Abstract
Objectives.
To compare 7-day readmission among patients with a non-English primary language in language-serving versus non–language-serving hospitals.
Methods.
We conducted a retrospective cohort study using the New Jersey State Inpatient Database (2010–2022). We identified 578 030 inpatient discharges home among patients with a non-English primary language. We used generalized linear mixed models to compare 7-day readmission between language-serving hospitals—defined as those with a higher proportion of patients with a non-English primary language—and non–language-serving hospitals.
Results.
Patients with a non-English primary language were less likely to be readmitted within 7 days at language-serving versus non–language-serving hospitals (4.8% vs 9.8%; adjusted odds ratio = 0.49; 95% confidence interval = 0.25, 0.96). This trend persisted when stratified into Spanish (5.5% vs 9.7%), common (4.0% vs 14.6%) and rare non-English primary languages (4.0% vs 8.8%).
Conclusions.
Patients with a non-English primary language discharged from language-serving hospitals were less likely to be readmitted within 7 days.
Public Health Implications.
To improve care for patients with a non-English primary language, future research should identify the hospital- and state-level language access policies that underpin these findings.
More than 26 million people in the United States speak English “less than very well,”1 defined as limited English proficiency.2 According to US census data, English proficiency varies among those who speak a non-English language, with a greater proportion of those who speak Spanish (39.6%) or an Asian/Pacific Islander language (44.1%) reporting limited English proficiency.3 Within health care, patients with a non-English primary language, used as a proxy for limited English proficiency because proficiency is difficult to measure, experience worse outcomes compared with their English-proficient counterparts, including more frequent emergency department visits and readmissions and longer stays.4 These trends persist across a range of surgical and medical specialties, including trauma, surgery, mental health, and medicine,4–6 indicating that linguistic barriers affect quality of care throughout the hospital system.
Patient–provider language concordance and professional medical interpretation improve comprehension, quality of care, and patient satisfaction and reduce length of stay and 30-day readmissions.7,8 The limited availability of multilingual physicians necessitates reliance on professional medical interpretation, but interpretation and translation resources vary widely by hospital.9,10 Although recipients of federal funds are required to provide reasonable access to services under Title VI of the Civil Rights Act of 1964 (Pub L No. 88–352) and Section 1557 of the Affordable Care Act (Pub L No. 111–148),11 it is ultimately left to hospitals to determine the scope of these services based on their patient population.
Hospital-level language access resources may influence early readmissions for patients with a non-English primary language.12 Language-serving hospitals are those that serve a relatively high proportion of patients with a non-English primary language.13 The term “language-serving hospitals” builds on existing institutional designations, including “minority serving hospitals,” “safety net hospitals,” and “Hispanic serving institutions,” that indicate that these institutions, by virtue of their larger volume of historically underserved populations, play an out-sized role in ensuring pathways to health care, education, and other resources.14–16 Additionally, the demographics of the hospital catchment area are used as proxies for demand for services, including interpreter services.17 As patient and catchment area demographics are key drivers of hospital investment decisions,17–19 the higher volume of patients with a non-English primary language at language-serving hospitals may lead to greater investment in interpreter and translation resources, increasing access to language-concordant care and improving outcomes for patients with a non-English primary language.
We compared 7-day readmission among patients with a non-English primary language discharged from language-serving versus non–language-serving hospitals, hypothesizing that readmission would be lower at language-serving hospitals.
METHODS
We conducted a retrospective cohort study following the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines (https://www.strobe-statement.org).
Study Design and Data Source
We used the 2010–2022 New Jersey State Inpatient Databases from the Healthcare Cost and Utilization Project (HCUP), a program of the Agency for Healthcare Research and Quality.20 The State Inpatient Databases capture all inpatient discharges from New Jersey hospitals each year. Seven other states also report primary language to HCUP, but we selected New Jersey because its hospitals report primary language with less missingness (<1%). New Jersey is also a linguistically diverse state, indicating that there is likely a high demand for language services.21 We linked the discharge-level data with hospital-level data from the 2016 American Hospital Association Annual Survey,22 used because it fell in the middle of our study period.
An estimated 12.3% of the New Jersey population has limited English proficiency.1 We defined language-serving hospitals as hospitals where the proportion of patients with a non-English primary language meets or exceeds the state census proportion of 12.3% across the inpatient (State Inpatient Database),20 ambulatory (State Ambulatory Surgery and Services Database),23 and emergency department (State Emergency Department Database)24 settings. We referred to hospitals that did not meet this threshold as “non–language-serving hospitals,” although all included hospitals served at least 1 patient with a non-English primary language.
Because HCUP data use agreements prevent identification of individual hospitals or direct assessment of language access resources, we used language-serving hospitals as a proxy for hospitals with more robust language access capacity based on previous work indicating that a higher volume of patients with non-English primary language drives linguistic diversity, demand for language access resources, and subsequent investments in language resources.17–19
Study Population
Of the 77 hospitals in the New Jersey HCUP State Inpatient Databases, we excluded those that could not be linked to the American Hospital Association data (n = 5), did not report primary language in all 3 settings (n = 7), did not report the readmission outcome (n = 4), did not report at least 1 patient with a non-English primary language during the study period (n = 1), had no readmission events (n = 2), or were outliers with respect to the hospital random effect (n = 2). Additional details are available in Figure A (available as a supplement to the online version of this article at http://www.ajph.org).
