Telemedicine Use in Orthopaedic Surgery Varies by Race, Ethnicity, Primary Language, and Insurance Status

Abstract

Background

Healthcare disparities are well documented across multiple subspecialties in orthopaedics. The widespread implementation of telemedicine risks worsening these disparities if not carefully executed, despite original assumptions that telemedicine improves overall access to care. Telemedicine also poses unique challenges such as potential language or technological barriers that may alter previously described patterns in orthopaedic disparities.

Questions/purposes

Are the proportions of patients who use telemedicine across orthopaedic services different among (1) racial and ethnic minorities, (2) non-English speakers, and (3) patients insured through Medicaid during a 10-week period after the implementation of telemedicine in our healthcare system compared with in-person visits during a similar time period in 2019?

Methods

This was a retrospective comparative study using electronic medical record data to compare new patients establishing orthopaedic care via outpatient telemedicine at two academic urban medical centers between March 2020 and May 2020 with new orthopaedic patients during the same 10-week period in 2019. A total of 11,056 patients were included for analysis, with 1760 in the virtual group and 9296 in the control group. Unadjusted analyses demonstrated patients in the virtual group were younger (median age 57 years versus 59 years; p < 0.001), but there were no differences with regard to gender (56% female versus 56% female; p = 0.66). We used self-reported race or ethnicity as our primary independent variable, with primary language and insurance status considered secondarily. Unadjusted and multivariable adjusted analyses were performed for our primary and secondary predictors using logistic regression. We also assessed interactions between race or ethnicity, primary language, and insurance type.

Results

After adjusting for age, gender, subspecialty, insurance, and median household income, we found that patients who were Hispanic (odds ratio 0.59 [95% confidence interval 0.39 to 0.91]; p = 0.02) or Asian were less likely (OR 0.73 [95% CI 0.53 to 0.99]; p = 0.04) to be seen through telemedicine than were patients who were white. After controlling for confounding variables, we also found that speakers of languages other than English or Spanish were less likely to have a telemedicine visit than were people whose primary language was English (OR 0.34 [95% CI 0.18 to 0.65]; p = 0.001), and that patients insured through Medicaid were less likely to be seen via telemedicine than were patients who were privately insured (OR 0.83 [95% CI 0.69 to 0.98]; p = 0.03).

Conclusion

Despite initial promises that telemedicine would help to bridge gaps in healthcare, our results demonstrate disparities in orthopaedic telemedicine use based on race or ethnicity, language, and insurance type. The telemedicine group was slightly younger, which we do not believe undermines the findings. As healthcare moves toward increased telemedicine use, we suggest several approaches to ensure that patients of certain racial, ethnic, or language groups do not experience disparate barriers to care. These might include individual patient- or provider-level approaches like expanded telemedicine schedules to accommodate weekends and evenings, institutional investment in culturally conscious outreach materials such as advertisements on community transport systems, or government-level provisions such as reimbursement for telephone-only encounters.

Level of Evidence

Level III, therapeutic study.

Introduction

Telemedicine has evolved across multiple specialties, including orthopaedics, as a practical means to improve access to care. It offers some demonstrable advantages including ease of use, lower cost [2, 16], and increased patient satisfaction [9] compared with in-person medical office visits. The coronavirus-19 (COVID-19) pandemic greatly accelerated adoption, lowered barriers to reimbursement, and reduced medicolegal concerns [8].

But the rapid growth of telemedicine outpaced the healthcare system’s ability to implement this technology. Although the potential to mitigate barriers to care are evident, there are several recognized challenges to the equitable application of telemedicine [7, 15]. Although telemedicine platforms circumvent the need for geographic proximity to medical centers or transportation resources, they create novel technological requirements that may affect access to and use by populations such as those with limited English-language proficiency, those who lack technological resources such as internet capabilities and access to virtual communication devices, or patients with limited access to expedited transportation options, especially in the context of orthopaedic injury or immobilization. Healthcare disparities have been shown to exist in orthopaedics across the domains of access [18], use [5], and outcomes [21, 27]. There is a concern that the broad implementation of telemedicine, without ensuring the equitable distribution of resources necessary for its use, could worsen many of these already existing disparities as well as exacerbate healthcare segregation [20]. Currently, disparities in telemedicine as a whole remain an understudied topic, and we are unaware of prior work in this arena specific to orthopaedic surgery.

