BMJ

Confidential: For Review Only

Refugee migration and risk of and other non - affective psychoses: a cohort study of 1.3m people in Sweden

Journal: BMJ

Manuscript ID BMJ.2015.029606

Article Type: Research

BMJ Journal: BMJ

Date Submitted by the Author: 26-Sep-2015

Complete List of Authors: Hollander, Anna-Clara; Karolinska Institutet, Dept. of Public Health Dal, Henrik; Centre for Epidemiology and Community Medicine Lewis, Glyn; UCL Psychiatric Epidemiology, Sciences Unit Magnusson, Cecilia; Karolinska Institutet, Public Health Sciences; Centre for Epidemiology and Community Medicine Kirkbride, James; UCL Psychiatric Epidemiology, Mental Health Sciences Unit Dalman, Christina; Karolinska Institutet, Public Health Sciences; Centre for Epidemiology and Community Medicine

refugees, migrants, schizophrenia, psychotic disorder, incidence, cohort Keywords: study

https://mc.manuscriptcentral.com/bmj Page 1 of 26 BMJ

1 2 3 Refugee migration and risk of schizophrenia and other non-affective 4 5 psychoses: a cohort study of 1.3m people in Sweden 6 Anna-Clara Hollander, Henrik Dal, Glyn Lewis, Cecilia Magnusson, James B Kirkbride, ‡ Christina 7 ‡ 8 Confidential:Dalman For Review Only 9 Department of Public Health Sciences, Karolinska Institutet, 171 77 Stockholm, Sweden: Anna-Clara 10 Hollander, postdoctoral researcher; Cecilia Magnusson, Professor of Public Health Epidemiology; 11 12 Christina Dalman, Professor of Public Health Epidemiology 13 14 Centre for Epidemiology and Community Medicine, Stockholm County Council, SE-171 77 Stockholm, 15 Sweden: Henrik Dal, statistician, Cecilia Magnusson, Professor of Public Health Epidemiology; 16 Christina Dalman, Professor of Public Health Epidemiology 17 18 Division of Psychiatry, UCL, London, W1T 7NF, UK: James B Kirkbride, Sir Henry Dale Fellow; Glyn 19 Lewis, Professor of Psychiatric Epidemiology 20 21 Correspondence to: Anna-Clara Hollander, [email protected] 22 ‡Joint senior authors 23

24 25 Key words: refugees, migrants, schizophrenia, psychotic disorder, incidence, cohort study 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 1

https://mc.manuscriptcentral.com/bmj BMJ Page 2 of 26

1 2 3 Abstract 4 5 Objective 6 To determine whether refugees are at elevated risk of schizophrenia and other non-affective 7 psychotic disorders, relative to non-refugee migrants from similar regions of origin and the Swedish- 8 Confidential:born population. For Review Only 9 10 11 Design 12 Cohort study of people living in Sweden, born after 1 January 1984, and followed from their 14 th 13 birthday or arrival in Sweden, if later, until diagnosis of a non-affective psychotic disorder, 14 emigration, death or 31 December 2011. 15 16 17 Setting 18 Linked Swedish national register data. 19

20 21 Participants 22 1 348 602 persons, including people born in Sweden to two Swedish-born parents (88.3%), refugees 23 (1.8%) and non-refugee migrants (9.9%) from four major refugee-generating regions: the Middle 24 East & North Africa, Sub-Saharan Africa, Asia, Eastern Europe and Russia. 25 26 27 Main outcome measures 28 Cox regression analysis was used to estimate adjusted hazard ratios (aHR) for non-affective 29 psychotic disorders by refugee status and region of origin, controlling for age-at-risk, sex and 30 31 disposable income. 32 33 Results 34 We identified 3 704 cases during 8.9m person-years of follow-up. Compared with the crude 35 36 incidence in the Swedish-born population (38.6 per 100 000 person-years; 95% confidence interval: 37 37.3 to 39.9), there were 41.8 excess cases (33.6 to 50.0) in non-refugee migrants and 87.8 (62.1 to 38 113.5) excess cases in refugees per 100 000 person-years. Refugees were at increased risk 39 compared with both the Swedish-born population (aHR: 2.9; 2.3 to 3.7) and non-refugee migrants 40 41 (aHR: 1.6; 1.3 to 2.0) in our adjusted models. The increased rate in refugees compared with non- 42 refugees was more pronounced in men (χ2-test: 13.3; p=0.001), and present for refugees from all 43 regions, except from Sub-Saharan Africa, where both groups experienced similarly high rates relative 44 to the Swedish-born population. 45 46 47 Conclusions 48 Refugees face an increased risk of schizophrenia and other non-affective psychotic disorders 49 compared with non-refugee immigrants from similar regions of origin and the native-born Swedish 50 51 population. Clinicians and health service planners in refugee-receiving countries should be aware of 52 raised psychosis risk in addition to other mental and physical health inequalities experienced by 53 refugees. 54 55 56 57 58 59 60 2

https://mc.manuscriptcentral.com/bmj Page 3 of 26 BMJ

1 2 3 What this paper adds 4 5 Section 1: What is already known on this subject 6 Immigrant populations are at elevated risk of schizophrenia and other non-affective psychotic 7 8 Confidential:disorders. It is unclear whether refugees experiencFore rates Review of these disorders over andOnly above those 9 typically observed in non-refugee immigrant groups. 10 11 12 13 Section 2: What this study adds 14 Using a national cohort study of over 1.3m people in Sweden, we found that the risk of a non- 15 16 affective psychotic disorder was 60% higher amongst refugees than non-refugees immigrants from 17 18 similar regions of origin, and nearly three times greater than in the native-born Swedish population. 19 These patterns were apparent for men and women, although were stronger in men. Refugees from 20 21 all regions of origin experienced higher rates of psychotic disorder than non-refugee migrants, 22 except for people from Sub-Saharan Africa, where both groups experienced similarly high rates 23 24 relative to the Swedish-born population. Clinicians and health service planners should be aware of 25 26 early signs of psychosis in vulnerable migrant populations, who may benefit from timely and early 27 interventions. More broadly, our research is consistent with the possibility that traumatic life events 28 29 contribute to the aetiology of schizophrenia and other non-affective psychoses. 30

31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 3

https://mc.manuscriptcentral.com/bmj BMJ Page 4 of 26

1 2 3 Introduction 4 5 Schizophrenia and other psychotic disorders lead to lifelong health and social adversities, 6 culminating in a reduced life expectancy of 10 to 25 years. 1 Immigrants and their descendants are, 7 8 Confidential:on average, 2.5 times more likely to experience For a psychotic Review disorder than the majority Only ethnic group 9 2 3 in a given setting, although exact risk varies by ethnicity and setting. For example, in Europe, 10 11 incidence rates for people of black Caribbean or African descent are approximately five times higher

12 2 4 13 than experienced in the white European population. These marked differences persist after 14 adjustment for age, sex and socioeconomic position,5 are maintained in the descendants of first 15 16 generation migrants, 2 and do not appear to be attributable to higher incidence rates in people’s 17 6 7 8 9 18 country of origin, or selective migration. Putative explanations centre on various social 19 determinants of health, including severe or repeated exposure to psychosocial adversities such as 20 21 trauma, abuse, socioeconomic disadvantage, discrimination and social isolation. If true, people 22 granted refugee status may be particularly vulnerable to psychosis, given their increased exposure to 23 24 these risk factors. 10 11 25 26 27 While refugees experience more mental health problems than their non-refugee counterparts, 11 12 28 29 including posttraumatic stress disorder (PTSD) 13 and common mental disorders, 14 little is known 30 about psychosis risk in refugees. One previous longitudinal study from Denmark observed that 31 32 refugees were at elevated psychosis risk compared with the native-born Danish population.15 33 34 However, the risk in refugees was not compared with other non-refugee migrants, known to be at 35 increased risk, 16 making it impossible to attribute this excess directly to a refugee effect. More 36 37 recently, a Canadian cohort study found that refugees had a modestly increased risk of

