Minorities’ Diminished Returns of Educational Attainment on Hospitalization Risk: National Health Interview Survey (NHIS)

Copyright © 2019 The Author(s). This is an open-access article distributed under the terms of the Creative Commons Attribution License (http:// creativecommons.org/licenses/by/4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. 1. Background High socioeconomic status (SES), particularly educational attainment, is protective against undesired health outcomes.1-3 Individuals with high educational attainment are less likely to develop cardiovascular disease (CVD), which is the main cause of morbidity.1,4-10 Racial and ethnic inequalities in health are due, in part, to racial and ethnic inequalities in SES and educational attainment.11 However, the protective effects of SES indicators on health outcomes differ among populations.12-15 The effects of SES indicators, particularly educational attainment, on CVDs risk are also shown to vary across populations.13,14 According to the Minorities’ Diminished Returns (MDRs) theory,16,17 SES indicators, particularly educational attainment, are less protective for racial and ethnic minority groups than for majority groups.18 Most of the literature on this issue, however, has focused on the effects of SES on health outcomes for Blacks18-20 and Hispanics21,22 compared to Whites. For example, various studies have shown that SES indicators show larger effects on smoking23 and drinking21 for non-Hispanic Whites than for Hispanics or Blacks. Very limited knowledge, however, exists on MDRs on the risk of hospitalization risk. One study has focused on MDRs on the use of healthcare services. In a study, family SES better reduced unmet health needs for Whites than for Hispanics and non-Hispanics.24 Thus, it seems that the very same MDRs that are established on health for Hispanics Abstract

and Blacks [18][19][20] may also be relevant to health service use.

Objective
This study explored racial and ethnic variations in the effects of educational attainment on the odds of hospitalization over the past 12 months among Americans in the United States.It was hypothesized that high educational attainment would be associated with lower hospitalization risk; however, this effect would be smaller for Hispanics and Blacks than for non-Hispanic Whites.

Design and Settings
The National Health Interview Survey (NHIS) of 2015 was used for this study.The NHIS is the primary source of information on the health and wellbeing of American adults.The sample was limited to the civilian noninstitutionalized population of the United States.Data was collected by the National Center for Health Statistics (NCHS), CDC.

Data Collection
The U.S. Census Bureau acts as the data collection agent for the NHIS under a contractual agreement between the two organizations.Interviews for the NHIS are conducted face-to-face and in the participants' residences.The interviews are sometimes followed with and, on rare occasions, replaced by telephone interviews.The NHIS data is collected continuously throughout each year.Data is collected on demographic factors, socioeconomic characteristics, health behaviors, mental health, physical health, and utilization of healthcare facilities.

Sample and Sampling
The NHIS sampling and sample design are described elsewhere.The NHIS applies a multi-stage sampling strategy.Stage 1 of the sample design was the sampling of 428 primary sampling units (PSUs) drawn from 1,900 geographically defined PSUs, which may be a county, a small group of contiguous counties, or a metropolitan statistical area.All 50 US states and the District of Columbia have representative PSUs in the study.
The NHIS draws samples from households and noninstitutional group quarters such as college dormitories and has four main cores: (1) the Household Composition section, (2) the Family Core, (3) the Sample Child Core, and (4) the Sample Adult Core.

Inclusion and Exclusion Criteria
Individuals were eligible for participation if they were adults (age 18 years or more) and US residents.Those institutionalized in a correctional or healthcare setting were ineligible to participate.
3.5.Measures 3.5.1.Predictor Educational Attainment.Educational attainment was a continuous variable ranging from 0 to 21. Participants were asked about the number of years of schooling they had.A higher score reflected a higher number of years of schooling.

Moderators
Race and Ethnicity.All participants self-identified their race and ethnicity.Races were Blacks/African Americans = 1 and Whites = 0. Ethnicity was a dichotomous variable: Hispanics = 1, Non-Hispanics = 0.

Covariates
Demographic Factors.Demographic data included participant age and gender.Age was continuous, but gender was dichotomous: males = 1, females = 0.
Marital status.A dichotomous variable was used for marital status: 1 married, 0 any other status.
Employment.A single item was used to measure employment over the prior week.This variable was a binary variable: employed = 1, unemployed = 0.
Region.Region was a four-level categorical variable coded as (1) Northeast, (2) Midwest, (3) South, and (4) West.This variable was treated as a nominal variable with West as the reference group.
Obesity.The current study defined obesity as a body mass index (BMI) equal to or larger than 30 kg/m 2 . 25This was based on data on individuals' self-reported weight and height.
Number of CVDs.Number of CVDs was measured using the self-reported history of doctor-diagnosed CVDs.Participants reported if a doctor had told them that they have 1) diabetes, 2) hypertension, 3) high cholesterol, 4) coronary heart disease, 5) angina pectoris, 6) heart attack, 7) heart condition/disease, or 8) stroke.A sum CVD score ranging from 0 to 8 was calculated, and a higher score indicated more CVDs.

Dependent Variable
Hospitalization.Participants were asked if they have been hospitalized overnight during the prior 12 months.Possible answers were yes and no.

Statistical Analysis
Data was analyzed using SPSS 23.0 (IBM Inc, NY, USA).For descriptive statistics, mean and proportion (frequencies) were reported.Two survey logistic regression models were used for multivariable analysis.In these models, educational attainment was used as the independent variable, hospitalization as the dependent variable, demographic factors, marital status, employment, region, obesity, and number of CVDs as covariates, and race and ethnicity as the focal moderators.Both logistic regression models were estimated in the pooled sample that included Blacks, Hispanics, and non-Hispanic Whites.Model 1 did not include race or ethnicity by educational attainment interaction terms, but Model 2 did.Cox & Snell R Square and Nagelkerke R Square were used to compare the fit of Model 1 and Model 2. Adjusted Odds Ratios (ORs), 95% Confidence Intervals (CI), and P values were reported.

