Race, Sex, and Total Knee Replacement Consideration: Role of Social Support

Authors


Arthritis Research Center, 3347 Forbes Avenue, Suite 220, Pittsburgh, PA 15213. E-mail: evina1@pitt.edu

Abstract

Objective

To determine whether there are racial differences in social support among patients with knee osteoarthritis (OA) and whether the impact of social support on patient preferences for total knee replacement (TKR) varies by race and sex.

Methods

A total of 514 white and 285 African American patients with knee OA were surveyed. Logistic regression models were performed to determine if the relationship between willingness to undergo TKR and the interaction of patient race and sex was mediated by social support.

Results

Compared to whites with knee OA, African American patients were less likely to be married (P < 0.001), reported less close friends/relatives (P < 0.001), and had lower Medical Outcomes Study Social Support Scale (MOS-SSS) scores (P < 0.001). African American patients were also less willing to undergo TKR (62% versus 80%; P < 0.001) than whites. The odds of willingness to undergo TKR were less in white females compared to white males when adjusted for recruitment site, age, income, and the Western Ontario and McMaster Universities Osteoarthritis Index score (odds ratio [OR] 0.57, 95% confidence interval [95% CI] 0.34–0.96). This difference was no longer significant when further adjusted for marital status, number of close friends/relatives, and MOS-SSS score, but the effect size remained unchanged (OR 0.60, 95% CI 0.35–1.02). The odds of willingness to undergo TKR remained much less in African American females (OR 0.35, 95% CI 0.19–0.64) and African American males (OR 0.28, 95% CI 0.14–0.54) compared to white males when controlled for sociodemographic, clinical, and social support measures.

Conclusion

African American patients reported less structural and functional social support than whites. Social support is an important determinant of preference for TKR surgery only among whites.

INTRODUCTION

According to an Institute of Medicine report on racial/ethnic inequities in US health care, variations in the utilization of medical procedures by race exist even after controlling for insurance status, income, age, and clinical factors ([1]). The report also acknowledges that disparities in utilization of medical procedures may be largely due to differences in patient preferences for care. Multiple studies have documented that total knee replacement (TKR), an effective and cost-effective intervention, is less utilized by African Americans as compared to whites ([2-4]), while others have reported that African American patients with osteoarthritis (OA) are less willing to have the procedure done than their white counterparts ([5-7]).

Similarly, reports have shown a significant underutilization of TKR surgery for severe OA in women as compared to men ([8]), and others have suggested that sex-specific patient preferences may influence decisions regarding utilization of the procedure ([9]). Studies have demonstrated that factors including patient psychosocial features, perceptions of procedure effectiveness and risk, and trust in physicians influence these racial and sex differences in willingness to have TKR ([6, 7]). Minimal attention, though, has been paid to the extent to which the quantity and quality of social support may affect these racial and sex differences in patients' preferences for TKR.

Social support has been conceptualized in different ways and can have both positive and negative effects on an individual's well-being ([10]). Most commonly, social support refers to a social network's provision of psychological and material resources intended to benefit an individual's ability to cope with stress ([11]). It has both structural and functional dimensions ([12, 13]). Structural support refers to the existence and quantity of social relationships and the interconnectedness of a person's social network ([12]). In contrast, functional support refers to the degree to which interpersonal relationships serve particular functions ([12]).

In general, African Americans are less likely to receive structural and functional social support than whites ([14-18]). In addition, women are more likely to have limited social support than men ([19-21]). Studies have also shown that higher social support is associated with better patient decision making in regard to medication adherence and other health-related behaviors (e.g., diet control) for a variety of conditions ([13, 22]). None of these studies in decision making, though, has stratified their results by race or sex. Moreover, no study has ever examined the importance of social support as a determinant of willingness to consider an elective surgical procedure.

The purpose of this study is to determine whether there are racial differences in social support among patients with knee OA, and whether social support contributes to the treatment preferences of African American and white patients for TKR. We will also examine whether the impact of social support on preference for TKR varies by race and sex. We hypothesize that compared to white patients, African American patients with knee OA will have a weaker structural and functional social support system and that social support will be strongly associated with TKR preferences, especially in women.

