Clinical appropriateness and not race predicted referral for joint arthroplasty




To understand the reasons behind racial disparities in the use of total joint arthroplasty (TJA), we sought to examine the predictors of time to referral to orthopedic surgery for consideration of joint replacement.


In this prospective, longitudinal study of 676 primary care clinic patients with at least a moderately severe degree of hip or knee osteoarthritis (OA), we examined the effects of race, health beliefs (i.e., perceived benefits and risks) of TJA, and clinical appropriateness of TJA on referral to orthopedic surgery.


The sample included 255 African Americans (38%) and 421 whites (62%); 523 patients had knee OA (78%) and 153 had hip OA (22%). Subjects were 60% male, with a mean ± SD age of 64 ± 9 years, a mean ± SD body mass index of 33.6 ± 8 kg/m2, and a mean ± SD summary Western Ontario and McMaster Universities Osteoarthritis Index score of 56 ± 14, suggesting moderately severe OA. At baseline, African Americans perceived fewer benefits and greater risk from TJA than whites. There were no significant racial group differences in the proportions of cases deemed clinically appropriate for TJA. After controlling for potential confounders, clinical appropriateness (hazard ratio [HR] 1.95, 95% confidence interval [95% CI] 1.15–3.32; P = 0.01) predicted referral to orthopedic surgery. Neither race (HR 1.30, 95% CI 0.94–2.05; P = 0.1) nor health beliefs (HR 1.0, P = 0.5) were associated with referral status.


In this sample of primary care clinic patients, African Americans and whites were equally likely to be referred by their physicians to orthopedic surgery. Clinical appropriateness predicted future referral to orthopedic surgery, and not race or TJA-specific health beliefs.


Osteoarthritis (OA), the most common form of arthritis, accounts for as much disability and dependency in lower extremity functioning among the elderly as any disease (1, 2). Treatment options are varied, and include nonpharmacologic approaches (e.g., weight loss), use of analgesic agents and nonsteroidal antiinflammatory drugs (NSAIDs) for relief of symptoms, and total joint arthroplasty (TJA) for end-stage disease. Based on existing research evidence, TJA is a relatively safe and cost-effective treatment for alleviating pain and restoring physical function in patients who do not respond to nonsurgical therapies (3, 4).

Despite these benefits, marked ethnic differences in the utilization of TJA are well documented. Over the past 15 years, numerous studies have reported that African Americans received TJA less often than whites (5–11). The reasons for these reported differences in utilization are likely multifactorial and include patient-level factors (e.g., health beliefs), system-level factors (e.g., access to specialist care), and provider-level factors (e.g., physician biases) (12–14).

Because referral to an orthopedic surgeon is the chief means by which the use of TJA is made available to patients with OA, it is important to study the mechanisms of racial disparities that may be operating at the patient–primary care physician level. As the medical gatekeeper, primary care physicians have an important role in deciding when to refer a patient to an orthopedic surgeon for consideration of TJA. Before making such a decision, the primary care physician must take into consideration 2 sets of issues: patient preferences (i.e., beliefs regarding the benefits and risks of TJA) and the clinical appropriateness of the procedure. In 2 survey studies of primary care physicians, physicians self-reported that clinically relevant variables, and not race, were the main determinant in their decision to refer a hypothetical patient to an orthopedic surgeon (15, 16). Because a physician response to a hypothetical clinical scenario may differ from his actual practice when confronted with a similar problem in the clinic (17), we deemed it necessary to evaluate how clinically relevant variables, race, and health beliefs may influence referral for TJA. Determining predictors of referral to an orthopedic surgeon is an important first step toward eliminating racial differences in the use of TJA.

In this longitudinal cohort study of 676 patients with OA, we sought to identify baseline patient characteristics that predicted time to referral to an orthopedic surgeon. Consistent with the known disparities in health care use in different medical and surgical specialties (18), we hypothesized that African Americans would be less likely to be referred to orthopedic surgery for knee or hip OA than their white counterparts. Second, because we and others have previously reported that African Americans were less likely than whites to perceive the benefits of TJA and more likely to recognize barriers to TJA (14, 19, 20), we hypothesized that these racial differences in TJA-specific health beliefs would influence referral to orthopedic surgery. Finally, we hypothesized that clinical appropriateness to undergo TJA would be a stronger predictor of referral to orthopedic surgery than race.


