At the intersection of ethnicity/race and poverty: knee pain and physical function

Abstract

Background:

Knee osteoarthritis (OA) disproportionately affects racial and ethnic minorities. non-Hispanic Blacks (NHB) report a higher prevalence and severity of knee OA symptoms than their non-Hispanic White (NHW) counterparts. The role of poverty in explaining this disparity remains unclear.

Objective:

The overall aim of this cross-sectional study was to determine whether ethnic/racial differences in knee pain and physical function varied according to poverty status.

Design:

NHB and NHW adults with or at risk of knee OA self-reported sociodemographic information, and completed the Western Ontario & McMaster Universities Osteoarthritis Index (WOMAC) and the Short Physical Performance Battery (SPPB). Annual income was adjusted for number of household occupants to determine poverty status (i.e., living above versus below poverty line).

Results:

Findings revealed 120 individuals living above the poverty line (49% NHB, 77% NHW) and 71 individuals living below the poverty line (51% NHB, 23% NHW). Adjusted multivariable models revealed significant ethnic/race by poverty status interactions for knee pain (p = .036) and physical function (p = .032) on the WOMAC, as well as physical function on the SPPB (p = .042). Post-hoc contrasts generally revealed that NHW adults living above the poverty line experienced the least severe knee pain and best physical function, while NHB adults living below the poverty line experienced the most severe knee pain and poorest physical function.

Conclusions:

Results of the present study add to the literature by emphasizing the importance of considering poverty and/or other indicators of socioeconomic status in studies examining ethnic/racial disparities in pain and physical function.

Introduction

Arthritis is a highly prevalent, disabling, and costly condition among adults in the United States. Arthritis that is diagnosed by a medical doctor has been reported to affect approximately 22% of American adults (52.5 million) over age 18 years [1]. Osteoarthritis (OA) is the most common of the arthritic conditions, with the knee being the most commonly affected joint [2]. Symptomatic knee OA affects roughly 16% of the American population over 45 years of age (>10 million) [3], and can be accompanied by experiences of intense pain and physical function limitations that are severe enough to produce significant emotional distress [4].

The literature addressing ethnic/racial disparities in the prevalence of symptomatic and radiographic knee OA has historically been mixed. Some studies have reported greater overall prevalence of radiographic and symptomatic knee OA in non-Hispanic Black (NHB) adults compared to non-Hispanic White (NHW) adults [5–7], while other studies have suggested that the prevalence rates are equal between NHB and NHW adults [8–10]. What has become increasingly clear in recent years is NHB adults experience a disproportionate burden of knee OA-related pain symptoms and poor physical function compared to their NHW counterparts [11, 12]. Our group and others have found significant ethnic/racial differences in knee OA pain intensity and related disability, such that NHB adults reported greater intensity of knee pain symptoms and greater physical functioning limitations compared to NHW adults [13–16].

It has been asserted that ethnic/racial differences in the experience of pain (including pain due to knee OA) cannot be fully understood without consideration of socioeconomic status (SES) [17]. In previous research, ethnic/racial differences in chronic pain and disability have traditionally been examined while statistically controlling for SES variables, with the assumption that SES is a confounding factor [18]. In some studies, significant ethnic/racial differences in pain disappear when statistical models control for SES [19], and this could lead to the tenuous conclusion that ethnicity/race is not a factor in knee OA outcomes. It is important to understand that ethnicity/race and SES are inextricably linked constructs that often interact to affect health (and pain) outcomes [20]. To illustrate, the U.S. Census Bureau has recently reported that over one in four (28%) NHB adults has an income below the national poverty line compared with only 11% of NHW adults [21, 22]. It is imperative to consider the interplay between ethnicity/race and SES to fully appreciate how they might uniquely and synergistically affect pain and physical function outcomes in people with knee OA.

The overall aim of this study was to examine differences in the experience of knee pain and physical functioning by ethnicity/race and poverty status. It was hypothesized that ethnicity/race and poverty status would be interactively related to reports of knee pain and physical function. Specifically, NHW adults living above the poverty line would report the lowest severity of knee pain and best physical functioning, while NHB adults living below the poverty line would report the most severe knee pain and poorest physical functioning.

Methods

Study Overview

The current study is part of a larger ongoing project that aims to enhance the understanding of mechanisms underlying ethnic/racial group differences in knee pain among adults with or at risk for knee OA; (Understanding Pain and Limitations in Osteoarthritic Disease - Second Cycle, UPLOAD-2). The UPLOAD-2 study addresses this aim by directly and longitudinally assessing the nature and evolution of altered central pain processing and its relationship with progression of pain and disability among middle-aged and older NHB and NHW adults. The UPLOAD-2 study is a multi-site investigation conducted at the University of Florida and the University of Alabama at Birmingham. The participants described in the current analysis were recruited at both study sites between August 2015 and May 2017. Our group has previously published a portion of the collected data addressing psychological resiliency factors in relation to experimental and clinical pain [23]. Findings from the present study were presented in poster format at the 37^th Annual Scientific Meeting of the American Pain Society [24]. Overlap between this previously published work and the study presented here is minimal. The measures and procedures described below are limited to those involved in the current study.

