Epidemiology of Untreated Psychoses in 3 Diverse Settings in the Global South

This cohort study investigates if the clinical and demographic profiles of individuals with, and rates of, untreated psychoses vary across diverse settings in the Global South.

CONCLUSIONS AND RELEVANCE This analysis adds to research that suggests that core aspects of psychosis vary by historic, economic, and social context, with far-reaching implications for understanding and treatment of psychoses globally. T here are striking global inequities in our knowledge and treatment of psychotic disorders. More than 80% of the world's population lives in the Global South, which refers broadly to the regions of Latin America, Asia, Africa, and Oceania, but less than 10% of research on psychotic disorders has been conducted in these settings. 1,2 This evidence gap is important because the epidemiology, etiology, and outcomes of psychoses vary by context and social group. 1, 3,4 In most settings in the Global South, we lack core epidemiological evidence that can inform policy and development of accessible, humane, and effective services.
A systematic review identified 15 population-based studies of the incidence of psychoses in 9 low-and middleincome countries. 5 Drawing on our earlier review of studies and a systematic review of all incidence studies between 2002 and 2017, 1 we identified a further 9 studies outside of Europe, North America, and Australasia. Only 7 of these 26 studies were conducted in the past 20 years (2 in Brazil, 2 in Taiwan, 1 in Suriname, 1 in South Africa, 1 in Israel). 2 The methods used were heterogenous, and none were conducted in more than 1 country. The World Health Organization (WHO) Determinants of Outcome of Severe Mental Disorders (DOSMeD) study, the last study that included multiple sites in the Global South, was conducted more than 40 years ago. 6 To address these evidence gaps, we established INTREPID II (International Research Program on Psychotic Disorders in Diverse Settings), a program of research on psychoses in 3 countries (India, Nigeria, Trinidad) in the Global South.

Aim and Hypotheses
The aim of INTREPID II is to investigate the incidence, presentation, outcomes, physical health, and impacts of untreated psychotic disorders in 3 diverse settings: Kancheepuram District, Tamil Nadu, India; Ibadan, Oyo State, Nigeria; and northern Trinidad. In this article, we describe and compare core demographic and clinical characteristics of cohorts of individuals with an untreated psychosis identified in INTREPID II and present findings on rates of untreated psychoses (to approximate incidence) across and within settings. We test 2 hypotheses: (1) demographic and clinical profiles of cases with an untreated psychotic disorder vary across settings and (2) rates of untreated psychotic disorders vary across and within settings by clinical and demographic group.

Methods
INTREPID II, started October 1, 2017, and currently ongoing, comprises incidence, case-control, cohort, and qualitative studies of psychoses in 3 settings, based on the identification, assessment, and follow-up of cohorts of individuals with an untreated psychotic disorder (cases), of population-based matched controls, and of relatives of cases. The program was implemented in economically and socially diverse settings in 3 countries (eTable 1 in the Supplement). The catchment areas comprise urban and rural areas with populations of approximately 500 000 adults aged 18 to 64 years (eAppendix 1 in the Supplement). This study followed the Strengthening the Re-porting of Observational Studies in Epidemiology (STROBE) reporting guidelines.

Case Ascertainment
To estimate rates of untreated psychoses, we sought to identify all individuals aged 18 to 64 years with an untreated International Statistical Classification of Diseases and Related Health Problems, Tenth Revision (ICD-10) psychotic disorder (ie, not treated with antipsychotic medication for more than 1 continuous month) over a 2-year period in each catchment area. Individuals were excluded if there was evidence of ICD-10defined moderate/severe learning disability or organic cerebral disorder. These criteria mirror those in population-based studies in the Global North. 7,8 We used a multipronged approach to identify cases. 9 First, in each catchment area, we established case-detection systems by mapping and engaging service providers and community informants, covering all sectors of the local health care systems (ie, professional [mental health services], folk [traditional, spiritual healers], and popular [community informants]). 9 Second, we gave all mental health service professionals, traditional and spiritual healers, and informants materials that described experiences and behaviors characteristic of psychoses, using local terms and language, to facilitate shared understandings. 10 Third, trained researchers conducted regular checks with mental health professionals, traditional and spiritual healers, and informants to identify potential cases. Finally, in rural villages in Kancheepuram and Ibadan, field workers engaged community informants to identify potential cases. At the end of case ascertainment, we conducted leakage studies by rechecking service registers and completing final checks with mental health professionals, traditional and spiritual healers, and informants. All potential cases were screened using the Screening Schedule for Psychosis. All identified cases were approached to participate in all aspects of INTREPID II. Written informed consent was obtained from those who agreed to participate in the casecontrol arm of the program; otherwise, ethical approval was obtained from local research ethics committees in each setting to collate basic demographic and clinical information from records (eAppendix 2 in the Supplement). 8

Key Points
Question Do the clinical and demographic profiles of individuals with (and rates of) untreated psychoses (a proxy for incidence) vary across diverse settings in the Global South?
Findings In this population-based cohort study of 1038 individuals with untreated psychoses, results suggest that there were variations across settings in the Global South in clinical, sex, and age profiles and in rates of untreated psychotic disorders (eg, high rates in Trinidad vs India and Nigeria; in African vs Indian Trinidadian).
Meaning Findings of this study add to research that suggests that core aspects of psychosis vary by economic and social context, with implications for understanding and treatment of psychoses globally.