38 17 39 schizophrenia compared with other migrants, but neither group were at elevated risk compared 40 with an ethnically-diverse Canadian-born background population, making this finding difficult to 41 42 interpret, and contrary to a large literature on immigration and psychosis. 2 43

44 45 Here, we clarify the risk of non-affective psychotic disorders, including schizophrenia, in refugees 46 47 compared with both non-refugee migrants and the native-born Swedish population in a national 48 population-based cohort of 1.3m people. Sweden experienced high levels of labour immigration 49 50 between 1940 and 1970, followed by substantial refugee immigration. On a per capita basis Sweden

51 18 52 grants more refugee applications than any other high income country, which combined with 53 national linked register data, makes it an excellent setting to conduct this research. We hypothesised 54 55 that refugees would have a higher risk of non-affective psychotic disorders than non-refugee 56 migrants, and that risk for both groups would be elevated compared with the Swedish-born 57 58 59 60 4

https://mc.manuscriptcentral.com/bmj Page 5 of 26 BMJ

1 2 3 population. We also hypothesised that risk in refugees compared with non-refugee migrants would 4 5 vary by region of origin, given putative differences in the pre-migratory experiences of non-refugee 6 immigrants from different regions. 7 8 Confidential: For Review Only 9 Methods 10 11 Study design and population 12 13 We established a retrospective cohort of 1 348 602 people born after 1 January 1984, who were 14 either (1) born in Sweden to two Swedish-born parents (N=1 191 071), (2) a refugee (N=24 127) or 15 16 (3) a non-refugee first generation migrant (N=133 404) granted residency in Sweden. To permit valid 17 18 comparisons between refugee and non-refugee migrants, we restricted the immigrant sample to 19 people born in geographical regions with at least 1 000 refugees in our cohort (see below). People 20 21 without an official residence permit in Sweden were excluded i.e. undocumented migrants or people 22 pending an official asylum decision. We followed participants from their 14 th birthday, or date of 23 24 arrival in Sweden if later, until diagnosis of an International Classification of Diseases, Tenth Revision 25 26 (ICD-10) non-affective psychotic disorder (F20-29), emigration, death or 31 December 2011, 27 whichever was sooner. We could not include people who immigrated to Sweden before 1 January 28 29 1998 (N=53 855), because refugee status was not sufficiently recorded in the Swedish national 30 registers prior to this date. Excluded participants did not differ from immigrants included in the 31 32 cohort by sex (51.0% vs. 50.7% men; χ2 p=0.21), but had a higher disposable income (11.0% vs 5.4% 33 34 were in the highest income quartile; χ2 p<0.001) and were more likely to come from the former 35 2 36 Yugoslavia (32.4% vs 8.5%; χ p<0.001) than other regions. Crude incidence rates were similar 37 between excluded (77.7 per 100 000 person-years; 70.4 to 85.8) and included immigrants (86.6; 79.1 38 39 to 94.7). 40 41 42 Data sources 43 44 Data were extracted from a large, longitudinal database of linked national registers, known as 45 Psychiatry Sweden, which included data on all people officially resident in Sweden after 1 January 46 47 1932, linked via a unique personal identity number and anonymized by Statistics Sweden for 48 research purposes. We obtained relevant outcome, exposure and covariate data from the following 49 50 registers: The Register of the Total Population to identify cohort participants and obtain basic 51 52 demographic data (birth date, sex, country of birth); the Multi-generation Register to link 53 participants to their parents for identification of the native-born Swedish population; the 54 55 Longitudinal integration database for health insurance and labour market studies (LISA) to obtain 56 57 data on disposable income; the immigration and emigration database (STATIV) to obtain migration 58 59 60 5

https://mc.manuscriptcentral.com/bmj BMJ Page 6 of 26

1 2 3 and refugee data; The National Patient Register to obtain outcome data, and; the Causes of Death 4 5 Register for data pertaining to mortality. 6 7 8 Confidential:Outcome For Review Only 9 Our primary outcome was an ICD-10 clinical diagnosis of schizophrenia or another non-affective 10 11 psychotic disorder (F20-29). We defined cases as cohort participants with a first recorded diagnosis 12 13 between 1 January 1998 and 31 December 2011 in the National Patient Register, which records 14 diagnoses following in- and out-patient admissions in Sweden (including privately-run public 15 16 healthcare settings). In-patient records are complete since 1987, while complete recording from out- 17 18 patient settings began in 2001. We excluded anyone with a recorded diagnosis of non-affective 19 psychotic disorder made before 14 years old (N=156). 20 21 22 Exposures 23 24 Our primary exposure was refugee status, defined as either: a refugee migrant, non-refugee 25 26 migrant, or Swedish-born to two Swedish-born parents, obtained from the STATIV database, which 27 records the reason why a residence permit was granted. Permanent residency for asylum in Sweden 28 29 is based on the Swedish Migration Agency’s definition of refugee status, 19 made in accordance with 30 Swedish law and the UN Refugee Convention, as someone who: "owing to a well-founded fear of 31 32 being persecuted […] is unable to, or owing to such fear, is unwilling to avail himself of the 33 20 (pp.14) 34 protection of that country." All other immigrants granted official residency were classified as 35 non-refugee migrants. We identified people born in Sweden to two Swedish-born parents 36 37 (henceforth the “Swedish-born” group) via linkage to the multi-generation register. 38 39 40 As a secondary exposure, we classified people according to region of origin, as defined by country of 41 42 birth. Although Statistics Sweden records data on specific country of birth, information is only 43 released for research purposes according to 13 larger geographical regions to ensure confidentiality. 44 45 From this variable, we derived a broader region of origin variable for analysis, which included 46 47 Sweden (Swedish-born only) and four other regions from which at least 1 000 refugees in our cohort 48 originated: Sub-Saharan Africa, Asia, Eastern Europe and Russia, and the Middle East & North Africa 49 50 (see Supplemental Table 1). 51 52 53 54 55 56 Covariates 57 58 59 60 6

https://mc.manuscriptcentral.com/bmj Page 7 of 26 BMJ

1 2 3 We included age-at-risk and sex as two a priori confounder variables in all analyses. We also included 4 5 individual disposable income in Sweden as a covariate, to control for possible differences in income 6 between refugees, non-refugee migrants and the Swedish-born population. Disposable income was 7 8 Confidential:defined as annual disposable income, based For on total family Review income from all registered Only sources, 9 including wages, welfare benefits, other social subsidies and pensions. Individual disposable income 10 11 was estimated by Statistics Sweden, weighting total family income according to household size and 12 13 composition, with younger children given lower weights than older household members. We 14 measured disposable income at the earliest point during follow-up (available in LISA at 16 years old, 15 16 or arrival in Sweden, if later). To account for inflation, individual disposable income was categorised 17 18 into quartiles, relative to all other cohort members assigned a disposable income score in the same 19 year. 20 21 22 Statistical analyses 23 24 We reported basic descriptive statistics and crude incidence rates for refugees, non-refugees and the 25 26 Swedish-born group. Next, we fitted Cox proportional hazard models to estimate hazard ratios (HR) 27 and 95% confidence intervals (95% CIs) according to each exposure variable. Follow-up time was 28 29 based on the earliest date of entry into the risk period (date of 14 th birthday, or for immigrants older 30 than 14 years on arrival, date of immigration) until exit from the cohort. Age-at-risk was modelled as 31 32 a time-varying covariate, using Lexis expansion to stratify each participant into N observations, 33 34 taking into account differing ages-at-risk over the follow-up period (14-16, 17-19, 20-22, 23-25, 26- 35 27; N =5). 36 max 37 38 39 We initially examined the effect of refugee status on risk of non-affective psychotic disorder, after 40 adjustment for age-at-risk, sex and their interaction, if statistically significant. In a second 41 42 adjustment we added disposable income. We tested whether the relationship between refugee 43 status and non-affective disorder differed between men and women, by fitting an interaction term 44 45 between refugee status and sex, with results presented separately for men and women, where 46 47 appropriate. We repeated these analyses for our secondary exposure variable, region of origin. Next, 48 to determine whether risk of non-affective psychotic disorder in refugees relative to non-refugees 49 50 differed by region of origin we fitted a Cox regression model to a subset of the cohort, excluding the 51 52 Swedish-born group who did not contribute information to these analyses. Given the small sample 53 of female refugees diagnosed with psychosis (N=27), these analyses were restricted to both sexes 54 55 and, separately, for men only. All statistical interactions were assessed using likelihood ratio tests 56 [LRT] against a model without the relevant interaction term. 57 58 59 60 7