Box 1. Significance & Innovations

  • African American patients with knee osteoarthritis (OA) had lower preference for total knee replacement (TKR) surgery compared with white patients.
  • African American patients with knee OA reported less structural and functional social support than white patients.
  • Social support appears to be an important determinant of TKR preference among white patients but not African American patients.
  • Social support does not seem to mediate the sex difference in willingness to consider TKR.

PATIENTS AND METHODS

Study sample

Participants were recruited from multiple sources. Direct mailing was implemented to recruit potential participants from research and clinic registries. Patients were also recruited from the University of Pittsburgh Medical Center Primary Care and Rheumatology Clinics, and the Veterans Affairs (VA) Pittsburgh Healthcare System Primary Care Clinics. Additional participants were recruited via advertisements (e.g., local newspaper advertisements). The study was approved by both University of Pittsburgh and VA Institutional Review Boards.

Individuals ages ≥50 years interested in participating in the study were asked to call a phone number to undergo a brief screening procedure. To assess the presence of chronic knee pain and the identification of individuals likely to have knee OA, we used 2 items from the arthritis supplement of the National Health and Nutrition Examination Survey I (NHANES-I) ([23]). Questions focused on the presence of pain in the knee on at least half of the days over the past month, and pain in either knee for more than 6 months. Candidate subjects that answered “yes” to both questions were categorized as having chronic, frequent knee pain. At this stage, patients were also asked to self-identify their race.

Individuals who screened positive for knee OA using the above questions proceeded to the next phase of screening, which used the Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC) to assess disease severity ([24]). Potential participants were also asked if they had a knee radiograph performed of the relevant joint within the past 2 years. Those without a recent radiograph were asked to have a knee radiograph performed. All knee radiographs were evaluated using the Kellgren/Lawrence (K/L) scoring system ([25]). Determination of the presence or absence of knee OA was based on the classification criteria developed by the American College of Rheumatology (ACR) ([26]).

Inclusion to the study was based on the following criteria: African American or white, age ≥50 years, presence of chronic, frequent knee pain based on the NHANES, moderate to severe knee OA based on a WOMAC score ≥39, radiographic evidence of OA (i.e., K/L grade ≥2), and presence of knee OA by ACR criteria. Exclusion criteria were as follows: prior history of any major joint replacement, terminal illness (e.g., end-stage cancer), inflammatory arthritis (i.e., rheumatoid arthritis, connective tissue disease, seronegative spondyloarthropathy, or crystal-induced arthropathy), or dementia.

Study measures

Patients meeting study criteria were invited to complete a face-to-face interview administered by trained research staff.

Sociodemographic data

Demographic characteristics included age, sex, educational attainment (≤ high school/general equivalency diploma, certificate program/vocational training/associate's or bachelor's degree, graduate degree, or other), employment status (full time, part time, unemployed, disabled, and retired), household income (<$4,999, $5,000–9,999, $10,000–14,999, $15,000–19,999, $20,000–29,999, $30,000–39,999, $40,000–49,999, and ≥$50,000), and type(s) of health insurance (Medicare, Medicaid, Medigap, private, health maintenance organization, and self-free care).

Social support

Three measures of structural social support were obtained: 1) current marital status (married/living with partner or not), 2) current household composition (alone, spouse/significant other, or child/other relative/nonrelative), and 3) number of close friends and relatives. Note that from this point forward, “married” will be defined as those who are legally married or living with a domestic partner. Functional social support was assessed using a 5-item modified version of the Medical Outcomes Study social support scale (MOS-SSS) that represents 4 dimensions of functional social support: emotional/informational, tangible, affectionate, and positive social interaction ([12]). The mean of responses to all items constituted the overall functional social support score. Range is from 0–20, with a higher score indicating more social support. This shortened MOS-SSS has a Cronbach's α of 0.90 and has been used in other behavioral studies, including those with chronic illness ([27, 28]).

Willingness to consider joint replacement

A single question, developed by Hawker et al for use in a population- based study ([8]), was used to determine willingness to consider joint replacement: “If your knee pain were ever to get severe, would you be willing to have surgery to replace your knee if your doctor recommended it?” This question has a 5-category ordinal response scale. For the current analysis, responses were dichotomized as willing (“definitely willing” and “probably willing”) or unwilling (“unsure,” “probably not willing,” and “definitely not willing”) as in previous studies ([7]).

