Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Self-Rated Health Trajectories among Married Americans: Do Disparities Persist over 20 Years?

Self-Rated Health Trajectories among Married Americans: Do Disparities Persist over 20 Years? Hindawi Journal of Aging Research Volume 2018, Article ID 1208598, 8 pages https://doi.org/10.1155/2018/1208598 Research Article Self-Rated Health Trajectories among Married Americans: Do Disparities Persist over 20 Years? 1 2 Terceira A. Berdahl and Julia McQuillan Division of Research and Modeling, Center for Financing, Access and Cost Trends, Agency for Healthcare Research and Quality, 5600 Fishers Lane, Rockville, MD 20852, USA University of Nebraska-Lincoln, 709 Oldfather Hall, Lincoln, NE 68588-0324, USA Correspondence should be addressed to Terceira A. Berdahl; terceira.berdahl@ahrq.hhs.gov Received 5 June 2017; Revised 27 October 2017; Accepted 22 November 2017; Published 11 January 2018 Academic Editor: F. R. Ferraro Copyright © 2018 Terceira A. Berdahl and Julia McQuillan. ,is is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. ,e purpose of this study is to understand self-rated health (SRH) trajectories by social location (race/ethnicity by gender by social class) among married individuals in the United States. We estimate multilevel models of SRH using six observations from 1980 to 2000 from a nationally representative panel of married individuals initially aged 25–55 (Marital Instability Over the Life Course Study). Results indicate that gender, race/ethnicity, and social class are associated with initial SRH disparities. Women are less healthy than men; people of color are less healthy than whites; lower educated individuals are less healthy than higher educated individuals. Women’s health declined slower than men’s but did not differ by race/ethnicity or education. Results from complex intersectional models show that white men with any college had the highest initial SRH. Only women with any college had significantly slower declines in SRH compared to white men with any college. For married individuals of all ages, most initial SRH disparities persist over twenty years. Intersecting statuses show that education provides uneven health benefits across racial/ethnic and gender subgroups. of health inequality to acknowledge diverse outcomes of 1. Introduction individuals who occupy different positions in social struc- ,ere are well-documented disparities in health by race/ tural hierarchies [1–3]. One useful approach is the inter- ethnicity, gender, and social class in the United States. Less sectionality paradigm [4–6], which emphasizes how social is known about the combined advantages and disadvantages at structures intersect to create the axis of advantage and the intersection of race/ethnicity, gender, and social class. disadvantage in multiplicative rather than only in separate or Questions also remain regarding whether or not initial self- additive ways. For example, there is evidence that the ex- rated health disparities exacerbate or diminish over time. It is periences of women of color cannot be adequately un- important to estimate, and not assume, whether social class derstood by studying race/ethnicity or gender in isolation effects on health are similar for men and women and white and [1, 7–9]. To assess particular policies or laws, it can be ap- nonwhite Americans. To advance knowledge of patterns of propriate to study only one axis of inequality; but to gain health disparities in the United States, we estimate self-rated a more comprehensive understanding of overall changes in health (SRH) trajectories over 20 years using six waves (ob- health with aging, comparing subgroups created by assessing servations) from a panel study of married individuals who were multiple axes of stratification simultaneously (e.g., social aged 25–55 in 1980 at the time of the first interview. class, race/ethnicity, and gender) is fruitful [10–12]. Prior work on health disparities finds that indicators of social location (e.g., race, gender, and education) contribute 2. Background to health inequality [10, 13]. Men have worse mortality than 2.1. Intersectionality and Health Disparities. A growing women but spend fewer years with comorbidity and dis- number of researchers are calling for more nuanced analyses ability [10, 14]. Compared to whites, racial-ethnic minorities 2 Journal of Aging Research continue to face lower life expectancy and worse health due to the intersections of gender, race, and education? If yes, outcomes from birth until death [14]. Studies of educational do disparities persist over decades? To answer these questions, disparities find that people with lower education have worse we examined self-rated health trajectories of a random sample health and faster health declines with aging compared to of individuals who were married in 1980 and followed over people with higher education [15, 16]. Recent studies have twenty years. We restricted the sample to adults who were examined the intersections of gender and social class or initially aged 25–55 years. We categorized individuals by race/ethnicity and social class and find that these intersections gender (men or women), race/ethnicity (white or nonwhite), matter for health changes associated with aging [10]. and level of education (up to less than a high school degree, Most longitudinal studies of aging include people as a high school degree, or any level of college education), thus young as 25 and follow people for decades in small geo- creating twelve subgroups with varying advantages and graphic areas with little racial diversity. We know of no disadvantages. studies that simultaneously analyze the intersections of class, race/ethnicity, and gender for twenty years. Instead, most 4. Data and Methods studies focus on older and elderly adults. ,e current study focuses on people who were married at the start of the study. For the current study, we used data from the Marital In- It is possible that the health protective effects of marriage stability Over the Life Course (MIOLC) study, a six-wave panel study of a national sample of initially married persons in the could minimize gender, education, and race/ethnicity health disparities [17, 18]. Yet, recent research highlighting the long United States [27]. Respondents were interviewed by tele- phone in 1980, 1983, 1988, 1992, 1997, and 2000 (waves 1–6). reach of inequality from grandparents and from early life suggests that even among the married, disparities are likely Only the wife or the husband was included in the study; the participant was selected by random assignment. During the to persist with aging [19–21]. first wave of the study, the survey involved random digit dialing and screening of individuals for inclusion in the study. 2.2. Health Protective Effects of Marriage. Marital status is ,e inclusion criteria required that individuals had to be a social resource that can protect against steep health de- married and between the ages of 18 and 55. Among eligible clines associated with aging. Among people who are mar- households, the rate of completed interviews for the initial ried, the influence of depressive symptoms and chronic wave of the study was 65%, yielding a sample size of 2,034. ,e conditions on subsequent functional limitations is weaker response rates varied among subsequent waves of data col- compared to nonmarried [22]. ,e end of a marriage is also lection (in the order: 78%, 84%, 88%, 90%, and 82%). ,e associated with higher allostatic load (an indicator of bi- analytical sample for this study includes those between the ological risk for worse health) [23]. ,e physiological impact ages of 25 and 55 at the initial interview in order to ensure that of marital disruption is larger for widowhood than divorce, they were likely to have finished their education. Multilevel as indicated by comparisons of inflammatory, metabolic, models incorporate data from as many waves as participants and cardiovascular functioning compared to those who completed; as with all longitudinal studies, there was attrition remain married [23]. over time. Comprehensive information about the data set and Rates of marriage, like health disparities, vary by race how well it represents the U.S. married population of 1980 and social class. Marriage rates and duration are lower and comparisons with formerly married adults in 2000 is among people with lower education and racial/ethnic mi- provided in the methodology report [28]. norities compared to people who have higher education and who are white [24]. ,erefore, health differences by 4.1. Variables/Measures race/ethnicity and education could be smaller among the married than the nonmarried because of selection into 4.1.1. Self-Rated Health (SRH). We measure self-rated a status with privileges [25]. In 1980, the difference in ever- health, the dependent variable, with the following ques- married rates among white women (95%) and black women tion: “Now I have some questions about your health. In (87%) was smaller than that in 2012 (87% for white women general, would you say your own health is excellent (3), good and 63% for black women) [24]. Because marriage was (2), fair (1), or poor (0)?” ,is single-item measure has ubiquitous in 1980, it is unclear if there will be health a strong association with mortality, suggesting that it is disparities in this more narrow group than the general a valid indicator of health [29]. Self-rated health is measured population, yet we argue that it is important to find out, in every wave of data collection in the same way. because marriage is sometimes promoted as a way to reduce social problems, even though there is evidence that marriage does not have the same privileges for all [26]. 