Racial-Ethnic Disparities in Outpatient Mental Health Visits to U.S. Physicians, 1993–2008

After the U.S. Surgeon General’s conclusion in 2001 (1) that access to mental health services was “plagued by disparities,” research on how race and ethnicity influence the probability, course, and outcome of mental health treatment has grown markedly. Studies have documented that racial-ethnic minority groups receive less mental health care overall (2–8), specifically for depressive (9–12), bipolar (13), anxiety (10–12,14), personality (15), and eating disorders (16). The care that persons from minority groups receive has been shown to be of lower quality for depression (9,12,17), substance use disorders (18–20), and schizophrenia (21), to be less likely to include psychotherapy (4,6,12,22), and to consist of shorter office visits (23). Racial-ethnic differences in the prevalence of mental disorders are generally not large enough to fully explain observed racial and ethnic patterns of care (9,24–26).

Identifying which mental disorders involve the largest or most persistent racial-ethnic differences in treatment is important for tailoring disparity reduction efforts, but diagnostic and temporal variations in treatment differences are difficult to assess with available evidence. Most studies of racial-ethnic differences in mental health treatment have aggregated all disorders (3–7,22) or have focused on selected diagnostic categories, such as depression (9,17,25) or substance use disorders (10,18–20,27). Few studies have used consistent methodology to examine treatment differences for a range of specific disorders (5,10–12), and these studies have relied on participants’ recall of past use of mental health care. Although such reports may be valid to enumerate recent visits and as crude measures of any prior care, more detailed information on the specific volume and characteristics of past treatment is vulnerable to recall bias and other error (28). Only a small number of studies have examined time trends in racial-ethnic differences, and available data are contradictory, indicating both the amelioration (12) and exacerbation (7) of differences over time.

We aimed to address some of these limitations in the literature on mental health care disparities by using a nationally representative, provider-verified sample of physician outpatient visits in the United States to examine racial-ethnic differences in the use of outpatient mental health care for a comprehensive range of specific diagnoses over nearly two decades.

Methods

Data source

Data from the National Ambulatory Medical Care Survey (NAMCS) and the National Hospital Ambulatory Medical Care Survey (NHAMCS) (29), which are conducted annually by the National Center for Health Statistics (NCHS), were combined to create a representative sample of adult outpatient visits to U.S. physician offices and hospital-based clinics from 1993 through 2008. The NAMCS samples visits to office-based physicians who are not employed by the federal government, and the NHAMCS samples visits to ambulatory care services in emergency and outpatient departments of noninstitutional, nonfederal hospitals. Although the two surveys differ in sampling design, valid estimates can be obtained from combined data sets because variance estimates rely on ultimate cluster design variables, which are shared from the first stage of the sample design for each survey (30,31). Complete documentation in regard to the NAMCS and NHAMCS methodologies, including questionnaires, sample design, data collection and processing procedures, and estimation procedures, is available online (32).

Census estimates of racial-ethnic group population sizes by age, sex, geographic region, and calendar year were obtained by using the Bridged-Race Population Estimator of the Centers for Disease Control and Prevention (33).

Sample

Mental health visits to physicians were the units of analysis. Mental health visits were identified from the sample of all physician visits on the basis of a primary diagnosis listed by the treating physician with ICD-9 codes, which were cross-walked to DSM-IV-TR codes 290–316 (34). A total of 754,497 visits were identified, representing an estimated 11,768,800,000 weighted visits in the United States over the 16-year period, or an average of 735,550,000 visits per year.

Visit characteristics

The main racial-ethnic categories—assigned by physicians, not patients—were non-Hispanic white, non-Hispanic black, and Hispanic. Findings for the category of non-Hispanic other were restricted to supplementary tables because the group’s size was small and its composition changed markedly over time. ICD-9 codes for the visit diagnosis were categorized into broad and specific diagnostic groups relevant to the DSM-IV-TR (34) by using a crosswalk developed previously. Broader categories were subdivided to present selected specific disorders with the highest treatment prevalence. Physician specialty was categorized as psychiatrist or nonpsychiatrist.

Analytic methods

Visit prevalence ratios.

First, each visit weight for a mental health visit was adjusted by dividing the weight by the corresponding U.S. Census population size, which was calculated on the basis of patient race-ethnicity, sex, age, region, and year. Insurance information was not available for U.S. Census estimates. The prevalence of visits by specific racial-ethnic groups with 95% confidence intervals (CIs) was then estimated as the adjusted, weighted sum of visits for all disorders for the racial-ethnic group, representing the population-adjusted volume of care that the group received for the disorder. Next, we calculated racial-ethnic visit prevalence ratios (VPRs) by using non-Hispanic white visit prevalences as the denominator. VPRs were considered statistically significant when the CIs for the VPRs excluded 1. Analyses were stratified by broad and narrow diagnostic groupings and by each of the patient characteristics used in the creation of the U.S. Census population sizes. VPRs and error estimates were calculated using the PROC RATIO procedure in SUDAAN, version 10.0 (35).