Design and patient population.

This was a 2-year prospective observational study designed to understand the reasons behind ethnic disparities in health care utilization among patients with OA. Patients were enrolled from the primary care clinics of the Roudebush Veterans Affairs Medical Center (VAMC) and the Wishard Hospital. The Wishard Hospital is a county hospital with an established primary care network consisting of its primary care centers and 6 community health centers located in neighborhoods throughout Indianapolis, Indiana. Approximately one-half mile from the county hospital, the Roudebush VAMC serves veterans from Indiana and surrounding states.

The initial identification of potential participants has been detailed elsewhere (19). To be eligible for this study, patients had to be ≥50 years of age, have radiographic evidence of OA on the symptomatic joint, and have a Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC) summary score ≥30. The WOMAC summary score of ≥30 was chosen based on a previous study that demonstrated that the mean ± SD WOMAC summary score for patients with knee OA undergoing preoperative evaluation for total knee arthroplasty was 27.6 ± 2.9 (21). Notably, subjects were not required to meet specific radiographic criteria (e.g., Kellgren/Lawrence grade ≥3) for OA. Patients who had already undergone knee/hip TJA or who had already been referred to an orthopedic surgeon for their knee or hip pain were excluded.

Between March 1, 2003 and September 30, 2006, 1,478 consecutive patients with radiographic evidence of OA were screened for the study. Of these, 740 patients met the study eligibility criteria. Of these, 64 patients (8.5%) declined to participate and 676 (91.4%) enrolled. After completion of baseline assessments, subjects received followup phone calls at the 12-month and 24-month study anniversary dates to assess their arthritis-related management since study entry or since the last phone contact. The mean ± SD length of followup was 1.8 ± 0.6 years.

Study procedures, including written informed consent, were approved by the Indiana University Purdue University Indianapolis and Veterans Administration Institutional Review Boards.

Baseline measures.

The arthritis-related health belief instrument.

The arthritis-related health belief instrument is a 16-item tool with 4 scales (19, 22, 23): 1) perceived severity of arthritis (an individual's perception of the medical and social consequences of having arthritis or of not treating such a condition); 2) perceived susceptibility for arthritis to progress (an individual's subjective perception of the risk or vulnerability for arthritis to progress); 3) benefits of arthroplasty (an individual's beliefs about the likelihood that having TJA will lead to effective treatment of arthritis); and 4) barriers/risks of arthroplasty (barriers are the potential negative aspects of TJA, including cost, amount of time required, inconvenience, side effects, and degree of unpleasantness [e.g., painful, upsetting, difficult]).

The 16 health belief items were measured on a 5-point Likert scale, where 1 = strongly disagree, 2 = disagree, 3 = undecided, 4 = agree, and 5 = strongly agree. For each scale, a higher composite score indicates a heightened perception of the specific latent construct (i.e., severity, susceptibility, benefits, and barriers/risks) being measured. We have previously reported the Cronbach's alpha reliability estimates for each of the scales ranged from 0.7–0.8 (23, 24). Factor analysis indicated that the dimensions of the arthritis-related health belief instrument are sufficiently distinct to be considered different beliefs (22, 23). Establishment of the factorial invariance of the arthritis-related health belief instrument lends further support for the use of the instrument in comparing mean scale scores between African Americans and whites (23). The survey items for each of the 4 health belief scales can be found in Appendix A.

Clinical appropriateness of TJA.

At study entry, clinical appropriateness for TJA was obtained for each patient using an algorithm that included 5 variables: adequacy of medical management, WOMAC pain severity, WOMAC functional limitation, age (50–70 years or >70 years), and medical comorbidity. Based on these 5 variables, the appropriateness algorithm classified each patient as appropriate, uncertain, or inappropriate for TJA (25–27). Appropriate characterizes someone who is severely symptomatic (according to pain and functional limitation scores) despite adequate medical management, and is also healthy enough to withstand the stress of surgery. The appropriateness algorithm has been validated in ≥1,500 patients with OA, and studies have suggested a direct relationship between appropriateness and better health-related quality of life outcomes 6 months after TJA (28).