4.1.2. Wave/Trajectory. To measure the changes in self-rated health, we include a variable that indicates the wave of observation. ,is variable is coded as 0 for the first interview 3. Current Study in 1980 and 5 for the last interview in 2000, with consecutive In the present study, we advance understanding of health numbers in between. ,ere were about three years between disparities by answering two important questions: In the interviews. With this variable in the model, the intercept United States, are there health disparities among people who indicates the average initial self-rated health, and the co- are married that are created by disadvantaged social locations efficient for “wave” indicates the average change in self-rated Journal of Aging Research 3 Table 1: Sample descriptive statistics for MIOLC wave 1, aged health between observations, or the trajectory of change in 25–55. health. ,e variance component for this variable indicates the amount of spread around the average trajectory. Mean (%) SD Self-rated health 2.26 0.75 Age (in years) 37.24 8.33 4.1.3. Age. We included age (measured in years) as an in- dividual level characteristic to adjust for differences in initial Race health status. We mean center age so that the intercept Nonwhite 11.9 — represents participants with the average age for the sample. White 88.1 — Gender Men 41.5 — 4.1.4. Social Location. We measure social class using edu- cation, a common practice in health research [30]. We use Women 58.5 — only the initial value of education because very few people add Education years of education beyond age 25 in the 1980s. Education was <HS 10.4 — measured in years of completed formal education. We High school graduate 63.1 — converted years into three categories. Because a high school Any college 26.5 — degree is a requirement for many jobs and a college degree is N � 1,785. required for another set of jobs, we divided years into less than a high school degree (<12 years) (LTHS), high school (12 years) (HS), and any college (13+ years). ,e three-category Table 2: Self-rated health by gender, race, and education for adults measure of education helped to make meaningful subgroups aged 25–55 in 1980. for the intersectionality analysis by combining education with Mean SD N gender (what is your sex? male/female) and race/ethnicity Less than high school 2.09 0.85 88 (what race do you consider yourself to be? recoded into White High school graduate 2.30 0.72 371 white and nonwhite) to create 12 total subgroups. Any college 2.49 0.67 455 Women Less than high school 1.85 0.81 20 Nonwhite High school graduate 2.15 0.70 52 4.2. Analysis Strategy. We use multilevel models to estimate Any college 2.05 0.71 58 the trajectories of self-rated health. ,e multilevel model Less than high school 1.88 1.00 57 trajectory approach appropriately estimates the nesting of White High school graduate 2.36 0.77 181 observations within individuals and attrition [31]. All Any college 2.55 0.62 420 models are estimated with random effects for the intercept Men Less than high school 2.10 0.89 21 (initial status) and slope parameters using HLM 7.03. Ad- Nonwhite High school graduate 2.22 0.67 23 ditionally, we control for initial age because the rate of health Any college 2.31 0.73 39 decline is likely to be steeper for those closer to 55 than to 25. MIOLC data, wave 1, N � 1,785. ,e time variable is centered at the initial observation to facilitate meaningful interpretation of the constant (e.g., average initial self-rated health) [32]. and the need to rotate out the reference category using Our analysis begins with a brief discussion of the a “simple slope” approach to hypothesis testing to de- sample in 1980 (see Tables 1 and 2). After presenting these termine which groups are different from each other [33]. descriptive statistics, we move to the HLM analysis. In Yet, the conventional approach also requires solving Table 3, we present Model 1, which includes the time equations to determine the total difference of each group variable and serves as a baseline trajectory model for self- from the others based upon the main effects, two-way rated health. ,is model contains an intercept and a slope interactions, and three-way interactions. We constructed for time (year) with random effects for these parameters as indicator variables for the subgroups to provide a mathe- well as an overall Level 1 error term. Next, we present matically equivalent way to determine subgroup differences in SRH and simultaneously increase the ease of interpreting estimates from Model 2 (Table 3) which includes a variable capturing respondent age at the first year of the survey coefficients summarizing those differences [8]. We also (initial age). In Models 3–5 (Table 4), we present separate estimated supplemental analyses by rotating the reference models with gender (Model 3), race (Model 4), and edu- group to compare each subgroup (available from the author cation (Model 5) to evaluate the effects of these charac- upon request). Our approach involves multiple significance teristics on the aging trajectories without other variables in tests, which in classical regression can increase the risk of the model. ,ese models illustrate how gender, race, and Type 1 errors (finding a significant association when none education modify self-rated health over time. Finally, in exists). By using HLM with empirical Bayes estimation Model 4, we explore the joint effects of gender, race, and methods, we reduce the risk of Type 1 error [34]. ,ere are practices for post hoc adjustments to reduce the risk of education by constructing comparative categories to cap- ture the three-way interaction of these variables (Table 5). Type 1 error (e.g., Bonferroni adjustments), but there are A conventional interaction approach would also gen- arguments against this practice [35]. Because we also have erate eleven coefficients with a comparison to the intercept some groups with small sample sizes (e.g., nonwhite 4 Journal of Aging Research Table 3: Models 1-2: self-rated health by age and self-rated health aging trajectory. Model 1 Model 2 b SE p value b SE p value Initial SRH (intercept) 2.358 0.016 <0.001 2.358 0.016 <0.001 × initial age — — — −0.011 0.002 <0.001 SRH trajectory (time slope) −0.061 0.005 0.005 −0.061 0.005 <0.001 × initial age — — — −0.001 0.001 0.034 Random effects SD VC p value SD VC p value Initial SRH intercept, u 0.561 0.311 <0.001 0.551 0.301 <0.001 SRH trajectory slope, u 0.111 0.011 <0.001 0.111 0.011 <0.001 Level 1, r 0.511 0.261 — 0.511 0.261 — Note. SD � standard deviation; VC � variance component; Level 2 MIOLC N � 1,785. Table 4: Models 3–5: self-rated health trajectories by gender, race, and education. Model 3 Model 4 Model 5 b SE p value b SE p value b SE p value Initial SRH (intercept) 2.410 0.025 <0.001 2.389 0.017 <0.001 2.462 0.020 <0.001 × initial age −0.012 0.002 <0.001 −0.012 0.002 <0.001 −0.010 0.002 <0.001 Women (men � reference group) −0.085 0.033 0.009 — — — — — — Nonwhite (white � reference group) — — — −0.266 0.050 <0.001 — — — Less than HS (any college � reference group) — — — — — — −0.440 0.065 <0.001 High school — — — — — — −0.168 0.034 <0.001 SRH trajectory (time slope) −0.083 0.008 <0.001 −0.064 0.005 <0.001 −0.062 0.006 <0.001 × initial age −0.001 0.001 0.064 −0.001 0.001 0.041 −0.001 0.001 0.020 Women (men � reference group) 0.037 0.010 <0.001 — — — — — — Nonwhite (white � reference group) — — — 0.022 0.020 0.289 — — — Less than HS (any college � reference group) — — — — — — 0.021 0.024 0.392 High school — — — — — — −0.007 0.010 0.496 Random effects SD VC p value SD VC p value SD VC p value Initial SRH intercept, u 0.546 0.298 <0.001 0.541 0.293 <0.001 0.532 0.282 <0.001 SRH trajectory slope, u 0.106 0.011 <0.001 0.108 0.012 <0.001 0.108 0.012 <0.001 Level 1, r 0.508 0.258 — 0.508 0.258 — 0.508 0.258 — Note. SRH � self-rated health; VC � variance component; SD � standard deviation; Level 2 MIOLC N � 1,785. women with less than a high school degree), we also risk In the initial survey, 1,785 participants responded to the making Type 2 errors (failing to find significance when variables in the analytical sample. Average self-rated health there really is an association in the population). We ranges from a low of 1.85 among nonwhite women with less therefore use the conventional 0.05 level of significance, than a high school degree (those with the most social dis- focus on the size and meaning of differences, and use the advantages) to a high of 2.55 among white men with any significance tests as a heuristic for interpreting patterns college education (those with the most social advantages) of associations. (Table 2). For every race by the gender group, those with any college education are the largest group. Almost half of the sample consists of people who are white with any college 5. Results education. ,e smallest groups are people who are nonwhite 5.1. Sample Descriptive Characteristics. We report sample with less than a high school degree. Only nonwhite women descriptive statistics for the sample in Table 1. ,e mean age do not have an education health gradient because those with for the analytical sample was 37.24 (SD � 8.33) at the initial any college education have lower self-rated health than those interview. Just over 10 percent of the sample is nonwhite with a high school degree. For all other gender by race (12%). Over half (58%) of the sample was women. Fewer groups, higher education is associated with higher self-rated participants have less than a high school degree (10%) than health. any college education (55%). On a scale from 0 to 3, with zero representing poor health and 3 representing excellent health, average initial self-rated health was above the mid- 5.2. Aging Trajectory Findings. We first describe the baseline point (M � 2.26 and SD � 0.75). model (no covariates; not shown in the table). ,e baseline Journal of Aging Research 5 Table 5: Model 6: self-rated health trajectories by gender, race, trajectory measure has the value 0 for the first year, 1980, and and education. 5 for the twentieth year, 2000. ,erefore, the intercept for this model provides the average self-rated health in the first Model 6 ∗∗∗ year of the study (1980). Initial self-rated health is 2.361 , b SE p value slightly higher (better) than the average for all years in the Initial SRH (intercept) 2.52 0.03 <0.001 baseline model. ,e SRH trajectory coefficient indicates that, × initial age −0.01 0.00 <0.001 on average, self-rated health declines between interviews ∗∗∗ Race by gender by education (b �−0.061 ). ,e standard deviation for self-rated health × nonwhite women LTHS −0.67 0.14 <0.001 is 0.75. ,erefore, over the six waves and twenty years, on average, self-rated health declines by −0.36 (−0.06 × 6), or × nonwhite women HS −0.42 0.08 <0.001 almost half a standard deviation. ,e rate of decline in × nonwhite women any college −0.47 0.09 <0.001 self-rated health varies across individuals, indicated by × nonwhite men LTHS −0.32 0.18 0.07 the significant variance component (variance components × nonwhite men HS −0.30 0.14 0.03 ∗∗∗ p value � 0.012 ). × nonwhite men any college −0.22 0.12 0.06 To adjust for the thirty-year range of initial ages, we × white women LTHS −0.47 0.09 <0.001 included an indicator of initial age in Model 2 (Table 3). × white women HS −0.22 0.05 <0.001 Consistent with the strong association between age and ∗∗∗ health, health declines with age (b �−0.012 ), and the self- × white women any college −0.05 0.04 0.19 rated health aging trajectory is steeper for participants who × white men LTHS −0.54 0.13 <0.001 were older at the start of the study (the coefficient is −0.001 × white men HS −0.19 0.06 0.00 larger for each additional year). ,e average age in the SRH trajectory (time slope) −0.082 0.010 <0.001 analytical sample is 37.24, with a standard deviation of 8.33 × initial age −0.001 0.001 0.047 years. ,erefore, the difference between a standard deviation