From the 56 remaining hospitals, we included all patients aged 18 to 95 years old who reported primary language and were not discharged from a specialized psychiatric or rehabilitation hospital. We excluded patients from psychiatric or rehabilitation hospitals because of significant differences in the readmission pathway compared with the acute care hospitals included in this study.25,26 We limited the cohort to patients who were not missing age, sex, insurance type, admission type, length of stay, or readmission data; had a length of stay in the 99th percentile or under (28 days); and were discharged home. The pathway to readmission may differ for patients discharged to another health care facility compared with those discharged home, including caregiver experience and access to linguistic resources after discharge. Therefore, we included only patients discharged home to reduce the potential effects that discharge to another health care facility might have on readmission and length of stay.
Variables
To investigate whether language-serving and non–language-serving hospitals were similar with respect to ownership, accreditations, volume, and services, we compared hospital characteristics from the American Hospital Association Annual Survey 2016.22 Selected variables included bed size, Joint Commission accreditation, American College of Surgeons Commission on Cancer approval, medical school affiliation, linguistic or translation services availability (yes, no, or unknown), total facility admissions, and total facility inpatient days.
We identified covariates a priori and included age, sex (male, female), race/ethnicity (non-Hispanic White, non-Hispanic Black, Hispanic, non-Hispanic Asian, non-Hispanic other/unknown [including American Indian/Alaska Native]), insurance type (commercial insurance, Medicare, Medicaid, self-pay, other), discharge year (2010–2022), Elixhauser comorbidity index score (0–1, 2–3, ≥ 4),27 admission urgency (emergency, urgent, elective), urbanicity (urban, rural), neighborhood median household income quartile, patient–hospital county concordance (yes, no), in-hospital complication (yes, no; Appendix A, available as a supplement to the online version of this article at http://www.ajph.org), length of stay, and principal diagnosis group (maternal/neonatal, mental health/substance abuse, injury, surgery, medicine). Sex and race/ethnicity are based on the value reported by the hospital and uniformly coded by HCUP.
We included race/ethnicity in this analysis as a nonexact proxy for multiple social and systemic factors that can affect health outcomes. We grouped urbanicity (urban, rural) using the 2013 Urban Influence Codes. Using county-level Federal Information Processing Standards codes, the patient–hospital county concordance variable assesses the impact of geography on readmission by indicating whether the patient’s home county is the same as the county of the hospital from which they were discharged. We assigned principal diagnosis group hierarchically to maternal or neonatal, mental health or substance use, injury, surgery, and medicine using codes from the International Classification of Diseases, Ninth Revision and 10th Revision (Geneva, Switzerland: World Health Organization; 1980 and 2019), Major Diagnostic Categories (Centers for Medicare & Medicaid Services MS-DRG system), and the Clinical Classifications Software Refined (https://hcup-us.ahrq.gov/toolssoftware/ccsr/dxccsr.jsp).
We compared the outcomes between language-serving and non–language-serving hospitals for all patients with a non-English primary language and then stratified by primary language into 3 groups to act as proxies for how commonly spoken a non-English language is across the study period in New Jersey. We used the stratification to assess whether patients with less commonly encountered languages benefit more from care at language-serving hospitals. Spanish is the most commonly spoken non-English language in New Jersey. Arabic, Portuguese, and Korean are the next most commonly spoken languages, and although we aggregated them to ensure adequate sample size, we suspect they are the best resourced non-English languages in New Jersey hospitals after Spanish. Finally, we combined the remaining 63 non-English languages into the “rare non-English, non-Spanish languages” group because of low sample size for individual languages.
The primary outcome was 7-day readmission to the index hospital. The reference group includes all other patients for whom the New Jersey Department of Health reported readmission data and could include those readmitted to the index hospital after 7 days, those readmitted to a different facility, or those who were never readmitted. In this data set, New Jersey reports only 7-day readmission. The data source did not report the reasons for missing values for readmission.
Statistical Analysis
We compared patient demographics, outcomes, and hospital characteristics between language-serving and non–language-serving hospitals using the Wilcoxon rank–sum test for nonnormally distributed continuous variables, the x2 test for categorical variables, and the Fisher exact test for categorical variables with cell sizes less than 5. We conducted generalized linear mixed models with a random intercept for hospital to assess the odds of 7-day readmission, comparing patients at language-serving hospitals to those at non–language-serving hospitals (reference group). We adjusted models for discharge year, age, sex, race/ethnicity, insurance coverage, Elixhauser comorbidity index score, admission urgency, neighborhood median household income, patient–hospital county concordance, in-hospital complication, length of stay, and principal diagnosis group. We tested variables for multicollinearity, which were low to moderate (variance inflation factors < 5). We excluded urbanicity from the models because of small sample size and lack of statistical significance (P > .05). We used 2-sided tests and considered results statistically significant at P < .05. We conducted all analyses in SAS version 9.4 (SAS Institute Inc., Cary, NC).
We conducted a sensitivity analysis to examine the effect of limiting the cohort to those discharged home on the results by examining 7-day readmission, in-hospital complication, length of stay, and mortality among those discharged home, to rehabilitation care, or transferred to additional inpatient care.