Time trends and other contrasts of VPRs.

VPR time trends were estimated by contrasting VPRs stratified into four-year intervals: 1993–1996, 1997–2000, 2001–2004, and 2005–2008. Weights were adjusted by using the total aggregate population of each period. Trends were tested by calculating absolute changes with CIs, with 1993–1996 as the reference, and trends were considered statistically significant if the CI for the absolute change excluded 0. Pairwise differences between other selected VPRs were similarly tested.

Imputation of race-ethnicity.

Race-ethnicity was imputed by the NCHS for a relatively large proportion of visits during later years of the surveys, ranging from <5% to 40% of values used in specific study analyses. Race-ethnicity was imputed by randomly assigning a value from a record with similar characteristics, with priority given to the patient’s locality (36). Analyses were conducted both with and without adjustment for this imputation. Consistent with NCHS research (37), adjustment did not meaningfully alter our findings, and unadjusted values are reported.

Results

Differences by diagnosis

Total annual prevalence rates for mental health-related physician visits were 197 visits per 1,000 population (CI=188–206) for non-Hispanic whites, 118 (CI=104–131) for non-Hispanic blacks, 114 (CI=99–130) for Hispanics, and 90 (CI=78–103) for non-Hispanic others. [A table presenting data on visit rates per 1,000 population by disorder and racial-ethnic group is available in an online data supplement to this article.] VPRs for outpatient mental health visits to physicians by racial-ethnic group stratified by diagnosis are shown in Table 1.

VPRs for any mental disorder over the entire time period were .60 (CI=.52–.68) for non-Hispanic blacks and .58 (CI=.50–.67) for Hispanics. Compared with non-Hispanic whites, members of racial-ethnic minority groups received significantly less outpatient mental health care from physicians for all diagnostic categories except psychotic and drug and alcohol use disorders.

Differences between non-Hispanic black patients and white patients were particularly large for obsessive-compulsive disorder (VPR=.11, CI=.01–.20), generalized anxiety disorder (VPR=.20, CI=.09–.31), attention-deficit hyperactivity disorder (VPR=.23, CI=.09–.38), personality disorders (VPR=.25, CI=.10–.40), dysthymia (VPR=.30, CI=.22–.39), agoraphobia (VPR=.33, CI=.11–.55), panic disorder (VPR=.36, CI=.19–.54), and nicotine use disorder (VPR=.36, CI=.11–.61). By contrast, non-Hispanic black patients received significantly more treatment for psychotic disorders (VPR=1.47, CI=1.16–1.78), including schizophrenia (VPR=2.07, CI=1.58–2.56); a nonsignificant trend was also noted toward more treatment for drug use disorders.

Hispanics received markedly less outpatient mental health treatment from physicians than non-Hispanic whites for impulse control disorder (VPR=.08, CI=0–.17), autism spectrum disorder (VPR=.16, CI=0–.35), personality disorders (VPR=.21, CI=.04–.39), obsessive-compulsive disorder (VPR=.26, CI=.08–.43), nicotine use disorder (VPR=.32, CI=.02–.63), and bipolar I disorder (VPR=.35, CI=.23–.46). A nonsignificant trend toward receipt of more treatment for drug use disorder was also noted.

Differences by specialty and demographic characteristics

Total annual prevalence rates for visits to psychiatrists were 121 visits per 1,000 population (CI=114–129) for non-Hispanic whites, 57 (CI=47–68) for non-Hispanic blacks, 56 (CI=44–68) for Hispanics, and 51 (CI=42–61) for non-Hispanic others. [A table presenting data on visit rates per 1,000 population by provider type and patient characteristic is available in the online data supplement.] Total annual prevalence rates for visits to nonpsychiatrist physicians were 90 visits per 1,000 population (CI=85–95) for non-Hispanic whites, 68 (CI=59–78) for non-Hispanic blacks, 67 (CI=57–76) for Hispanics, and 45 (CI=36–54) for non-Hispanic others. VPRs for outpatient mental health care by racial-ethnic group stratified by provider specialty and by patient demographic characteristics are presented in Table 2.

