Association of structural factors with provision of non-English language services in substance use disorder treatment facilities in the USA: a cross-sectional study

Abstract

Objectives

The substance use disorder (SUD) crisis continues to be a public health challenge in the USA, disproportionately affecting culturally and linguistically diverse (CALD) populations. Understanding how structural factors contribute to providing linguistically competent SUD treatment can guide efforts to incorporate treatment provisions tailored to the specific needs of diverse CALD communities. This study examines the association of structural factors, namely ownership type and receipt of government funding, with facility-level provision of SUD treatment in non-English languages in the USA.

Design

A cross-sectional study using deidentified facility-level data from the 2019 National Survey of Substance Abuse Treatment Services (N-SSATS) dataset.

Setting

SUD treatment facilities from all 50 US states and the District of Columbia.

Participants

Of 15 961 facilities in the 2019 N-SSATS dataset, 14 993 were selected for inclusion in our study.

Main outcomes and measures

The main outcome was a binary indicator of the provision of SUD treatment services in languages other than English. To assess the association between structural factors and the provision of SUD treatment services in non-English languages, we ran generalised linear mixed models, controlling for state- and facility-level variables, to estimate adjusted odds ratios (aORs).

Results

Of 14 993 SUD treatment facilities nationally, 7864 (52.45%) offered non-English language SUD treatment services. Private for-profit facilities had significantly lower odds of offering non-English language services than private non-profit facilities (aOR=0.59; 95% CI 0.54 to 0.65). Facilities that received government funding had significantly higher adjusted odds of providing non-English language services compared to those that did not (aOR=1.45; 95% CI 1.32 to 1.59).

Conclusions and relevance

Structural factors, including ownership type and government funding, were associated with the provision of SUD treatment services in non-English languages. Our findings highlight the potential role of these structural factors in shaping the capacity to provide services in non-English languages in SUD care settings. Future research is needed to better understand how changes in funding mechanisms or ownership type may facilitate the provision of non-English-language treatment services across SUD facilities.

STRENGTHS AND LIMITATIONS OF THIS STUDY.

• By linking national-level substance use disorder (SUD) treatment facility data with state-level demographic and policy indicators, this study provides a comprehensive perspective on the structural and contextual factors shaping the provision of non-English language services in SUD care in the USA.

• Examining the provision of SUD treatment services—particularly those offered in languages other than English—is especially timely, as limited English-speaking communities may potentially face heightened risks of communication barriers in healthcare settings following the designation of English as the official language of the US.

• The study’s cross-sectional design limits causal inference.

Introduction

Substance use disorders (SUDs) continue to have a profound impact on public health in the USA, with an estimated 48.5 million people nationally diagnosed with one in 2023.¹ Although SUDs are treatable, few people receive adequate treatment.² In 2023, of the 48.5 million people aged 12 and older requiring SUD treatment, only 14.6% received treatment.¹ Untreated SUDs can lead to adverse consequences for individuals, families and communities.³ Among those who do not receive treatment, nearly 12 million (28.7%) report the reason for non-treatment is that they, ‘could not find a treatment program or healthcare professional they wanted to go to’.¹ The inability to locate a healthcare provider that aligns with one’s preferences and needs reflects a gap in patient-centered SUD care.

Patient-centered treatment designed to incorporate the unique needs, values and preferences of a patient is essential to improve SUD treatment utilisation.^{4 5} However, implementation of patient-centered SUD care is often constrained by barriers at the provider and system levels.⁶ These barriers include, but are not limited to, social stigma, discrimination, limited availability or affordability of services and limited availability of culturally and linguistically appropriate treatment programmes and providers.7,9

Many of these barriers disproportionately affect SUD treatment accessibility and experiences of culturally and linguistically diverse (CALD) individuals. CALD populations include individuals from diverse racial, ethnic and cultural backgrounds, defined as those born in non-English speaking countries or who do not speak English at home.¹⁰ Linguistic or cultural disconnects can hinder access to suitable treatment services, cause treatment delays, lead to miscommunication and often result in dissatisfaction with the healthcare provider.^{11 12} Further, CALD individuals are less likely to initiate and maintain treatment and experience worse outcomes such as higher rates of morbidity, mortality and adverse social and legal consequences compared with non-CALD peers.13,17

To address these disparities and meet the unique needs of CALD individuals, provision of culturally and linguistically competent treatment that respects patients’ beliefs, preferred languages and communication needs is essential.¹⁸ A widely accepted definition of cultural competence in clinical practice is ‘a set of congruent behaviors, attitudes, and policies that come together in a system, agency, or among professionals and enable the system, agency, or professionals to work effectively in cross-cultural situations’.¹⁹ Different interpretations of this definition have been adopted, and some have attempted to expand, elaborate or refine the understanding of cultural competence.20,22 In this paper, we focus on the provision of treatment services in non-English languages as a marker of linguistically competent SUD care,^{23 24} a core dimension of patient-centered care,²⁵ especially among CALD communities.

