A Naturalistic Study of Racial Disparities in Diagnoses at an Outpatient Behavioral Health Clinic

Rates of diagnoses of schizophrenia among African Americans are elevated compared with rates for other racial and ethnic groups in U.S. clinical settings, contradicting expectations from structured epidemiology surveys (1). Specifically, previous investigations have consistently suggested that African Americans with mood disorders are much more likely than other groups to be misdiagnosed with schizophrenia (2–5). A meta-analysis by Olbert and colleagues (5) concluded that “racial diagnostic disparity in schizophrenia represents a robust albeit heterogeneous clinical phenomenon that has been stable for the past three decades.”

Some of the racial differences in diagnosis rates could be attributable to factors other than clinician bias. For example, a systematic review and meta-analysis of incidence rates of schizophrenia and other psychotic disorders among black Caribbean-born migrants to England suggested that the greater incidence of psychosis in this group relative to English whites has persisted for more than 60 years (6). Moreover, etiological explanations may account for these ethnic differences more so than methodological biases. For example, some etiological explanations include greater poverty, discrimination, and cumulative social defeat, which have been hypothesized to increase dopaminergic activity in the mesolimbic dopamine system via sensitization (7). Biological factors include genetics, infections, and malnutrition (6).

Nevertheless, controlling for unconscious bias in clinical assessment is difficult, and for members of racial minority groups, being assessed or treated by a member of a racial majority group may have effects on the individual’s clinical presentation (for example, mistrust or hopelessness) (8). Consequently, racial differences in clinical diagnoses may reflect two sets of factors (and perhaps the interaction between the two): biological and social environmental factors that affect the likelihood that a given racial-ethnic group may be more susceptible than other groups to a specific disorder, such as schizophrenia; and factors that affect judgment when a clinician is making a diagnostic decision. The focus of this study was on the latter.

Studies by Strakowski and colleagues (2, 9–13) have suggested the hypothesis that racial bias in the diagnosis of schizophrenia spectrum disorders in part results from clinicians’ overemphasizing the relevance of psychotic symptoms relative to mood symptoms among African Americans, compared with other racial or ethnic groups. For example, a study by Strakowski and colleagues (2) focused on a group of 79 patients diagnosed according to expert consensus as having mood disorders. These patients were significantly more likely to be diagnosed as having schizophrenia spectrum disorders by unblinded assessors who used structured interviews than by expert diagnosticians who used transcripts of the same interviews but who were blind to patients’ race-ethnicity (racial-ethnic cues were eliminated from transcripts). In addition, the unblinded clinician evaluators, in sharp contrast to the blinded ones, assessed a greater number of first-rank psychotic symptoms among African-American versus white patients. This finding suggested that the psychotic symptoms were given more weight in the unblinded assessment. Finally, in a comparison of clinical and structured interview assessments by Strakowski and colleagues (10), it was found that mood symptoms identified in the structured interviews were more commonly not recorded in the clinical assessments of African Americans, compared with the assessments of white patients.

Another study extended the work of Strakowski and colleagues through multiethnic comparisons at six racially diverse regional sites in the United States by using multiple standardized assessments under two conditions: blinded and unblinded to race-ethnicity (4). This large-scale study (N=610) also sought to determine whether apparent biases in diagnostic assignment pertained to all people of color or were specific to African Americans. It was found that African American and white individuals did not differ significantly in blinded expert ratings of depression and manic symptoms; however, African Americans received higher ratings of positive symptoms, particularly hallucinations and delusions, but not bizarre behavior. These observations suggested that among African Americans, psychotic symptoms were excessively weighted by clinicians, skewing diagnoses toward schizophrenia spectrum disorders, even when the patients had ratings of depression and manic symptoms similar to those of white patients. These results extended and validated previous studies and supported our hypothesis. They further suggested that exclusionary criteria in the diagnosis of schizophrenia may not be correctly applied. Effects were robust in that they persisted across different types of assessments and whether or not the patients given diagnoses of schizophrenia spectrum disorders included those diagnosed as having schizoaffective disorder. Finally, these results also suggested that this diagnostic effect was specific to African Americans because it did not extend to Latino patients.

The purpose of this study was to extend this prior work into a large clinical system to determine whether clinicians failed to effectively weigh mood symptoms when diagnosing schizophrenia or schizoaffective disorder among African Americans. The opportunity for this study arose fortuitously as a result of Rutgers University Behavioral Health Care’s (RUBHC’s) participation in the certified community behavioral health clinics (CCBHC) initiative, which required, among other things, that RUBHC screen all outpatients for depression. It was predicted that African Americans with schizophrenia or schizoaffective disorder would be more likely than similarly diagnosed whites to screen positive for major depression.

