that forms part of the focus of the present study—limited English proficiency has been considered as a major stressor that poses physical and mental health risks (Finch and Vega 2003; Kandula, Lauderdale, and Baker 2007; Masel, Howrey, and Peek 2011), but its impact on self-ratings of health in Hispanic older adults and within subgroups has been understudied (Jerant, Arellanes, and Franks 2008).
It is important to note that Hispanics consist of individuals of diverse background in nativity and language proficiency, and treating Hispanic populations as a single entity may lead to an overgeneralization of findings (Jerant et al. 2008). The present study included Cubans, who have been considered as distinct from other Hispanic populations in several ways. As the fourth largest Hispanic population in the United States, Cuban Americans in general are older, are more often married, and have higher levels of education and income compared to other Hispanic groups (Pew Hispanic Center 2011). In 2009, approximately 18% of Cuban Americans were aged 65 years or older, compared to about 13% of the U.S. general population and 5% of other Hispanic groups. Cubans are also geographically concentrated; more than two-thirds live in Florida (Pew Hispanic Center 2011). One unique aspect of the present study is its capability of comparing Cuban older adults to non-Cuban Hispanics as well as to non-Hispanic Whites and African Americans.
Purpose of the Study
As noted, the goal of the study was to identify factors that predict self-rated health among four racial/ethnic groups of older adults (Whites, African Americans, Cubans, non-Cuban Hispanics). Although self-reported health has been widely examined across racial/ethnic groups, the majority of studies focus on Blacks and Whites or treat Hispanics as one homogeneous group. Little is known about the within-group variations in self-rated health and its predictors. To fill the gap, we explored both direct and interactive roles of predictive variables that included sociodemographics (age, gender, education, marital status, and income), physical health indicators (chronic conditions and functional disability), and sociocultural resources (social support, religious attendance, and English proficiency) within racial/ethnic group. The approach was designed to help identify group-specific risk and protective factors that could guide the development of targeted interventions to enhance older adults’ health status.
Method
Participants
Data for the present study were derived from the Survey of Older Floridians, which conducted a series of computer-assisted telephone interviews in 2004 and 2005. Two different sampling methods were developed to recruit participants aged 65 years and older. The first sampling method was used to collect a statewide representative sample of 437 older adults using random-digit dialing; the sample consisted of 382 Whites, 37 African Americans, 15 Cuban Americans, and 3 non-Cuban Hispanics. All older adults who were members of one of the four racial/ethnic groups had an equal chance of being selected.
The second sampling method was designed to recruit the racial/ethnic minority groups of interest. Random-digit dialing was used with telephone exchanges with high proportions of the racial/ethnic minority groups. This sampling yielded 996 participants composing 122 Whites, 323 African Americans, 313 Cubans, and 238 non-Cuban Hispanics. Combing the two samples, the total sample included 1,433 participants: 504 Whites, 360 African Americans, 328 Cuban Americans, and 241 non-Cuban Hispanics.
In addition to being a member of one of the study populations, to be included for the study participants had to be 65 years or older. When more than one individual in the household was age 65 or older, the interviewer asked to speak to the oldest member. In addition, individuals had to score 5 or higher on the Short Portable Mental Status Questionnaire (Pfeiffer 1975), a level indicative of only moderate cognitive impairment. To calculate response rates, we used the Response Rate 3 (RR3) formula created by the American Association for Public Opinion Research Standards (2000). RR3 contrasts the total number of persons who agree to be interviewed with an estimate for the number of eligible respondents. Our approach to calculating the number of eligible respondents included refusals and no contacts as well as those whose physical or mental capacity precluded participation. Using this calculation, the response rate for the statewide sample was 62%; RR3 for the White subsample was 61%, and it was 55% to 57% for the remaining groups.
Measures
Dependent variable
Participants were asked how they would rate their health, and their responses were coded as poor (1), fair (2), good (3), or excellent (4). Self-rated health has been shown to be a valid and reliable measure of general health in numerous population-based studies (August and Sorkin 2010; Cummings and Jackson 2008; Idler and Benyamini 1997; Ware 2004).
Sociodemographics
Sociodemographic characteristics included age (in years), gender (0 = male, 1 = female), education (0 = high school or less, 1 = beyond high school), marital status (0 = not married, 1 = married), and annual household income (0 = less than $30,000, 1 = beyond $30,000).