RESULTS
We included 56 hospitals, of which 17 (30.4%) met the definition of a language-serving hospital.
Hospital and Patient Characteristics
The median proportion of patients with a non-English primary language in the inpatient setting was 22.0% (interquartile range [IQR] = 16.1%–32.3%) at language-serving hospitals and 4.9% (IQR = 3.2%–8.2%) at non–language-serving hospitals. A greater proportion of language-serving hospitals had a medical school affiliation compared with non–language-serving hospitals (82.4% vs 46.2%; P = .018) and a lower proportion reported American College of Surgeons Commission on Cancer approval (47.1% vs 74.4%; P = .047). We did not observe any other hospital-level differences between language-serving and non–language-serving hospitals (Table A, available as a supplement to the online version of this article at http://www.ajph.org).
We included 578 030 patients with a non-English primary language (Figure 1): 64.8% were discharged from language-serving hospitals and 35.2% from non–language-serving hospitals (Table 1). Patients at language-serving hospitals were slightly older (language-serving hospitals: median age = 55 years; IQR = 35–72; non–language-serving hospitals: median age = 48 years; IQR = 32–69) and a greater proportion were non-Hispanic (44.5% vs 35.2%), were insured with Medicare (18.7% vs 13.2%), were admitted emergently (70.3% vs 62.5%), and had more comorbidities (4 or more: 31.3% vs 26.6%).
FIGURE 1— Study Sample Selection for Patients With a Non-English Primary Language: New Jersey State Inpatient Database, 2010–2022.
Note. FIPS = Federal Information Processing Standards.
TABLE 1—
Demographics of Patients With a Non-English Primary Language at Language-Serving and Non–Language-Serving Hospitals: New Jersey State Inpatient Database, 2010–2022
| No. (%) or Median (IQR) | ||||
|---|---|---|---|---|
| Variable | Overall (n = 578 030) | Language-Serving Hospitals (n = 374620) | Non-Language-Serving Hospitals (n = 203410) | P a |
| Age at admission, y | 53 (34–70) | 55 (35–71) | 48 (32–69) | ≤ .001 |
| Non-English primary language | ≤ .001 | |||
| Spanish | 337 371 (58.4) | 198 159 (52.9) | 139 212 (68.4) | |
| Common non-English, non-Spanish languagesb | 38 188 (6.6) | 25 087 (6.7) | 13 101 (6.4) | |
| Rare non-English, non-Spanish languagesc | 202 471 (35.0) | 151 374 (40.4) | 51 097 (25.1) | |
| Sex | ≤ .001 | |||
| Male | 215 752 (37.3) | 145 074 (38.7) | 70 678 (34.8) | |
| Female | 362 278 (62.3) | 229 546 (61.3) | 132 732 (65.3) | |
| Race/ethnicity | ≤ .001 | |||
| Hispanic | 321 023 (55.5) | 189 291 (50.5) | 131 732 (64.8) | |
| Non-Hispanic White | 102 462 (17.7) | 73 436 (19.6) | 29 026 (14.3) | |
| Non-Hispanic Black | 22 833 (4.0) | 16 860 (4.5) | 5 973 (2.9) | |
| Non-Hispanic Asian | 52 035 (9.0) | 35 293 (9.4) | 16 742 (8.2) | |
| Non-Hispanic other/unknownd | 79 677 (13.8) | 59 740 (16.0) | 19 937 (9.8) | |
| Insurance type | ≤ .001 | |||
| Commercial | 253 695 (43.9) | 166 952 (44.6) | 86 743 (42.6) | |
| Medicare | 96 894 (16.8) | 69 968 (18.7) | 26 926 (13.2) | |
| Medicaid | 78 097 (13.5) | 48 052 (12.8) | 30 045 (14.8) | |
| Self-pay | 72 426 (12.5) | 40 406 (10.8) | 32 020 (15.7) | |
| Othere | 76 918 (13.3) | 49 242 (13.1) | 27 676 (13.6) | |
| Admission type | ≤ .001 | |||
| Emergency | 390 444 (67.6) | 263 324 (70.3) | 127 120 (62.5) | |
| Urgent | 58 527 (10.1) | 28 416 (7.6) | 30 111 (14.8) | |
| Elective | 129 059 (22.3) | 82 880 (22.1) | 46 179 (22.7) | |
| Elixhauser comorbidity scoref | ≤ .001 | |||
| 0–1 | 249 922 (43.2) | 151 614 (40.5) | 98 308 (48.3) | |
| 2–3 | 156 848 (27.1) | 105 750 (28.2) | 51 098 (25.1) | |
| ≥ 4 | 171 260 (29.6) | 117 256 (31.3) | 54 004 (26.6) | |
| Principal diagnosis group | ≤ .001 | |||
| Maternal and neonatal | 139 761 (24.2) | 80 846 (21.6) | 58 915 (29.0) | |
| Mental health/substance abuse | 23 539 (4.1) | 15 902 (4.2) | 7 637 (3.8) | |
| Injury | 5 721 (1.0) | 3 458 (0.9) | 2 263 (1.1) | |
| Surgery | 109 127 (18.9) | 71 218 (19.0) | 37 909 (18.6) | |
| Medicine | 299 882 (51.2) | 203 196 (54.2) | 96 686 (47.5) | |
| Neighborhood median household income, quartile | ≤ .001 | |||