Methods

Electronic medical record (EMR) data were obtained from several RUBHC outpatient offices in Middlesex County, New Jersey, that became CCBHCs on July 1, 2017. The CCBHCs were located in Edison, New Brunswick, and Monmouth Junction. Launching CCBHCs across nine states, including New Jersey, was an initiative funded by the Substance Abuse and Mental Health Services Administration. CCBHCs are specifically designed to provide a community with an all-inclusive range of evidence-based services for mental and substance use disorders, especially for individuals who have the most complex needs. Such services include medication, evidence-based psychotherapy, case management, physical health screening (for example, blood pressure and body mass index), primary care referral, and vocational and housing services. CCBHCs also have specific quality objectives and associated metrics designed to place a renewed focus on person-centered care as well as the quality of the care delivered.

Because this was a retrospective study based on data gathered in the EMR as part of the normal clinical operations of the CCBHC clinic, the study was exempt from institutional review board review. Personal health information was strictly protected, and individual patient data were concealed in the aggregate analysis. As part of a commitment to use high-quality metrics for the CCBHC, the RUBHC adopted several performance measures in the outpatient clinics. One measure in particular provided an opportunity to unobtrusively determine whether clinicians appropriately weighed mood symptoms when making a diagnosis in the schizophrenia spectrum. That is, clinicians were required by the CCBHC to screen new outpatients (or returning outpatients evidencing a new episode) for depression at the first visit, by using the nine-item Patient Health Questionnaire (PHQ-9) (14). Important for this study, screening for depression was to take place regardless of diagnosis. Primary ICD-10 diagnoses as entered in the EMR by attending psychiatrists (not residents) were used to operationalize “diagnosis” in this study.

To test the hypothesis that African Americans with diagnoses of schizophrenia or schizoaffective disorders were more likely than similarly diagnosed non-Latino whites to screen positive for major depression, we used Fisher’s exact test to test the difference between the percentages of African Americans and non-Latino whites with a schizophrenia diagnosis who also screened positive for major depression on the PHQ-9. Because diagnostic disparities have been found to be specific to African Americans, compared with whites, and have not been observed among Latino patients (4), we did not include Latino patients in statistical testing, regardless of their race. Finally, logistic regression analysis was conducted to examine the impact of covariates, such as gender and age.

It is important to note that the cutoff score (≥10) generally used for the PHQ-9 tends to screen as positive persons who may be only mildly depressed and persons who present with negative symptoms common in schizophrenia, such as low motivation and anhedonia. Therefore, we used a PHQ-9 cutoff score of ≥15, which screens positive individuals with moderately severe to severe major depression (14).

Results

The CCBHC outpatient sample included 599 African Americans and 1,058 non-Latino whites. There were 246 (41%) African-American males and 353 (59%) African-American females in the sample and 508 (48%) non-Latino white males and 550 (52%) non-Latino white females. Table 1 presents data for each group on the number of patients screening negative and positive for major depression (moderately severe to severe) in four diagnostic groups (mood disorders, schizophrenia, schizoaffective disorder, and other). The percentage of African Americans diagnosed as having schizophrenia who also screened positive for major depression (N=8 of 43, 19%) was significantly larger (p=.003, Fisher’s exact test) than the percentage of whites diagnosed with schizophrenia who also screened positive for major depression (N=2 of 80, 3%). In other words, among patients diagnosed as having schizophrenia, African Americans were significantly more likely than non-Latino whites to screen positive for major depression. The between-group difference in positive depression screens among those with schizoaffective disorder was not significant (African Americans, 13%; non-Latino whites, 10%).

In a logistic regression (Table 2), we examined a binary dependent variable (scored 1) for a diagnosis of schizophrenia comorbid with moderately severe to severe depression (PHQ score ≥15) and otherwise (scored 0). The independent variables were race-ethnicity (African American and non-Latino white), age, and gender. The effect of race-ethnicity remained significant (Wald χ²=6.74, df=1, p<.01), whereas age and gender were nonsignificant covariates, despite the fact that the proportion of males was larger among whites than among African Americans (48% versus 41%) and the sample of whites was older than the sample of African Americans (mean±SD=45±16.8 versus 39±15.4).

