Racial discrimination, the superwoman schema, and allostatic load: exploring an integrative stress-coping model among African American women

Abstract

Racial discrimination has been linked to allostatic load (i.e., cumulative biological stress) among African American women. Coping is a central component of the stress response process. However, limited attention has been given to psychosocial processes involved in the stress response—critical for understanding biological pathways to health—in studies examining racial discrimination as a social determinant of health. We examined whether the superwoman schema (SWS), a multidimensional culture-specific framework characterizing psychosocial responses to stress among AA women, modifies the association between racial discrimination and allostatic load. We used purposive sampling to recruit a community sample of African American women ages 30–50 from five San Francisco Bay Area counties (n = 208). Path analysis was used to test for interactions while accounting for the covariance among SWS subscales using both linear and quadratic models. Significant interactions were observed between racial discrimination and four of the five SWS subscales. Feeling obligated to present an image of strength and an obligation to suppress emotions were each protective whereas feeling an intense motivation to succeed and feeling an obligation to help others exacerbated the health risk associated with experiencing racial discrimination. Our findings affirm the need to consider individual variability in coping and potentially other psychosocial processes involved in the stress response process; and offer several insights that may help elucidate the mechanisms by which racial discrimination gets “under the skin”.

Introduction

Social stress and integrated physiology

Chronic social stress plays a critical role in the progression of dysregulated physiologic systems via prolonged activation of the body’s primary stress response system—combined activity of the hypothalamic–pituitary–adrenal (HPA) axis and sympatho-adrenal-medullary (SAM) axis.^1–5 Though short-term stress can have beneficial/protective effects, prolonged activation of the body’s stress response can lead to sustained arousal and impaired plasticity across a number of physiologic systems (i.e., neuroendocrine, cardiovascular, metabolic, and immune) compromising the body’s ability to adaptively respond to stressors.^6–12 Allostasis is an active regulatory process involving adaptation among integrated biological systems intended to meet changing environmental demands.^13–18 In such a model, physiologic flexibility rather than rigidity is favorable.^13–18 However, prolonged adaptation to sustained arousal precipitated by chronic stress can lead to increased allostatic load.^1,2,14,16

Studies show significant associations between various forms of psychosocial stress and a variety of biological parameters responsible for maintaining physiologic balance.^9,12,19–21 Although definitions of stress have varied over time, contemporary scholars generally understand stress as the dynamic interplay (i.e., transactions) between a person and their environment, involving environmental demands, psychosocial processes that determine appraisal of threat/harm/loss ([di]stress), or alternatively motivation/challenge ([eu]stress), and concomitant biological adaptation.^22–24 Hence, although stress is not inherently positive or negative, scholars using a risk-based (versus resilience-based) model emphasize the imbalance between perceived demands and perceived resources resulting in a deleterious cognitive, affective, and ultimately physiologic state. Accordingly, Cohen et al. define stress as the “process in which environmental demands tax or exceed the adaptive capacity of an organism, resulting in psychological and biological changes that may place a person at risk for disease”.²² Similar to allostatic load, a key aspect of this definition and, respectively, the concept of person-environment fit, is the idea of adaptive capacity, one’s actual and perceived capacity for lessening the harmful impact of environmental stressors, i.e., coping.

Coping involves two components, dispositional coping-styles, relatively stable traits (cognitive and affective) that determine one’s usual interactions with the environment, and context-dependent coping – conscious strategies to regulate emotion, cognition, behavior, and/or physiology in response to specific stressful encounters.^25–28 Studies show little correlation between measures capturing these two components of coping.²⁷ Hence, not only do these components capture different aspects of coping, but they may intersect in ways that have implications for health. However, research on coping has generally fallen into one of the two camps without considering the ways in which the two interrelate. Understanding how dispositional coping styles relate to different types of stressors may elucidate understanding of biological sensitivity to social context.

Racial discrimination, coping, and integrated [psychobiological] specificity

Social evaluative threat (SET), such as that experienced with racial discrimination, is a particular form of stress that studies show is associated with biological disruption via dysregulation of both the HPA and SAM axes.^29,30 Until recently, research on stress and health has been largely based on Selye’s general adaptation syndrome, which suggests that the physiologic stress response is non-specific and includes a common set of physiologic responses for any given stressor.³¹ This would suggest that experiencing a form of SET, such as racial discrimination, and something more innocuous such as giving a speech would elicit the same physiologic stress response. However, studies show this is not the case.^29,32 Evidence suggests that rather than having uniform effects on physiology, stressors are met with integrated cognitive and affective responses that determine the physiologic response to stress.^14,29,30,33 In particular, studies suggest that specific stress exposures (e.g., racial discrimination) in combination with specific cognitive stress appraisals (e.g., threat vs. challenge) and coping-styles determines the biological stress response (i.e., integrated specificity).³⁰ Additionally, coping-style influences cognitive stress appraisal and thus may have a particularly meaningful impact on physiologic adaptation.^24,26,34,35

Numerous studies show that African American (AA) women report racial discrimination as a unique and salient form of social stress,^36–39 and report using a combination of active and passive as well as problem- and emotion-focused coping-styles in response to racial discrimination relative to other forms of social stress.^36,37,40,41 This finding is novel given dominant views of coping as categorical (e.g., active versus passive, emotion- versus problem-focused) suggesting that AA women do not fit neatly into one category; and calls attention to the need for measures that capture the synergy of coping-styles among AA women, especially in relation to context-dependent situations such as racial discrimination, which may engender particularly intense cognitive appraisals and affective responses.

African American women, schema development, and gendered racial identity

Schemas are an important aspect of social cognition and play a fundamental role in shaping the coping response.^42–47 Several studies have documented adoption of what has become known as the strong black woman role or, more recently, the superwoman schema (SWS) conceptual framework among AA women.^40,48–51 Schemas are internal representations of the self, of the social world and of the self in relation to the social world.^42,43,45 Schemas are mental models— a network of core beliefs— derived from past experience, consisting of generally stable patterns of cognitions, emotions, and memories that aid in the processing and interpretation of information. Schemas are often activated and operate subconsciously (i.e., habitual), especially within the context of chronic or repeated activation.^45,46 Triggered by situational cues in the environment, schemas aid in stress appraisal and influence coping responses (e.g., integrating dispositional- and context-dependent coping); and ultimately affect behavior and psychophysiologic stress responses, as well as future expectancies (e.g., anticipatory threat).^{28,42,43,45,46,52,53} Schemas play a fundamental role in how people experience, interpret and respond to the social world. Hence, they are an essential component of various mechanisms linking stress to health.

Culture influences how people experience, interpret and respond to the social world.^42,46,47 Thus, certain schemata may be common to members of a specific cultural group due to “shared histories.”^28,46,47 Schemas also reflect individual differences in information processing (e.g., beliefs, values, exposures, coping resources, personal salience).^28,47,54 Hence, schemata are shaped by both individual and social context, and provides “interpretive context”⁴² for assessing experiences, particularly those appraised as personally relevant.^{42,45–47,55}

African Americans report greater overall stress exposure relative to their white counterparts, including experiences of racial discrimination. African American women have reported racial discrimination as a particularly salient form of psychosocial stress. Specifically, experiences of racial discrimination, especially those experienced early in life, have been described as pivotal to their understanding of themselves in relation to the broader social world––that is, stigmatized, stereotyped, excluded, and devalued.³⁷ African American women identify racial discrimination as a persistent stressor occurring throughout their life across numerous life domains.³⁷ These experiences are described as being particularly painful, having long-lasting effects on their self and group identities and on how they perceive situations involving encounters with others, particularly whites.³⁷ They describe ruminating on past experiences, putting on what has been called their “armor” in anticipation of future threats, and feeling the need to overcompensate for negative stereotypes about African Americans and about AA women in particular (e.g., working harder to prove themselves, suppressing emotions, and code switching).^37,56–58 These cognitive, affective, and behavioral responses are, therefore, triggered by a combination of past experiences and current situational/environmental cues that activate expectancies about the present situation.^37,42,43 These expectancies convey for many the need to engage in what some have referred to as the “invisible labor” that characterizes AA women’s lives across a variety of social contexts.^59,60 The SWS framework provides cultural context for the unique ways AA women experience and respond to life stress, including but not limited to racism, sexism, and gendered racism. However, the personal relevance of gendered racism for AA women may intensify and prolong schema activation potentially resulting in impaired cognitive flexibility and hence limit capacity for adaptive coping.^{37,43,44,55,58,61–64}