In this context, we sought to evaluate changes in patients’ use of telemedicine for elective orthopaedic care based on race or ethnicity, primary language, and insurance status.

We asked whether the proportions of telemedicine use across orthopaedic services would be different for (1) racial and ethnic minorities, (2) non-English speakers, and (3) patients insured through Medicaid during a 10-week period after the implementation of telemedicine in our healthcare system compared with in-person visits during a similar time period in 2019.

Patients and Methods

Study Design

We conducted a retrospective study comparing all telemedicine encounters for elective orthopaedic care conducted between March 24, 2020 and May 18, 2020 with in-person encounters that occurred over the same time period in 2019. From March to May 2020, telemedicine services were used exclusively in our healthcare system to provide elective orthopaedic care because of statewide (Massachusetts) regulations in response to the COVID-19 pandemic.

The clinical data for this work were accessed using the research patient data registry of Mass General Brigham, an integrated healthcare network that comprises several hospitals. Data for this study were extracted from encounters performed by physicians primarily affiliated with Massachusetts General Hospital or Brigham and Women’s Hospital, both academic referral centers located in Boston, MA, USA. This data repository automatically captures sociodemographic and clinical details for all patients presenting for care in the healthcare system and has been used to study aspects of orthopaedic healthcare policy [6, 24].

Inclusion and Exclusion Criteria

We included patients aged 16 years and older who had a new outpatient orthopaedic encounter via telemedicine between March 24, 2020 and May 18, 2020 at Massachusetts General Hospital or Brigham and Women’s Hospital (both in Boston, MA, USA). These dates were chosen because of their correspondence with the abrupt transition to telemedicine-only platforms secondary to local government ordinances. Comparison patients who received in-person outpatient care between March 24, 2019 and May 18, 2019 were used as a control group. Orthopaedic outpatient visits were queried by Current Procedural Terminology code for all orthopaedic surgeons at both hospitals. In-person visits were defined as any visit for which the patient and surgeon met in person in an outpatient clinic. Telemedicine visits were conducted virtually, either by video or telephone. The use of telemedicine was confirmed by reviewing the documentation of the clinical encounter or a review of clinician schedules in the electronic medical record. Interpreter services, if used, were provided with the same service for both groups with no change in languages available. To focus only on patients who established care with a new orthopaedic surgeon during the study period, patients were excluded if they had an outpatient visit with the same orthopaedic surgeon within the previous 3 years or if the encounter was conducted solely by an advanced practice provider or a resident physician. Follow-up visits were excluded to avoid confounding from patients who had established care via an in-person visit. To ensure that both populations consisted of patients with procedures that would be considered elective, we excluded all patients seen in the orthopaedic trauma and oncology departments as well as those who received an in-person evaluation during the COVID-19 pandemic.

Patients

We identified 11,056 encounters for inclusion in this investigation. These were derived from 11,483 total encounters performed during the study period, of which 324 were excluded in the 2020 group for not being telemedicine visits and 103 patients excluded for being younger than 16 years (92 in-person, 11 telemedicine). Of the final included patients, 84% (9296 of 11,056) were in-person visits conducted in 2019 and 16% (1760 of 11,056) were telemedicine visits performed in 2020. On unadjusted analysis, the telemedicine group was younger, with a median (range) age of 57 years (16 to 97) versus 59 years (16 to 102) (p < 0.001). These age differences are slight, and we believe they do not undermine the study findings. There were no differences in gender (telemedicine group 56% [979 of 1760] women, in-person 56% [5223 of 9296]; p = 0.66). The proportion of white patients was higher in the telemedicine group (83% [1460 of 1760] versus 81% [7511 of 9296] p = 0.03), and the proportion of Hispanic patients was lower in the telemedicine group (1% [26 of 1760] versus 2% [231 of 9296]; p = 0.01) than in the in-person group. The proportion of English speakers was higher in the telemedicine group (97% [1702 of 1760] versus 94% [8762 of 9296]; p < 0.001) and the proportion of participants who spoke other languages was lower (1% [10 of 1760] versus 2% [180 of 9296]; p < 0.001). In terms of insurance, the proportion of privately insured patients was higher in the telemedicine group (65% [1147 of 1760] versus 59% [5517 of 9296]; p < 0.001), and the proportion of Medicare patients was lower in the telemedicine group (21% [372 of 1760] versus 25% [2309 of 9296]; p = 0.001). Other insurance categories were not different. Finally, the proportion of subspecialty hand patients was higher in the telemedicine group (22% [381 of 1760] versus 19% [1781 of 9296]; p = 0.02) as well as subspecialty spine patients (19% [341 of 1760] versus 14% [1273 of 9296]; p < 0.001), whereas the proportion of sports patients was lower (22% [386 of 1760] versus 27% [2478 of 9296]; p < 0.001), with no differences in other specialties (Table 1).