Race by gender by education below and above the mean age is 16.66 years translating into × nonwhite women LTHS 0.057 0.128 0.653 24% of a standard deviation in initial self-rated health × nonwhite women HS 0.042 0.035 0.226 (−0.011 ×16.66 �−0.18/.75 � 24%). Initial self-rated health is × nonwhite women any college 0.096 0.035 0.006 arranged as we would expect, with those who are 25 reporting the best health and those age 55 the worst health. × nonwhite men LTHS 0.064 0.070 0.361 Twenty years later, all groups decline, but the decline is × nonwhite men HS −0.041 0.043 0.345 steeper for those who were initially older, resulting in larger × nonwhite men any college −0.045 0.051 0.383 differences between the age groups at the end of the study. × white women LTHS 0.050 0.030 0.099 Building on the model with initial age and the self-rated × white women HS 0.022 0.014 0.105 health trajectory, we next separately explore if there are × white women any college 0.029 0.013 0.022 disparities by gender, race/ethnicity, or level of education for × white men LTHS 0.015 0.045 0.745 the initial or the aging trajectory of self-rated health (Table 4, Models 3–5). Women have lower average initial self-rated × white men HS −0.012 0.021 0.550 ∗∗ health (b �−0.085 , or 12% of a standard deviation lower), Random effects SD VC p value yet the decline in health for women is less steep than it is for Initial SRH intercept, u 0.525 0.275 <0.001 ∗∗∗ men (b � 0.037 ; therefore, the slope for women is SRH trajectory slope, u 0.106 0.011 <0.001 ∗∗∗ ∗∗∗ −0.083 + 0.037 �−0.05). Women initially have worse Level 1, r 0.508 0.258 — health, but because their health declines more slowly, over Note. ,e reference group is white men with any college; LTHS � less than time, they end up with better health than men twenty years high school; HS � high school; AC � any college; Level 2 MIOLC N � 1,785. later. ,ere is also an initial difference between nonwhite and model provides an estimate of the intercept and variance white individuals (Table 4, Model 4). ,e race/ethnicity ∗∗∗ components. ,e intercept indicates that the average self- difference is larger than the gender effect (b �−0.266 ). ∗∗∗ rated health across all participants and all waves is 2.252 , ,e declines in self-rated health are not as steep for nonwhite a value that is above the midpoint of the 0 to 3 range for the compared to white individuals (white �−0.064; ∗∗∗ variable (3 indicates excellent health). Average self-rated nonwhite �−0.064 + 0.022 �−0.040), yet this difference is health varies significantly between people (the variance not significant. component for the intercept in the baseline model is Consistent with prior research, there is a steep health ∗∗∗ 0.267 , and for within people, the Level 1 error is 0.308). gradient by education (Table 4, Model 5). Individuals ,e variance components indicate that just under half of the with any college have the highest self-rated health ∗∗∗ variance in self-rated health is between individuals (47%, or (intercept � 2.462 ). ,ose with a high school level of (0.267/(0.267 + 0.307))) and just over half is within in- education have about a fifth of a standard deviation lower ∗∗∗ dividuals over time (53%). self-rated health (b �−0.168 ), and the difference is even Model 1 in Table 3 provides an estimate of the aging greater for those with less than a high school degree ∗∗∗ trajectory in self-rated health, which is measured by the (b �−0.440 ). Although education has strong associations time variable (one unit is approximately three years). ,e with initial SRH, level of education does not modify the rate 6 Journal of Aging Research of decline, adjusted for age. ,ose who start the study in their did not observe similar effects of education, however, across older age do have slightly but significantly steeper declines in self-rated health trajectories for all racial/ethnic and gender ∗ ∗∗∗ health (age b �−0.001 + trajectory coefficient �−0.062 ; groups. For lower educated women, declines in health are therefore, for each additional year, the rate of decline is similar to higher educated white men. For higher educated −0.001 larger). women, there are slower declines in self-rated health ,e separate models of gender, race/ethnicity, and ed- compared to higher educated white men. ,us, the rate of ucation also show that none of these indicators of social decline in self-rated health is conditioned by gender and status alone explain the variance in initial self-rated health or education. Similar to Liang et al. we find racial differences in in the change in self-rated health over time, indicated by the the intercept, or average health at the beginning of the study, significant variance components in all of the models. with nonwhites scoring significantly worse on health We next examine the combined effect of all three in- measures [19]. Additionally, that study found few signifi- dicators of social location simultaneously (Table 5, Model 6). cant racial differences in aging trajectories over time, which Most of the subgroups differ in initial self-rated health is also similar to our study (although not directly com- relative to white men with any college, the reference group. parable because they did not evaluate racial effects across All groups have negative coefficients, indicating lower initial gender and education) [19]. Consistent with our study self-rated health. ,ree groups do not have significantly findings, another study of aging among African Americans lower self-rated health than white men with any college found that in older adults, self-rated health did not decline (nonwhite men with less than a high school level of edu- significantly over time [37]. cation and nonwhite men and white women with any ,e current study provides evidence that among college). ,e differences in initial self-rated health for most married individuals, there are several initial differences in groups persist with age because the rate of decline is not self-rated health, and for all but higher educated women, significantly different from the most privileged group (white those differences persist over twenty years. Consistent ∗∗∗ men with any college: b �−0.08 ). Two groups, however, with other studies, older age was associated with worse did have significantly slower declines (positive coefficients) health initially and a steeper decline in health over time. ∗∗ Not surprisingly, for all age groups, self-rated health in self-rated health with age: nonwhite women (b � 0.100 ) and white women (b � 0.029 ) with any college. declined with aging. We were surprised to find that even among married individuals, there are important dispar- In addition to the comparisons with the most privi- leged group (white men who have any college education), ities in health and persistence in disparities with declines it is useful to determine if there are significant differences in aging. ,erefore, neither selection into nor the pro- among the groups with varying levels of privilege and tective effects of marriage eliminate the effects of struc- disadvantage as indicated by gender, race/ethnicity, and tural inequality on health. education. We therefore reran the final model and rotated Data limitations require some caution in over- the comparison group until we estimated all possible generalizing these results. One limitation is that the study comparisons (detailed results available upon request). was designed to focus on married individuals. We do not Unlike the story from the separate gender, race/ethnicity, have a comparison group of unmarried individuals during the same time period. We cannot generalize and education models, all women are not different from all men. White women at any level of education do not beyond the ever-married population in 1980. ,is lim- itation is somewhat mitigated by historical context be- differ from less-educated nonwhite men (less than HS and HS). Education does not have the same strong association cause marriage was more common for whites and blacks with self-rated health among nonwhite men and women; in 1980 [24]. We expect that health status, however, there are no education differences among those who are would be worse among unmarried individuals; therefore, nonwhite, but there is an education health gradient these results should not be generalized beyond the among those who are white. White men have higher levels married [17]. of self-rated health than nonwhite women but only if they ,e MIOLC data set has smaller sample sizes for the have at least a high school level of education. ,ere are no sample of nonwhite men and women. Because of the smaller gender differences in self-rated health between white men relative sample of nonwhites, it is possible our study lacks power to find statistically significant differences for sub- and women who have the same level of education. ,erefore, the main effects of gender, race, and education groups with small samples in our data (more likely to have do not hold among all of the subgroups. a type 2 error). As described above, there is a risk that we will falsely claim significance when nonexists with the many comparisons involved in subgroup analysis to model 6. Discussion intersectionality. Most prior research on health disparities focuses on either In a recent study, Ferarro et al. reviewed several possible racial-ethnic, gender, or social class as separate forms of mechanisms for long-standing racial and ethnic health social stratification and focuses on older adults. Our findings disparities (e.g., differential exposure to environmental provide support for the value of recognizing multiple hazards, poverty, higher rates of smoking, more dangerous intersecting systems of advantage and disadvantage simul- jobs, less access to health care, accumulation of disadvan- tages, and weathering) [3]. A central idea in health dis- taneously [1, 5, 6, 11]. For example, similar to prior studies, we find strong associations of education and health [36]. We parities research is that differences between groups Journal of Aging Research 7 constructed from different locations in social hierarchies References reflect modifiable characteristics potentially amenable to [1] L. Bowleg, “,e problem with the phrase women and mi- interventions. Even with our limitations, the current study norities: intersectionality—an important theoretical frame- provides valuable evidence that for most groups, health work for public health,” American Journal of Public Health, disparities persist with age, yet for women, higher education vol. 102, no. 7, pp. 1267–1273, 2012. can be protective (but not for nonwhite men). In addition, [2] G. R. Bauer, “Incorporating intersectionality theory into we have an unusually wide range of ages and participants are population health research methodology: challenges and the followed for two decades, providing considerable