Inequalities in the Incidence of Psychotic Disorders Among Racial and Ethnic Groups

The primary sample consisted of members who received care at KPNC any time between January 1, 2009, and December 31, 2019. A retrospective cohort was constructed (N=5,994,758), for which at least 1 year of continuous enrollment was required. Follow-up started on January 1, 2009, or after 1 year of continuous enrollment, whichever came later. Individuals who received a psychotic disorder diagnosis before the follow-up period were excluded. Incidence was calculated by using the date of the first diagnosis as the date of onset. KPNC is a large, integrated health care delivery system that provides comprehensive medical care to more than 4 million individuals. The population of KPNC members is generally representative of the overall regional population, although income distribution extremes may be underrepresented (25). The study design was reviewed and approved by the KPNC institutional review board and included a waiver of informed consent, because the research involved no more than a minimal risk to participants.

Electronic health records were used to extract demographic and clinical information, including age at study entry, sex, race/ethnicity, address, body mass index, marital status, employment status, service utilization, psychiatric diagnoses, and negative outcomes or comorbidities identified during outpatient and inpatient visits. Race and ethnicity were self-identified and categorized for analysis as African American or Black (hereafter “Black”), American Indian or Alaskan Native (hereafter “AIAN”), Native Hawaiian or other Pacific Islander (hereafter “NHPI”), Latino or Hispanic (hereafter “Hispanic”), non-Hispanic White (hereafter “White”), Asian American (hereafter “Asian”), or “other” (which included multiple races or races that do not fit well in any other category and instances where race and ethnicity were not reported). Psychiatric and physical comorbidities were identified based on diagnostic codes from outpatient and inpatient visits. We calculated a dichotomous variable for whether participants had one or more health care visits (inpatient or outpatient) during the year prior to entering the study as a measure of health care access (26). Because individual-level indicators of socioeconomic status were not available (KPNC does not request information about socioeconomic status or parental education), a census-tract-level neighborhood deprivation index (NDI), created by Messer et al. (27), was calculated based on the member’s address using eight census indicators from the 2019 American Community Survey, with higher NDI values indicating greater deprivation or lower neighborhood socioeconomic status. Demographic characteristics were continually updated in the KPNC system until the data were extracted, and thus the temporal ordering of covariates in relation to outcomes was unknown.

We identified psychotic disorder diagnoses from inpatient and outpatient encounters based on ICD-9 and ICD-10 diagnostic codes recorded in the KPNC system, including nonaffective psychotic disorders (ICD-10: F20.x, F22.x, F23.x, F24.x, F25.x, F28.x, and F29.x; ICD-9: 295.x, 297.x, and 298.x) and affective psychotic disorders (ICD-10: F30.2, F31.2, F31.5, F31.64, F32.3, and F33.3; ICD-9: 296.04, 296.14, 296.24, 296.34, 296.44, 296.54, and 296.64). These case definitions are similar to those used in previous population-based registry studies (28) and were intentionally broad (e.g., unspecified diagnoses were included) to capture the full range of psychosis illness. Diagnosis of psychotic disorders in the KPNC system is typically based on medical history, chief concerns, mental status examination, functional ability, and a clinical interview with a qualified, licensed mental health clinician, which included a history of present illness, psychiatric review of systems, psychiatric history, family psychiatric history, substance use history, and social history. Although diagnoses in the KPNC system may be made by various types of health care providers, a pilot study conducted among 100 randomly selected patients with at least five psychotic disorder diagnoses found that 97% of diagnoses were made by a licensed mental health practitioner (results available upon request).

We estimated annual incidence of affective and nonaffective psychotic disorders, diagnosed by licensed mental health clinicians, among race/ethnicity groups for each year in the study period (2009–2019). Unadjusted Poisson regression models with a log-link function were used to assess overall linear changes in annual incidence rates over time and linear changes by race/ethnicity during the study period. The statistical significance of these trends was tested with Wald tests.

