Race-Related Differences in Acute Pain Complaints Among Inner-City Women: The Role of Socioeconomic Status

Abstract

Previous research has shown that African Americans (AA) report higher pain intensity and pain interference than other racial/ethnic groups as well as greater levels of other risk factors related to worse pain outcomes, including PTSD symptoms, pain catastrophizing, and sleep disturbance. Within a Conservation of Resources theory framework, we tested the hypothesis that socioeconomic status (SES) factors (i.e., income, education, employment, perception of income meeting basic needs) largely account for these racial/ethnic differences. Participants were 435 women [AA, 59.1%; Hispanic/Latina (HL), 25.3%; Non-Hispanic/White (NHW), 15.6%] who presented to an Emergency Department (ED) with an acute pain-related complaint. Data were extracted from psychosocial questionnaires completed at the participants’ baseline interview. Structural Equation Modeling was used to examine whether racial/ethnic differences in pain intensity and pain interference were mediated by PTSD symptoms, pain catastrophizing, sleep quality, and sleep duration, and whether these mediation pathways were, in turn, accounted for by SES factors. Results indicated that SES factors accounted for the mediation relationships linking AA race to pain intensity via PTSD symptoms and the mediation relationships linking AA race to pain interference via PTSD symptoms, pain catastrophizing, and sleep quality. Results suggested that observed racial/ethnic differences in AA women’s pain intensity, pain interference, and common risk factors for elevated pain may be largely due to racial/ethnic differences in SES. These findings highlight the role of social inequality in persistent health disparities facing inner-city, AA women.

INTRODUCTION

Pain conditions affect approximately 69.6 million Americans in a given year (Dahlhamer et al., 2018) and have been shown to differ considerably based on social and demographic factors, including race/ethnicity and socioeconomic status (SES; Riskowski, 2014). African Americans (AAs), in particular, report greater acute pain intensity and pain-related interference than other racial/ethnic groups (Booker, 1989; Meints, Miller & Hirsh, 2016; Portenoy et al., 2004; Kim et al., 2017; Rahim‐Williams et al., 2012; Ostom et al., 2017). In addition, low socioeconomic status (SES; as measured by factors including income, employment status, education level, and neighborhood poverty level) has also been consistently associated with worse pain outcomes, including higher pain severity and more pain-related interference (Davies et al, 2009; Portenoy et al., 2004; Brekke, Hiortdahl, Kvien, 2002).

Importantly, national data consistently show that inequalities in SES are strongly patterned by race (Williams, Mohammed, Leavell, & Collins, 2010). Likely due to longstanding a history of systemic disadvantage, AAs, in particular, have persistent levels of overall poverty that are higher than Whites, and despite increases in education across all racial/ethnic groups since the 1960s (i.e., high school graduation, college graduation), AAs continue to have lower rates of educational attainment when compared to Whites (U.S. Census Bureau, 2017). Further, these increases in educational attainment among AAs have not led to commensurate increases in their earning potential; AAs earn approximately 59 cents for every dollar of income that Whites receive (Ryan & Siebens, 2012). Collectively, these persistent and pervasive racial inequalities in SES require that exploration of race-related differences in pain outcomes address concurrent race-related associations with SES.

Previous work that has sought to disentangle the effects of race from SES in pain outcomes has generated mixed results. Some studies have found that low SES explains adverse pain outcomes in AAs (Urwin et al., 1998; Elliot et al., 1999; Portenoy et al., 2004; Schneider et al., 2005; Fuentes, Hart-Johnson, & Green, 2007) while other studies have found that AAs have significantly worse pain outcomes even after controlling for SES (Plesh, Crawford, & Gansky, 2002; Day & Thorn, 2010; Green & Hart-Johnson, 2012). These discrepancies could be due to a number of methodological factors, including variations in the definition and assessment of SES and race (Edwards, Fillingim, & Keefe, 2001; Poleshuck & Green, 2008; Chan, Hamamura, & Janschewitz, 2013; Baldassari et al., 2016; Kim et al., 2017). For example, some studies have measured SES at the individual level, with only one indicator [e.g., individual income; Day & Thorn, 2010] while others have measured SES with multiple indicators, at the neighborhood level [e.g., percent of homes living below poverty line, percent of homes with high school education or higher, percent of homes with employed labor force; Green & Hart-Johnson, 2012]. Other studies have been limited methodologically by failing to distinguish acute and chronic pain processes (Riskowski, 2014), which is problematic, given that chronic pain has been defined as an acute pain complaint lasting longer than 3–6 months (Steingrímsdóttir et al., 2017). Thus, studies that conflate acute and chronic pain complaints lack specificity as to which risk factors are associated with the development of chronic pain from those pain complaints that may go away on their own. Still other studies have lacked assessment of other potential explanatory variables contributing to worse pain outcomes among AAs. Specifically, well-known pain-related risk factors, including symptoms of Posttraumatic Stress Disorder (PTSD), pain catastrophizing, and indicators of poor or insufficient sleep, have also been shown to differ by race (Burns et al, 2003; Smith, 2004; Beck & Clapp, 2011).

As many as 80% of individuals with PTSD report co-morbid pain (Otis, Keane, Kerns, 2003; Beck & Clapp, 2011; Brennstuhl, Tarquinio & Montel, 2015) and AAs have consistently been shown to have a higher lifetime prevalence of PTSD than other racial/ethnic groups, despite not necessarily experiencing more traumatic events than other racial/ethnic groups (Alim, Charney, & Mellman, 2006; Roberts, Gilman, Breslau, Breslau & Koenen, 2011; McLaughlin et al., 2018). Existing theories that have sought to explain the overlap among PTSD and pain-related complaints point to common underlying physiological and psychological factors [i.e., hyperarousal, attentional biases, and anxiety sensitivity; (Holley, Wilson, Noel, & Palermo, 2016; Beck & Clapp, 2011; Sharp & Harvey, 2001)]. These factors, in turn, have been highlighted as potential mechanisms underlying the persistent associations observed between higher PTSD symptom severity and both increased pain intensity and pain-related interference (Geisser, et al., 1996; Jenewein et al, 2009; Sherman et al., 2000; Phifer et al., 2011).