There has been limited research examining the factors associated with the current landscape of SUD treatment provided in languages other than English for CALD individuals. Structural factors, such as facility ownership type, organisational structure of treatment facilities, workforce diversity and receipt of funding, play a critical role.26,28 Resource dependency theory helps explain how external factors may influence the behaviour and structure of organisations, including healthcare facilities.^{29 30} From a resource dependency theoretical lens, previous research has shown that ownership type and availability of government funding can influence a facility’s capacity to provide SUD services in non-English languages.^{28 31} However, these studies have been limited in generalisability as they have focused on specific geographic locations³² or primarily examined the provision of SUD care only in the Spanish language.²⁸ Moreover, previous research has not accounted for a wide range of state-level factors that may also influence SUD facility behaviour, including but not limited to the state’s cultural competency training requirements, adoption of Medicaid³³ expansion and SUD prevalence.³⁴

In our research, we examine the association between the provision of SUD care in all non-English languages, including but not limited to Spanish, and structural factors, particularly facility ownership type and receipt of government funding. Investigating this association sheds light on the landscape of provision of non-English language services necessary to respond to the needs of the increasingly diverse SUD population.

Materials and methods

This study follows the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guidelines for cross-sectional studies. The University of Maryland, Baltimore Institutional Review Board determined that this study was exempt from review.

Data sources

This cross-sectional study was conducted using data from multiple publicly available sources.35,43 The primary data source was the 2019 National Survey of Substance Abuse Treatment Services (N-SSATS), an annual census-based survey conducted by the Substance Abuse and Mental Health Services Administration (SAMHSA).³⁵ This dataset includes information on alcohol and drug abuse treatment facilities and services across all 50 US states, the District of Columbia and other US territories. We used N-SSATS data to identify treatment facilities that reported offering services in a language other than English, their location, accepted forms of payment, ownership type, receipt of government funding and levels of care.

We used the state variable in the N-SSATS data to link each facility to seven additional publicly available data sources. (1) The Office of Minority Health sponsored Think Cultural Health’s Tracking Culturally and Linguistically Appropriate Services online tool was used to identify state-level adoption of cultural competency training in 2019.³⁶ (2) SAMHSA’s National Survey on Drug Use and Health State Estimates tables from 2018 and 2019 were used to obtain percentages of SUD among adults (aged ≥18 years) by state during the 2018–2019 year.³⁷ (3) The 2019 American Community Survey (ACS) was used to obtain the percentage estimates of households with limited English-speaking ability and diversity index by state.^{38 39} (4) The State Health Access Data Assistance Center map was used to capture the state-level Gini coefficients for 2019.⁴⁰ (5) The political orientation of state legislatures was determined using data from the National Conference of State Legislatures website.⁴¹ (6) The status of Medicaid expansion for each state in 2019 was determined using the Kaiser Family Foundation interactive map.⁴² (7) Lastly, the state-rurality percentage in 2023—the closest year with available data—was derived using America’s Health Rankings data.⁴³ The values for all the state-level variables extracted from various publicly available sources are provided in online supplemental table 1.

Measures

Dependent variable: The outcome variable of interest in our study was a binary indicator of whether an SUD treatment facility offered services in a language other than English, sourced from N-SSATS using the ‘LANG’ variable.

Primary independent variables: The primary structural predictor variables were facility ownership type derived from the ‘OWNERSHP’ variable in the N-SSATS data. We created a categorical indicator representing facility ownership type: (i) private for-profit; (ii) private non-profit; and (iii) public. Facilities operated by state, local, tribal, community or federal governments were combined to form the ‘public’ category. Facility receipt of government funds was identified using the ‘EARMARK’ variable from the N-SSATS data, indicating whether facilities receive funding or grants from federal, state, county or local governments to support their substance use treatment programmes, and was operationalised as a binary indicator.