data to potential to advance health equity,” Social Science & Medicine, estimate health trajectories. Many studies of aging tend to vol. 110, pp. 10–17, 2014. start when people are in the 50s, because most studies rely [3] K. F. Ferraro, B. R. Kemp, and M. M. Williams, “Diverse aging upon data sources such as the Health and Retirement Survey, and health inequality by race and ethnicity,” Innovation in Aging, vol. 1, no. 1, pp. 1–11, 2017. which is limited to individuals aged 50 or older [19]. [4] P. H. Collins, “Toward a new vision: race, class, and gender as categories of analysis and connection,” Race, Sex & Class, 7. Conclusion vol. 1, no. 1, pp. 25–45, 1993. [5] K. Crenshaw, “Demarginalizing the intersection of race and Our findings join the growing effort to provide more context sex: a black feminist critique of antidiscrimination doctrine, for health inequality and to address groups that have been feminist theory and antiracist politics,” University of Chicago historically marginalized (e.g., women of color). ,is ap- Legal Forum, vol. 1989, p. 139, 1989. proach also allows us to compare groups that are simulta- [6] P. H. Collins, Black Feminist Aought: Knowledge, Con- neously advantaged and disadvantaged, such as lower sciousness, and the Politics of Empowerment, Routledge, New educated white men and women, groups recently receiving York, NY, USA, 2002. increased research and policy attention [38, 39]. As urged by [7] L. McCall, Complex Inequality: Gender, Class, and Race in the Bauer, our study illustrates how using an intersectionality New Economy, Psychology Press, New York, NY, USA, 2001. framework can further disparities research, especially among [8] L. McCall, “,e complexity of intersectionality,” Signs, vol. 30, married individuals [2]. Future studies should evaluate no. 3, pp. 1771–1800, 2005. [9] I. Browne and J. Misra, “,e intersection of gender and race in whether or not health advantages for married individuals the labor market,” Annual Review of Sociology, vol. 29, no. 1, intersect with race, gender, and education to explain the pp. 487–513, 2003. effect of marital status on health. Our findings suggest that [10] T. M. Calasanti and K. F. Slevin, Gender, Social Inequalities, policy efforts to promote marriage are not likely to reduce and Aging, AltaMira Press, Walnut Creek, CA, USA, 2001. health disparities, but discovering if disparities exist even [11] T. M. Calasanti and K. F. Slevin, Age Matters: Realigning Feminist among the married will add to efforts to reduce disparities by Ainking, Taylor & Francis, New York, NY, USA, 2006. focusing efforts on factors that matter the most [26]. In [12] S. W. Hinze, J. Lin, and T. E. Andersson, “Can we capture the addition to healthier people “selecting into” marriage, there intersections? Older Black women, education, and health,” is also evidence that marriage has health benefits [18], Women’s Health Issues, vol. 22, no. 1, pp. e91–e98, 2012. particularly for men [40]. However, this benefit may not [13] D. R. Williams and C. Collins, “US socioeconomic and racial extend to nonwhite men given lower marriage and higher differences in health: patterns and explanations,” Annual Review of Sociology, vol. 21, no. 1, pp. 349–386, 1995. cohabitation rates among African Americans compared to [14] United States Center for Disease Control NCHS, Health, white Americans [24]. United States, 2015: With Special Feature on Racial and Ethnic Examining multiple subgroups with varying levels of Health Disparities, CreateSpace Independent Publishing privilege and disadvantage reveals that the most structurally Platform, Hyattsville, MD, USA, 2016. advantaged group did indeed have the best initial self-rated [15] M. E. Dupre, “Educational differences in age-related patterns health. Most, but not all, subgroups declined in health at of disease: reconsidering the cumulative disadvantage and similar rates. ,erefore, we cannot expect that health dis- age-as-leveler hypotheses,” Journal of Health and Social parities will diminish with time, as would occur if age was Behavior, vol. 48, no. 1, pp. 1–15, 2007. a leveler [15, 16]. Instead, gaps persist, and therefore efforts [16] D. Dannefer, “Cumulative advantage/disadvantage and the to improve health disparities by gender, race, and class are life course: cross-fertilizing age and social science theory,” still needed. Journals of Gerontology Series B: Psychological Sciences and Social Sciences, vol. 58, no. 6, pp. S327–S337, 2003. [17] L. J. Waite, “Does marriage matter?,” Demography, vol. 32, Disclosure no. 4, pp. 483–507, 1995. [18] J. S. Wong and L. J. Waite, “Marriage, social networks, and ,e findings and conclusions in this paper are those of the health at older ages,” Journal of Population Ageing, vol. 8, authors and do not necessarily represent the views of the no. 1-2, pp. 7–25, 2015. Department of Health and Human Services or the Agency [19] J. Liang, A. R. Quiñones, J. M. Bennet et al., “Evolving self- for Healthcare Research and Quality. rated health in middle and old age: how does it differ across Black, Hispanic, and White Americans?,” Journal of Aging and Health, vol. 22, no. 1, pp. 3–26, 2010. Conflicts of Interest [20] B. J. Goosby, J. E. Cheadle, and T. McDade, “Birth weight, early life ,e authors declare that there are no conflicts of interest course BMI, and body size change: chains of risk to adult in- regarding the publication of this paper. flammation?,” Social Science & Medicine, vol. 148, pp.102–109, 2016. 8 Journal of Aging Research [21] K. M. Walsemann, B. J. Goosby, and D. Farr, “Life course SES and cardiovascular risk: heterogeneity across race/ethnicity and gender,” Social Science & Medicine, vol. 152, pp. 147–155, 2016. [22] B. L. Kail, “Marital status as a moderating factor in the process of disablement,” Journal of Aging and Health, vol. 28, no. 1, pp. 139–164, 2016. [23] R. Sunshine and S. Rote, “Marital disruption and allostatic load in late life,” Journal of Aging and Health, vol. 29, no. 4, pp. 688–707, 2016. [24] R. K. Raley, M. M. Sweeney, and D. Wondra, “,e growing racial and ethnic divide in US marriage patterns,” Ae Future of Children, vol. 25, no. 2, pp. 89–109, 2015. [25] K. Edin and J. M. Reed, “Why don’t they just get married? Barriers to marriage among the disadvantaged,” Ae Future of Children, vol. 15, no. 2, pp. 117–137, 2005. [26] K. Edin and M. J. Kefalas, Promises I Can Keep: Why Poor Women Put Motherhood before Marriage, University of California Press, Berkeley, CA, USA, 2011. [27] P. R. Amato, D. R. Johnson, A. Booth, and S. J. Rogers, “Continuity and change in marital quality between 1980 and 2000,” Journal of Marriage and Family, vol. 65, no. 1, pp. 1–22, 2003. [28] A. Booth and D. R. Johnson, “Tracking respondents in a tele- phone interview panel selected by random digit dialing,” So- ciological Methods & Research, vol. 14, no. 1, pp. 53–64, 1985. [29] E. L. Idler and Y. Benyamini, “Self-rated health and mortality: a review of twenty-seven community studies,” Journal of Health and Social Behavior, vol. 38, no. 1, pp. 21–37, 1997. [30] I. Kawachi, B. P. Kennedy, K. Lochner, and D. Prothrow-Stith, “Social capital, income inequality, and mortality,” American Journal of Public Health, vol. 87, no. 9, pp. 1491–1498, 1997. [31] S. W. Raudenbush and A. S. Bryk, Hierarchical Linear Models: Applications and Data Analysis Methods, Vol. 1, Sage Publi- cations, ,ousand Oaks, CA, USA, 2002. [32] J. D. Singer and J. B. Willett, Applied Longitudinal Data Analysis: Modeling Change and Event Occurrence, Oxford University Press, Oxford, UK, 2003. [33] G. N. Holmbeck, “Post-hoc probing of significant moderational and mediational effects in studies of pediatric populations,” Journal of Pediatric Psychology, vol. 27, no. 1, pp. 87–96, 2002. [34] A. Gelman, J. Hil, and M. Yajima, “Why we (usually) don’t have to worry about multiple comparisons,” Journal of Re- search on Educational Effectiveness, vol. 5, no. 2, pp. 189–211, [35] T. Perneger, “What’s wrong with Bonferroni adjustments,” British Medical Journal, vol. 316, no. 7139, pp. 1236–1238, [36] J. Mirowsky and C. E. Ross, “Education, health, and the default American lifestyle,” Journal of Health and Social Behavior, vol. 56, no. 3, pp. 297–306, 2015. [37] F. D. Wolinsky, T. R. Miller, T. K. Malmstrom et al., “Self- rated health: changes, trajectories, and their antecedents among African Americans,” Journal of Aging and Health, vol. 20, no. 2, pp. 143–158, 2008. [38] J. K. Montez and A. Zajacova, “Why is life expectancy de- clining among low-educated women in the United States?,” American Journal of Public Health, vol. 104, no. 10, pp. e5–e7, [39] A. Case and A. Deaton, “Rising morbidity and mortality in midlife among white non-Hispanic Americans in the 21st century,” Proceedings of the National Academy of Sciences, vol. 112, no. 49, pp. 15078–15083, 2015. [40] R. M. Stolzenberg and K. Williams, “Gendered reciprocity: work discontent and the household production of health,” Social Science Research, vol. 37, no. 1, pp. 180–201, 2008. MEDIATORS of INFLAMMATION The Scientific Gastroenterology Journal of World Journal Research and Practice Diabetes Research Disease Markers Hindawi Hindawi Publishing Corporation Hindawi Hindawi Hindawi Hindawi www.hindawi.com Volume 2018 http://www www.hindawi.com .hindawi.com V Volume 2018 olume 2013 www.hindawi.com Volume 2018 www.hindawi.com Volume 2018 www.hindawi.com Volume 2018 International Journal of Journal of Immunology Research Endocrinology Hindawi Hindawi www.hindawi.com Volume 2018 www.hindawi.com Volume 2018 Submit your manuscripts at www.hindawi.com BioMed PPAR Research Research International Hindawi Hindawi www.hindawi.com Volume 2018 www.hindawi.com Volume 2018 Journal of Obesity Evidence-Based Journal of Journal of Stem Cells Complementary and Ophthalmology International Alternative Medicine Oncology Hindawi Hindawi Hindawi Hindawi Hindawi www.hindawi.com Volume 2018 www.hindawi.com Volume 2018 www.hindawi.com Volume 2018 www.hindawi.com Volume 2018 www.hindawi.com Volume 2013 Parkinson’s Disease Computational and Behavioural Mathematical Methods AIDS Oxidative Medicine and in Medicine Neurology Research and Treatment Cellular Longevity Hindawi Hindawi Hindawi Hindawi Hindawi www.hindawi.com Volume 2018 www.hindawi.com Volume 2018 www.hindawi.com Volume 2018 www.hindawi.com Volume 2018 www.hindawi.com Volume 2018 http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Journal of Aging Research Hindawi Publishing Corporation

Self-Rated Health Trajectories among Married Americans: Do Disparities Persist over 20 Years?

Loading next page...
 
/lp/hindawi-publishing-corporation/self-rated-health-trajectories-among-married-americans-do-disparities-1P3u1iDHdC

References

References for this paper are not available at this time. We will be adding them shortly, thank you for your patience.

