Association Between Limited English Language Proficiency and Disparities in Length of Stay and Discharge Disposition After Total Shoulder Arthroplasty: A Retrospective Cohort Study

Abstract

Background: Limited English language proficiency in patients undergoing total shoulder arthroplasty (TSA) may make treatment more challenging. Purpose: We sought to investigate the potential association between TSA patients’ use of a language interpreter and 2 outcomes: hospital length of stay (LOS) and discharge disposition. Methods: We conducted a retrospective cohort study comparing LOS and discharge disposition after TSA for patients who required interpreter services and patients who did not at a single institution in an urban setting between 2016 and 2020. Consecutive patients requiring interpreter services who underwent TSA were matched 1:1 to patients who did not require an interpreter by age, body mass index (BMI), sex, and procedure. Multivariate regression models controlling for age, BMI, sex, smoking, opioid use, white or non-white race, procedure, and diagnosis were constructed to determine associations between interpreter use, LOS, and discharge disposition. Results: Forty-one patients were included in each cohort, exceeding the minimum number required per an a priori power analysis. Mean hospital LOS was longer in the interpreter cohort than in the non-interpreter cohort (2.8 ± 2.4 vs 1.8 ± 1.0 days, respectively). Multivariate linear regression demonstrated interpreter use was the strongest predictor of LOS, with the effect estimate indicating an additional 0.88-day LOS per patient. A greater proportion of patients from the interpreter cohort were discharged to an acute/subacute rehabilitation facility than patients from the non-interpreter cohort (n = 8 [19.5%] vs n = 2 [4.9%], respectively). Patients from the interpreter cohort were 454% more likely to be discharged to acute/subacute rehabilitation facilities. Conclusions: Our retrospective analysis of patients undergoing TSA suggests that the need for interpreter services may be associated with increased LOS and discharge to a facility. More rigorous study is needed to identify the factors that influence these outcomes and to avoid disparities in hospital stay and discharge.

Introduction

Projections for total shoulder arthroplasty (TSA) procedures performed in the United States annually are estimated to increase by 235% by the year 2025, outpacing the growth rates of hip and knee total joint arthroplasty (TJA) [20]. Based on these increasing demands, clinical and financial pressures secondary to value-based health care movements will necessitate practices that minimize costs while producing clinically meaningful outcomes [1,4]. Perioperative and clinical outcomes are considered an important metric to define quality of care and value [15,17]. Thus, it is imperative to better understand which patient-specific factors are associated with poor outcomes or episodes of care that may result in greater costs to patients, hospitals, and surgeons. Indeed, there has been substantial research on identifying patient risk factors and their impact on clinical outcomes after TSA [5,12,13].

Several investigations have focused on the influence of race, ethnicity, and socioeconomic status on patient outcomes and quality of care. Poor outcomes among racial and ethnic minority groups and those with lower socioeconomic status have been well-documented in the shoulder literature [8], as well as in the hip and knee literature [7,9,14,16]. Such factors may result in disproportionate hospital length of stay (LOS) and influence discharge disposition for patients who are at a socioeconomic disadvantage, possibly leading to substantial differences in the quality and costs for care.

A related potential risk factor is limited English language proficiency. A report from 2015 estimated that as of 2013, a total of 25.1 million individuals in the United States had limited English language proficiency [21]. Bernstein et al [2] reported that patients requiring an interpreter experienced significantly longer hospital LOS and were less likely to be discharged home following total hip arthroplasty (THA) and total knee arthroplasty (TKA), despite widely available interpreter services. Given the increasing surgical volume of TSA as well as the growing importance of value-based care and alternative payment models, it is imperative to understand whether similar relationships exist for patients undergoing TSA. The purpose of the current study was to investigate the potential influence that the use of an interpreter had on patients’ LOS and discharge disposition after TSA. We hypothesized that after TSA, those who required the use of language interpreter services during their care would have longer hospital LOS and fewer discharges home than those who did not.

