Depression is a leading cause of disability worldwide, with an estimated 17.3 million (7.1%) U.S. adults ages ≥18 years reporting at least one depressive episode in the past year (
1). Veterans are at high risk for developing depression, given the stressors associated with military service (
2). According to a recent study, the prevalence of depression among U.S. veterans in 2005–2016 was roughly 9.6% (
3).
Primary care serves as an initial point of contact for patients and a setting in which depression is often first recognized and treated. Antidepressant medications, a treatment recommended for patients with moderate to severe depression, are commonly prescribed in primary care (
4). Prescription of antidepressants in the general U.S. population and in primary care has increased markedly in recent decades (
5,
6). Access to antidepressant treatment is essential, especially for individuals with severe depression and when such treatment is preferred by the patient (
7). Many primary care providers cannot provide psychotherapy; therefore, antidepressant medications can be an important early intervention for patients who have limited access to mental health care (
8).
Despite the recent increase in antidepressant prescription, rates of antidepressant treatment are not equal across racial-ethnic groups (
9,
10). Black and Hispanic people in the general, non–treatment-seeking population are consistently prescribed antidepressants at lower rates compared with White individuals (
10,
11). Patients belonging to minority groups are also prescribed antidepressants at a lower rate in primary care and psychiatric settings compared with White patients. However, these data are from ≥10 years ago and may not represent current patterns of care. The data also were based on medical charts for diagnosis of depression and did not capture patients’ self-reports of symptoms (
12–
14).
Racial-ethnic disparities in depression treatment also exist in the veteran population, such that veterans belonging to minority groups are less likely to be prescribed adequate antidepressant treatment (
15–
17). These findings, however, may not represent current patterns in the Veteran Health Administration (VHA) system. In addition to the limitations of existing data noted above (e.g., reliance on medical chart review), the data for these studies were collected in the context of ongoing psychiatric care and not in primary care settings. They were also collected either before or shortly after the implementation of initiatives within the VHA system to improve access to and delivery of mental health care, such as the Primary Care Mental Health Integration (PCMHI) program initiated in 2007. PCMHI and other collaborative care models may address long-standing inequities by providing opportunities for early detection of mental health problems and by increasing access to mental health care to all veterans (
18–
20).
Much is still unknown about the extent to which racial disparities in prescription of antidepressants remain in the current integrated primary care system of the VHA. The aim of this study was to examine associations between current antidepressant prescription rates and race for White and Black veterans who were referred to the collaborative care program from VHA primary care within an integrated primary care system. Recognizing and demonstrating the scope of racial disparities is an important first step to addressing them.
Methods
Participants
This study included adult primary care patients (Black, N=4,120; White, N=4,372) referred to the Behavioral Health Laboratory (BHL), an integrated behavioral health program embedded within the Corporal Michael J. Crescenz Veterans Affairs (VA) Medical Center (
21). Patients are referred from primary care to the BHL on the basis of the results of annual screening measures (e.g., 2-item Patient Health Questionnaire) or the provider’s clinical judgment. The first step of a referral to the BHL is a structured assessment that acts as a gateway and triage to mental health care. The assessment is conducted with a mental health technician to assess self-reported symptoms with a series of validated questionnaires (described below). Information on the BHL, including how it functions, has been previously published (
21,
22).
Data from the initial assessment, which include patients’ symptoms and history of primary care use before mental health treatment, were utilized in the study. Patients with complete data for medications prescribed, race, and depression symptoms were included. Clinical patient record data collected over 6 years (January 1, 2015–December 22, 2020) were extracted, and duplicates were removed (i.e., each patient was unique, and data from each patient’s first entry into the collaborative care program was used in this study). All principles outlined in the Declaration of Helsinki were followed. Because these data were part of initiatives to conduct equitable clinical care and program evaluation, patient consent was not needed. This protocol was approved by the institutional review board of Corporal Michael J. Crescenz VA Medical Center.