| 1 | 112 889 (19.5) | 73 727 (19.7) | 39 162 (19.3) | |
| 2 | 141 250 (24.4) | 96 674 (25.8) | 44 576 (21.9) | |
| 3 | 140 542 (24.3) | 97 143 (25.9) | 43 399 (21.3) | |
| 4 | 183 349 (31.7) | 107 076 (28.6) | 76 273 (37.5) | |
| Urbanicityg | .28 | |||
| Urban | 577 950 (100.0) | 374 563 (100.0) | 203 387 (100.0) | |
| Rural | 80 (0.0) | 57 (0.0) | 23 (0.0) | ≤ .001 |
| Patient-hospital county concordance | ||||
| Yes | 435 047 (75.3) | 283 538 (75.7) | 151 509 (74.5) | |
| No | 142 983 (24.7) | 91 082 (24.3) | 51 901 (25.5) | |
| Discharge year | ≤ .001 | |||
| 2010 | 33 269 (5.8) | 20 759 (5.5) | 12 510 (6.2) | |
| 2011 | 31 250 (5.4) | 19 228 (5.1) | 12 022 (5.9) | |
| 2012 | 31 733 (5.5) | 19 147 (5.1) | 12 586 (6.2) | |
| 2013 | 31 202 (5.4) | 18 928 (5.1) | 12 274 (6.0) | |
| 2014 | 30 665 (5.3) | 18 644 (5.0) | 12 021 (5.9) | |
| 2015 | 34 996 (6.1) | 23 430 (6.3) | 11 566 (5.7) | |
| 2016 | 41 837 (7.2) | 29 289 (7.8) | 12 548 (6.2) | |
| 2017 | 49 172 (8.5) | 34 799 (9.3) | 14 373 (7.1) | |
| 2018 | 55 374 (9.6) | 39 333 (10.5) | 16 041 (7.9) | |
| 2019 | 60 404 (10.5) | 41 901 (11.2) | 18 503 (9.1) | |
| 2020 | 58 970 (10.2) | 37 257 (10.0) | 21 713 (10.7) | |
| 2021 | 60 928 (10.5) | 37 561 (10.0) | 23 367 (11.5) | |
| 2022 | 58 230 (10.1) | 34 344 (9.2) | 23 886 (11.7) | |
| In-hospital complication | ≤ .001 | |||
| Yes | 30 777 (5.3) | 20 479 (5.5) | 10 298 (5.1) | |
| No | 547 253 (94.7) | 354 141 (94.5) | 193 112 (94.9) | |
| Length of stay, d | 3 (2–5) | 3 (2–5) | 3 (2–4) | ≤ .001 |
Note. IQR = interquartile range.
We calculated P values using the Wilcoxon rank–sum test for nonnormal continuous variables and the x2 test for categorical variables.
We considered Arabic, Portuguese, and Korean to be common non-English, non-Spanish languages.
We considered all reported languages other than English, Spanish, Arabic, Portuguese, and Korean to be rare non-English, non-Spanish languages.
The other/unknown category included non-Hispanic patients who identified as American Indian, Alaska Native, multiracial, or other or who declined to answer.
Other insurance includes no charge, Workers Compensation, and other government programs (e.g., from Department of Corrections, Department of Defense, the Indian Health Service or Tribe).
The Elixhauser comorbidity index is a calculated score.
We assessed urbanicity using a 4-category simplified adaptation of the 2013 Urban Influence Codes.
Primary Outcome
At language-serving hospitals, a lower proportion of patients with a non-English primary language were readmitted within 7 days compared with those at non–language-serving hospitals (4.8% vs 9.8%; P ≤ .001; Table 2). In unadjusted models, patients with a non-English primary language at language-serving hospitals were 28% less likely to be readmitted within 7 days (odds ratio [OR] = 0.72; 95% confidence interval [CI] = 0.40, 1.28), compared with those at non–language-serving hospitals, although this result was not statistically significant (Table 3).
TABLE 2—
Proportion of Patients With a Non-English Primary Language Readmitted to the Index Hospital Within 7 Days at Language-Serving vs Non–Language-Serving Hospitals: New Jersey State Inpatient Database, 2010–2022
| Overall, No. (%)a | Language-Serving Hospitals, No. (%) | Non-Language-Serving Hospitals, No. (%) | P | |
|---|---|---|---|---|
| All patients with a non-English primary language | ≤ .001 | |||
| Readmitted | 37 850 (6.6) | 17 956 (4.8) | 19 894 (9.8) | |
| Not readmitted | 540 180 (93.5) | 356 664 (95.2) | 183 516 (90.2) | |
| Spanish is primary language | ≤ .001 | |||
| Readmitted | 24 405 (7.2) | 10 941 (5.5) | 13 464 (9.7) | |
| Not readmitted | 312 966 (92.8) | 187 218 (94.5) | 125 748 (90.3) | |
| Common non-English, non-Spanish primary languageb | ≤ .001 | |||
| Readmitted | 2 926 (7.7) | 1 014 (4.0) | 1 912 (14.6) | |
| Not readmitted | 35 262 (92.3) | 24 073 (96.0) | 11 189 (85.4) | |
| Rare non-English, non-Spanish primary languagec | ≤ .001 | |||
| Readmitted | 10 519 (5.2) | 6 001 (4.0) | 4 518 (8.8) | |
| Not readmitted | 191 952 (94.8) | 145 373 (96.0) | 46 579 (91.2) |
Percentages may not add to exactly 100% because of rounding.
