Disparities in mental health services and outcomes among persons from racial minority groups across a range of psychiatric conditions (
1,
2), including schizophrenia, have been widely documented (
3). These disparities have been particularly prominent among African Americans, who receive fewer mental health services (
4), more coercive methods of care (
5–
8), and poorer quality of treatment (
9,
10). Not surprisingly, such differences in the quality and nature of mental health services received by African Americans have led to significantly poorer outcomes (
11).
One of the most striking and consistent disparities among African Americans seeking mental health services is the increased diagnosis of schizophrenia (
12–
17). This phenomenon exists despite the absence of unique genetic evidence to indicate a true increase in prevalence in this population (
18,
19), although some increase in prevalence may be due to environmental risk factors (
20). Bias in the diagnosis of schizophrenia among African Americans has the potential to have a monumental impact on care, given that approaches indicated for treatment of psychotic versus mood disorders vary considerably and the receipt of an unneeded antipsychotic treatment regimen may have severe adverse consequences (
21). As such, an important priority in reducing disparities in mental health outcomes among African Americans has been understanding and reducing diagnostic bias.
Unfortunately, remarkably little is known about the factors or mechanisms that contribute to the overrepresentation of schizophrenia diagnoses affecting the African-American community. Efforts to pinpoint the sources of diagnostic bias in this population have primarily examined differences in the sociodemographic characteristics and the clinical presentations of individuals seeking care. After adjustment for differences in socioeconomic status (
12,
22) and symptomatology (
17), these studies have continued to find increased diagnoses of schizophrenia among African Americans.
Recently, investigators of potential racial biases in the diagnosis of schizophrenia have begun to turn their attention toward clinician behavior and characteristics. Evidence suggests that clinicians use different symptom criteria when making schizophrenia diagnoses among African Americans (
23), a practice that is not remediated by the use of a structured clinical interview (
24). Further, increased diagnosis of schizophrenia among African Americans by African-American and white clinicians appears to be equally prevalent (
25). Clinicians also appear to perceive African-American clients as more paranoid and suspicious in general (
17,
26), which could contribute to elevated rates of psychotic diagnoses in this population. Indeed, recent popular literature has extensively discussed how schizophrenia became a “black” diagnosis, in part because of cultural mistrust between the two racial groups and what whites perceived as obstinacy by blacks fighting for civil rights (
27). Currently, however, few empirical data exist about the impact of the patient-clinician relationship, such as clinician perceptions of a patient's honesty and trust, on disparities in the diagnosis of schizophrenia among African Americans.
This study investigated the degree to which sociodemographic characteristics, clinical presentation, and interviewer perceptions of patient honesty served as mechanisms of racial disparities in diagnostic rates of schizophrenia. We hypothesized that differences in the sociodemographic characteristics and clinical presentation of African Americans and whites upon hospital admission would contribute to the overrepresentation of African Americans among patients diagnosed as having schizophrenia. In addition, we hypothesized that independent of any differences in clinical presentation, perception by the interviewer of the honesty of the patient would also contribute significantly to disparities in schizophrenia diagnoses.
Methods
Participants
Participants consisted of 752 individuals with a major depressive disorder (57%, N=432), schizophrenia or schizoaffective disorder (26%, N=198), or bipolar disorder (16%, N=122) enrolled in the multisite MacArthur Violence Risk Assessment Study (
28). Completed in 1995, data collection by the study involved uniquely in-depth research assessments of a large number of participants, including assessments by the research interviewers of their interactions with participants. The study oversampled African Americans and collected both independent research data and diagnostic data from participants' charts. Together these data allowed for a comprehensive examination of racial disparities in the diagnosis of schizophrenia using multiple methodologies.
Individuals were recruited from psychiatric inpatient units located in the major metropolitan areas of Pittsburgh, Pennsylvania; Worcester, Massachusetts; and Kansas City, Missouri. Eligibility criteria consisted of a DSM-III-R diagnosis in the participant's chart of a severe mental illness (schizophrenia, schizophreniform disorder, schizoaffective disorder, major depressive disorder, dysthymic disorder, bipolar disorder, brief reactive psychosis, delusional disorder, substance abuse or dependence, or personality disorder), current hospitalization of fewer than 21 days, age 18 to 40 years, and the ability to read and speak fluent English. The study recruited 1,695 participants, and 1,136 (67%) were enrolled.