African American women have been characterized by what has been called “psychological androgyny”, embodying at once the traditionally masculine traits of self-reliance, independence, hard work and achievement, assertiveness, and obstinate strength⁶⁵ while also assuming the role of nurturer and caregiver.^66–69 These “habits of surviving”⁷⁰ form the “armor”⁵⁷ AA women have been socialized to wear, as a strategy for maintaining psychological resilience in socially threatening situations.^{68,69,71–73} These gendered racial socialization experiences begin early in life and are reinforced throughout the lifecourse.^{50,71,72,74–77} Hence, the strong black woman role, or SWS framework, has been described as a “collective [and ubiquitous] feature of African American womanhood”.⁶⁶ One study examining racial and gender identity among AA women asked women to rate the importance of being “Black,” being a “woman,” and being a “Black woman” to their identity. Results showed that being a Black woman was significantly more important that being a woman and marginally more important than being Black,⁵⁴ suggesting a unique and salient identity related to the intersection of race and gender among AA women (for detailed discussion, see Supplemental Information, online only).

Being a superwoman: tonic or toxic?

The strong black woman or SWS framework has its roots in the black feminist movement of the 1960s and 70s. During this time, social movements tended to focus on unitary systems of power (e.g., civil rights, black power, feminism/women’s liberation, worker’s rights), subordinating the unique challenges black^a women faced across axes of race, class, and gender.⁷⁸ Within this context, the core beliefs of intersectionality were born. Among these core beliefs are notions of power (interpersonal, cultural, structural), inequality, relationality, and social context. As both a conceptual and analytic tool, intersectionality situates black women in an autonomous social position and motivates understanding of their experiences from what Collins calls a “distinctive angle of vision” where black women are at once gendered and racialized.^78,79 Referring to this intersection, the Combahee–River–Collective’s “A Black Feminist Statement” reads, “The synthesis of these oppressions creates the condition of our lives.”⁷⁸ Stemming from the mutually constructed disadvantages associated with these intersecting systems of power and privilege, the strong black woman or superwoman schema is understood as a strategy for withstanding the social adversity related to being a member of a dually-oppressed group.^78–81 Accordingly, AA women were urged to “possess the spirit of independence…possess the spirit of men, bold and enterprising, fearless and undaunted”.⁷⁹

Dating back to the mid-19^th century Sojurner Truth became a symbol of strength and resilience among AA women.^82,83 Her speech, “Ain’t I a Woman?” (c.1851), reflects her social location as an AA woman as she describes her need to develop a toughness that enabled her to “work [the fields]…as much as a man….and bear the lash [endure the pain/whippings of slavery] as well!” while simultaneously bearing the responsibility of motherhood and caretaking; despite being denied the simple pleasures of womanhood such as being “helped into carriages” and “lifted over ditches”.⁸⁴ As reflected in these excerpts, black women have historically been denied the normal conveniences of womanhood due to their inferior racial status, often treated like their male counterparts, and prohibited from expressions of weakness or vulnerability for the sake of survival. This image of the strong black woman has persisted throughout generations and is described, even today, as an asset that allows AA women to manage their lives in a race- and gender-conscious society.^36,37,40,48

According to the SWS framework, the coping styles adopted by AA women exist within a unique sociocultural and historical context and influence the ways in which they experience stress, including but not limited to race-related stress (i.e., racial discrimination, gendered racism). The framework is based on previously collected qualitative data among a socioeconomically diverse sample of AA women ages 19–72.⁴⁰ In response to questions about experiencing and coping with stress, participants commented on the need to be strong indicating that it was “just part of a woman’s life, and more specifically, an African American woman’s life”. Among AA women, being strong was described as (1) feeling an obligation to present an image of strength, even when one didn’t feel strong (SWS_strength), (2) feeling an obligation to suppress emotions (SWS_emosupp), (3) resistance to being vulnerable or dependent on others (SWS_vuln), (4) determination to succeed despite limited resources (SWS_succeed), and (5) obligation to help others (SWS_help). Hence, the SWS is conceptualized as a double-edged sword for AA women, providing them with the fortitude and determination to withstand and support others in the context of race and gender specific stressors, while simultaneously placing them at greater risk for stress-related health outcomes. For example, some aspects of the SWS (e.g., resistance to being vulnerable or dependent on others such as not asking for help) may lead to excessive or chronic use of internal resources resulting in the “wear and tear” of physiological mechanisms meant to be compensatory in acute stressful situations.⁸⁵ Whereas the SWS may be adaptive initially, it has been described as ”an overused asset that develops uncritically without ongoing evaluation and attention to changing needs and demands [that] runs the risk of becoming a liability”.⁸⁶

Although some research suggests that caregiving and volunteering positively impact overall health and longevity,^87,88 caregiving activities—particularly within a context of inadequate resources, perceived obligations to suppress emotion and resistance to vulnerability—that require AA women to habitually and chronically delay their own self-care may have deleterious effects on health.^40,89 For example, prioritization of caregiving can result in excessive fatigue and inadequate time to engage in health promoting activities such as exercise and preparing meals. Importantly, although these characteristics and perceived obligations may contribute to limited self-care, they are critical and central components of self-concept for many AA women, which may itself be protective.⁴⁰ However, aspects of the SWS may also lead to chronic psychological distress, which is associated with physiological processes, including activation of the hypothalamic–pituitary–adrenal (HPA) axis, chronic inflammation, and abdominal obesity, which increases risk for cardiometabolic diseases.^1,2,40,90

Some studies suggest that AA women may be particularly vulnerable to the psychological and physiologic effects of racial discrimination due to their endorsement of the caretaker role. For example, AA women report significant psychological distress as a result of their own discrimination experiences as well as the discrimination experiences of others in their social network, especially their spouses and children.^{36,37,41,91–93} Indeed, studies show that whereas men tend to be more “egoistic”, women report significantly higher levels of “network distress,” thereby increasing their overall stress burden.^94–96 Studies show that chronic stress reduces one’s capacity for coping with future stressors.^42,61,97,98 Thus, the heightened stress experienced by AA women due to both self and network stress has important implications for reducing their capacity for coping effectively and for their long-term health and well-being. In their seminal study of network stress, Kessler and McLeod (see Ref. 64) found that network events accounted for 75% of the difference in distress between men and women in the labor force, leading researchers to conclude that considering their role as caretakers, nurturers, mothers and “other mothers” (i.e., mothers to the extended AA community) is essential when contextualizing the stress experiences of AA women.^95,96

Accordingly, AA women describe the SWS as a liability, in part, due to the postponement of self-care. In one study of AA women in the Southern United States, one woman stated, “I would work long hard hours because I wanted to make a good appearance but it didn’t make a difference, I was still a black person.”³⁶ In another study, AA women described the need to “work harder” to prove themselves because of their low social status. They also described the need to suppress their emotions when anticipating racial discrimination: “I know this person is going to say something that’s going to make me, my heart rate go up, or maybe have to hold back my tears while I’m talking to them”.³⁷ In other work, when AA women were asked about factors contributing to their adoption of the superwoman role, they identified several factors including: the historical legacy of racial and gender stereotyping, lessons from foremothers, past history of mistreatment or abuse, and spiritual values. One woman noted, “Our past makes us have to be a strong woman and it’s annoying as hell”. Another commented, “People always say, ‘Oh you look so calm…and I’m thinking I’m just about to crumble in two seconds and I think a lot of people don’t know when black women are stressed because of the Superwoman Syndrome…”.⁴⁰ These quotes illustrate the dissonance between the image AA women feel they must portray and what they feel they can truly bear. In a qualitative study examining gendered racism among AA women, one woman explained: “We feel an obligation not only to ourselves but our family and our community to keep it strong and to keep it together”.³⁶ These quotes represent samples from three different qualitative studies conducted with AA women in various regions of the U.S. suggesting that being a superwoman is not a singular phenomenon, but perhaps a schema shared by many.