In addition, AAs have been found to be more likely than other racial/ethnic groups to report engaging in pain catastrophizing, a maladaptive cognitive style characterized by inflexibility, rumination, and increased vigilance to painful stimuli (Aldrich, Eccleston & Crombez, 2000; Carty, O-Donnell, Evans, Kazantzis, & Creamer, 2011). Pain catastrophizing has consistently been linked with worse pain outcomes, including increased pain interference and greater pain intensity (Forsythe et al., 2011; Meints, Miller & Hirsh, 2016). Furthermore, reflecting shared attentional biases and vigilance to potentially threatening stimuli, previous research has also highlighted increased vulnerability towards the use of pain catastrophizing among individuals with comorbid PTSD and pain (Alschuller & Otis, 2012).

Finally, AAs are nearly twice as likely as other racial/ethnic groups to report short sleep durations and low-quality sleep (Stamatakis, Kaplan, & Roberts, 2007; Krueger & Friedman, 2009; Hale & Do., 2007; Grandner et al., 2010). Previous research has linked shortened sleep times (< 6.5 hours per night) and reports of low-quality sleep with more severe pain intensity and pain-related interference (Affleck et al., 1996; Raymond et al. 2001; Finan, Goodin, Smith, 2013; Koffel et al., 2016). In addition, PTSD symptoms and sleeping difficulties are often comorbid, with difficulty falling and/or staying asleep and nightmares representing two diagnostic criteria of PTSD (American Psychological Association, 2013). Recent studies have also implicated disrupted sleep as a potential explanatory mechanism linking symptoms of PTSD to increased pain-related complaints (Bigatti et al, 2008; Lillis et al., 2018a; Aaron et al, 2019).

Collectively, these findings may indicate that differences in pain intensity and pain interference among AAs could be explained, in part, by underlying differences in PTSD symptoms, pain catastrophizing, and sleep. That is, AAs may have worse pain than other racial/ethnic groups because they have worse PTSD, pain catastrophizing, and sleep. However, such findings would still leave the question as to why AAs differ from other racial/ethnic groups on pain intensity and pain interference, and on PTSD, pain catastrophizing and sleep.

Accordingly, from the perspective of the Conservation of Resource Theory (COR Theory; Hobfoll, 1989), we sought to test the hypothesis that socioeconomic disadvantage (i.e., being low SES) largely accounts for both observed AA-race-related differences in self-reported pain outcomes (i.e., pain intensity and pain interference) and the impact of known risk factors associated with poor self-reported pain outcomes (i.e., PTSD symptoms, pain catastrophizing, poor sleep quality, and short sleep duration). We tested this hypothesis in a sample of racially diverse, inner-city women presenting to the Emergency Department with an acute pain complaint.

COR theory was chosen as our conceptual framework for this study as it has been applied in previous research exploring health outcomes following stress (Schumm, Hobfoll, & Keogh, 2004; Hou et al., 2010; Cook, Aten, Moore, Hook, & Davis, 2013; Lillis et al., 2018b) and provides a theoretical bridge across the PTSD, pain, and health disparities literature to implicate (both historical and current) socioeconomic disadvantage as a potential mechanism by which AAs would be disproportionately affected by pain and pain-related risk factors. Specifically, COR theory’s central tenant is that individuals are motivated to protect their resource (personal, social, material, financial, etc.) and that psychological distress arises when actual or threatened resource loss occurs. COR theory further asserts that those who have the fewest resources are the most vulnerable to further loss and distress because they lack sufficient resources to apply towards management of acute stressors.

Thus, viewed through the lens of COR theory, AAs who are also low SES would have fewer resources to recover from previous stressors, which may make them more vulnerable to developing PTSD symptoms. PTSD symptoms, in turn, could increase vulnerability to worse pain intensity and pain interference directly (Geisser, et al., 1996; Jenewein et al, 2009; Sherman et al., 2000; Phifer et al., 2011; Lillis et al., 2018b) as well as indirectly through increasing the use of maladaptive cognitive coping strategies, like pain catastrophizing (Alschuller & Otis, 2012), and through disrupting sleep (Lillis et al., 2018a, Aaron et al., 2019, Bigatti et al., 2008).

Importantly, COR theory could also potentially explain worse pain outcomes among economically disadvantaged AAs independent of PTSD symptoms. Specifically, AAs who are low SES may be more likely to be vigilant to threats of further resource loss and thus engage in more pain catastrophizing (leading to higher pain intensity and pain-interference). In addition, AAs who are low SES may also lack the resources needed to provide a sufficient opportunity for sleep. For example, previous research has shown that occupational instability, whether from working multiple part-time jobs or jobs with long or unpredictable hours, may restrict the amount of time available to obtain a sufficient amount of sleep (Basner et al., 2007; Luckhaupt, Tak, & Calvert, 2010). Further, living in low-income neighborhoods has also been shown to have a negative impact on sleep through increased exposure to nighttime noise and light (Hill et al., 2009; Hale et al., 2013; Johnson, Brown, Morgenstern, Meurer, & Lisabeth, 2015). These SES-related intrusions on sleep duration and sleep quality, in turn, could lead to higher pain intensity and pain-interference.