We considered Arabic, Portuguese, and Korean to be common non-English, non-Spanish languages.
We considered all reported languages other than English, Spanish, Arabic, Portuguese, and Korean to be rare non-English, non-Spanish languages.
TABLE 3—
Odds of 7-Day Readmission for Patients With a Non-English Primary Language at Language-Serving and Non–Language-Serving Hospitals: New Jersey State Inpatient Database, 2010–2022
| Language Spoken | Unadjusted OR (95% CI) | AOR (95% CI)a |
|---|---|---|
| All non-English languages | 0.72 (0.40, 1.28) | 0.49 (0.25, 0.96) |
| Spanish | 0.71 (0.37, 1.35) | 0.52 (0.25, 1.08) |
| Common non-English, non-Spanish languagesb | 0.63 (0.31, 1.25) | 0.58 (0.28, 1.18) |
| Rare non-English, non-Spanish languagesc | 0.69 (0.39, 1.23) | 0.67 (0.38, 1.16) |
Note. AOR = adjusted odds ratio; CI = confidence interval; OR = odds ratio. Reference group is non–language-serving hospitals.
Adjusted for age, sex, race/ethnicity, discharge year, insurance type, Elixhauser comorbidity index score, admission urgency, neighborhood median household income, and patient–hospital county concordance.
We considered Arabic, Portuguese, and Korean to be common non-English, non-Spanish languages.
We considered all reported languages other than English, Spanish, Arabic, Portuguese, and Korean to be rare non-English, non-Spanish languages.
After adjustment for age, sex, race/ethnicity, insurance type, discharge year, Elixhauser comorbidity index score, admission urgency, neighborhood median household income, patient–hospital county concordance, and a random effect for hospital, patients with a non-English primary language had 51% lower odds of 7-day readmission at language-serving versus non–language-serving hospitals (adjusted OR [AOR] = 0.49; 95% CI = 0.25, 0.96; Table 3). Adjustment for additional covariates did not substantially change the ORs (Table B, available as a supplement to the online version of this article at http://www.ajph.org). Hospital-level variables did not improve model fit, and we excluded them from the models.
When we stratified by language group (Spanish, common non-English non-Spanish languages, and rare non-English non-Spanish languages; Table 3), we observed the greatest decrease in the odds of 7-day readmission in the Spanish primary language group (AOR = 0.52; 95% CI = 0.25, 1.08), followed by the common non-English, non-Spanish languages group (AOR = 0.58; 95% CI = 0.28, 1.18) and rare non-English, non-Spanish languages (AOR = 0.67; 95% CI = 0.38, 1.16). However, none of these results was statistically significant.
Secondary and Sensitivity Analyses
We also examined the characteristics and outcomes of 4 570 040 patients with English as their primary language at language-serving versus non–language-serving hospitals (Table C, available as a supplement to the online version of this article at http://www.ajph.org). In adjusted models, patients with English as their primary language at language-serving hospitals were less likely to be readmitted, but this difference was not statistically significant (4.6% vs 7.4%; AOR = 0.60; 95% CI = 0.32, 1.10).
When examining all discharge dispositions, mortality did not differ significantly for patients with a non-English primary language at language-serving versus non–language-serving hospitals (2.7% vs 2.4%; AOR = 0.99; 95% CI = 0.85, 1.17). There was also no significant difference for in-hospital complications (8.5% vs 8.2%; AOR = 0.98; 95% CI = 0.84, 1.15) or length of stay (adjusted incidence rate ratio = 0.99; 95% CI = 0.94, 1.04). In the sensitivity analysis examining all discharge dispositions, the odds of 7-day readmission for patients with a non-English primary language were similar to the result observed when limited to those discharged home (all discharge dispositions: AOR = 0.49; 95% CI = 0.27, 0.91; discharged home only: AOR = 0.49; 95% CI = 0.25, 0.96).
DISCUSSION
We evaluated inpatient discharges home from New Jersey hospitals and found that patients with a non-English primary language cared for at language-serving hospitals were 51% less likely to be readmitted within 7 days than those at non–language-serving hospitals (AOR = 0.49; 95% CI = 0.25, 0.96). When stratified by language, the decrease in likelihood of readmission at language-serving hospitals was greatest for patients speaking Spanish. Hospitals with a greater proportion of their patient population composed of patients with a non-English primary language may have practices and resources that reduce readmissions more effectively than hospitals with a lower proportion of patients with a non-English primary language. However, additional research is needed to elucidate the specific factors driving the improvements at language-serving hospitals.
More comprehensive access to interpreter and translation services may play a role in mitigating 7-day readmission rates at language-serving hospitals.12 It is possible that the use of interpreter services increases utilization of services and facilitates a more extensive workup for the patient, relative to what they would otherwise receive without high-quality interpretation.28 Previous research indicates that patients who speak a non-English primary language were more likely to be readmitted than were their counterparts with an English primary language and that this difference is mitigated with confirmed interpreter services use.29,30 We observed that patients with a non-English primary language and those with English as their primary language both experienced fewer 7-day readmissions at language-serving hospitals.