Adults who only reported their ethnicity as not Hispanic or Latino and their race as either Black or White were recruited. Annual household income adjusted for number of household occupants was used to determine poverty status (i.e., above or below poverty line) according to the 2017 Health and Human Services poverty guidelines [25]. While educational attainment and occupation capture individually based dimensions of SES, annual household income is more indicative of standard of living and life chances household members experience through sharing goods and services [26]. Low SES, including living below the poverty line, is consistently associated with increased risk for musculoskeletal pain [27]. With musculoskeletal pain conditions such as knee OA, physical activity is often the primary driver of pain symptoms (i.e., movement-evoked pain) [28]. Therefore, we assessed movement-evoked pain and physical function in this study by having participants complete a standardized short physical performance battery [29]. A validated self-report measure of knee OA-related pain and physical function was also completed [30]. All procedures were reviewed and approved by the Institutional Review Boards at the University of Florida and University of Alabama at Birmingham.

Participants

A total of 200 participants aged 45 years and older with or at risk of knee OA were consecutively enrolled into this study. Specifically, all individuals enrolled in the present study were positive for OA symptoms on a telephone screening questionnaire that has previously shown 87% sensitivity and 92% specificity for radiographically confirmed symptomatic knee OA [31]. In addition, all participants reported current knee pain and were negative for other rheumatologic conditions (e.g. rheumatoid arthritis, fibromyalgia) that could explain knee pain. Given that there is widespread variability in definitions of OA [32], we adopted this approach to be as inclusive as possible in recruitment because our primary focus is on understanding factors associated with knee pain rather than OA pathophysiology itself. Participants were recruited from the community via posted fliers and radio and print media advertisements, as well as from orthopedic and rheumatology clinics, and by word-of-mouth referral. All participants provided informed consent and were compensated for their participation.

Procedures

Initial Screening.

All participants completed initial screening via telephone to determine eligibility for study inclusion. The following sociodemographic data were obtained as part of the screening: self-reported sex, age, ethnic/racial identity, annual household income, and current number of household occupants. A brief health history was completed to assess physical health, including symptoms of knee OA. Initial criteria for participant inclusion into the study were as follows: 1) between 45 and 85 years of age; 2) having or being at risk for unilateral or bilateral symptomatic knee OA based upon American College of Rheumatology criteria [33]; and 3) availability to complete multiple study sessions. Individuals were excluded from participation if they met any of the following criteria: 1) prosthetic knee replacement or other clinically significant surgery to the arthritic knee; 2) heart disease, congestive heart failure, or history of acute myocardial infarction; 3) peripheral neuropathy; 4) systemic rheumatic disorders including rheumatoid arthritis, systemic lupus erythematosus, and fibromyalgia; 5) chronic daily opioid use; 6) excessive anxiety regarding protocol procedures (e.g., refusal to complete short physical performance battery); or 7) hospitalization within the preceding year for psychiatric illness. Individuals who endorsed knee pain and met initial study inclusion criteria presented approximately one to two weeks later for a laboratory session.

Laboratory Session.

During this session, participants completed physical and mental health assessments to ensure participant safety and confirm ongoing study inclusion. For example, three consecutive blood pressure readings were obtained to assess for uncontrolled hypertension. Individuals with blood pressure readings that were > 150/95 were excluded from further study participation to minimize risk of potential cardiovascular events associated with study completion. Furthermore, all individuals underwent a bilateral knee joint evaluation by an experienced examiner (i.e., the study rheumatologists or nurse practitioner). Participants were categorized as either having, or being at risk for, knee OA using the American College of Rheumatology criteria for symptomatic knee OA [33]. Based upon the knee joint evaluation and the participant’s self-report, the affected knee for participants with unilateral knee pain or the most symptomatic knee for those with bilateral knee pain was selected for knee radiography. Knees were graded by a rheumatologist (RS, LBH) using the Kellgren-Lawrence scale (scores ranging from 0 = normal to 4 = severe) [34]. Height and weight was collected from each participant to calculate body mass index. Participants then completed self-report measures of knee OA pain and physical function, positive and negative affect, pain coping strategies, and somatic symptom severity prior to a short physical performance battery.

Measures

Ethnicity/Race.

Participants self-reported their ethnic and racial background using response options consistent with the United States census survey. All participants enrolled into the study identified their ethnic background as non-Hispanic and their racial background as either Black or White.

Poverty Status.

Annual household income was assessed in increments of $9,999, starting at “$0-$9,999” and continuing in this manner until the last category: “$150,000 or higher.” Participants then reported the total number of individuals currently residing in their household. This information was used to determine whether a participant fell above or below the national poverty line based upon the 2017 U.S. Health and Human Services poverty guidelines [25]. To illustrate, a household with four occupants reporting a collective annual income less than $24,600 would fall below the national poverty line for a household of that size. Conversely, if this household reports a collective annual income greater than $24,600, then they are above the national poverty line. We chose to focus on poverty status in this study given its practical relevance and utilization by the Department of Health and Human Services for determining eligibility for certain federal programs (e.g., Supplemental Nutrition Assistance Program, Low-Income Home Energy Assistance Program).

Body Mass Index.

Using the same scale for all participants, weight was measured without shoes to the nearest 0.1 kilogram (kg). Height was measured to the nearest 0.1 centimeter (cm) using a wall-mounted stadiometer which was calibrated daily with a standardized measuring rod. Body mass index (BMI) was calculated using the following algorithm: weight in kg/height in m².

