Quantile regressions as a tool to evaluate how an exposure shifts and reshapes the outcome distribution: A primer for epidemiologists.

Quantifying how an exposure affects the entire outcome distribution is often important, e.g., for outcomes such as blood pressure which have non-linear effects on long-term morbidity and mortality. Quantile regressions offer a powerful way of estimating an exposure's relationship with the outcome distribution but remain underused in epidemiology. We introduce quantile regressions with a focus on distinguishing estimators for quantiles of the conditional and unconditional outcome distributions. We also present an empirical example in which we fit mean and quantile regressions to investigate educational attainment's association with later-life systolic blood pressure (SBP). We use data on 8,875 US-born respondents aged 50+ years from the US Health and Retirement Study. More education was negatively associated with mean SBP. Conditional and unconditional quantile regressions both suggested a negative association between education and SBP at all levels of SBP, but the absolute magnitudes of these associations were higher at higher SBP quantiles relative to lower quantiles. In addition to showing that educational attainment shifted the SBP distribution left-wards, quantile regression results revealed that education may have reshaped the SBP distribution through larger protective associations in the right tail, thus benefiting those at highest risk of cardiovascular diseases.

Code Review as a Simple Trick to Enhance Reproducibility, Accelerate Learning, and Improve the Quality of Your Team's Research.

Programming for data wrangling and statistical analysis is an essential technical tool of modern epidemiology, yet many epidemiologists receive limited formal training in strategies to optimize the quality of our code. In complex projects, coding mistakes are easy to make, even for skilled practitioners. Such mistakes can lead to invalid research claims that reduce the credibility of the field. Code review is a straightforward technique used by the software industry to reduce the likelihood of coding bugs. The systematic implementation of code review in epidemiologic research projects could not only improve science but also decrease stress, accelerate learning, contribute to team building, and codify best practices. In the present article, we argue for the importance of code review and provide some recommendations for successful implementation for 1) the research laboratory, 2) the code author (the initial programmer), and 3) the code reviewer. We outline a feasible strategy for implementation of code review, though other successful implementation processes are possible to accommodate the resources and workflows of different research groups, including other practices to improve code quality. Code review isn't always glamorous, but it is critically important for science and reproducibility. Humans are fallible; that's why we need code review.

Does the Type and Timing of Educational Attainment Influence Physical Health? A Novel Application of Sequence Analysis.

Nontraditional education trajectories are common, but their influence on physical health is understudied. We constructed year-by-year education trajectories for 7,501 National Longitudinal Survey of Youth 1979 participants aged 14 to 48 years (262,535 person-years of education data from 1979 to 2014). We characterized trajectory similarity using sequence analysis and used hierarchical clustering to group similar educational trajectories. Using linear regression, we predicted physical health summary scores of the participants at age 50 years from the 12-item Short-Form Survey, adjusting for available confounders, and evaluated effect modification by sex, race/ethnicity, and childhood socioeconomic status. We identified 24 unique educational sequence clusters on the basis of highest level of schooling and attendance timing. General education development credentials predicted poorer health than did high school diplomas (ß = -3.07, 95% confidence interval: -4.07, -2.07), and bachelor's degrees attained at earlier ages predicted better health than the same degree attained at later ages (ß = 1.66, 95% confidence interval: 0.05, 3.28). Structurally marginalized groups benefited more from some educational trajectories than did advantaged groups (e.g., Black vs. White Americans with some college; those of low vs. high childhood socioeconomic status who received an associate's or bachelor's degree). Both type and timing of educational credentials may influence physical health. Literature to date has likely underestimated the impact of educational trajectories on health.

Homelessness, housing instability, and abortion outcomes at an urban abortion clinic in the United States.