Because this study investigated racial disparities among patients with diagnoses of schizophrenia-spectrum versus major mood disorders, we included only the 752 individuals diagnosed as having schizophrenia or schizoaffective, major depressive, or bipolar disorder. The 21 Hispanic individuals in the sample were also excluded because of their limited numbers. On average, participants were 29.99±6.20 years old; half (54%, N=406) were male, and 29% (N=215) were African American. Over half of the individuals (58%, N=438) had never been married, most (75%, N=561) had been hospitalized for a psychiatric condition previously, and a majority (72%, N=542) were currently hospitalized voluntarily.
Measures
Psychiatric diagnosis.
Principal psychiatric diagnoses were obtained by using two separate methods. First, a thorough review of the psychiatric records of the participant was conducted to obtain the latest clinic diagnosis. Second, participants completed a research diagnostic interview using the DSM-III-R Criteria Checklist (
29), a structured interview and checklist of diagnostic symptoms of
DSM disorders. The interview covers all axis I diagnostic categories and, most important, does not allow interviewers to circumvent certain diagnostic categories on the basis of their own judgments or information they receive during the interview. This is accomplished by requiring interviewers to review all inclusion and exclusion criteria for all axis I disorders in the
DSM during the interview. The DSM-III-R Checklist has been validated for diagnosis of psychiatric disorders (
29–
31) and has been used successfully in previous studies of racial disparities in psychiatric diagnosis (
23).
Psychiatric symptomatology.
Psychopathology was assessed by using the Brief Psychiatric Rating Scale (BPRS) (
32) and the Auditory Hallucinations Schedule (
28). The BPRS is a widely used measure of psychiatric symptomatology that has good psychometric properties (
33), with higher scores reflecting greater levels of symptomatology. The Auditory Hallucinations Schedule, a semistructured interview designed to characterize auditory hallucinations among individuals with psychotic conditions, was used to establish the presence of auditory hallucinations. Finally, the presence of delusions was assessed by the DSM-III-R Checklist.
Perceived honesty.
To obtain information on the interaction between the interviewer and the participant during diagnostic and clinical assessments, interviewers completed a questionnaire after conducting each interview. This questionnaire was originally completed for quality assurance purposes and consisted of ratings of the perceived honesty of the participant during the interview. Such information provided a unique opportunity to begin to examine the impact of the patient-clinician interaction on diagnostic decisions. Perceived honesty was rated on a 5-point scale, ranging from 1 (honest) to 5 (untruthful), in response to the question, “Did the subject appear to be answering honestly?” Separate ratings were obtained for the participant's responses during assessment of diagnosis, delusions, and hallucinations. Ratings of perceived honesty across participants were highly interrelated and, therefore, were averaged. The resulting interviewer-perceived honesty scale combining these ratings demonstrated excellent internal consistency (α=.87).
Socioeconomic status.
Hollingshead and Redlich's index (
34) was used to measure socioeconomic status by taking into account prehospitalization education and occupation.
Procedures
Research interviewers assessed the participants during their stay at one of the inpatient units. Careful attention was paid to ensure that individuals were capable of providing informed consent to participate in the study as well as capable of completing the study assessments. This study was approved and reviewed annually by the institutional review board of each study site, and all enrolled individuals provided written, informed consent prior to participation.
Data analysis
Disparities in rates of diagnosis of schizophrenia among African Americans and whites were investigated. Systematic differences in the demographic and clinical characteristics and in the interviewers' perceptions of honesty of participants of different races were examined with independent t tests or with Fisher's exact test (for categorical variables). A series of logistic regression models was used to examine whether variables that were significantly associated with race contributed to the effects of race on the diagnosis of schizophrenia. Finally, multiple mediation models using Baron and Kenny's (
35) approach examined whether sociodemographic, clinical, and perceived honesty characteristics that demonstrated significant associations with both race and a schizophrenia diagnosis mediated the effect of race on schizophrenia diagnosis. The indirect effect of race on outcome through the mediator was subsequently calculated by using the product-of-coefficients method (
35), and the significance of indirect effects was tested by using the asymptotic z test for indirect effects outlined by MacKinnon and colleagues (
36).
