Examining educational attainment and allostatic load in non-Hispanic Black women | BMC Women’s Health

Study design and participants

We performed statistical analysis with cross-section data from NHANES between 1999 and 2018. NHANES is a nationally representative sample of US adults, where persons aged 60 and older, Latinos and Blacks are oversampled, and weighted analysis generates generalizable population estimates [22]. Since 1959, The National Center for Health Statistics (NCHS) of the Centers for Disease Control and Prevention (CDC) has collected, analyzed, and disseminated data on the health status of US residents [23]. Using stratified multi-stage probability sampling, NHANES enrolls a nationally representative sample of about 5000 non-institutionalized civilians annually. Those selected to participate are initially interviewed in their homes by trained NHANES personnel, who administer questionnaires using computer assisted technology for standardization. One to two weeks after the household interview, participants are requested to visit a Mobile Examination Center (MEC) to complete additional interviews, examinations, and laboratory assessments. NHANES collects demographic, socioeconomic, dietary, health-related questionnaires, and includes clinical measures of blood pressure, fasting blood glucose, triglycerides, and HDL cholesterol, in addition to self-reported medication use for health conditions. We used NHANES years with consistent data on the component variables of allostatic load, i.e., 1999 through 2018 [24]. We excluded participants younger than 18 or pregnant from study. This analysis included all Black women aged 18 and older with available data on allostatic load components, and no missing information on educational attainment, and household income. The resulting 4,177 participants over the 20-year study period served as our analytic population.

Ethics and consent to participate/informed consent

This study was considered exempt by the Institutional Review Boards of Augusta University and St. Cloud State University due to use of existing secondary data that are publicly available and non-identifiable. Original NHANES investigators maintained informed consent for all participants surveyed. Health information collected in the NHANES is kept in strictest confidence. During the informed consent process, survey participants are assured that data collected will be used only for research purposes and will not be disclosed or released to others without the consent of the individual or the establishment in accordance with section 308(d) of the Public Health Service Act (42 U.S.C. 242m).

Outcome variable: allostatic load

Definitions of AL vary though most incorporate biomarker measures from three categories of physiologic functioning: cardiovascular, metabolic, and immune systems [25]. As there is no consensus definition, we elected to define AL using the Geronimus et al. (2006) and Mays et al. (2018) taxonomies [9, 26]. To determine the high-risk thresholds for each AL component, we examined the distribution of each component among Black women for whom complete biomarker data was available in the NHANES study sample (N = 4177 participants). High-risk thresholds were determined as either (a) above the 75th percentile for body mass index (BMI), diastolic blood pressure (DBP), systolic blood pressure (SBP), glycated hemoglobin, total cholesterol, and serum triglycerides; or (b) below the 25th percentile for serum albumin and serum creatinine. Each participant was scored as either 1 (high risk) or 0 (low risk) based on the following cutoffs for each component (See Additional file 1: Table S1): (1) serum albumin ≤ 4.03 g/dL, (2) BMI ≥ 32.1 kg/m², (3) creatinine ≤ 81 μmol/L, (4) DBP ≥ 81.00 mmHg, (5) glycohemoglobin % ≥ 5.86, (6) SBP ≥ 135.00, (7) total cholesterol ≥ 212.00 mg/dL, and (8) serum triglycerides ≥ 145.00 mg/dL. We calculated a total AL score, ranging from 0 to 8, by summing the individual components based on the high-risk thresholds. We further categorized participants with AL scores greater than or equal to 4 as having high AL [25, 26].

Independent variable and other variables of interest

Our independent variable of interest was educational attainment. We categorized NHANES educational attainment into four groups: (1) less than a high school education; (2) high school graduate, GED, or equivalent; (3) some college; and (4) college graduate or above. We were unable to differentiate by degree types due to the NHANES data collection categorization.