Western Ontario and McMaster Universities Index of Osteoarthritis (WOMAC).

The WOMAC is frequently used in clinical and research settings to assess individuals’ retrospective self-report of knee osteoarthritis symptoms over the preceding 48 hours [30]. The WOMAC includes 24 items and can be divided into subscales of pain, stiffness, and physical function. The WOMAC pain (range 0–20) and physical function (range 0–68) subscales were included in data analysis. Higher scores on these two subscales indicate greater pain and poorer physical functioning, respectively. In the current study, Cronbach’s α for the WOMAC pain subscale was .89 and .97 for the WOMAC physical function subscale.

Positive and Negative Affective Scale (PANAS).

The PANAS is a 20-item scale that assesses positive and negative affect [35]; however, the present study utilized only the 10 items corresponding to the negative affect subscale. Participants rated their affect on a 5-point scale (1 = very slightly or not at all; 5 = extremely). Negative affect was calculated by summing all 10 items, resulting in scores ranging from 10 to 50. Higher scores suggest more negative affect. In this study, the internal consistency of the negative affect subscale was excellent (Cronbach’s α = 92).

Coping Strategies Questionnaire-Revised (CSQ-R).

The CSQ-R assesses the extent to which individuals use pain coping strategies across 7 possible strategies [36]; however, the present study only used the catastrophizing subscale. Additionally, the CSQ-R was chosen over a larger, more comprehensive scale (e.g. Pain Catastrophizing Scale) to adequately balance assessment needs while minimizing time and burden of participants. Participants rated pain-related catastrophic thoughts on a scale from 0 (never do that) to 6 (always do that). Responses to the catastrophizing items of the CSQ-R are averaged and range from 0 to 6, with higher scores suggesting greater catastrophizing. The internal consistency of the CSQ-R catastrophizing subscale was good (Cronbach’s α = 85).

Patient Health Questionnaire - 15 (PHQ-15).

The PHQ-15 is one of the most frequently used instruments to identify people at risk for somatization, which refers to the production of recurrent and multiple medical symptoms with no discernible organic cause [37]. The PHQ-15 assesses the presence and severity of common somatic symptoms such as fatigue, gastrointestinal, musculoskeletal, pain, and cardiopulmonary symptoms within the last four weeks using 15 items. Responses are summated and range from 0 to 30. Higher scores on the PHQ-15 are suggestive of higher self-rated somatic symptom burden (0–4 no-minimal; 5–9 low; 10–14 medium; 15–30 high). The internal consistency of the PHQ-15 was good (Cronbach’s α = 82).

Short Physical Performance Battery (SPPB).

The SPPB assesses lower extremity function with balance, chair, and walking tests [29]. Specifically, participants were asked to: 1) stand with their feet together in the side-by-side, semi-tandem, and tandem positions for up to ten seconds each; 2) rise from a seated position in a chair and return to a seated position five times; and, 3) walk a 4-meter course twice. If the participant did not feel it was safe to perform the activity, they received a score reflecting non-participation. For each category, based on their performance, they received a score of 0–4 (total score 0–12). A lower score indicates worse function and greater likelihood of disability. Participants also used a 0–100 (0 = no pain, 100 = the most intense pain imaginable) numeric rating scale to indicate the intensity of any movement-evoked pain experienced during completion of the balance, chair, and walking tests. Pain intensity ratings were averaged across the three tests. The SPPB has been standardized and previously used in older populations with knee OA as a measure of lower extremity function [38, 29].

Data Analysis

In this study missing data was analyzed separately for each variable using frequency distributions. Past literature suggests that 5 to 10% of cases missing data is reasonable for consideration of data imputation methods, as long as data are missing at random [39]. Less than 5% of participants (9 of 200) declined to provide information pertaining to annual household income. Because these data were not missing at random, data imputation methods were not completed and these 9 participants were omitted from further analysis. All other key study variables each had less than 10% missing data, which was determined to be missing at random. Therefore, a simple data imputation method was completed using the macro for Hot Deck imputation [40], which resulted in complete data for the remaining 191 participants. The Hot Deck data imputation method is well validated and accepted in the statistical community [41]. Differences among continuously measured variables were assessed by one-way analysis of variance (ANOVA), while categorical variables were analyzed using Chi-square (χ²). Pearson correlations were used to evaluate the associations among continuously measured variables. General Linear Models incorporating a 2×2 factorial design for Analysis of Covariance (ANCOVA) were used to examine race by poverty status differences in self-reported and movement-evoked pain and physical function. Significant ethnicity/race by poverty status interaction effects were probed using pair-wise contrasts adjusted for multiple comparisons with a Holm-Bonferroni correction. Statistical significance is reported with p values and 95% confidence intervals (CI) of mean differences (M_diff) where appropriate. Variance inflation factor (VIF) statistics were used to explore whether multicollinearity between ethnicity/race and poverty status might have negatively impacted results of the ANCOVA analyses. VIF > 5 is commonly accepted to be suggestive of multicollinearity.