Summary and study hypotheses

Previous studies have demonstrated associations between discrimination and a range of individual biomarkers, most consistently cortisol and measures of inflammation. The majority of these studies have examined general discrimination without attention to the specific type of discrimination reported.^{6,9,11,99–101} However, previous evidence indicates that the specific type of stressor matters in determining the biological response to stress (i.e., integrated specificity).³⁰ AAs attribute the majority of their discrimination experiences to race.¹⁰² Moreover, AA women report racial discrimination as a particularly salient and chronic psychosocial stressor over their lifecourse, and as distinct from other forms of unfair treatment.^36,37,40 Additionally, Previous research shows associations between racial discrimination and allostatic load among AA women.¹⁰³ Accordingly, researchers have called for measuring racial discrimination as a particular form of SET, and considering intersectionality in both measuring and assessing the health effects of racism.¹⁰⁴ Scholars also suggest a need to study simultaneous dysregulation across multiple systems rather than individual biomarkers to better understand how psychosocial factors contribute to stress-related illness.⁷ Last, the transactional model of stress (TMS) suggests that coping is a central component of the stress response process.^22,24 However, limited attention has been given to the role of coping or schemas, which are central to coping processes, in studies examining racial discrimination as a social determinant of health.

The strong black woman or SWS develops over time starting early in life.^{43,46,50,71,72} Though schemas are activated by situational cues, studies also suggest that (1) schemas can become stabilized over time and are activated not only as a result of situational cues but become enduring aspects of the “self” (i.e., dispositions) that are invariant across situations;^44,45,47 (2) racial discrimination is a chronic stressor among AA women;^{37,103,105,106} (3) when salient and frequent, discrimination can make relevant schemas more accessible, resulting potentially in chronic accessibility or activation;⁴³ and (4) schemas influence threat appraisals,^43,61 which influence the biological stress response.^22–24 Given our sample of midlife AA women, many of whom report having experienced frequent racial discrimination over their lifecourse, we hypothesize the SWS as a relatively stable disposition that moderates the association between racial discrimination and allostatic load. Hence, the purpose of this study was to examine whether Superwoman Schema modifies the association between racial discrimination and allostatic load; and whether those effects are uniform or differ by the SWS dimension. As a multidimensional construct with potentially protective as well as harmful effects, we hypothesize that the various SWS dimensions will have differing effects on the association between racial discrimination and allostatic load. Given that exhibiting strength has both health protective and health damaging properties,⁴⁰ we did not specify hypotheses about the direction of association for SWS_strength. However, previous research indicates that anger is the most common response to racial discrimination¹⁰⁷ and has further shown the particularly health damaging effects of anger, versus other emotional responses (e.g., sadness).¹⁰⁸ Hence, within the context of racial discrimination we hypothesize that emotion suppression will be health protective. Additionally, for the reasons noted above, we hypothesize that SWS_vuln (e.g., not asking for help), SWS_succeed (e.g., “working harder”) and SWS_help (i.e., taking care of others) will have unfavorable health consequences within the context of increasing racial discrimination. Given the exploratory nature of this study, our aim was to explore, for the first time, whether the SWS modifies the association between racial discrimination and allostatic load among midlife AA women, and if so, the nature of those relationships. Therefore, these are preliminary hypotheses only, meant to inform the interpretation of our study findings and inform the generation of hypotheses to guide future work.

Methods

Study design and participants

Data are from the African American Women’s Heart & Health Study, an observational cross-sectional study designed to examine associations between socio-environmental stressors and mental and physical health among a community sample of midlife AA women residing in the San Francisco Bay area (n = 208). The study design has been described in detail elsewhere.¹⁰⁹ Briefly, we (first author) used purposive sampling to recruit a socioeconomically and geographically diverse sample of AA women from five Bay area counties: Alameda, Contra Costa, San Francisco, and Solano. Eligibility criteria were (1) self-identify as AA, (2) female gender since birth, (3) ages 30–50, (4) U.S.-born, (5) parent(s)/primary caregiver(s) U.S.-born AA, and (6) English literacy. Women who were pregnant or lactating, or who self-reported a physician-diagnosed inflammatory or auto-immune disease were excluded to minimize confounding. Recruitment took place March 2012 through March 2013.

Participation consisted of two study visits. Visit 1 included an interviewer-administered questionnaire—health history, housing and other sociodemographics, and general health status—and a computer-assisted self-interview (CASI). The CASI was used to enhance confidentiality and minimize potential response bias (i.e., acquiescence and social desirability bias). Visit 2 included a physical exam and fasting venous blood draw. Participants were compensated with a $70 gift card, study t-shirt, heart-healthy cookbook, and a packet of health promoting materials. Informed consent was obtained from all study participants. The study was approved by the Office for the Protection of Human Subjects at the University of California, Berkeley.

Measures

Allostatic load.

Allostatic load was calculated as a composite of 15 biomarkers—primary mediators and secondary outcomes associated with stress-related pathophysiologic processes—reflecting activity of the HPA axis, SAM axis, inflammatory, cardiovascular and metabolic systems.^2,5 Primary mediators include cortisol, epinephrine and norepinephrine, Interleukin-6, and C-reactive protein. Cortisol reflects HPA-axis activity and epinephrine and norepinephrine represent activity of the SAM-axis, which together comprise the body’s primary stress response system (“fight or flight”). Interleukin-6 (IL-6) is a pro-inflammatory cytokine commonly elevated as a result of HPA and SAM activity. C-reactive protein (hsCRP) is an acute phase reactant released from the liver and found in plasma in response to inflammation. The remaining 10 biomarkers represent secondary outcomes—resulting from the cumulative effect of primary mediators and primary outcomes (i.e., cellular events). These include: systolic and diastolic blood pressure (SBP and DBP, respectively) reflecting cardiovascular activity; and waist circumference, body mass index (BMI), glucose, glycosylated hemoglobin (HbA1c; average glucose elevation over previous 2–3 months), high- and low-density lipoprotein (HDL and LDL, respectively), triglycerides, and total cholesterol, all indicative of metabolic function.

Resting SBP and DBP were measured from the sitting position after a 5-minute rest period using an automated oscillometric blood pressure monitor. We took four consecutive measures in 1-minute intervals. The first measure was discarded and the average of the last three measures was recorded.¹¹⁰ Waist circumference was measured just above the top of the iliac crest, according to standard protocol.¹¹¹ Body mass index was assessed via bioelectrical impedance using a handheld body fat analyzer. All other biomarkers were assayed from fasting venous blood specimens stored at –80 °C by one of two CLIA certified laboratories.

Table 1 displays the cutpoint for each biomarker. We used a combination of clinically-based cutpoints and cutpoints based on the distribution of the study sample to ascertain level of risk.¹¹² Consistent with the concept of allostatic load, established cutpoints for subclinical risk were used where available (e.g., pre-hypertension: 120 for SBP and 80 for DBP; pre-diabetes: HbA1c = 5.7 and glucose = 100). Otherwise, risk was defined as being above the 75^th-percentile for the selected biomarker as has been done in prior research.^4,5,113 Each biomarker was dichotomized at the specified cutpoint (0 = not at risk, 1 = at risk) and then summed. Scores range from 0–15, where higher scores reflect higher levels of allostatic load.

