Underrepresentation of Non-White Participants in the American Academy of Orthopaedic Surgeons Guidelines for Surgical Management of Knee Osteoarthritis

Abstract

Background:

The aim of this study was to examine the racial and ethnic representation in studies included in the 2015 American Academy of Orthopaedic Surgeons Surgical Management of the Knee Evidence-Based Clinical Practice Guideline relative to their representation of the United States (US).

Methods:

The demographic characteristics reported in articles included in the 2015 American Academy of Orthopaedic Surgeons Surgical Management of the Knee Evidence-Based Clinical Practice Guideline were analyzed. The primary outcome of interest was the representation quotient, which is the ratio of the proportion of a racial/ethnic group in the guideline studies relative to their proportion in the US. There were 211 studies included, of which 15 (7%) reported race. There were 35 studies based in the US and 7 of the US-based studies reported race.

Results:

No US-based studies reported race and ethnicity separately, no studies reported American Indian/Alaska Native participants and no US-based studies reported Asian participants. The representation quotient of US-based studies was 0.66 for Black participants, 0.33 for Hispanic participants, and 1.30 for White participants, which indicates a relative over-representation of White participants compared to national proportions.

Conclusion:

This study illustrated that the evidence base for the surgical management of knee osteoarthritis has been constructed from studies which fail to consider race and ethnicity. Of those US-based studies which do report race or ethnicity, study cohorts do not reflect the US population. These results illustrate a disparity in clinical orthopedic surgical evidence and highlight the need for improved research recruitment strategies.

Among adults over the age of 65 years, osteoarthritis (OA) of the knee is one of the leading causes of disability worldwide and has an estimated prevalence of 33.6% in the United States [1]. Surgical management of knee OA, including total knee arthroplasty, is increasingly common, both in the United States and worldwide [2].

The American Academy of Orthopaedic Surgeons (AAOS) published the Surgical Management of OA of the Knee Evidence-Based Clinical Practice Guidelines in 2015 [3]. This guideline was formulated using 224 full text articles which provides a summary of recommendations regarding best practices in surgical management of knee OA, with accompanying descriptions of the strength and quality of the overall evidence base for each recommendation [3]. Multiple organizations, including the American Association of Hip and Knee Surgeons, the American Society of Anesthesiologists, and the American College of Radiology have endorsed this guideline [3]. As such, this document forms the foundation for how knee OA is understood and managed in the United States. It is therefore important to understand how these recommendations and their evidence bases are generalizable to the patient population in the United States.

The generalizability of data is predicated on the inclusion of trial participants which reflect the larger study population. Additionally, the increasingly evident orthopedic health disparities in the United States require an understanding of how different health conditions affect racial and ethnic minorities [4–6]. In the United States and worldwide, equitable clinical trial enrollment methods are an essential step to both improving the quality of data available for clinical decision making and for addressing the gaps in health care for underserved groups [6]. To our knowledge, there have been no studies which have examined the representation of race and ethnicity within the published evidence base for surgical management of knee OA. The aim of this study was to compare the racial and ethnic representation of the 2015 AAOS Surgical Management of the Knee Evidence-Based Clinical Practice Guidelines to the larger racial and ethnic breakdown of the United States. We hypothesized that, similarly to other areas of medicine, racial and ethnic minorities would be underrepresented in the evidence base relative to their national distribution in the United States.

Methods

We obtained the full-text journal articles for each nonduplicate cited study included in the final recommendations of the 2015 Surgical Management of the Knee Evidence-Based Clinical Practice Guidelines. We excluded any citation that was included more than once in the recommendations or those for which the full text article was not available. We also excluded systematic and nonsystematic reviews or retracted studies. In those situations, in which more than 1 study used the same study population, we excluded all but the most recent study to avoid duplicating participants.

