Although mental illness treatment rates have increased recently (
1,
2), many persons with mental illness still do not receive psychiatric care (
3,
4). Further, the ability to access services says little about whether the care is evidence based or of sufficient quality or quantity to promote improvement (
5). Evidence-based treatment guidelines have improved adherence to medications (
6,
7) and are associated with improved care quality and outcomes (
8).
Limited English proficiency (LEP) represents a particularly important barrier to the ability to access general medical care (
9–
12) and is an especially great barrier to accessing psychiatric care (
13–
16). Because many mental health evaluations and treatments, such as psychological “talk therapies,” rely on strong communication, persons with LEP may be especially inhibited from seeking mental health treatment and from receiving treatments of minimally adequate quality (
17).
Yet few prior studies of LEP and mental health treatment were large in scale, generalizable, or quasi-experimental (
17). Offering interpreter services or other language access programming should, according to prior research, improve the quality of specialty mental health services provided to persons with LEP. However, implementing language access programming in resource-strapped settings may diminish quality of care if measures are not taken to address overcrowding, allocate stretched resources, or educate providers about new challenges likely to be encountered in caring for clients with LEP.
California’s Medicaid mental health care system represents a key opportunity to study implementation of language access programming on a large scale. Title VI of the 1964 Civil Rights Act prohibits recipients of federal funds from providing care of more limited scope or of lower quality to persons with LEP (
9,
18,
19). Since then, Title VI has been reinforced by federal and state rulings and by passage of complementary laws in all 50 states (
20,
21).
In 1997, the California Department of Mental Health (DMH) adopted a “threshold language access policy” to meet its Title VI obligations. The policy directs county-operated Medicaid mental health plans to provide a minimum level of language access programming for populations whose language meets threshold language status. Threshold status occurs either when 3,000 county residents or when 5% of a county’s residents speak the same non-English language. Minimum programming consists of four parts: a 24-hour, toll-free phone line with linguistic capability; translated written materials to assist beneficiaries in accessing medically necessary specialty mental health services, including personal correspondence; linguistically capable staff or interpreters at key points of contact; and free information to consumers and communities about the availability of these linguistic services (
22).
Adoption of language access programs has been shown to increase the penetration of mental health services among some LEP populations (
13,
16,
23,
24). However, penetration rates—the rate at which persons who are eligible for services receive care—do not capture whether persons with LEP are more or less likely to receive adequate or recommended care. By providing linguistically capable staff or interpreters at key points of contact, the language access programming required by the threshold language access policy potentially sets the stage for improving the quality of care provided to persons with LEP.
Thus we studied the effects of implementing language access programming on quality of care, using medication follow-up visits during the early stages of treatment as an indicator of quality. Medication follow-up visits serve as a strong measure of the adequacy of psychiatric care, given that many psychiatric medications require several visits to adjust dosing, to switch drugs as needed, and to deal with side effects (
25,
26). Multiple medication visits following a new diagnosis or medication regimen are especially important, because long-term management of illness and a good understanding by patients of their medications are associated with stronger medication adherence and improved outcomes (
27,
28).
We tested conflicting hypotheses about the impact of implementing language access programs on Spanish-speaking LEP clients of county-operated mental health departments in California, where they had received diagnoses of schizophrenia, major depression, and bipolar disorder. On the one hand, consistent with prior literature, we expected that language access programs would improve the likelihood that persons with LEP who were eligible for Medicaid would receive adequate follow-up medication visits from county-run mental health providers in California. On the other hand, it is possible that mental health programs would be unable to adequately serve increased numbers of clients with LEP, especially in a context of clinical complexity and competing clinical demands, making receipt of high-quality care less likely.
Methods
Study Design
This study used a longitudinal design with a nonequivalent control group. County-level fixed effects controlled for time-invariant differences between counties, whether observed or unobserved. In addition, because they were observed before and after the implementation of language access programming, county plans served as their own control groups. Counties that did not implement language access programming in a given quarter year served as a nonequivalent control group. The onset of language access programming was staggered over 38 quarters from 1997 to 2006, with some counties waiting to fully implement programming until required. The research design exploited California’s decentralized Medicaid structure, in which county plans act independently.
This study was approved both by the Committee for the Protection of Human Subjects, University of California, Berkeley, and by California’s State Committee for the Protection of Human Subjects. The data transmitted to the investigators lacked information identifying individual people.
Data Sources
We assembled five distinct panel data sets: county-by-quarter data from the DMH, containing quarterly counts of each county’s Medi-Cal mental health clients who declared Spanish as a primary language; county-by-quarter data from the California Department of Health Services containing quarterly counts of each county’s Medi-Cal enrollees who declared Spanish as a primary language; county-by-quarter data from a survey of county Medi-Cal mental health plans describing the language access programs offered in each county; county-by-year data from the DMH indicating the counties that had been notified of having met language threshold status; and county-by-year census data for county-level demographic characteristics. The first two data sets were collected from counties, the third was collected from DMH records available online, and the fourth was collected through a survey of mental health departments. We gathered the fifth data set from census data.