BACKGROUND: Adverse reproductive health outcomes are well documented among people experiencing homelessness or housing instability. Little is known about abortion outcomes among this population. OBJECTIVE: This study aimed to investigate the relationship between housing status and abortion outcomes and whether gestational age mediates this relationship. STUDY DESIGN: Our sample comprised 1903 individuals who had abortions at an urban clinic in San Francisco, CA, from 2015 to 2017. We defined homelessness or housing instability as a binary exposure, which included staying outside, with friends and/or family, or in a tent, vehicle, shelter, transitional program, or hotel. We evaluated gestational duration of ≥20 weeks as a mediator variable. Our primary outcome was any abortion complication. Logistic regression models were adjusted for age, race, substance use, mental health diagnoses, and previous vaginal and cesarean deliveries. RESULTS: Approximately 19% (n=356) of abortions were among people experiencing homelessness or housing instability. Compared with those with stable housing, people experiencing homelessness or housing instability presented later in pregnancy (mean gestational duration, 13.3 vs 9.5 weeks; P<.001) and had more frequent complications (6.5% vs 2.8%; P<.001; odds ratio, 2.2; 95% confidence interval, 1.2-3.9). Adjusting for race, substance use, mental health diagnoses, and previous cesarean deliveries, individuals experiencing homelessness or housing instability were more likely to have abortion complications (odds ratio, 2.3; 95% confidence interval, 1.3-4.0). However, the relationship was attenuated after adjusting for gestational duration (odds ratio, 1.4; 95% confidence interval, 0.7-2.6), suggesting that gestational duration mediates the relationship between housing status and abortion complications. CONCLUSION: Patients experiencing homelessness or housing instability presented later in gestation, which seems to contribute to the increased frequency of abortion complications.

Short-term effects of the earned income tax credit on mental health and health behaviors.

Poverty has consistently been linked to poor mental health and risky health behaviors, yet few studies evaluate the effectiveness of programs and policies to address these outcomes by targeting poverty itself. We test the hypothesis that the earned income tax credit (EITC)-the largest U.S. poverty alleviation program-improves short-term mental health and health behaviors in the months immediately after income receipt. We conducted parallel analyses in two large longitudinal national data sets: the National Health Interview Survey (NHIS, 1997-2016, N = 379,603) and the Panel Study of Income Dynamics (PSID, 1985-2015, N = 29,808). Outcomes included self-rated health, psychological distress, tobacco use, and alcohol consumption. We employed difference-in-differences analysis, a quasi-experimental technique. We exploited seasonal variation in disbursement of the EITC, which is distributed as a tax refund every spring: we compared outcomes among EITC-eligible individuals interviewed immediately after refund receipt (Feb-Apr) with those interviewed in other months more distant from refund receipt (May-Jan), "differencing out" seasonal trends among non-eligible individuals. For most outcomes, we were unable to rule out the null hypothesis that there was no short-term effect of the EITC. Findings were cross-validated in both data sets. The exception was an increase in smoking in PSID, although this finding was not robust to sensitivity analyses. While we found no short-term "check effect" of the EITC on mental health and health behaviors, others have found long-term effects on these outcomes. This may be because recipients anticipate EITC receipt and smooth their income accordingly.

Performance of Matching Methods as Compared With Unmatched Ordinary Least Squares Regression Under Constant Effects.

Matching methods are assumed to reduce the likelihood of a biased inference compared with ordinary least squares (OLS) regression. Using simulations, we compared inferences from propensity score matching, coarsened exact matching, and unmatched covariate-adjusted OLS regression to identify which methods, in which scenarios, produced unbiased inferences at the expected type I error rate of 5%. We simulated multiple data sets and systematically varied common support, discontinuities in the exposure and/or outcome, exposure prevalence, and analytical model misspecification. Matching inferences were often biased in comparison with OLS, particularly when common support was poor; when analysis models were correctly specified and common support was poor, the type I error rate was 1.6% for propensity score matching (statistically inefficient), 18.2% for coarsened exact matching (high), and 4.8% for OLS (expected). Our results suggest that when estimates from matching and OLS are similar (i.e., confidence intervals overlap), OLS inferences are unbiased more often than matching inferences; however, when estimates from matching and OLS are dissimilar (i.e., confidence intervals do not overlap), matching inferences are unbiased more often than OLS inferences. This empirical "rule of thumb" may help applied researchers identify situations in which OLS inferences may be unbiased as compared with matching inferences.

Is it possible to overcome the 'long arm' of childhood socioeconomic disadvantage through upward socioeconomic mobility?