Descriptive Statistics
A total number of 28,959 individuals entered this analysis.Table 1 summarizes the descriptive statistics of the participants overall.On average, the participants were 51 years old.Most of the sample were White (83.8%) and non-Hispanic (88.4%).Most participants were female (55.3%).From all participants, 54.2% were employed, 43.4% were married, 34.1% were obese, and 10.1% had been hospitalized during the 12 months prior to the study (Table 1). 2 showed statistically better fit compared to Model 1. Model 1 only included the main effects; however, model 2, also included two interaction terms between race and ethnicity with educational attainment.Model 1 showed that high educational attainment had a protective effect against hospitalization (odds ratio [OR] = 0.99, 95% CI =0.97-1.00)above and beyond our covariates.Model 2 showed significant interactions between the effects of race (OR =1.04, 95% CI = 1.01 -1.08) and ethnicity (OR = 1.04, 95% CI =1.01 -1.07) and educational attainment on hospitalization, suggesting that the inverse association between educational attainment and odds of hospitalization is significantly smaller for Blacks and Hispanics than for non-Hispanic Whites (Table 2).

Discussion
Higher educational attainment was inversely associated with hospitalization in the pooled sample; however, this effect was smaller for Blacks and Hispanics than for non-Hispanic Whites.
It was found that, as educational attainment increases, people's risk of hospitalization decreases.This is, however, less true for Hispanic and Black Americans than for non-Hispanic Whites.This is a replication of MDRs 16,17 of the educational attainment in terms of hospitalization.
The results are in line with other observations that SES indicators, particularly educational attainment, generate less than expected health outcomes for marginalized groups than for mainstream ones.Similar patterns are shown for Blacks, [26][27][28][29] Hispanics, [21][22][23]30 and sexual minorities, 31,32 meaning that minority status, regardless of whether it is based on race, 33 ethnicity, 21,22 or sexual orientation, 31 reduces the health gains of SES resources. 33It is not only educational attainment, 34 but any SES indicator that generates less health for Blacks and Hispanics than for non-Hispanic Whites.The very same patterns are shown for income, 19 occupation, 35 and marital status. 36 Theffects of education and income on chronic medical conditions such as CVDs, 19 asthma, 37 obesity, 18,38 and ADHD 20 are also shown to be smaller for Hispanics and Blacks than for non-Hispanic Whites.Finally, education and other SES indicators have a larger effect on reducing mental health problems such as depression, [39][40][41] suicide, 41 and anxiety for non-Hispanic Whites than Blacks.36 Education also better reduces the risk of mortality for Blacks than for Whites.35 None of these patterns have been previously shown for Hispanics compared to non-Hispanics. The contibution of this study is to extend this literature to hospitalization as the outcome to both racial and ethnic groups.
The systemic nature of the MDRs suggests that it is the society that differentially rewards and differently treats racial and ethnic groups. 16,17This study argues that it is probably the marginalization processes that reduce full participation in society, and it is the social structure and function that do not allow non-White groups to fully leverage their human and economic resources and turn them into the highest levels of tangible outcomes. 16,17

Future Research
There is a need to study the contextual, economic, and behavioral mediators of MDRs of educational attainment on the risk of hospitalization.One potential mechanism for this observation is through CVDs 19 as well as asthma, 37 hypertension, 30 obesity, 18,38 depression, [39][40][41] and suicidality. 42Another mechanism that may increase the risk of hospitalization in high SES Blacks and Hispanics is high risk behavior such as poor diet, 43 poor exercise, 27 smoking, 23,44 and alcohol consumption. 21High SES Blacks and Hispanics are also exposed to higher levels of stress 41,[45][46][47] and more second-hand smoke. 48Research on public and health policies that reduce the MDRs of educational attainment are also needed.Future research should attempt to replicate and validate these findings in other contexts, settings, and age groups.

Limitations
This study has some methodological, conceptual, and statistical limitations.With a cross-sectional design, no causal inferences are drawn between educational attainment and risk of hospitalization.The study could not include data on a wide range of potential covariates such as healthcare access and mental health.Other SES indicators, such as wealth and childhood SES indicators, should also be measured.All study variables were measured at the individual level.Structural factors such as availability and density of healthcare services as well as area-level SES and racial and ethnic compositions of the neighborhoods were not included in this study.Despite these limitations, this is the first of its kind to investigate the relevance of MDRs to the risk of hospitalization.

Conclusion
The protective effects of educational attainment on hospitalization seem to be smaller for Hispanics and Blacks compared to non-Hispanic Whites.To prevent racial and ethnic health disparities, policies and programs should go beyond equal access and focus on equal outcomes.For such a goal, policies that minimize diminished returns of SES for racial and ethnic minorities are needed.Innovative and bold economic, public, and social policies are needed for the members of racial and ethnic minorities to leverage their resources and gain tangible health outcomes.Researchers may also explore mediators and moderators of MDRs for Black and Hispanic people.Review Board (IRB).According to the NIH guidelines and the decision tool regarding human subject involvement in secondary analyses of existing data, the current study was found to be a "Not Human Subject Research".The definition and decision tool which were applied are available here: https://grants.nih.gov/policy/humansubjects/hs-decision.htm.This study was found to be exempt from the IRB approval process for human involvement in research by the CDU IRB.

Table 2
presents the results of 2 pooled sample logistic regression models.Both models were statistically significant at the 0.001 level.Cox & Snell R Square for Model 1 and Model 2 were 0.057 and 0.058.Nagelkerke R square were 0.120 and 0.121 for Model 1 and Model 2, respectively.These fit statistics suggest that Model

Table 1 .
Descriptive Statistics of the Participants Overall