Visit rates to psychiatrists and nonpsychiatrist physicians were significantly lower among members of all minority groups compared with non-Hispanic whites, although differences were smaller for visits to nonpsychiatrist physicians. Differences declined with advancing age and were not significant among elderly (over age 65) Hispanic persons for visits to all physician specialty categories, as well as among elderly blacks for visits to nonpsychiatrists. Differences for Hispanics were generally greater in the West than in other regions, whereas differences for non-Hispanic blacks were greater in the South than the Midwest, with the exception of differences in visits to psychiatrists, which were greatest for non-Hispanic blacks in the Northeast. Differences between Hispanics and non-Hispanic blacks were far larger in the West than in other regions. Variation in VPRs by patient sex was small and not statistically significant.

Time trends

Table 3 shows racial-ethnic VPRs stratified by provider specialty and time period. Over the 16 years examined (1993–2008), racial-ethnic differences in any mental health care significantly worsened for Hispanics; differences worsened for both Hispanics and non-Hispanic blacks for visits to nonpsychiatrists. Racial-ethnic differences were otherwise stable over the period.

Discussion

We found large and persisting racial-ethnic differences in use of mental health treatment provided by physicians. Differences varied by disorder and appear to have increased between 1993 and 2008. Differences were particularly marked for a number of diagnoses that have received little attention in the health disparities literature. Compared with non-Hispanic whites, racial-ethnic minority groups received less outpatient mental health care from physicians for most disorders. However, non-Hispanic blacks received more treatment than whites for psychotic disorders, and both minority racial-ethnic groups showed a nonsignificant trend toward more treatment for drug and alcohol use disorders. Differences in visits to psychiatrists were larger than differences in visits to nonpsychiatrist physicians, and racial-ethnic differences were less pronounced among older patients.

Our diagnosis-specific analysis found differences that were largest and most consistent across racial-ethnic groups for anxiety disorders—particularly obsessive-compulsive disorder—as well as dysthymia, personality disorders, nicotine use disorder, and attention-deficit hyperactivity disorder. Little direct attention has been paid to disparities in the treatment of anxiety disorders in general, and particularly for obsessive-compulsive disorder, although one study found that African Americans and Caribbean blacks with obsessive-compulsive disorder were highly unlikely to receive evidence-based treatment, even though both groups had high levels of illness severity and functional impairment (14). A recent comprehensive review of racial-ethnic differences in the prevalence, diagnosis, and treatment of personality disorders (38) identified only three studies that examined treatment, one of which was conducted in the United States. This study found that adults from racial-ethnic minority groups, especially Hispanics, were significantly less likely than whites to receive a range of outpatient services and that those with the most severe personality disorders were the least likely to receive treatment (15). Adults from minority groups are less likely to receive nicotine replacement therapy (39), and persons with mental illnesses are far more likely than others to smoke (40). Little attention has been paid to adult treatment disparities for attention-deficit hyperactivity disorder or for dysthymia. However, unlike this study, a previous study found little evidence of racial-ethnic differences for these disorders (41).

As Table 4 shows, racial-ethnic differences in the prevalence of mental disorders that have been observed in national epidemiologic samples (24,42–46) are too modest to fully explain the treatment differences that we found. For instance, although prevalence rates of anxiety disorders are roughly equal across racial-ethnic groups and there is evidence that certain disorders such as obsessive-compulsive disorder may be more prevalent among minority groups, we found that persons from racial-ethnic minority groups received far less treatment for most anxiety disorders.

Studies have sought other explanations for observed racial-ethnic differences in mental health treatment, including language barriers and immigration status (47–49); patient preferences (50); sociodemographic factors, such as poverty and insurance status (3,17,18,51,52); clinician-related factors (53); and structural factors, such as hospital segregation (54,55) and geographical location (56). Our stratified analyses may help to further explain observed differences. The regional variation in treatment differences that we observed may arise from factors related to the health care delivery system and from cultural factors. For instance, research has shown regional variation between Hispanic subgroups in use of mental health care (57). The finding that racial-ethnic differences were smaller for visits to nonpsychiatrist physicians supports the integration of mental health care into nonspecialty health care services as a potentially important avenue for disparity reduction (58). Our finding that differences ameliorated with age—particularly for elderly Hispanics—suggests that access to public insurance such as Medicare may mitigate disparities in mental health care. Research has indicated that entry into Medicare narrows differences in health care use and health between previously insured and uninsured individuals (59), although recent work has also documented racial-ethnic differences in depression treatment among Medicare beneficiaries (60).