Results

Participant characteristics

In Table 1, participant characteristics are presented for the overall sample participants, as well as separately for NHB adults and NHW adults. Within the full sample of 191 participants, the mean age of 57.7 years and 62% were women (38% men). Fifty-one percent of the sample identified as NHB adults, while the remaining participants (49%) identified as NHW adults. Overall, the majority of participants (63%) were living above the national poverty line according to annual household income and number of occupants in the home. Occupational status was reported as 41% currently employed (full-time or part-time), 20% currently unemployed, 23% retired, and 16% disabled (supplemental security income benefits). The majority of participants obtained a high school degree or more education (93%), while the remaining 7% of individuals reported less than a high school diploma. Overall BMI for this sample was in the obese range (>30 kg/m2). According to American College of Rheumatology criteria, 64% of this sample met criteria for symptomatic knee OA while the remaining 36% were at risk for knee OA. As seen in Table 1, 44% of the overall sample had no radiographic evidence (Grade 0) or doubtful radiographic evidence of knee OA (Grade 1). Of the remaining participants, 23% demonstrated definite radiographic findings (Grade 2), 17% demonstrated moderate radiographic findings (Grade 3), and 16% demonstrated severe radiographic findings (Grade 4). Neither knee pain or physical function on the SPPB and WOMAC significantly differed according to Kellgren-Lawrence grade (p’s > .05), or between participants who met criteria for symptomatic knee OA and those who were at risk for knee OA (p’s > .05).

Table 1:

Participant characteristics and descriptive data.

	Overall (n = 191)	NHB (n = 98)	NHW (n = 93)
Variable	Mean (SD) or N (%)	Mean (SD) or N (%)	Mean (SD) or N (%)
Age (years)	57.7 (7.5)	56.2 (6.4)	59.3 (8.3)
Sex
Men	73 (38%)	42 (43%)	31 (33%)
Women	118 (62%)	56 (57%)	62 (67%)
Income
$0–$9,999	58 (30%)	41 (42%)	17 (18%)
$10,000 – $19,999	26 (14%)	16 (17%)	10 (11%)
$20,000 – $29,999	26 (14%)	15 (15%)	11 (12%)
$30,000 – $39,999	8 (4%)	5 (5%)	3 (3%)
$40,000 – $49,999	14 (7%)	3 (3%)	11 (12%)
$50,000 – $59,999	17 (9%)	5 (5%)	12 (13%)
$60,000 – 79,999	14 (7%)	6 (6%)	8 (9%)
$80,000 – $99,999	11 (6%)	4 (4%)	7 (7%)
$100,000 – $149,999	12 (6%)	2 (2%)	10 (11%)
$150,000 or Higher	5 (3%)	1 (1%)	4 (4%)
Poverty Status
Above Poverty Line	120 (63%)	48 (49%)	72 (77%)
Below Poverty Line	71 (37%)	50 (51%)	21 (23%)
Occupation
Currently employed	79 (41%)	42 (43%)	37 (40%)
Currently unemployed	37 (20%)	19 (19%)	18 (19%)
Retired	44 (23%)	14 (14%)	30 (32%)
Disabled	31 (16%)	23 (24%)	8 (9%)
Education
High school diploma or more	178 (93%)	87 (89%)	91 (98%)
Less than high school diploma	13 (7%)	11 (11%)	2 (2%)
Study Site
Site 1	123 (64%)	58 (59%)	65 (70%)
Site 2	68 (36%)	40 (41%)	28 (30%)
Body Mass Index (BMI)	31.7 (7.3)	32.6 (7.3)	30.8 (7.3)
PANAS - Negative Affect	16.5 (6.7)	17.5 (7.5)	15.5 (5.5)
CSQ-R - Catastrophizing	1.3 (1.3)	1.7 (1.3)	0.9 (1.1)
PHQ-15	8.1 (4.2)	8.4 (4.7)	7.7 (3.6)
SPPB Pain	22.0 (25.2)	28.5 (28.8)	15.1 (18.5)
SPPB Physical Function	9.4 (1.7)	9.1 (1.8)	9.8 (1.5)
WOMAC Pain	7.8 (4.3)	9.0 (4.1)	6.6 (4.1)
WOMAC Physical Function	25.1 (14.6)	29.9 (14.0)	19.9 (13.5)
Kellgren Lawrence
Grade 0–1	84 (44%)	44 (45%)	40 (43%)
Grade 2	43 (23%)	21 (21%)	22 (24%)
Grade 3	33 (17%)	19 (20%)	14 (15%)
Grade 4	31 (16%)	14 (14%)	17 (18%)

Note: NHB = non-Hispanic Blacks and NHW = non-Hispanic Whites; SD = standard deviation; PANAS = Positive and Negative Affect Scale; CSQ-R = Coping Strategies Questionnaire-Revised; PHQ-15 = Patient Health Questionnaire-15; SBBP = Short Performance Physical Battery; WOMAC = Western Ontario McMasters Universities Osteoarthritis Index.