Table 1.

Allostatic load biomarker cutpoints

Biomarker	Guideline used	AL cutpoints	% Above cutpoint
Metabolic system
HDL (mg/dL)	ATPIII	<50	41.06
LDL (mg/dL)	ATPIII	≥100	52.17
Waist Circumference (in)	ATPIII	>35	72.46
Glucose (mg/dL)	ATPIII	≥100	17.39
HbA1c (mmol/mol)	ADA	≥5.7	19.32
Total Cholesterol (mg/dL)	ATPIII	≥160	66.18
Triglycerides (mg/dL)	ATPIII	≥150	7.25
BMI (kg/m2)	ATPIII	≥85.51
Cardiovascular system
Systolic BP (mm Hg)	AHA	≥120	47.34
Diastolic BP (mm Hg)	AHA	≥80	47.83
Neuroendocrine system
Cortisol (μg/dL) a	n/a	>12.69	25.12
Epinephrine (pg/mL) a	n/a	>77.70	24.15
Norepinephrine (pg/mL) a	n/a	>686.30	25.12
Inflammatory system
IL-6 (pg/mL)a	n/a	>7.85	29.47
hsCRP (mg/L)	AHA	>3	49.28

Biomarker abbreviations: HDL = High-density lipoprotein; LDL = Low-density lipoprotein; HbA1c = Hemoglobic A1c (glycated hemoglobin); BMI = Body mass index; Systolic BP = Systolic blood pressure; Diastolic BP = Diastolic blood pressure; IL-6 = Interlukin 6. hsCRP = High-sensitivity C-reactive protein.

SI conversion factors: To convert HDL to mmol/L, multiply values by 0.0259; LDL to mmol/L, multiply values by 0.0259; glucose to mmol/L, multiply values by 0.0555; Total Cholesterol to mmol/L, multiply values by 0.0259; Triglycerides to mmol/L, multiply by 0.0113; Cortisol to nmol/L, multiply by 27.588; Epinephrine to pmol/L, multiply by 5.459; norepinephrine to pmol/L, multiply by 5.911; IL-6 mmol/L, multiply by 2.539 e¹³; hsCRP to nmol/L, multiply by 9.524.

^a75th percentile cutpoints used for biomarkers that don’t have established clinical guidelines.

Racial discrimination.

We used a modified version of the Experiences of Discrimination scale.¹¹⁴ Eight items assessed how often respondents “have ever been treated unfairly, judged differently than others, prevented from doing something, or hassled or made to feel inferior because of their race, ethnicity, or skin color” in any of eight different situations (i.e., at school, getting hired or getting a job, at work, getting housing, getting medical care, getting credit, on the street or in a public setting, from police or in the courts; α = 0.92). Responses were coded on a 5-point Likert scale ranging from 1 = “Never” to 5 = “6 or more times”. Summary scores were averaged generating a final score ranging from 1 to 5.

The superwoman schema.

The SWS is a 35-item scale comprising five subscales, each representing a different dimension of SWS: (1) obligation to present an image of strength (6 items, e.g., “I try to present an image of strength”, α = 0.76); (2) obligation to suppress emotions (7 items, e.g., “My tears are a sign of weakness”, α = 0.87); (3) resistance to being vulnerable (7 items, e.g., “Asking for help is difficult for me”, α = 0.84); (4) intense motivation to succeed (6 items, e.g., “No matter how hard I work, I feel like I should do more”, α = 0.77); and (5) obligation to help others (9 items, e.g., “I put everyone else’s needs before mine”, α = 0.89).^115,116 Relevant items were reverse coded for consistent valence. Responses for each subscale were coded on a 4-point Likert scale ranging from 0 = “This is not true for me” to 3 = “This is true for me all the time”, summed across items, and averaged to generate a final score.

Covariates.

Potential confounders were theoretically informed and based on prior literature: age in years, educational attainment (≤ versus > high school diploma), poverty status categorized as ≤ or > 100% of the federal poverty threshold (FPT), employment status (employed versus not employed), marital status (married/domestic partner versus not married/no domestic partner), health insurance (insured versus uninsured), and the neuroticism sub-scale of the Big Five Personality Inventory. We also adjusted for medication use (yes/no) to ensure accurate risk assessment. Health behaviors such as smoking, alcohol use, diet and physical activity were conceptualized as mediators and therefore excluded from the moderation analysis to avoid overcontrolling as we were interested in assessing total (direct) effects. All dichotomous variables were coded 0 for the not-at-risk group and 1 for at-risk.

Analytic strategy

Multiple imputation was used to account for missing data. The majority of the sample (86%) had less than 5% missing data. One observation, with 32% missing, was excluded because it had missing data on both the right- and left-hand side of the imputation model, and therefore had missing data post-imputation (n = 207). Relative efficiency ranged from 98–100%, and relative variance increase and percent increase in standard error were both within 5%.

Univariate and bivariate analyses were conducted to assess variable distributions, crude associations between study variables, test for confounders, assess collinearity among study variables and ensure enough unique variance among SWS subscales to estimate independent effects. Correlations between SWS subscales were in the low to moderate range (Table 2). Next, we conducted path analysis within a structural equation modeling framework to test for interactions according to the specification in Figure 1. We tested interactions between racial discrimination and each SWS subscale concurrently, while accounting for the covariance among subscales. We first tested a linear interaction model, estimating the main effects of racial discrimination and the five SWS subscales, centered and their means, as well as the interaction between racial discrimination and each of the five SWS subscales on allostatic load. Next, given prior evidence of a non-linear association between racial discrimination and allostatic load, we tested a quadratic interaction model by introducing a quadratic term for racial discrimination and testing the interaction effects between the quadratic racial discrimination term and each of the SWS subscales. For all significant interaction effects, we conducted simple slope analyses to examine the extent to which racial discrimination was associated with allostatic load when a given SWS subscale was high (i.e., one standard deviation above the mean) versus low (i.e., one standard deviation below the mean).¹¹⁷

Table 2.

Zero-order correlations of SWS subscales, discrimination, allostatic load, and covariates