Data Collected
The primary outcome was whether the patient had a current antidepressant prescription at the time of assessment. To conduct the initial assessment, the health technician accessed the patient’s medical record, which included prescription information. The technician verified the medications in the medical record with the patient, and, if there was a discrepancy, the patient’s self-reported information was entered into the database. Patients were coded as either prescribed or not prescribed an antidepressant medication. “Antidepressant” was defined as a serotonin-norepinephrine reuptake inhibitor, selective serotonin reuptake inhibitor, tricyclic antidepressant, or other type of medication commonly considered to be an antidepressant. Trazadone is used for conditions other than depression, such as sleep disorders, so separate analyses were conducted with trazadone coded as an antidepressant and not coded as an antidepressant.
The primary predictor was race, categorized as either Black or White. In the assessment, participants were limited to a single selection among the following: White, Black or African American, Asian or Pacific Islander, Native American or Alaskan, refused disclosure, or other or mixed. Only veterans who selected White or Black or African American were included in the study because of small samples in the other categories, limiting statistical power. Ethnicity was examined as a covariate (non-Hispanic or non-Latino vs. Hispanic or Latino).
Depression symptoms were measured with the 9-item Patient Health Questionnaire (PHQ-9) (
23). Possible scores on the PHQ-9 range from 0 to 27, with higher scores indicating greater severity of depression symptoms. The symptoms were categorized on the basis of score ranges: no-to-mild depression symptoms (PHQ-9 score of 0–9), moderate depression symptoms (score of 10–19), and severe depression symptoms (score of 20–27) (
23,
24).
Providers sometimes prescribe antidepressant medication for conditions other than depression (e.g., generalized anxiety). To account for such prescriptions, patients’ scores on additional mental health assessments conducted during the assessment were compiled into a composite item: patients meeting criteria for one or more other mental health condition for which there is an evidence base for treatment with antidepressants were categorized as either having or not having a comorbid mental health condition with an evidence base for an antidepressant prescription. This composite variable was included as a covariate. Mental health conditions included in this categorization were anxiety (measured with the 7-item Generalized Anxiety Disorder scale [GAD-7]) (
25) and posttraumatic stress disorder (measured with the Posttraumatic Stress Disorder Checklist for DSM-5 [PCL-5]) (
26). Participants were coded as having a comorbid mental health condition that has an evidence base for an antidepressant prescription if they had a GAD-7 score ≥10 or a PCL-5 score ≥35. Scores on the GAD-7 range from 0 to 21, with higher scores indicating higher levels of anxiety; scores on the PCL-5 range from 0 to 81, with higher scores indicating greater severity of trauma-related symptoms.
Other demographic and psychosocial variables collected included gender (male or female), marital status (married/partnered or other), age at time of assessment, employment status (unemployed or employed), and self-rated health (Veterans RAND 12-Item Health Survey; participants are asked, “In general, would you say your health is excellent, very good, good, fair, or poor?” and responses are scored 1–5, with lower scores denoting better health) (
27,
28).
Statistical Analysis
SPSS, version 27, was used for statistical analysis. We examined the bivariate association between race and antidepressant prescription (no antidepressant prescription or antidepressant prescription). We also examined the association between race and antidepressant prescription in a series of stepped multivariate models that adjusted for depression symptoms (i.e., the sum of PHQ-9 scores, analyzed as a continuous variable), demographic and psychosocial variables (see above), and other related symptoms (dichotomous score of “comorbid mental health condition” or “no comorbid mental health condition”).
Next, we stratified patients according to depression symptom severity to examine differences in clinical guideline implementation, by race. Among adults with no-to-mild symptoms of depression, antidepressants are recommended only if patients were originally prescribed antidepressants during treatment and are in early remission (
4). Among adults with moderate depression symptoms, either antidepressants or evidence-based psychotherapy are recommended. For adults with severe depression symptoms, both antidepressants and evidence-based psychotherapy are recommended. Multivariate models were run, adjusted for demographic and psychosocial variables (see above) and other related symptoms (comorbid mental health condition or no comorbid mental health condition), to examine the relationship between race and antidepressant symptoms among patients with no-to-mild, moderate, and severe depression symptoms (
23).
Results
Black patients composed slightly less than half of the total sample (N=4,120, 48.5%). Of the total sample, 3.9% identified as Hispanic or Latino (see
Table 1 for additional demographic characteristics). We noted significant differences between Black and White patients, including that Black patients had higher depression scores compared with White patients (see
Table 1 for additional comparisons). Overall, at least one antidepressant was prescribed to 1,854 patients, and two antidepressants were prescribed to 340 patients (
Table 2).