Ethnic/racial and poverty status differences in participant characteristics

A significantly greater proportion of NHB adults were living below the poverty line compared to their NHW counterparts (χ² = 16.53, p < .001). There were no significant ethnic/racial or poverty status differences observed across Kellgren-Lawrence grades (p’s > .05). Likewise, the proportions of participants meeting criteria for symptomatic knee OA versus those at risk for knee OA did not significantly differ according to ethnicity/race or poverty status (p’s > .05). NHW adults were significantly older than NHB adults (F(1, 189) = 8.27, p = .004; M_diff = 3.08, 95% CI = 0.97, 5.19). The proportions of men and women did not significantly differ across NHB and NHW adults (χ² = 1.83, p = .176); however, a greater proportion of women compared to men were living above the poverty line (χ² = 5.87, p = .015). BMI did not significantly differ between NHB and NHW adults (F(1, 189) = 2.97, p = .087; M_diff = 1.82, 95% CI = −0.25, 3.90), or those living above versus below the poverty line (F(1, 189) = 1.37, p = .243; M_diff = 1.28, 95% CI = −0.88, 3.44). Incorporating a Holm-Bonferroni adjustment, NHB adults reported significantly greater negative affect (F(1, 189) = 4.28, p = .04; M_diff = 1.98, 95% CI =0.09, 3.86) and pain catastrophizing ((F(1, 189) = 21.13, p < .001; M_diff = 0.80, 95% CI = 0.46, 1.15), but not somatic symptoms ((F(1, 189) = 1.51, p = .221; M_diff = 0.74, 95% CI = −0.45,1.93), compared to NHW adults. Participants living below the poverty line reported significantly greater negative affect ((F(1, 189) = 9.12, p = .003; M_diff = 2.95, 95% CI = 1.02, 4.87), pain catastrophizing ((F(1, 189) = 13.87, p < .00; M_diff = 0.69, 95% CI = 0.32, 1.05), and somatic symptoms ((F(1, 189) = 4.34, p = .038; M_diff = 1.29, 95% CI = 0.69, 2.52) compared to their counterparts living above the poverty line.

Correlations and selection of covariates

Pearson correlations are shown in Table 2. Greater pain and poorer physical function on the SPPB and WOMAC were all generally and consistently associated with younger age, higher BMI, greater negative affect, greater pain catastrophizing, and greater somatic symptoms. Accordingly, all analyses described below included the following variables as statistical covariates: age, BMI, negative affect, pain catastrophizing, and somatic symptoms. Study site was also included as a covariate given a significant site difference in movement-evoked knee pain on the SPPB ((F(1, 189) = 7.34, p = .006; M_diff = 10.40, 95% CI = 3.03, 17.77). Despite lack of significant associations with race/ethnicity, poverty status, knee pain, or physical function, radiographic findings according to Kellgren-Lawrence grades were included in subsequent analyses as a covariate given the putative role of knee joint pathology in the manifestation of clinical knee pain. The decision to include these variables as statistical covariates was made to better appreciate whether ethnicity/race, poverty status, and/or their interaction was significantly associated with pain and physical function over and above the influence of covariates.

Table 2:

Pearson’s correlations among continuously-measured variables.

Variable	1	2	3	4	5	6	7	8
1. Age	—
2. BMI	−.235**	—
3. PANAS Negative Affect	−.306**	.106	—
4. CSQ-R Catastrophizing	−.302**	.202**	.517**	—
5. PHQ-15	−.298**	.215**	.437**	.329**	—
6. SPPB Pain	−.177*	.129	.214**	.352**	.206**	—
7. SPPB Physical Function	.018	−.241**	−.295**	−.319**	−.203**	−.147*	—
8. WOMAC Pain	−.286**	.234**	.375**	.517**	.347**	.522**	−.301**	—
9. WOMAC Physical Function	−.236**	.304**	.386**	.528**	.281**	.565**	−.436**	.809**

^*= p < .05,

^**= p < .01

Note: BMI = Body mass index; PANAS = Positive and Negative Affect Scale; CSQ-R = Coping Strategies Questionnaire-Revised; PHQ-15 = Patient Health Questionnaire-15; SBBP = Short Performance Physical Battery; WOMAC = Western Ontario McMasters Universities Osteoarthritis Index.

SPPB pain and physical function

A 2×2 factorial ANCOVA was conducted to evaluate the effects of ethnicity/race (NHB versus NWH adults) and poverty status (above poverty line versus below poverty line) on reports of movement-evoked pain and physical function in response to the SPPB. Despite unequal sample sizes, homogeneity of variance was not violated according to Levene’s test (p’s > .05). The distributions of movement-evoked pain and physical function were approximately normal across racial groups and poverty status as indicated by Shapiro-Wilk statistics (p’s > .05). These findings indicate that the parametric assumptions for the conduct of ANCOVA were not violated. After adjustment for covariates, results revealed a significant ethnic/racial difference for movement-evoked pain (F(1, 180) = 3.956, p = .048; M_diff = 13.30, 95% CI = 6.36, 20.25), such that NHB adults reported significantly greater knee pain in response to the SPPB than their NHW counterparts (Table 3). There was a non-significant difference for poverty status (F(1, 180) = 1.433, p = .233; M_diff = 10.92, 95% CI = 3.63, 18.20), and a non-significant ethnicity/race by poverty status interaction for movement-evoked knee pain on the SPPB (F(1, 180) = .359, p = .534). Results further revealed a significant ethnicity/race by poverty status interaction for physical function on the SPPB (F(1, 180) = 4.953, p = .027). In light of this significant interaction, pairwise contrasts were completed to examine differences in physical function on the SPPB according to ethnicity/race and poverty status. As shown in Figure 1, NHW adults above the poverty line demonstrated significantly better physical function on the SPPB in comparison to both NHW adults (p < .001; M_diff = 1.59, 95% CI = 0.91, 2.27) and NHB adults (p < .001; M_diff = 1.33, 95% CI = 0.80, 1.86) below the poverty line. Findings remained significant even after adjustment for multiple comparisons using the Holm-Bonferroni correction. There was no significant difference in physical function on the SPPB between NHB and NHW adults above the poverty line (p = .191; M_diff = 0.76, 95% CI = 0.18, 1.34). The VIFs in these analyses ranged from 1.05 to 1.21, which are well below the commonly accepted cutoff of 5. This suggests that multicollinearity, particularly that between ethnicity/race and poverty status, did not unduly influence the findings for movement-evoked knee pain and physical function on the SPPB.