For adequacy of medical management, data on drugs (acetaminophen, NSAIDs, tramadol, and narcotic analgesics) and referrals to allied health specialists (physical therapy, occupational therapy, and nutrition) were abstracted from the electronic medical record (EMR) databases. Previous medical management was considered adequate if any of the following conditions were met within the 24-month period prior to study entry: 1) filled prescriptions for ≥3 different drugs (≥30 day supply) for hip or knee pain, 2) filled prescriptions for 2 different drugs (≥30 day supply) and received a referral to 1 allied health specialist for hip/knee pain or weight reduction, or 3) filled a prescription for 1 drug (≥30 day supply) and received referrals to ≥2 allied health specialists. If none of the conditions were met, previous medical management was considered inadequate.

For WOMAC disease severity, lower extremity OA disease severity was assessed using the WOMAC, which includes 24 items that probe pain (score range 0–20), stiffness (score range 0–8), and functional limitation (score range 0–68). The WOMAC has been extensively validated and shown to be a reliable and responsive instrument (29). Depending on the WOMAC scores, participants were classified as having a mild, moderate, or severe level of symptoms. For WOMAC pain, scores of 0–8, 9–14, and 15–20 correspond to mild, moderate, and severe pain, respectively. For WOMAC functional limitation, scores of 0–22, 23–45, and 46–68 correspond to mild, moderate, and severe functional limitation, respectively.

Medical comorbidity, which is an indicator of surgical risk, was assessed using the modified Deyo-Charlson comorbidity index (30, 31). The index assigned nonzero weights to 19 conditions based on their risk of mortality (32, 33). The weights can take on values of 1, 2, 3, or 6 and are then summed for each patient. The modified Deyo-Charlson comorbidity index has been shown to predict mortality in a cohort of community-dwelling older adults attending a large primary care practice (34). Depending on the total comorbidity score, participants in the study were classified either in the low (score ≤1) or high (score ≥2) surgical risk category at study entry. The relevant diagnoses were electronically abstracted from the medical record database.


Age, sex, educational level, employment status, marital status, and adequacy of financial situation (“Considering your regular household members' income, would you say that you are… comfortable, just able to make ends meet, unable to make ends meet”) were assessed. Patients were asked to self-identify their ethnicity. EMR abstractions provided information on patients' height, weight, and number and types of insurance coverage.

Outcome measures.

The primary outcome was the length of time in months from study entry to orthopedic surgery referral by the primary care physician. The referral date and the reason for referral were abstracted from the EMR. For the purpose of the study, participants were considered referred if the primary care physician filled out a consult specifically for consideration of TJA, or for treatment of hip or knee OA. We did not consider referrals that were made for reasons such as evaluation and treatment of joint instability (n = 2), meniscal injury (n = 1), fracture (n = 4), avascular necrosis (n = 0), joint injection (n = 2), or arthroscopy (n = 2). It is important to note that this group of participants (n = 11) was not excluded from the cohort, but that they continued to be followed until the end of the 2-year study or until they were referred back to orthopedic surgery for the appropriate reasons.

The secondary outcome was self-reported referral to an orthopedic surgeon as obtained by the research assistant during the month 12 or month 24 followup phone call interviews.

Statistical analysis.

Chi-square tests were used for all categorical variables and t-tests were used for the continuous variables. Cox proportional hazards regression was used to model time to referral (in months) as abstracted from the EMR. All participants were captured in the EMR. Participants who were not referred were censored at the last completed phone interview, the 2-year study anniversary date, or their date of death, whichever came first. The months to referral time variable was available for all 676 participants, but 16 had missing values for some of the baseline variables. Thus, the Cox models were based upon 660 individuals. Three separate models were estimated in which the African American indicator for race was entered first, followed by arthritis-related health beliefs and then by the clinical appropriateness indicators. All 3 models were controlled or adjusted for potential confounders, including sex, education, adequacy of income, insurance coverage, recruitment site (county hospital versus VAMC), index joint (hip versus knee), and body mass index (BMI). With this approach, we were able to reduce or eliminate the confounding effects of demographic and health conditions to obtain an adjusted effect of race on time to referral. Additionally, interactions of race with arthritis-related health beliefs and with clinical appropriateness were examined.