Clinical covariates

Disease severity

Severity of disease was assessed both radiologically and clinically. Knee radiographs, using the K/L grading system, were used to assess radiologic severity ([29]). The WOMAC index was utilized to assess clinical severity. This reliable and validated scale was designed specifically to assess lower extremity pain and function in OA ([24]). Summary scores range from 0–100, where a higher score indicates increased pain, stiffness, and functional limitations.

Comorbidity

Medical comorbidity was assessed using an interviewer-based modification of the Charlson Comorbidity Index ([30]). This measure is a weighted index designed to evaluate the longitudinal risk of mortality attributable to comorbid disease.

Quality of life

Overall quality of life was assessed using the Short Form 12 health survey, version 2, from which the physical and mental component summary scores were calculated ([31]). This 3- to 4-minute survey has been well-validated with large and heterogeneous samples of adults in the US. It has excellent reliability with a Cronbach's α ranging from 0.81–0.88 for each subscale ([32]). Higher scores indicate better health.

Mental health

Presence of major depression was assessed using the Patient Health Questionnaire 9 (PHQ-9). The PHQ-9 consists of 9 items based on the Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition, depressive disorders criteria. The PHQ-9 total score is a sum of all items, where higher scores indicate increased severity of depressive symptoms. It is a reliable (Cronbach's α = 0.86–0.89) and valid (overall accuracy 85%, sensitivity 75%, and specificity 90%) measure of depression severity ([33]).

Statistical analysis

Demographic, psychosocial, and clinical characteristics were compared by race using t-tests or Wilcoxon's rank sum tests for continuous variables, and Pearson's chi-square tests for categorical variables. In the same manner, these characteristics were then contrasted by willingness to undergo TKR surgery.

Separate logistic regression models were used to assess the association between willingness to undergo TKR surgery and each measure of social support (marital status, number of close friends or family, and MOS-SSS score). The number of close friends/family was transformed into quartiles and entered into the logistic models as a linear categorical variable. Both unadjusted and adjusted logistic regression models were performed. The following covariates were selected a priori for inclusion in the adjusted models: site (VA or non-VA), age, sex, household income, and WOMAC total score. Statistical significance for the association of each social support measure on willingness was calculated using likelihood ratio tests. The above analyses were repeated after stratifying by race to investigate possible effect modification.

Logistic regression models were also performed to determine if the relationship between willingness to undergo TKR surgery and the interaction of patient race and sex is mediated by social support. The initial model in these analyses included race by sex categories only (white male, white female, African American female, and African American male), entered as dummy variables as the independent variable. Sociodemographic and clinical variables were then added to the initial model. For models 3 and 4, structural (marital status, number of close friends or family) and functional (MOS-SSS score) measures of social support were then serially added to determine whether these covariates may explain the association among the individual race and sex categories with willingness to undergo TKR. Statistical significance of the association of the combined race by sex variable with willingness in each model was determined using likelihood ratio tests for heterogeneity. A significant change in odds ratio (OR; >20%) after the addition of the potential mediating variable(s) suggests mediation.

In all regression models, 95% confidence intervals (95% CIs) were calculated using robust variance estimates. Statistical significance was also set with an alpha level of 0.05. All analyses were performed using STATA software, version 12.0.

RESULTS

Demographic characteristics

A total of 514 white and 285 African American patients participated in the study. Compared to white patients, African American patients were younger (P < 0.001) and more likely to be female (P < 0.001). They were also less likely to have a graduate degree, less likely to be employed, and more likely to have lower income (Table 1).