Results
Racial disparities
As shown in
Table 1, 19% of the white sample received a research diagnosis of schizophrenia or schizoaffective disorder compared with nearly half (45%) of the African-American sample. This difference represented over a threefold increase in the rates of schizophrenia spectrum disorders among African Americans compared with whites. Differences in the two groups' chart diagnoses were of a similar magnitude (data not shown). Significant differences between the races in a variety of sociodemographic and clinical characteristics were also present. Compared with whites, the African Americans in the sample were slightly older and had lower socioeconomic status. Further, they tended to experience greater rates of hallucinations, delusions, and substance use problems and had greater levels of psychiatric symptomatology, particularly thought disturbance. In addition, interviewers consistently perceived African Americans to be less honest than whites during diagnostic interviews.
Mechanisms of racial disparities
A series of logistic regression models were conducted to examine the degree to which disparities in sociodemographic and clinical characteristics contributed to the increased diagnosis of schizophrenia among African Americans. Results revealed numerous predictors of the diagnosis of schizophrenia, both from the chart review and the research diagnostic assessment (
Table 2). Older individuals who had never married were significantly more likely to be diagnosed as having schizophrenia, but socioeconomic status was an inconsistent predictor of a schizophrenia diagnosis. As expected, individuals experiencing delusions and other schizophrenia symptomatology, such as thought disturbance and anergia, were more likely to be diagnosed as having the disorder. However, even after adjustment for demographic and clinical characteristics that were highly significant predictors of a schizophrenia diagnosis, being African American continued to place individuals at an over threefold risk of receiving a research diagnosis of schizophrenia.
Given that clinical and demographic characteristics could not account for the increased diagnosis of schizophrenia among African Americans, we studied the contribution of perceived honesty during the research interview. As shown in
Table 2, individuals were nearly one-and-a-half times as likely to receive a schizophrenia diagnosis if the interviewer perceived them to be dishonest during diagnostic assessments. Further, after adjustment for the impact of perceived honesty, the effect of being African American on receiving a diagnosis of schizophrenia in one's chart was no longer statistically significant, and its effect on receipt of a research interview diagnosis of schizophrenia was markedly reduced.
Mediator analyses indicated that few sociodemographic or clinical characteristics significantly mediated the effect of race on the diagnosis of schizophrenia (
Table 3). The presence of auditory hallucinations significantly mediated the effect of race on research interview diagnoses, but not on chart diagnoses, of schizophrenia. The absence of substance abuse significantly mediated the effect of race on chart diagnoses, but not on research interview diagnoses, of schizophrenia. In contrast, interviewers' perceptions of participants' honesty represented the strongest and only consistent mediator of racial disparities in the diagnosis of schizophrenia by both diagnostic methodologies. Taken together, these findings indicated that clinician perceptions of honesty and the relationship between the interviewer and the participant were important and potentially unique contributors to racial disparities in the diagnosis of schizophrenia.
Discussion
Disparities in access, availability, and quality of mental health services and treatment among African Americans have been repeatedly documented (
1,
3). One of the most consistent disparities, an increase in the diagnosis of schizophrenia, has the potential to significantly impact mental health outcomes of this population (
16,
17). To date, however, few studies have examined the potential mechanisms driving this diagnostic disparity.
We investigated the degree to which sociodemographic and clinical characteristics and interviewer-perceived honesty independently served as mechanisms contributing to the diagnosis of schizophrenia among African Americans. As predicted, among the strongest contributors to racial disparities in the diagnosis of schizophrenia was whether the research interviewer perceived the participant to be honest and trustworthy during their interaction. Further, interviewer-perceived honesty proved to be the only consistent mediator of the increased diagnosis of schizophrenia among African Americans. These findings point to the potentially important contribution of the patient-clinician relationship to race-related disparities in the diagnosis of schizophrenia. This relationship appears to affect such disparities as much as, if not more than, the clinical presentation of the patient.