Study Population
To examine care quality, we measured medication follow-up visits for Spanish-speaking clients with LEP ages 19–64 who had been diagnosed as having schizophrenia, major depression, or bipolar disorder, because effective psychotropic medications exist for these relatively common conditions. County mental health departments identify Spanish-speaking LEP clients through clients’ self-reported inability “to speak, read, write, or understand the English language to interact effectively with county and local welfare/health agencies” (
25). This study used the same definition, because only persons with self-identified LEP had a right to services in their preferred language under the threshold language access policy. To minimize extreme fluctuations stemming from fractions with small denominators, we omitted counties with fewer than 50 enrolled Medi-Cal beneficiaries. In addition, the analysis was restricted to counties with at least one new client who met our focal criteria during the analysis period and to counties with over 50 enrolled beneficiaries, reducing to 30 the number of counties represented in the analysis. Because only county-quarter observations with at least one new client could be analyzed, not all 30 counties had observations for each quarter between 1996 and 2006, resulting in the inclusion of 911 county-quarter observations in the final analysis.
Measures
Dependent variables.
Because of the importance of medications in the treatment of many psychiatric diagnoses and the strong side effects often resulting from psychiatric medications, timely receipt of an adequate number of medication follow-up visits by patients diagnosed as having psychiatric disorders is a strong measure of the quality of mental health care (
26–
29). Prior studies have used measures of follow-up medication visits in assessing the quality of mental health care; for example, Wang and colleagues (
5,
30) defined adequate care as receipt of four or more visits within 12 months following the prescription of a suitable medication.
To capture county-level rates of medication follow-up visits, we used two similar measures: two or more visits within 90 days and three or more follow-up visits within 180 days of an initial medication visit following recent entry or reentry into the Medi-Cal system. These measures concentrated on early phases of treatment, when careful monitoring has the potential to increase the likelihood of prescribing an appropriate medication and dosage level.
The measures were created by counting the number of new Medi-Cal clients with LEP in a given county-quarter who had been diagnosed as having schizophrenia, major depression, and bipolar disorder and who received at least a minimum number of medication visits within 90 or 180 days of their first system contact (the numerator). Clients were considered new if they had not had any services for a given (“clean”) period prior to the initial medication visit (the denominator). The 90-day follow-up measure used a clean period of 90 days, and the 180-day measure used a clean period of 180 days.
Sensitivity tests for the construction of the outcome measures were conducted by systematically varying three parameters: the length of the clean period, the length of the follow-up period, and the number of medication follow-up visits. [The results of the sensitivity tests are available online as a data supplement to this article.] Although alternative criteria for the number of required visits and the length of the clean and follow-up periods predictably altered follow-up visit rates and the number of eligible clients, multivariate findings did not vary substantially in magnitude or at statistically significant levels across different constructions.
Explanatory variable: language access programming.
To assess whether each county had implemented required language access programming for Spanish speakers, we constructed a binary variable reflecting implementation in a given quarter of all four components required by the threshold language access policy (
22,
23). A quarterly time trend variable captured ongoing longitudinal trends, and an indicator interacting quarter by language programming status captured trends after counties implemented programming.
To assess implementation, we sent surveys by e-mail to all 57 county mental health departments. Counties were asked to report on services implemented from 1997 to 2006 for each language with threshold status. Completed surveys were received from 35 (61%) counties with at least one language (often, though not necessarily Spanish) that surpassed threshold levels.
Control variables.
County fixed effects controlled for time-invariant county characteristics. We also controlled for several time-variant characteristics of counties and county plans. The number of eligible beneficiaries per 1,000 residents accounted for variation in county plan burden over time. The number of bilingual staff accounted for the capacity of counties to treat persons with LEP regardless of whether the county’s language access programming met the requirements of the threshold language access policy. Spanish-language penetration rate—the proportion of all Spanish speakers with LEP and eligibility for services who received care (
13,
24)—accounted for language-specific use of services. To control for case complexity, we measured the percentage of clients ages 46–64 and the percentage of clients with a disability. The same outcome measures used to describe rates of adequate follow-up medication visits by Spanish-speaking clients were also used for English speakers to account for temporal changes in overall plan activity and resource levels.
Analysis.
To assess the impact of the threshold language access policy on whether new clients received adequate follow-up medication visits, we used linear regression with county fixed effects at the county-quarter unit of analysis. County fixed effects controlled for time-consistent differences between counties, thereby reducing the likelihood that our analysis would be subject to omitted variable bias (
30). Standard errors were adjusted for clustering within counties.
Results
Language programming was implemented in 697 (77%) of the 911 county-quarter observations included in the analysis (
Table 1). On average, only 34% of Spanish-speaking clients with LEP and 30% of English-speaking clients received at least two follow-up medication visits within 90 days (median=31% and 29%, respectively). For speakers of both languages, rates were slightly higher for receipt of three or more visits within 180 days than for receipt of two visits within 90 days. Large differences between the 25th and 75th percentiles of all variables indicated marked differences across county-quarters in quality of care, need for and supply of services, and case mix complexity.