Covariates: The selection of covariates for inclusion in our study was informed by a review of prior literature^{28 33 44} and reviewer feedback. Facility-level characteristics included the accepted forms of payment at the facility, categorised into six binary indicators: cash, self-pay, Medicare, Medicaid, state-financed other than Medicaid, military insurance and private insurance. Similarly, levels of care comprised three categories: (i) hospital inpatient; (ii) non-hospital residential; and (iii) outpatient services, all of which were coded as binary indicators as obtained from N-SSATS data. Facilities missing data on any facility-level variables from N-SSATS were excluded.

We created nine state-level variables. (1) Cultural competency training requirement was categorised as follows: (i) requires mandatory cultural competency training; (ii) proposed but does not have mandatory cultural competency training; or (iii) does not require cultural competency training. (2) Tertiles of the proportion of limited English-speaking households was defined as low (0.3–1.6), medium (1.7–2.6) or high (2.8–8.6). We used tertiles to classify the proportion of limited-English speaking households, as it allowed us to capture variation across states with differing levels of linguistic distribution while having adequate sample sizes per group to enable analysis. (3) State-level SUD prevalence was categorised as above or below the national average of 7.74%.³⁷ (4) State rurality was categorised as having low or high rurality using the national average of 20.4% as the cut-off.⁴³ (5) Income inequality was measured using the Gini coefficient and categorised as low inequality (<0.48) or high inequality (≥0.48), using the national average of 0.48 as the cut-off.⁴⁰ For variables such as state SUD prevalence, rurality and the Gini coefficient, we used the national average as a cut-off for classification so as to have a policy-relevant reference to aid interpretation of our study findings. (6) The diversity index was categorised into five groups using fixed percentage cut points to reflect incremental variations in racial and ethnic diversity. (7) The US census region of the state was categorised into Northeast, Midwest, South or West. (8) Political orientation was coded as Democratic or Republican based on the political affiliation of the state’s legislature during the study period (2019).⁴¹ For the District of Columbia, the Mayor’s political party of 2019 was used; given the divided legislature in Minnesota, the Governor’s party (Democratic) was used; given the unicameral legislature in Nebraska, the Governor’s political party (Republican) was used to represent the state’s political orientation. Finally, (9) Medicaid expansion status was coded as a binary variable indicating if the state had implemented Medicaid expansion as of 2019.

Statistical analysis

We generated descriptive statistics to characterise the facilities providing SUD treatment in our sample, overall and stratified by facilities offering services in non-English languages. We used Pearson’s χ² test or the Wilcoxon rank-sum test to examine the statistical significance of differences in distribution between facilities that offered non-English language services and those that did not. We performed correlation analysis to avoid multicollinearity and removed one variable from each pair of highly correlated predictors (correlation ≥0.80).⁴⁵ For example, the diversity index and the tertiles of the proportion of limited English-speaking households were highly correlated (r=0.89). Thus, in the final model, we retained the variable for the tertiles of limited English-speaking households because it better captures the linguistic diversity relevant to our study outcome. Multicollinearity among predictors was further assessed using variance inflation factors (VIFs), and all values were well below the commonly used threshold (VIF<5).

Next, we used a generalised linear mixed model (GLMM) with a logit link to model the association of facilities offering services in non-English languages with ownership type and receipt of government funding, controlling for covariates. We chose a GLMM including state as a random intercept to account for the correlation among facilities within the same state, based on the assumption that facilities within a state may be more similar to each other than those located in other states.⁴⁶ Our approach assumes that the unobserved state-level factors captured by the random intercept are not correlated with facility-level predictors. Additionally, GLMM allows efficient estimation of facility-level effects using a parsimonious model, avoiding the need to estimate numerous state-level parameters generated by a state fixed-effects model.⁴⁷ For the crude analysis, a separate bivariate regression model was run to estimate the association between each predictor and the outcome. The multivariable model included the structural predictors and all the covariates that were retained after accounting for high correlations. We report the crude and adjusted odds ratios (aORs) and precision around the ORs using a two-sided 95% CI. All the analyses were performed using SAS V.9.4 (SAS Institute, Cary, North Carolina, USA). Because N-SSATS is a census survey of SUD treatment facilities, and the public-use dataset does not include survey weights, analyses were conducted using unweighted facility-level observations. Additionally, to help visualise data demonstrating the need (limited English-speaking households) and availability (number of SUD treatment facilities providing treatment in non-English languages) at the state level, a map representing rates of facilities offering services in a language other than English per 1000 limited English-speaking households displayed as quintiles was created using the Datawrapper visualisation tool⁴⁸ (figure 1). The state specific rates of facilities offering services in non-English languages per 1000 limited English-speaking households are provided in online supplemental table 2. In addition, in online supplemental table 3, we provide a frequency distribution of non-English languages spoken by the staff counsellors providing care at the SUD treatment facilities, offering context on the workforce’s linguistic capacity.