Table 1.

Patient characteristics by visit type

Parameter	In-person visit (n = 9296)	Telemedicine visit (n = 1760)	p value
Age in yearsa	59 (16-102)	57 (16-97)	< 0.001
Household income in USDa	95,447 (17,000-250,000)	95,368 (30,551-235,714)	0.04
Women (n = 6202), % (n)	56 (5223)	56 (979)	0.66
Race or ethnicity, % (n)
White (n = 8971)	81 (7511)	83 (1460)	0.03
Black (n = 652)	6 (545)	6 (107)	0.72
Hispanic (n = 257)	2 (231)	1 (26)	0.01
Asian (n = 366)	3 (316)	3 (50)	0.23
Other (n = 456)	4 (389)	4 (67)	0.47
Unknown (n = 354)	3 (304)	3 (50)	0.35
Language, % (n)
English (n =10,464)	94 (8762)	97 (1702)	< 0.001
Spanish (n = 231)	2 (204)	2 (27)	0.08
Other (n = 190)	2 (180)	1 (10)	< 0.001
Unknown (n = 171)	2 (150)	1 (21)	0.19
Insurance type, % (n)
Private (n = 6664)	59 (5517)	65 (1147)	< 0.001
Medicare (n = 2681)	25 (2309)	21 (372)	0.001
Medicaid (n = 1358)	12 (1157)	11 (201)	0.23
Workers compensation (n = 145)	1 (121)	1 (24)	0.83
Veterans (n = 47)	0.5 (42)	0.3 (5)	0.32
Self-pay (n = 58)	0.6 (54)	0.2 (4)	0.15
Other (n = 31)	0.3 (29)	0.1 (2)	0.15
Missing (n = 72)	0.7 (67)	0.3 (5)	0.04
Subspecialty, % (n)
Arthroplasty (n = 2169)	20 (1838)	19 (331)	0.35
Foot and ankle (n = 1752)	16 (1496)	15 (256)	0.10
Hand (n = 2162)	19 (1781)	22 (381)	0.02
Shoulder and elbow (n = 495)	5 (430)	4 (65)	0.08
Spine (n = 1614)	14 (1273)	19 (341)	< 0.001
Sports (n = 2864)	27 (2478)	22 (386)	< 0.001

Missing data are presented where applicable.

^aData are presented as median (range); Ref = referent group.

Study Variables

We abstracted data from full records across all encounters for eligible patients and obtained sociodemographic and clinical characteristics including age, self-reported race or ethnicity, self-reported gender, self-reported primary language, insurance status, ZIP code of residence, and orthopaedic subspecialty (arthroplasty, foot and ankle, hand, spine, shoulder and elbow, and sports surgery) in which the visit occurred.