Methods

Institutional review board approval was obtained prior to the commencement of this study. This study was performed in accordance with the Strengthening the Reporting of Observational studies in Epidemiology (STROBE) statement and REporting of studies Conducted using Observational Routinely-collected Data (RECORD) guidelines [10,19]. It was a single-institution study in an urban setting using an institutional TSA registry with contributions from 12 fellowship-trained sports medicine and shoulder surgeons. The registry was queried for consecutive patients who underwent reverse or anatomic TSA between January 2016 and March 2020. Patients were excluded if one of the following items pertained to their episode of care: (1) underwent shoulder arthroplasty for infection including 2-stage exchange or irrigation and debridement procedures or (2) did not have data available pertaining to their preferred language or use of a language interpreter. All language interpreters were licensed experts fluent in a particular language of interest, including Spanish, Arabic, Cantonese, Russian, Italian, Greek, and Polish. Interpreters were used in the outpatient clinic setting, during the inpatient stay, and to discuss discharge instructions using either on-site in-person, videoconferencing, or telephone interpretation services provided free of charge by the institution. At a minimum, videoconferencing or telephone services were available at any time for all languages listed.

An a priori power analysis was performed to determine an appropriate number of patients to include in both the interpreter and the non-interpreter cohorts to be adequately powered. Assuming a mean LOS of 2 ± 1.5 days for the non-interpreter cohort and 3 days for the interpreter cohort with an alpha of 0.05 and power of 80%, this analysis indicated that a minimum of 35 patients in each group would be required. We then identified patients who were required the use of a language interpreter throughout their hospital stay from a repository. Patients requiring an interpreter were identified through documented translator use, preferred language documentation, and use of non–English language forms in electronic medical records. This cohort was subsequently matched in age, body mass index (BMI), sex, and procedure to patients who did not require the use of a language interpreter.

The primary outcome of interest was LOS until discharge after TSA measured in days. Secondary outcomes included discharge disposition and 30- and 90-day complications. Discharge disposition was classified as discharge to the patient’s home, to a subacute rehabilitation facility, or to an acute inpatient rehabilitation facility. Demographic and preoperative variables were also recorded for all patients and included (1) age, (2) BMI, (3) sex, (4) American Society of Anesthesiologists (ASA) score, (4) diagnosis, (5) procedure, (6) laterality, (7) white or non-white race, (8) insurance type, (9) medical comorbidities, (10) smoking status, and (11) preoperative opioid use. Race was extracted from the electronic medical record and subsequently dichotomized into white versus non-white race for the purpose of analysis only.

Normality of the primary and secondary outcomes was determined prior to analysis using the Shapiro-Wilk test. Descriptive statistics were used to quantify data with means and standard deviations or frequencies with percentages where appropriate. Baseline characteristics and outcomes of interest were compared between the cohorts with the appropriate parametric or nonparametric testing. Categorical or ordinal data were compared using the χ² test of association or Fisher’s exact test. Multivariate linear regression models were constructed to determine the potential association between interpreter use and LOS while controlling for potential confounding variables; multivariate logistic regression models were constructed to investigate the potential association between interpreter use, discharge disposition, and short-term complication rates. All statistical analyses were performed using Stata Version 16.1. A P value of less than .05 was used for all analyses to indicate statistical significance.

Results

A total of 41 patients who underwent TSA and required the use of a language interpreter (interpreter cohort) were identified. A total of 41 patients who did not required the use of a language interpreter (non-interpreter cohort) and underwent TSA during the same period were matched by age, BMI, sex, and procedure. Baseline characteristics used for matching did not differ between these cohorts, indicating appropriate matching. Additional baseline characteristics of the study population are listed in Table 1.

Table 1.

Baseline characteristics of study population.