We recorded the study characteristics, demographic characteristics, race reporting, and race/ethnicity of all participants for each included study. We defined race reporting as the listing of participants’ race or ethnicity in the body, tables, or figures of a paper. When categorizing papers that reported race, a subcategory of papers referenced race in their methods without reporting data and this was treated as a separate category of race reporting. In analyzing those studies that referenced race with data, we utilized the Office of Management and Budget (OMB) standard categories to define races, which included Black, White, Asian, and American Indian/Native Alaskan [7]. An additional category of “other” was created to account for any studies which added a category of “other” to their demographic data. Although Hispanic/Latino/a/x is an ethnicity and not a race, for the purpose of this study the decision was made to include Hispanic/Latino/a/x as a race category in order to reflect the categorization methods of the included studies. However, the authors explicitly acknowledge that race and ethnicity are not interchangeable concepts and there were varying definitions for these concepts. One definition of race is that it refers to social groups which identify by external characteristics, such as skin color, while ethnicity refers to social groups which identify by shared cultural characteristics, such as language, religion, and customs. It is possible for individuals to belong to multiple racial and ethnic groups simultaneously [8]. Furthermore the term Latino/a/x will be used in this paper, however the authors acknowledge that at the time of writing, there is no consensus over gender-inclusive terminology when referring to Spanish-speaking or Latin American ethnicities in the United States, and that different demographic groups may prefer to self-identify with a different term [9]. To calculate the overall representation of each racial and ethnic group relative to the US population, we utilized the US population estimates from 2005 to 2015 from the US Census Bureau’s American Community Survey in 5-year increments. This increment was selected in order to account for variability in census reporting over time; the average for the decade was calculated from this 5-year increment data.

Our primary outcome of interest was the representation quotient (RQ) for each category of race or ethnicity in US-based studies in 5-year increments. For the purposes of this study, the RQ was defined as the ratio of the proportion of a racial/ethnic group in the AAOS clinical practice guideline relative to the estimated proportion of that racial/ethnic group nationally, as reported in the US Census Bureau’s American Community Survey [10,11]. Previous studies have utilized RQ’s to represent the quotient of a particular race category in a population of interest versus the demographic representation in the national US population [11,12]. An RQ less than 1 means that a group is underrepresented in the guideline data relative to their distribution in the United States, and an RQ large than 1 indicates that the group is overrepresented [10–12]. More granularly, the RQ can be further interpreted to assess the magnitude of mismatch; an RQ of 0.25 indicates that a group is underrepresented in the population of interest by 75% and vice versa [11]. Importantly, the RQs were calculated using only studies that recruited participants in the United States, since both the OMB race/ethnicity categories and the US Census Bureau race category data are based on the population of the United States [7]. We censored any study that did not report race or did not recruit primarily from the United States prior to calculating the RQ for any given group. We calculated this RQ for each 5-year increment using the reported years of data collection as well as overall. When data collection time periods were not available, we used the publication year as a substitute. Secondary outcomes for this study included the proportion of studies which reported race, the proportion of studies that referenced adjusting for race without providing data, and the proportion of participants from each group that were enrolled in the studies after censoring those studies which did not report participant race or ethnicity.

We calculated the study characteristics and racial/ethnic group characteristics using descriptive statistics. We performed group comparisons using the Pearson χ² test. All tests were 2-sided with α set at 0.05. This study was exempt from institutional review board approval because it exclusively utilized published studies and publicly available data. Informed consent was also waived as a result. Where applicable, the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines were followed in this study [13].We reviewed 224 studies, of which 211 met inclusion criteria (Figure 1). The publication years spanned 1985 to 2015, with the majority (54%) published between 2010 and 2014. Most studies were randomized controlled trials (140 of 211, 66%) and funded from academic only sources. The mean number of study participants was 12,888 (Standard deviation [SD] 49,616) with a range of 20 to 528, 495 participants. Of the 211 articles included, the majority were based outside of the United States (80%) and 35 were based within the United States (17%). There were 7 US-based studies and 4 international studies that reported race, and an additional 4 international studies referenced race in the discussion or statistical analysis without reporting data, with a total of 15 (7%) papers reporting race. All studies reporting race were published after 2005. Of all the studies reporting race regardless of study location, 7 studied modes of anesthesia and analgesia in knee OA surgery. No US-based studies reported race and ethnicity separately; 1 international study reported ethnicity only, but not race and 1 international study reported nationality, but not race or ethnicity (data not shown). The remainder of study characteristics are reported in Table 1.