Study Hypotheses

Following Figure 1 (Model #1), generalizing from the PTSD and pain literature, we first hypothesized that AA race would be significantly related to pain intensity and pain interference through a direct positive relationship with PTSD, an indirect positive relationship with pain catastrophizing (via PTSD), and indirect negative relationships with sleep duration and sleep quality (via PTSD). Second, following Figure 2 (Model #2), based on U.S. Census data (2017) and previous research, we expected a negative direct relationship between AA race and SES. Third, as informed by COR theory, we hypothesized that SES would be linked indirectly with pain intensity and pain interference through a negative direct relationship with PTSD, a negative direct relationship with pain catastrophizing, and positive direct relationships with sleep duration and sleep quality. Finally, in the models reflected in both figures, we expected positive direct relationships from pain catastrophizing to pain intensity and pain interference, and negative direct relationships from sleep duration and sleep quality to pain intensity and interference, respectively.

Figure 1.

Conceptual Model #1 linking African American race to pain intensity and pain interference via direct relationship with PTSD Symptoms and indirect relationships with Pain Catastrophizing, Sleep Quality, and Sleep Quantity.

Figure 2.

Conceptual Model #2 linking African American race to pain intensity and pain interference via direct relationship with SES factors and indirect relationships with PTSD symptoms, Pain Catastrophizing, Sleep Quality, and Sleep Quantity.

METHODS

Participants

The data from the current study are part of an ongoing longitudinal study exploring the relationship between trauma and pain among women who presented to our institution’s inner-city Chicago ED with an acute pain complaint. At the time of the current analysis, 1,921 women had been approached for recruitment. Of those 1,921 women, 1,702 (88.6%) expressed interest in participating. Of the 1702 interested 663 (39.1%) were screened eligible to participate. Of those 663 eligible women, 477 (72%) completed a baseline interview.

In the current study, only data collected at the baseline interview were used for analyses (which excluded 25 cases who listed their race as “other” or who identified as more than one race as these totals were too small to analyze separately). This baseline sample (n=435) was approximately 28 years-old (SD = 6.19) and predominantly AA (59.1%), followed by Hispanic/Latina (HL; 25.3%), and Non-Hispanic/White (NHW; 15.6%). Racial/ethnic composition of the sample of women who completed baseline interviews was not significantly different from the sample of women who did not complete baseline interviews (i.e., those who refused to participate, were ineligible, or were lost to follow up). Demographic characteristics are presented in Table 1 by race/ethnicity and for the total sample.

Table 1.

Characteristics of Total Sample and by Race.

	African American n = 257		Hispanic/Latina n = 110		Non-Hispanic/White n = 68		Total Sample N = 435		Significance Tests
Variables	M	SD	M	SD	M	SD	M	SD	F	p
Age (18–40)	28.49	6.31	28.46	6.27	28.78	5.70	28.53	6.19	.071	n.s.
	n	%	n	%	n	%	n	%	χ2	p
Education									91.52	.00
High School or Less	108a	42.0	44a	40.0	10b	14.7	162	37.2
Some college	115a	44.7	49a	44.5	13b	19.1	177	40.7
College degree or Higher	34a	13.2	17a	15.5	45b	66.2	96	22.1
Employment									18.59	.01
Unemployed	76a	29.6	28a	25.5	8b	11.8	112	25.7
Part-Time/Multiple Jobs	66a,b	25.7	39a	35.5	14b	20.6	119	27.4
Full-Time	115a	44.7	43a	39.1	46b	67.6	204	46.9
Annual Incomec									72.32	.00
< $10,000	110a	43.8	23b	21.3	9b	13.2	142	33.3
$10,000–$39,999	95a	37.8	56b	51.9	15c	22.1	166	38.9
> $40,000	46a	18.3	29a	26.9	44b	64.7	119	27.9
Income meets basic needs?									57.07	.00
Not enough	160a	62.3	53b	48.2	14c	20.6	227	52.2
Enough	89a	34.6	48a,b	43.6	37b	54.4	174	40.0
More than enough	8a	3.1	9b	8.2	17c	25.0	34	7.8
	M	SD	M	SD	M	SD	M	SD	F	p
PTSD Symptoms									7.69	.00
PCL-5 Total Scored	39.48a	17.50	37.71a	15.06	30.93b	10.52	37.69	16.24
	n	%	n	%	n	%	n	%	χ2	p
									8.72	.01
PCL-5 Total Score >33e	144a	56.0	58a,b	52.7	24b	35.8	226	52.1
	M	SD	M	SD	M	SD	M	SD	F	p
Pain Catastrophizing									9.99	.00
PCS Total Scoref	29.95a	12.58	29.39a	11.60	22.65b	11.18	28.66	12.39
Sleep
PROMIS Sleep Duration	6.12	1.71	6.35	1.34	6.58	1.29	6.25	1.57	2.68	n.s.
PROMIS Sleep Quality	1.95a	1.05	2.05a	.83	2.28b	.88	2.03	.99	3.04	.04
Pain
Pain Intensity rating	2.60a	2.62	2.46a	2.45	1.31b	1.47	2.36	2.47	7.69	.00
PROMIS Pain Interference	16.62a	9.17	15.94a	8.14	12.47b	5.87	15.80	8.59	6.45	.00

^a,b,c= each unique subscript letter denotes a race/ethnicity category whose column proportions differ significantly from each other at the p =.05 level.

^c= Income data missing for n = 7.

^d= (α=.93).

^e= a score greater than 33 is the clinical cutoff for a total score consistent with a PTSD diagnosis on the PCL-5.

^f= (α=.94).

^g= (α=.94).