OBJECTIVES: Socioeconomically disadvantaged children have worse adult health; we test if this 'long arm' of childhood disadvantage can be overcome through upward socioeconomic mobility in adulthood. METHODS: Four SES trajectories (stable low, upwardly mobile, downwardly mobile and stable high) were created from median dichotomized childhood socioeconomic status (SES; childhood human and financial capital) and adult SES (wealth at age 67) from Health and Retirement Study respondents (N = 6669). Healthy ageing markers, in tertiles, were walking speed, peak expiratory flow (PEF), and grip strength measured in 2008 and 2010. Multinomial logistic regression models, weighted to be nationally representative, controlled for age, gender, race, birthplace, outcome year and childhood health and social capital. RESULTS: Upwardly mobile individuals were as likely as the stable high SES group to be in the best health tertile for walking speed (OR = 0.81; 95% CI: 0.63, 1.05; P = 0.114), PEF (OR = 0.97; 95% CI: 0.78, 1.21; P = 0.810) and grip strength (OR = 0.97; 95%CI: 0.74, 1.27; P = 0.980). DISCUSSION: Findings suggest the 'long arm' of childhood socioeconomic disadvantage can be overcome for these markers of healthy ageing through upward socioeconomic mobility.

Trends for Reported Discrimination in Health Care in a National Sample of Older Adults with Chronic Conditions.

BACKGROUND: Discrimination in health care settings is associated with poor health outcomes and may be especially harmful to individuals with chronic conditions, who need ongoing clinical care. Although efforts to reduce discrimination are growing, little is known about national trends in discrimination in health care settings. METHODS: For Black, White, and Hispanic respondents with chronic disease in the 2008-2014 Health and Retirement Study (N = 13,897 individuals and 21,078 reports), we evaluated trends in patient-reported discrimination, defined based on frequency of receiving poorer service or treatment than other people from doctors or hospitals ("never" vs. all other). Respondents also reported the perceived reason for the discrimination. In addition, we evaluated whether wealth predicted lower prevalence of discrimination for Blacks or Whites. We used generalized estimating equation models to account for dependency of repeated measures on individuals and wave-specific weights to represent the US non-institutionalized population aged 54+ . RESULTS: The estimated prevalence of experiencing discrimination in health care among Blacks with a major chronic condition was 27% (95% CI: 23, 30) in 2008 and declined to 20% (95% CI: 17, 22) in 2014. Reports of receiving poorer service or treatment were stable for Whites (17%, 95% CI: 16, 19 in 2014). The Black-White difference in reporting any health care discrimination declined from 8.2% (95% CI: 4.5, 12.0) in 2008 to 2.5% (95% CI: -1.1, 6.0) in 2014. There was no clear trend for Hispanics. Blacks reported race and Whites reported age as the most common reason for discrimination. CONCLUSIONS: Findings suggest national declines in patient-reported discrimination in health care among Blacks with chronic conditions from 2008 to 2014, although reports of discrimination remain common for all racial/ethnic groups. Our results highlight the critical importance of monitoring trends in reports of discrimination in health care to advance equity in health care.

Does the "widowhood effect" precede spousal bereavement? Results from a nationally representative sample of older adults.

OBJECTIVE: Increased mortality risk following spousal bereavement (often called the "widowhood effect") is well documented, but little prior research has evaluated health deteriorations preceding spousal loss. DESIGN: Data are from the Health and Retirement Study, a nationally representative sample of Americans over 50 years old. METHOD: Individuals who were married in 2004 were considered for inclusion. Outcome data from 2006 on mobility (walking, climbing stairs), number of depressive symptoms, and instrumental activities of daily living (IADLs) were used. Exposure was characterized based on marital status at the time of outcome measurement: "recent widows" (N=396) were bereaved between 2004 and 2006, before outcomes were assessed; "near widows" (N=380) were bereaved between 2006 and 2008, after outcomes were assessed; "married" individuals (N=7,330) remained married from 2004 to 2010, the follow-up period for this analysis. Linear regression models predicting standardized mobility, depressive symptoms, and IADLs, were adjusted for age, race, gender, birthplace, socio-economic status, and health at baseline. RESULTS: Compared to married individuals, recent widows had worse depressive symptoms (ß=0.71, 95% confidence interval (CI): [0.57, 0.85]). Near widows had worse depressive symptoms (ß=0.21, 95% CI: [0.08, 0.34]), mobility (ß=0.14, 95%CI: [0.01, 0.26]), and word recall (ß=-0.13, 95%CI: [-0.23, -0.02]) compared to married individuals. CONCLUSIONS: Health declines before spousal death suggests some portion of the "widowhood effect" may be attributable to experiences that precede widowhood and interventions prior to bereavement might help preserve the health of the surviving spouse.