Data
Data on sociodemographic characteristics and symptoms, including duration, were collated from cases, relatives, and clinical records (where available) using translated versions of the Medical Research Council Sociodemographic Schedule, 11 the WHO Personal and Psychiatric History Schedule (PPHS), 12 and the Schedules for Clinical Assessment in Neuropsychiatry (SCAN). 13 Duration of untreated psychosis was assessed using the PPHS and defined as the time between onset of psychotic symptoms (ie, symptoms meeting criteria for a rating of 2 [clinically relevant] in the psychosis sections of the SCAN) and date of identification. Date of onset was derived from this, by subtracting the duration of untreated psychosis from age at identification. Diagnoses were determined by consensus based on SCAN data.
Assessments were conducted by researchers fluent in the local language. Researchers underwent extensive training. For relevant assessments, we conducted interrater reliability (IRR) exercises. Researchers rated videos of assessments; these were compared with ratings developed by the principal investigators. For the assessments in this article, researcher ratings were within acceptable margins of principal investigator ratings: SCAN 87% (range, 85%-88%) and PPHS 76% (range, 73%-84%).

Populations at Risk
We estimated the populations at risk (aged 18-64 years) in each catchment area using the most recent data from official statistics in each country, projected from previous countrywide censuses to the year INTREPID II began or the nearest possible (India: 2011 Census; Nigeria, 2010 Census; Trinidad, 2011 Census). Population data were stratified by age (18-19 years, then 5-year bands), sex, and self-defined ethnic group (ie, Tamil, Hausa, Yoruba, African Trinidadian, Indian Trinidadian, mixed Trinidadian). We multiplied population at risk by period of case ascertainment to estimate person-years at risk (Table 1). 14-17

Statistical Analyses
We compared demographic and clinical characteristics of cases across and within sites using descriptive statistics and univariable tests of association. We used direct standardization to estimate sex-and age-standardized rates of untreated psychoses (per 100 000 person-years) using the World (WHO: 2000-2025) Standard Population (https://seer.cancer.gov/ stdpopulations/world.who.html). To test hypotheses relating to differences in rates of untreated psychosis across and within settings, we used Poisson regression to model rate ratios adjusted for sex and age. We fit interaction terms, as

Results
In each catchment area, the population at risk included more than 500 000 individuals (

Demographic and Clinical Characteristics
There were variations across settings in the demographic and clinical profiles of the cohorts (
In all settings, differences by sex varied by age at detection, with evidence of sex by age interactions in each setting ( Figure 2; eTables 3-4 and 9 in the Supplement). In Kancheepuram, rates were marginally higher among men than women in younger groups (18-29 years) and lower among men in older groups (over 30 years) (likelihood ratio test for interaction: χ 2 9 = 17.12; P for interaction = .047). In Ibadan, rates were also higher among men than women in younger groups; similar between ages 30 and 49; and lower among men in older groups (likelihood ratio test for interaction: χ 2 9 = 20.51; P for interaction = .015). In Trinidad, rates were substantially higher among men in the younger groups. From age 35 years, rates were lower for both men and women (likelihood ratio test for interaction: χ 2 9 = 44.88; P for interaction <.001). These patterns were broadly similar when we used age at onset (eFigure and eTables 5, 6, and 10 in the Supplement).

Discussion
INTREPID II is the first program in a generation to investigate the epidemiology, onset, outcomes, and impacts of psychotic disorders in multiple countries in the Global South. Using methods comparable with population-based studies in the Global North, we found considerable heterogeneity in the demographic and clinical profiles and in rates of untreated psychoses in settings in India, Nigeria, and Trinidad.

Variations in Rates: Place
The patterns of variation in rates of psychoses challenge some accepted assumptions about the epidemiology of psychoses. Age-and sex-standardized rates were approximately 3 times higher in Trinidad than in Kancheepuram and Ibadan. There are 2 sides to this. First, rates in Kancheepuram and Ibadan were relatively low. There are few previous studies for comparison. The 2 that we are aware of in India (Chennai, 18 Chandigarh 6 ) were conducted in the 1980s, and rates of nonaffective psychoses or schizophrenia in these studies were higher (approximately 40-60 per 100 000), albeit the number of cases was small (ie, <125). There are several possible reasons for these differ-   tions (EU-GEI) study, the highest rate was similarly approximately 60 per 100 000 in London, UK. 8 Further, the only other study to publish data on incidence in Trinidad, conducted in the 1990s, reported a lower rate of approximately 20 to 25 per 100 000. 19 It is possible that rates have increased over time. In this context, it is notable that both substance use and levels of community violence and crime-both of which have been posited as contributory causes of psychoses [20][21][22][23] have risen markedly in Trinidad since 1999. 24,25 These increases have been greatest in urban areas, eg, the capital Port of Spain, which forms part of our catchment area.