Procedure

The following study procedures were approved by our institution’s Institutional Review Board. Study staff approached women at the ED with information about participating in a study about trauma and pain. For women who expressed interest in participating, study staff collected contact information and then conducted a brief telephone-screening interview within 72-hours of the ED visit to determine eligibility to participate in the study. Inclusion criteria were as follows: (1) female, (2) 18–40 years old, (3) premenopausal, (4) able to read and write English sufficiently to provide informed consent, and (5) present to our institution’s ED with an acute pain complaint of the chest, abdomen/pelvis, neck/shoulder, or back (i.e., not extremity or head pain). Exclusion criteria were as follows: (1) pain intensity or any injury or illness great enough to impair concentration or capacity to understand study instructions or the nature of being in the study, (2) current chronic illness that involved constant or frequent pain, (3) history of chronic pain on presentation in ED or documented in the Electronic Medical Record (EMR), (4) appearing intoxicated or under the influence of drugs at the ED visit, (5) self-reported or EMR-documented daily opiate use over the prior 3 months, or (6) the presenting ED pain complaint was due to a traumatic circumstance (e.g., a motor vehicle accident (MVA), physical assault, sexual assault, etc.). This latter exclusionary criterion was established in order to avoid the confounding effects of the presenting pain complaint and any reported PTSD symptoms being from the same event [e.g., a MVA survivor may have both pain and PTSD from the MVA, not because pain and PTSD impact each other].

Following completion of the telephone screening interview, eligible participants were scheduled for a baseline interview where they completed the informed consent. The baseline interview collected information on participants’ demographic characteristics, psychosocial functioning, pain intensity, pain-related interference, pain catastrophizing, current/past PTSD symptoms, and self-reported sleep characteristics. Participants were compensated for their time in the form of gift cards from a local retail store and could earn $150 for completion of all baseline measures.

Measures

Socioeconomic Status (SES)

Participant SES was measured via four demographic variables: highest education level, current employment status, current annual household income, and perception of current annual household income’s ability to meet basic needs. Highest education level was assessed with the question, “What is the highest grade of school you have completed?” Composition for highest education level was recoded for subsequent analyses as follows: (1) high school diploma/G.E.D. or less; (2) some college or technical school; and (3) college degree or higher. Current employment status was assessed with the question, “Which of these categories best describes your employment status: Full-time; Part-time; Working multiple jobs; Disability/SSI; Unemployed?” Composition for current employment status was recoded for subsequent analyses as follows: (1) unemployed or on disability/SSI; (2) part-time employment/multiple jobs; and (3) full-time employment. Current annual household income was assessed with the question, “Which category best represents your total household income before taxes: less than $10,000; $10,000-$29,999; $30,000-$39,999; $40,000-$49,999; $50,000-$59,999; $60,000-$69,999; $70,000-$79,999; greater than $80,000.” Composition for current employment status was recoded for subsequent analyses as follows: (1) <$10,000; (2) $10,000-$39,999; and (3) >$40,000. Composition of perception of income’s ability to meet basic needs was assessed with the following question “Do you consider your household income to be…:” (1) not enough to meet basic needs; (2) enough to meet basic needs; and (3) more than enough to meet basic needs. This latter measure was included as several past studies have found this to be a key indicator of SES (Shi & Stevens, 2005; Sachs-Ericsson et al., 2006; Nobles, Weintraub, & Adler, 2013; Präg, Mills, & Wittek, 2016).

Sleep Duration & Sleep Quality

Self-reported sleep duration and sleep quality were measured with two items from the PROMIS Sleep Disturbance Short-Form (Yu et al., 2011). Participants were asked to report the average number of hours they typically sleep per night and rate their average nightly sleep quality over the previous seven days on a scale from 1 (very poor) to 5 (very good). These items were developed using rigorous psychometric testing methods (i.e., Classical Test Theory, Item Response Theory), comprehensive literature reviews, qualitative item reviews, and focus groups and were found to have good face and construct validity as well as internal consistency (α = .95; Buysse et al., 2010).

Pain Catastrophizing

The total score from the 13-item Pain Catastrophizing Scale [(PCS); Sullivan, Bishop, & Pivik, 1995] was used to measure symptoms of pain catastrophizing. Respondents were asked to rate, from 0 (not at all) to 4 (all the time), the degree to which they had thoughts and feelings related to rumination, magnification, or helplessness when experiencing pain. Total scores can range from 0–52, with higher scores indicating greater catastrophic thinking. The measure has adequate reliability and validity [(α = .94); Osman et al., 2000].

PTSD Symptoms

The total score from the 20-item PTSD Checklist for DSM-5 [(PCL-5); Weather et al., 2013] was used to measure PTSD symptoms. Prior to administration of the PCL-5, participants were asked to reflect on their worst or most distressing trauma (i.e., their index trauma) and were then asked to rate the degree to which over the previous month, on a 1 (not at all) to 5 (extremely) scale, they had experienced PTSD symptoms related to re-experiencing, avoidance, hyperarousal, and negative alterations in cognition and mood. Total scores range from 20–100 with higher scores indicating more severe symptoms of PTSD. Scores above 33 are suggestive of a potential PTSD diagnosis. The measure has adequate reliability and validity [(α = .93); Blevins et al., 2015].

Pain Intensity

Pain intensity was measured on 11-point numeric rating scale (NRS-11; Farrar et al., 2001) of how much pain they were experiencing at that moment (0 = none at all – 10 = in extreme pain) in the same body area as presented to the ED. A systematic review of different clinical rating scales of pain intensity (Hjermstad et al., 2008) highlighted the superior reliability and face, convergent, divergent, and criterion-related validity of the NRS-11 when compared with visual analog and verbal pain rating scales.

Pain Interference

Pain interference was measured with the PROMIS 8-item short form scale (Amtmann et al., 2010) that asks respondents to rate, on a scale from 1 (not at all) to 5 (very much), how much their pain that initially brought them to the ED had interfered with their engagement in and enjoyment of daily work, home and social-related activities at the time of the interview. Total raw scores were converted into T-scores, with higher scores indicating higher pain-related interference. The measure has adequate reliability and validity [(α = .94); Broderick et al., 2013].