Physical health indicators
Chronic health conditions were measured with a checklist that asked participants whether they had ever experienced or been diagnosed with nine specific conditions or diseases: heart attack, stroke, high blood pressure, Parkinson’s disease, respiratory problems, cancer, diabetes, osteoporosis, and arthritis. A total count of the reported diseases and conditions was used in the analyses. Functional disability was measured with a composite score of six items of activities of daily living (Katz 1983) and three items of instrumental activities of daily living (Lawton and Brody 1969). Using a yes–no response format, participants were asked whether they needed help with each of the nine activities. The potential range of total scores was 0 (functional independence on all items) to 9 (functional dependence on all items). Cronbach’s alpha was .78.
Sociocultural resources
Social support was measured with a question, “In times of trouble, can you count on at least some of your family or friends?” with three response options: (1) hardly ever, (2) some of the time, and (3) most of the time. A single-item measure of religious attendance asked how often the individual attends mass or religious services (1 = never or almost never, 5 = more than once a week). For those who chose Spanish for their interviews, English proficiency was measured with a single item asking how well the respondent spoke English (1 = not at all, 4 = very well). Those who chose English for their language of interview were given the highest score. In the present article, this measure was included only in analyses of the Hispanic subgroups.
Data Analysis
Descriptive statistics were used to describe the sample. The sample characteristics were then compared across four racial/ethnic groups using an analysis of variance, t test for English proficiency, and chi-square tests. Hierarchical multiple regressions were used to examine the direct and interaction effects of the independent variables on self-rated health within each racial/ethnic group. Each block of variables (sociodemographics, physical health indicators, sociocultural resources, and an interaction term) was entered in sequential order, and the model’s R2 change was calculated for the blocks of the variables. Each interaction term was entered into the direct effect model separately and its statistical significance was evaluated.
Results
Sample Characteristics
Table 1 shows the descriptive information of the four racial/ethnic groups. For all participants, the average age was 73.73 (SD = 6.74). Approximately 65% were female (n = 934), 45% had beyond a high school education (n = 639), 42% were married (n = 596), and 26% had annual incomes greater than $30,000 (n = 309). Compared to Whites, the three other groups tended to be younger and had lower levels of education and income. African Americans were most likely to be female and least likely to be married or to have income greater than $30,000. Cubans were most likely to be married.
Sample Characteristics.
p < .01. ***p < .001.
There was no difference in the number of chronic conditions across groups, but minority groups were all significantly more likely to experience functional disabilities. Minority groups reported less social support but were equally or more likely to attend religious services compared to Whites. Non-Cuban Hispanics had a higher level of English proficiency than Cubans. Last, minority older adults rated their health more negatively than did Whites, with Cubans being the most negative in their health rating.
Regression Models of Self-Rated Health
The results of hierarchical regression analyses are presented in Table 2. The sociodemographic variables explained 4% to 10% of the total variance in the four racial/ethnic groups. Older age was associated with more positive ratings of health for Cubans, female non-Cuban Hispanics had significantly lower self-rated health than their male counterparts, and higher education was associated with higher self-rated health among Whites. Except for African Americans, higher income predicted more positive ratings of health among other racial/ethnic groups.
Regression Models of Self-Rated Health.
p < .05. **p < .01. ***p < .001.
It is not surprising that health-related variables added a substantial amount (16% to 22%) of the explained variance to the model. Both chronic conditions and functional disability were significantly associated with self-rated health for all groups. Older adults who had more chronic conditions and more functional disabilities rated their health poorer. The third block of variables, sociocultural resources, explained an additional 1% to 6% of the variance. Given the literature suggesting its importance for several racial/ethnic groups, it was surprising that religious attendance was associated with self-rated health only for Whites. Among both Cuban and non-Cuban Hispanics, higher English proficiency predicted positive ratings of health.
When the interaction terms were included in the final model, each of the interaction terms added 1% to 2% of the explained variance to the model. Statistical significance was found for age × religious attendance and income × religious attendance in Whites; age × functional difficulty and female × functional difficulty for African Americans; age × chronic conditions, education × functional difficulty, and English proficiency × functional difficulty for Cubans; and female × English proficiency and English proficiency × chronic conditions for non-Cuban Hispanics.