Figure 1. Rates of SUD treatment facilities offering services in non-English languages per 1000 limited English-speaking households in the USA, 2019. The rates of SUD treatment facilities offering services in non-English languages per 1000 limited English-speaking households categorised into quintiles (range min−max): 0 (0.39–1.63), 1 (1.65–2.37), 2 (2.46–3.92), 3 (4.12–6.18) and 4 (6.42–24.17). SUD, substance use disorder.

Patient and public involvement

None.

Results

Characteristics of all SUD treatment facilities

Of the 15 961 facilities in the 2019 N-SSATS dataset, 14 993 were included in this study. These facilities provide SUD treatment across the 50 states and the District of Columbia and had complete information on key study variables (figure 2). About 681 (4.27%) facilities were excluded due to missingness. We examined whether missingness differed by the dependent variable and found that the prevalence of facilities offering non-English language services was similar between the facilities included (52.45%) in our study sample and those excluded (50.68%). Regarding ownership, 7463 (49.78%) facilities were private non-profit, 6087 (40.60%) were private for-profit and 8026 (53.53%) received government funding. Over 90% (n=13 720; 91.51%) of facilities accepted cash or self-payment for treatment. Less than two-fifths of the facilities accepted Medicare (5516; 36.79%), whereas a higher proportion accepted Medicaid (10 242; 68.31%) or private insurance (11 000; 74.03%) for payment. Most facilities were located in states that had adopted Medicaid expansion (10 742; 71.65%) (table 1).

Figure 2. Treatment facilities selected for analysis. SUD, substance use disorder.

Table 1. Characteristics of substance use disorder treatment facilities in the USA, 2019 (N-SSATS), n=14 993.

Characteristics	All facilities n=14 993	Offers services in non-English languages n=7864; 52.45%	Does not offer services in non-English languages n=7129; 47.55%	P value*
Facility ownership
Private non-profit	7463 (49.78)	4588 (58.34)	2875 (40.33)	<0.01
Private for-profit	6087 (40.60)	2368 (30.11)	3719 (52.17)
Public†	1443 (9.62)	908 (11.55)	535 (7.5)
Government funding
Yes	8026 (53.53)	5000 (63.58)	3026 (42.45)	<0.01
Accepted forms of payment (yes)
Cash or self-payment	13 720 (91.51)	7287 (92.66)	6433 (90.24)	<0.01
Medicare	5516 (36.79)	3700 (47.05)	1816 (25.47)	<0.01
Medicaid	10 242 (68.31)	6272 (79.76)	3970 (55.69)	<0.01
Federal military insurance	5664 (37.78)	3507 (44.60)	2157 (30.26)	<0.01
Private health insurance	11 000 (74.03)	6059 (77.05)	5041 (70.71)	<0.01
Other state-financed insurance	7315 (48.79)	4496 (57.17)	2819 (39.54)	<0.01
Level of care (yes)
Hospital inpatient	735 (4.9)	478 (6.08)	257 (3.60)	<0.01
Non-hospital residential	3484 (23.24)	1426 (18.13)	2058 (28.87)	<0.01
Outpatient	12 452 (83.05)	6816 (86.67)	5636 (79.06)	<0.01
State-level cultural competency training requirement
Mandatory	4693 (31.30)	2584 (32.86)	2109 (29.58)	<0.01
Proposed, not mandatory	4476 (29.85)	2623 (33.35)	1853 (25.99)
Not required	5824 (38.84)	2657 (33.79)	3167 (44.42)
Tertile of limited English-speaking households in the state‡
Low (0.30%–1.60%)	3145 (20.98)	1725 (21.94)	1420 (19.92)	<0.01
Medium (1.70%–2.60%)	4429 (29.54)	1989 (25.29)	2440 (34.23)
High (2.80%–8.60%)	7419 (49.48)	4150 (52.77)	3269 (45.85)
Prevalence of substance use disorder in state
Below or equal to national average§	7649 (51.02)	3911 (49.73)	3738 (52.43)	<0.01
Above national average	7344 (48.98)	3953 (50.27)	3391 (47.57)
US census region
Northeast	2720 (18.14)	1499 (19.06)	1221 (17.13)	0.02
Midwest	3620 (24.14)	1883 (23.94)	1737 (24.37)
South	4434 (29.57)	2285 (29.06)	2149 (30.14)
West	4219 (28.14)	2197 (27.94)	2022 (28.36)
Gini coefficient¶
Low inequality (<0.48)	8233 (54.91)	4090 (52.01)	4143 (58.11)	<0.01
High inequality (≥0.48)	6760 (45.09)	3774 (47.99)	2986 (41.89)
Diversity index
<35	1237 (8.25)	571 (7.26)	666 (9.34)	0.01
35–44.9	2967 (19.79)	1463 (18.6)	1504 (21.10)
45–54.9	2403 (16.03)	1390 (17.68)	1013 (14.21)
55–64.9	4317 (28.79)	2338 (29.73)	1979 (27.76)
≥65	4069 (27.14)	2102 (26.73)	1967 (27.59)
Medicaid expansion status of state
Expanded	10 742 (71.65)	5588 (71.06)	5154 (72.30)	0.09
Non-expanded	4251 (28.35)	2276 (28.94)	1975 (27.70)
Political orientation of state
Democratic	6903 (46.04)	3612 (45.93)	3291 (46.16)	0.77
Republican	8090 (53.96)	4252 (54.07)	3838 (53.84)
Rurality**
Low rurality (<20.4)	7967 (53.14)	4380 (55.7)	3587 (50.32)	<0.01
High rurality (≥20.4)	7026 (46.86)	3484 (44.30)	3542 (49.68)