Data Analytic Strategy

Descriptive statistics and bivariate correlations were first examined to characterize our primary study variables. Structural equation modeling (SEM) was conducted with Mplus v.8.0 software (Muthen & Muthen, 2017) to examine the hypothesized direct and indirect relationships among our primary study variables (see Figures 1 and 2). In addition, SEM techniques afforded the opportunity to create and analyze a latent SES variable, which represents the common variance underlying our four observed SES variables: highest education level, current employment status, current annual household income, and perception of current annual household income’s ability to meet needs. Race/ethnicity was entered as a dichotomous variable in the SEM, where 1 = AA (n = 257, 59.1%) and 0 = Non-AA (n = 178, 40.9%). All other continuous variables in the model were measured at the indicator level as described above in the Measures section.

Bootstrapping techniques and maximum likelihood estimation were used in the estimation of our SEMs. Bootstrapping techniques provided 95% confidence intervals to determine the significance of indirect effects. Maximum likelihood estimation allowed all cases in the dataset to be analyzed, even those with missing data (of which < .03% were missing). Model fit was determined via several fit indices, including the root mean square error of approximation (RMSEA), the comparative fit index (CFI), and the standardized root mean square residual (SRMR). Adequate model fit was determined based on published recommendations (Marsh, Hau, & Wen, 2004) with RMSEA values < .06, CFI values > .95 and SRMR values < .09. In addition, prior to inclusion in the structural model, we evaluated the fit of our latent SES variable in a measurement model, which demonstrated good fit [(X₂ (2) = 2.08, p = .35, CFI = .99, RMSEA = .01, 90% C.I. [.00, .09], SRMR = .01] and was retained in subsequent analyses.

RESULTS

Sample Characteristics

Table 1 provides descriptive statistics for demographic and primary study variables by race/ethnicity and for the total sample. Less than half (46.9%) of the total sample reported full-time employment outside the home and 37.2% reported obtaining a high school education or less. Just over a third (37.2%) reported an annual household income of less than $10,000 and over half (52.2%) described their annual income as not enough to meet their basic needs. Relative to AA and HL participants, NHW participants had significantly higher annual incomes, higher levels of education, higher levels of employment, and were more likely to report that their income was enough to meet their basic needs. NHW participants also had significantly lower- PCL-5 scores, PCS scores, pain intensity ratings, and pain interference ratings, and significantly higher sleep quality ratings than AA and HL participants. Just over half (52.1%) of the sample scored above the clinical cutoff that would be consistent with a PTSD diagnosis and significantly smaller proportion of NHW’s scored above the clinical cutoff for a suspect PTSD diagnosis as compared to AA participants (all p’s < .05). Table 2 provides bivariate correlations for the primary study variables. AA race was significantly correlated with all primary study variables.

Table 2.

Bivariate correlations among primary study variables.

Study Variable	1	2	3	4	5	6	7	8	9	10	11	12
1. African American	1
2. White/Non-Hispanic	−.53**	1
3. Hispanic/Latina	−.69**	−.24**	1
4. Education	−.25**	.39**	−.04	1
5. Employment	−.12*	.17**	−.01	.34**	1
6. Annual Income	−.35**	.37**	.09	.51**	.47**	1
7. Income Meets Basic Needs	−.29**	.35**	.04	.28**	.20**	.42**	1
8. PTSD	.13*	−.19**	.01	−.22**	−.07	−.12*	−.20**	1
9. Pain Catastrophizing	.12*	−.21**	.04	−.18**	−.11*	−.15**	−.17**	.38**	1
10. Sleep Quantity	−.15*	.13*	.07	.03	−.03	−.06	.09	−.21**	−.13*	1
11. Sleep Quality	−.13*	.14*	.04	.06	.01	.01	.11*	−.34**	−.19**	.61**	1
12. Pain Intensity	.13*	−.21**	.02	−.12*	−.08	−.14**	−.17**	.30**	.20**	−.16**	−.21**	1
13. Pain Interference	.13*	−.21**	.02	−.13*	−.08	−.14**	−.14**	.33**	.31**	−.16**	−.26**	.70

Note. Significant bivariate correlations are bolded and denoted *p<.05 and **p < .01

Hypothesized Model #1 Fit and Results

Model #1, in which SES variables were not included, demonstrated good fit [(X₂ (7) = 8.67, p = .27, CFI = .99, RMSEA = .02, 90% C.I. [.00, .06], SRMR = .03; See Table 3 and Figure 3]. As hypothesized, there was a positive direct relationship between AA race and PCL-5 scores, the indicator of PTSD symptoms (β = .132). In addition, PCL-5 scores accounted for the direct relationships between: AA race and PCS scores (the indicator of Pain Catastrophizing), AA race and Sleep Quality, and AA race and Sleep Duration, respectively. Significant direct effects also indicated that PCL-5 scores were positively related to PCS scores (β = .399), and negatively related to Sleep Quality (β = −.297) and Sleep Duration ((β = −.181). Pain Interference scores, in turn, were positively related to PCL-5 scores (β = .187), PCS scores (β = .180), and negatively related to Sleep Quality (β = −.141). The only significant direct association with Pain Intensity was a positive relationship with PCL-5 scores (β = .196).

Table 3.

Standardized path coefficients of direct and indirect effects from Conceptual Model #1.