We also included other variables typically connected to our area of inquiry. NHANES calculated income relative to federal poverty line (PIR) as the ratio of total family income to poverty threshold values. Persons who reported having had no income were assigned a zero value for PIR. PIR values less than 1 are considered below the official poverty line, whereas PIR values greater than 1 are above the poverty level, and values near 5 are considered very high income. We evaluated personal characteristics that may influence AL score, including age at first menarche, total number of pregnancies, parity, depression, smoking status, waist circumference, and comorbidities. Age at first menarche was defined by self-reported response to “How old were you when you had your first menstrual period?” Total number of pregnancies was determined by self-reported response to “How many times have you been pregnant? Be sure to count all pregnancies including live births, miscarriages, stillbirths, tubal pregnancies or abortions.” We defined parity as the total number of pregnancies resulting in live births. Twins and multiple births counted as a single delivery. During 1999–2004 NHANES survey periods, diagnostic modules were administered that addressed diagnoses of depressive disorders within the past 12 months. Participants were diagnosed with depression according to definitions and criteria of the tenth revision of the International Classification of Diseases (ICD-10) and the fourth edition of the American Psychiatric Association’s Diagnostic and Statistical Manual of Mental Disorders (DSM-IV).

However, for years 2005 through 2018, NHANES obtained depressive disorder through the Patient Health Questionnaire-9 (PHQ-9) [27,28,29]. The PHQ-9 was similarly administered through face-to-face MEC and assessed depressive symptoms over the past two weeks [27,28,29]. Participants indicated on a scale from 0 to 3, the frequency with which they experienced the following symptoms: (1) inability to feel pleasure, (2) depressed mood, (3) sleep disturbance, (4) fatigue, (5) appetite changes, (6) low self-esteem, (7) concentration problems, (8) psychomotor disturbances, and (9) suicidal ideation [27,28,29]. PHQ-9 scores range from 0 to 27 with scores ≥ 10 representing clinically significant depressive symptoms [29]. Using these definitions, we categorized participants in NHANES 1999–2004 as living with depressive disorder if diagnosed based on ICD-10 and DSM-IV, and participants within NHANES 2005–2018 as living with depressive disorder if having a PHQ-9 score ≥ 10. Smoking status was categorized as a discrete, mutually exclusive, three-level variable. Participants who had smoked fewer than 100 cigarettes in their lifetime were categorized as never smokers; participants who had smoked at least 100 cigarettes in their lifetime but were not current smokers were categorized as past smokers. Participants who had smoked at least 100 cigarettes in their lifetime and continued to smoke were categorized as current smokers. NHANES purposefully collected high quality body measurement data participants including waist circumference (in centimeters). We examined these variables as continuous variables. NHANES questioned “Has a doctor or other health professional ever told you that you had … (cancer, angina, congestive heart failure, or heart attack?” Thus, we included any self-reported response to a physician-diagnosed history of cancer, angina, congestive heart failure, or heart attack as comorbidities. These comorbidities were examined as discrete binary variables (i.e., yes or no for each condition).

Statistical analysis

We performed analyses using NHANES generated sampling statistical strata, clusters, and weights as designated and described in detail within the NHANES methodology handbook [22]. We reported categorical variables as weighted row percentages, continuous variables as mean and associated standard errors (SEs), and non-normal continuous variables as median and associated first (Q1) and third quartiles (Q3). The primary outcomes of interest were: (1) dichotomous AL (i.e., high or low) and AL total scores (ranging from 0 to 8). We performed several sequential adjusted modified Poisson regression models for estimating relative prevalence of high AL by educational attainment [30,31,32]. We adjusted for potential confounders including: age groups, total number of pregnancies, age at menarche, PIR, depressive disorder, smoking status, any history of cancer, congestive heart failure, and heart attack [5, 7, 33]. We estimated the variance for point estimates within the modified Poisson regression utilizing the delete-1 jackknife (resampling) method [30,31,32]. Next, we conducted weighted generalized linear models to examine the association between the AL total score and educational levels adjusted for age [34, 35]. We conducted sensitivity analysis to examine whether the association between educational attainment and prevalence of high allostatic load were modified by age groups. Estimates derived from log-binomial are presented as prevalence ratios (PRs) and associated 95% confidence intervals (CIs), and estimates derived from generalized linear regression models are presented as mean allostatic load estimates and associated 95% CIs. We considered p values less than 0.05 as statistically significant. We used SAS software version 9.4 (copyright © 2013 SAS Institute Inc., Cary, NC, USA) for all analyses.