Our finding that members of racial-ethnic minority groups, particularly non-Hispanic blacks, received more mental health care for psychotic disorders represented a striking divergence from the pattern of less treatment for most other disorders. Several explanations are possible. Previous work has shown that compared with whites, African Americans are more likely to be diagnosed as having psychotic disorders (61–64), to receive higher dosages of antipsychotic medications and more depot antipsychotics (65–67), and to be prescribed second-generation antipsychotics and clozapine less often (66,68–71). Such differences could be influenced by provider diagnostic bias (72) and by treatment setting, because African Americans are more likely to use inpatient and emergency services (67,73–75). We also found that minority groups received more treatment for certain substance use disorders, which are not more common in minority groups than among non-Hispanic whites (42). Overrepresentation of minority groups in the U.S. criminal justice system may result in increased enrollment in mandated substance abuse treatments (76). There is some evidence that members of minority groups may receive more restrictive treatment for substance use disorders than non-Hispanic whites (65,77).

The study had a number of important limitations in generalizability and potential sources of error. Our findings of differences in treatment volume are not sufficient to characterize disparities in treatment because NAMCS and NHAMCS data are restricted to outpatient treatment users and do not provide information from standardized diagnostic assessments or about patient preferences and prior treatment. It is therefore important to interpret these results only as racial-ethnic differences in treatment. Because the NAMCS and NHAMCS sample individual visits and not treatment courses, the VPRs we calculated did not distinguish between treatment entry and intensity and might have been driven by racial-ethnic differences in either component of care. Therefore, our findings can be precisely interpreted only as average volumes of physician treatment provided to population groups. However, research has found striking racial-ethnic differences in both initiation and delay of mental health treatment among individuals with diagnoses of anxiety and mood disorders (41), and many studies have shown differences in treatment intensity and quality (9,12,25).

Although the data captured a substantial portion of outpatient mental health treatment, some important treatment settings, such as community mental health centers, other public facilities, and nonphysician care were excluded. Members of racial-ethnic minority groups are relatively more likely to receive care in settings not captured in the data set used for this study (78), and the differences in physician care that we found likely exceed differences in total mental health care received. The impact of this omission may have declined recently because public mental health care funding has been subject to continued cuts and cost-shifting. For instance, although care in certain state-funded mental health specialty settings may be more intensive (5), public mental health care funding continues to shift from state-funded specialty care to federally insured (i.e., Medicaid and Medicare) care in general medical settings (79).

Although sampling provider visits directly avoids error arising from participant recall of mental health care use, provider reports are subject to diagnostic error, which may differ by patient race-ethnicity, and to error in assigning race-ethnicity, which would bias our VPRs toward underestimating true racial-ethnic differences. The frequency of missing information on race and ethnicity in the NAMCS and NHAMCS data has risen, requiring an increasing reliance on imputation, which may increase potential bias toward underestimation of differences. To the extent that particular physicians seen by patients from racial-ethnic minority groups are different from those seen by white patients and that systematic differences may exist in survey response rates between these groups of physicians, survey data could bias our VPRs toward overestimating differences.

We used broad racial-ethnic categories, obscuring variation in factors such as country of origin or immigration status, which have been shown to influence receipt of treatment. Our analyses could not be adjusted for a number of important sociodemographic characteristics, including socioeconomic and insurance status, nor could we adjust for differences in treatment preferences. Experts disagree on how best to conceptualize racial-ethnic disparities. The Institute of Medicine defines disparities as treatment differences unexplained by differences in illness or preferences, which argues against adjusting for sociodemographic factors (80). Finally, our trend data did not measure the potential influence on care of events after 2008, including an extended economic recession, growing emphasis on integrating medical and mental health care, mental health parity legislation, and passage of the Affordable Care Act.

Treatment disparities for specific disorders may arise from different causes and be amenable to different strategies of disparity reduction. The diagnosis-specific patterns we identified highlight disorders worthy of particular focus for disparity reduction efforts and inform speculation about underlying systemic phenomena, which may be helpful to clinicians and policy makers. For instance, it is particularly striking that racial-ethnic minority groups received more treatment than non-Hispanic whites for psychotic and substance use disorders, which are socially disruptive and may carry legal implications, whereas they received markedly less care for a range of anxiety, mood, and personality disorders, which are nonetheless associated with significant distress and functional impairment. These findings might suggest that the contexts and points of entry that often lead to identification of and treatment for psychotic and substance use disorders, such as emergency departments, inpatient units, and law enforcement, could benefit from implementing additional screening for depression and anxiety. At the same time, the settings that commonly provide for identification of mood and anxiety disorders for non-Hispanic whites, such as primary care offices, could better target screening programs for minority groups.

Conclusions

We found substantial racial-ethnic differences in the use of outpatient mental health treatment provided by U.S. physicians. The differences varied substantially by disorder, provider type, and patient age. Many differences were large and did not show evidence of improvement from 1993 to 2008. Findings highlight potentially important diagnostic and sociodemographic patterns. The persistence and magnitude of racial-ethnic differences in care suggest that remediation remains an urgent public health task.