Publisher
Hindawi Publishing Corporation
Copyright
Copyright © 2018 Terceira A. Berdahl and Julia McQuillan. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
ISSN
2090-2204
eISSN
2090-2212
DOI
10.1155/2018/1208598
Publisher site
See Article on Publisher Site

Abstract

Hindawi Journal of Aging Research Volume 2018, Article ID 1208598, 8 pages https://doi.org/10.1155/2018/1208598 Research Article Self-Rated Health Trajectories among Married Americans: Do Disparities Persist over 20 Years? 1 2 Terceira A. Berdahl and Julia McQuillan Division of Research and Modeling, Center for Financing, Access and Cost Trends, Agency for Healthcare Research and Quality, 5600 Fishers Lane, Rockville, MD 20852, USA University of Nebraska-Lincoln, 709 Oldfather Hall, Lincoln, NE 68588-0324, USA Correspondence should be addressed to Terceira A. Berdahl; terceira.berdahl@ahrq.hhs.gov Received 5 June 2017; Revised 27 October 2017; Accepted 22 November 2017; Published 11 January 2018 Academic Editor: F. R. Ferraro Copyright © 2018 Terceira A. Berdahl and Julia McQuillan. ,is is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. ,e purpose of this study is to understand self-rated health (SRH) trajectories by social location (race/ethnicity by gender by social class) among married individuals in the United States. We estimate multilevel models of SRH using six observations from 1980 to 2000 from a nationally representative panel of married individuals initially aged 25–55 (Marital Instability Over the Life Course Study). Results indicate that gender, race/ethnicity, and social class are associated with initial SRH disparities. Women are less healthy than men; people of color are less healthy than whites; lower educated individuals are less healthy than higher educated individuals. Women’s health declined slower than men’s but did not differ by race/ethnicity or education. Results from complex intersectional models show that white men with any college had the highest initial SRH. Only women with any college had significantly slower declines in SRH compared to white men with any college. For married individuals of all ages, most initial SRH disparities persist over twenty years. Intersecting statuses show that education provides uneven health benefits across racial/ethnic and gender subgroups. of health inequality to acknowledge diverse outcomes of 1. Introduction individuals who occupy different positions in social struc- ,ere are well-documented disparities in health by race/ tural hierarchies [1–3]. One useful approach is the inter- ethnicity, gender, and social class in the United States. Less sectionality paradigm [4–6], which emphasizes how social is known about the combined advantages and disadvantages at structures intersect to create the axis of advantage and the intersection of race/ethnicity, gender, and social class. disadvantage in multiplicative rather than only in separate or Questions also remain regarding whether or not initial self- additive ways. For example, there is evidence that the ex- rated health disparities exacerbate or diminish over time. It is periences of women of color cannot be adequately un- important to estimate, and not assume, whether social class derstood by studying race/ethnicity or gender in isolation effects on health are similar for men and women and white and [1, 7–9]. To assess particular policies or laws, it can be ap- nonwhite Americans. To advance knowledge of patterns of propriate to study only one axis of inequality; but to gain health disparities in the United States, we estimate self-rated a more comprehensive understanding of overall changes in health (SRH) trajectories over 20 years using six waves (ob- health with aging, comparing subgroups created by assessing servations) from a panel study of married individuals who were multiple axes of stratification simultaneously (e.g., social aged 25–55 in 1980 at the time of the first interview. class, race/ethnicity, and gender) is fruitful [10–12]. Prior work on health disparities finds that indicators of social location (e.g., race, gender, and education) contribute 2. Background to health inequality [10, 13]. Men have worse mortality than 2.1. Intersectionality and Health Disparities. A growing women but spend fewer years with comorbidity and dis- number of researchers are calling for more nuanced analyses ability [10, 14]. Compared to whites, racial-ethnic minorities 2 Journal of Aging Research continue to face lower life expectancy and worse health due to the intersections of gender, race, and education? If yes, outcomes from birth until death [14]. Studies of educational do disparities persist over decades? To answer these questions, disparities find that people with lower education have worse we examined self-rated health trajectories of a random sample health and faster health declines with aging compared to of individuals who were married in 1980 and followed over people with higher education [15, 16]. Recent studies have twenty years. We restricted the sample to adults who were examined the intersections of gender and social class or initially aged 25–55 years. We categorized individuals by race/ethnicity and social class and find that these intersections gender (men or women), race/ethnicity (white or nonwhite), matter for health changes associated with aging [10]. and level of education (up to less than a high school degree, Most longitudinal studies of aging include people as a high school degree, or any level of college education), thus young as 25 and follow people for decades in small geo- creating twelve subgroups with varying advantages and graphic areas with little racial diversity. We know of no disadvantages. studies that simultaneously analyze the intersections of class, race/ethnicity, and gender for twenty years. Instead, most 4. Data and Methods studies focus on older and elderly adults. ,e current study focuses on people who were married at the start of the study. For the current study, we used data from the Marital In- It is possible that the health protective effects of marriage stability Over the Life Course (MIOLC) study, a six-wave panel study of a national sample of initially married persons in the could minimize gender, education, and race/ethnicity health disparities [17, 18]. Yet, recent research highlighting the long United States [27]. Respondents were interviewed by tele- phone in 1980, 1983, 1988, 1992, 1997, and 2000 (waves 1–6). reach of inequality from grandparents and from early life suggests that even among the married, disparities are likely Only the wife or the husband was included in the study; the participant was selected by random assignment. During the to persist with aging [19–21]. first wave of the study, the survey involved random digit dialing and screening of individuals for inclusion in the study. 2.2. Health Protective Effects of Marriage. Marital status is ,e inclusion criteria required that individuals had to be a social resource that can protect against steep health de- married and between the ages of 18 and 55. Among eligible clines associated with aging. Among people who are mar- households, the rate of completed interviews for the initial ried, the influence of depressive symptoms and chronic wave of the study was 65%, yielding a sample size of 2,034. ,e conditions on subsequent functional limitations is weaker response rates varied among subsequent waves of data col- compared to nonmarried [22]. ,e end of a marriage is also lection (in the order: 78%, 84%, 88%, 90%, and 82%). ,e associated with higher allostatic load (an indicator of bi- analytical sample for this study includes those between the ological risk for worse health) [23]. ,e physiological impact ages of 25 and 55 at the initial interview in order to ensure that of marital disruption is larger for widowhood than divorce, they were likely to have finished their education. Multilevel as indicated by comparisons of inflammatory, metabolic, models incorporate data from as many waves as participants and cardiovascular functioning compared to those who completed; as with all longitudinal studies, there was attrition remain married [23]. over time. Comprehensive information about the data set and Rates of marriage, like health disparities, vary by race how well it represents the U.S. married population of 1980 and social class. Marriage rates and duration are lower and comparisons with formerly married adults in 2000 is among people with lower education and racial/ethnic mi- provided in the methodology report [28]. norities compared to people who have higher education and who are white [24]. ,erefore, health differences by 4.1. Variables/Measures race/ethnicity and education could be smaller among the married than the nonmarried because of selection into 4.1.1. Self-Rated Health (SRH). We measure self-rated a status with privileges [25]. In 1980, the difference in ever- health, the dependent variable, with the following ques- married rates among white women (95%) and black women tion: “Now I have some questions about your health. In (87%) was smaller than that in 2012 (87% for white women general, would you say your own health is excellent (3), good and 63% for black women) [24]. Because marriage was (2), fair (1), or poor (0)?” ,is single-item measure has ubiquitous in 1980, it is unclear if there will be health a strong association with mortality, suggesting that it is disparities in this more narrow group than the general a valid indicator of health [29]. Self-rated health is measured population, yet we argue that it is important to find out, in every wave of data collection in the same way. because marriage is sometimes promoted as a way to reduce social problems, even though there is evidence that marriage does not have the same privileges for all [26]. 