However, the result for the English primary language group was not statistically significant. Language-serving hospitals had a smaller disparity between patients with a non-English primary language and those with English as their primary language (4.8% vs 4.6%). They also showed a greater reduction in 7-day readmission compared with non–language-serving hospitals for patients with a non-English primary language (−5.0 percentage points) than for those with English as their primary language (−2.8 percentage points). Although other hospital initiatives and practices may contribute to the overall lower proportions of patients readmitted at language-serving hospitals, the smaller disparity between patients with a non-English primary language and those with English as their primary language at language-serving hospitals (0.2 percentage points) versus non–language-serving hospitals (2.4 percentage points) indicates that language is playing an important role in both lowering overall readmission and reducing disparities.
Language access resources vary widely by hospital; hospitals may have differences in the size of the interpreter services department, operating hours, and modalities available. The American Hospital Association Annual Survey collects a binary yes or no response to whether the hospital offers linguistic or translation services, and we found no significant differences between language-serving and non–language-serving hospitals. The effectiveness of these resources in each hospital is unknown, and the accessibility of resources, such as easily accessible telephone interpreters,31 has been found to affect readmission. Providing sufficient resources to meet the needs of a hospital’s unique patient population is an important factor in reducing the likelihood of readmission.
Importantly, hospitals do not provide equivalent resources for all non-English languages. In previous research, those who spoke Spanish, Portuguese, or other Asian/Pacific Island languages reported worse experiences than did their English-preferring peers of the same race/ethnicity for care coordination. The results, however, were mixed with regard to discharge and medical information, indicating that the use of language services may vary significantly by language and context.32 This would support our own finding that the context of care at language-serving hospitals may play a role in improving outcomes for patients with a non-English primary language. In our analysis, those with Spanish as their primary language saw the largest decrease in odds of 7-day readmission between language-serving and non–language-serving hospitals. As the Spanish-speaking population of New Jersey is large (16.6%, nearly twice that of any other group with a non-English language in New Jersey),1 language-serving hospitals in New Jersey may be focusing their investments on the most commonly encountered languages of the surrounding population.1 Future research should focus on the factors driving these different patient experiences and how improved linguistic resources (e.g., discharge summaries, coordination between clinicians) or other clinical practices could mitigate these differences.
Strengths and Limitations
We are among the first to examine the role that hospital-level language data plays in patient outcomes, and we used data from a comprehensive database from New Jersey, one of the most linguistically diverse states in the United States. Although our analysis focused on New Jersey, the methodology could be applied to other states by adapting the threshold for language-serving hospitals to that state’s population with limited English proficiency. The language-serving hospital methodology can be used with administrative claims data when researchers are otherwise unable to measure language access resources. However, it is important to note that populations are in flux. Therefore, the exact cutoff should be adapted over time to reflect the population of interest.
Our analysis has several limitations. We used primary language as a proxy for limited English proficiency, but this may not reflect true language proficiency. Hospitals report the primary language data to New Jersey, and there may be misclassification of primary language, especially for those we classified as “other” primary language. Furthermore, as this analysis focused on all patients, the study population was heterogeneous with respect to admission diagnosis; however, adjusting for principal diagnosis group did not influence our results. Additionally, interpreter services are a hospital-level resource, and availability should not be affected by diagnostic group or service. Our primary outcome had a high level of missingness, and the reasons for readmission were not reported. Finally, we were limited to 7-day readmissions at the index hospital, as we could not track patients within or across states and years.
Public Health Implications
We evaluated hospital readmission rates and found that these varied between language-serving and non–language-serving hospitals. The approximately 4 million adult hospital readmissions that occur annually expose patients to a risk of additional in-hospital complications or infections, stress, or lost days of work,33,34 and both patients and hospitals experience increased financial burdens.31 Readmissions cost an average of $16 300 and, in New Jersey, may result in denied payment claims.33,35 The use of language-concordant care and investment in robust interpreter services mitigates the drivers of preventable readmissions, including discharge summary errors, insufficient follow-up care, and lack of patient education; it may ultimately provide cost savings, relative to the costs incurred by preventable readmissions for patients with a non-English primary language.31
Future research should seek to understand reasons for readmissions and how interpreter services influence utilization and readmissions in language-serving and non–language-serving hospitals. By understanding the hospital’s role in language-concordant care and patient outcomes, we can influence policy changes at the hospital, state, and federal levels to reduce readmission rates for patients with a non-English primary language.
Supplementary Material
ACKNOWLEDGMENTS
This work was supported by the National Institute on Minority Health and Health Disparities, National Institutes of Health (NIH; to G. O.; award K23MD016129).
The New Jersey Department of Health provided New Jersey state data to the Agency for Healthcare Research and Quality and the Healthcare Cost and Utilization Project.
Note. The contents of this article are solely the responsibility of the authors and do not necessarily represent the official views of the NIH.
Footnotes
CONFLICTS OF INTEREST
The authors declare that they have no affiliations with or involvement in any organization or entity with any financial interest in the subject matter or materials discussed in this article.