Table 3:

General linear models examining ethnicity/race by poverty interactions in relation to movement-evoked pain and physical function assessed by the short physical performance battery (SPPB).

	SPPB Pain					SPPB Physical Function
	SS	DF	MS	F	Sig.	SS	DF	MS	F	Sig.
Variables
Age	23.033	1	23.033	.043	.837	17.528	1	17.528	8.414	.004
Study site	2299.646	1	2299.646	4.2486	.041	15.164	1	15.164	7.279	.008
BMI	36.657	1	36.657	.068	.795	19.897	1	19.897	9.551	.002
PANAS negative	.402	1	.402	.001	.978	11.504	1	11.504	5.522	.020
CSQ catastrophizing	3557.156	1	3557.156	6.570	.011	5.480	1	5.480	2.630	.107
PHQ-15	579.474	1	579.474	1.070	.302	1.662	1	1.662	.798	.373
Kellgren-Lawrence grade	172.514	1	172.514	.319	.573	8.514		8.514	4.087	.045
Ethnicity/race	2141.821	1	2141.821	3.956	.048	1.020	1	1.020	.490	.485
Poverty	775.845	1	775.845	1.433	.233	31.196	1	31.196	14.975	<.001
Ethnicity/race X poverty	210.464	1	210.464	.389	.534	10.318	1	10.318	4.953	.027

Note: Ethnicity/race coded as 1 = non-Hispanic Black, 2 = non-Hispanic White; SS = sum of squares, DF = degrees of freedom, MS = mean square

Figure 1:

Mean differences in SPPB physical function according to ethnicity/race and poverty.

WOMAC pain and physical function

Self-reported knee pain and physical function on the WOMAC were each found to be approximately normally distributed with homogenous variances across ethnicity/race and poverty status groups as indicated by Shapiro-Wilk statistics (p’s > 0.05) and Levene’s test (p’s > 0.05), respectively. Adjusted analyses revealed significant ethnicity/race by poverty status interactions in relation to both knee pain (F(1, 180) = 4.438, p = .037) and physical function (F(1, 180) = 5.290, p = .023) on the WOMAC (Table 4). Additional pairwise contrasts were completed to examine differences in knee pain and physical function on the WOMAC according to ethnicity/race and poverty status. As shown in Figure 2, NHW adults above the poverty line reported significantly less knee pain on the WOMAC in comparison to NHW adults below the poverty line (p < .01; M_diff = 3.46, 95% CI = 1.56, 5.36), NHB adults above the poverty line (p < .01; M_diff = 2.66, 95% CI = 1.24, 4.08), and NHB adults below the poverty line (p < .001; M_diff = 3.72, 95% CI = 2.27, 5.17). Reported knee pain on the WOMAC did not significantly differ between NHB adults living above versus below the poverty line (p = .561; M_diff = 1.06, 95% CI = −0.59, 2.71). Similarly, as shown in Figure 3, NHW adults above the poverty line reported significantly better physical function on the WOMAC in comparison to NHW adults below the poverty line (p < .01; M_diff = 12.40, 95% CI = 6.24, 18.56), NHB adults above the poverty line (p < .01; M_diff = 10.96, 95% CI = 6.25, 15.68), and NHB adults below the poverty line (p < .001; M_diff = 14.47, 95% CI = 9.78, 19.16). Reported physical function on the WOMAC did not significantly differ between NHB adults living above versus below the poverty line (p = .558; M_diff = 3.51, 95% CI = −2.09, 9.11). All significant findings remained that way even after adjustment for multiple comparisons using the Holm-Bonferroni correction. The VIFs in these analyses ranged from 1.04 to 1.57, suggesting that multicollinearity did not bias the findings for self-reported knee pain and physical function on the WOMAC.

Table 4:

General linear models examining ethnicity/race by poverty interactions in relation to self-reported pain and physical function assessed by the Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC).