Interpretation of Interaction Effects
Following the guidelines suggested by Aiken and West (1991), we examined the interaction effects, and further analyses were conducted for each of the significant interaction effects. Table 3 presents the results of correlational analyses between predictors and self-rated health stratified by each of the moderating variables. In the White sample, the positive effect of religious attendance on self-rated health was particularly high among those who were older and of lower levels of income. In the African American sample, the inverse relationship between functional disability and self-rated health was stronger for individuals with advanced age than their younger counterparts. Also, the association between functional disability and self-rated health was significant among females but not among males. Among Cubans, older age seemed to reduce the effects of chronic conditions on self-rated health. The significant inverse relationships between functional difficulty and self-rated health were observed for Cubans with less than high school education and those with lower English proficiency. In the non-Cuban Hispanic sample, older age lessened the effects of chronic conditions on self-rated health similar to Cubans, and the inverse relationship between chronic conditions and self-rated health was stronger for those with less English proficiency than for their counterparts. In addition, the correlation between English proficiency and self-rated health was stronger for men than for women.
Correlational Analysis of Interaction Effects.
Age was split into a younger group (50.4%) and an older group (49.6%) based on the median score.
English proficiency was divided into a lower English proficiency group (speaking English not at all or not too well; 25%) and a higher English proficiency group (speaking English pretty well or very well; 75%).
p < .05. **p < .01. ***p < .001.
Discussion
Although racial/ethnic disparities have been reported in physical health, little is known about group-specific predictors of self-rated health in diverse populations. Using samples representative of community-dwelling older adults in Florida, the present study explored how sociodemographics, physical health indicators, and sociocultural resources predict self-rated health for Whites, African Americans, Cubans, and non-Cuban Hispanics. In addition, the study examined the interactive effects of the predictors to identify group-specific risk-reducing or health-enhancing factors. Findings show similarities and differences in predictors of self-rated health across diverse racial/ethnic groups and suggest the importance of understanding group-specific factors in efforts to develop health intervention strategies tailored to each group.
Consistent with previous research, minority older adults in this study had poorer ratings of health and were of lower socioeconomic status and had greater functional difficulties than their White counterparts (Cummings and Jackson 2008; Louie and Ward 2011; Schoeni et al. 2005; Yao and Robert 2008). These differences reflect important issues facing our health care system. Despite efforts to reduce health disparities, a recent report by the Agency for Healthcare Research and Quality (2011) suggests that disparities in health care quality among minority and low-income people persist across all states in the United States. As scholars report, sociodemographic disadvantages of minority populations may persist into late adulthood, and these disparities negatively affect their physical health and disability (Cummings and Jackson 2008; Louie and Ward 2011; Read and Gorman 2006; Schoeni et al. 2005).
Greater functional difficulties and a greater number of chronic health problems were associated with lower ratings of health across all groups in the direct effects model. However, the health variables remained significant only for minority older adults when interactions were considered. Notably, the relationship between functional difficulty and self-rating of health was strong for African Americans, especially for older African American women, in line with previous research reporting poor physical health of African American women (August and Sorkin 2010; Cummings and Jackson 2008; Louie and Ward 2011; Read and Gorman 2006; Warner and Brown 2011). Future studies should further examine the burden of functional disability among African American elders, and efforts should be made to improve their physical function.
The interaction effects further underscore the importance of examining within-group variations. For example, we found that positive effects of religious involvement were most salient for older and poor Whites. Although salutary effects of religious attendance on physical health have been widely reported (Krause 2002; Krause and Bastida 2011; Musick 1996; Reyes-Ortiz et al. 2007), the literature on racial differences in religiousness is somewhat mixed. Generally, minority elders, African Americans in particular, tend to be highly involved with religious activities and enjoy health benefits from the strong ties with the religious community (Krause 2002, 2006); yet because of the generally higher levels of religious involvement among minority elders, the beneficial effect of religious involvement could be more salient for Whites (Ellison 1995). The latter argument reflects our findings, and future studies should further explore the moderating role of religious involvement on physical health across diverse racial/ethnic groups.