HUMAN PARTICIPANT PROTECTION
This study was deemed exempt by the Mass General Brigham institutional review board (IRB no. 2019P001325).
Contributor Information
Alyssa N. Jones, Center for Surgery and Public Health, Department of Surgery, Mass General Brigham, Boston, MA..
Brittany M. Dacier, Center for Surgery and Public Health, Department of Surgery, Mass General Brigham, Boston, MA..
Jesus Mejia, Tufts School of Medicine, Boston..
Benjamin G. Allar, Center for Surgery and Public Health, Department of Surgery, Mass General Brigham, Boston, MA..
Rachel E. Murphy, Center for Surgery and Public Health, Department of Surgery, Mass General Brigham, Boston, MA..
Regan W. Bergmark, Center for Surgery and Public Health, Department of Surgery, Mass General Brigham, Boston, MA..
Esteban F. Gershanik, Department of Medicine, Mass General Brigham..
Alicia Fernandez, Department of Medicine, University of California, San Francisco..
Andrea L. Pusic, Patient-Reported Outcomes, Value & Experience Center, Department of Surgery, Mass General Brigham..
Gregory L. Peck, Department of Surgery, Rutgers Robert Wood Johnson Medical School, New Brunswick, NJ..
David W. Bates, Department of Medicine, Mass General Brigham..
Gezzer Ortega, Center for Surgery and Public Health, Department of Surgery, Mass General Brigham, Boston, MA..
REFERENCES
- 1.US Census Bureau. Language spoken at home: S1601. American Community Survey. 2022: ACS 5-year estimates subject tables. Available at: https://data.census.gov/table/ACSST5Y2022.S1601?q=S1601%20new%20jersey. Accessed December 20, 2024. [Google Scholar]
- 2.US Department of Justice. Commonly asked questions and answers regarding limited English proficient (LEP) individuals. July 17, 2025. Available at: https://LEP.gov. Accessed April 28, 2026.
- 3.US Census Bureau. Language spoken at home: S1601. American Community Survey. 2023: ACS 5-year estimates subject tables. Available at: https://data.census.gov/table/ACSST5Y2023.S1601?q=Language+Spoken+at+Home. Accessed April 21, 2025. [Google Scholar]
- 4.Twersky SE, Jefferson R, Garcia-Ortiz L, Williams E, Pina C. The impact of limited English proficiency on healthcare access and outcomes in the US: a scoping review. Healthcare (Basel). 2024;12(3): 364. 10.3390/healthcare12030364 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Joo H, Fernández A, Wick EC, Moreno Lepe G, Manuel SP. Association of language barriers with perioperative and surgical outcomes: a systematic review. JAMA Netw Open. 2023;6(7):e2322743. 10.1001/jamanetworkopen.2023.22743 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Maurer LR, Eruchalu CN, Gaitanidis A, et al. Trauma patients with limited English proficiency: outcomes from two level one trauma centers. Am J Surg. 2023;225(4):769–774. 10.1016/j.amjsurg.2022.10.043 [DOI] [PubMed] [Google Scholar]
- 7.Lindholm M, Hargraves JL, Ferguson WJ, Reed G. Professional language interpretation and inpatient length of stay and readmission rates. J Gen Intern Med. 2012;27(10):1294–1299. 10.1007/s11606-012-2041-5 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Diamond L, Izquierdo K, Canfield D, Matsoukas K, Gany F. A systematic review of the impact of patient–physician non-English language concordance on quality of care and outcomes. J Gen Intern Med. 2019;34(8):1591–1606. 10.1007/s11606-019-04847-5 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Khoong EC, Fernandez A. Addressing gaps in interpreter use: time for implementation science informed multi-level interventions. J Gen Intern Med. 2021;36(11):3532–3536. 10.1007/s11606-021-06823-4 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Diamond LC, Wilson-Stronks A, Jacobs EA. Do hospitals measure up to the national culturally and linguistically appropriate services standards? Med Care. 2010;48(12):1080–1087. 10.1097/MLR.0b013e3181f380bc [DOI] [PubMed] [Google Scholar]
- 11.National Archives. Nondiscrimination in health programs or activities; 42 USC 18116. April 27, 2026. Available at: https://www.ecfr.gov/current/title-45/subtitle-A/subchapter-A/part-92. Accessed August 6, 2025.