As a secondary outcome, self-reported referral to orthopedic surgery was also examined using logistic regression. Self-report data were available for 609 participants, and 7 of these had missing values at baseline. Therefore, the logistic models were based upon 602 participants. Similar to the Cox regression models, we estimated 3 separate models and controlled for the potential confounders in each model.


Baseline characteristics of the sample.

The sample of 676 patients consisted of 255 African Americans (38%) and 421 whites (62%); 523 (78%) had knee OA and 153 (22%) had hip OA. The mean ± SD age for the entire sample was 64.4 ± 9 years; 60% were men, 64% had at least a high school education, 20% were employed (either full or part time), and 55% reported an annual household income <$15,000. The mean ± SD summary WOMAC score was 56 ± 14, suggesting at least moderately severe symptomatic OA. The 2 groups were similar with respect to educational attainment, employment status, insurance coverage, joint type, WOMAC pain and function scores, and status of clinical appropriateness (Table 1). Compared with whites, African Americans were younger (mean ± SD age 62.6 ± 8.8 years versus 65.7 ± 10.0 years), had higher BMI (mean ± SD 34.4 ± 8.4 kg/m2 versus 33.1 ± 8.0 kg/m2) and fewer comorbid illnesses, and were more likely to be women (58% versus 29%), to report inadequate income (73.5% versus 66.0%), and to be recruited from county hospital–affiliated clinics (63.9% versus 29.7%). Interestingly, fewer African American patients met the preset definition of having received adequate medical management (22.8% versus 37.5%). African Americans perceived less benefit (mean ± SD 9.5 ± 2.4 versus 10.1 ± 2.0) and greater risks (mean ± SD 15.1 ± 3.7 versus 14.1 ± 3.5) from TJA than whites, but did not differ in their perceptions of arthritis severity or susceptibility to progression.

Table 1. Baseline characteristics comparing African Americans and whites*
VariablesAfrican Americans (n = 255)Whites (n = 421)P
  • *

    Values are the mean ± SD unless otherwise indicated. BMI = body mass index; TJA = total joint arthroplasty; OA = osteoarthritis; WOMAC = Western Ontario and McMaster Universities Osteoarthritis Index.

  • From the WOMAC. Greater scores indicate more severe pain/dysfunction.

  • From the arthritis-related health belief instrument; greater scores indicate heightened perception of the target construct.

 Age, years62.6 ± 8.865.7 ± 10.0< 0.0001
 Female, %58.029.0< 0.0001
 ≤12 years of education, %63.564.60.7
 Adequacy of income, % uncomfortable73.566.00.04
 Employed, %19.620.00.9
 ≥2 sources of insurance coverage, %50.649.80.8
 Recruited from county hospital, %63.929.7< 0.0001
Clinical factors   
 BMI, kg/m234.4 ± 8.433.1 ± 8.00.03
 Knee as index joint, %
 Comorbidity index score, %  0.04
 Knee/hip pain (range 0–20)11.8 ± 3.411.5 ± 3.50.2
 Lower extremity function (range 0–68)40.0 ± 11.140.3 ± 10.40.7
 Medical management adequate, %22.837.5< 0.0001
 TJA appropriateness, %  0.04
Health beliefs and perceptions   
 Benefits of TJA9.5 ± 2.410.1 ± 2.00.001
 Barriers to/risks of TJA15.1 ± 3.714.1 ± 3.50.0006
 Severity of OA10.9 ± 2.610.9 ± 2.60.9
 Susceptibility to OA progression17.7 ± 3.817.6 ± 3.40.5

Comparison of subjects based on referral status.