Short- and long-term associations between widowhood and mortality in the United States: longitudinal analyses.

BACKGROUND: Past research shows that spousal death results in elevated mortality risk for the surviving spouse. However, most prior studies have inadequately controlled for socioeconomic status (SES), and it is unclear whether this 'widowhood effect' persists over time. METHODS: Health and Retirement Study participants aged 50+ years and married in 1998 (n = 12 316) were followed through 2008 for widowhood status and mortality (2912 deaths). Discrete-time survival analysis was used to compare mortality for the widowed versus the married. RESULTS: Odds of mortality during the first 3 months post-widowhood were significantly higher than in the continuously married (odds ratio (OR) for men = 1.87, 95% CI: 1.27, 2.75; OR for women = 1.47, 95% CI: 0.96, 2.24) in models adjusted for age, gender, race and baseline SES (education, household wealth and household income), behavioral risk factors and co-morbidities. Twelve months following bereavement, men experienced borderline elevated mortality (OR = 1.16, 95% CI: 1.00, 1.35), whereas women did not (OR = 1.07, 95% CI: 0.90, 1.28), though the gender difference was non-significant. CONCLUSION: The 'widowhood effect' was not fully explained by adjusting for pre-widowhood SES and particularly elevated within the first few months after widowhood. These associations did not differ by sex.

First-generation college graduates have similar depressive symptoms in midlife as multi-generational college graduates.

Purpose: Higher education may protect an individual against depressive symptoms, yet, disadvantaged socioeconomic status (SES) during childhood, often measured by lower parental education, may put them at higher risk for depressive symptoms later in life. This study evaluates if midlife depression is similar for first-generation and multi-generation college graduates. Methods: For US Health and Retirement Study (HRS) participants ages 55-63 (N = 16,752), we defined a 4-category exposure from parents' (highest of mother or father's) and participant's own years of education, with 16 years indicating college completion: multi-gen (both ≥ 16 years: reference); first-gen (parents <16; own ≥ 16); only parent(s) (parents ≥ 16; own <16); and neither (both <16) college graduates across three birth cohorts. We used linear regressions to evaluate relationships between college completion and depressive symptoms measured by an 8-item Center for Epidemiologic Studies - Depression (CES-D) scale. Models pooled over time evaluated differences by sex, race/ethnicity, and birthplace. Results: First-gen and multi-gen college graduates averaged similar depressive symptoms in midlife (ß: 0.01; 95% CI: 0.15, 0.13). Results were similar by sex and race/ethnicity. Conclusion: Consistent with resource substitution theory, college completion may offset the deleterious effects of lower parental education on midlife depressive symptoms for first-generation graduates.

Undergoing transformation to the patient centered medical home in safety net health centers: perspectives from the front lines.