Table 1. Demographic, clinical, psychosocial, and social support characteristics by race*
CharacteristicsWhite (n = 514)African American (n = 285)Pa
  1. Values are the number (percentage) unless indicated otherwise. GED = general equivalency diploma; HMO = health maintenance organization; VA = Veterans Affairs; WOMAC = Western Ontario and McMaster Universities Osteoarthritis Index; PHQ-9 = Patient Health Questionnaire 9; SF-12 = Short Form 12 health survey; MOS-SSS = Medical Outcomes Study social support scale.
  2. aBy Pearson's chi-square or t-test.
  3. bIncludes certificate program or vocational training.
  4. cIndividuals may be included in multiple categories.
  5. dLives with child, other relative, or nonrelative.
Demographic   
Age, mean ± SD years64.54 ± 9.3958.68 ± 8.13< 0.001
Female sex302 (58.8)207 (72.6)< 0.001
Education  < 0.001
≤ High school or GED200 (39.1)170 (60.1) 
Postsecondary training, associate's/ bachelor's degreeb194 (38.0)98 (34.6) 
Graduate degree117 (22.9)15 (5.3) 
Income   
<$10,00032 (6.9)84 (31.7)< 0.001
$10,000–19,99972 (15.6)83 (31.3) 
$20,000–29,99975 (16.2)35 (13.2) 
$30,000–39,99950 (10.8)17 (6.4) 
$40,000–49,99951 (11.0)12 (4.5) 
≥$50,000183 (39.5)34 (12.8) 
Employment status  < 0.001
Full time145 (28.4)70 (24.6) 
Part time51 (10.0)23 (8.1) 
Unemployed36 (7.0)46 (16.2) 
Disabled53 (10.4)81 (28.5) 
Retired226 (44.2)64 (22.5) 
Health insurancec   
Medicare225 (46.2)98 (35.5)0.004
Medicaid53 (11.2)63 (22.9)< 0.001
HMO or private/group381 (77.6)161 (57.7)< 0.001
No insurance12 (2.6)21 (7.7)0.001
Recruitment site   
Non-VA467 (90.9)278 (97.5)< 0.001
VA47 (9.1)7 (2.5) 
Clinical characteristics, mean ± SD   
No. of comorbidities3.22 ± 1.713.56 ± 1.690.007
WOMAC score42.98 ± 15.2753.98 ± 15.10< 0.001
Psychosocial characteristics, mean ± SD   
PHQ-9 depression4.77 ± 4.676.63 ± 5.5< 0.001
SF-12 physical health38.50 ± 11.0136.25 ± 10.000.004
SF-12 mental health53.58 ± 8.9648.80 ± 10.91< 0.001
Social support measures   
Currently married275 (53.5)65 (22.8)< 0.001
Household composition   
Lives alone172 (33.5)142 (49.8)< 0.001
Lives with spouse/significant other284 (55.3)62 (21.8)< 0.001
Otherd96 (18.7)94 (33.0)< 0.001
No. close friends/relatives, mean ± SD10.31 ± 13.137.52 ± 8.88< 0.001
MOS-SSS score, mean ± SD15.17 ± 4.7913.44 ± 5.26< 0.001

Clinical and psychosocial characteristics

African American knee OA patients had worse OA severity (P < 0.001), depressive symptoms (P < 0.001), physical health (P = 0.005), and mental health (P < 0.001) than white patients. In addition, African American patients, compared to white patients, had worse social support, being less likely to be married (P < 0.001), having fewer numbers of close friends/relatives (P < 0.001), and being more likely to live alone (P < 0.001). Finally, mean MOS-SSS scores were also lower in African Americans as compared with white participants (P < 0.001).

Association of social support with willingness to undergo TKR

Most patients were willing to consider undergoing TKR (n = 585 of 793) (Table 2). In the full sample, 83% of OA patients who were married were willing to undergo TKR, compared with 67% among those who were not married (crude OR 2.30, 95% CI 1.63–3.24). After adjusting for recruitment site, sex, age, income, and WOMAC score, this OR was attenuated but remained statistically significant (adjusted OR 1.72, 95% CI 1.14–2.60). Our regression analysis did not show a statistically significant association with willingness and number of close friends or family (adjusted OR 1.05, 95% CI 0.91–1.21). Although the mean ± SD MOS-SSS scores of those who were willing and those who were unwilling to have TKR were 15.0 ± 4.8 and 13.5 ± 5.5, respectively (P < 0.001), the association between willingness to have the procedure done and MOS-SSS score did not persist after controlling for sociodemographic and clinical variables (adjusted OR 1.03, 95% CI 1.00–1.07).