Fig. 1.

Study design. Of the 224 studies included in the guidelines, 7 duplicate citations were identified and 6 studies were excluded. Of the included studies, 7 US-based studies reported race/ethnicity data and 8 international studies either reported race/ethnicity data or referenced race/ethnicity within their paper.

Table 1.

Characteristics of Studies Included From the 2015 American Academy of Orthopaedic Surgeons Surgical Management of the Knee Evidence-Based Clinical Practice Guideline.

Variables	All Studies	US-Based Studies Reporting Race	International Studies Reporting Race	Studies Referencing Race Without Reporting Data
Study characteristics	n = 211	n = 7	n = 4	n = 4
Publication year, n (%)
1985 to 1989	1 (0%)	0	0	1 (25%)
1990 to 1994	8 (4%)	0	0	0
1995 to 1999	13 (6%)	0	0	0
2000 to 2004	32 (15%)	0	0	0
2005 to 2009	40 (19%)	1 (14%)	1 (25%)	0
2010 to 2014	113 (54%)	6 (86%)	2 (50%)	3 (75%)
2015 +	4 (2%)	0	1 (25%)	0
Study type, n (%)
Randomized controlled trial	140 (66%)	2 (29%)	3 (75%)	2 (50%)
Prospective cohort	50 (24%)	0	0	1 (25%)
Retrospective cohort	16 (8%)	5 (74%)	1 (25%)	1 (25%)
Case-control	4 (2%)	0	0	0
Cross-sectional	1 (0%)	0	0	0
Geographic location, n (%)
US only	35 (17%)	7 (100%)	–	0
US primary with other countries	1 (0%)	–	–	2 (50%)
Other country/countries primary with US	2 (1%)	–	–	0
No US involvement	169 (80%)	–	4 (100%)	2 (50%)
Not reported	5 (2%)	–	–	0
Funding, n (%)
Academic only	102 (48%)	6 (86%)	2 (50%)	2 (50%)
Industry only	11 (5%)	0	0	0
Both	17 (8%)	0	0	0
Not reported	81 (38%)	1 (14%)	2 (50%)	2 (50%)
Number of study participants
Mean (SD)	12,888 (49,616)	86,313	137	117,032
Median (range)	90 (20 to 528,495)	15,687 (31 to 528,495)	72 (47 to 357)	164 (21 to 467,779)
Guideline section
Risk stratification	26 (12%)	0	0	1 (25%)
Tourniquet use	5 (4%)	0	0	0
Anesthesia and nerve blockade	48 (22%)	5 (71%)	2 (50%)	0
Patellar resurfacing	16 (8%)	0	0	0
Cemented versus cementless components versus hybrid fixation	15 (8%)	0	0	0
Timing TKA	2 (0%)	0	0	0
UKA	8 (4%)	1 (14%)	0	1 (25%)
Early-stage versus late-stage supervised exercise and physical therapy	24 (11%)	1 (14%)	0	0
Postop mobilization	3 (1%)	0	0	0
Patient-specific instrumentation	7 (3%)	0	0	0
Surgical navigation	15 (7%)	0	1 (25%)	0
Tranexamic acid	9 (4%)	0	0	0
Antibiotic bone cement	3 (1%)	0	0	0
Cruciate retaining arthroplasty	7 (3%)	0	0	0
Polyethylene tibial component	4 (2%)	0	0	0
Drains	7 (3%)	0	0	0
Cryotherapy devices	4 (2%)	0	0	1 (25%)
Continuous passive motion	8 (4%)	0	0	1 (25%)
Race reported, n (%)
Yes	15 (7%)	7 (100%)	4 (100%)	4
No	196 (93%)	–	–	0
Gender reported, n (%)
Yes	198 (94%)	7 (100%)	4 (100%)	3 (75%)
No	13 (6%)	0	0	1 (25%)
Participant Characteristics	n = 1,121,269	n = 604,195	n = 548	n = 117,032
Age, y
Mean of meansa	68.0	66.0	69.9	67.3
Mean of SDsa	8.39	10.25	–	7.00
Female, n (%)b	[502,962/856,128] (58%)	372,099 (62%)	400 (73%)	[141/99] 70%
Race, n (%)c
White	–	452,573 (75%)	[7/357] (2%)	–
Black	–	41,037 (7%)	–	–
Asian	–	–	[446/458] (97%)	–
Other	–	[19,090/604,122] (3%)	[5/411] (1%)	–
Hispanic ethnicity	–	[1,155/16,586] (7%)	–	–