https://mc.manuscriptcentral.com/bmj BMJ Page 8 of 26

1 2 3 4 5 To minimise the possibility that immigrants diagnosed with non-affective psychotic disorder may 6 have been prevalent (i.e. existing) cases on arrival to Sweden, we conducted sensitivity analyses on 7 8 Confidential:all models, excluding any refugee or non-refugee For migrant Review diagnosed within 12 months Only of 9 immigration. Finally, we checked our main models (via LRT) for departure from proportional hazards. 10 11 We used Stata v13 to analyse the data. 12 13 14 Results 15 16 We identified 3,704 cases during more than 8.9 m person-years of follow-up (Table 1). Median age- 17 18 at-first-diagnosis in the Swedish-born population was 20.1 years old (interquartile range: 18.3 to 19 22.3), younger than for refugees (21.0; 19.2 to 23.7; Mann-Whitney p<0.001) and non-refugees 20 21 (20.9; 18.7 to 23.6; Mann-Whitney p<0.001), for whom age-at-first-diagnosis was similar (Mann- 22 Whitney p=0.30). Following arrival in Sweden, time to first diagnosis was sooner for refugees 23 24 (median 2.8 years: 0.7 to 5.6) than non-refugees (3.9 years: 1.2 to 7.0; Mann-Whitney p=0.02). 25 26 27 28 29 30 The crude incidence of non-affective psychotic disorders was 38.6 per 100,000 person-years in the 31 32 Swedish-born population (95% CI: 37.3 to 39.9), 80.3 in non-refugee migrants (72.7 to 88.9) and 33 34 126.3 in refugee migrants (103.1 to 154.8). This corresponded to an absolute rate difference of 46.0 35 extra cases per 100 000 person-years (19.1 to 72.9) in refugees compared with non-refugee 36 37 migrants, in addition to an extra 41.8 cases per 100 000 person-years (33.6 to 50.0) in non-refugee 38 39 migrants compared with the Swedish-born population. Compared with the Swedish-born 40 population, rates of non-affective psychotic disorder were 2.92 (2.33 to 3.67) times higher in 41 42 refugees and 1.83 (1.59 to 2.11) higher in non-refugees, after adjustment for age, sex, their 43 interaction and disposable income (Table 2). Refugees were 1.60 (1.27 to 2.01) times more likely to 44 45 be diagnosed with non-affective psychotic disorders than non-refugee migrants. These associations 46 47 were more pronounced in men than women (LRT p-value for interaction = 0.001; Table 2 and Figure 48 1). 49 50 51 52

53 54 55 Taking refugees and non-refugee migrants together, immigrants from all regions of origin had 56 increased rates of disorder relative to the Swedish-born population, after adjustment for age-at-risk 57 58 59 60 8

https://mc.manuscriptcentral.com/bmj Page 9 of 26 BMJ

1 2 3 and sex (Supplemental Table 2). Hazard ratios were most pronounced for immigrants from Sub- 4 5 Saharan Africa (HR: 5.23; 4.32 to 6.34), observed for both men (HR: 6.68; 5.32 to 8.37) and women 6 (HR: 3.64; 2.68 to 4.94) separately. These patterns persisted after adjustment for disposable income, 7 8 Confidential:ranging from 1.45 (1.14 to 1.83) in people fromFor Eastern EuropeReview and Russia, to 4.26 Only(3.51 to 5.17) in 9 people from Sub-Saharan Africa, relative to the Swedish-born population. 10 11 12 13 14 15 16 We next investigated whether the elevated rates of non-affective psychotic disorders in refugees 17 18 compared with non-refugees differed by region of origin, excluding the Swedish-born population 19 who did not contribute to these analyses. For men and women combined, there was evidence that 20 21 the rate of non-affective psychosis in refugees compared with non-refugee migrants varied by region 22 of origin (Table 3; LRT p=0.05). This finding was even more pronounced in men (LRT p=0.008), such 23 24 that rates of non-affective psychotic disorder were elevated in refugees compared with non-refugee 25 26 migrants from all regions of origin, except Sub-Saharan Africa (HR: 0.70; 0.41 to 1.18), after adjusting 27 for age-at-risk, sex and disposable income (Table 3). Male refugees from Eastern Europe and Russia 28 29 were at greatest risk compared with their non-refugee counterparts (HR: 2.91; 1.23 to 6.87). In 30 general, the rate of psychotic disorders in refugees relative to non-refugees became smaller as the 31 32 crude incidence rate in non-refugees from each region of origin increased (Table 3). No attempt to 33 34 examine this effect in women was made, given insufficient numbers of refugee migrants (N=27). 35 36 37

38 39 40 Sensitivity analyses, excluding potentially prevalent cases amongst immigrants, did not appreciably 41 42 alter estimates of associations for our main exposures (Supplemental Tables 3 and 4). The 43 assumption of proportional hazards was not violated (p=0.83 and p=0.13 for analyses of refugee 44 45 status and region of origin, respectively). 46 47 48 Discussion 49 50 In this cohort study, we have demonstrated that refugees granted asylum in a high income setting 51 52 were, on average, 60% more likely to develop schizophrenia or another non-affective psychotic 53 disorder than non-refugee migrants from the same regions of origin, and up to 3.6 times more likely 54 55 to do so than the Swedish-born population. 56

57 58 59 60 9

https://mc.manuscriptcentral.com/bmj BMJ Page 10 of 26

1 2 3 Strengths and weaknesses 4 5 This study has several methodological strengths. It was based on a large, national population-based 6 cohort of over 1.3m people, followed for more than 8.9m person-years using linked Swedish register 7 8 Confidential:data. This research has not previously been For possible due Review to a lack of information onOnly the reason for 9 migration in official Swedish registers; one earlier attempt to investigate this issue in Sweden could 10 11 not distinguish between refugee and non-refugee migrants from the same region.21 Swedish register

12 22 23 13 data is known to be reliable for research purposes, and diagnosis of psychotic disorders recorded 14 in the National Patient Register has good validity and positive predictive value. 24-26 This register is 15 16 highly complete, recording all psychiatric contacts from in-patient settings from 1988 onwards, and 17 18 from out-patient settings since 2001. While this may have led to slight under-ascertainment from 19 out-patient settings between 1998 and 2000, we have no reason to believe this would have 20 21 introduced differential bias by refugee status or region of origin. We cannot exclude the possibility 22 that we under-estimated the true incidence of non-affective psychoses in Sweden, particularly for 23 24 certain groups, such as recent immigrants or refugees, who may have been unfamiliar with the

25 27 26 Swedish healthcare system or who had poor health literacy. Any resultant effects are likely to have 27 reduced our hazard ratios. Sensitivity analyses did not suggest that our results were attributable to 28 29 prevalent cases amongst refugees and non-refugees. Although it has been suggested that diagnostic 30 bias might explain excess rates of psychotic disorders observed in immigrant groups, 8 there is little 31 32 evidence to support this in general,8 and such bias is unlikely to account for observed differences in 33 34 risk between refugees and other migrants from the same regions of origin. Refugees are also at 35 elevated risk of PTSD,13 which can present with psychotic features; however, our findings are 36 37 unlikely to be attributable to misdiagnosed cases of PTSD amongst refugees, since these disorders