Studies of racial disparities in mental health diagnoses have consistently underscored the potential importance of the patient-clinician relationship (
17,
23,
25). This research is the first to document directly that perceived honesty during diagnostic interviews is an important mediator of racial disparities in the diagnosis of schizophrenia. The findings indicate that a fundamental challenge to psychiatric assessment of African Americans is developing a trusting, open, and collaborative relationship, which has been identified as essential to effective care of minority populations (
37,
38). Unfortunately, the context of a diagnostic assessment often poses some unique challenges to developing a trusting therapeutic relationship with patients, given that assessments are commonly conducted in a single session by a clinician who has only just met the patient and frequently must be rendered quickly to support an inpatient admission or other immediate referral. Future research efforts might profitably employ a longitudinal diagnostic assessment system that emphasizes developing an empathic trusting relationship between the patient and the clinician.
Although this research has several implications for how to address racial disparities in the diagnosis of schizophrenia, such implications need to be understood in the context of a number of limitations. First, assessments of patient-clinician trust and interactions were limited to completion by the interviewer of a single measure of the participant's perceived honesty, which was based on a small number of ratings from diagnostic and assessment interviews. Although this measure demonstrated good internal consistency, an important limitation was the absence of information about the participant's perceptions of the interaction. There are many reasons why a patient might be perceived as not completely honest and forthcoming, and without the perspective of the patient and further examination of the patient-clinician interaction, such reasons remain elusive.
One possibility is that African Americans were completely honest during the interviews but that clinicians held a racial bias in perceptions of honesty. It is also possible that the ratings of greater dishonesty among African Americans reflect not so much a perception of untruthfulness but an acknowledgment of the increased presentation of unusual or inconsistent information common in schizophrenia. It will be important for future studies to examine more comprehensively the patient-clinician relationship to clarify sources of mistrust and interviewer biases. In addition, this study examined perceived honesty among only research interviewers, and thus these findings may not generalize to nonresearch settings. Although interviewer-perceived honesty was related to increased diagnoses of schizophrenia both during structured research interviews and in chart diagnoses, future studies are needed to examine the impact of perceptions of honesty and the patient-clinician relationship outside a research setting.
Second, psychiatric diagnoses were based upon
DSM-III-R, the diagnostic standard at the time of the MacArthur study, which may have affected our findings. Although
DSM-IV introduced important changes, few differences exist between the diagnostic classification of schizophrenia in
DSM-III-R and the current version,
DSM-IV-TR. In addition, studies based on
DSM-IV also continue to document disparities in the diagnosis of schizophrenia among African Americans at rates similar to those observed by this study (
39). As such, these findings are not likely to be appreciably affected by changes in the diagnostic criteria for schizophrenia. Third, disparities in quality of care also could have contributed to the overrepresentation in this inpatient sample of African Americans diagnosed as having schizophrenia, although previous studies have also suggested racial disparities in schizophrenia diagnoses in outpatient settings (
14).
Finally, this study was limited to an inpatient sample, which may affect the generalizability of the findings to individuals receiving psychiatric assessments in outpatient settings. To date, a majority of studies have focused on the diagnosis of schizophrenia in inpatient or psychiatric emergency settings. Whether racial disparities in diagnoses exist in outpatient settings, where patients may have the opportunity to develop a more personal relationship with the diagnostician, remains an important question.
Conclusions
This study revealed that interviewer perceptions of patients' honesty are significant and strong mediators of racial disparities in the diagnosis of schizophrenia among African Americans. Despite limited assessment of the patient-clinician relationship and restriction to an inpatient sample, our findings highlight the importance of understanding and seeking ways to improve the clinician-patient working alliance with members of underserved ethnic and racial minority groups during psychiatric assessment to improve the care of such individuals during treatment.
Acknowledgments and disclosures
Dr. Eack has served as a scientific consultant for Abbott Laboratories. The other authors report no competing interests.