TKA, total knee arthroplasty; UKA, unicompartmental knee arthoplasty.

^aAmong studies reporting variable.

^bAmong studies reporting gender and including female participants.

^cAmong studies reporting race and including participants of that race.

Results

Overall, the 7 US-based studies reporting race data enrolled a total of 604,195 participants of whom 75% of participants were White, 7% were Black, 7% were Hispanic/Latino/a/x, and 3% were reported as “other” (Table 1, Table 2). Of the studies which included an “other” category, 1 study defined this category as Native Hawaiian, Asian, Pacific Islander, or American Indian/Alaska Native, 1 defined it as any category other than White or Black, and the remainder did not define this category (Table 2). No studies reported American Indian/Alaska Native participants as a separate category and no US-based studies reported Asian participants (Table 2). The overall RQ of US-based studies was 0.66 for Black participants, 0.33 for Hispanic/Latino/a/x participants, 0 for American Indian/Alaska Native and Asian participants and 1.30 for White participants (Figure 2). These RQs indicate that White participants were over-represented compared to their proportion of the US population by 30% (Figure 2). All other races were underrepresented in the studies relative to their US population proportions (Figure 2). Figure 2 also illustrates the RQ for all groups over time. The RQ for Hispanic/Latinx groups decreased from 2005 to 2010, while that of Black and White participants increased during this time interval (Figure 2). Of note, the study reporting Hispanic/Latino/a/x participants from 2010 to 2014 was smaller in size than the study included in the 2005 time period (Table 2). No studies reporting Hispanic/Latinx participants clarified how this category was defined and no studies distinguished between race and ethnicity. Race reporting was associated with retrospective study design, academic funding sources, and US-based recruitment (P < .05).

Table 2.

Characteristics of US-Based Studies Reporting Race Data.

Citation	Publication Year	Years of Data Collection (5-y Increments)	Study Design	Guideline Section	Retrospective versus Prospective	Level of Data	Funding Source	Number of Participants	Race Reporting Method
Duchman et al [14]	2014	2005 to 2014	Cohort study	Unicompartmental knee arthroplasty: revisions	Retrospective	3	Academic	29,333	Non-OMB categories
Liu et al [15]	2013	2005 to 2009	Cohort study	Neuraxial anesthesia	Retrospective	Not reported	Academic	16,555	Non-OMB categories
Memtsoudis et al [16]	2013	2005 to 2009	Cohort study	Neuraxial anesthesia	Retrospective	Not reported	Academic	528,495	Non-OMB categories
Pugely et al [17]	2013	2005 to 2009	Cohort study	Neuraxial anesthesia	Retrospective	3	Academic	14,052	Non-OMB categories
Stundner et al [18]	2012	2005 to 2009	Cohort study	Neuraxial anesthesia	Retrospective	Not reported	Academic	15,687	Non-OMB categories
Brown et al [19]	2014	Not reported	Randomized control trial	Preoperative physical therapy	Prospective	Not reported	Academic	31	Non-OMB categories
Good et al [20]	2007	Not reported	Randomized control trial	Peripheral nerve blockade	Prospective	Not reported	Unknown	42	Non-OMB categories