We categorized race using five mutually exclusive categories: non-Hispanic white, Black, Hispanic, Asian, and other (including Native American, Pacific Islander, mixed race, and other race). The race categories were modeled off of primary designated race options as denoted by the U.S. Census Bureau. The data breakdown by majority of self-identified race responses included white, Black or African American, Hispanic, and Asian; these also reflected the most commonly reported racial and ethnic groups on the U.S. Census and were therefore used as the primary groupings in this study. The primary language was classified as English, Spanish, or other (that is, a non-English-language speaker with a designated primary language other than Spanish). Insurance was stratified as Medicaid, Medicare, private insurance, workers compensation, veterans, or self-pay or other, with Medicaid and self-pay or other considered underinsured populations. Patients’ listed ZIP code of residence was linked to publicly available geospatial annual median household income from the U.S. Census based on mean 5-year data from 2013 to 2018 [26]. This geographical approach is an established method of approximating socioeconomic status [4].

Ethical Approval

Ethical approval for this study was obtained from Mass General Brigham, Boston, MA, USA (approval number 2020P002564).

Statistical Analysis

In this study, changes in the proportion of patients accessing orthopaedic care via telemedicine in 2020 compared with those receiving in-person visits in 2019 were considered the primary outcome. We used race or ethnicity as our primary independent variable, with primary language and insurance status considered secondarily. In evaluations of race, the category of white was used as the referent, with English and private insurance used as the referents for analyses involving primary language and insurance status, respectively. All other variables we abstracted were used as covariates in our statistical tests.

Bivariate unadjusted analyses were conducted using chi-square testing for categorical variables and the Wilcoxon rank-sum tests for age and median household income by ZIP code. Multivariable-adjusted analyses were performed for our primary and secondary predictors using a logistic regression analysis. We assessed interaction between race or ethnicity, primary language, and insurance type. If interactions were detected, we planned a priori to run separate analyses for predictors while excluding the interaction term. Goodness-of-fit was assessed for all final models using the Hosmer-Lemeshow test. Results are presented using odds ratios (ORs) and 95% confidence intervals, with point estimates and 95% CIs exclusive of 1.0. p values < 0.05 were defined a priori as statistically significant.

Interactions were detected between race or ethnicity and primary language. As a result, we conducted separate multivariable analyses for these variables. Although no interactions were detected between race or ethnicity and insurance, or between the primary language and insurance, these combinations resulted in numerous empty cells because of perfect prediction and a number of factors being dropped from the model. As a result, we included insurance as a covariate in the analyses for race or ethnicity and primary language but conducted a separate analysis for insurance status as a predictor without race or ethnicity or primary language included as covariates. There was no evidence of lack of fit among any of the final regression models. All levels of missing data were well below a 5% threshold where sensitivity analysis would have been indicated. Statistical testing was conducted using STATA version 15.1 (STATA Corp).

Results

Telemedicine Use by Racial and Ethnic Minorities

After controlling for age, gender, subspecialty, insurance, and median household income, patients who were Hispanic were less likely to be seen through telemedicine during the study period than were patients who were white (OR 0.59 [95% confidence interval 0.39 to 0.91]; p = 0.02). After controlling for those same confounders, we also found that Asian patients were also less likely to be seen via telemedicine than were white patients (OR 0.73 [95% CI 0.53 to 0.99]; p = 0.04) (Table 2).

Table 2.

Results of the multivariable regression analysis evaluating the adjusted association between telemedicine visit type and race or ethnicity

Parameter	Adjusted OR (95% CI)	p value
Primary outcome
Race
White	Ref	Ref
Black	0.97 (0.77-1.22)	0.81
Hispanic	0.59 (0.39-0.91)	0.02
Asian	0.73 (0.53-0.99)	0.04
Other	0.85 (0.64-1.13)	0.27
Covariates
Age	0.99 (0.99-0.99)	< 0.001
Gender	1.00 (0.90-1.11)	0.98
Specialty
Arthroplasty	Ref	Ref
Foot and ankle	0.84 (0.69-1.01)	0.07
Hand surgery	1.14 (0.96-1.35)	0.13
Shoulder and elbow	0.82 (0.61-1.11)	0.20
Spine surgery	1.42 (1.19-1.70)	< 0.001
Sports surgery	0.75 (0.63-0.90)	0.002
Insurance
Private	Ref	Ref
Medicare	0.87 (0.75-1.03)	0.10
Medicaid	0.85 (0.71-1.02)	0.08
Workers compensation	0.95 (0.59-1.51)	0.82
Veterans	0.49 (0.17-1.38)	0.18
Self-pay	0.54 (0.19-1.51)	0.24
Other	0.76 (0.17-3.41)	0.72
Median household income	1.00 (1.00-1.00)	0.14

OR = odds ratio; Ref = referent group.