^*Pearson’s χ2 test; Wilcoxon rank sum test, p<0.05 is the cut-off value established for significance.

^†Publicly owned facility can be operated by either state/local/tribal/community/federal governments.

^‡The tertiles were created based on the 2019 sample distribution of limited English-speaking households and categorised the states into low (0.3–1.6), medium (1.7–2.6) and high (2.8–8.6) groups.

^§The national average percentage of individuals with substance use disorder was 7.74 in 2019.

^¶The Gini coefficient of the USA in 2019 was 0.4843 (min: 0.43; max: 0.51).

^**In 2023, 20.4% of the US population lived in rural areas.

N-SSATS, National Survey of Substance Abuse Treatment Services.

Characteristics of facilities providing SUD treatment services in non-English languages

Of the 14 993 facilities in our sample, 7864 (52.45%) offered non-English language SUD treatment services. Most of the facilities offering non-English language SUD treatment services were predominantly private non-profit (4588; 58.34%), and most received some form of government funding (5000; 63.58%). Many of the facilities providing non-English language services, totaling 5588 (71.06%), were located in Medicaid expansion states. A little over half (52.77%) of these facilities were also in states in the highest tertile (2.80%–8.60%) of limited English-speaking households (table 1).

Association between structural factors, facility characteristics and provision of non-English language treatment services

After adjusting for facility-level and state-level characteristics, we found a significant association between structural factors and the provision of non-English language services. Specifically, we found that private for-profit facilities had significantly lower adjusted odds (aOR=0.59; 95% CI 0.54 to 0.65) of offering non-English language services compared with private non-profit facilities. Facilities that received government funding had significantly higher adjusted odds (aOR=1.45; 95% CI 1.32 to 1.59) of providing non-English language services than those that did not. Facilities accepting Medicaid (aOR=2.21; 95% CI 2.00 to 2.44) and Medicare (aOR=1.48; 95% CI 1.35 to 1.62) as sources of payment had significantly higher adjusted odds of providing non-English language services than those that did not accept Medicaid and Medicare, respectively. Conversely, facilities accepting private insurance had lower odds (aOR=0.80; 95% CI 0.73 to 0.89) of offering non-English language services relative to those that did not. Facilities in states with mandatory (aOR=2.08; 95% CI 1.37 to 3.16) or proposed (aOR=1.81; 95% CI 1.30 to 2.53) cultural competency training requirements were more likely to provide non-English language services than those in states without such requirements. Notably, facilities in Medicaid non-expansion states had higher odds (aOR=1.62; 95% CI 1.22 to 2.13) of providing non-English language services than those in expansion states (table 2).

Table 2. Association between facility and state factors and provision of substance use disorder treatment services in non-English languages.