Unadjusted means of explanatory and control factors were stratified by whether county-quarters were above or below the median for the quality measures. Observations below the median were associated with serving over twice as many eligible beneficiaries compared with observations above the median (
Table 2). County-quarters above the median were actually less likely to implement language programming, but they had moderately higher Spanish-language penetration rates and substantially higher rates of adequate follow-up visits among English speakers.
In multivariate analyses, language access programming was not significantly associated with receipt of at least two medication follow-up visits within 90 days (
Table 3) or at least three medication follow-up visits within 180 days (
Table 4) among new Spanish-speaking clients. Both outcomes had confidence intervals that widely encompassed zero. The time trend–programming interaction variable also was not significantly associated with both the 90- and the 180-day follow-up outcomes.
Discussion
We found no evidence that rates of follow-up visits differed for Spanish speakers diagnosed as having schizophrenia, major depression, and bipolar disorder on the basis of whether county mental health plans implemented mandatory language access programming. For both outcome measures studied, the estimated coefficient departed little from zero, and the confidence interval spanned well beyond zero. Nor was employment of bilingual providers, another measure widely believed to promote return visits, linked with rates of medication follow-up visits.
Sufficient medication follow-up visits for patients with schizophrenia, major depression, or bipolar disorder are important to adjust dosing, to switch drugs as needed, and to deal with side effects (
26–
29), yet fewer than four of every ten clients on average received three or more follow-up medication visits within 180 days, regardless of language proficiency. Threshold language policy and language access programming have proven to be successful for increasing treatment participation rates for many persons with LEP who need treatment (
13,
16). However, as more new clients draw upon limited language access resources, fewer resources, including translators and bilingual staff, may be available to treat clients who have sought treatment. Under the threshold language access policy, translators, bilingual staff, telephone assistance, and other forms of capacity must increase with the supply of LEP clients, but the policy itself appears not to provide resources to meet these needs nor to suggest how needed resources might be acquired.
The Spanish-speaking persons who were eligible for this study were moderately more likely to receive adequate follow-up visits than English-proficient clients. However, the difference between the groups was relatively small, and ample room existed for improvements in the quality of care received by both language groups. In addition, bivariate descriptive statistics comparing low- and high-quality county-quarters suggested that receipt of higher-quality care among both English and Spanish speakers was associated with smaller populations of beneficiaries. County-quarters with high rates of adequate follow-up visits among Spanish-speaking clients also had substantially higher rates of adequate follow-up visits for English speakers. These findings suggest that agencies serving relatively large populations may face greater challenges in scheduling follow-up visits for Spanish-speaking and English-speaking clients alike. This hypothesis merits further study; the role of language access programming may be better understood in the wider context of key system-related influences on treatment quality.
Furthermore, moderately greater rates of adequate follow-up among Spanish-speaking clients should be understood in light of the very low treatment penetration rate for this population. English-speaking clients were roughly five times more likely to appear in treatment than Spanish speakers, a disproportionate underrepresentation of Spanish speakers that likely did not reflect treatment need. If the Spanish-speaking clients who received services were especially in need of care or otherwise highly motivated, they also may have been more likely to have been scheduled for follow-up appointments and to have returned. Even so, rates of adequate follow-up visits were low, less than 40% on average.
This study had limitations. First, although sensitivity analyses demonstrated stable findings across several variable specifications, the possibility remains that the thresholds tested did not capture truly beneficial effects. Second, rates of adequate follow-up medication visits are only one dimension of quality. More direct measures of quality, such as the receipt of at least two antipsychotic medications (
31) or polypharmacy (
32), may have been more sensitive to language programming. In addition, although we capitalized on Medi-Cal’s psychiatric medication reporting requirements in constructing this quality indicator, we were unable to assess whether prescribed medications were appropriate for the mental illness being treated. Still, attendance of follow-up visits almost certainly increases the probability that patients will be prescribed appropriate medications. Third, although the staggered implementation of language access programming across counties, along with use of county-level fixed effects, reduced the likelihood that omitted variables influenced these findings, the possibility remains that an unobserved time-changing factor may have confounded these results. Fourth, other barriers aside from language proficiency, such as stigma and cultural attitudes (
33) and the sociocultural context (
34), may also prevent persons with LEP from gaining access to high-quality mental health services. Fifth, some beneficiaries who declared Spanish as their preferred language may have had partial English fluency, which could reduce the effect of language programming observed in this study. However, even persons with partial English proficiency may benefit from language programming, because of the need for especially strong communication in many mental health evaluations and treatments. Finally, the study may not apply to smaller, poorer, and more conservative counties, given that they were less likely to respond to the survey or to meet the study inclusion criteria.
Conclusions
This study found no evidence that language access programming led to significantly increased rates of follow-up medication visits for Spanish-speaking clients with LEP. However, as a silver lining, language programming did not adversely affect the quality of follow-up medication visits. Under California’s threshold language access policy, translators, bilingual staff, telephone assistance, and other forms of care assistance must increase with the supply of clients with LEP; still, this policy did not appear to substantially alter the quality of care received by Spanish-speaking patients. Additional research is needed to better understand factors associated with the quality of behavioral health care received by persons with LEP.
Acknowledgments
This study was funded by a grant from the National Institute of Mental Health (R01 MH0709452-01).
The authors report no competing interests.