Characteristic	Interpreter	Non-interpreter	All	P value
Age, years	72.1 ± 11.7 (39–95)	72.1 ± 11.4 (42–94)	72.1 ± 11.5 (39–94)	.98
BMI, kg/m2	31.5 ± 7.4 (21.9–54.7)	30.2 ± 6.5 (20.5–50.7)	30.8 ± 6.9 (20.5–54.7)	.42
Sex				1.00
Female	27 (66.6%)	27 (66.6%)	54 (65.9%)
Male	14 (33.3%)	14 (33.3%)	28 (34.1%)
Race				<.001
White	22 (53.7%)	37 (90.2%)	59 (72.0%)
Non-white	19 (46.3%)	4 (9.8%)	23 (28.1%)
Insurance type				.031
Private	12 (29.3%)	11 (26.8%)	23 (28.1%)
Medicare/Medicaid	23 (56.1%)	30 (73.2%)	53 (64.6%)
International/other	6 (14.6%)	0 (0%)	6 (7.3%)
Smoking				1.00
Current	3 (7.3%)	3 (7.3%)	6 (7.3%)
Former/never	38 (92.7%)	38 (92.7%)	76 (92.7%)
Preoperative opioid use				.58
Yes	7 (17.1%)	9 (22.0%)	16 (19.5%)
No	34 (82.9%)	32 (78.0%)	66 (80.5%)
ASA score				.87
I	1 (2.4%)	1 (2.4%)	2 (2.4%)
II	30 (73.2%)	32 (78.0%)	62 (75.6%)
III	10 (24.4%)	8 (19.5%)	18 (22.0%)
Diagnosis				.010
Primary osteoarthritis with or without rotator cuff tear	32 (78%)	41 (100%)	73 (89.0%)
Secondary osteoarthritis	7 (17.1%)	0 (0%)	7 (8.5%)
Proximal humerus fracture	2 (4.9%)	0 (0%)	2 (2.4%)
Procedure				1.00
Anatomic TSA	10 (24.4%)	10 (24.4%)	20 (24.4%)
Reverse TSA	31 (75.6%)	31 (75.6%)	62 (75.6%)

TSA total shoulder arthroplasty, BMI body mass index, ASA American Society of Anesthesiologists.

Data presented as means with standard deviations (and ranges) or frequencies with percentages.

Bolded P values indicate statistical significance at P < .05 level.

The mean hospital LOS among all patients was 2.3 ± 2.1 days. Patients in the interpreter cohort had significantly longer hospital LOS on average compared with those in the non-interpreter cohort (2.8 ± 2.4 vs 1.8 ± 1.0 days, P = .048). Multivariate linear regression analysis demonstrated that the need for an interpreter was the strongest predictor of hospital LOS, with the effect estimate indicating an additional 0.88-day LOS per patient compared with patients who did not need a language interpreter (Table 2).

Table 2.

Multivariate linear regression analysis estimates for predicting hospital length of stay after TSA.

Independent variable	Effect estimate (β)	95% confidence interval	P value
Interpreter use	0.88	0.28–1.81	.008
Age	0.06	−0.036 to 0.050	.75
BMI	−0.028	−0.09 to 0.031	.34
Female sex	0.74	−0.25 to 1.73	.14
Non-white race	0.82	−0.24 to 1.89	.13
Insurance type	0.71	0.12–1.64	.022
Current smoking	1.89	−0.081 to 5.64	.74
Preoperative opioid use	0.89	−0.16 to 1.95	.097
Diagnosis	0.46	−0.071 to 3.69	.25
Anatomic TSA	0.12	−0.94 to 1.18	.82

BMI body mass index, TSA total shoulder arthroplasty.

Interpreter use compared with no interpreter use; female sex compared with male sex; non-Caucasian race compared with Caucasian/white race; current smoking compared with never/former smokers; preoperative opioid use compared with no opioid use history; diagnosis other than primary osteoarthritis compared with diagnosis of primary osteoarthritis; anatomic TSA compared with reverse TSA.

Bolded P values indicated statistical significance at P < .05 level.

A total of 72 (87.8%) patients were discharged home following TSA, while the remaining 10 patients were discharged to an acute or subacute rehabilitation facility. Comparison of study cohorts revealed that the need of a language interpreter was associated with discharge disposition; a greater proportion of patients in the interpreter cohort were discharged to an acute or subacute rehabilitation facility compared with patients in the non-interpreter cohort (n = 8 [19.5%] vs n = 2 [4.9%], P = .033). Multivariate logistic regression demonstrated that the need for an interpreter was the strongest predictor of discharge disposition, with patients who required an interpreter being 454% more likely to be discharged to a subacute or acute rehabilitation facility compared with patients who did not need an interpreter (Table 3).

Table 3.

Multivariate logistic regression analysis estimates for predicting discharge disposition after TSA.