Multivariable Cox proportional hazards regression analyses were conducted to estimate hazard ratios for outcomes of incident diagnosis of affective and nonaffective psychotic disorders by race/ethnicity (with White participants serving as the reference group because they represented >40% of cases) and other demographic factors. We first present the association between each demographic covariate and psychosis, adjusted for age and sex only (referred to as model 1). In addition to age and sex, model 2 also adjusted for marital status, employment status, health care utilization, and neighborhood deprivation. We then present logistic regression results that assessed the association of incident psychotic disorder diagnosis with other psychiatric diagnoses, including depressive, bipolar, anxiety, personality, and alcohol and drug use disorders, posttraumatic stress disorder (PTSD), and pervasive developmental disorders, occurring at any time during the study period. Incident psychotic disorder diagnosis was the predictor in these models, and each comorbid diagnosis was modeled as a separate outcome. Model 1 was adjusted for age and sex only. Model 2 was also adjusted for other psychiatric diagnoses occurring during the study period. Model 3 was further adjusted for race/ethnicity, marital status, employment status, and NDI.

Finally, we examined associations of psychotic disorder diagnosis with health consequences and comorbidities, including premature death (defined as death occurring before the average age of death in the U.S. population), obesity, diabetes, smoking status, sexually transmitted infection, HIV, stroke, coronary artery disease, myocardial infarction, and service utilization, by using logistic regression analyses with psychotic disorder diagnosis as the predictor. Model 1 was adjusted for demographic characteristics, including age at study entry, sex, race/ethnicity, marital status, employment status, and NDI. Model 2 was further adjusted for psychiatric comorbidities, including depressive, bipolar, anxiety, psychotic, personality, alcohol use, and drug use disorders, PTSD, and pervasive developmental disorders.

Statistical analyses were performed with SAS, version 9.4 (SAS Institute, Cary, N.C.). The p values from both multivariable Cox proportional hazards models and logistic regression models were calculated based on Wald chi-square statistics. Multiple comparisons were not accounted for, and therefore nominal p values are reported. A two-sided p value less than 0.05 was considered significant.

Table 1 shows the distribution of demographic characteristics and psychiatric comorbidities by psychotic disorder diagnosis. Among 5,994,758 KPNC members cared for between January 1, 2009, and December 31, 2019, 0.22% (N=13,303) were diagnosed with a nonaffective psychotic disorder, and 0.098% (N=5,888) were diagnosed with an affective psychotic disorder.

The overall incidence of nonaffective psychosis diagnosis (N=13,303) per 10,000 person-years was 2.84 in 2009 and 2.33 in 2019, with a peak of 3.92 in 2013, and the overall incidence of affective psychosis diagnosis (N=5,888) per 10,000 person-years was 1.25 in 2009 and 1.00 in 2019, with a peak of 1.86 in 2016. Figure 1 shows yearly incidence rates of nonaffective and affective psychosis diagnoses over the study period for each racial/ethnic group. Black participants generally had the highest incidence rates, and Asian participants had the lowest incidence rates, of both nonaffective and affective psychosis diagnoses throughout the study period. Specifically, the incidence of nonaffective psychosis diagnosis per 10,000 person-years among Black participants was 5.99 in 2009 and 4.91 in 2019, with a peak of 8.82 in 2015, and the incidence of affective psychosis diagnosis per 10,000 person-years among Black participants was 2.28 in 2009 and 1.99 in 2019, with a peak of 4.02 in 2015. The incidence of nonaffective psychosis diagnosis per 10,000 person-years among Asian participants was 2.19 in 2009 and 1.47 in 2019, with a peak of 2.88 in 2011, and the incidence of affective psychosis diagnosis per 10,000 person-years among Asian participants was 1.02 in 2009 and 0.77 in 2019, with a peak of 1.60 in 2017.

