| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
| ||||||||||||||||||||||||||||||||
RESEARCH ARTICLE |
1 Indiana University Center for Aging Research, Indianapolis.
2 Regenstrief Institute, Inc., Indianapolis, Indiana.
3 Department of Medicine, Division of General Internal Medicine and Geriatrics, Indiana University School of Medicine, Indianapolis.
4 Departments of Sociology and Department of Statistics, Indiana University, Bloomington.
Address correspondence to Daniel O. Clark, PhD, Regenstrief Institute, Inc., 1050 Wishard Boulevard RG-6, Indianapolis, IN 46202. E-Mail: dclark{at}regenstrief.org
 |
Abstract |
|---|
 TOP Abstract MethodsResultsDiscussionReferences |
|---|
Methods. Data came from the 1998 and 2000 waves of the Health and Retirement Study, a nationally representative cross-sectional and prospective cohort study of community-based adults aged 65 and older. We created a physical vulnerability score from age, gender, and self-reported disability measures and measured socioeconomic status via educational attainment. Mortality data came from the National Death Index.
Results. In the 1998 cohort, high physical vulnerability was more than 3 times more prevalent in individuals with less than 12 years of education compared to those with 16 or more years of education. Although less educated older adults had a higher probability of death overall, evidence of educational differences in the mortality consequence of high physical vulnerability was limited. In 2000, 2.16 million older adults had high physical vulnerability, and more than one half (53%) of these adults had less than 12 years of education.
Discussion. In persons 65 years of age or older, educational differences are more apparent in the prevalence of physical vulnerability than in the mortality consequence of that vulnerability.
TWO recent articles have developed and validated short, simple scoring tools for identifying community-dwelling older adults at significantly elevated risk of disability and death (Carey, Walter, Lindquist, & Covinsky, 2004
; Saliba et al., 2001). These tools draw on age, gender, and self-reported health and functional status to identify persons at elevated risk. Older adults with the highest scores on these tools have 4 to 7 times greater risk of 2-year mortality (Carey et al., 2004) and functional decline (Saliba et al., 2001) than comparable persons with the lowest scores.
Saliba and colleagues (2001) characterized these risk assessment tools as indicators of medical vulnerability, but we prefer the label physical vulnerability because the tools focus on physical health as both the cause and consequence of vulnerability. The literature has defined vulnerability as a process of negative adaptation that results in a reduced capacity to cope with and recover from stressors (Kuh, Ben-Shlomo, Lynch, Hallqvist, & Power, 2003). Physical health represents just one from among many domains of possible causes and consequences of vulnerability.
It is well-established, for example, that socioeconomic status (SES) is a cause of vulnerability: Lower SES individuals suffer higher rates of disease, disability, and mortality (Antonovsky, 1967; Crimmins & Saito, 2001; Guralnik, Land, Blazer, Fillenbaum, & Branch, 1993; House, Lantz, & Herd, 2005; Li et al., 2005; Lynch, Kaplan, & Shema, 1997; Maddox & Clark, 1992); and there is evidence that health disparities associated with SES are increasing in the United States (Mensah, Mokdad, Ford, Greenlund, & Croft, 2005; Pappas, Queen, Hadden, & Fisher, 1993; Shi & Stevens, 2005).
To conceptualize the socioeconomic distribution of health, we drew on a recent model of the social basis of health disparities (Diderichsen, Evans, & Whitehead, 2001) that recognizes prevalence and consequence of illness as separate outcomes resulting from social contexts experienced over the life course. This model assumes that social contexts vary by socioeconomic position and that these varied contexts produce both differential exposure and susceptibility to health risks. The social stratification of exposure and susceptibility produces a social stratification of both prevalence and consequence of illness.
George (2005) noted that there is limited information about how SES affects illness onset versus survival, and that separating the two would do much to address the problems posed by selective survival. We do not estimate illness or disability onset in this article, but we do provide separate estimates of the prevalence and consequence of physical vulnerability. We present nationally representative estimates of the prevalence of physical vulnerability and the mortality consequence of that vulnerability by educational attainment among older adults. We used educational attainment as a marker of SES; there are a number of reasons to focus on education alone to represent SES in older adults (Cairney & Krause, 2005). Compared to education, income is more frequently missing and can misrepresent the financial situation of older adults for whom net worth may be at least as important for gauging one's ability to mobilize resources to respond to challenges (Smith, 1999). Occupation may be less relevant than education to persons older than age 65 years, as a vast majority have retired from their lifetime primary occupation and many older women were never employed in the labor force. Mirowsky and Ross (2003) viewed formal education as a fundamental determinant of health status and as a resource that accumulates and grows over the life course. Finally, education is unlikely to be influenced by middle- and later life illness, and characterization of SES by educational attainment is easy to replicate. Others have noted these observations as well (see House et al., 2005), and, for these reasons, the analyses, findings, and discussion that follow focus on SES as characterized by educational attainment.