38 28 39 often present comorbidly in people exposed to trauma. 40 41 42 We were unable to include immigrants who arrived in Sweden before 1998, because data on refugee 43 status was unavailable before this year. These groups would have been at risk during the follow-up 44 45 period, and were more likely to come from the former Yugoslavia, reflecting geopolitical conflicts of 46 47 the time. This may have reduced our power to detect differences between refugees and other 48 migrants from Eastern Europe, but we have no reason to believe their exclusion would have 49 50 otherwise biased our estimates; the crude incidence in this group was comparable with that for 51 52 included immigrants, despite higher income amongst excluded migrants. Finally, despite our large 53 cohort size, the number of cases in refugees was small, which limited our power to detect effects in 54 55 certain groups, most notably women, for whom risk of non-affective psychotic disorders is, on 56 average, half that of men. 29 57 58 59 60 10

https://mc.manuscriptcentral.com/bmj Page 11 of 26 BMJ

1 2 3 4 5 While we adjusted for possible differences between refugees, non-refugees and the Swedish-born 6 population with regard to age, sex and disposable income, we did not include other markers of social 7 8 Confidential:disadvantage measured following arrival inFor Sweden; such Review factors were likely to have Only been on the 9 causal pathway between reason for immigration and psychosis risk, making adjustment 10 11 inappropriate. Disposable income did not appreciably confound psychosis risk between refugees and 12 13 non-refugees, suggesting that post-migratory social disadvantage following immigration was similar 14 for refugees and non-refugee migrants from the regions we included. 15 16 17 18 Clinical and public health implications of the study 19 Contemporary humanitarian crises in Europe, the Middle East, North Africa and central Asia have 20 21 contributed to more displaced persons, asylum seekers and refugees worldwide than at any time 22 since World War II.30 The severe social, economic and health inequalities faced by displaced 23 24 populations arising from these crises are often compounded by national immigration policies and 25 26 structural constraints in receiving countries. In turn, exposure to these ongoing adversities appears 27 likely to contribute to the increased risk of PTSD and common mental disorders experienced by 28 29 refugees.11-13 Our data highlight further mental health inequalities facing such groups. Clinicians and 30 service planners in high income settings should be aware of the early signs of psychosis in refugees, 31 32 for whom median presentation to services after arrival to Sweden was over a year sooner than for 33 34 other migrant groups. Just as for the general population, refugees and their families will benefit from 35 timely and early interventions and care, particularly in those exposed to severely traumatic events. 36 37 Our findings are consistent with the hypothesis that increased risk of non-affective psychotic

38 31 39 disorders among immigrants is due to exposure to traumatic experiences prior to migration. 40 Further studies of the pre-migratory experiences of refugees and other migrants will be required to 41 42 confirm this possibility. In general population samples, exposure to traumatic events in childhood 43 has also been associated with later schizophrenia risk, 32 and violence experienced by children 33 and 44 45 adults 11 who flee persecution has been linked to worse subsequent mental health in general. 46 47 Intriguingly, there was some evidence to suggest that psychosis risk in refugees relative to other 48 migrants varied by region of origin. While this finding requires replication in larger samples, it 49 50 suggests that in addition to refugee status, context matters. For example, amongst immigrants from 51 52 Sub-Saharan Africa we observed no differences in psychosis risk between refugees and non-refugee 53 migrants; both groups experienced high rates of disorder (over 165 new cases per 100 000 person- 54 55 years), over four times greater than in the Swedish-born population, despite after adjustment for 56 age-at-risk, sex and disposable income. One parsimonious explanation for this finding is that a larger 57 58 59 60 11

https://mc.manuscriptcentral.com/bmj BMJ Page 12 of 26

1 2 3 proportion of all Sub-Saharan Africa immigrants will have been exposed to psychosocial adversities 4 5 prior to emigration, irrespective of refugee status. By contrast, pre-migratory social adversities 6 experienced by refugees from Eastern Europe and Russia vis-à-vis non-refugee migrants from these 7 8 Confidential:countries, may have been substantially different, For thus confiningReview excess psychosis riskOnly to the refugee 9 group from such regions. 10 11 12 13 Other issues, including difficulties in the asylum process, also warrant further investigation. For 14 example, women seeking asylum are less likely to be granted refugee status than men, given greater 15 16 structural and cultural barriers in the asylum process. 34 In our study, such an effect would have led 17 18 to a higher proportion of women being classified as non-refugee migrants, which may have partially 19 explained why differences in incidence between female refugees and non-refugees were less 20 21 pronounced than for their male counterparts. Another important avenue will be to investigate 22 whether post-migratory risk and protective factors affect the risk and course of disorder in refugees 23 24 and other immigrants following resettlement. In general population samples, there is some evidence

25 35 36 26 to suggest that perceived discrimination and ethnic density (proximity to one’s own ethnic group) 27 are, respectively, risk and protective factors for psychosis. Issues including cultural distance, social 28 29 network size, employment opportunities and the host population’s perceptions about immigration 30 may also influence exposure to ongoing social adversity experienced by different immigrant 31 32 populations. 33 34 35 Conclusion 36 37 Our study shows that, on average, refugees in a high income setting face substantially elevated rates 38 39 of schizophrenia and other non-affective psychoses, in addition to the array of other mental, 40 physical and social inequalities which already disproportionately affect these vulnerable populations. 41 42 This risk exceeded the well-established excess burden of psychosis experienced in immigrant and 43 ethnic minority groups more generally, and thus emphasises the need to take the early signs and 44 45 symptoms of psychosis into account in refugee populations, as part of any clinical mental health 46 47 service responses to current global humanitarian crises. More broadly, our findings support the 48 possibility that exposure to psychosocial adversity contributes to excess psychosis observed for 49 50 some immigrant groups. 51 52 53

54 55 56 57 58 59 60 12

https://mc.manuscriptcentral.com/bmj Page 13 of 26 BMJ

1 2 3 Details of contributors 4 5 A-CH and JK had full access to the final data, performed statistical analyses, drafted the tables of the 6 data, co-wrote the manuscript, had final responsibility for content, and the decision to submit for 7 8 Confidential:publication. A-CH, JK and HD prepared the Fordata. A-CH and Review CD conceived of the study. Only A-CH, JK and 9 CD designed the study. CD and HD acquired all other cohort data. A-CH, CD and, HD acquisitioned 10 11 the migration data. A-CH, JK and CD obtained funding for the study. A-CH, JK, CD, CM, GL interpreted 12 13 statistical analyses. HD coordinated the data management and A-CH wrote the study protocol. All 14 authors critically revised the paper for important intellectual content and approved the final version. 15 16 17 18 Declaration of interests 19 All authors have completed the Unified Competing Interest form at 20 21 www.icmje.org/coi_disclosure.pdf (available on request from the corresponding author) and declare 22 that (1) A-CH, JK, HD, CD, CM and GL have no support from any company for the submitted work; (2) 23 24 A-CH, JK, HD, CD, CM and GL have no relationships with any companies that might have an interest 25 26 in the submitted work in the previous 3 years; (3) their spouses, partners, or children have no 27 financial relationships that may be relevant to the submitted work; and (4) A-CH, JK, HD, CD, CM and 28 29 GL have no non-financial interests that may be relevant to the submitted work. 30