Based on our review of the medical records, 119 subjects (17.6%) were referred by their primary care physician to orthopedic surgery during the followup period. The overall mean ± SD time from study entry to referral was 20.1 ± 7.2 months. Compared with the nonreferred subjects, referred subjects were more likely to be African American (50.4% versus 35.0%), female (52.1% versus 37.3%), younger (mean ± SD 62.1 ± 8.7 years old versus 65.0 ± 9.8 years old), and recruited from the county hospital–affiliated primary care clinics (60.5% versus 38.8%). Of the socioeconomic status variables, only perceived adequacy of income was associated with referral status (Table 2). Compared with subjects who were comfortable with their household income, subjects who were uncomfortable (i.e., “just able to make ends meet” or “unable to make ends meet”) were more likely to be referred (78.2% versus 66.8%).

Table 2. Relationship between baseline variables and status of referral (observed on medical records) to orthopedic surgery*
VariablesReferred (n = 119)Not referred (n = 557)P
  • *

    Values are the mean ± SD unless otherwise indicated. VAMC = Veterans Affairs Medical Center. See Table 1 for additional definitions.

  • From the arthritis-related health belief instrument; greater scores indicate heightened perception of the target construct.

  • From the WOMAC. Greater scores indicate more severe pain/dysfunction.

 Race, %  0.001
  African Americans50.435.0 
 Age, years62.1 ± 8.765.0 ± 9.80.002
 Sex, %  0.002
 Education, %  0.8
  ≤12 years64.764.1 
  >12 years35.335.9 
 Adequacy of income, %  0.01
 Employment status, %  0.7
  Not employed79.080.4 
 Insurance coverage, %  0.4
  ≥2 sources4751 
  ≤1 source5349 
 Recruitment site, %  < 0.0001
  County hospital affiliated60.538.8 
Clinical factors   
 BMI, kg/m235.7 ± 9.233.1 ± 7.90.006
 Knee as index joint, %80.776.70.3
 Comorbidity index, %  < 0.0003
 Knee/hip pain12.6 ± 3.511.4 ± 3.40.0006
 Lower extremity function42.2 ± 11.339.7 ± 10.50.02
 Medical management adequate, %34.531.40.5
 TJA appropriateness, %  0.003
Health beliefs   
 Benefits of TJA10.0 ± 2.49.8 ± 2.20.3
 Barriers to/risks of TJA14.3 ± 4.114.5 ± 3.50.6
 Severity of OA11.2 ± 2.610.8 ± 2.60.1
 Susceptibility to OA progression18.1 ± 3.817.5 ± 3.50.08

Clinically, the referred group had worse WOMAC pain (mean ± SD score 12.6 ± 3.5 versus 11.4 ± 3.4) and function (mean ± SD score 42.2 ± 11.3 versus 39.7 ± 10.5), higher BMI (mean ± SD 35.7 ± 9.2 kg/m2 versus 33.1 ± 7.9 kg/m2), fewer comorbid conditions, and was more clinically appropriate for TJA (36.1% versus 22.8%) at study entry than the nonreferred group. Joint type and adequacy of medical management did not influence referral status. Compared with the nonreferred subjects, referred subjects considered it more likely that their arthritis would progress (mean ± SD 18.1 ± 3.8 versus 17.5 ± 3.5; P = 0.08). However, the referred and nonreferred group did not differ in terms of their perception of the benefits and risks of TJA or in their perception of the severity of arthritis.

Multivariate predictors of referral.

After controlling for potential confounders, race did not predict referral to an orthopedic surgeon (Table 3). None of the health beliefs measures predicted referral. As we expected, clinical appropriateness was an important predictor of referral. Compared with subjects who were inappropriate for TJA at study entry, subjects who were appropriate were almost twice as likely to be referred to orthopedic surgery. The only other variable that was associated with time to referral was recruitment site. Subjects recruited from the county hospital–affiliated clinics were more likely (hazard ratio [HR] 2.7, 95% confidence interval [95% CI] 1.4–5.1, P = 0.0026) than the VAMC participants to be referred to orthopedic surgery. Sex (HR 0.6, 95% CI 0.3–1.2), education (HR 1.2, 95% CI 0.8–1.8), adequacy of income (HR 1.3, 95% CI 0.8–2.1), insurance coverage (HR 0.9, 95% CI 0.6–1.3), joint type (HR 0.9, 95% CI 0.6–1.5), and BMI (HR 1.0, 95% CI 0.9–1.1) were not associated with future referral status. Moreover, the interactions of race with recruitment site (HR 0.8, 95% CI 0.4–1.8, P = 0.7), arthritis-related health beliefs (P = 0.11–0.82), and clinical appropriateness (HR 0.7, 95% CI 0.2–1.9, P = 0.5) were not significant.