One of the strengths of the present study is that the sample included two groups of Hispanics: Cubans and non-Cuban Hispanics. Inclusion of these two groups allowed us to consider differences across subpopulations of what is often treated as a homogeneous group of Hispanics and to underscore the importance of explaining heterogeneity among Hispanics. Results related to English proficiency highlight the subgroup differences. The findings that Cubans have lower English proficiency but that English proficiency was a significant predictor only for non-Cuban Hispanics might be the result of the geographic concentration of Cuban Americans (Jang et al. 2009). Since Cuban Americans are likely to reside in Cuban enclaves, such as Little Havana in Miami, English-language proficiency may not be as significant as it is to other Hispanics who are dispersed geographically. Such geographic concentration may delay English acquisition of Cuban elders, as evidenced in other studies on immigrants (Chiswick and Miller 2001; Hochhausen, Perry, and Le 2010).
It is worth noting the differences between Cubans and non-Cuban Hispanics in interpreting interaction effects. These differences stressed the importance of within-group variations that have been reported among Hispanic populations (Borrell and Dallo 2008). First, the effect of chronic conditions on self-rated health was lessened with advancing age in both groups. Second, English proficiency reduced the effect of functional difficulty and chronic conditions on self-rated health among Cubans and non-Cuban Hispanics, respectively. Third, for Cubans only, individuals with a higher level of education handled functional difficulties better than their less educated counterparts. These findings support the importance of acquiring English proficiency to lessen the burden of poor health conditions (Dunlop et al. 2007; Masel et al. 2011). Although it is not clear why English proficiency moderates different health variables (functional difficulty and chronic conditions for Cubans and non-Cuban Hispanics, respectively), this finding confirms that functional disability and chronic conditions tap different perceptions of health (Kendig, Browning, and Young 2000).
Thus, the different main effects and interaction effects across groups highlight the importance of examining intergroup differences from a holistic perspective that integrates objective and subjective information. Although objective indicators of health such as chronic conditions and functional disability were associated with their perceived health across all groups, differences in the interactive effects of health variables with socioeconomic variables and English proficiency were found across minority groups. Differences in self-rated health may reflect a cumulative disparity of opportunity and health across the life course in minority population (Kelley-Moore and Ferraro 2004), as opposed to current functioning. It is also important to consider that certain minority older adults (female, less educated, and low English proficiency) may be more disadvantaged.
The present study has several limitations. First, the study involves analyzing the cross-sectional data, which does not allow us to draw any causal inferences about the relationship between covariates and self-rated health. Second, the study sample was drawn in the state of Florida. This may limit generalizing the findings to other geographic areas. Third, although a strength of the present study is the comparison of Cuban Hispanics and other Hispanics, Hispanic populations consist of diverse groups of nativity and English-language proficiency with different cultural backgrounds (Acevedo-Garcia et al. 2010; Borrell and Dallo 2008). Research suggests that reports of health could be affected by language and cultural differences and beliefs (Bzostek, Goldman, and Pebley 2007; Jerant et al. 2008; Kandula et al. 2007; Masel et al. 2011). Future studies should include more subgroups of Hispanic populations and take into account different cultural experiences and meanings. Fourth, social resources such as social support and religiousness were measured using single items. Given the multidimensionality of sociocultural constructs, future studies need to employ validated multiple-item instruments for the assessment. Last, the present study focuses on individual-level factors predicting self-rated health. Studies have shown that health advantages or disadvantages may stem also from contextual effects of neighborhoods (Kawachi et al. 1999; Yao and Robert 2008). Future studies should investigate larger social and environmental contexts surrounding individuals.
Despite the limitations, the present study adds to the literature on specific racial/ethnic group factors contributing to predicting self-rated health. Findings of the study suggest that efforts to develop intervention programs should attend to specific needs of racial/ethnic groups. For example, language barriers may prevent navigating the health care systems for Hispanic elders with limited English proficiency, which can negatively affect their maintenance of their physical health; thus, language services in the health care systems should be provided. Research also shows that members of different racial/ethnic groups are differentially affected by risk factors (Jang et al. 2008). Efforts should be made to address group-specific risk factors and address health-enhancing strategies for each racial/ethnic group. For example, intervention on health behaviors (e.g., increasing physical activity, maintaining healthy diet) could help minority older adults reduce their disability and enhance their physical health (August and Sorkin 2010; Dunlop et al. 2007). In addition, culture-specific programs to increase functional health and decrease chronic conditions should promote physical health for minority older adults.
Footnotes
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: The project was supported by the Administration on Aging Research (Grant 90AM2750; Jennifer Salmon, principal investigator; David A. Chiriboga, co–principal investigator).