4.1.2. Wave/Trajectory. To measure the changes in self-rated health, we include a variable that indicates the wave of observation. ,is variable is coded as 0 for the first interview 3. Current Study in 1980 and 5 for the last interview in 2000, with consecutive In the present study, we advance understanding of health numbers in between. ,ere were about three years between disparities by answering two important questions: In the interviews. With this variable in the model, the intercept United States, are there health disparities among people who indicates the average initial self-rated health, and the co- are married that are created by disadvantaged social locations efficient for “wave” indicates the average change in self-rated Journal of Aging Research 3 Table 1: Sample descriptive statistics for MIOLC wave 1, aged health between observations, or the trajectory of change in 25–55. health. ,e variance component for this variable indicates the amount of spread around the average trajectory. Mean (%) SD Self-rated health 2.26 0.75 Age (in years) 37.24 8.33 4.1.3. Age. We included age (measured in years) as an in- dividual level characteristic to adjust for differences in initial Race health status. We mean center age so that the intercept Nonwhite 11.9 — represents participants with the average age for the sample. White 88.1 — Gender Men 41.5 — 4.1.4. Social Location. We measure social class using edu- cation, a common practice in health research [30]. We use Women 58.5 — only the initial value of education because very few people add Education years of education beyond age 25 in the 1980s. Education was <HS 10.4 — measured in years of completed formal education. We High school graduate 63.1 — converted years into three categories. Because a high school Any college 26.5 — degree is a requirement for many jobs and a college degree is N � 1,785. required for another set of jobs, we divided years into less than a high school degree (<12 years) (LTHS), high school (12 years) (HS), and any college (13+ years). ,e three-category Table 2: Self-rated health by gender, race, and education for adults measure of education helped to make meaningful subgroups aged 25–55 in 1980. for the intersectionality analysis by combining education with Mean SD N gender (what is your sex? male/female) and race/ethnicity Less than high school 2.09 0.85 88 (what race do you consider yourself to be? recoded into White High school graduate 2.30 0.72 371 white and nonwhite) to create 12 total subgroups. Any college 2.49 0.67 455 Women Less than high school 1.85 0.81 20 Nonwhite High school graduate 2.15 0.70 52 4.2. Analysis Strategy. We use multilevel models to estimate Any college 2.05 0.71 58 the trajectories of self-rated health. ,e multilevel model Less than high school 1.88 1.00 57 trajectory approach appropriately estimates the nesting of White High school graduate 2.36 0.77 181 observations within individuals and attrition [31]. All Any college 2.55 0.62 420 models are estimated with random effects for the intercept Men Less than high school 2.10 0.89 21 (initial status) and slope parameters using HLM 7.03. Ad- Nonwhite High school graduate 2.22 0.67 23 ditionally, we control for initial age because the rate of health Any college 2.31 0.73 39 decline is likely to be steeper for those closer to 55 than to 25. MIOLC data, wave 1, N � 1,785. ,e time variable is centered at the initial observation to facilitate meaningful interpretation of the constant (e.g., average initial self-rated health) [32]. and the need to rotate out the reference category using Our analysis begins with a brief discussion of the a “simple slope” approach to hypothesis testing to de- sample in 1980 (see Tables 1 and 2). After presenting these termine which groups are different from each other [33]. descriptive statistics, we move to the HLM analysis. In Yet, the conventional approach also requires solving Table 3, we present Model 1, which includes the time equations to determine the total difference of each group variable and serves as a baseline trajectory model for self- from the others based upon the main effects, two-way rated health. ,is model contains an intercept and a slope interactions, and three-way interactions. We constructed for time (year) with random effects for these parameters as indicator variables for the subgroups to provide a mathe- well as an overall Level 1 error term. Next, we present matically equivalent way to determine subgroup differences in SRH and simultaneously increase the ease of interpreting estimates from Model 2 (Table 3) which includes a variable capturing respondent age at the first year of the survey coefficients summarizing those differences [8]. We also (initial age). In Models 3–5 (Table 4), we present separate estimated supplemental analyses by rotating the reference models with gender (Model 3), race (Model 4), and edu- group to compare each subgroup (available from the author cation (Model 5) to evaluate the effects of these charac- upon request). Our approach involves multiple significance teristics on the aging trajectories without other variables in tests, which in classical regression can increase the risk of the model. ,ese models illustrate how gender, race, and Type 1 errors (finding a significant association when none education modify self-rated health over time. Finally, in exists). By using HLM with empirical Bayes estimation Model 4, we explore the joint effects of gender, race, and methods, we reduce the risk of Type 1 error [34]. ,ere are practices for post hoc adjustments to reduce the risk of education by constructing comparative categories to cap- ture the three-way interaction of these variables (Table 5). Type 1 error (e.g., Bonferroni adjustments), but there are A conventional interaction approach would also gen- arguments against this practice [35]. Because we also have erate eleven coefficients with a comparison to the intercept some groups with small sample sizes (e.g., nonwhite 4 Journal of Aging Research Table 3: Models 1-2: self-rated health by age and self-rated health aging trajectory. Model 1 Model 2 b SE p value b SE p value Initial SRH (intercept) 2.358 0.016 <0.001 2.358 0.016 <0.001 × initial age — — — −0.011 0.002 <0.001 SRH trajectory (time slope) −0.061 0.005 0.005 −0.061 0.005 <0.001 × initial age — — — −0.001 0.001 0.034 Random effects SD VC p value SD VC p value Initial SRH intercept, u 0.561 0.311 <0.001 0.551 0.301 <0.001 SRH trajectory slope, u 0.111 0.011 <0.001 0.111 0.011 <0.001 Level 1, r 0.511 0.261 — 0.511 0.261 — Note. SD � standard deviation; VC � variance component; Level 2 MIOLC N � 1,785. Table 4: Models 3–5: self-rated health trajectories by gender, race, and education. Model 3 Model 4 Model 5 b SE p value b SE p value b SE p value Initial SRH (intercept) 2.410 0.025 <0.001 2.389 0.017 <0.001 2.462 0.020 <0.001 × initial age −0.012 0.002 <0.001 −0.012 0.002 <0.001 −0.010 0.002 <0.001 Women (men � reference group) −0.085 0.033 0.009 — — — — — — Nonwhite (white � reference group) — — — −0.266 0.050 <0.001 — — — Less than HS (any college � reference group) — — — — — — −0.440 0.065 <0.001 High school — — — — — — −0.168 0.034 <0.001 SRH trajectory (time slope) −0.083 0.008 <0.001 −0.064 0.005 <0.001 −0.062 0.006 <0.001 × initial age −0.001 0.001 0.064 −0.001 0.001 0.041 −0.001 0.001 0.020 Women (men � reference group) 0.037 0.010 <0.001 — — — — — — Nonwhite (white � reference group) — — — 0.022 0.020 0.289 — — — Less than HS (any college � reference group) — — — — — — 0.021 0.024 0.392 High school — — — — — — −0.007 0.010 0.496 Random effects SD VC p value SD VC p value SD VC p value Initial SRH intercept, u 0.546 0.298 <0.001 0.541 0.293 <0.001 0.532 0.282 <0.001 SRH trajectory slope, u 0.106 0.011 <0.001 0.108 0.012 <0.001 0.108 0.012 <0.001 Level 1, r 0.508 0.258 — 0.508 0.258 — 0.508 0.258 — Note. SRH � self-rated health; VC � variance component; SD � standard deviation; Level 2 MIOLC N � 1,785. women with less than a high school degree), we also risk In the initial survey, 1,785 participants responded to the making Type 2 errors (failing to find significance when variables in the analytical sample. Average self-rated health there really is an association in the population). We ranges from a low of 1.85 among nonwhite women with less therefore use the conventional 0.05 level of significance, than a high school degree (those with the most social dis- focus on the size and meaning of differences, and use the advantages) to a high of 2.55 among white men with any significance tests as a heuristic for interpreting patterns college education (those with the most social advantages) of associations. (Table 2). For every race by the gender group, those with any college education are the largest group. Almost half of the sample consists of people who are white with any college 5. Results education. ,e smallest groups are people who are nonwhite 5.1. Sample Descriptive Characteristics. We report sample with less than a high school degree. Only nonwhite women descriptive statistics for the sample in Table 1. ,e mean age do not have an education health gradient because those with for the analytical sample was 37.24 (SD � 8.33) at the initial any college education have lower self-rated health than those interview. Just over 10 percent of the sample is nonwhite with a high school degree. For all other gender by race (12%). Over half (58%) of the sample was women. Fewer groups, higher education is associated with higher self-rated participants have less than a high school degree (10%) than health. any college education (55%). On a scale from 0 to 3, with zero representing poor health and 3 representing excellent health, average initial self-rated health was above the mid- 5.2. Aging Trajectory Findings. We first describe the baseline point (M � 2.26 and SD � 0.75). model (no covariates; not shown in the table). ,e baseline Journal of Aging Research 5 Table 5: Model 6: self-rated health trajectories by gender, race, trajectory measure has the value 0 for the first year, 1980, and and education. 5 for the twentieth year, 2000. ,erefore, the intercept for this model provides the average self-rated health in the first Model 6 ∗∗∗ year of the study (1980). Initial self-rated health is 2.361 , b SE p value slightly higher (better) than the average for all years in the Initial SRH (intercept) 2.52 0.03 <0.001 baseline model. ,e SRH trajectory coefficient indicates that, × initial age −0.01 0.00 <0.001 on average, self-rated health declines between interviews ∗∗∗ Race by gender by education (b �−0.061 ). ,e standard deviation for self-rated health × nonwhite women LTHS −0.67 0.14 <0.001 is 0.75. ,erefore, over the six waves and twenty years, on average, self-rated health declines by −0.36 (−0.06 × 6), or × nonwhite women HS −0.42 0.08 <0.001 almost half a standard deviation. ,e rate of decline in × nonwhite women any college −0.47 0.09 <0.001 self-rated health varies across individuals, indicated by × nonwhite men LTHS −0.32 0.18 0.07 the significant variance component (variance components × nonwhite men HS −0.30 0.14 0.03 ∗∗∗ p value � 0.012 ). × nonwhite men any college −0.22 0.12 0.06 To adjust for the thirty-year range of initial ages, we × white women LTHS −0.47 0.09 <0.001 included an indicator of initial age in Model 2 (Table 3). × white women HS −0.22 0.05 <0.001 Consistent with the strong association between age and ∗∗∗ health, health declines with age (b �−0.012 ), and the self- × white women any college −0.05 0.04 0.19 rated health aging trajectory is steeper for participants who × white men LTHS −0.54 0.13 <0.001 were older at the start of the study (the coefficient is −0.001 × white men HS −0.19 0.06 0.00 larger for each additional year). ,e average age in the