Table 3. Predictors of referral (observed on medical records) to an orthopedic surgeon*
 Referral status, HR (95% CI)P
  • *

    All models controlled for potential confounders, including sex, education, adequacy of income, insurance coverage, recruitment site (county hospital–affiliated versus Veterans Affairs Medical Center–affiliated clinics), symptomatic joint (hip versus knee), and body mass index. HR = hazard ratio; 95% CI = 95% confidence interval.

  • Reference group: whites.

  • Reference group: inappropriate.

  • §


Model 1  
 African Americans1.36 (0.93–2.00)0.11
Model 2  
 African Americans1.41 (0.96–2.07)0.08
 Arthritis-specific health beliefs  
  Benefits of arthroplasty1.05 (0.97–1.15)0.2
  Risks of arthroplasty0.98 (0.93–1.04)0.5
  Severity of arthritis1.03 (0.94–1.12)0.5
  Susceptibility of arthritis to get worse1.01 (0.95–1.08)0.6
Model 3  
 African Americans1.39 (0.94–2.05)0.1
 Arthritis-specific health beliefs  
  Benefits of arthroplasty1.05 (0.96–1.15)0.2
  Risks of arthroplasty0.99 (0.94–1.05)0.6
  Severity of arthritis1.01 (0.93–1.10)0.7
  Susceptibility of arthritis to get worse1.00 (0.94–1.07)0.9
 Clinical appropriateness  
  Appropriate1.95 (1.15–3.32)0.01§
  Uncertain1.28 (0.78–2.09)0.3

When self-reported referral was used as the outcome, the findings were similar to the primary analyses except that the recruitment site was no longer significant (odds ratio 1.2, 95% CI 0.6–2.5, P = 0.5) (Table 4). Being appropriate for TJA at study entry was still a significant predictor. Additionally, subjects who were classified as uncertain were more likely than those who were inappropriate to report that they had been referred to orthopedic surgery for their hip or knee OA.

Table 4. Predictors of referral (self-reported) to an orthopedic surgeon*
 Referral status, OR (95% CI)P
  • *

    All models controlled for potential confounders, including sex, education, adequacy of income, insurance coverage, recruitment site (county hospital–affiliated versus Veterans Affairs Medical Center–affiliated clinics), symptomatic joint (hip versus knee), and body mass index. OR = odds ratio; 95% CI = 95% confidence interval.

  • Reference group: whites.

  • Reference group: inappropriate.

  • §


Model 1  
 African Americans1.30 (0.89–1.91)0.16
Model 2  
 African Americans1.35 (0.92–1.99)0.12
 Arthritis-specific health beliefs  
  Benefits of arthroplasty1.03 (0.94–1.13)0.4
  Risks of arthroplasty0.97 (0.92–1.03)0.3
  Severity of arthritis0.99 (0.91–1.07)0.7
  Susceptibility of arthritis to get worse1.04 (0.98–1.10)0.2
Model 3  
 African Americans1.35 (0.92–1.98)0.12
 Arthritis-specific health beliefs  
  Benefits of arthroplasty1.03 (0.94–1.12)0.5
  Risks of arthroplasty0.98 (0.93–1.03)0.4
  Severity of arthritis0.98 (0.90–1.06)0.5
  Susceptibility of arthritis to get worse1.03 (0.97–1.09)0.3
 Clinical appropriateness  
  Appropriate1.96 (1.17–3.28)0.01§
  Uncertain1.57 (1.01–2.46)0.04§


In our sample of 676 primary care patients with moderately severe OA, race was not a predictor of the time to consult to orthopedic surgery. Surprisingly, although African Americans perceived fewer benefits and recognized more barriers to TJA than whites, TJA health beliefs did not influence referral to orthopedic surgery. Moreover, the clinical appropriateness of TJA at study entry was as an independent predictor of referral to surgery. Specifically, subjects who were clinically appropriate to undergo TJA were more likely to be referred to orthopedic surgery than subjects who were considered inappropriate for the procedure.