Discussion

As hypothesized, we found that among individuals diagnosed as having schizophrenia, African Americans were significantly more likely than non-Latino whites to screen positive for moderately severe to severe depression. This effect was not found for schizoaffective disorder and appeared to be limited to schizophrenia, presumably because an appreciation of mood symptoms is built into the diagnosis of schizoaffective disorder.

These results are consistent with findings from a six-site study (4) and the hypothesis by Strakowski and colleagues (2, 9–13, 15) suggesting that racial bias in the diagnosis of schizophrenia in part results from clinicians underemphasizing the relevance of mood symptoms among African Americans compared with other racial or ethnic groups. Why this phenomenon occurs remains unknown, but it may represent either conscious or unconscious bias toward the meaning of psychiatric symptoms of African Americans or the effects on clinical presentation of being a patient from a racial-ethnic minority group (for example, mistrust leading to less than full disclosure of presenting problems) (8).

The disparity between whites and African Americans in the diagnosis of schizophrenia when a major mood disorder may also be present is troubling, given that schizophrenia is by definition a diagnosis of exclusion. That is, schizophrenia requires other potential causes of presenting symptoms to be ruled out, including and especially mood disorders, before the diagnosis of schizophrenia is applied. An alternative explanation is that African Americans with schizophrenia have higher rates of co-occurring major depression that is secondary to or independent of a primary diagnosis of schizophrenia. To our knowledge, there are no epidemiological or other reports supporting this possibility. Nonetheless, to discount this possibility, it would be necessary to clarify the temporal relationships between the psychotic and mood symptoms for these individuals, which was not possible from the records available.

Inaccurate diagnosis can have serious consequences. Treatments for mood disorders or substance use disorders differ from those for schizophrenia in many important aspects, and the prognosis for these conditions is typically more positive than for schizophrenia. Selecting ineffective treatments for such wrongly diagnosed patients can produce such consequences as prolonged patient frustration, treatment nonadherence or dropout, delays in treatment response, and worsening morbidity and suicide (16–25). Moreover, failure to recognize alternative diagnoses may lower treatment expectations, leading to inadequate breadth and depth of therapeutic interventions. Because the side effects of antipsychotic medications can be serious (for example, metabolic syndrome, weight gain, diabetes, and movement disorders), the risks of serious side effects cannot outweigh the benefits when a diagnosis is incorrect.

This study was retrospective in nature and thus had the advantage of being unobtrusive. That is, neither clinicians nor patients were aware of being studied or of the nature of the hypothesis explored. As for limitations, the data collection arose from the organization’s desire to be compliant with the CCBHC performance measures requirement. We had no input regarding what data were to be collected. Had we explicitly designed a study to test the main hypothesis, we might have insisted on the inclusion of many other covariates, such as clinician ratings of psychotic symptoms; assessments of temporal relationships between psychotic and affective symptoms; measures of poverty, discrimination, and nutrition; and other variables suggested by Tortelli and colleagues (6), as well as the use of structured interviews to diagnose major depressive disorder. It should be noted that ratings of psychotic symptoms and structured interviews were in fact used in the six-site study (4). Also, prior work suggests that about half to two-thirds of patients with PHQ-9 scores of 15 or more actually meet criteria for major depression disorder (26). Therefore, we cannot be sure whether a person who screened positive for major depression in this study would have also met ICD-10 criteria for the disorder, and we cannot rule out racial differences in responses to this self-report instrument.

It is also unknown whether these results would generalize to comparable clinics. However, it is possible in principle to replicate this study in CCBHCs across the United States because all are required to screen for depression.

Conclusions

A case could be made that routine screening for major depression in a setting such as ours could reduce racial inequities in schizophrenia diagnoses, given the results of this study. Such screening might help because the cooccurrence of schizophrenia and major depression seems to be overrepresented only among African Americans. If physicians screened every patent with putative schizophrenia for major depression, they might find, after further scrutiny of symptoms and functioning, that some patients would ultimately not meet ICD-10 criteria for major depression. However, some patients would meet these criteria, and in these cases the physicians might consider diagnostic alternatives, such as major depression with psychotic features, schizoaffective disorder, or bipolar disorder if there is a history of manic or hypomanic symptoms. Regardless, careful follow-up and assessment following a positive PHQ-9 screen might lead to better personalized treatment and ultimately better clinical outcomes. We hope that this report will encourage clinical sites to consider these types of interventions.