OBJECTIVES: Safety net health centers (SNHCs), which include federally qualified health centers (FQHCs) provide primary care for underserved, minority and low income patients. SNHCs across the country are in the process of adopting the patient centered medical home (PCMH) model, based on promising early implementation data from demonstration projects. However, previous demonstration projects have not focused on the safety net and we know little about PCMH transformation in SNHCs. DESIGN: This qualitative study characterizes early PCMH adoption experiences at SNHCs. SETTING AND PARTICIPANTS: We interviewed 98 staff (administrators, providers, and clinical staff) at 20 of 65 SNHCs, from five states, who were participating in the first of a five-year PCMH collaborative, the Safety Net Medical Home Initiative. MAIN MEASURES: We conducted 30-45 minute, semi-structured telephone interviews. Interview questions addressed benefits anticipated, obstacles encountered, and lessons learned in transition to PCMH. RESULTS: Anticipated benefits for participating in the PCMH included improved staff satisfaction and patient care and outcomes. Obstacles included staff resistance and lack of financial support for PCMH functions. Lessons learned included involving a range of staff, anticipating resistance, and using data as frequent feedback. CONCLUSIONS: SNHCs encounter unique challenges to PCMH implementation, including staff turnover and providing care for patients with complex needs. Staff resistance and turnover may be ameliorated through improved health care delivery strategies associated with the PCMH. Creating predictable and continuous funding streams may be more fundamental challenges to PCMH transformation.

Food Insecurity, Memory, and Dementia Among US Adults Aged 50 Years and Older.

Importance: Despite existing federal programs to increase access to food, food insecurity is common among US older adults. Food insecurity may affect Alzheimer disease and Alzheimer disease-related dementias via multiple mechanisms, yet there is almost no quantitative research evaluating this association. Objective: To examine whether food insecurity in older adults is associated with later-life cognitive outcomes. Design, Setting, and Participants: This cohort study of US residents aged 50 years and older from the US Health and Retirement Study was restricted to respondents with food insecurity data in 2013 and cognitive outcome data between calendar years 2014 and 2018. Analyses were conducted from June 1 to September 22, 2023. Exposure: Food insecurity status in 2013 was assessed using the validated US Department of Agriculture 6-item Household Food Security Module. Respondents were classified as being food secure, low food secure, and very low food secure. Main Outcomes and Measures: Outcomes were dementia probability and memory score (standardized to 1998 units), estimated biennially between 2014 and 2018 using a previously validated algorithm. Generalized estimation equations were fit for dementia risk and linear mixed-effects models for memory score, taking selective attrition into account through inverse probability of censoring weights. Results: The sample consisted of 7012 participants (18 356 person-waves); mean (SD) age was 67.7 (10.0) years, 4131 (58.9%) were women, 1136 (16.2%) were non-Hispanic Black, 4849 (69.2%) were non-Hispanic White, and mean (SD) duration of schooling was 13.0 (3.0) years. Compared with food-secure older adults, experiencing low food security was associated with higher odds of dementia (odds ratio, 1.38; 95% CI, 1.15-1.67) as was experiencing very low food security (odds ratio, 1.37; 95% CI, 1.11-1.59). Low and very low food security was also associated with lower memory levels and faster age-related memory decline. Conclusions and Relevance: In this cohort study of older US residents, food insecurity was associated with increased dementia risk, poorer memory function, and faster memory decline. Future studies are needed to examine whether addressing food insecurity may benefit brain health.

Lifecourse Educational Trajectories and Hypertension in Midlife: An Application of Sequence Analysis.

BACKGROUND: Higher educational attainment predicts lower hypertension. Yet, associations between nontraditional educational trajectories (eg, interrupted degree programs) and hypertension are less well understood, particularly among structurally marginalized groups who are more likely to experience these non-traditional trajectories. METHODS: In National Longitudinal Survey of Youth 1979 cohort data (N = 6 317), we used sequence and cluster analyses to identify groups of similar educational sequences-characterized by timing and type of terminal credential-that participants followed from age 14-48 years. Using logistic regression, we estimated associations between the resulting 10 educational sequences and hypertension at age 50. We evaluated effect modification by individual-level indicators of structural marginalization (race, gender, race and gender, and childhood socioeconomic status [cSES]). RESULTS: Compared to terminal high school (HS) diploma completed at traditional age, terminal GED (OR: 1.32; 95%CI: 1.04, 1.66) or Associate degree after

Development of a safety net medical home scale for clinics.