- 12.Graham KL, Auerbach AD, Schnipper JL, et al. Preventability of early vs late hospital readmissions in a national cohort of general medicine patients. Ann Intern Med. 2018;168(11):766–774. 10.7326/M17-1724 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Dacier BM, Jones AN, Allar BG, et al. Language-serving hospitals and their impact on readmission for surgical patients with a non-English primary language: a retrospective cohort study. Ann Surg. 2025. 10.1097/SLA.0000000000006707 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Concha JB, Umucu E, Duarte-Gardea M, et al. Hispanic-serving institutions contributing to the training and diversity of the public health and health care workforce. Am J Public Health. 2024;114(suppl 6):S472–S477. 10.2105/AJPH.2024.307655 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Chen YW, Kim TD, Molina RL, Chang DC, Oseni TO. Minority-serving hospitals are associated with low within-hospital disparity. Am Surg. 2024;90(4):567–574. 10.1177/00031348231175117 [DOI] [PubMed] [Google Scholar]
- 16.Dhar VK, Kim Y, Wima K, Hoehn RS, Shah SA. The importance of safety-net hospitals in emergency general surgery. J Gastrointest Surg. 2018;22(12): 2064–2071. 10.1007/s11605-018-3885-8 [DOI] [PubMed] [Google Scholar]
- 17.Schiaffino MK, Nara A, Mao L. Language services in hospitals vary by ownership and location. Health Aff (Millwood). 2016;35(8):1399–1403. 10.1377/hlthaff.2015.0955 [DOI] [PubMed] [Google Scholar]
- 18.Wu S, Ridgely MS, Escarce JJ, Morales LS. Language access services for Latinos with limited English proficiency: lessons learned from Hablamos Juntos. J Gen Intern Med. 2007;22(suppl 2): 350–355. 10.1007/s11606-007-0323-0 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Nandy B Hospital investment decisions and prioritization of clinical programs. Cureus. 2025; 17(3):e79998. 10.7759/cureus.79998 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Agency for Healthcare Research and Quality. Healthcare Cost & Utilization Project State Inpatient Databases (SID). April 2026. Available at: http://www.hcup-us.ahrq.gov/sidoverview.jsp. Accessed April 28, 2026.
- 21.Dietrich S, Hernandez E. Language use in the United States: 2019. August 2022. Available at: https://www.census.gov/content/dam/Census/library/publications/2022/acs/acs-50.pdf. Accessed August 6, 2025.
- 22.American Hospital Association. AHA annual survey database. 2026. Available at: https://www.ahadata.com/aha-annual-survey-database. Accessed April 28, 2026.
- 23.Agency for Healthcare Research and Quality. Healthcare Cost & Utilization Project. State Ambulatory Surgery and Services Databases (SASD). April 2026. Available at: https://hcup-us.ahrq.gov/sasdoverview.jsp. Accessed April 28, 2026. [Google Scholar]
- 24.Agency for Healthcare Research and Quality. Healthcare Cost & Utilization Project. State Emergency Department Databases (SEDD). April 2026. Available at: https://hcup-us.ahrq.gov/db/state/sedddbdocumentation.jsp. Accessed April 28, 2026. [Google Scholar]
- 25.Ottenbacher KJ, Karmarkar A, Graham JE, et al. Thirty-day hospital readmission following discharge from postacute rehabilitation in fee-for-service Medicare patients. JAMA. 2014;311(6): 604–614. 10.1001/jama.2014.8 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Owusu E, Oluwasina F, Nkire N, Lawal MA, Agyapong VIO. Readmission of patients to acute psychiatric hospitals: influential factors and interventions to reduce psychiatric readmission rates. Healthcare (Basel). 2022;10(9):1808. 10.3390/healthcare10091808 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Quan H, Sundararajan V, Halfon P, et al. Coding algorithms for defining comorbidities in ICD-9-CM and ICD-10 administrative data. Med Care. 2005; 43(11):1130–1139. 10.1097/01.mlr.0000182534.19832.83 [DOI] [PubMed] [Google Scholar]
- 28.Karliner LS, Jacobs EA, Chen AH, Mutha S. Do professional interpreters improve clinical care for patients with limited English proficiency? A systematic review of the literature. Health Serv Res. 2007;42(2):727–754. 10.1111/j.1475-6773.2006.00629.x [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Chu JN, Wong J, Bardach NS, et al. Association between language discordance and unplanned hospital readmissions or emergency department revisits: a systematic review and meta-analysis. BMJ Qual Saf. 2024;33(7):456–469. 10.1136/bmjqs-2023-016295 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Squires A, Ma C, Miner S, Feldman P, Jacobs EA, Jones SA. Assessing the influence of patient language preference on 30 day hospital readmission risk from home health care: a retrospective analysis. Int J Nurs Stud. 2022;125:104093. 10.1016/j.ijnurstu.2021.104093 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Karliner LS, Pérez-Stable EJ, Gregorich SE. Convenient access to professional interpreters in the hospital decreases readmission rates and estimated hospital expenditures for patients with limited English proficiency. Med Care. 2017;55(3): 199–206. 10.1097/MLR.0000000000000643 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Quigley DD, Elliott MN, Hambarsoomian K, et al. Inpatient care experiences differ by preferred language within racial/ethnic groups. Health Serv Res. 2019;54(suppl 1):263–274. 10.1111/1475-6773.13105 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Jiang HJ, Hensche M. Characteristics of 30-day all-cause hospital readmissions, 2016–2020. September 2023. Available at: https://hcup-us.ahrq.gov/reports/statbriefs/sb304-readmissions-2016-2020.pdf. Accessed April 28, 2026.
- 34.Beauvais B, Whitaker Z, Kim F, Anderson B. Is the hospital value-based purchasing program associated with reduced hospital readmissions? J Multi-discip Healthc. 2022;15:1089–1099. 10.2147/JMDH.S358733 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35.N.J. Admin. Code § 10:52–14.16. Payment for readmissions. 2016. Available at: https://www.law.cornell.edu/regulations/new-jersey/N-J-A-C-10-52-14-16. Accessed August 1, 2025.
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.