31 32 Funding and role of funder 33 34 The work for this paper was supported by FORTE (dnr 2014-1430) and FORTE (dnr 2014-2678). Dr 35 Kirkbride is supported by a Sir Henry Dale Fellowship jointly funded by the Wellcome Trust and the 36 37 Royal Society (Grant number: 101272/Z/13/Z). 38 39 40 Ethics approval 41 42 This research has ethical approval as part of Psychiatry Sweden “Psykisk ohälsa, psykiatrisk sjukdom: 43 förekomst och etiologi”, approved by the Stockholm Regional Ethical Review Board Dnr 2010/1185- 44 45 31/5. The funders had no involvement in any aspect of the design of this study, preparation of 46 47 results, or decision to submit for publication. 48 49 50 Access to data 51 52 All authors had full access to all of the data (including statistical reports and tables) in the study and 53 can take responsibility for the integrity of the data and the accuracy of the data analysis. 54 55 56 Data sharing 57 58 59 60 13

https://mc.manuscriptcentral.com/bmj BMJ Page 14 of 26

1 2 3 Statistical code is available from the corresponding author. Under Swedish law and ethical approval, 4 5 patient-level data cannot be made available. 6 7 8 Confidential:Transparency declaration For Review Only 9 Anna-Clara Hollander (the manuscript's guarantor) affirms that the manuscript is an honest, 10 11 accurate, and transparent account of the study being reported; that no important aspects of the 12 13 study have been omitted; and that any discrepancies from the study as planned (and, if relevant, 14 registered) have been explained. 15 16 17 18 Copyright for authors 19 The Corresponding Author, Anna-Clara Hollander, has the right to grant on behalf of all authors and 20 21 does grant on behalf of all authors, a worldwide licence to the Publishers and its licensees in 22 perpetuity, in all forms, formats and media (whether known now or created in the future), to i) 23 24 publish, reproduce, distribute, display and store the Contribution, ii) translate the Contribution into 25 26 other languages, create adaptations, reprints, include within collections and create summaries, 27 extracts and/or, abstracts of the Contribution, iii) create any other derivative work(s) based on the 28 29 Contribution, iv) to exploit all subsidiary rights in the Contribution, v) the inclusion of electronic links 30 from the Contribution to third party material where-ever it may be located; and, vi) licence any third 31 32 party to do any or all of the above 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 14

https://mc.manuscriptcentral.com/bmj Page 15 of 26 BMJ

1 2 3 References 4 5 6 1. Laursen TM, Munk-Olsen T, Vestergaard M. Life expectancy and cardiovascular mortality in 7 persons with schizophrenia. Curr Opin Psychiatry 2012; 25 (2):83-8. 8 Confidential:2. Bourque F, van der Ven E, Malla A. A meta-analys Foris of Reviewthe risk for psychotic disorders Only among first- 9 and second-generation immigrants. Psychological Medicine 2011; 41 (5):897-910. 10 3. Cantor-Graae E, Selten JP. Schizophrenia and migration: a meta-analysis and review. Am J 11 Psychiatry 2005; 162 (1):12-24. 12 4. Kirkbride JB, Errazuriz A, Croudace TJ, et al. Incidence of schizophrenia and other psychoses in 13 England, 1950-2009: a systematic review and meta-analyses. PLoS One 2012; 7(3):e31660. 14 5. Kirkbride JB, Barker D, Cowden F, et al. Psychoses, ethnicity and socio-economic status. Br J 15 Psychiatry 2008; 193 (1):18-24. 16 17 6. Mahy GE, Mallett R, Leff J, et al. First-contact incidence rate of schizophrenia on Barbados. Br J 18 Psychiatry 1999; 175 :28-33. 19 7. Jablensky A, Sartorius N, Ernberg G, et al.Schizophrenia: manifestations, incidence and course in 20 different cultures. A World Health Organization ten-country study. Psychol Med Monogr 21 Suppl. 1992; 20 :1-97. 22 8. Selten JP, Hoek H. Does misdiagnosis explain the schizophrenia epidemic among immigrants from 23 developing countries to Western Europe? Social Psychiatry and Psychiatric Epidemiology 24 2008; 43 (12):937-39. 25 9. van der Ven E, Dalman C, Wicks S, et al. Testing Odegaard's selective migration hypothesis: a 26 longitudinal cohort study of risk factors for non-affective psychotic disorders among 27 prospective emigrants. Psychological Medicine 2015;45 (4):727-34. 28 10. Steel Z, Chey T, Silove D, et al. Association of torture and other potentially traumatic events with 29 mental health outcomes among populations exposed to mass conflict and displacement: a 30 31 systematic review and meta-analysis. JAMA 2009; 302 (5):537-49. 32 11. Porter M, Haslam N. Predisplacement and postdisplacement factors associated with mental 33 health of refugees and internally displaced persons: a meta-analysis. JAMA 2005; 294 (5):602- 34 12. 35 12. Hollander AC, Bruce D, Burstrom B, et al. The association between immigrant subgroup and poor 36 mental health: a population-based register study. J Nerv Ment Dis 2013; 201 (8):645-52. 37 13. Fazel M, Wheeler J, Danesh J. Prevalence of serious in 7000 refugees resettled in 38 western countries: a systematic review. Lancet 2005; 365 (9467):1309-14. 39 14. Hollander AC. Social inequalities in mental health and mortality among refugees and other 40 immigrants to Sweden - epidemiological studies of register data. Glob Health Action 41 2013; 6:21059. 42 15. Norredam M, Garcia-Lopez A, Keiding N, et al. Risk of mental disorders in refugees and native 43 44 Danes: a register-based retrospective cohort study. Soc Psychiatry Psychiatr Epidemiol 45 2009; 44 (12):1023-9. 46 16. Cantor-Graae E, Pedersen CB, McNeil TF, et al. Migration as a risk factor for schizophrenia: a 47 Danish population-based cohort study. Br J Psychiatry 2003; 182 :117-22. 48 17. Anderson K, Cheng J, Susser E, et al. The incidence of psychotic disorders among first generation 49 immigrants and refugees to Ontario. Canadian Medical Association Journal 2015. 50 18. The World Bank. http://databank.worldbank.org/data/home.aspx, 2015. Accessed 22nd 51 September 2015 52 19. Swedish Migration Agency. http://www.migrationsverket.se/english. Secondary 53 http://www.migrationsverket.se/english 2015. Accessed 22nd September 2015 54 20. United Nations High Commissioner for Refugees UNHCR. Convention and protocol relating to the 55 status of refugees, 1951, 1967. 56 57 21. Leao TS, Sundquist J, Frank G, et al. Incidence of schizophrenia or other psychoses in first- and 58 second-generation immigrants: a national cohort study. J Nerv Ment Dis 2006; 194 (1):27-33. 59 60 15