Characteristics*	Unadjusted OR (95% CI)	Adjusted OR (95% CI)
Facility ownership
Private non-profit	1 (Reference)	1 (Reference)
Private for-profit	0.41 (0.38 to 0.44)	0.59 (0.54 to 0.65)
Public	1.16 (1.02 to 1.31)	1.10 (0.96 to 1.26)
Government funding
Yes	2.38 (2.22 to 2.55)	1.45 (1.32 to 1.59)
Sources of payment (yes)
Cash or self-payment	1.40 (1.24 to 1.59)	1.64 (1.42 to 1.89)
Medicare	2.64 (2.45 to 2.85)	1.48 (1.35 to 1.62)
Medicaid	3.86 (3.55 to 4.20)	2.21 (2.00 to 2.44)
Federal military insurance	1.89 (1.76 to 2.03)	1.28 (1.17 to 1.41)
Private health insurance	1.39 (1.28 to 1.50)	0.80 (0.73 to 0.89)
Other state-financed insurance	2.23 (2.08 to 2.40)	1.16 (1.06 to 1.27)
Level of care
Hospital inpatient	1.62 (1.38 to 1.91)	1.99 (1.66 to 2.38)
Outpatient	1.86 (1.69 to 2.04)	1.82 (1.64 to 2.03)
State-level cultural competency training requirement
Not required	1 (Reference)	1 (Reference)
Mandatory	1.59 (1.16 to 2.17)	2.08 (1.37 to 3.16)
Proposed, not mandatory	1.50 (1.11 to 2.02)	1.81 (1.30 to 2.53)
Tertile of limited English-speaking households in the state
Low (0.30%–1.60%)	1 (Reference)	1 (Reference)
Medium (1.70%–2.60%)	0.77 (0.57 to 1.05)	0.83 (0.63 to 1.09)
High (2.80%–8.60%)	1.24 (0.91 to 1.69)	1.02 (0.70 to 1.48)
Prevalence of substance use disorder in state
Below or equal to national average	1 (Reference)	1 (Reference)
Above national average	1.06 (0.81 to 1.39)	1.12 (0.88 to 1.43)
US census region
West	1 (Reference)	1 (Reference)
Northeast	1.20 (0.78 to 1.83)	1.01 (0.69 to 1.46)
Midwest	0.88 (0.60 to 1.30)	0.72 (0.51 to 1.01)
South	0.94 (0.66 to 1.35)	1.12 (0.78 to 1.60)
Gini coefficient
Low inequality (<0.48)	1 (Reference)	1 (Reference)
High inequality (≥0.48)	1.31 (0.98 to 1.74)	1.13 (0.85 to 1.51)
Medicaid Expansion status of state
Expanded	1 (Reference)	1 (Reference)
Non-expanded	1.08 (0.81 to 1.44)	1.62 (1.22 to 2.13)

Note: The reference group is ‘No’ where not specified.

^*All variables reported in this table were included in the final model.

Geographic variation in the availability of non-English SUD treatment services for CALD individuals

We observed that the rates of SUD treatment facilities providing non-English language services per 1000 limited English-speaking households are heterogeneous across states, with rates ranging from 0.39 to 24.17 per 1000 limited English-speaking households (figure 1).

Sensitivity and post hoc analyses

The GLMM with state as a random intercept is based on the assumption that state-level unobserved factors are independent of the included facility-level covariates.⁴⁷ To examine the robustness of our findings to this assumption, we also estimated a model with state fixed effects. The results from the fixed-effects model produced substantively similar results to those from the GLMM with state as a random intercept, with no meaningful changes in the magnitude or statistical significance of the associations between facility ownership, government funding and the provision of non-English language services (online supplemental table 4). Additionally, to assess whether the categorisation of certain state-level variables influenced the results, we conducted a sensitivity analysis in which we modelled the percentage of limited English-speaking households, SUD prevalence and the Gini coefficient as continuous measures. The results were consistent with no changes in the magnitude or statistical significance of the associations between facility ownership type, receipt of government funding and the provision of non-English language services (online supplemental table 5).

To further explore potential differences in the receipt of government funding by ownership type and the provision of non-English language services, we conducted post hoc subgroup analyses stratified by ownership type. Separate models were estimated for each ownership category using the same covariates for adjustment as in the primary analysis. We found that across all facility ownership types, receipt of government funding was associated with significantly higher odds of offering non-English language services compared with those without government funding. The adjusted odds were highest for publicly owned facilities (aOR=1.68; 95% CI 1.18 to 2.41), followed by private non-profit (aOR=1.51; 95% CI 1.33 to 1.72) and private for-profit facilities (aOR=1.26; 95% CI 1.08 to 1.47) (online supplemental table 7). Also, the distribution of government funding receipt and the provision of non-English language services by facility ownership type are presented in online supplemental table 6.