Black Participants (#)	White Participants (#)	Asian Participants (#)	Native American or Alaska Native Participants (#)	Hispanic/Latinx Participants (#)	Included Category of “Other”?	Description of “Other” Category	“Other” Participants (#)
1,900	23,298	Not reported	Not reported	NR	Yes	“other than Black or white”	4,133
1,186	13,002	Not reported	Not reported	1,154	Yes	“Includes patients listed as Native Hawaiian or Pacific Islander, Asian or Pacific Islander, Asian, and American Indian or Alaska Native in NSQIP database”	355
36,162	393,148	Not reported	Not reported	NR	Yes	No description provided	11,933
950	11,151	Not reported	Not reported	NR	Yes	No description provided	1,951
833	11,908	Not reported	Not reported	NR	Yes	No description provided	718
3	27	Not reported	Not reported	1	No	NA	NA
3	39	Not reported	Not reported	NR	No	NA	NA

Fig. 2.

Representation quotient (RQ) of race/ethnicity reported in US-based guideline studies relative to the US population. White participants had an RQ over 1 overall and in each date range studied. The RQ for White participants increased over time. All other race/ethnic groups had RQ’s less than 1 and Asian and American Indian Alaska Native (AIAN) participants had RQs of 0. Overall, only Hispanic/Latinx participant RQ decreased over time.

Discussion

Our study illustrates that, in those 7 United States papers where race was reported, all racial and ethnic groups other than White are underrepresented relative to their US proportions in the evidence base for the 2015 Surgical Management of OA of the Knee Evidence-Based Clinical Practice Guidelines. In particular, Asian and American Indian/Native Alaskan participants are absent from the US evidence base entirely for the past several decades. Furthermore, race overall is underreported in the evidence base, and there are no distinctions made between race and ethnicity in any US-based studies supporting the AAOS clinical practice guidelines. These findings threaten the external validity and generalizability of the AAOS guidelines and highlight a disparity in the research enrollment strategies that have been utilized to form this evidence base for the past several decades.

The degree of under-representation of Black, Asian, Native American, and Hispanic/Latino/a/x populations in clinical research has become increasingly recognized throughout the field of medicine, including orthopedics. Similar results have been found in systematic reviews of orthopedic research, at times reporting underrepresentation of Black participants by 3.5-fold and Hispanic participants by 2-fold relative to US Census data [21]. Our study additionally highlights that the distinction between race and ethnicity is absent within AAOS evidence, which is consistent with recent studies of published trials funded by the National Institute of Health [11,22]. This is of particular importance when discussing the health of Hispanic and Latino/a/x populations, in which the distinction between ethnicity and race is an essential step to understanding how social determinants of health act within this large and increasingly diverse population in the United States [23].

Strategies to improve the design and conduct of future clinical studies to assure representative participation can and should address several facets of the research process which have been implicated as potential causes for underrepresentation. Study enrollment conducted in partnership with community-based organizations with established relationships with diverse communities may increase participation [24]. Although the details of different community participation strategies are out of the scope of this paper, these methods allow researchers to begin addressing the breach of trust brought on by ongoing and historical structural racism [24–31]. There are a wealth of publications describing community-based health promotion and research development strategies and implementation plans for those interested in incorporating these strategies [24–31]. It should be noted that research that oversamples from historically underrepresented populations is needed to inform evidence-based care in order to account for historical gaps [31]. Methodology should include power analyses that include race/ethnicity in sample size calculations to assure outcomes can be defined for participants of diverse race and ethnicity [23]. Also, diversification of the work force is associated with increased studies focusing on the same population that authors are members of, suggesting that efforts to expand orthopedic pipeline programs for individuals underrepresented in medicine could also affect research enrollment diversity in the future [23]. The AAOS and other knee OA care guidelines will be limited until broader research exists that include representative populations.