	1		2		3		4		5		6		7		8		9		10		11		12		13		14		15		16		17		18		19		20		21
1. Strength
2. Emotion	.19	**
3. Vuln	.26	***	.57	***
4. Succeed	.47	***	.43	***	.60	***
5. Help	.32	***	.45	***	.59	***	.59	***
6. EOD	.15	*	.03		.17	*	.18	**	.27	***
7. AL	.01		.01		−.05		−.03		.01		−.06
8. Med_CV	.04		−.01		−.03		.00		.02		.04		.24	**
9. Med_DB	.00		.03		.08		.03		.06		−.02		.23	***	.13
10. Education	−.03		.17	*	−.05		−.04		.07		−.12		.22	**	−.01		−.04
11. Poverty	−.11		−.17	*	−.19	**	−.15	*	−.08		.08		.06		.00		−.01		.05
12. Age	.05		−.13		−.10		−.03		−.09		.13		.27	***	.23	***	.09		.05		−.05
13. Smoking	−.01		.10		.01		−.05		.11		.12		.09		.04		.04		.30	***	.16	*	−.05
14. Alc use	−.03		.05		−.01		−.02		.04		.01		.08		.16	*	.10		.09		.03		−.10		.09
15. Exercise	−.07		.10		.08		.06		.01		−.03		−.01		.01		−.03		.01		−.26	***	.00		−.21	**	−.06
16. Coping	.07		.06		.21	**	.06		.24	***	.29	***	.13		.09		.00		−.05		.16	*	.08		.27	***	.12		−.03
17. Fruit	.08		.11		.09		.03		.18	**	.04		.02		−.08		−.03		.14	*	−.07		−.03		.18	**	.02		−.22	**	.02
18. Employ	.13		.06		.04		−.01		−.07		−.09		−.01		−.02		−.11		−.27	***	−.09		−.05		−.35	***	−.05		.14	*	−.10		−.19	**
19. Marital status	−.02		−.10		−.08		−.10		.05		.00		.00		.14		−.02		.08		−.04		.05		.23	**	−.01		−.09		.03		−.01		−.14	*
20. Health ins	−.01		−.05		−.02		−.07		.00		.06		−.02		.09		.01		−.18	*	−.16	*	−.05		−.07		−.08		.12		−.04		−.01		.05		.13
21. Neurotisism	.07		.12		.30	***	.19	**	.36	***	.21	**	−.02		.14	*	.08		.08		.09		−.10		.16	*	.09		−.02		.35	***	.06		−.13		−.08		.05
Mean	2.31		1.33		1.56		1.88		1.61		2.41		6.02		.43		.18		.64		2.57		.21		.06		.33		.19		.00		.30		.55		.30		.73		3.07
Variance	.38		.55		.56		.47		.64		1.27		4.46		.25		.15		.23		1.22		.17		.06		.22		.15		34.62		.21		.25		.21		.20		.57
SD	.61		.74		.75		.69		.80		1.13		2.11		.49		.39		.48		1.10		.41		.23		.47		.39		5.88		.46		.50		.46		.44		.76

Note. Vuln = vulnerability. EOD = racial discrimination. Med_CV = medication for cardiovascular diseases. Med_DB = medication for diabetes. Alc use = alcohol use. Employ = employment. Health ins = health insurance.

^*p < .05.

^**p < .01.

^***p < .001.

Figure 1.

Path analysis model.

Models were adjusted for age, educational attainment, poverty status and medication use regardless of significance to account for known or theoretical confounders and ensure accurate risk assessment. Remaining covariates were included in final models if significant at p < 0.10. Employment, poverty status, marital status, health insurance, and neuroticism were not significantly associated with allostatic load in bivariate analysis (p ranged from 0.29 to 0.95) and were not statistically significant in multivariable models. Additionally, results were unchanged after including these variables. For parsimony, we excluded them in the final models.

As a secondary aim, we assessed whether smoking, alcohol use, diet and physical activity mediated the association between racial discrimination, SWS and allostatic load to better understand the potential mechanisms driving associations with health.¹¹⁸ Additionally, to test an alternative hypothesis also supported by the literature, we assessed whether SWS may operate as a mediator rather than a moderator of the association between racial discrimination and allostatic load. Finally, we conducted a sensitivity analysis related to risk classification for allostatic load where participants were coded as at-risk for values considered clinically low or high, given the health risk associated with low values for select biomarkers: BMI (<18.5/>25), glucose (<70/>100), systolic blood pressure (<90/>120) and diastolic blood pressure (<60/>80). There were no appreciable differences in the study findings (data available upon request). Hence, consistent with previous work,^4,85 we present findings for allostatic load coded based on high values only.

The recommended sample size for path analysis is >200. Hence, all models had adequate power to detect differences.¹¹⁹ Two observations for SWS_strength were removed because they were outside of the 3 SD range with z values of –3.76 and –3.22; and one for allostatic load with a z value of 3.28 was also removed. Multiple imputation was conducted using Stata SE v14. All other analyses were conducted using Mplus 8. In Mplus, the non-normality in the data is handled by the maximum likelihood parameter estimates with standard errors (MLR) and is robust to outliers.¹²⁰ Significance was assessed at p < 0.05.

Results

Sample description

Sample characteristics are shown in Table 3. Mean age was 42 years. Approximately two-thirds of the study sample had more than a high school education, over half were employed, 73% had health insurance and 81% were above the FPT. The sample was generally healthy with a mean level of allostatic load in the low to moderate range (6.1 + 2.23). Mean level of racial discrimination was also in the moderate range (2.41 + 1.13). Racial discrimination experienced “on the street or in a public setting” was the most common form of racial discrimination reported followed by “from the police or in the courts” and “at work”, “getting hired or getting a job”, and “at school”: 73%, 67%, 67%, 65%, and 61% respectively (Supplementary Fig. S1, online only). The majority of women reported moderate to high levels of SWS across the five subscales (Mean ranged from 1.33 to 2.31).

Table 3.

Sample characteristics (n = 207)

Covariates	n	%
Age (mean, SD)	41.72	5.90
Educational Attainment
> High School Diploma	138	66.67
≤ High School Diploma	69	33.33
Poverty Status
> 100% FPL	168	81.16
≤ 100% FPL	39	18.84
Employment Status
Employed	114	55.07
Not Employed	93	44.93
Health Insurance
Insured	152	73.43
Not Insured	55	26.57
Marital/Domestic Partnership Status
Married/Domestic Partnership	61	29.47
Not Married	146	70.53
Taking Cardiovascular Medication
No	164	79.23
Yes	43	20.77
Taking Diabetes Medication
No	195	94.20
Yes	12	5.80
Neuroticism (mean, SD)	3.08	0.75
Racial discrimination
	Mean	SD
	2.41	1.13
Super woman schema
	Mean	SD
Strength	2.31	0.61
Suppress emotions	1.33	0.75
Resistance to Vulnerability	1.56	0.75
Motivation to Succeed	1.88	0.69
Help Others	1.61	0.80
Allostatic load
Allostatic Load (mean, SD)	6.02	2.11

Exploratory associations between study variables

Table 2 shows zero-order correlations between the study variables. Correlations between SWS measures are in the low to moderate range (r = 0.18 to 0.60).¹²¹ Being in poverty showed a significant but weak negative correlation with SWS (SWS_emosupp, SWS_vuln, SWS_succeed). There was also a weak but significant correlation between lower educational attainment and both higher SWS_emosupp and higher allostatic load. Higher levels of SWS_vuln and SWS_help were correlated with using food or substances to cope with stress (r = 0.21 and 0.24, respectively), as was racial discrimination (r = 0.29). With the exception of SWS_emosupp, racial discrimination showed a weak but significant positive correlation with SWS. In these bivariate correlations, neither racial discrimination nor any of the SWS measures was significantly correlated with allostatic load.

Linear assessment of racial discrimination × SWS

Table 4 displays results of the two path models, one based on the linear assessment of racial discrimination and the other based on the quadratic assessment of racial discrimination, with all SWS subscales estimated concurrently. For the linear assessment of racial discrimination, there was a significant interaction between racial discrimination and SWS_emosupp. Higher levels of racial discrimination were associated with lower allostatic load among those reporting higher (versus lower) levels of SWS_emosupp (B = −0.41, SE = 0.17, p = 0.02, β = −0.17).

Table 4.

Interaction effects between racial discrimination and SWS subscales on allostatic load

	Linear model (R2 = 0.219)						Quadratic model (R2 = 0.295)
	B	SE	p		β		B	SE	p		β
Discrimination	−.20	.12	.09		−.11		.05	.15	.73		.03
Discrimination (Q)							−.28	.10	.00		−.21
SWS
Strength	.11	.26	.67		.03		−.32	.34	.34		−.09
Emotion	.04	.22	.87		.01		−.42	.32	.18		−.15
Vulnerability	−.15	.24	.54		−.05		−.08	.34	.81		−.03
Succeed	−.03	.29	.92		−.01		.60	.33	.07		.19
Help	.12	.20	.55		.05		.39	.25	.11		.15
Interaction
Discrimination × Strength	.00	.25	.99		.00		.00	.28	1.00		.00
Discrimination × Emotion	−.41	.17	.02	*	−.17		−.87	.27	.00		−.35
Discrimination × Vulnerability	.07	.20	.74		.03		.17	.29	.56		.07
Discrimination × Succeed	.10	.25	.70		.04		.34	.32	.28		.13
Discrimination × Help	−.03	.17	.85		−.01		.23	.21	.28		.10
Interaction (Q)
Discrimination (Q) × Strength							.56	.22	.01	*	.29
Discrimination (Q) × Emotion							.32	.15	.03	*	.24
Discrimination (Q) × Vulnerability							−.17	.16	.28		−.13
Discrimination (Q) × Succeed							−.45	.17	.01	**	−.30
Discrimination (Q) × Help							−.28	.12	.02	*	−.23

Note. Q = quadratic term. All models adjusted for age, educational attainment, poverty status and medication use.