https://mc.manuscriptcentral.com/bmj BMJ Page 16 of 26

1 2 3 22. Abel KM, Heuvelman HP, Jorgensen L, et al. Severe bereavement stress during the prenatal and 4 childhood periods and risk of psychosis in later life: population based cohort study. BMJ 5 2014; 348 :f7679. 6 23. Rai D, Lee BK, Dalman C, et al. Parental depression, maternal antidepressant use during 7 pregnancy, and risk of autism spectrum disorders: population based case-control study. BMJ 8 Confidential:2013; 346 :f2059. For Review Only 9 24. Dalman C, Broms J, Cullberg J, et al. Young cases of schizophrenia identified in a national 10 inpatient register--are the diagnoses valid? Soc Psychiatry Psychiatr Epidemiol 11 12 2002; 37 (11):527-31. 13 25. Jorgensen L, Allebeck P, Dalman C. Prevalence of psychoses in Stockholm County-A population- 14 based study using comprehensive healthcare registers. Nord J Psychiatry 2013. 15 26. Ludvigsson JF, Andersson E, Ekbom A, et al. External review and validation of the Swedish 16 national inpatient register. BMC Public Health 2011; 11 :450. 17 27. Wangdahl J, Lytsy P, Martensson L, et al. Health literacy among refugees in Sweden - a cross- 18 sectional study. Bmc Public Health 2014; 14 . 19 28. OConghaile A, DeLisi LE. Distinguishing schizophrenia from posttraumatic stress disorder with 20 psychosis. Curr Opin Psychiatry 2015; 28 (3):249-55. 21 29. Kirkbride JB, Fearon P, Morgan C, et al. Heterogeneity in incidence rates of schizophrenia and 22 other psychotic syndromes: findings from the 3-center AeSOP study. Arch Gen Psychiatry 23 2006; 63 (3):250-8. 24 25 30. UNHCR. Global Appeal 2015 Update. http://www.unhcr.org/ga15/index.xml. Accessed 22nd 26 September 2015 27 31. Kirkbride JB, Jones PB. Epidemiological aspects of migration and mental illness. In: Bhugra D, 28 Gupta S, eds. Migration and mental health. Cambridge: Cambridge University Press, 2011:xv, 29 350 p. 30 32. Hjern A, Wicks S, Dalman C. Social adversity contributes to high morbidity in psychoses in 31 immigrants--a national cohort study in two generations of Swedish residents. Psychol Med 32 2004; 34 (6):1025-33. 33 33. Fazel M, Reed RV, Panter-Brick C, et al. Mental health of displaced and refugee children resettled 34 in high-income countries: risk and protective factors. Lancet 2012; 379 (9812):266-82. 35 34. United Nations High Commissioner for Refugees UNHCR. Beyond Proof, Credibility Assessment in 36 EU Asylum Systems. http://www.unhcr.org/51a8a08a9.html, 2013. Accessed 22nd 37 38 September 2015 39 35. Veling W, Selten JP, Susser E, et al. Discrimination and the incidence of psychotic disorders 40 among ethnic minorities in The Netherlands. International Journal of Epidemiology 41 2007; 36 (4):761-68. 42 36. Shaw RJ, Atkin K, Becares L, et al. Impact of ethnic density on adult mental disorders: narrative 43 review. Br J Psychiatry 2012; 201 (1):11-9. 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 16

https://mc.manuscriptcentral.com/bmj Page 17 of 26 BMJ

1 2 3 4 5 Table 1: CohortConfidential: characteristics by migrant status; refugees, non-refugee For migrants andReview the Swedish-born population Only 6

7 Refugee status 8 9 Swedish -born population Non -refugee migrants Refugee migrants 1 1 1 10 Category Case (%) Person -years (%) Case (%) Person -years (%) Case (%) Person -years (%) 11 Total 3 232 (100.0) 8 385 059 (100.0) 379 (100.0) 471 691 (100.0) 93 (100.0) 73 613 (100.0) 12 13 Sex Men 1 778 (55.0) 4 311 099 (51.4) 234 (61.7) 232 302 (49.3) 66 (71.0) 41 078 (55.8) 14 Women 1 454 (45.0) 4 073 960 (48.6) 145 (38.3) 239 390 (50.7) 27 (29.0) 32 535 (44.2) 15

16 Birth year 1984 -86 1 279 (39.5) 2 928 490 (34.9) 175 (46.2) 185 267 (39.3) 35 (37.6) 23 829 (32.4) 17 1987 -89 1 111 (34.4) 2 510 875 (29.9) 107 (28.2) 125 887 (26.7) 28 (30.1) 19 093 (25.9) 18 1990 -92 649 (20.1) 1 896 928 (22.6) 74 (19.5) 92 004 (19.5) 22 (23.7) 16 837 (22.9) 19 1993 -95 174 (5.4) 903 854 (10.8) 19 (5.0) 56 248 (11.9) 8 (8.6) 11 728 (15.9) 20 21 1996 -97 19 (0.6) 144 912 (1.7) 4 (1.1) 12 287 (2.6) 0 (0.0) 2 127 (2.9) 22 Region of origin Sweden 3 232 (100.0) 8 385 059 (100.0) - - - - 23 24 Sub -Saharan Africa - - 111 (29.3) 59 485 (12.6) 31 (33.3) 18 671 (25.3) 25 Asia - - 66 (17.4) 105 731 (22.4) 15 (16.1) 12 929 (17.6) 26 Eastern Europe - - 80 (21.1) 134 241 (28.5) 7 (7.5) 6 547 (8.9) 27 Middle East - - 122 (32.2) 172 234 (36.5) 40 (43.0) 35 467 (48.2) 28 29 Income 1. Lowest quartile 1 156 (35.8) 2 161 497 (25.8) 264 (69.7) 339 445 (72.0) 63 (67.7) 51 962 (70.6) 30 2. Second quartile 830 (25.7) 2 185 386 (26.1) 52 (13.7) 63 153 (13.4) 12 (12.9) 10 486 (14.2) 31 3. Third quartile 679 (21.0) 2 073 842 (24.7) 45 (11.9) 35 919 (7.6) 13 (14.0) 6 768 (9.2) 32 33 4. Highest quartile 567 (17.5) 1 964 334 (23.4 ) 18 (4.8 ) 33 174 (7.0 ) 5 (5.4 ) 4 398 (6.0 ) 34 35 1Rounded to nearest integer 36 37 38 39 40 41 42 17 43 44 45 46 https://mc.manuscriptcentral.com/bmj 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Page 18 of 26

1 2 3 4 5 Table 2: Confidential:Risk of non-affective psychoses by migrant status after adjustmentFor for confounders Review Only 6 7 8 All Men Women 9 10 Category Model 1 Model 2 Model 1 Model 2 Model 1 Model 2

11 HR (95% CI) HR (95% CI) HR (95% CI) HR (95% CI) HR (95% CI) HR (95% CI) 12 Swedish -born 1 1 1 1 1 1 13 Non -refugee migrant 2.28 (1.99 -2.62) 1.83 (1.59 -2.11) 2.61 (2.22 -3.07 ) 2.10 (1. 78 -2.48) 1.91 (1.5 8-2.3 1) 1.51 (1.24 -1.84 ) 14 Refugee migrant 3.61 (2.87 -4.53) 2.92 (2.33 -3.67) 4.28 (3.28 -5.58 ) 3.51 (2.69 -4.58) 2.66 (1.80 -3.92 ) 2.09 (1.42 -3.10 ) 15 16 17 Non -refugee migrant 1 1 1 1 1 1 18 Refugee migrant 1.58 (1.26 -1.99) 1.60 (1.27 -2.01) 1.64 (1.25 -2.16 ) 1.67 (1.27 -2.20 ) 1.39 (0.92 -2.10 ) 1.38 (0.92 -2.09 ) 19 20 21 Legend: HR: Hazard Ratio; 95% CI: 95% confidence interval. Model 1 is adjusted for age-at-risk, sex and their interaction. Model 2 is also adjusted for disposable income. A 2 2 22 likelihood ratio test on four degrees of freedom confirmed statistical interaction between sex and age-at-risk in Model 1 ( χ : 71.5; p<0.001) and Model 2 ( χ : 72.9; 23 p<0.001). A LRT, on two degrees of freedom, also confirmed statistical interaction between sex and refugee status in Model 1 ( χ2: 11.7; p=0.003) and Model 2 ( χ2: 13.3; 24 p=0.001). Hazard ratios by refugee status are therefore presented separately for men and women. 25 26 27

28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 18 43 44 45 46 https://mc.manuscriptcentral.com/bmj 47 48 49 50 51 52 53 54 55 56 57 58 59 60 Page 19 of 26 BMJ

1 2 3 Table 3: Risk of non-affective psychoses in refugees relative to non-refugees, by region of origin 4 5 6 All Men 7 Crude incidence Model 2 Crude incidence Model 2 8 Confidential:rate per 100,000 For HR (95% Review CI) rate per 100,000 Only HR (95% CI) 9 PYAR (95% CI) PYAR (95% CI) 10 Swedish -born 38.5 (37.2 -39.9) - 41.2 (39.4 -43.2) - 11 12 13 Eastern Europe 14 Non -refugees 59.6 (47.9 -74.2) 1 62.5 (45.8 -85.2) 1 15 Refugees 106.9 (51.0 -224.3) 1.77 (0.82 -3.85 ) 184.1 (82.7 -409.7) 2.91 (1.2 3-6.87 ) 16