Telemedicine Use by Non-English Speakers

After controlling for race, age, gender, specialty, insurance type, and median household income, participants whose primary language was one other than English or Spanish were less likely to access orthopaedic services via telemedicine than patients who spoke English or Spanish as their primary language (OR 0.34 [95% CI 0.18 to 0.65]; p = 0.001) (Table 3).

Table 3.

Results of the multivariable regression analysis evaluating the adjusted association between telemedicine visit type and language

Parameter	Adjusted OR (95% CI)	p value
Primary outcome
Language
English	Ref	Ref
Spanish	0.73 (0.47-1.11)	0.14
Other	0.34 (0.18-0.65)	0.001
Covariates
Age	0.99 (0.99-0.99)	< 0.001
Gender	1.00 (0.91-1.12)	0.88
Specialty
Arthroplasty	Ref	Ref
Foot and ankle	0.83 (0.68-1.00)	0.06
Hand surgery	1.12 (0.95-1.33)	0.16
Shoulder and elbow	0.82 (0.61-1.10)	0.19
Spine surgery	1.43 (1.20-1.70)	< 0.001
Sports surgery	0.75 (0.63-0.89)	0.001
Insurance
Private	Ref	Ref
Medicare	0.86 (0.74-1.01)	0.07
Medicaid	0.89 (0.75-1.07)	0.23
Workers compensation	0.97 (0.61-1.53)	0.90
Veterans	0.48 (0.17-1.35)	0.17
Self-pay	0.58 (0.20-1.64)	0.31
Other	0.88 (0.19-3.96)	0.87
Median household income	1.00 (0.99-1.00)	0.13

OR = odds ratio; Ref = referent group.

Telemedicine Use by Those Insured Through Medicaid

After controlling for age, gender, subspecialty, and mean household income, patients insured through Medicaid had a lower likelihood of receiving a telemedicine evaluation than did patients who were privately insured (OR 0.83 [95% CI 0.69 to 0.98]; p = 0.03) (Supplementary Table 1; Supplemental Digital Content 1, http://links.lww.com/CORR/A554).

Discussion

Disparities in orthopaedic care are well known from prior work, including reduced access to orthopaedic services, lower use of surgery, and a higher likelihood of adverse events postoperatively [5, 18, 21, 27]. Telemedicine has been touted as a promising platform to mitigate various barriers to care, including enhanced access to higher-volume centers by people living in rural communities [3, 12, 17]. Although prior orthopaedic studies on telemedicine have focused on patient satisfaction, efficiency of care, and cost [11, 13], comparatively little consideration has been given to health disparities. Notably, concerns from other specialties have signaled that relying on telemedicine without considering differential resources among patients could exacerbate disparities [7, 15].

Limitations

We acknowledge several limitations to this study. First, because the study was retrospective and observational, no causality can be inferred from the described associations. Because of the study’s design, there were also inherent limitations on inferring associations with the telemedicine modality versus in-person visits owing to the overall decrease in utilization of orthopaedic care during the COVID-19 pandemic, as evidenced by the decrease in new patient visits from 2019 to 2020. Patients might have hesitated to seek clinical care during the initial phases of the COVID-19 pandemic and may have been unwilling to interact with the medical system except for more urgent conditions. We did not assess the urgency of the clinical question and thus do not know how this may have disproportionately impacted different racial, ethnic, language, or insurance groups, which is an area for further study. The findings of the study are still valid, as our goal was to describe differences in proportions; however, further research should be undertaken to determine contributing factors. Socioeconomic stressors might also play a role and can disproportionately affect certain social or ethnic groups differently during a pandemic. Most likely these stressors exacerbate preexisting barriers to care; however, our findings should be revisited once telemedicine becomes more broadly accepted and the threat of COVID-19 recedes.