Independent variable	Odds ratio	95% confidence interval	P value
Interpreter use	4.54	2.13–8.61	.037
Age	1.12	0.98–1.26	.12
BMI	0.97	0.85–1.10	.63
Female sex	2.8	0.53–4.32	.16
Non-white race	2.93	0.38–22.74	.30
Insurance type	0.14	0.016–1.24	.078
Current smoking	1.02	0.17–4.62	.46
Preoperative opioid use	1.31	0.14–12.54	.68
Diagnosis	0.52	0.06–6.01	.60
Anatomic TSA	0.47	0.064–3.50	.46

BMI body mass index, TSA total shoulder arthroplasty.

Interpreter use compared with no interpreter use; female sex compared with male sex; non-Caucasian race compared with white race; current smoking compared with never/former smokers; preoperative opioid use compared with no opioid use history; diagnosis other than primary osteoarthritis compared with diagnosis of primary osteoarthritis; anatomic TSA compared with reverse TSA.

Bolded P values indicated statistical significance at P < .05 level.

The overall 30- and 90-day complication rates were 6.1% and 9.8%, respectively (Table 4). Analysis of 30-day complication rates demonstrated a trend toward a higher frequency of complications in the interpreter cohort, with 1 complication in the non-interpreter cohort and 4 complications in the interpreter cohort, though this failed to reach statistical significance (P = .36). The complication in the non-interpreter cohort was an upper extremity deep vein thrombosis (DVT), whereas the complications in the interpreter cohort included periprosthetic joint infection (PJI) necessitating 2-stage exchange; PJI necessitating irrigation and debridement; chronic regional pain syndrome in the ipsilateral upper extremity; and glenosphere migration on radiograph, with inability to use upper extremity, requiring revision. Likewise, no statistically significant association was found between 90-day complication rates and interpreter use, with a total of 6 cumulative complications documented for the interpreter cohort and 2 cumulative complications for the non-interpreter cohort (P = .26). Complications in the non-interpreter cohort included delayed onset median nerve neuropathy confirmed by electromyography, whereas those in the interpreter cohort included pseudoparalysis, rotator cuff insufficiency, and subjective instability requiring revision.

Table 4.

Cohort-specific 30- and 90-day complications and cumulative complication rates.

Adverse Event	Interpreter	No interpreter	P value
Cumulative 30-day complication rate	4 (9.8%)	1 (2.4%)	.36
Cumulative 90-day complication rate	6 (14.6%)	2 (4.9%)	.26

Discussion

Our retrospective cohort study of patients who underwent TSA found an association between the use of language interpreter services and the longer hospital LOS, as well as discharge to subacute or acute rehabilitation facilities. We found no association between the need for interpreter services and an increase in 30- and 90-day complication rates. These findings suggest the importance of interventions beyond the patient-physician interaction, including those that promote trust throughout the health care system, such as efforts toward addressing language barriers.

A few limitations of this study warrant discussion. First, though patients were matched by several potential confounding factors and baseline characteristics were similar, given the retrospective nature of this study, potential confounding factors that were not controlled for may contribute to the differences between cohorts. For example, information on conditions such as obstructive sleep apnea diagnosis, which has implications for LOS, was not available to be studied. Second, our study was not powered to detect differences in complication rates, and future studies with larger study populations may report conflicting data. Third, we did not account for day of the week the TSA procedures were performed, though matched patients were randomly selected, which could minimize a skewed distribution of earlier or later procedural days. Fourth, although the use of a registry allowed for a greater number of subjects, it also introduces a limitation in that each surgeon has different thresholds and protocols for postoperative LOS, and we neither quantified nor studied this inherent variation. Fifth, we chose to incorporate race in our analysis as a binary variable (white or non-white), and we acknowledge that this grouping may neglect important cultural and racial differences that exist among patients. Last, we did not report on the following social factors that may contribute to LOS: (1) a patient’s cultural expectations of hospital stay when that person does not normally seek care in the United States and (2) a patient’s home situation, which may influence safe discharge home postoperatively. These variables are important but difficult to quantify and were not available in a retrospective chart review.