Variations in Rates: Diagnosis
It may be that variations in the distribution of diagnoses by setting reflect variations in the distribution of risks in each population. For example, in Trinidad, a context with high levels of trauma and substance use (which have been linked to acute positive and affective symptoms 26,27 ), there were relatively high rates of affective and brief psychoses. This said, there is a need for some caution. The low rates of affective and brief psychoses in Kancheepuram and in Ibadan contrast with some previous reports of high rates of brief psychoses in low-and middle-income countries, up to 10 times higher than in highincome countries. 28 Variations in Rates: Sex, Age, and Ethnic Group There were further variations in the sex and age distributions of rates of untreated psychoses by setting. Typically, in studies in the Global North, rates of psychoses tend to be higher among men, particularly in younger age groups. 7,8 This is what we found in Trinidad. By contrast, the sex and age distributions in Kancheepuram and Ibadan differed from this pattern. In particular, in Kancheepuram, rates were higher among women and in older age groups, a pattern that held when analyses were restricted to those with a short duration of psychoses and when repeated using age at onset. There are possible methodological explanations, eg, bias in case finding, and it is possible that subsequent prospective studies would produce findings more similar to those in high-income countries. This noted, it is also possible that sex and age distributions and profiles vary by context. This makes sense if rates overall are influenced by environmental risks and protective factors. There are, however, very few previous studies for comparison. In Trinidad, rates were approximately 2 times higher in the African Trinidadian population compared with Indian Trinidadian and mixed. The only previous study in Trinidad did not report rates by ethnic group. Our findings, consequently, need to be considered cautiously and any proposed explanations are speculative. This noted, it may be relevant that there are marked variations by ethnic group in area of residence in Trinidad, with the African population more concentrated in urban settings (eg, Port of Spain: approximately 55% African vs approximately 5% Indian), in which there are higher levels of exposure to established risks for psychosis.

Limitations
This study has several limitations. First, in each setting we sought to establish case-detection systems tailored to local health contexts. This goes beyond previous studies. However, we still cannot exclude the possibility that we missed cases and that this varied across and within settings. It is possible, for example, that folk and traditional healers and informants more often missed cases. Further, in Kancheepuram, where approximately 10% of the population is non-Tamil, we did not identify any non-Tamil cases. Differential case ascertainment may, therefore, partly explain variations in rates and absence of non-Tamil cases in Kancheepuram.
Second, in line with previous studies, in primary analyses we did not restrict inclusion based on duration of psychosis, and we based age-sex specific rates on age at detection. [6][7][8] This ensures consistency with previous research and means we can consider the full spectrum of psychotic disorders and of clinical, social, and service-use histories. Still, in the analyses presented here, the variations in duration of psychosis by site mean that there is a need for caution in comparing rates of untreated psychoses, particularly age-sex-specific rates. In sites where duration is relatively long, such as Kancheepuram, the estimated rates may be less accurate proxies for incidence. In this article, we do also provide estimated age-sexspecific rates based on age at onset and estimated rates of untreated psychoses with a duration less than 2 years. However, rates of short-duration psychosis may underestimate incidence, as it may take time for individuals with a psychotic disorder to become visible to detection systems.
Third, in each setting, we relied on projections from previous censuses to estimate populations at risk. We do not know to what extent any inaccuracies in projections varied by site and to what extent, if any, this distorted rate ratios. However, it seems implausible that this could account for large observed differences, eg, between Trinidad and both Kancheepuram and Ibadan. Further, it is in Ibadan, where projections were available only to 2016, that the denominator is most likely underestimated and therefore the rate overestimated.

Conclusion
Findings of this cohort study add to research that suggests core aspects of psychosis are shaped by historic, economic, and social context. It follows that we can only fully understand the etiology, manifestations, and outcomes of psychoses-indeed the very nature of psychoses-if we research psychoses in context. We cannot assume that what we learn in high-income countries can be generalized elsewhere. In addition, our findings show that the nature and extent of needs for care and support vary across contexts; this points to the necessity of grounding the development and delivery of services in locally contextualized knowledge. Of course, services in all contexts must provide care for a wide range of people. But this does not negate the general point that, on average, needs will vary, and services need to orientate toward this. It is therefore essential that future hypothesis-driven research on psychoses is broadened to encompass a wider range of settings to deepen our understandings of psychoses and of how to deliver humane, accessible, and effective services in diverse contexts.