Overall, there was evidence for a slightly decreasing linear trend in incidence of nonaffective psychosis diagnosis over time (incidence rate ratio=0.990, 95% CI=0.98–1.00, p=0.0002), which was also observed among White (incidence rate ratio=0.981, 95% CI=0.97–0.99, p≤0.0001) and Asian (incidence rate ratio=0.976, 95% CI=0.96–0.99, p=0.00198) members. Overall, no linear trend in incidence of affective psychosis over time was observed (incidence rate ratio=1.006, 95% CI=0.99–1.01, p=0.1188). Among members whose race/ethnicity was categorized as “other,” there was evidence for slightly increasing linear trends in diagnosis of both nonaffective psychosis (incidence rate ratio=1.033, 95% CI=1.05–1.05, p=0.0009) and affective psychosis (incidence rate ratio=1.043, 95% CI=1.01–1.08, p=0.0063). No other significant linear trends were observed.

Table 2 shows associations between demographic characteristics and incident nonaffective and affective psychosis diagnosis: model 1 adjusted for age and sex, and model 2 adjusted for age, sex, race/ethnicity, neighborhood deprivation, marital status, employment status, and health care utilization. Note that these covariates may capture mechanisms for racial and ethnic differences in health, such as socioeconomic status. Furthermore, because the demographic information in the KPNC system was continually updated, temporal ordering between some covariates and other study variables is unknown. We therefore focus our interpretation on model 1, although model 2 results are also provided for readers interested in associations accounting for other covariates. In model 1, compared with White participants, both Black and AIAN participants had a higher risk of nonaffective psychosis diagnosis (Black: hazard ratio=2.13, 95% CI=2.02–2.24, p<0.001; AIAN: hazard ratio=1.85, 95% CI=1.53–2.23, p<0.001). In contrast, Hispanic, Asian, and “other” participants had lower risk compared with White participants (Hispanic: hazard ratio=0.91, 95% CI=0.87–0.96, p<0.001; Asian: hazard ratio=0.72, 95% CI=0.68–0.76, p<0.001; “other”: hazard ratio=0.92, 95% CI=0.86–0.99, p=0.018). Risk of diagnosis among NHPI participants did not significantly differ from that of White participants (hazard ratio=0.85, 95% CI=0.69–1.06, p=0.152). In model 2, associations for Asian (hazard ratio=0.78, 95% CI=0.74–0.83, p<0.001), Black (hazard ratio=1.73, 95% CI=1.64–1.82, p<0.001), NHPI (hazard ratio=0.82, 95% CI=0.66–1.02, p=0.074), and AIAN (hazard ratio=1.67, 95% CI=1.38–2.02, p<0.001) participants were similar to those calculated with model 1, but slightly reduced in magnitude. The lower risk of diagnosis among Hispanic participants compared with that among White participants was slightly higher in model 2 (hazard ratio=0.85, 95% CI=0.81–0.90, p<0.001), and the risk among “other” participants was not significantly different from that among White participants (hazard ratio=1.03, 95% CI=0.96–1.10, p=0.453).

Results for affective psychosis diagnosis were similar in direction to results for nonaffective psychosis diagnosis for Black (model 1: hazard ratio=1.76, 95% CI=1.62–1.91, p<0.001; model 2: hazard ratio=1.43, 95% CI=1.32–1.56, p<0.001), Asian (model 1: hazard ratio=0.77, 95% CI=0.71–0.83, p<0.001; model 2: hazard ratio=0.84, 95% CI=0.78–0.92, p<0.001), and “other” (model 1: hazard ratio=0.86, 95% CI=0.77–0.96, p=0.006; model 2: hazard ratio=1.00, 95% CI=0.90–1.12, p=0.947) participants compared with White participants. However, the higher risk among Black participants compared with White participants was less pronounced for affective psychoses than for nonaffective psychoses, based on nonoverlapping confidence intervals. Risk of affective psychosis diagnosis was significantly higher in model 1, but not in model 2, for Hispanic (model 1: hazard ratio=1.09, 95% CI=1.02–1.16, p=0.016; model 2: hazard ratio=1.01, 95% CI=0.94–1.08, p=0.79) and AIAN (model 1: hazard ratio=1.38, 95% CI=1.00–1.90, p=0.048; model 2: hazard ratio=1.26, 95% CI=0.91–1.74, p=0.159) participants, and the risk was lower among NHPI participants (model 1: hazard ratio=0.69, 95% CI=0.48–0.99, p=0.046; model 2: hazard ratio=0.67, 95% CI=0.47–0.97, p=0.035).