BACKGROUND: Existing tools to measure patient-centered medical home (PCMH) adoption are not designed for research evaluation in safety-net clinics. OBJECTIVE: Develop a scale to measure PCMH adoption in safety-net clinics. RESEARCH DESIGN: Cross-sectional survey. SUBJECTS: Sixty-five clinics in five states. MAIN MEASURES: Fifty-two-item Safety Net Medical Home Scale (SNMHS). The total score ranges from 0 (worst) to 100 (best) and is an average of multiple subscales (0-100): Access and Communication, Patient Tracking and Registry, Care Management, Test and Referral Tracking, Quality Improvement, and External Coordination. The scale was tested for internal consistency reliability and tested for convergent validity using The Assessment of Chronic Illness Care (ACIC) and the Patient-Centered Medical Home Assessment (PCMH-A). The scale was applied to centers in the sample. In addition, linear regression models were used to measure the association between clinic characteristics and medical home adoption. RESULTS: The SNMHS had high internal consistency reliability (Cronbach's alpha = 0.84). The SNMHS score correlated moderately with the ACIC score (r = 0.64, p < 0.0001) and the PCMH-A (r = 0.56, p < 0.001). The mean SNMHS score was 61 ± SD 13. Among the subscales, External Coordination (66 ± 16) and Access and Communication (65 ± 14) had the highest mean scores, while Quality Improvement (55 ± 17) and Care Management (55 ± 16) had lower mean scores. Clinic characteristics positively associated with total SNMHS score were having more providers (ß 15.8 95% CI 8.1-23.4 >8 provider FTEs compared to <4 FTEs) and participation in financial incentive programs (ß 8.4 95% 1.6-15.3). CONCLUSION: The SNMHS demonstrated reliability and convergent validity for measuring PCMH adoption in safety-net clinics. Some clinics have significant PCMH adoption. However, room for improvement exists in most domains, especially for clinics with fewer providers.

Implications of the new definition of diabetes for health disparities.

In July 2009, an international committee announced a new diagnostic criterion for diabetes based on hemoglobin Alc (HbA1c) values. Our objective was to estimate how the new diabetes diagnostic criterion will affect the prevalence of diabetes among different race, age, and gender subpopulations, compared to the previously used fasting plasma glucose (FPG) criterion. We analyzed nationally representative data from The National Health and Nutrition Examination Survey (NHANES), aggregated from 1999 to 2006. We estimated the prevalence of known diabetes (prevalence static across either diagnostic criterion), unknown, and no diabetes (prevalence variable by criterion). We tested statistical significance of prevalence differences for unknown diabetes between the prior diagnostic criterion--FPG of at least 126 mg/dL--and the new diagnostic criterion--HbA1c of at least 6.5%--using conditional logistic regression. We further tested the association of these differences with demographic factors. The new HbA1c diagnostic criterion differentially affects different racial/ethnic groups. For non-Hispanic whites, the prevalence of undiagnosed diabetes was more than halved from 2.6% (95% confidence interval [CI], 2.2-3.1) with FPG diagnosis to 1.3% (95% CI, 1.0-1.7), P<.001 with HbAic diagnosis. For Hispanics and non-Hispanic blacks, the differences in prevalence by the 2 criteria were smaller and nonsignificant. Racial differences by diagnostic criteria were most pronounced among people aged over 55 years. Overall, the new definition of diabetes differentially affects ethnic groups, especially for older people. If the new criterion is widely adopted, over time, we may see an apparent widening of racial/ethnic disparities in diabetes prevalence.

Understanding the benefits of different types and timing of education for mental health: A sequence analysis approach.

OBJECTIVES: Individuals increasingly experience delays or interruptions in schooling; we evaluate the association between these non-traditional education trajectories and mental health. METHODS: Using year-by-year education data for 7,501 National Longitudinal Survey of Youth 1979 participants, ages 14-48 (262,535 person-years of education data), we applied sequence analysis and a clustering algorithm to identify educational trajectory groups, incorporating both type and timing to credential. Linear regression models, adjusted for early-life confounders, evaluated relationships between educational trajectories and mental health component scores (MCS) from the 12-item short form instrument at age 50. We evaluated effect modification by race, gender, and race by gender. RESULTS: We identified 24 distinct educational trajectories based on highest credential and educational timing. Compared to high school (HS) diplomas, < HS (beta=-3.41, 95%CI:-4.74,-2.07) and general educational development credentials (GEDs) predicted poorer MCS (beta=-2.07,95%CI:-3.16,-0.98). The following educational trajectories predicted better MCS: some college immediately after High School (beta=1.52, 95%CI:0.68,2.37), Associate degrees after long interruptions (beta=1.73, 95%CI:0.27,3.19), and graduate school soon after Bachelor's completion (beta=1.13, 95%CI:0.21,2.06). Compared to White men, Black women especially benefited from educational credentials higher than HS in predicting MCS. CONCLUSIONS: Both type and timing of educational credential predicted mental health. Black women's mental higher especially benefited from higher educational credentials.