17 18 Asia 19 Non -refugees 62.4 (49.0 -79.5) 1 67.0 (48.3 -92.8) 1 20 Refugees 116.0 (69.9 -192.4) 1. 81 (1.03 -3. 17 ) 146.1 (83.0 -257.3) 2.25 (1.17 -4.34 ) 21 22 23 Middle East & North Africa 24 Non -refugees 70.8 (59.3 -84.6) 1 94.4 (75.9 -117.3) 1 25 Refugees 112.8 (82.7 -153.8) 1. 57 (1. 10 -2.25 ) 143.4 (100.3 -205.2) 1. 56 (1. 03 -2.38 ) 26 27 Sub -Saharan Africa 28 29 Non -refugees 186.6 (154.9 -224.8) 1 268.7 (214.9 -336.0) 1 30 Refugees 166.0 (116.8 -236.1) 0. 83 (0. 55 -1. 24 ) 207.1 (130.5 -328.8) 0. 70 (0. 41 -1. 18 ) 31 32 33 Legend: HR: Hazard Ratio; 95% CI: 95% confidence interval. PYAR: Person-years at-risk. Estimates from Model 34 1 and Model 2 were similar, and only data from Model 2, adjusted for age-at-risk, sex, their interaction (for 2 35 both sexes combined) and disposable income is reported. LRT χ p-values, on 3 degrees of freedom, for 36 statistical interaction between refugee status and region of origin were χ2: 7.9; p=0.05 for the full sample, and 37 χ2: 11.8; p=0.008 in an analysis restricted to men. Given the small number of refugee women with the 38 outcome (N=27) no attempt was made to inspect risk by region of origin separately for women. 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 19

https://mc.manuscriptcentral.com/bmj BMJ Page 20 of 26

1 2 3 Figure 1: Hazard ratios for schizophrenia and other non-affective psychotic disorders by refugee 4 status and sex 5 6 7 6 8 Confidential: For Review Only 9 10 11 5 12 13 14 4 15 16 17 3 18 19 20 2 21

22 CI) (95% ratio Hazard 23 1 24 25 26 27 0 28 All Men Women All Men Women 29 Model 1 Model 2 30 31 Swedish-born Non-refugee migrant 32 Refugee migrant Refugee vs non-refugee migrant 33 34 35 Legend: Model 1 is adjusted for age-at-risk, sex and their interaction (where appropriate). Model 2 is 36 additionally adjusted for disposable income. The Swedish-born group provide the reference category, 37 except for the hashed columns, which show hazard ratios for refugees relative to non-refugee 38 migrants. Error bars represent 95% confidence intervals. 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 20

https://mc.manuscriptcentral.com/bmj Page 21 of 26 BMJ

1 2 3 Supplementary Table 1: Region of origin classification and basic sample characteristics 1

4 2 3 4 5 Region of origin Area by Statistics Sweden Countries included 6 Cases (%) [Person-years at-risk, (%)] Cases (%) [Person-years at-risk, (%)] 7 Sweden Sweden Sweden 8 Confidential:3 233 (87.3) [8 385 059, (93.9)] 3 233For (87.3) [8 385Review 059, (93.9)] Only 9 10 Sub -Saharan Africa West Africa Benin 11 139 (3.8) [78 156, (0.9)] 20 (0.5) [10 261, (0.1)] Burkina Faso 12 Cape Verde 13 Gambia 14 15 Ghana 16 Guinea 17 Guinea Bissau 18 Ivory Coast 19 Liberia 20 Mali 21 Mauritania 22 Niger 23 24 Nigeria 25 Senegal 26 Sierra Leone 27 Togo 28 Somalia -Eritrea -Ethiopia -Djibouti Djibouti 29 93 (2.5) [50 562, (0.6)] Eritrea 30 Ethiopia 31 Somalia 32 33 Other Africa Angola 34 29 (0.8) [17 333, (0.2)] Botswana 35 Burundi 36 Cameroon 37 Central African Republic 38 Chad 39 Comor os 40 Congo 41 42 Democratic republic of 43 Congo 44 Equatorial Guinea 45 Gabon 46 Kenya 47 Lesotho 48 Madagascar 49 Malawi 50 Mauritius 51 52 Mozambique 53 Namibia 54 Rwanda 55 Sao Tome and Principe 56 Seychelles 57 South Africa 58 59 60 https://mc.manuscriptcentral.com/bmj BMJ Page 22 of 26

1 2 3 Swaziland 4 Tanzania 5 Uganda 6 Zambia 7 Zanzibar 8 Confidential: For Review Only 9 10 Asia Central Asia Afghanistan 11 81 (2.2) [116 680, (1.3)] 42 (1.1) [57 020, (0.6)] Armenia 12 Azerbaijan 13 Bangladesh 14 Bhutan 15 Georgia 16 India 17 18 Sikkim (India, state in) 19 Kazakhstan 20 Kyrgyzstan 21 Maldives 22 Nepal 23 Pakistan 24 Sri Lanka 25 Tajikistan 26 27 Turkmenistan 28 Uzbekistan 29 Northeast Asia China 30 14 (0.4) [22 714, (0.3)] Korea, People’s Republic of 31 Korea, South 32 Japan 33 Mongolia 34 Taiwan 35 Southeast Asia Brunei 36 37 25 (0.7) [36 946, (0.4)] Cambodia 38 East Timor 39 Hong Kong 40 Indonesia 41 Laos 42 Malaysia 43 Myanmar 44 Philippines 45 46 Singapore 47 Thailand 48 Vietnam 49 50 Eastern Europe & Russia Eastern Europe Albania 51 87 (2.3) [140 788, (1.6)] 36 (1.0) [52 957, (0.6)] Belarus 52 Bulgaria 53 Czech republic 54 55 Hungary 56 Moldova 57 Poland 58 Romania 59 60 https://mc.manuscriptcentral.com/bmj Page 23 of 26 BMJ

1 2 3 Slovakia 4 Ukraine 5 Former Yugoslavia Bosnia Herzegovina 6 36 (1.0) [58 534, (0.7)] Croatia 7 Kosovo 8 Confidential: For Review Only 9 Macedonia 10 Montenegro 11 Serbia 12 Slovenia 13 Russia and the Baltic states Estonia 14 15 (0.4) [29 297, (0.3)] Latvia 15 Lithuania 16 Russia 17 18 19 Middle East & North Africa Iraq 20 162 (4.4) [207 700, (2.3)] 105 (2.8) [138 638, (1.6)] 21 Iran Iran 22 21 (0.6) [20 057, (0.2)] 23 Other Middle Eastern countries Bahrain 24 30 (0.8) [42 253 , (0.5) ] Cyprus 25 Israel 26 Jordan 27 28 Kuwait 29 Lebanon 30 Oman 31 Palestine 32 Qatar 33 Saudi Arabia 34 Syria 35 United Arab Emirates 36 37 Yemen 38 Turkey 39 North Africa Algeria 40 6 (0.2) [6 752 , (0.1) ] Egypt 41 Libya 42 Morocco 43 Tunisia 44 1Sample characteristics provided where available. Statistics Sweden do not provide data on a per country basis 45 2 46 Author-defined 47 3Defined by Statistics Sweden. Iran & Iraq are presented separately from other Middle Eastern countries given 48 particularly large migration flows from these countries to Sweden. 49 4 50 Register data not made available by country by Statistics Sweden 51 52 53 54 55 56 57 58 59 60 https://mc.manuscriptcentral.com/bmj BMJ Page 24 of 26