Table 3 shows associations between incident psychosis diagnosis and other psychiatric disorders occurring at any time during the study period. When the model was adjusted for age and sex (model 1), patients with psychotic disorder diagnoses had higher odds of receiving any of the other psychiatric diagnoses assessed here, with odds ratios ranging from 4.83 (95% CI=4.45–5.25) for pervasive developmental disorder to 38.58 (95% CI=37.35–39.86) for bipolar disorder. After additional adjustment for the other psychiatric diagnoses (model 2), odds ratios were greatly reduced in magnitude, but still elevated, for all diagnoses. The largest odds ratios were observed for bipolar disorder (odds ratio=7.54, 95% CI=7.25–7.84), drug use disorder (odds ratio=3.26, 95% CI=3.13–3.39), and depressive disorder (odds ratio=3.00, 95% CI=2.88–3.12). Further adjustment for demographic characteristics (model 3) slightly changed the magnitude of odds ratios without altering inferences.

Table 4 shows associations between incident psychosis diagnosis and negative outcomes occurring at any time during the study period. In the model adjusted for age and sex only, individuals with psychotic disorder diagnoses had higher odds of all outcomes investigated. After additional adjustment for psychiatric comorbidities (model 2), significant positive associations were observed for death by suicide (odds ratio=2.51, 95% CI=2.03–3.10, p<0.001), premature death (odds ratio=1.53, 95% CI=1.43–1.63, p<0.001), stroke (odds ratio=1.94, 95% CI=1.84–2.04, p<0.001), coronary artery disease (odds ratio=1.28, 95% CI=1.20–1.36, p<0.001), myocardial infarction (odds ratio=1.30, 95% CI=1.20–1.41, p<0.001), obesity (odds ratio=1.13, 95% CI=1.09–1.17, p<0.001), and diabetes (odds ratio=1.36, 95% CI=1.30–1.43, p<0.001). In contrast, significant negative associations were observed for smoking (odds ratio=0.76, 95% CI=0.73–0.79, p<0.001) and service utilization (odds ratio=0.50, 95% CI=0.48–0.53, p<0.001). Further adjustment for other demographic characteristics (model 3) altered the magnitude of some odds ratios but did not change inferences.

In this study of over 5 million KPNC health plan members, we found evidence for a slightly decreasing linear trend in the incidence of treated psychosis between 2009 and 2019, although the underlying trends may be curvilinear. The incidence rates confirm those reported from several other large studies of both community and hospital samples in U.S. and European registries (29) and confirm a lack of consistent changes in incident psychotic disorder diagnosis rates over previous time periods in the United States (30). We found that Black participants generally had the highest rates of incident psychotic disorder diagnosis throughout the 10-year study period. In contrast to some previous U.S.-based studies (31–33), Hispanic participants had a lower risk of incident nonaffective psychotic disorder diagnosis compared with White participants, and Asian participants had the lowest rates of incident psychotic disorders. Likewise, schizophrenic disorders were more common among Black participants and less common among Hispanic participants, compared with White participants, prior to adjustment for current sociodemographic characteristics in the Epidemiologic Catchment Area study, a direct interview study of a nationally representative sample of the United States conducted over 30 years ago (34). Our study is the first to show that, compared with White participants, AIAN participants are at higher risk of incident nonaffective psychosis and that Asian participants are at lower risk of affective and nonaffective psychosis. Although Black participants had the highest risk of both nonaffective and affective psychosis, the excess risk compared with that of White participants was less pronounced for affective psychosis compared with nonaffective psychosis, consistent with previous findings that Black participants are more likely to be diagnosed with nonaffective psychosis and less likely to be diagnosed with affective illness (5).