^*p < 0.05;

^**p < 0.01.

Quadratic assessment of racial discrimination × SWS subscales

Model fit improved in the quadratic model from R² = 0.219 to R² = 0.295. In the quadratic model, significant interactions were observed between racial discrimination (Q) and four of the five SWS subscales (Table 4: strength, emotion suppression, intense motivation to succeed, and obligation to help others). The nature of those interactions is described below and displayed in Figure 2 below and Supplementary Table S1 (online online).

Figure 2.

Interaction effects between racial discrimination and SWS subscales on allostatic load. Low and high SWS corresponds to 1 SD below and above the mean, respectively. Solid lines indicate significant slopes; dashed lines indicate non-significant slopes.

Strength.

Simple slope analysis shows that among those reporting high SWS_strength (Mean + 1SD = 0.61), AL did not vary significantly with increasing racial discrimination. Among those reporting low SWS_strength (Mean–1SD = –0.61), a negative curvilinear relationship was observed (B = –0.62, SE = 0.19, p = 0.00, β = –0.47). When plotted, the highest allostatic load was observed among those reporting moderate levels of racial discrimination and the lowest allostatic load among those reporting either very low or very high levels of racial discrimination.^b

Emotion suppression.

Among those reporting high emotion suppression, we did not observe a significant quadratic effect of racial discrimination, but rather a significant linear effect (B = –0.60, SE = 0.26, p = 0.02, β = –0.32). When plotted, there was an inverse linear relationship such that higher levels of racial discrimination were associated with lower allostatic load. A significant quadratic effect of racial discrimination on allostatic load was observed among those reporting low emotion suppression (B = –0.49, SE = 0.14, p = 0.001, β = –0.37). When plotted, higher levels of discrimination were associated with higher AL; however, there appears to be a threshold beyond which further increases in racial discrimination are not met with additional increases in allostatic load but rather show a slight decrease.

Intense Motivation to Succeed.

A negative curvilinear relationship between racial discrimination and allostatic load was also observed among those reporting high SWS_succeed (B = –0.58, SE = 0.16, p <.001, β = –0.45), indicating that, among this group, allostatic load was the highest at moderate to high levels of racial discrimination. Specifically, there was a positive association between racial discrimination and allostatic load among those reporting high levels of SWS_succeed. However, as racial discrimination increased, a reversal of that association was observed. Allostatic load did not vary with increasing levels of racial discrimination among those reporting low SWS_succeed

Obligation to help others.

We found a negative curvilinear relationship between racial discrimination and allostatic load among those reporting high SWS_help (B = –0.50, SE = 0.15, p =.001, β = –0.38), indicating that allostatic load was the highest at moderate to high levels of racial discrimination.

Mediation analysis.

As in previous work, we found significant associations between racial discrimination, SWS dimensions and a variety of health behaviors: physical activity, diet, using food to cope with stress, smoking, and alcohol use. The findings vary by SWS subscale with the most consistent findings for SWS_emosupp and SWS_help (data available upon request): the interaction between racial discrimination and SWS_emosupp was associated with smoking, alcohol use, and physical activity; the interaction between racial discrimination and SWS_help was significantly associated with using food or substances to cope with stress, and showed a marginal association with smoking and physical activity. However, none of these behaviors showed statistically significant associations with allostatic load and the test of mediation failed (β range: –0.03 to 0.05; p-value range: 0.15 to 0.95).

Finally, as an alternative hypothesis, we assessed whether SWS may mediate rather than moderate the racial discrimination– allostatic load relationship. Although, racial discrimination significantly predicted each SWS dimension, none of the SWS dimensions were associated with allostatic load, and the test of mediation failed (β range: –0.02 to 0.01; p-value range: 0.58 to 0.87) (data available upon request).

Summary of findings.

We found significant interactions for four of the five SWS subscales. Overall, SWS_strength and SWS_emosupp were protective and SWS_succeed and SWS_help exacerbated the negative effects of racial discrimination on allostatic load. The strongest effects were observed for SWS_succeed

Discussion

Summary of study findings

We investigated whether SWS modifies the association between racial discrimination and AL among a community sample of midlife AA women. Our findings indicate that among this group: (1) the relationship between racial discrimination and allostatic load is nonlinear, (2) the SWS modifies the relationship between racial discrimination and allostatic load, and (3) the effects of the SWS vary with some dimensions acting as a safeguard and others exacerbating the harmful effects of racial discrimination on allostatic load. These findings add to a growing body of literature elucidating the mechanisms by which racial discrimination becomes biologically embedded.

Interpretation of study findings

Our study findings offer several insights that may help elucidate the mechanisms by which racial discrimination gets “under the skin”.¹²² First, our findings provide support for the transactional nature of the stress process. Examination of stress exposure alone (i.e., racial discrimination)—captured in our independent effects—while neglecting to capture other key elements of the stress response process (i.e., coping) will likely result in biased effect estimates. As described above, the TMS defines stress as a process involving exposure to an environmental demand, appraisal of the degree to which that demand is threatening, appraisal of one’s actual and perceived capacity to cope with or manage that demand, and the physiologic response corresponding to that appraisal.²⁴ Though our findings suggest more complex relationships between stressor exposure and coping than proposed in this classic stress framework, they affirm the need to consider individual variability in coping and potentially other psychosocial processes involved in the stress response process (e.g., threat appraisal). Second, our findings suggest that studies of racial discrimination, particularly those using the EOD scale, should consider non-linear effects to more fully capture the range of associations with health across the exposure distribution. Previous studies have similarly found non-linear associations between the EOD scale and physical health outcomes.^6,9,105,123

Third, as hypothesized, the effects of SWS were not uniform. Feeling an obligation to present an image of strength and an obligation to suppress emotions were generally protective. The association between racial discrimination and allostatic load was null among those in the high strength group. On the contrary, higher levels of racial discrimination was associated with higher allostatic load in the low strength group. These findings suggest that presenting an image of strength buffers the harmful effects of racial discrimination on health. However, the reversal observed at the higher end of the exposure distribution suggests other factors that may be salutogenic. This pattern may also reflect a system no longer able to mount an appropriate stress response.^1,2 According to the Superwoman Schema framework, SWS_strength is described as feeling an obligation to present an image of strength even when one does not feel strong.⁴⁰ In addition to being strong, the concept of strength in discourse about the strong black woman in AA culture is about presenting a façade of strength, focusing on outward behavior despite one’s actual emotional and physical condition.^37,40,72,124 In their paper, “…Grace of a Lady and the Grit of a Warrior”, Abrams et al. discuss the ubiquitous nature of the strong black woman schema in AA culture and how AA girls are socialized by their mothers and other caretakers to adopt this schema early in life.⁵⁰ Others have similarly described the early adoption of this gendered racial identity.^65,71,72,77 Strength has been described as a core value characterized by “obligatory and volitional independence”, “learned and compulsory resilience”, “matriarchal leadership” and “maintaining a sense of control, confidence, pride, and self-sufficiency.”⁵⁰ As a “collective feature of African American womanhood”,⁶⁶ living up to these ideals is often described as a source of pride that bolsters psychological resilience increasing one’s sense of self-control, belonging, meaning and purpose, and ultimately their ability to withstand external threats.^54,73 Further research is needed to better understand the mechanisms driving this curvilinear relationship.