Our a priori hypotheses were that (a) high physical vulnerability would be more prevalent among less educated older adults than among their more educated counterparts, and (b) less educated older adults would demonstrate higher consequence of physical vulnerability than their more educated counterparts in terms of 2-year mortality.
|
METHODS |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
. Using HRS-supplied sampling weights, we were able to work with a nationally representative sample of persons aged 65 years and older. We used the 10,313 respondents in the 1998 wave to replicate the Carey and colleagues (2004) work. (Note that 21 respondents had missing data, which left 10,292 for analyses.) We used 2-year follow-up data (year 2000) to estimate mortality as a consequence of high physical vulnerability. We also used the 2000 HRS data to present the estimated number of adults 65 years of age or older who were physically vulnerable and the distribution of vulnerability by education level.
Measures
Physical vulnerability
We operationalized physical vulnerability following the work of Carey and colleagues (2004), who used the 1993 Asset and Health Dynamics Among the Oldest Old (AHEAD) data with mortality follow-up to 1995. We developed the physical vulnerability score to efficiently predict mortality. Potential predictors of mortality included age, gender, and functional status measures. We assigned a score to final predictors based on coefficients from logistic regression models. Higher scores were assigned for older age, male gender, bathing and shopping dependence, and difficulty walking and pulling or pushing large objects. Based on logical breaks, we characterized scores of 0 to 2 as low vulnerability, 3 to 6 as medium, and 7 to 10 as high.
Our physical vulnerability estimates based on the 1998–2000 HRS data were very comparable to the results obtained by Carey and colleagues (2004) using the 1993–1995 AHEAD data. We identified 54%, 39%, and 7% of respondents to be low, medium, and high in physical vulnerability, respectively (compared to 44%, 43%, and 12%, respectively, in Carey et al.). We found that 3%, 11%, and 34% of participants in the low, medium, and high physical vulnerability categories, respectively, had died by 2-year follow-up (compared to 4%, 12%, and 34%, respectively, in Carey et al.). Finally, we compared the individual physical vulnerability item odds ratios for mortality (see Table 1). Again, the results were very consistent with those of Carey and colleagues, which we had expected given the significant overlap of AHEAD and HRS data.
|
) and education at less than 12 years (Seeman et al., 2004). We also explored the incorporation of net worth as a third criterion for SES. Findings from these more complex models did not differ substantively from those of the simpler model that used education only.
2-year mortality
Mortality data came from the National Death Index.
Analyses
First, we estimated the prevalence of high physical vulnerability by educational level using 1998 data. We then explored differential consequence in terms of 2-year mortality. We created dummy variables for the three levels of physical vulnerability and educational attainment. We used logit regression models to regress 2-year mortality on physical vulnerability, education, and the interaction of vulnerability and education. Finally, we estimated the number of older adults in 2000 that were in each physical vulnerability level by education. We incorporated the sampling design weights and stratification variables supplied by the HRS into all analyses using the survey commands in Stata Version 9.2 (StataCorp, 2005).
|
RESULTS |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
|
, noted that standard tests comparing logit coefficients across group are inappropriate because they confound the magnitude of the effect with differences in residual variance across groups. However, in our analyses we compared predicted probabilities, which are unaffected by this problem [Long, 2006].) Table 2 shows estimates from the logit model of mortality, whereas Figure 2 presents the predicted probabilities of mortality from this model. Mortality was strongly related to physical vulnerability; there was a weaker relationship between mortality and education. In the low physical vulnerability group, the risk of death decreased as education increased, F(2, 10239) = 5.04, p <.01, such that those with less than 12 years of education had more than twice (0.04) the unadjusted probability of death of the high education group (0.02). This pattern was not apparent in the medium vulnerability group, F(2, 10239) = 1.20, p =.30, or the high vulnerability group, F(2, 10239) = 1.06, p =.34, with no significant difference in mortality by education.
|
|
|
DISCUSSION |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
). Although physical vulnerability is clearly a strong predictor of mortality, our findings suggest that it may be important for interventions to address multiple domains of vulnerability. This is because people who are most physically vulnerable are often also vulnerable in terms of SES, and socioeconomic vulnerability predicts outcomes beyond what can be expected from physical vulnerability alone. In the nationally representative data used in this study, more than 1 of every 2 (53%) older adults with high physical vulnerability were also socioeconomically vulnerable in terms of low education. Although we focused on education in this article, it is not likely to be the only resource limitation of consequence in the physically vulnerable population. Material resources vary dramatically by education level. The 2000 HRS data showed that the median income for individuals with less than 12 years of education was $16,700 versus $54,500 for those with 16 or more years of education. Similarly, median net worth was $10,500 versus $259,500 for the less and more educated groups, respectively.