Another limitation is that our work captured only patients who completed an outpatient visit. We were not able to characterize patients who declined to be seen through telemedicine or who were not even made aware of the option. This presents a possible source of bias as certain patients, such as non-English-speaking patients or patients from certain cultural backgrounds, may have been less likely to know about the possibility of telemedicine or inquire about its availability. Nonetheless, standardized department protocol during the study period necessitated that any patient inquiring for an appointment was asked to conduct a telemedicine visit, somewhat mitigating this bias. We were further unable to characterize the patients’ reasons for declining telemedicine encounters if the option was presented. For example, the patients in both groups had a wide variety of private insurance coverage, which may have covered telemedicine at different times that may have impacted individual patients’ willingness to have a telemedicine visit, a factor that was not accounted for in this study. Although this limitation does not disqualify our results as we present a comparative description of overall trends without causality, this represents a further line of research that would benefit from qualitative interviews with patients regarding their perceptions of the telemedicine encounter and potential barriers to care. Further work is also needed to identify specific targets in the chain of care and opportunities for improvement.

Furthermore, the current work relied on self-reported measures of mutually exclusive race and primary language, as available through the electronic medical record. The issue of race and ethnicity is complex, and with the growing diversity in our society it is challenging to create an exhaustive list that is inclusive of all racial and ethnic origins. The designations used in this study were limited by the retrospective observational nature of the study, and they rely on self-reported social definitions of race and do not attempt to define any relationship between racial or ethnic associations genetically or anthropologically. Furthermore, some patients may define themselves as more than one race or language, but for this study only the primary self-reported designation was used. There may be interactions between racial or ethnic groups that are not described in this study for those reasons.

Finally, this work was conducted at two medical centers in a single city on the east coast of the United States. Although we believe we accrued a sufficiently representative sample to evaluate the associations of certain sociodemographic factors with telemedicine use, our findings may not be translatable across all clinical contexts, especially in settings where there is less variation in the factors we considered and increased homogeneity between healthcare providers and the population served. The differences presented between the two comparison groups represent changes over a specific baseline in a particular catchment area for two hospitals. Furthermore, there may be preexisting disparities even in the 2019 comparison group; therefore, the results should be interpreted only as changes over time rather than an absolute measure of inequity.

Telemedicine Use and Orthopaedic Healthcare Disparities

We found concerning disparities in access to orthopaedic care via telemedicine for patients, and we observed that those disparities were associated with race or ethnicity, primary language, and insurance status. We believe this investigation is strengthened by a large and diverse sample, including a broad range of racial or ethnic designations, primary languages, insurance carriers, and orthopaedic subspecialties. Specifically, the odds of a Hispanic patient having a telemedicine consultation were 41% lower compared with that of a white patient, while the odds of a telemedicine encounter for Asian patients was similarly 27% lower compared with white patients. Similar changes were encountered for patients with Medicaid having a 15% lower odds of having a telemedicine visit, while the most drastic change in our comparison occurred for patients whose primary language is not English or Spanish, who had 66% lower odds of being seen via telemedicine.

There likely are several reasons for these observed differences. Some of these may be patient-facing, such as access to and use of the healthcare system, cultural norms regarding when it is appropriate to seek healthcare, and trust of the medical system during a pandemic. Others may be institution-facing, perhaps including the variety of technological platforms used, outreach efforts for non-English speakers, and scheduling availability when many patients had increased responsibility at home such as childcare. It is also important to note that this study is observational, and further research is needed to determine causality for any of these observed associations. Further, these scenarios may already be artifacts of systemic barriers and long-standing socioeconomic disparities that existed before the pandemic but were also exacerbated by it. For example, during the COVID-19 pandemic, “essential workers” were required to be physically at their place of work. Because many nonwhite, non-English-speaking individuals who are employed fall under this designation [19], their ability to access telemedicine visits during regular business hours would be diminished. Innate concerns among patients regarding communication and language would further complicate a desire for engagement through telemedicine [22]. Moreover, disproportionate rates of COVID-19 among these populations may have diminished the relative importance of musculoskeletal complaints by comparison.