In the logistic regression, in which we controlled for depressive symptoms as a continuous variable, White patients were significantly more likely than Black patients to be prescribed an antidepressant (SE=0.05, Wald χ
2=116.82, df=1, p<0.001, odds ratio [OR]=1.80, 95% confidence interval [CI]=1.62–2.00) (see
Table 3 for additional statistics). Race was again significantly associated with antidepressant prescription when we controlled for demographic characteristics and depression symptoms (SE=0.06, Wald χ
2=140.84, df=1, p<0.001, OR=1.96, 95% CI=1.75–2.19) (
Table 3). Race remained a significant predictor when we controlled for depression symptoms, demographic and psychosocial variables, and comorbid mental health conditions, with White patients being significantly more likely than Black patients to be prescribed an antidepressant (SE=0.06, Wald χ
2=141.72, df=1, p<0.001, OR=1.96, 95% CI=1.75–2.19) (
Table 3). The pattern of results was the same with and without trazadone coded as an antidepressant; results are presented with trazadone coded as an antidepressant.
Analyses were also conducted by stratifying patients on the basis of severity of depression symptoms. Among patients not meeting criteria for depression (i.e., PHQ-9 score <10), 21% (N=406) of White patients (N=1,917) were prescribed an antidepressant, compared with 13% (N=201) of Black patients (N=1,510). Among patients meeting criteria for moderate to severe depression (PHQ-9 score ≥10), 20% (N=513) of Black patients (N=2,610) were prescribed an antidepressant, compared with 30% (N=744) of White patients (N=2,455). In binary regressions, among patients with no-to-mild depression symptoms (PHQ-9 score 0–9), race was a significant predictor of antidepressant prescription, such that White patients were more likely than Black patients to have an antidepressant prescription (SE=0.100, Wald χ
2=51.79, df=1, p<0.001, OR=2.05, 95% CI=1.69–2.49) (
Table 4). The pattern was the same among patients with moderate depression symptoms (PHQ-9 score 10–19) (SE=0.08, Wald χ
2=67.73, df=1, p<0.001, OR=1.90, 95% CI=1.63–2.21) and severe depression symptoms (PHQ-9 score 20–27) (SE=0.15, Wald χ
2=19.52, df=1, p<0.001, OR=1.87, 95% CI=1.40–2.50) (
Table 4).
Discussion
In this study, we examined the association between race and likelihood of having an antidepressant prescription among U.S. veterans in primary care who were referred to a collaborative care program. Race was consistently associated with likelihood of receiving an antidepressant prescription, with White patients being almost twice as likely as Black patients to have an antidepressant prescription. These results are particularly notable because, in this sample, Black patients had significantly more severe depression symptoms than White patients, and significantly more Black patients had symptoms of moderate and severe depression.
Stratified analyses of patients on the basis of depression severity were also conducted to understand racial disparities in implementation of clinical guidelines. Among patients with no-to-mild symptoms of depression, for whom antidepressants are recommended only as maintenance therapy for patients with major depressive disorder who responded well to the medication (
4), White patients were more likely to be prescribed an antidepressant. It is possible that these patients met criteria for major depressive disorder and that antidepressants caused a reduction in depression symptoms. However, most patients in primary care who are newly prescribed an antidepressant have minimal symptoms of depression (
29); thus, our results may indicate an overprescription of antidepressants for White patients who do not meet criteria for depression.
These results held across categories of depression severity. Among patients with moderate depression symptoms, for whom clinical guidelines recommend either evidence-based psychotherapy or antidepressants according to a patient’s choice, Black patients were significantly less likely to have an antidepressant prescription compared with White patients (
4). Of note, patients in this study were unlikely to have had access to psychotherapy because they were newly referred to the collaborative care program, and psychotherapy was not available outside of the program. There is evidence to suggest that Black and other minority patients prefer psychotherapy. This study did not examine patient preference; thus, the cause of the observed racial disparity in depression treatment is unknown, and this disparity may be rectified if patients were engaged in care where psychotherapy is offered (
30,
31). In this study, Black patients with moderate depression symptoms in primary care were less likely than White patients to have an antidepressant prescription and unlikely to be receiving psychotherapy; therefore, it is unlikely that their care was in accordance with clinical guidelines (
4).