SRH trajectory (time slope) −0.082 0.010 <0.001 analytical sample is 37.24, with a standard deviation of 8.33 × initial age −0.001 0.001 0.047 years. ,erefore, the difference between a standard deviation Race by gender by education below and above the mean age is 16.66 years translating into × nonwhite women LTHS 0.057 0.128 0.653 24% of a standard deviation in initial self-rated health × nonwhite women HS 0.042 0.035 0.226 (−0.011 ×16.66 �−0.18/.75 � 24%). Initial self-rated health is × nonwhite women any college 0.096 0.035 0.006 arranged as we would expect, with those who are 25 reporting the best health and those age 55 the worst health. × nonwhite men LTHS 0.064 0.070 0.361 Twenty years later, all groups decline, but the decline is × nonwhite men HS −0.041 0.043 0.345 steeper for those who were initially older, resulting in larger × nonwhite men any college −0.045 0.051 0.383 differences between the age groups at the end of the study. × white women LTHS 0.050 0.030 0.099 Building on the model with initial age and the self-rated × white women HS 0.022 0.014 0.105 health trajectory, we next separately explore if there are × white women any college 0.029 0.013 0.022 disparities by gender, race/ethnicity, or level of education for × white men LTHS 0.015 0.045 0.745 the initial or the aging trajectory of self-rated health (Table 4, Models 3–5). Women have lower average initial self-rated × white men HS −0.012 0.021 0.550 ∗∗ health (b �−0.085 , or 12% of a standard deviation lower), Random effects SD VC p value yet the decline in health for women is less steep than it is for Initial SRH intercept, u 0.525 0.275 <0.001 ∗∗∗ men (b � 0.037 ; therefore, the slope for women is SRH trajectory slope, u 0.106 0.011 <0.001 ∗∗∗ ∗∗∗ −0.083 + 0.037 �−0.05). Women initially have worse Level 1, r 0.508 0.258 — health, but because their health declines more slowly, over Note. ,e reference group is white men with any college; LTHS � less than time, they end up with better health than men twenty years high school; HS � high school; AC � any college; Level 2 MIOLC N � 1,785. later. ,ere is also an initial difference between nonwhite and model provides an estimate of the intercept and variance white individuals (Table 4, Model 4). ,e race/ethnicity ∗∗∗ components. ,e intercept indicates that the average self- difference is larger than the gender effect (b �−0.266 ). ∗∗∗ rated health across all participants and all waves is 2.252 , ,e declines in self-rated health are not as steep for nonwhite a value that is above the midpoint of the 0 to 3 range for the compared to white individuals (white �−0.064; ∗∗∗ variable (3 indicates excellent health). Average self-rated nonwhite �−0.064 + 0.022 �−0.040), yet this difference is health varies significantly between people (the variance not significant. component for the intercept in the baseline model is Consistent with prior research, there is a steep health ∗∗∗ 0.267 , and for within people, the Level 1 error is 0.308). gradient by education (Table 4, Model 5). Individuals ,e variance components indicate that just under half of the with any college have the highest self-rated health ∗∗∗ variance in self-rated health is between individuals (47%, or (intercept � 2.462 ). ,ose with a high school level of (0.267/(0.267 + 0.307))) and just over half is within in- education have about a fifth of a standard deviation lower ∗∗∗ dividuals over time (53%). self-rated health (b �−0.168 ), and the difference is even Model 1 in Table 3 provides an estimate of the aging greater for those with less than a high school degree ∗∗∗ trajectory in self-rated health, which is measured by the (b �−0.440 ). Although education has strong associations time variable (one unit is approximately three years). ,e with initial SRH, level of education does not modify the rate 6 Journal of Aging Research of decline, adjusted for age. ,ose who start the study in their did not observe similar effects of education, however, across older age do have slightly but significantly steeper declines in self-rated health trajectories for all racial/ethnic and gender ∗ ∗∗∗ health (age b �−0.001 + trajectory coefficient �−0.062 ; groups. For lower educated women, declines in health are therefore, for each additional year, the rate of decline is similar to higher educated white men. For higher educated −0.001 larger). women, there are slower declines in self-rated health ,e separate models of gender, race/ethnicity, and ed- compared to higher educated white men. ,us, the rate of ucation also show that none of these indicators of social decline in self-rated health is conditioned by gender and status alone explain the variance in initial self-rated health or education. Similar to Liang et al. we find racial differences in in the change in self-rated health over time, indicated by the the intercept, or average health at the beginning of the study, significant variance components in all of the models. with nonwhites scoring significantly worse on health We next examine the combined effect of all three in- measures [19]. Additionally, that study found few signifi- dicators of social location simultaneously (Table 5, Model 6). cant racial differences in aging trajectories over time, which Most of the subgroups differ in initial self-rated health is also similar to our study (although not directly com- relative to white men with any college, the reference group. parable because they did not evaluate racial effects across All groups have negative coefficients, indicating lower initial gender and education) [19]. Consistent with our study self-rated health. ,ree groups do not have significantly findings, another study of aging among African Americans lower self-rated health than white men with any college found that in older adults, self-rated health did not decline (nonwhite men with less than a high school level of edu- significantly over time [37]. cation and nonwhite men and white women with any ,e current study provides evidence that among college). ,e differences in initial self-rated health for most married individuals, there are several initial differences in groups persist with age because the rate of decline is not self-rated health, and for all but higher educated women, significantly different from the most privileged group (white those differences persist over twenty years. Consistent ∗∗∗ men with any college: b �−0.08 ). Two groups, however, with other studies, older age was associated with worse did have significantly slower declines (positive coefficients) health initially and a steeper decline in health over time. ∗∗ Not surprisingly, for all age groups, self-rated health in self-rated health with age: nonwhite women (b � 0.100 ) and white women (b � 0.029 ) with any college. declined with aging. We were surprised to find that even among married individuals, there are important dispar- In addition to the comparisons with the most privi- leged group (white men who have any college education), ities in health and persistence in disparities with declines it is useful to determine if there are significant differences in aging. ,erefore, neither selection into nor the pro- among the groups with varying levels of privilege and tective effects of marriage eliminate the effects of struc- disadvantage as indicated by gender, race/ethnicity, and tural inequality on health. education. We therefore reran the final model and rotated Data limitations require some caution in over- the comparison group until we estimated all possible generalizing these results. One limitation is that the study comparisons (detailed results available upon request). was designed to focus on married individuals. We do not Unlike the story from the separate gender, race/ethnicity, have a comparison group of unmarried individuals during the same time period. We cannot generalize and education models, all women are not different from all men. White women at any level of education do not beyond the ever-married population in 1980. ,is lim- itation is somewhat mitigated by historical context be- differ from less-educated nonwhite men (less than HS and HS). Education does not have the same strong association cause marriage was more common for whites and blacks with self-rated health among nonwhite men and women; in 1980 [24]. We expect that health status, however, there are no education differences among those who are would be worse among unmarried individuals; therefore, nonwhite, but there is an education health gradient these results should not be generalized beyond the among those who are white. White men have higher levels married [17]. of self-rated health than nonwhite women but only if they ,e MIOLC data set has smaller sample sizes for the have at least a high school level of education. ,ere are no sample of nonwhite men and women. Because of the smaller gender differences in self-rated health between white men relative sample of nonwhites, it is possible our study lacks power to find statistically significant differences for sub- and women who have the same level of education. ,erefore, the main effects of gender, race, and education groups with small samples in our data (more likely to have do not hold among all of the subgroups. a type 2 error). As described above, there is a risk that we will falsely claim significance when nonexists with the many comparisons involved in subgroup analysis to model 6. Discussion intersectionality. Most prior research on health disparities focuses on either In a recent study, Ferarro et al. reviewed several possible racial-ethnic, gender, or social class as separate forms of mechanisms for long-standing racial and ethnic health social stratification and focuses on older adults. Our findings disparities (e.g., differential exposure to environmental provide support for the value of recognizing multiple hazards, poverty, higher rates of smoking, more dangerous intersecting systems of advantage and disadvantage simul- jobs, less access to health care, accumulation of disadvan- tages, and weathering) [3]. A central idea in health dis- taneously [1, 5, 6, 11]. For example, similar to prior studies, we find strong associations of education and health [36]. We parities research is that differences between groups Journal of Aging Research 7 constructed from different locations in social hierarchies References reflect modifiable characteristics potentially amenable to [1] L. Bowleg, “,e problem with the phrase women and mi- interventions. Even with our limitations, the current study norities: intersectionality—an important theoretical frame- provides valuable evidence that for most groups, health work for public health,” American Journal of Public Health, disparities persist with age, yet for women, higher education vol. 102, no. 7, pp. 1267–1273, 