Table 2. Logistic regression analyses of willingness to undergo arthroplasty and the individual social support measures, stratified by race*
 WillingUnwillingUnadjustedAdjusteda
OR (95% CI)bPcOR (95% CI)bPc
  1. OR = odds ratio; 95% CI = 95% confidence interval; MOS-SSS = Medical Outcomes Study social support scale.
  2. aAdjusted for site (Veterans Affairs [VA] or non-VA), sex, age, income, and Western Ontario and McMaster Universities Osteoarthritis Index total score.
  3. b95% CIs are calculated using robust variance estimates.
  4. cBy likelihood ratio test.
  5. dTransformed into quartiles and entered into models as linear categorical variable.
All patients585208    
Married, no.279592.30 (1.63–3.24)< 0.0011.72 (1.14–2.60)0.010
No. close friends/familyd      
Quartile 114055    
Quartile 2146521.07 (0.94–1.22)0.3381.05 (0.91–1.21)0.489
Quartile 38333    
Quartile 421668    
Mean ± SD MOS-SSS score15.0 ± 4.813.5 ± 5.51.06 (1.03–1.09)< 0.0011.03 (1.00–1.07)0.079
White patients409102    
Married, no.237372.42 (1.54–3.79)< 0.0011.87 (1.11–3.15)0.029
No. close friends/familyd      
Quartile 19322    
Quartile 290241.01 (0.84–1.20)0.9521.03 (0.84–1.26)0.801
Quartile 36317    
Quartile 416339    
Mean ± SD MOS-SSS score15.5 ± 4.513.7 ± 5.51.08 (1.03–1.13)< 0.0011.06 (1.00–1.12)0.036
African American patients176106    
Married, no.42221.20 (0.67–2.15)0.5441.15 (0.61–2.17)0.671
No. close friends/familyd      
Quartile 14733    
Quartile 256281.05 (0.85–1.28)0.6691.02 (0.82–1.27)0.843
Quartile 32016    
Quartile 45329    
Mean ± SD MOS-SSS score13.6 ± 5.013.3 ± 5.61.01 (0.97–1.06)0.6011.01 (0.96–1.06)0.750

Among white patients with knee OA, 87% who were married were willing to undergo TKR, compared with 66% of patients who were not married. After adjusting for sociodemographic and clinical variables, the odds of willingness to undergo TKR were nearly twice as high among those who were married compared to those who were not (adjusted OR 1.87, 95% CI 1.11–3.15). In this group of patients, the mean ± SD MOS-SSS score was also higher among those who were willing to have TKR, compared to those unwilling to have the procedure (15.5 ± 4.5 and 13.7 ± 5.5, respectively; P < 0.001). The association between willingness and MOS-SSS score remained significant even after controlling for sociodemographic and clinical variables (adjusted OR 1.06, 95% CI 1.00–1.12). Among African American patients with knee OA, willingness to undergo TKR was not statistically significantly associated with marital status, number of close friends/relatives, or MOS-SSS score.

Association of race/sex with willingness to undergo TKR

Associations between the individual races by sex categories with willingness to undergo TKR are presented in Table 3. The odds of willingness to undergo TKR were less in white females compared to white males when adjusted for recruitment site, age, income, and WOMAC score (model 2: adjusted OR 0.6, 95% CI 0.3–1.0). This difference was no longer statistically significant when further adjusted for marital status, number of close friends/relatives, and MOS-SSS score (model 4: adjusted OR 0.6, 95% CI 0.4–1.0), but the OR did not change appreciably.

Table 3. Relationships between patient race/sex and willingness to undergo arthroplasty*
 Willing, no.Unwilling, no.OR (95% CI)a
  1. OR = odds ratio; 95% CI = 95% confidence interval.
  2. a95% CIs are calculated using robust variance estimates.
  3. bModel 1 = race by sex only; likelihood ratio test for heterogeneity; P ≤ 0.001.
  4. cModel 2 = model 1 plus site (Veterans Affairs [VA] or non-VA), age, income, and Western Ontario and McMaster Universities Osteoarthritis Index total score; likelihood ratio test for heterogeneity; P ≤ 0.001.
  5. dModel 3 = model 2 plus marital status plus number of close friends or family; likelihood ratio test for heterogeneity; P ≤ 0.001.
  6. eModel 4 = model 3 plus Medical Outcomes Study social support scale score; likelihood ratio test for heterogeneity; P ≤ 0.001.
Model 1b   
White male (referent)17735 
White female232670.68 (0.44–1.08)
African American female131750.35 (0.22–0.55)
African American male45310.29 (0.16–0.51)
Model 2c   
White male (referent)16328 
White female205570.57 (0.34–0.96)
African American female119640.33 (0.18–0.60)
African American male44310.26 (0.13–0.52)
Model 3d   
White male (referent)16328 
White female205570.59 (0.35–1.01)
African American female119640.36 (0.20–0.65)
African American male44310.28 (0.14–0.55)
Model 4e   
White male (referent)16328 
White female205570.60 (0.35–1.02)
African American female119640.35 (0.19–0.64)
African American male44310.28 (0.14–0.54)