Our first hypothesis was supported. High physical vulnerability was 3 times more prevalent in the lower compared to higher education group. Our second hypothesis was not supported. Although we did find a main effect for education on mortality, and an interaction between physical vulnerability and education was significant, greater consequence of high physical vulnerability was not apparent for the less educated population. Rather, greater consequence of education was apparent for persons with low physical vulnerability.
We used a large, nationally representative survey weighted for sampling design, but there are still many limitations to our study. First, despite the large sample size, the number of individuals in some categories was small. For example, there were 64 individuals with both high physical vulnerability and high education. Second, we were unable to include any community-level measures of socioeconomic factors. A number of studies have shown that community-level measures operate independently of individual-level measures of SES to influence health disparities (Adler & Newman, 2002; Seeman & Crimmins, 2001; Taylor, Lerner, Sage, Lehman, & Seeman, 2004). Third, age and gender, which we included in the physical vulnerability score, were ascribed statuses with a mix of social, biological, and historical implications. As a result, there was some confounding of education and the physical vulnerability scale. Older generations achieved less education, and older age gave a high physical vulnerability score, but these may have been offset by male gender, which gave a higher physical vulnerability score and was associated with higher education. Fourth, we elected to present mortality simply as a dichotomous event. We also ran survival analyses incorporating time to death, but these did not lead to conclusions any different from those presented here (data not shown). Fifth, data (except for mortality) were based on self-report. The physical vulnerability score would have been affected if individuals of different educational levels had interpreted and reported function differently. However, our education-stratified analyses of outcomes did not show large differences in persons with high physical vulnerability, which suggests the predictive validity of the physical vulnerability score was similar across education levels. Sixth, because education alone proved to be a valid indicator of disparities and was less likely to be an outcome of physical vulnerability, we focused our analyses on SES as measured by education. We explored incorporating income and net worth and using continuous measures, but this did not alter the overall findings shown here. Finally, because we did not have incident data in this study, we did not know how long respondents had been physically vulnerable. We plan further research on prevalence and consequence using younger age groups and longer follow-up, and estimating the timing of onset and consequences of physical vulnerability.
A large literature explores the causes of SES disparities. We focused here on describing the disparities. Continued work to address the causes of disparities is clearly needed. The model of the social basis of health disparities noted in the introduction to this article suggests that opportunities for intervention exist at several different levels. For example, our preliminary analyses suggest that addressing SES disparities through a focus on prevalence may be the best use of limited resources. This will require a greater effort to target and tailor interventions geared to the prevention of physical vulnerability among persons who have limited education and material resources. Of course we are not the first to find support for such a focus. Murray, Kulkarni, and Ezzati (2005) showed that SES disparities in mortality are greatest from early to middle adulthood and that the deaths are largely due to chronic illnesses. Similarly, House and colleagues (2005) showed that disparities in health and disability are greatest in middle age.
As Alwin and Wray (2005) and George (2005) have noted, social context varies for individuals over time, and individual susceptibility will depend in part on personal life course trajectories, the historical period, and the timing of exposures relative to one's life stage. Identifying early, middle, and lifelong causes of health disparities will continue to be important. Alwin and Wray also noted that greater effort to translate such information into interventions is clearly needed. We would add that society needs interventions designed for and tailored to the socioeconomically vulnerable. With increased attention to issues of health literacy (Pignone, DeWalt, Sheridan, Berkman, & Lohr, 2005) and federal funding for research on health disparities, there is a promising movement in the direction of developing and testing interventions targeted to less educated, lower SES populations.
|
Acknowledgments |
|---|
|
Footnotes |
|---|
Received for publication February 1, 2006. Accepted for publication November 3, 2006.
|
References |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
This article has been cited by other articles:
 |
|
 |
|
  D. O. Clark, R. M. Frankel, D. L. Morgan, G. Ricketts, M. J. Bair, K. A. Nyland, and C. M. Callahan The Meaning and Significance of Self-Management Among Socioeconomically Vulnerable Older Adults J. Gerontol. B. Psychol. Sci. Soc. Sci., September 1, 2008; 63(5): S312 - S319. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||
| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