	Standardized Beta (SE)	Two-Tailed p-value
Direct Effects
Effects on PCL-5 Scores
AA Race	.132 (.04)	.00
Effects on Pain Catastrophizing
PTSD symptoms	.399 (.04)	.00
AA Race	.073 (.05)	.12
Effects on Sleep Quality ratings
PTSD symptoms	−.297 (.05)	.00
AA Race	−.053 (.05)	.24
Effects on Sleep Quantity
PTSD symptoms	−.181 (.05)	.00
AA Race	−.078 (.05)	.10
Effects on Pain Intensity
PTSD symptoms	.196 (.06)	.00
Pain Catastrophizing	.081 (.05)	.09
Sleep Quality	−.080 (.06)	.21
Sleep Quantity	−.014 (.05)	.44
Effects on Pain Interference
PTSD symptoms	.187 (.06)	.00
Pain Catastrophizing	.180 (.05)	.00
Sleep Quality	−.141 (.06)	.01
Sleep Quantity	.014 (.05)	.79
Correlation of Sleep Quality with Sleep Quantity	.559 (.04)	.00
Correlation of Pain Intensity with Pain Interference	.661 (.03)	.00
	Standardized Beta (SE)	95% Confidence Interval
Indirect Effects
AA Race to Pain Intensity
via PTSD Symptoms	.026 (.01)	[.010, .048]
via PTSD Symptoms and Pain Catastrophizing	.004 (.01)	[.000, .010]
via PTSD Symptoms and Sleep Quality	.003 (.01)	[.000, .010]
via PTSD Symptoms and Sleep Quantity	.001 (.0)	[−.001, .005]
Total Indirect Effect	.034 (.01)	[.014, .059]
AA Race to Pain Interference
via PTSD Symptoms	.025 (.01)	[.009, .048]
via PTSD Symptoms and Pain Catastrophizing	.010 (.03)	[.004, .018]
via PTSD Symptoms and Sleep Quality	.006 (.01)	[.001, .013]
via PTSD Symptoms and Sleep Quantity	.001 (.01)	[−.003, .002]
Total Indirect Effect	.039 (.02)	[.016, .067]

Note. Significant direct effects bolded when p <.05 and significant indirect effects bolded with 95%.C.I.

Figure 3.

Visual depiction of results from evaluation of Conceptual Model #1. Solid dark lines represent significant direct effects (p < .05) and dashed light lines represent non-significant direct paths (p >.05). Please refer to Table 3 for corresponding standardized path coefficients of direct and indirect effects.

Significant indirect effects were also observed, such that the indirect association of AA race on Pain Intensity ratings through PCL-5 scores was significant (β = .026), as was the indirect effect on Pain Interference scores through PCL-5 scores (β = .025). In addition, the indirect association of AA race on Pain Interference scores through PCL-5 scores and PCS scores was significant (β = .010) as was the indirect association through PCL-5 scores and Sleep Quality (β = .006). Collectively, these results suggested that race-related differences in pain intensity were explained by PCL-5 scores and differences in pain interference scores were explained by PCL-5 scores, PCS scores, and Sleep Quality. No significant direct or indirect effects through Sleep Duration on either pain outcome were observed.

Hypothesized Model #2 Fit and Results

The second hypothesized model, in which the latent SES variable was included, demonstrated good fit [(X₂ (30) = 57.13, p = .01, CFI = .98, RMSEA = .04, 90% C.I. [.02, .06], SRMR = .04; See Table 4 and Figure 4]. As expected, there were significant negative direct relationships between AA race and SES factors (β = −.368) and SES factors and PCL-5 scores (β = −.212). Consistent with our hypothesis, SES factors accounted for the direct relationship between AA race and PCL-5 scores. In addition, SES factors were negatively related to PCS scores (β = −.153), but not were not significantly related to Sleep Quality or Sleep Duration. Similar to Model #1, PCL-5 scores were positively related to PCS scores (β = .366), and negatively related to Sleep Quality (β = −.298) and Sleep Duration (β = −.194). Pain Interference scores, in turn, were positively related to PCL-5 scores (β = .187) and PCS scores (β = .180) and negatively related to Sleep Quality (β = −.141). The only significant direct association with Pain Intensity was a positive relationship with PCL-5 scores (β = .196).

Table 4.

Standardized path coefficients of direct and indirect effects from Conceptual Model #2.

	Standardized Beta (SE)	Two-Tailed p-value
Latent SES Variable
Highest Education Level	.607 (.04)	.00
Annual Household Income	.829 (.04)	.00
Employment Status	.528 (.04)	.00
Perception of Income Meeting Basic Needs	.511 (.05)	.00
Direct Effects
Effects on SES Factors
AA Race	−.368 (.05)	.00
Effects on PTSD Symptoms
SES Factors	−.212 (.05)	.00
AA Race	.067 (.05)	.17
Effects on Pain Catastrophizing
SES Factors	−.153 (.05)	.01
PTSD symptoms	.366 (.04)	.00
Effects on Sleep Quality
SES Factors	−.004 (.04)	.94
PTSD symptoms	−.298 (.05)	.00
Effects on Sleep Quantity
SES Factors	−.063 (.06)	.27
PTSD Symptoms	−.194 (.05)	.00
Effects on Pain Intensity
PTSD symptoms	.196 (.05)	.00
Pain Catastrophizing	.081 (.05)	.08
Sleep Quality	−.080 (.06)	.20
Sleep Quantity	−.047 (.06)	.44
Effects on Pain Interference
PTSD symptoms	.187 (.05)	.00
Pain Catastrophizing	.180 (.05)	.01
Sleep Quality	−.141 (.06)	.01
Sleep Quantity	.014 (.06)	.79
Correlation of Sleep Quality with Sleep Quantity	.550 (.04)	.00
Correlation of Pain Intensity with Pain Interference	.661 (.03)	.00
	Standardized Beta (SE)	95% Confidence Interval
Indirect Effects
AA Race to Pain Intensity
via SES Factors and PTSD Symptoms	.015 (.01)	[.006, .029]
via SES Factors and Pain Catastrophizing	.005 (.01)	[.000, .012]
via SES Factors, PTSD Symptoms, and Pain Catastrophizing	.002 (.01)	[.000, .006]
via SES Factors and Sleep Quality	.000 (.01)	[−.004, .003]
via SES Factors, PTSD Symptoms, and Sleep Quality	.002 (.01)	[.000, .005]
via SES Factors and Sleep Quantity	−.001 (.01)	[−.006, .001]
via SES Factors, PTSD Symptoms, and Sleep Quantity	.001 (.01)	[−.001, .003]
Total Indirect Effect	.023 (.01)	[.010, .041]
AA Race to Pain Interference
via SES Factors and PTSD Symptoms	.015 (.01)	[.006, .028]
via SES Factors and Pain Catastrophizing	.010 (.01)	[.003, .021]
via SES Factors, PTSD Symptoms, and Pain Catastrophizing	.005 (.01)	[.002, .010]
via SES Factors and Sleep Quality	.000 (.01)	[−.005, .005]
via SES Factors, PTSD Symptoms, and Sleep Quality	.003 (.01)	[.001, .007]
via SES Factors and Sleep Quantity	.000 (.01)	[−.001, .005]
via SES Factors, PTSD Symptoms, and Sleep Quantity	.000 (.01)	[−.002, .001]
Total Indirect Effect	.033 (.01)	[.016, .054]

Note. Significant direct effects bolded when p <.05 and significant indirect effects bolded with 95%.C.I.