Discussion

As the US demographic landscape is becoming more diverse, creating effective pathways for accessible SUD treatment for CALD individuals is crucial. Despite disparities in access to linguistically competent SUD treatment services for the CALD population, few studies have examined the role of underlying factors that are associated with these inequities nationally. In this cross-sectional study, we found that providing SUD treatment services in a non-English language is strongly associated with facility ownership type and receipt of government funding.

Our findings show that private for-profit ownership was associated with lower odds of providing non-English language services. This finding is consistent with previous studies on the provision of behavioural health services for SUD, which found that private for-profit facilities were less likely to provide services in multiple languages compared with non-profit facilities.⁴⁹ There may be several factors driving this practice. For example, these facilities may be deprioritising services in non-English languages since such services require significant investment in hiring multilingual staff, language interpreter services and cultural competency training.⁵⁰ In addition, as others have noted, private for-profit facilities may be prioritising higher-margin services.⁵¹ Considering the rising trend of privatisation (including increasing private equity ownership) in SUD treatment facilities,^{34 52} with private for-profit facilities accounting for 43% of the facilities nationally, greater attention should be paid to the consequences of such transformation to CALD individuals.⁵³

One potential approach to support the expansion of non-English language services may be for state and/or local governments to offer financial incentives⁵⁴ to for-profit facilities to boost their capacity and willingness to provide linguistically competent services. The incentivisation can take place through private-public partnerships or through a contractual mechanism, such as reimbursement⁵⁵ or bonuses⁵⁶ for employing bilingual staff who reflect the needs of the communities they serve.

We also found that more than 60% of facilities providing non-English language services received government funding, underscoring the importance of continued subsidies to support the provision of culturally competent services. However, qualitative research indicates that there are major challenges to reliance on government funding, both federal and state, due to funding streams being fragmented, inconsistent and insufficient to address the diverse needs of communities.⁵⁷ Since facilities across all ownership types rely significantly on government funding for the provision of non-English language services, increasing the availability of flexible funding streams is vital for their improvement and sustainability. Government funding may be used to support interpreter services, hire bilingual staff, train healthcare providers in linguistic competencies to build trust and encourage CALD individuals to seek SUD treatment.

Nearly 47% of facilities with non-English language services accepted Medicare, around 80% accepted Medicaid and 77% accepted private insurance. A related study had similar findings, reporting that among SUD facilities that provided treatment in the Spanish language, 77.4% accepted Medicaid as a form of payment.²⁸ A potential reason for the higher acceptance of Medicaid may be that Medicaid offers extensive coverage for SUD treatment,⁹ whereas Medicare had very restrictive reimbursement rules up until 2020.⁵⁸ This potentially prevents the SUD facilities from accepting Medicare as insurance compared with Medicaid or private insurance,⁵⁹ eventually limiting the availability of non-English language services for Medicare beneficiaries. Additionally, we also noted that only 3503 (23%) of SUD facilities in our cohort reported offering programmes tailored to seniors and older adults, which could further explain their lower likelihood of Medicare acceptance. The 2024 rules from the Centers for Medicare and Medicaid,^{60 61} which expanded access to SUD treatment for Medicare beneficiaries, may positively impact the prospects of care for Medicare beneficiaries with SUD. However, their impact on the availability of non-English services for Medicare beneficiaries remains unclear. Future research should explore this possibility.

The map in figure 1 illustrates the need and availability of non-English language services at SUD treatment facilities. We found that California (0.76), Connecticut (1.93), Hawaii (2.46), Illinois (1.87), Massachusetts (2.03), Nevada (0.76) and Rhode Island (1.76) had lower rates of non-English language services at SUD treatment facilities per 1000 limited English-speaking households. Despite having higher percentages of limited English-speaking households and surpassing the national average for SUD prevalence, these states did not appear to have a sufficient number of facilities with non-English language services–highlighting a potential gap in the provision of patient-centered care. This finding should be interpreted with some caution. The lower rate of non-English language services at SUD treatment facilities in these states likely reflects multiple intersecting factors, some of which we could not account for due to data limitations. While facilities in other states may provide care in languages other than English, these services are often predominantly in Spanish (online supplemental table 3). Given their linguistic diversity, states like California, Connecticut and Massachusetts may require services in a broader range of languages, such as Chinese or Polish, making it more challenging for facilities to meet all diverse language needs. Additionally, not all limited English-speaking households may seek or require SUD treatment.