Likewise, for patients with severe depression symptoms, for whom both psychotherapy and antidepressants are recommended, race significantly predicted antidepressant prescription patterns, indicating that Black patients are less likely to receive the recommended care compared with White patients. Overall, White patients were more likely to be prescribed an antidepressant, regardless of the severity of their depression. These results underscore the importance of examining patterns of racial disparities in all settings in which mental health care is provided to identify areas for improvement.
Our analyses focused on prescription patterns at the time of referral to a collaborative care program, and this study could not address why these differences in prescribing patterns emerged. Therefore, several patient, provider, and contextual factors could explain these results. One explanation to consider is racial bias among medical providers in offering antidepressant prescriptions. Extensive evidence indicates that implicit racial bias negatively affects treatment provided to Black patients across medical conditions (
32–
34). This study sample was from a well-established integrated primary care system; similar integrated care systems have shown a reduction in some mental health care delivery disparities (
18–
20). The patients in this study, however, were at the entry point to an integrated behavioral health program when the data were collected. Integrated primary care services may address racial inequities in the patient population they directly treat, although it may be necessary to expand offerings or resources that address racial disparities to other staff, such as primary care providers, who address mental health concerns in a larger patient population (
22).
Another potential explanation of these results is that patient-provider communication during primary care encounters may vary between Black and White patients, affecting treatment recommendations. A previous study reported that, despite having similar severities in depression symptoms, Black patients experienced less rapport building and communication about depression with their primary care providers compared with White patients (
35). In this study sample, Black patients had more severe depressive symptoms than White patients. The full measure of depression symptoms (i.e., the PHQ-9), however, was not administered until the patient was referred to the collaborative care program, and, therefore, symptom scores were not available to providers before this assessment. Additional use of measurement-based care earlier in the treatment process may help address differences in patient-provider communication about depression symptoms. Black and other minority patients may also prefer psychotherapy (
30,
31) or may distrust medical providers or treatments because of the history of systemic racism and unethical treatment of Black people in medicine and scientific trials (
36–
38).
This study had several limitations. We examined prescription rates among U.S. veterans referred from primary care for assessment of a mental health need; this study did not examine other related data, such as information about communication, prescriber characteristics, medications not in the medical record, history of mental health care or other conditions (e.g., panic disorder), or adherence to the medications prescribed, thereby limiting the scope of our conclusions. The specific setting of this study, a well-established integrated primary care model in the VHA system, may limit generalizability. Although the present sample had strengths (e.g., its size and a high percentage of Black patients), it also had limitations, such as the low number of veterans in racial-ethnic groups other than non-Hispanic White and Black. In addition, most patients were men; women are more frequently prescribed antidepressants than men (
39,
40), and the racial disparities in this sample may be different from those in samples with different gender compositions. Assessments for mental health and mental health treatment at this VHA facility were available only in English, potentially limiting access to care for non–English-speaking veterans. The assessments utilized to control for comorbid conditions focused on conditions for which there is an evidence base for antidepressant use; however, they may not capture all conditions for which a provider may prescribe an antidepressant. Finally, the sample represented patients who were identified and referred for further assessment, and, thus, it was a sample of convenience, not a random or complete sample of patients prescribed antidepressants within a primary care setting.
These results inform several recommendations. First, we recommend conducting a regular review of prescription rates in clinics by race to identify racial disparities. Future research on the drivers of inequality in antidepressant rates can help inform the next steps in improving treatment of patients from minority groups. For example, additional research on patient-provider communication could help identify provider bias in recommending medications or patient hesitancy in accepting the medication; recommendations to address disparities can then be developed. Other options include examining the impact of clinical assessments on prescriptions, examining differences in health outcomes after prescription, and including medical record data to provide additional detail on prescribing time lines. Inequities in the provision of mental health care are likely the symptoms of systemic racism in both health care and society at large and likely need to be addressed in systematic, structural ways that were beyond the scope of this study (
41,
42).