Figure 4.

Visual depiction of results from evalution of Conceptual Model #2. Solid dark lines represent significant direct effects (p < .05) and dashed light lines represent non-significant direct paths (p >.05). Please refer to Table 4 for corresponding standardized path coefficients of direct and indirect effects.

Significant indirect effects were found, such that the indirect association of AA race on Pain Intensity through SES factors and PCL-5 scores was significant (β = .015). In addition, a number of significant indirect effects of AA race on Pain Interference scores were observed: 1) through SES factors and PCL-5 scores (β = .015); 2) through SES factors and PCS scores (β = .010); 3) through SES factors, PCL-5 scores, and PCS scores (β = .005); and 4) through SES factors, PCL-5 scores, and Sleep Quality (β = .0003). Collectively, these results suggest that SES factors accounted for race-related differences in Pain Intensity ratings through PCL-5 scores and race-related differences in Pain Interference scores through PCL-5 scores, PCS scores, and Sleep Quality. As with Model #1, no significant direct or indirect effects through Sleep Duration on either pain outcome were observed.

DISCUSSION

Previous research has shown that AAs report worse pain intensity and pain-related interference than other racial/ethnic groups as well as greater levels of risk factors commonly associated with poor pain outcomes, including higher PTSD symptoms and pain catastrophizing, shorter sleep duration, and lower sleep quality. From the perspective of COR theory (Hobfoll, 1989), we sought to explore whether SES factors explained AA-race-related differences in self-reported pain outcomes (i.e., pain intensity and pain interference) and known risk factors associated with worse self-reported pain outcomes (i.e., greater PTSD symptoms and pain catastrophizing, short sleep duration, and low sleep quality).

Consistent with past findings (Rahim‐Williams et al., 2012; Booker, 2016; Ostom et al., 2017), we found AA women to report higher pain intensity and pain-related interference than non-AA women. As expected, these differences were mediated by AA women’s higher PTSD symptoms and pain catastrophizing symptoms and lower sleep quality. In partial contrast to our original hypotheses, differences in AA women’s pain intensity and pain interference were not accounted for by differences in their sleep duration. In examination of our model with the inclusion of SES factors, we found that SES factors accounted for the mediation relationship linking AA race to pain intensity via PTSD symptoms and the mediation relationship linking AA race to pain interference via PTSD symptoms, pain catastrophizing, and sleep quality.

Our findings indicate that SES factors may play a role in explaining racial differences in risk factors associated with worse pain outcomes. To our knowledge, our study is the first to show that SES factors largely account for the relationship between AA race and PTSD symptoms. This is important given that previous research often highlights the increased severity in reported PTSD symptoms among AAs when compared to other racial/ethnic groups, but does not further contextualize these findings in terms socioeconomic differences among racial/ethnic minorities. As such, viewed through the lens of COR theory (Hobfoll, 1989), this finding may indicate that socioeconomically disadvantaged AAs may lack the resources needed to effectively recover from an acute stressor and, thus, are more vulnerable to developing PTSD.

Our results also indicated that SES factors explained the relationship linking AA race to higher pain intensity and pain inference, in part, via their association with higher PTSD symptoms. Given that low SES neighborhoods in Chicago have a high incidence of violent crime, it is possible that our participants of low SES were likely to experience greater exposure and threatened exposure to traumatic events, which may result in more severe symptoms of PTSD. The increased vigilance and attentional biases towards further threats of violent crime may, in turn, amplify pain intensity and efforts to cope effectively with pain (i.e., increase pain-related impairment; Geisser, et al., 1996; Jenewein et al, 2009; Sherman et al., 2000; Phifer et al., 2011). Importantly, although we did not measure neighborhood-level SES characteristics in the current study, the ED from which we recruited participants primarily serves the historically racially-segregated south and west-side neighborhoods of Chicago, where the annual rate of violent crime is nearly three times higher than the national average (Bureau of Justice Statistics, 2015; Berman, 2017) and the median household income is below $10,000 (Census Reporter, 2016).

Finally, previous studies have found that racial disparities in self-reported sleep quality and sleep duration are at least partly accounted for by SES factors (Stamatakis, Kaplan, & Roberts, 2007; Hale & Do., 2007; Grandner et al., 2010). For the most part, our sleep-related hypotheses were unsupported. SES factors were only found to partially account for the relationship between AA race and pain interference through a small indirect association with higher sleep quality (via PTSD). This finding may reveal more about the influence of PTSD symptoms on the relationship between sleep quality and pain interference than influences specific to socioeconomic factors. Indeed, our initial model (without SES factors) showed a slightly larger indirect effect of sleep quality on the relationship between AA race and pain interference (via PTSD).

In addition, we observed no significant direct or indirect effects through sleep quality on pain outcomes nor did we observe and significant direct or indirect links between SES factors and sleep duration. Although our study is not the first to demonstrate a lack of relationship between SES factors and self-reported sleep duration (Okun, Tolge, & Hall, 2014), our descriptive data paint a picture of relatively short self-reported sleep durations across racial/ethnic groups (Range Average hrs/night = 6.12 – 6.58). Consistent with the number of non-significant bivariate correlations between sleep duration and our primary study variables, there may not be enough variability in our sample’s self-reported sleep duration to pick up on differences in pain outcomes or unique associations with SES variables.