Lower access for those insured through Medicaid is aligned with prior studies, which have shown restricted or delayed orthopaedic visits for those with this insurance type [1, 10, 23]. All previous investigations of which we are aware, however, focused on disparities that arose from orthopaedic practices that did not accept Medicaid. Throughout the study period, telemedicine clinic visits were eligible for expanded Medicaid coverage under a Massachusetts state-of-emergency waiver [25]. Despite the increased reimbursement available for patients insured through Medicaid, these individuals were still less likely to be seen. The technological requirements of reliable internet access and devices may have contributed to this finding.

These observations culminate in our cautionary stance: Although broad expansion of telemedicine services might improve access to care and efficient delivery for certain individuals, it also may worsen healthcare segregation for certain racial and ethnic minorities, patients with limited English-language proficiency, and patients who are underinsured. This could stem from two pernicious scenarios. First, telemedicine expansion by healthcare providers may not be equitably accessible to Hispanic patients or those who do not speak English or Spanish. Second, some patients may only be able to access telemedicine services through safety-net hospitals or receive care through in-person visits, with the attendant increased costs of travel and waiting times. Previous, well-intentioned health policy initiatives that were anticipated to provide ancillary benefits to nonwhite populations, such as accountable care organizations and centers of excellence, were also found to exacerbate healthcare disparities [14]. These findings place the onus on providers, institutions, and local governance such as state Medicaid agencies to facilitate equal access for certain patients, particularly for nonmodifiable risk factors such as race, as highlighted in this study. For example, this might include expanded reimbursement for telemedicine conducted through a wider variety of telemedicine platforms to accommodate patients with limited technological resources. Notably, Hispanic patients had lower odds of having a telemedicine visit compared with white patients, however Spanish speakers did not have lower odds compared with English speakers. This suggests that the Hispanic community should not be treated as a Spanish-speaking monolith when designing pathways for care access (that is, challenges for native Spanish speakers may be different from those experienced by other Hispanic-Americans), and highlights one example of how a nonmodifiable factor such as race or ethnicity may be associated with care in ways separate from language.

In light of our findings, we believe we can propose several potential solutions. Healthcare systems expanding their outreach through telemedicine should ensure that information regarding this platform is appropriately disseminated among minority and low-income communities. Healthcare systems can ensure that their telemedicine modalities serve a population with diverse technological needs for patients who may not have access to a computer with a camera, such as telephone-only visits or community kiosks with equipped technology. Furthermore, local governance such as state Medicaid agencies should also ensure that these alternative telemedicine modalities are eligible for reimbursement to ensure that technology in the home is not an artificial prerequisite to access care. Expanding hours for telemedicine encounters outside the normal daily work schedule may also improve access for disadvantaged patients, as would appropriate availability of interpreter services. The onus rests on healthcare systems and their orthopaedic clinics to develop language- and culturally-conscious materials to advise patients of telemedicine and to help make this resource available through appropriate interpreter support.

Conclusion

We found that lower proportions of Hispanic, Asian, underinsured, and non-English speakers established orthopaedic care via telemedicine during the COVID-19 pandemic compared with a similar patient group one year prior. Although the adoption of telemedicine may mitigate geographic barriers to care and broaden healthcare markets, our findings raise concerns that it could widen gaps in healthcare use for specific racial/ethnic, language, or socioeconomic groups. As healthcare moves toward increased telemedicine use, care must be taken to ensure that certain populations are not left behind. For individual surgeons and practices, this might involve incorporating language about interpreter availability into administrative scripts concerning telemedicine booking. This might also include expanding telemedicine office hours to evenings and weekends to allow patients who work in-person jobs to still take advantage of telemedicine visits. On the institution level, partnering with or creating community initiatives can reduce inequity across different departments and specialties, such as culturally-conscious, language-specific advertisements on public transportation or social media and targeted outreach to community primary care providers. Investment in a variety of telemedicine platforms such as HIPAA-compliant smartphone apps for patients and providers could also reduce this gap. Finally, at the level of public policy, possibilities include expansion of reimbursement for a variety of visit types such as telephone visits, or removing the requirement to list patient location during a visit. Although telemedicine bears great opportunity for increased access and scaling more efficient and higher-quality care programs, it also brings with it a great responsibility of ensuring a future that includes health equity.