The odds of willingness to undergo TKR were also less in African American females (adjusted OR 0.3, 95% CI 0.2–0.6) and African American males (adjusted OR 0.3, 95% CI 0.1–0.5) compared to white males when adjusted for recruitment site, age, income, and WOMAC score (model 2). These differences in odds remained significant when further adjusted for marital status, number of close friends/relatives, and MOS-SSS score (model 4: adjusted OR 0.4, 95% CI 0.2–0.6 for African American females and adjusted OR 0.3, 95% CI 0.1–0.5 for African American males).

Association of race by depression level and physical health with willingness to undergo TKR

Relationships between race by level of depression and by physical health level with willingness to undergo TKR surgery are presented in Supplementary Tables 1 and 2, respectively (available in the online version of this article at http://onlinelibrary.wiley.com/doi/10.1002/acr.21925/abstract). The results are similar to each other. In summary, among whites, the odds of willingness to undergo TKR were no different regardless of depression or physical health level. In addition, the odds of willingness to undergo TKR remained lower among African American patients with low and high levels of depression (or physical health) compared to whites with low depression (or high physical health).

DISCUSSION

In this diverse group of patients with moderate to severe knee OA, we demonstrated that African American patients were less likely to have structural and functional social support than white patients. We have also shown that in this sample, as in prior studies, African American patients were less likely than white patients to consider TKR as a treatment option. Our study is also the first to show that structural social support, measured by marital status, and functional support appear to be important determinants of preference for TKR in white patients, but not African American patients. Finally, our study demonstrates a significant race and sex interaction in willingness to undergo TKR.

Besides confirming a previous report that African American male veterans with OA tend to have less structural and functional support than white male veterans ([15]), our study extends the generalizability of this finding to female and nonveteran patients with OA. This racial/ethnic disparity in measures of structural and functional support has also been observed in other study populations ([14, 16-18]). Among younger adults in the general population, African Americans were found to be less likely to receive social support, more likely to have smaller social networks, and more likely to have support come from family members compared to whites ([14, 18]).

Indeed, the finding that African American patients with OA have weaker support systems than white OA patients has important implications. Low social support has been related to greater symptoms from depression in patients with OA ([34]). Conversely, greater support has been related to lower depressive symptoms, greater life satisfaction, and more enhanced health-related quality of life measures in OA ([35, 36]). Social support also moderates the effects of pain and functional limitation due to OA ([36]). Clinical trials have, indeed, shown that support groups for OA patients improve health and limit health care costs ([37, 38]). In addition, spousal support has been found to predict better improvement in knee limitations following TKR ([39]). The perceived adequacy of assistance from others has also been significantly associated with WOMAC function months after joint replacement ([40]).

Other studies, though, have observed similar levels of structural and functional support in African American and white adults in the general population ([41, 42]). In a study of patients with hypertension, African American patients were less likely to be married than whites, while there was a similar distribution of emotional and financial support across groups ([43]). One study of patients with diabetes mellitus demonstrated no differences in social support index (assessed by the number of sources of support in one's life) by race/ethnicity, whereas another study showed that, compared with white patients, African American patients with diabetes mellitus had higher frequencies of contacts with relatives, friends, and neighbors ([44, 45]). However, comparison of studies in social support is often difficult due to the focus on select populations such as the elderly or pregnant women ([16, 17, 44]), failure to adjust for confounding variables such as social class ([18]), and variations in the measures of social support used.

Different pathways may explain the relationship between social support and health decision making ([13]). Structural support may provide individuals with a sense of belonging that can lead them to feel greater self-worth, thereby motivating them to care for themselves ([46]). This motivation can then lead them to make more appropriate health decisions. Having a large number of structural supports may also mean greater access to sources of information, which increases the likelihood of having access to accurate and relevant information sources ([10]). Peer pressure can also influence health decisions and engagement in positive health-relevant activities ([11]).