2012. can be protective (but not for nonwhite men). In addition, [2] G. R. Bauer, “Incorporating intersectionality theory into we have an unusually wide range of ages and participants are population health research methodology: challenges and the followed for two decades, providing considerable data to potential to advance health equity,” Social Science & Medicine, estimate health trajectories. Many studies of aging tend to vol. 110, pp. 10–17, 2014. start when people are in the 50s, because most studies rely [3] K. F. Ferraro, B. R. Kemp, and M. M. Williams, “Diverse aging upon data sources such as the Health and Retirement Survey, and health inequality by race and ethnicity,” Innovation in Aging, vol. 1, no. 1, pp. 1–11, 2017. which is limited to individuals aged 50 or older [19]. [4] P. H. Collins, “Toward a new vision: race, class, and gender as categories of analysis and connection,” Race, Sex & Class, 7. Conclusion vol. 1, no. 1, pp. 25–45, 1993. [5] K. Crenshaw, “Demarginalizing the intersection of race and Our findings join the growing effort to provide more context sex: a black feminist critique of antidiscrimination doctrine, for health inequality and to address groups that have been feminist theory and antiracist politics,” University of Chicago historically marginalized (e.g., women of color). ,is ap- Legal Forum, vol. 1989, p. 139, 1989. proach also allows us to compare groups that are simulta- [6] P. H. Collins, Black Feminist Aought: Knowledge, Con- neously advantaged and disadvantaged, such as lower sciousness, and the Politics of Empowerment, Routledge, New educated white men and women, groups recently receiving York, NY, USA, 2002. increased research and policy attention [38, 39]. As urged by [7] L. McCall, Complex Inequality: Gender, Class, and Race in the Bauer, our study illustrates how using an intersectionality New Economy, Psychology Press, New York, NY, USA, 2001. framework can further disparities research, especially among [8] L. McCall, “,e complexity of intersectionality,” Signs, vol. 30, married individuals [2]. Future studies should evaluate no. 3, pp. 1771–1800, 2005. [9] I. Browne and J. Misra, “,e intersection of gender and race in whether or not health advantages for married individuals the labor market,” Annual Review of Sociology, vol. 29, no. 1, intersect with race, gender, and education to explain the pp. 487–513, 2003. effect of marital status on health. Our findings suggest that [10] T. M. Calasanti and K. F. Slevin, Gender, Social Inequalities, policy efforts to promote marriage are not likely to reduce and Aging, AltaMira Press, Walnut Creek, CA, USA, 2001. health disparities, but discovering if disparities exist even [11] T. M. Calasanti and K. F. Slevin, Age Matters: Realigning Feminist among the married will add to efforts to reduce disparities by Ainking, Taylor & Francis, New York, NY, USA, 2006. focusing efforts on factors that matter the most [26]. In [12] S. W. Hinze, J. Lin, and T. E. Andersson, “Can we capture the addition to healthier people “selecting into” marriage, there intersections? Older Black women, education, and health,” is also evidence that marriage has health benefits [18], Women’s Health Issues, vol. 22, no. 1, pp. e91–e98, 2012. particularly for men [40]. However, this benefit may not [13] D. R. Williams and C. Collins, “US socioeconomic and racial extend to nonwhite men given lower marriage and higher differences in health: patterns and explanations,” Annual Review of Sociology, vol. 21, no. 1, pp. 349–386, 1995. cohabitation rates among African Americans compared to [14] United States Center for Disease Control NCHS, Health, white Americans [24]. United States, 2015: With Special Feature on Racial and Ethnic Examining multiple subgroups with varying levels of Health Disparities, CreateSpace Independent Publishing privilege and disadvantage reveals that the most structurally Platform, Hyattsville, MD, USA, 2016. advantaged group did indeed have the best initial self-rated [15] M. E. Dupre, “Educational differences in age-related patterns health. Most, but not all, subgroups declined in health at of disease: reconsidering the cumulative disadvantage and similar rates. ,erefore, we cannot expect that health dis- age-as-leveler hypotheses,” Journal of Health and Social parities will diminish with time, as would occur if age was Behavior, vol. 48, no. 1, pp. 1–15, 2007. a leveler [15, 16]. Instead, gaps persist, and therefore efforts [16] D. Dannefer, “Cumulative advantage/disadvantage and the to improve health disparities by gender, race, and class are life course: cross-fertilizing age and social science theory,” still needed. Journals of Gerontology Series B: Psychological Sciences and Social Sciences, vol. 58, no. 6, pp. S327–S337, 2003. [17] L. J. Waite, “Does marriage matter?,” Demography, vol. 32, Disclosure no. 4, pp. 483–507, 1995. [18] J. S. Wong and L. J. Waite, “Marriage, social networks, and ,e findings and conclusions in this paper are those of the health at older ages,” Journal of Population Ageing, vol. 8, authors and do not necessarily represent the views of the no. 1-2, pp. 7–25, 2015. Department of Health and Human Services or the Agency [19] J. Liang, A. R. Quiñones, J. M. Bennet et al., “Evolving self- for Healthcare Research and Quality. rated health in middle and old age: how does it differ across Black, Hispanic, and White Americans?,” Journal of Aging and Health, vol. 22, no. 1, pp. 3–26, 2010. Conflicts of Interest [20] B. J. Goosby, J. E. Cheadle, and T. McDade, “Birth weight, early life ,e authors declare that there are no conflicts of interest course BMI, and body size change: chains of risk to adult in- regarding the publication of this paper. flammation?,” Social Science & Medicine, vol. 148, pp.102–109, 2016. 8 Journal of Aging Research [21] K. M. Walsemann, B. J. Goosby, and D. Farr, “Life course SES and cardiovascular risk: heterogeneity across race/ethnicity and gender,” Social Science & Medicine, vol. 152, pp. 147–155, 2016. [22] B. L. Kail, “Marital status as a moderating factor in the process of disablement,” Journal of Aging and Health, vol. 28, no. 1, pp. 139–164, 2016. [23] R. Sunshine and S. Rote, “Marital disruption and allostatic load in late life,” Journal of Aging and Health, vol. 29, no. 4, pp. 688–707, 2016. [24] R. K. Raley, M. M. Sweeney, and D. Wondra, “,e growing racial and ethnic divide in US marriage patterns,” Ae Future of Children, vol. 25, no. 2, pp. 89–109, 2015. [25] K. Edin and J. M. Reed, “Why don’t they just get married? Barriers to marriage among the disadvantaged,” Ae Future of Children, vol. 15, no. 2, pp. 117–137, 2005. [26] K. Edin and M. J. Kefalas, Promises I Can Keep: Why Poor Women Put Motherhood before Marriage, University of California Press, Berkeley, CA, USA, 2011. [27] P. R. Amato, D. R. Johnson, A. Booth, and S. J. Rogers, “Continuity and change in marital quality between 1980 and 2000,” Journal of Marriage and Family, vol. 65, no. 1, pp. 1–22, 2003. [28] A. Booth and D. R. Johnson, “Tracking respondents in a tele- phone interview panel selected by random digit dialing,” So- ciological Methods & Research, vol. 14, no. 1, pp. 53–64, 1985. [29] E. L. Idler and Y. Benyamini, “Self-rated health and mortality: a review of twenty-seven community studies,” Journal of Health and Social Behavior, vol. 38, no. 1, pp. 21–37, 1997. [30] I. Kawachi, B. P. Kennedy, K. Lochner, and D. Prothrow-Stith, “Social capital, income inequality, and mortality,” American Journal of Public Health, vol. 87, no. 9, pp. 1491–1498, 1997. [31] S. W. Raudenbush and A. S. Bryk, Hierarchical Linear Models: Applications and Data Analysis Methods, Vol. 1, Sage Publi- cations, ,ousand Oaks, CA, USA, 2002. [32] J. D. Singer and J. B. Willett, Applied Longitudinal Data Analysis: Modeling Change and Event Occurrence, Oxford University Press, Oxford, UK, 2003. [33] G. N. Holmbeck, “Post-hoc probing of significant moderational and mediational effects in studies of pediatric populations,” Journal of Pediatric Psychology, vol. 27, no. 1, pp. 87–96, 2002. [34] A. Gelman, J. Hil, and M. Yajima, “Why we (usually) don’t have to worry about multiple comparisons,” Journal of Re- search on Educational Effectiveness, vol. 5, no. 2, pp. 189–211, [35] T. Perneger, “What’s wrong with Bonferroni adjustments,” British Medical Journal, vol. 316, no. 7139, pp. 1236–1238, [36] J. Mirowsky and C. E. Ross, “Education, health, and the default American lifestyle,” Journal of Health and Social Behavior, vol. 56, no. 3, pp. 297–306, 2015. [37] F. D. Wolinsky, T. R. Miller, T. K. Malmstrom et al., “Self- rated health: changes, trajectories, and their antecedents among African Americans,” Journal of Aging and Health, vol. 20, no. 2, pp. 143–158, 2008. [38] J. K. Montez and A. Zajacova, “Why is life expectancy de- clining among low-educated women in the United States?,” American Journal of Public Health, vol. 104, no. 10, pp. e5–e7, [39] A. Case and A. Deaton, “Rising morbidity and mortality in midlife among white non-Hispanic Americans in the 21st century,” Proceedings of the National Academy of Sciences, vol. 112, no. 49, pp. 15078–15083, 2015. [40] R. M. Stolzenberg and K. Williams, “Gendered reciprocity: work discontent and the household production of health,” Social Science Research, vol. 37, no. 1, pp. 180–201, 2008. MEDIATORS of INFLAMMATION The Scientific Gastroenterology Journal of World Journal Research and Practice Diabetes Research Disease Markers Hindawi Hindawi Publishing Corporation Hindawi Hindawi Hindawi Hindawi www.hindawi.com Volume 2018 http://www www.hindawi.com .hindawi.com V Volume 2018 olume 2013 www.hindawi.com Volume 2018 www.hindawi.com Volume 2018 www.hindawi.com Volume 2018 International Journal of Journal of Immunology Research Endocrinology Hindawi Hindawi www.hindawi.com Volume 2018 www.hindawi.com Volume 2018 Submit your manuscripts at www.hindawi.com BioMed PPAR Research Research International Hindawi Hindawi www.hindawi.com Volume 2018 www.hindawi.com Volume 2018 Journal of Obesity Evidence-Based Journal of Journal of Stem Cells Complementary and Ophthalmology International Alternative Medicine Oncology Hindawi Hindawi Hindawi Hindawi Hindawi www.hindawi.com Volume 2018 www.hindawi.com Volume 2018 www.hindawi.com Volume 2018 www.hindawi.com Volume 2018 www.hindawi.com Volume 2013 Parkinson’s Disease Computational and Behavioural Mathematical Methods AIDS Oxidative Medicine and in Medicine Neurology Research and Treatment Cellular Longevity Hindawi Hindawi Hindawi Hindawi Hindawi www.hindawi.com Volume 2018 www.hindawi.com Volume 2018 www.hindawi.com Volume 2018 www.hindawi.com Volume 2018 www.hindawi.com Volume 2018

Journal

Journal of Aging ResearchHindawi Publishing Corporation

Published: Jan 11, 2018

References