We found an association between requiring a language interpreter and an additional 1 day LOS on average. This was confirmed on regression analysis, where interpreter use was associated with an additional 0.88-day LOS per patient while controlling for several potential socioeconomic confounding factors (insurance type, race, gender, smoking status, and ASA score). These findings are in accordance with previous literature suggesting that patients’ limited English language proficiency may be linked to worse outcomes. Bernstein et al [2] performed a retrospective cohort study of primary THA and TKA patients who required a language interpreter and those whose primary language was English and found a significantly longer mean hospital LOS for patients who required an interpreter who underwent THA (1 day longer on average) but not TKA. They suggested that this effect was likely the result of communication disruptions that could be attributed to a language barrier. We found that patients who required an interpreter were more likely to be discharged to facilities rather than home after TSA. Similarly, Bernstein et al [2] found that non–English-speaking patients were discharged to facilities rather than home more frequently than English-speaking patients after both TKA and THA. Previous literature has reported that patients who required an interpreter after TKA experienced worse functional outcomes compared with their English-speaking counterparts [6]; whether this may also be applicable to patients undergoing TSA is worthy of further study. Another study found that patients who required a language interpreter had to wait 19 minutes on average for in-person translation services [3], which may significantly subtract from the time a physical or occupational therapist has to work with a patient.

These findings suggest that across orthopedic specialties, disparate outcomes are not infrequent for patients who require an interpreter. This is despite interpreter services being free of charge and widely available at our institution, indicating that differences in outcomes persist despite access to inpatient resources. One plausible explanation is that language barriers and inadequate communication between hospital staff and patients may prevent some patients from expressing their readiness for discharge or confidence in their postoperative function. Communication barriers may also affect instructions between physical therapy staff and patients, which may affect the final assessment of functional progress and lead to a higher risk of discharge to rehabilitation facilities. The additional time necessary for care coordination may subsequently lead to a decrease in surgical utilization and increase the gap in health care for these patients. To mitigate this risk, health care institutions may benefit from examining such cases in detail and adjusting compensation to not further existing disparities in these patients. It is also important to acknowledge that racial and cultural differences exist among patients, and systemic racism and clinicians’ implicit biases may contribute to differences in LOS and discharge disposition. We propose that preoperative education of patients with limited English proficiency may help to mitigate challenges that may arise regarding hospital stay and discharge. With recent shifts toward the use of telemedicine and widespread availability of interpreter services, this may be a feasible intervention that could benefit patients’ limited English proficiency and hospital systems. Furthermore, institutions may benefit from comprehensive implicit bias training to address the potential for racism, bias, and cultural differences that contribute to disparate outcomes. Pandya et al [11] recommended the use of a 10-item checklist of actions that can be used by individual surgeons to aid in the lessening of care disparities in their practice. Turcotte et al [18] evaluated the effect of an enhanced preoperative education pathway (EPrEP) on outcomes after TJA and found that although no differences in LOS or discharge disposition occurred based on race, use of an EPrEP resulted in a 51% reduction in 30-day postoperative emergency room visits. Unfortunately, despite increasing evidence highlighting persistent disparities in orthopedic care, few studies have investigated the efficacy of interventions used to reduce such disparities. Future studies are warranted to determine the factors contributing to disparities in hospital LOS and inpatient outcomes in orthopedics, as well as to investigate the interventions that are most effective at reducing racial and ethnic disparities in orthopedic patients.

Short-term complication rates were statistically similar among both cohorts, though the interpreter cohort experienced higher cumulative complication rates at both 30 and 90 days postoperatively. It is unlikely that the need for an interpreter in the inpatient setting would predispose patients to surgical complications within 3 months postoperatively, although difficulty interpreting discharge instructions could contribute. For example, 1 patient in this cohort experienced an upper extremity DVT, while 2 experienced surgical site infections. If the instructions on DVT prophylaxis or wound care were not clearly explained or understood due to either a language barrier or limited time available to discuss instructions while waiting for an interpreter, this could contribute to risk of DVT or surgical site infection. Of course, this is speculation; our retrospective study cannot establish causation. Further study may be warranted to investigate complication rates between cohorts.

In conclusion, we found an association in patients undergoing TSA between the need for interpreter services and the increased LOS or discharge to a facility, but not short-term complication rates. We recommend further study of patients requiring interpreter services to better understand the challenges that may contribute to disparities in hospital stay and discharge.