Building on prior work on the relationship between Medicaid expansion and provision of non-English language services, our findings differ from Lee and Dugan,³¹ who found no association between Medicaid expansion and the availability of non-English language treatment services. In contrast, we observed significantly higher odds of non-English language service provision among facilities in non-expansion states compared with the expansion states. This counterintuitive finding should be interpreted with caution and may reflect underlying state differences rather than the impact of Medicaid expansion in terms of expanding access to SUD care. While Medicaid expansion likely increased insurance coverage among diverse, non-English speaking populations, the demand for non-English language services may have outpaced the available supply of such services. Also, to counter our findings, a recent study using data from 2010 to 2020 showed that under Medicaid expansions, there has been a modest increase in the supply of SUD treatment facilities providing services in Spanish in areas with the highest concentrations of Spanish speakers,³³ thus meeting the linguistic needs of the CALD population. Additional research, leveraging more granular and updated data, may shed further light on some of these contradictions.

This study has several limitations. Importantly, we acknowledge that our utilisation of the ‘LANG’ variable in NSSATS only captures whether services are provided in non-English languages, not whether these services are provided by bilingual staff or through online or in-person interpreters. In addition, interpretation services may be offered by persons other than staff counsellors or on-call interpreters, such as community health workers,³¹ which we could not capture. Hence, despite controlling for various state-level factors affecting the provision of non-English language services, our study does not account for all barriers to such services, such as shrinking community support⁶² or workforce shortages,⁶³ which may lead to residual confounding.

Because this study is cross-sectional, we could not examine the temporal associations between the changing structural factors and the provision of non-English language services. In addition, some state-level covariates were derived from adjacent years when 2019 estimates were unavailable. Because these indicators represent relatively stable factors at the state level, we assumed minimal variation around the 2019 study year; however, the use of non-contemporaneous estimates may introduce some degree of misclassification. As these variables were measured independently of the outcome, any resulting misclassification would likely be non-differential. Also, 2019 data may not fully represent the current availability of non-English language services due to the rise in telehealth services during COVID-19.⁶⁴ However, this study remains timely and particularly relevant given recent shifts in federal policy, including the 2025 executive order designating English as the official language of the USA and revoking Executive Order 13166.⁶⁵ The revocation of the 13166 order, which previously required language access services for federal agencies and recipients of federal funding, eliminates government-wide guidance on language access policies. These policy changes may have significant implications⁶⁶ for culturally and linguistically appropriate SUD services, highlighting the need for continued research into how such policies impact healthcare access and outcomes. Our analysis was conducted by adjusting for various sociodemographic and policy factors at the state level; however, a more granular assessment, particularly at the county level, may add additional context for our findings. For example, among the limited English-speaking population, a breakdown at the county level can reveal the dominant language spoken and help inform SUD treatment facilities to offer services in that language. Furthermore, although we accounted for mandatory cultural competency training requirements and Medicaid expansion status, we did not control for state-specific SUD treatment policies or initiatives due to their complexity and heterogeneity. Future research should examine how these policies influence the provision of non-English language services. Finally, we used ACS data to identify limited English-speaking households as a proxy estimate for capturing state-level percentages of the CALD population. While the ACS data may represent the CALD population, it does not represent all individuals who require SUD treatment. Therefore, the rate of availability of treatment facilities offering non-English language services per 1000 limited English-speaking households might have been overestimated or underestimated in some states. Lastly, the outcome in our analysis was relatively common; the OR may overstate relative risks and should therefore be interpreted as measures of association rather than risk ratios.

Conclusions

Our cross-sectional study showed that structural factors—namely, facility ownership type and government funding—are associated with the availability of services in non-English languages required by the CALD population. SUD treatment facilities aiming to enhance their patient-centered care practices may consider exploring available funding opportunities and partnerships to build the capacity for services that meet the linguistic needs of the increasingly diverse communities. Further research using longitudinal or quasi-experimental designs is needed to better understand how changes in structural factors, including ownership type and government funding, influence the availability of services in non-English languages at SUD treatment facilities.

Data availability statement

Data are available upon reasonable request.