Similar curvilinear associations were found among those reporting low emotion suppression. However, this was coupled with a very strong inverse linear association among those reporting high levels of emotion suppression. These findings contradict prior literature showing the deleterious health effects of emotion suppression. However, research suggests that coping strategies may have differing effects depending upon the specific context in which they are deployed (i.e., context-dependent coping).²⁷ Racial discrimination refers to unfair treatment due to immutable characteristics (e.g., skin color) of one’s social identity.³⁷ In previous work, AA women refer to the experience of racial discrimination as chronic and beyond their control: as just “another part of your life”.³⁷ Within this context of high demand (i.e., requirement of constant adaptation in the context of repeated exposure to psychosocial stress) and low perceived control, previous studies have shown associations with higher levels of psycho-physiologic stress indicators,¹²⁵ namely cortisol and markers of immune function. Evidence also shows significant associations of high emotional demand (i.e., stressors that are emotionally demanding) with both psychological symptoms and emotional exhaustion or “burnout”.¹²⁵

Additionally, numerous studies have identified anger as among the most common emotional responses to rejection or “threatened belonging” along with hurt, sadness, loneliness, and general feelings of upset or distress.¹⁰⁷ However, within the context of unjust rejection or devaluation based on group membership such as in the case of racial discrimination, anger is the most common and is often coupled with antisocial or avoidance behavior such as expressive suppression (i.e., inhibition of emotion-expressive behavior) as a form of emotion regulation; particularly when the rejection is chronic.¹⁰⁷ Emotion suppression has generally been associated with poor health outcomes.^108,126 Indeed, AA women report “pent up” anger due to emotion suppression, due in part to their efforts to avoid being labeled as the “angry black woman” and describe this form of emotion suppression as distressing.⁶⁶ However, studies also report anger as a particularly damaging emotion, which has been associated with numerous poor outcomes.^108,127,128 Hence, within the context of anger, emotion suppression may be protective.

In a meta-analysis of 83 studies and 295 effect sizes, Jorgensen et al. (1996) found low levels of anger expression was more strongly associated with higher blood pressure in Blacks versus Whites; and one study found that low discrimination-related anger expression was associated with high diastolic blood pressure among black women.¹²⁹ However, the results of the meta-analysis suggest these findings may be moderated by age. Study results showed an inverse association between anger expression and blood pressure among younger groups but a positive association among older samples, such that anger-in was protective among older samples, which consistent with our sample were primarily middle-age adults. Moreover, the authors note that findings across the 83 studies varied greatly and may be related to methodological issues including, but not limited to, exposure assessment strategies. Last, curvilinear associations were not examined but, consistent with our study findings, may help explain the mixed findings. Studies have previously shown that AA women report anger as their primary emotional response to racial discrimination; and that holding “anger in” is associated with positive blood pressure responses whereas “anger out” is associated with maladaptive responses such as higher resting blood pressure, incident hypertension and greater systolic blood pressure and heart rate reactivity.¹²⁸

Additionally, although active coping (i.e., trying to change something about the encounter) is generally seen as more health protective than passive coping (i.e., managing/minimizing the distress associated with the experience such as with emotion suppression), studies suggest that chronic deployment of active coping strategies may be health damaging, especially when the stressor is perceived as particularly stressful and/or uncontrollable.^130–132 AA women have reported racial discrimination as a particularly salient form of psychosocial stress and note that “racism is going to be there, no matter what you do it’s going to be there.”³⁷ Hence, in such cases, where racial discrimination is perceived as chronic and uncontrollable, the benefits of active coping may be weakened.¹³⁰ Thus, studies suggest that perception of chronic racial discrimination coupled with chronic expressions of anger may be associated with repeated and prolonged sympathetic activation.¹³¹ Indeed, one recent study found that anger mediated the association between racial discrimination and allostatic load.¹³³ Hence, suppressing anger may be health protective, particularly within the context of frequent or chronic experiences of racial discrimination. Thus, emotion suppression may buffer against the harmful effects of emotional engagement in the context of racial discrimination, which may help explain the inverse association observed among the high emotion suppression group.

On the contrary, having an intense motivation to succeed and feeling an obligation to help others exacerbated the health risk associated with racial discrimination. African American women commonly report the need to “work harder” to prove themselves, which has been attributed to being both a woman and AA (i.e., gendered racism).^37,134 One of the contextual factors AA women describe as contributing to their motivation to succeed is the historical legacy of racial and gender stereotyping and oppression.⁴⁰ Women described having to “work harder than others” in order to overcome racial bias (i.e., stereotype threat).¹³³ Prior evidence has shown the deleterious health effects of sustained effortful coping.¹³⁶ Our findings suggest this may be particularly harmful in the context of high levels of psychosocial stress. Additionally, postponement of self-care has been described as a liability associated with being more nurturant versus “fixed” or egoistic.^96,137 Indeed, previous studies have shown that adopting a more nurturant stance is associated with fatigue and both mental and physical morbidity.

We hypothesized that the association between racial discrimination and allostatic load would be greater among women reporting high (versus low) levels of SWS_help. Our findings for SWS_help are considered in light of the literature on network stress where tending to the stress experiences of others in addition to one’s own stress has been associated with poor health.⁹⁴ Additionally, in previous qualitative work with AA women, the obligation they feel to care for others is described within the context of neglecting their own self-care.⁴⁰ One study found that the frequency of network stress events was associated with emotional distress among AA women, whereas perception of the stressfulness of either self- or network-stress events was not.⁹⁶ These findings may point to an overall lack of attentiveness to one’s own well-being, which some studies suggest is characteristic of the strong black woman. SWS_vuln was not a significant effect modifier in the present analysis. Further research is needed in larger and more representative samples spanning a wider age range.

As in previous work, we found significant associations between racial discrimination, various SWS dimensions and a variety of health behaviors.¹¹⁸ The most consistent findings were for emotion suppression and feeling an obligation to help others. However, there was no evidence of mediation given that none of these behaviors were significant predictors of AL. Future research examining the SWS framework should test the mediating role of health behaviors in relation to a broader set of physical health outcomes. As a measure of cumulative physiologic dysregulation, allostatic load may not be sensitive to specific health behaviors. More targeted outcomes such as hypertension or diabetes, or specific biomarkers such as blood pressure, HbA1c, glucose, and blood lipids may be more appropriate when examining mediation by health behaviors. Certain behavioral outcomes such as sleep quality may also be suitable health outcome targets.

Finally, although, racial discrimination significantly predicted each SWS dimension, none of the SWS dimensions were associated with allostatic load, and the test of mediation failed. Thus, our study findings suggest that the SWS is a stable disposition among this sample of midlife AA women. Whether the SWS mediates or moderates the racial discrimination–health association likely depends on the timing of exposure. It is possible that the SWS operates as a mediator earlier in life when schemas are still developing and before discrimination has become chronic. Further research is needed to assess these associations at various life stages and in experimental versus observational study designs.