Underlying the lack of representation in knee OA guidelines, this study contributes to the growing body of research which suggests that the reporting of race and ethnicity in surgical specialty research, including orthopedics, is deficient. A recent review of articles published in high impact surgical journals in 2019 found that the frequency and quality of race reporting in surgical research was low [32]. Systematic reviews of orthopedic publications from 2000 to 2020 have shown that orthopedic surgery in particular underreports race and ethnicity compared to other areas of medicine [6]. Our study further adds to the body of evidence by characterizing studies that are most likely to report race and ethnicity – namely, those with retrospective study design, academic funding, and location in the United States. Additionally, our study suggests that the research topic influences the likelihood of race reporting – of the 15 studies reporting race, 7 (47%) were focused on anesthesia and analgesia. Although our study is not designed to evaluate the reasons for race reporting across research topics, there are well-documented inequities in pain control within anesthesia and analgesia literature, suggesting that race reporting is of particular importance within this area of interest [33]. We suggest that race and ethnicity reporting should become a requirement of high quality research, much like sex assigned at birth or age is currently considered a standard measure. While there are several potential approaches, making race and ethnicity data a requirement to publish in a high impact journal and providing exemplary data collection methods would likely be effective in encouraging improved reporting. For consistency, OMB categories could be the minimum reporting standard and avoid collapsing categories such as “other.” Whenever possible in prospective studies, self-identification with multiple categories should be the ideal. The accurate reporting and equitable representation of racial and ethnic groups within clinical research is an essential step in understanding the health disparities that continue to be revealed in orthopedic surgery. Particularly within the management of knee OA, evidence suggests that Black and/or Hispanic/Latinx patients are less likely to receive an orthopedic referral, are more likely to receive surgical interventions later in the disease course, receive less aggressive pain management, and have higher risks for some outcomes, including aseptic revisions and readmissions [4, 34–36]. Furthermore, evidence suggests that these disparities are becoming more pronounced over time across the field of medicine and within total joint arthroplasty, specifically [37,38]. The underrepresentation of patients of minority race and ethnicity receiving care in elective orthopedic clinical practice adds another barrier to assuring that research to inform best practices will include diverse participants. These studies all highlight the need for increased recruitment of and characterization of minorities in orthopedics research so as to further elucidate the social and systematic mechanisms for these outcomes [37]. Additionally, equitable recruitment strategies which reflect the diversity of the knee OA population are an investment in the health of all people, an important aspect in the health care system gaining trust among communities affected by structural racism [11,31].

There are several potential limitations to this study. Because the degree of race reporting was low among our included studies, it is likely that our RQ’s underestimate the true degree of underrepresentation in AAOS guideline studies. The use of collapsing categories such as “other” in the data collection process also limits our ability to draw conclusions regarding the absence of certain racial and ethnic groups – such as Asian or Native American participants – from the data set. Because the “other” category lacks a true denominator, no true comparisons can be drawn to census data – it is possible that certain groups are identified within the “other” subset, but we cannot hypothesize on the demographic characteristics of this group with the data available. Additionally, this study focused specifically on the publications included in forming the 2015 AAOS knee OA guidelines, and therefore the results cannot be generalized to all studies of knee OA. This study is not designed to examine why race reporting was or was not included in any given study, and as such we cannot comment on the motivations for excluding race and ethnicity data.

Conclusion

Knee OA is a chronic, painful condition which affects the quality of life for a large proportion of the US population. The studies included in the guidelines for surgical management of this problem do not reflect the diversity and complexity of the nation. Therefore, improved research recruitment strategies are necessary to assure representative populations inform evidence-based care guidelines for all patients to benefit equitably from orthopedic care of knee OA in the United States.