Contrary to our first hypothesis, we found no difference in the rates of referral to an orthopedic surgeon between African Americans and whites. There are several potential explanations for our null finding. First, the presence or absence of racial differences in health care use depends significantly on the state and the type of conditions being studied. For example, Skinner et al reported significantly lower rates of TJA among African Americans than whites in Atlanta, Georgia, but not in Bronx, New York (8). Although racial disparities exist in invasive cardiac care, others have found no differences in cancer screening procedures and diabetes mellitus care between African Americans and whites (18). Second, the lack of racial difference in referral rates could have been because our study subjects were in equally accessible health care systems. Other authors have suggested that the provision of insurance coverage is a key pathway toward equalizing access to the health care system (35–38). Third, an increased primary care physician density in central Indiana may also explain our findings. Basu and Clancy have reported that an increased primary care physician supply was associated with reductions in racial disparities in referral patterns for certain high-cost surgical procedures for African Americans (39).

Similar to previous studies, African American patients in our sample perceived fewer benefits and greater risks from TJA than white patients. Despite this, TJA-specific health beliefs were not shown to predict referral to orthopedic surgery. We postulate 3 potential reasons. First, the lack of association between health beliefs and referral status could be a reflection of poor communication between primary care physicians and their African American patients. For instance, African Americans may not be adequately expressing their health beliefs regarding TJA during the clinic encounter, or primary care physicians may not be eliciting or listening to the concerns of their African American patients. Street et al have reported that miscommunication between providers and patients regarding concerns about TJA was greater when patients were African American than when patients were white (40). Second, although TJA-specific health beliefs did not influence referral to an orthopedic surgeon, they may very well predict the receipt of TJA. The role of health beliefs in the actual use of TJA deserves further scrutiny. Lastly, the Health Belief Model, which is the theoretical basis of the arthritis-related health belief instrument, has been challenged by other experts in the field (41). Historically, the Health Belief Model is based upon the assumption that direct health concerns are the reasons for change. However, other studies have documented the importance of social or other motivating factors rather than health concerns as driving behavioral change (42). Moreover, the Health Belief Model further assumes a rational basis of human behavior, which may not reflect the human experience in the real world where social, political, and environmental factors can influence behavior (43).

Our study is the first longitudinal study to our knowledge to investigate TJA-specific health beliefs as potential predictors of referral to an orthopedic surgeon. In a population-based cohort from Canada, Hawker et al found that the strongest predictor of the time to first TJA was a person's willingness to consider TJA (44). Although willingness has been linked to one's perceived risks and benefits of TJA (45), the lack of association between TJA-related health beliefs and surgical referral in the current study suggests that other factors that could affect willingness (e.g., knowing someone else who underwent TJA and had a good or bad outcome, or mistrust with the health care provider) may explain why some individuals get surgery and others do not. Further study is needed to better understand the factors that impact willingness.

Furthermore, this is the first study to introduce the concept of clinical appropriateness in relation to the use of TJA. An appropriate procedure is one in which “the expected health benefit exceeds the expected negative consequences by a sufficiently wide margin that the procedure is worth doing, exclusive of cost” (46). In contrast to 2 previous prospective studies that showed symptom severity as predictors of later use of TJA (47, 48), clinical appropriateness considers several clinically-related factors concurrently, thereby capturing the complexities of real-life clinical situations. Clinical appropriateness has been effectively applied in other fields of health care, including coronary revascularization, carotid endarterectomy, and renal transplantation, among other procedures (49–51). In the context of disparity in the use of TJA, it was reassuring to note that clinical appropriateness, and not sociodemographic factors, predicted referral for surgery.