Differential associations between state-level educational quality and cardiovascular health by race: Early-life exposures and late-life health.

Cardiovascular diseases (CVD) are patterned by educational attainment but educational quality is rarely examined. Educational quality differences may help explain racial disparities. Health and Retirement Study respondent data (1992-2014; born 1900-1951) were linked to state- and year-specific educational quality measures when the respondent was 6 years old. State-level educational quality was a composite of state-level school term length, student-to-teacher ratio, and per-pupil expenditure. CVD-related outcomes were self-reported (N = 24,339) obesity, heart disease, stroke, ever-smoking, high blood pressure, diabetes and objectively measured (N = 10,704) uncontrolled blood pressure, uncontrolled blood sugar, total cholesterol, high-density lipoprotein cholesterol (HDL), and C-reactive protein. Race/ethnicity was classified as White, Black, or Latino. Cox models fit for dichotomous time-to-event outcomes and generalized estimating equations for continuous outcomes were adjusted for individual and state-level confounders. Heterogeneities by race were evaluated using state-level educational quality by race interaction terms; race-pooled, race by educational quality interaction, and race-specific estimates were calculated. In race-pooled analyses, higher state-level educational quality was protective for obesity (HR = 0.92; 95%CI(0.87,0.98)). In race-specific estimates for White Americans, state-level educational quality was protective for high blood pressure (HR = 0.95; 95%CI(0.91,0.99). Differential relationships among Black compared to White Americans were observed for obesity, heart disease, stroke, smoking, high blood pressure, and HDL cholesterol. In race-specific estimates for Black Americans, higher state-level educational quality was protective for obesity (HR = 0.88; 95%CI(0.84,0.93)), but predictive of heart disease (HR = 1.07; 95%CI(1.01,1.12)), stroke (HR = 1.20; 95%CI(1.08,1.32)), and smoking (HR = 1.05; 95%CI(1.02,1.08)). Race-specific hazard ratios for Latino and Black Americans were similar for obesity, stroke, and smoking. Better state-level educational quality had differential associations with CVD by race. Among minorities, better state-level educational quality was predominately associated with poorer CVD outcomes. Results evaluate the 1900-1951 birth cohorts; secular changes in the racial integration of schools since the 1950s, means results may not generalize to younger cohorts.

Do the health benefits of education vary by sociodemographic subgroup? Differential returns to education and implications for health inequities.

PURPOSE: Evidence suggests education is an important life course determinant of health, but few studies examine differential returns to education by sociodemographic subgroup. METHODS: Using National Longitudinal Survey of Youth 1979 (n = 6158) cohort data, we evaluate education attained by age 25 years and physical health (PCS) and mental health component summary scores (MCS) at age 50 years. Race / ethnicity, sex, geography, immigration status, and childhood socioeconomic status (cSES) were evaluated as effect modifiers in birth year adjusted linear regression models. RESULTS: The association between education and PCS was large among high cSES respondents (ß = 0.81 per year of education, 95% CI: 0.67, 0.94), and larger among low cSES respondents (interaction ß = 0.39, 95% CI: 0.06, 0.72). The association between education and MCS was imprecisely estimated among White men (ß = 0.44; 95% CI: -0.03, 0.90), while, Black women benefited more from each year of education (interaction ß = 0.91; 95% CI: 0.19, 1.64). Similarly, compared to socially advantaged groups, low cSES Blacks, and low and high cSES women benefited more from each year of education, while immigrants benefited less from each year of education. CONCLUSIONS: If causal, increases in educational attainment may reduce some social inequities in health.