	WOMAC Pain					WOMAC Physical Function
	SS	DF	MS	F	Sig.	SS	DF	MS	F	Sig.
Variables
Age	7.855	1	7.855	.640	.425	18.027	1	18.027	.136	.713
Study site	.429	1	.429	.035	.852	3.747	1	3.747	.028	.867
BMI	22.951	1	22.951	1.871	.173	1089.089	1	1089.089	8.201	.005
PANAS negative	13.008	1	13.008	1.060	.305	482.265	1	482.265	3.632	.058
CSQ catastrophizing	231.124	1	231.124	18.839	<.001	2259.524	1	2259.524	17.015	<.001
PHQ-15	59.076	1	59.076	4.815	.029	87.284	1	87.284	.657	.419
Kellgren-Lawrence grade	.589	1	.589	.048	.827	297.125	1	297.125	2.238	.136
Ethnicity/race	7.625	1	7.625	.622	.432	401.113	1	401.113	3.021	.084
Poverty	61.129	1	61.129	4.983	.027	1003.772	1	1003.772	7.559	.007
Ethnicity/race X poverty	54.449	1	54.449	4.438	.037	702.513	1	702.513	5.290	.023

Note: Ethnicity/race coded as 1 = non-Hispanic Black, 2 = non-Hispanic White; SS = sum of squares, DF = degrees of freedom, MS = mean square

Figure 2:

Mean differences in WOMAC pain according to ethnicity/race and poverty.

Figure 3:

Mean differences in WOMAC physical function according to ethnicity/race and poverty.

Discussion

Findings from this study revealed a significant ethnic/racial difference in movement-evoked pain on the SPPB between NHB and NHW adults, such that NHB adults reported greater knee pain than their NHW counterparts. This finding is consistent with previously published research showing that NHB adults with knee OA tend to experience a much higher degree of movement-evoked pain in comparison to their NHW counterparts with knee OA [42]. Interestingly, the remaining findings from the current study demonstrated that ethnic/racial differences in knee pain and physical function could not be fully appreciated without consideration of poverty status. To illustrate, NHW adults living above the poverty line demonstrated significantly better physical functioning on the SPPB in comparison to their NHW counterparts living below the poverty line, as well as NHB adults living below the poverty line. However, there was no significant difference in physical function on the SPPB between NHB and NHW adults both living above the poverty line. Regarding reported knee pain and physical function on the WOMAC, NHW adults living above the poverty line reported doing significantly better (i.e., less knee pain, better physical function) than NHW adults living below the poverty line, as well as NHB adults living above and below the poverty line. Considered another way, NHW adults living above the poverty line reported significantly less knee pain and better physical function on the WOMAC compared to NHW adults living below the poverty line. However, knee pain and physical function on the WOMAC did not significantly differ according to poverty status among NHB adults. NHB adults living above the poverty line have knee pain and physical function that appears to be much worse than what might be expected based upon how well their NHW counterparts living above the poverty line seem to be doing. These results suggest that living above the poverty line may be more advantageous for knee pain and physical function outcomes for NHW adults relative to NHB adults. As such, future research efforts are needed specifically focused on knee OA outcomes in NHB adults living above the poverty line in an effort to determine the reasons for their relatively poor clinical outcomes. Based on results from the present study, it appears that disparities in knee pain and physical function are not simply a function of differences between NHB and NHW adults, but rather these ethnic/racial differences are affected by important indicators of SES such as poverty.

Radiographic findings according to Kellgren-Lawrence grading did not significantly differ between NHB and NHW adults, or between those living above versus below the poverty line. This suggests that differences in knee pain and physical function observed on the basis of ethnicity/race and poverty status are not likely reflective of differential disease severity in the knee joint. Furthermore, the severity of radiographic findings was not overwhelmingly associated with knee pain severity or physical function on the SPPB or WOMAC. Radiographic evidence (particularly X-rays) and corresponding Kellgren-Lawrence grades have variable predictive validity as a marker of self-reported knee pain symptoms, with some studies reporting moderate to strong correlations [43,44] between the two and others reporting weak correlations. [45,46]. This does not mean that joint pathology is unimportant for understanding clinical symptoms of knee OA. It may just be that radiographs (i.e., X-rays) are not the best means for assessing clinically relevant knee joint pathology. Indeed, more recent studies using MRI suggest that bone marrow lesions and synovitis are more reliably associated with knee pain symptoms [47,48]; neither of which can be appreciated using radiographs. It seems very likely, however, that factors above and beyond pathoanatomy must contribute to knee pain severity, as well as ethnic/racial and SES disparities therein. The associations among ethnicity/race, poverty status, and knee OA clinical symptoms observed in this study may be due to other circumstances associated with ethnic/racial minority status and low SES. For example, the average BMI for both NHB and NHW adults in this study met criteria for “obese” classification (>30); however, NHB adults tended to have greater average BMI than their NHW counterparts. Other circumstances that could possibly affect knee OA clinical symptoms include lifestyle choices (smoking, diet, physical activity engagement), neighborhood characteristics (proximity to affordable medical services, walkability), and/or psychosocial factors (perceived helplessness, social support, and perceived discrimination). Whether interventions that target lifestyle choices, neighborhood characteristics, and psychosocial factors might help alleviate disparities in knee pain and physical function is a topic worthy of additional research.