Surprisingly, the only other predictor we found to be significant was recruitment site. Subjects recruited from the Indianapolis VAMC had a longer time to referral to orthopedic surgery than those recruited from the county-affiliated primary care clinics. Based on our personal communication with the local VAMC physicians, there was a 6-month waiting list to see an orthopedic surgeon on multiple occasions during the study period. We speculate that the prolonged waiting time might have resulted in outsourcing TJA to surgeons outside the VAMC health care system. If so, such referral would have not been captured by the VAMC EMR, but only via self-report. Importantly, when we used self-report referral as the outcome measure, VAMC and non-VAMC patients were equally likely to have been referred to orthopedic surgery. Recruitment site was no longer associated with referral status.

Several limitations must be taken into account when interpreting the results of our study. First, the study took place in a narrow geographic distribution, which can decrease the generalizability. Second, although we did not observe any differences in referral rates, racial disparities in the receipt of joint arthroplasty may still exist. Clearly, orthopedic surgery referral does not always result in joint replacement (e.g., for whatever reason, the patient did not show up for his/her orthopedic surgery appointment). However, because referral to an orthopedic surgeon is the principal means by which the use of TJA is made available to patients with OA, we felt it necessary to determine the predictors of referral as the first step toward identifying the reasons behind the racial disparities in TJA use. Third, we have no followup data on the actual receipt of TJA. Future studies should determine whether a patient's health beliefs impact the use of TJA. Fourth, in our study, patients were considered to have been referred if the reason for consult was for joint arthroplasty, or for the treatment of knee or hip OA. For the latter reason, we assumed that evaluation for joint arthroplasty was the intended end point, which may not be entirely accurate. In an attempt to decrease the likelihood of including patients referred for reasons other than arthroplasty, we did not consider those referred for other indications. Finally, we have no data on dual users (i.e., someone who uses health services within and outside of the VAMC health care system), which could be a potential source of bias. The bias resulting from the lack of data on dual users was likely minimal, considering that our findings were similar whether we used referral data from the medical record or self-report from the participants.

In summary, in this cohort of primary care clinic patients with at least moderately severe knee or hip OA, African Americans and whites were equally likely to be referred to an orthopedic surgeon for consideration of TJA. Health beliefs specific to the risks and benefits of TJA did not influence referral status. Importantly, clinical appropriateness was an independent predictor of referral. In the context of racial disparity in TJA use, future explanatory studies should prospectively follow patients with OA from the time that referrals are made to orthopedic surgeons. In particular, determining the reasons why patients would accept or decline TJA offered by the orthopedic surgeon, or why patients would or would not show up in the orthopedic clinic, could potentially unravel the underlying mechanisms of the reported disparity in TJA use. Moreover, future studies should also examine the roles of the patient's trust, physician–patient connectedness, and the patient's knowledge of someone who had undergone surgery as potential predictors of referral to or use of TJA.


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. Ang 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. Ang.

Acquisition of data. Ang.

Analysis and interpretation of data. Ang, James, Stump.



Perceived benefits of arthroplasty:

  • 1Joint replacement surgery would get rid of my joint pain.
  • 2Joint replacement surgery would allow me to do things that I like to do.
  • 3Joint replacement surgery will make me feel better.

Perceived barriers of arthroplasty:

  • 1The joint replacement surgery I am aware of is too risky and too time consuming.
  • 2I would have to change too many daily activities to undergo joint replacement surgery for my arthritis.
  • 3It is too inconvenient for me to get joint replacement surgery.
  • 4I would experience too much pain several weeks after joint replacement surgery.
  • 5To get joint replacement surgery would cause too much inconvenience to my immediate family members.

Perceived severity of arthritis:

  • 1My arthritis keeps me from doing things I want to do.
  • 2My arthritis limits my daily activities.
  • 3My arthritis interferes with my going to work or school.

Perceived susceptibility of arthritis to get worse:

  • 1Due to the condition of my physical health, my arthritis is likely to get worse.
  • 2My chances that my arthritis will get worse are great.
  • 3Within the next year my arthritis will get worse.
  • 4I worry a lot about my arthritis getting worse.
  • 5It worries me to think about the effect my arthritis will have on my health.