Study limitations

Several methodological considerations should be noted. First, we restricted our sample to AA women. We did this for several reasons: (1) to avoid conflating the racial discrimination experiences reported by AA women and white women, thus avoiding what scholars have noted as the incommensurability of measures across groups, which compromises internal validity;^{109,138–140} (2) to acknowledge the heterogeneity that exists among AA women both with respect to the experience of racial discrimination and the SWS, thereby avoiding the conventional practice of using averages to describe the experience of a heterogeneous group (i.e., common when examining differences between groups) and better isolating variability in the exposure of interest (i.e., providing a more appropriate counterfactual quantity);¹⁴⁰ and (3) to avoid treating AAs as a monolith and instead use intersectionality as an analytic strategy to explore the unique experiences of AA women.^78,140 Second, the cross-sectional nature of this study and purposive sampling strategy precludes our ability to infer causality and generalize our study findings. However, the purpose of this study was not to establish causality or generalize but to explore associations to inform hypothesis generation. Compared to AA women in the 2013 American Community Survey ages 30–50 residing in the same Bay Area counties, our sample has a comparable distribution of educational attainment, employment and marital status but has a higher % not insured and a lower % in poverty.¹⁴¹ We also compared the socio-demographics of our sample to that of several national samples of AA women. Compared to the National Survey of American Life (NSAL), a national household probability sample of n = 6,082 non-Hispanic Whites, non-Hispanic AAs, and Caribbean Blacks;¹⁴² the Black Women’s Health Study, a geographically diverse sample of AA women ages 21–69 with approximately equal numbers from the four major US regions (n = 59,000);¹⁴³ and the CARDIA study, a sample of black and white men and women ages 18–30 at baseline (1985–1986) at four US sites (south, west, Midwest, and upper Midwest): our study participants were similar in terms of age (AAWHHS mean age = 42 years versus 43 years in NSAL and 40 years in the BWHS and CARDIA), but were less educated (33% with ≤HS diploma in AAWHHS versus 24%, 18%, and 28.3% in the NSAL, BWHS, and CARDIA, respectively), had a lower rate of employment (45% unemployed in AAWHHS versus 10.1% in NSAL), and a lower prevalence of being married (30% married/partnered in AAWHHS vs. 42% in NSAL; 42% in the BWHS and 46% in CARDIA). Although the study findings cannot be generalized beyond our study sample, the similar sociodemographic makeup of our sample compared to several other samples of midlife AA women improves the relevance and applicability of these findings and the hypotheses generated from the study results. Further research, in larger probability-based longitudinal samples is warranted. As with all survey research, there is a possibility of reporting bias although we used well-validated survey instruments and took steps to minimize such bias. To confirm our study findings, we also performed sensitivity analyses using the expanded set of controls without regard to statistical significance, which did not appreciably change our study results.

Future research

Given our results, several hypotheses have emerged that warrant further investigation. First, in relation to emotion suppression, we hypothesized that within the context of anger, emotion suppression may be health protective given the particularly pernicious health impacts of anger. An alternative hypothesis, also consistent with our results, is that anger suppression is a learned behavior among those reporting chronic experiences of racial discrimination;¹⁰⁸ and within this context has become part of the automatic and largely unconscious cognitive-affective processing system.^46,47 Previous studies have shown that automatic (versus deliberate) emotion regulation tends to be unconscious and largely effortless, thereby achieving the goals of the present situation (i.e., minimize distress) without the physiologic cost of the strain associated with more active regulation efforts.¹⁰⁸ Automatic emotion regulation has been described as an adaptive regulatory strategy learned from past experience. These habitual responses are activated by situational triggers,²⁸ which actuate schemas and shape various psychological functions, often without conscious awareness.¹⁰⁸ Experimental studies have shown inverse associations of emotion control with anger experience and anger behavior; and with physiologic responses consistent with challenge (versus threat) appraisals.¹⁰⁸ This is particularly the case in contexts where emotion control is valued. For example, studies show in cultures that esteem emotional restraint, emotional control is associated with physiologic challenge as opposed to threat responses, and otherwise has protective health effects.^145,146 This is consistent with other work suggesting that culture informs the way people experience, process, and respond to the social world.⁴⁷ The strong black woman role has been endorsed broadly with some describing their enactment of the superwoman role as unconscious.⁶⁷ Hence, our study findings raise questions about whether the emotion suppression reported by AA women is a learned and hence habitual and largely automatic response to chronic racial discrimination (i.e., protective) or more deliberate in nature. African American women have described the need to actively hold anger-in during what they describe as racist experiences.³⁷ Hence, anger suppression may have both protective and damaging properties. Further research is needed to better understand these nuances as well as whether timing plays a role in the health impacts of these responses. For example, are strength and emotion suppression helpful in both the short and long term; or does being a superwoman exact a toll over time decreasing the capacity of AA women to cope with stress. The Environmental Affordances model suggests that within the context of environmental resources and constraints, strategies are used to cope with situations in ways that preserve mental health and well-being in the present but may be harmful in the long-term.¹⁴⁷ This is consistent with Mendoza-Denton’s culturally-constituted cognitive-affective processing system framework, which posits that having an internal locus of control in the short term leads to feelings of agency, but that in the long-term the “consequences of swimming upstream in the face of systematic discrimination can lead to chronic stress and deteriorating health”.⁴⁷

Several additional questions have emerged from these findings. If not allostatic load, what specific physical health outcomes are relevant for mediation by health behaviors when considering the role of the SWS?, Does mental health mediate the association between the SWS and physical health outcomes?, Does the SWS moderate associations with non-racial stressors and if so does the magnitude of those interactions differ from those observed in relation to race-related stress?

Last, our findings suggest that different forms of coping may have differential health effects based on the severity or chronicity of the stressor. Further, given the context-dependent nature of stress and coping, those relationships may vary by social context (e.g., institutional versus public setting).¹⁴⁸ This may help explain the mixed findings in the literature on racism, coping and health; and the inability to identify coping strategies that are adaptive in relation to racism stress. Further research is needed to better understand how varying coping strategies may be more or less adaptive in different contexts, including within the context of differing levels of racism.

Conclusions

Our findings suggest a complex set of interactions between the various dimensions of SWS and racial discrimination in relation to allostatic load. Studies examining racial discrimination without considering the interacting influence of coping and potentially other aspects of the stress-coping process likely results in biased effect estimates and a distorted understanding of how racism operates to impact health. Our findings underscore the need for scholarship exploring the complexities of how and for whom racial discrimination impacts health. Feeling an obligation to present an image of strength and an obligation to suppress emotions was protective among those reporting high levels of racial discrimination; whereas having an intense motivation to succeed and feeling an obligation to nurture and help others was associated with greater health risk. Associations were curvilinear suggesting that the influence of coping differs by stressor severity/frequency. However, the majority of studies examining the health risk associated with racial discrimination models linear effects, a strict and, it appears, unrealistic assumption. Notably, this study aimed to examine individual variability in the SWS and the degree to which the SWS moderated the association between racial discrimination and allostatic load. However, the SWS may be examined in relation to a wide variety of environmental stressors, given its potential role in the embodiment of stress more broadly; and its ability to guide a more culturally informed examination of the psychobiological mechanisms linking stress to health.

We restricted our sample to U.S.-born AA women because of the unique history of racial discrimination in the United States. The majority of research in the U.S. focused on health disparities tends to dichotomize risk examining for example men vs. women or AAs versus whites without attention to the ways in which the lived and social experience of multiple social identities (e.g., race × gender) intersect to regulate exposure to risk. Research, and in particular public health research, has long documented the persistence of health disparities. We aim to distinguish between health disparities and health inequities.^149,150 Rather than asking do disparities exist, we aim to better understand why disparities exist. In so doing, we endeavor to move beyond rudimentary notions of race and difference and instead pose a challenge to engage in a more critical understanding of both the historical and contemporary social determinants of racial health inequities.

Supplementary Material

Supp FigS1

Supplementary Figure S1. Frequency (%) of racial discrimination across EOD dimensions (n = 207).

Supp TableS1

Supplementary Table S1. Estimates from simple slope analyses for effects of racial discrimination on allostatic load at high versus low levels of the SWS.

Supp TableS2

Supplementary Table S2. Interaction effects between racial discrimination and SWS subscales on allostatic load based on alternative coding scheme.