At the same time, it is also possible that the indicators of SES selected for this study (i.e., income, education, employment, perception of income meeting basic needs) were inadequate to explain these specific differences in self-reported sleep duration and sleep quality. It may be that factors related to lower SES in the urban context, like indices of household crowding (Rona et al., 1998) and nighttime exposure to light and neighborhood noise (Hill, Burdette, & Hale, 2009; Casey et al., 2017) better account for the relationship between race and sleep characteristics reported here.

A major strength of this study is the contextualization of race-related difference in pain risk-factors and pain outcomes with concurrent SES associations. Previous studies have reported independent effects of race on pain even after controlling for SES factors (Plesh, Crawford, & Gansky, 2002; Day & Thorn, 2010; Green & Hart-Johnson, 2012) while others have found mediational effects of SES similar to the ones presented here (Urwin et al., 1998; Elliot et al., 1999; Portenoy et al., 2004; Schneider et al., 2005; Fuentes, Hart-Johnson, & Green, 2007). Although there is no agreed upon “gold standard” measurement of SES in the literature (Shavers, 2007), a major strength of the current study is the use of multiple indicators of SES measured at the individual level, including perception of income meeting basic needs. Previous studies (Shi & Stevens, 2005; Sachs-Ericsson et al., 2006; Nobles, Weintraub, & Adler, 2013; Präg, Mills, & Wittek, 2016) have highlighted the value of measuring individual perception of SES and resource deficits alongside more traditional indicators of SES (i.e., annual income, education history) in predicting a variety of health and functional disability outcomes. At the same time, given the numerous ways in which SES can and has been measured, it is possible that our results could be different had we used different or additional indicators of SES.

A number of important limitations in the current study should be noted. Methodologically, the cross-sectional and self-reported nature of our data precludes drawing causal inferences from our findings. In addition, mediation, as described in the current study, was in the statistical sense only; replication with prospective measurement of the mediators and outcomes would be needed in order to validate these models. Along these lines, the conclusions of our study can only speak to associations of our primary study variables with acute pain processes, as we are not yet able to investigate whether and how our variables of interest relate to the transition from acute to chronic pain. From a study sample perspective, although a major strength of our study is its representation of low-income, inner-city women, this sample representation may also limit the generalizability of our findings to men, in general, women who reside outside low-income, inner-city environments, and individuals with acute pain complaints who did not present to the ED. Similarly, as our sample was restricted to pre-menopausal-age women (18–40), our results could differ with the inclusion of post-menopausal women, given known age-related changes in the frequency and intensity of pain-related complaints (Braden et al., 2012; El Tumi et al, 2017), age-related changes in sleep quality and quantity (Krystal et al., 1998), and age-related differences in financial stability and/or educational attainment.

Finally, a key limitation of our study is that we could not address the systemic effects of institutional racism that likely drive the relationships observed between low SES and poor pain outcomes among AAs in our sample. For example, we did not include a measurement of participants’ neighborhood-level health resources (i.e., distance from medical facilities, availability of pharmacies, access to fresh food, pollution index, etc.), which would be influenced by institutional policies and practices like residential segregation. Given Chicago’s long history of residential segregation that largely persists today (i.e., 13_th most racially segregated city in the United States; Chicago Tribune, 2019), our participants’ health outcomes may have differed as a function of their disproportionate neighborhood access to health-related resources. Along these lines, an additional byproduct of institutional racism is the well-documented non-equivalence of SES indicators among racial/ethnic minorities, which we did not adjust for in our analyses. Specifically, AAs have been shown to receive less income at the same education levels as Whites and accumulate less wealth at equivalent income levels to Whites (Williams, Mohammed, et al., 2010). In this way, although we found low SES to explain some of the associations between AA race and worse pain-related risk factors and pain outcomes, we would not necessarily expect incremental gains in SES to result in commensurate improvements in AA’s risk-factor profiles or pain outcomes.

This study has important clinical implications. Foremost among them is the finding that SES factors largely explained associations among AA race, higher PTSD symptoms, higher pain catastrophizing, and worse subjective pain outcomes. These findings underscore the role of economic inequality in persistent health disparities facing racial/ethnic minorities and emphasize the need for population-level interventions improves access to health-related resources for racial/ethnic minorities and low-income populations. Importantly, many interventions aimed at improving population health actually widen the disparity between advantaged and disadvantaged groups because advantaged groups start at a higher level of health than disadvantaged groups and, thus, stand to reap greater benefits from these interventions (Mechanic, 2002). Accordingly, any population-level intervention that proposes to improve access to health-related resources for marginalized groups should be aimed specifically at marginalized groups and accelerate improvement in health outcomes more rapidly than for the rest of the population (Williams, Priest, & Anderson, 2016).

In addition, our findings also extend the literature on the impact of PTSD symptoms, pain catastrophizing, and poor sleep on subjective pain complaints by demonstrating these associations in a historically understudied population of racially diverse, low-income women. In addition to improving equity in access to medical resources and treatment among economically disadvantaged groups and racial/ethnic minorities, these findings may also indicate the potential utility of prophylactic, cognitive-behavioral interventions for PTSD symptoms, pain catastrophizing and poor sleep (e.g., Pigeon et al., 2012). Such an approach may reduce the likelihood of pain problems persisting, which could obviate treatment escalation to consideration of pharmacological pain management strategies with known risk profiles (i.e., opioid-based medications).

In summary, our findings highlight the critical role of social inequality in explaining persistent racial disparities in subjective pain complaints. These data provide an important perspective on the health-related burden affecting economically disadvantaged AA women and underscore the importance of assessing concurrent socioeconomic factors in race/ethnicity-related health research.