Functional support, on the other hand, may have a positive effect on health decision making during periods of distress ([11, 13]). When an individual perceives high levels of social support, he or she may reassess a stressful event as less worrisome and become more capable to deal with decisions related to the disease ([47]). The perceived availability of individuals can reduce the presence of intrusive thoughts, which can also lead to better decision making ([48]). If people offer advice, material aid, or behavioral assistance (i.e., informational and tangible support), then a person may also feel more capable to make health-related decisions ([10]). In fact, in a meta-analysis that examined the association between social support and patient treatment adherence, functional measures of social support tend to be more strongly related to adherence than to structural measures ([22]).

Among white patients in our sample, women were less willing to undergo TKR than men; however, social support does not seem to significantly influence this sex difference. Sex population studies have suggested that social supports and networks may be more important for women's mental health than men's mental health ([27, 49]). Yet, conflicting studies also exist, suggesting that sex differences in the importance of social support may be explained in part by situational differences and the ways that culture-specific sex roles are affected by the situations ([27, 50]). Regardless, the decreased willingness of African Americans of both sexes to undergo TKR compared to white males persisted despite social support quantity and quality.

Although we and others have described multiple factors that contribute to racial/ethnic differences in willingness to consider joint replacement, which may also contribute to racial/ethnic differences in the utilization of total knee/hip joint replacement ([5-7]), we did not find that lower social support was associated with decreased preference for TKR among African Americans of either sex. These results seem to indicate that decreased social support is not an important determinant for TKR decision making among African American patients. In contrast, our study demonstrates that structural and functional social support were important determinants in white patients' decisions regarding TKR. Moreover, neither depression nor physical health seems to mediate this racial difference in willingness to undergo TKR. These findings support the idea that other sociocultural beliefs and attitudes impact treatment preferences among different racial groups. In previous studies, we and others found that racial/ethnic variations in willingness to undergo total knee/hip joint replacement may also be largely attributed to varying expectations in procedure benefit and outcome, different perceptions of usefulness of self-care methods such as prayer, and different levels of trust in the medical provider ([5-7]).

Our study has several limitations. First, our sample is composed primarily of African American and white patients; OA patients in other racial/ethnic groups were not represented in our sample. Therefore, our findings may not be generalizable to other racial/ethnic groups. Yet, we also recruited patients from multiple sources, not relying on clinic-based recruitment, thereby strengthening our study's external validity. In fact, we found that our study cohort's sociodemographic profile is similar to that of US Census Bureau participants living in Allegheny County, Pennsylvania, and a comparable cohort of knee OA patients recruited in another study (see Supplementary Table 3, available in the online version of this article at http://onlinelibrary.wiley.com/doi/10.1002/acr.21925/abstract) ([51, 52]). Second, subjects volunteered for our study. Patients with less social support may not have the time and energy to participate in our study. Therefore, the measured relationship between social support and willingness to undergo TKR may be underestimated. Lastly, our study may be subject to residual confounding. Not all potential confounding variables that may affect the association between race/sex and willingness to undergo TKR were assessed. For instance, we did not examine other possible determinants of OA care preferences such as patient risk aversion or provider bias.

In summary, in this sample of patients with OA who are potential candidates for TKR, social support may in part explain variations in patient preference regarding TKR. There is also a significant race and sex interaction in patient preferences for TKR surgery. Social support does not seem to mediate the sex difference in willingness to consider TKR. On the other hand, social support impacts patient preferences for TKR surgery among whites, but not among African Americans. Future studies should evaluate the specific aspects of social support that could be targeted for intervention in the national effort to reduce or eliminate disparities in access and utilization of TKR in the management of end-stage OA.

AUTHOR CONTRIBUTIONS

All authors were involved in drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be submitted for publication. Dr. Vina had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study conception and design. Vina, Cloonan, Ibrahim, Hannon, Boudreau, Kwoh.

Acquisition of data. Vina, Cloonan, Ibrahim, Hannon, Boudreau, Kwoh.

Analysis and interpretation of data. Vina, Cloonan, Ibrahim, Hannon, Boudreau, Kwoh.

Ancillary