This study of treated incidence cannot address whether disparities in diagnosis reflect true differences in the risk of psychotic disorder, noncausal factors, such as misdiagnosis and differential treatment access and utilization, or a combination of the two. In addition, the relative influence of these factors may differ across the groups considered. With respect to noncausal factors, although differences in treatment access related to insurance status might be minimized among KPNC members, enabling resources may be unequally distributed among groups, and differences in attitudes toward mental health treatment and expression of emotional distress may lead to lower rates of clinical diagnosis among some groups. Racial/ethnic differences could also reflect bias in the perception of symptoms across patients of different races (35–37), the underdiagnosis of mood disorders among Black patients (38), or the unconscious bias that has been documented in studies evaluating the influence of a clinician’s race (39).

Furthermore, true differences in the risk of psychotic disorder may exist. As recently reviewed by Anglin et al. (7), systemic racism in the United States has resulted in the unequal distribution of social, economic, and health-related disadvantages across racial and ethnic groups, resulting in greater exposure to risk factors for psychosis among Black Americans. Indeed, racism has been proposed as a fundamental cause of health inequalities that operates largely, but not entirely, through racial inequalities in socioeconomic status (40). Similarly, explanations for higher risk of psychosis among immigrant and ethnic minority groups in Europe have centered on the effects of social disadvantage, rather than noncausal explanations such as clinical bias (22, 23, 41). Such explanations deserve greater attention in the U.S.-based literature given the pervasive and cross-cutting nature of health disparities in the U.S. context.

Our results regarding comorbidity of psychotic disorders across all diagnostic categories were consistent with patterns of comorbidity in community-based samples (9). Some of the “comorbidity” observed here may more accurately reflect heterotypic continuity, clinical phenomena that occur during earlier stages of illness before the disorder fully manifests (42). Nonetheless, the pervasive comorbidity and association between comorbid conditions with poorer outcomes document the need for comprehensive care across mental health, medical, and substance abuse domains.

Our findings of increased suicide, premature death, cardiovascular disease, and diabetes and lower odds of health care utilization, even after adjustments for comorbidities, confirm the results of previous population-based studies documenting the excessive morbidity and mortality associated with a psychotic disorder diagnosis (43, 44). The fact that outcomes such as premature death, stroke, coronary artery disease, and diabetes, in this study and in previous research (45), exhibit racial and ethnic disparities in the United States further underscores the unequal health burden faced by some disadvantaged groups. Our finding of lower odds of smoking among patients with a psychotic disorder diagnosis, after adjustment for other psychiatric comorbidities, does not confirm the consistent association between smoking and schizophrenia observed in clinical data. The link between smoking and psychosis in previous research could reflect the influence of other psychiatric conditions that were not controlled for in those studies (20).

The strengths of this study include the availability of comprehensive clinical information and the large sample size that provided the power to examine ethnic and racial subgroups included in this uniquely diverse Northern California population. Nevertheless, a key limitation of this study is the use of broad racial and ethnic groups that may mask important differences between smaller groups (46). In addition, we did not have direct measures of disorder onset and instead relied on administrative records, which do not constitute validated cases. We also cannot rule out possible bias in our estimates due to informative censoring. Furthermore, our sample was derived from a single health care system in a specific U.S. state, and the findings may not generalize to other regions. Previous work has shown that the overall racial/ethnic composition of KPNC members does not differ from nonmember residents of California, although there is a lower percentage of Latinas among women (25). Very-low-income and high-income Californians were also underrepresented in the KPNC system, and KPNC members had higher average education and were more likely to be employed than nonmembers (25). Thus, some of the racial and ethnic differences we observed could potentially be attributed to lower case ascertainment among those with lower socioeconomic status. Because demographic information is continually updated in the KPNC system, we were unable to determine temporal ordering between psychosis and covariates that change over time, such as employment status. We also lacked information that could have enriched the study, such as measures of acculturation, immigrant status, and experiences of discrimination. Finally, a comprehensive investigation of racial and ethnic differences in diagnosis across the domains of mental and physical health disorders was beyond the scope of this study; however, such comparisons may be useful in evaluating explanations for differences in rates of diagnosis.