SES is considered one of the most robust predictors of health outcomes [22]. Accordingly, disparities in chronic pain, knee OA, and treatments exist both in terms of racial background and SES. Historically, the effects of ethnic/racial background and SES on health outcomes including knee OA, pain, and physical function have been examined independently in data analytic models by statistically controlling for the effect of the other (e.g., examining ethnic/racial differences in pain while controlling for SES). See [19], for example. Alternatively, recent literature suggests examining the combined, or interactive, effects of ethnicity/race and SES given their inextricable linkage in the United States today [18]. The present study provides support for this suggestion given the significant interaction effects between ethnicity/race and poverty status that we found in relation to knee pain and physical function. Our current findings compliment a growing body of literature addressing the intersectionality of ethnicity/race and SES in explaining chronic pain disparities. For example, in previously conducted research utilizing data from the nationally-representative Health and Retirement Study, it was revealed that NHB adults and individuals in the lowest wealth quartile reported the highest degree of pain-related disability across activity domains [49]. In two separate studies generated from the Johnston County Osteoarthritis cohort, which was comprised of >1/3 older NHB adults, indicators of individual and community SES demonstrated significant associations with radiographic and symptomatic knee OA [50, 51]. Among patients awaiting total knee arthroplasty (TKA), those with higher SES, including greater household income, had lower knee pain and better physical function than low SES patients [52]. Further, patients with higher SES demonstrated a greater likelihood of utilizing TKA, and subsequently having better TKA outcomes following surgery compared to those with low SES [53].

Living in poverty is associated with less access to healthcare, greater engagement in unhealthy behaviors, poorer mental health, and higher rates of morbidity and mortality [54–56]. However, it has previously been reported that the conventional explanations - the poor have less access to healthcare and unhealthy lifestyles - do not fully account for the negative impact of poverty on health outcomes like pain and knee OA [57]. Rather, studies suggest that the psychosocial stresses associated with living in poverty increase the risk for negative health outcomes [58], and the health of NHB adults appears especially vulnerable to poverty-related psychosocial stresses [59]. This may be attributable to the idea that “living in poverty” does not necessarily mean the same thing for NHB adults as it does for NHW adults. Williams and colleagues have reported that, “compared to NHW adults, NHB adults receive less income at the same education levels, have markedly less wealth at equivalent income levels, and have less purchasing power due to higher costs of goods and services in the residential environments where they are disproportionately located” [60]. Furthermore, previous research has shown that when NHB and NHW adults climb the socioeconomic ladder, discriminatory organizational structures often prevent NHB adults from making similar gains to NHW adults despite comparable effort [61]. Realizing this inequality is likely to exacerbate the stress of poverty for NHB adults, and it is already well established that psychosocial stress has a deleterious impact on knee OA outcomes [62–64].

Several limitations should be considered when interpreting the results of the present study. First, the present study focused exclusively on annual household income as an indicator of SES to determine poverty status. Annual household income may be an inadequate representation of the standard of living of retired individuals (23% retired in this study) because it may not reflect available financial resources, and it does not take into consideration the cumulative effects of a lifetime of deprivation or privilege [26]. Further, utilizing current annual household income to determine poverty status does not take into account financial liabilities (i.e., debts) that a household may possess. Thus, a family of three with an annual household income of $30,000 was living above the poverty line according to 2017 Department of Health and Human Services Guidelines; however, if this same family was also experiencing significant debt burden, then they may encounter financial struggles that are similar to a family living below the poverty line. Second, although poverty status is a commonly used indicator of SES throughout the literature, evidence suggests that SES is a multifactorial construct encompassing knowledge, prestige, power, wealth, social connectedness, and various other resources available to an individual [65]. Our sole focus on poverty status, therefore, may not have fully captured the ability of SES to affect ethnic/racial differences in knee pain and physical function. Future research examining associations among ethnicity/race, knee pain, and physical function in populations with or at risk for knee OA should consider examining a wide array of individual-level SES factors (e.g., income, education, employment and type of occupation) as well as community-level SES factors (e.g., neighborhood property values and walkability). Third, the 191 participants included in the final study sample were not evenly distributed across ethnic/racial and poverty status groupings. To illustrate, only 21 NHW adults reported living below the poverty line, relative to 72 NHW adults who reported living above the poverty line, 50 NHB adults who reported living below the poverty line, and 48 NHB adults who reported living above the poverty line. The small sub-sample of NHW adults living below the poverty line in this study may limit the generalizability of our findings to the broader population of these individuals. Replication of our study findings among large sub-samples of individuals according to ethnic/racial background and SES indicators like poverty status seems pertinent. Lastly, the present study included NHB adults as the only ethnic/racial minority in which to compare to NHW adults. Future research examining disparities in knee OA and pain would benefit from including other ethnic/racial minority groups in addition to NHB individuals.

Despite these study limitations, it appears that NHW adults living above the poverty line likely experience low severity knee pain while demonstrating the best physical function. Conversely, NHB adults living below the poverty line appear to be at greatest risk for experiencing severe knee pain and poor physical function. It has been argued that ethnicity and race are social constructs that are influenced to a greater extent by culture and societal norms than genetic diversity [66]. Therefore, to fully appreciate ethnic/racial disparities in pain outcomes, one must consider and appreciate the social context in which such disparities manifest. Results of the present study add to the existing literature by emphasizing the importance of considering SES in studies of racial and ethnic pain disparities. Consistent with the recommendations of Meghani and Chittams [18], it is prudent to examine the interaction between ethnicity/race and SES, not just the unique effect of one while statistically controlling for the other. Along this line, it is important to identify and disentangle contributors to pain disparities, such as ethnicity/race and SES, in order to target appropriate interventions for those at highest risk for poor pain and physical function outcomes. For example, NHB adults living below the poverty line may especially benefit from targeted healthcare approaches to help prevent the development of painful knee OA, or mitigate the pain and accompanying symptoms of knee OA once it has already developed [67].