1 2 3 4 5 SupplementaryConfidential: Table 2: Risk of non-affective psychoses by region ofFor origin for all immigrantReview groups Only 6 7 All Men Women 8 Category Model 1 Model 2 Model 1 Model 2 Model 1 Model 2 9 HR (95% CI) HR (95% CI) HR (95% CI) HR (95% CI) HR (95% CI) HR (95% CI) 10 Sweden 1 1 1 1 1 1 11 12 Sub -Saharan Africa 5.2 3 (4.3 2-6.3 4) 4.26 (3.51 -5.17 ) 6. 68 (5. 32 -8. 37) 5.45 (4.34 -6.84 ) 3. 64 (2. 68 -4. 94 ) 2.94 (2.16 -3.99 ) 13 Asia 1.9 6 (1.54 -2.51) 1. 55 (1.21 -2.00 ) 2.0 4 (1.5 0-2. 77 ) 1.61 (1.18 -2.20 ) 1.8 7 (1. 31 -2. 69 ) 1. 49 (1.04 -2.14 ) 14 Eastern Europe 1.74 (1.38 -2. 20 ) 1. 45 (1.14 -1. 83 ) 1.7 4 (1. 28 -2. 37 ) 1. 46 (1.07 -1. 98 ) 1. 74 (1. 26 -2. 41 ) 1. 43 (1.03 -1. 98 ) 15 Middle East & North Africa 2.16 (1.81 -2.58) 1. 69 (1.41 -2.02 ) 2.6 4 (2.1 5-3. 25 ) 2.10 (1. 70 -2.58 ) 1. 55 (1.1 6-2.0 7) 1.19 (0. 88 -1. 59 )

16 17 Legend: HR: Hazard Ratio; 95% CI: 95% confidence interval. Model 1 was adjusted for age-at-risk, sex and their interaction (where appropriate). Model 2 was additionally 18 adjusted for disposable income. Likelihood ratio tests, on four degrees of freedom, confirmed statistical interaction between age-at-risk and sex on the risk of non-affective 19 psychotic disorder in Model 1 ( χ2: 72.2; p<0.001) and Model 2 ( χ2: 73.4; p<0.001). LRTs, on four degrees of freedom, also confirmed statistical interaction between sex and 20 2 2 region of origin on psychosis risk in Model 1 ( χ : 20.1; p<0.001) and Model 2 ( χ : 22.5; p<0.001). 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 https://mc.manuscriptcentral.com/bmj 47 48 49 50 51 52 53 54 55 56 57 58 59 60 Page 25 of 26 BMJ

1 2 3 4 5 SupplementaryConfidential: Table 3: Sensitivity analysis of psychosis risk by exposure For status, excluding Review refugee and other migrants Only who were diagnosed with a non- 6 affective psychotic disorder within 12 months of arrival in Sweden 7 8 All Men Women 9 Category N (excluded N, Model 1 Model 2 Model 1 Model 2 Model 1 Model 2 10 11 %) 12 HR (95% CI) HR (95% CI) HR (95% CI) HR (95% CI) HR (95% CI) HR (95% CI) 13 Refugee status Swedish -born 3 232 (0, 0 .0) 1 1 1 1 1 1 14 Non -refugee 288 (91, 24 .0) 2.0 8 (1.79 -2.41) 1. 65 (1.41 -1. 92 ) 2.32 (1.94-2.77 ) 1.85 (1 .54 -2.21 ) 1.81 (1 .47-2.24 ) 1.42 (1.15 -1.76) 15 Refugee 63 (30, 32 .2) 3.1 0 (2. 38 -4.0 5) 2.47 (1. 89 -3.22 ) 3.74 (2 .73-5.12) 3.01 (2.20-4.12 ) 2.21 (1 .38-3.54 ) 1.07 (1.07-2.74) 16 17 Non -refugee 288 (91, 24 .0) 1 1 1 1 1 1 18 Refugee 63 (30, 32 .2) 1. 49 (1.1 4-1.9 6) 1. 50 (1.1 4-1. 97 ) 1.61 (1.16 -2.24) 1. 63 (1.1 8-2. 27 ) 1.2 0 (0.7 3-1.98 ) 1.14 (0.69 -1.88) 19

20 21 Region of origin Sweden 3 232 (0, 0 .0) 1 1 1 1 1 1 22 Sub -Saharan 104 (38, 26 .8) 4.7 9 (3.86 -5.9 5) 3.85 (3.10 -4.78 ) 6.09 (4 .72 -7.87 ) 4.91 (3.80 -6.35 ) 3.32 (2 .34 -4.72 ) 2.65 (1.86 -3.77 ) 23 Africa 24 Asia 59 (22, 27 .2) 1.8 3 (1.39 -2.4 2) 1. 45 (1.09 -1. 92 ) 1.71 (1 .18 -2.48 ) 1.35 (0 .92 -1.96 ) 2.00 (1 .36 -2.94 ) 1.58 (1.07 -2.33 ) 25 Eastern Europe 67 (20, 23 .0) 1.6 3 (1.2 6-2.1 1) 1. 34 (1.03 -1. 74 ) 1.73 (1 .24-2.42) 1.44 (1.03 -2.01 ) 1.51 (1.03 -2.20 ) 1.23 (0 .84 -1.80 ) 26 Middle East & 85 (41, 32 .5) 1.89 (1.55 -2.30) 1. 45 (1.19 -1. 77 ) 2.21 (1 .74-2.80) 1.72 (1.35 -2.19 ) 1.47 (1 .07-2.02 ) 1.10 (0 .80 -1.52 ) 27 North Africa 28 29 30 Legend: HR: Hazard Ratio; 95% CI: 95% confidence interval. Model 1 was adjusted for age-at-risk, sex and their interaction (where appropriate). Model 2 was additionally 31 adjusted for disposable income. LRT χ2 p-values, on two degrees of freedom, for interaction between sex and refugee status, were χ2: 7.2; p=0.03 (Model 1) and χ2: 8.3; 32 p=0.02 (Model 2) and, on four degrees of freedom, between sex and region of origin, were χ2: 13.4; p=0.01 (Model 1) and χ2: 14.6; p=0.01 (Model 2). 33 34 35 36 37 38 39 40 41 42 43 44 45 46 https://mc.manuscriptcentral.com/bmj 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Page 26 of 26

1 2 3 Supplemental Table 4: Sensitivity analysis of risk of non-affective psychoses in refugees relative to 4 non-refugees for men by region of origin, excluding immigrants who were diagnosed within 12 5 months of arrival in Sweden 6 7 8 Confidential: All ( Model 2)For MenReview ( Model 2 ) Only 9 Refugee vs non -refugee HR ( 95% CI) HR ( 95% CI) 10 Sub -Saharan Africa 0.92 (0 .54 -1.56 ) 0.90 (0 .45-1.78) 11 Asia 1.86 (0.96 -3. 58) 2.49 (1.11 -5.57 ) 12 13 Eastern Europe 1.69 (0 .73 -3.92 ) 2.46 (0 .96-6.31 ) 14 Middle East & North Africa 1.58 (1 .04-2.39 ) 1.72 (1 .05-2.8 2) 15 16 17 Legend: HR: Hazard Ratio; 95% CI: 95% confidence interval. All models were conducted on a restricted cohort, 18 excluding the Swedish-born population and refugee or migrant cases presenting within 12 months of arrival in 19 Sweden. Baseline groups are non-refugees from each country of origin. For all people, Model 2 was adjusted 20 for age-at-risk, sex, their interaction and disposable income. For men, it was adjusted for age-at-risk and 21 2 22 disposable income. LRT χ p-values, on three degrees of freedom, for interaction between refugee status and 2 2 23 region of origin, were χ : 3.7; p=0.29 (all people) and χ : 5.0; p=0.17 (men). Given the small sample size for 24 women, no attempt to analyse whether hazard ratios by refugee status differed by region of origin. 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 https://mc.manuscriptcentral.com/bmj