Comparison of Hospitalization for Nonaffective Psychotic Disorders Among Refugee, Migrant, and Native-Born Adults in Sweden and Denmark

This cohort study investigates if second-generation, nonrefugee, and refugee migrants in Sweden and Denmark experience more days hospitalized for nonaffective psychotic disorders during the first 5 years of illness than their native-born peers.


Introduction
Meta-analyses indicate that first-and second-generation migrants are at increased risk for developing schizophrenia and other nonaffective psychotic disorders (NAPDs) [1][2][3][4] and that refugees are at even greater risk of these disorders than nonrefugee migrants. 5It is likely that both shared factors (eg, discrimination, language difficulties affecting health care access, socioeconomic disadvantage, and social isolation) 6 and distinct factors (eg, exposure to war, conflict, famine, persecution, family fragmentation, and institutional accommodation on arrival, common among refugees) 7 contribute to the excess risk of NAPD in these groups.Although these factors may plausibly influence illness course and lead to poorer outcomes among migrants, previous studies have yielded mixed findings for most outcomes (including functioning, remission, and engagement) with few examining refugees specifically. 8The paucity of studies examining refugees is particularly concerning given that their number worldwide has risen exponentially, increasing from 25.7 million to 32 million during the first half of 2022 alone. 9Studies examining long-term outcomes among refugees with NAPDs are essential to ensure adequate health care provision to these disadvantaged populations.
One important outcome is psychiatric hospitalization, which can be a distressing and stigmatizing experience 10 and inevitably causes disruption to education, employment, and family life.Psychiatric hospitalization is driven by clinical need (ie, when individuals are deemed to be a danger to themselves or others), which may in turn be influenced by sociodemographic factors and delays or difficulties in accessing adequate health care in the early course of illness.Indeed, factors such as age, relationship status, ethnicity, and longer duration of untreated psychosis (DUP) have been associated with risk of hospitalization and longer length of stay among patients with psychosis. 11Given that refugees and nonrefugee migrants appear to underutilize specialist health care for several psychiatric disorders that are prevalent during the early stages of psychosis (eg, affective, anxiety, and substance use disorders), 12,13 we might expect an excess risk of hospitalization for first-episode NAPD among migrants.Consistent with this hypothesis, 2 studies 14,15 observed that refugee and first-generation, nonrefugee migrant groups were both more likely to be hospitalized during the first psychotic episode when compared with their native-born peers.Whether these disparities extend to subsequent admissions (when individuals are within the health care system) or length of stay (which might be prolonged due to clinical and/or sociodemographic factors) is less clear.Moreover, to our knowledge, no study has investigated whether second-generation migrants are equally disadvantaged.
To address these knowledge gaps, we compared patterns of hospitalization for NAPDs during the early course of illness among refugee, nonrefugee, and second-generation migrants and their native-born peers in Sweden and Denmark.Although there are some similarities in the general health care systems in these Scandinavian countries (both have decentralized health care systems with publicly financed psychiatric services 16 and a similar number of psychiatric hospital beds per 100 000 inhabitants), 17 these countries differ substantially in terms of the availability of early

JAMA Network Open | Psychiatry
Hospitalization for Nonaffective Psychotic Disorders Among Refugee, Migrant, and Native-Born Adults  20 health care disparities among migrants may be reduced in countries with early intervention (although if these groups have more difficulty accessing early intervention services, disparities may be further exacerbated).A further difference between these countries during the study period concerns immigration and integration policies, with Sweden historically implementing a more inclusionist approach (characterized by a higher immigrant population) than Denmark. 21 the present study, we used Swedish and Danish national registries to examine hospitalization for NAPDs during the first 5 years of illness; to reduce potential heterogeneity in illness course related to age of onset, [22][23][24] study cohorts were restricted to those aged 18 to 35 years at first diagnosis.Our primary aim was to determine if refugee, nonrefugee, and second-generation migrants experienced more days hospitalized for NAPDs during the 5-year follow-up than their native-born peers and if patterns were similar in Sweden and Denmark.In subgroup analyses, we explored the impact of region of birth and duration of residence.Our secondary aim was to investigate whether any observed differences in hospital days were due to these migrant groups experiencing more frequent and/or longer admissions.

Data Sources
This study was approved by the regional ethical review board in Sweden.Data were obtained from multiple, nationwide registries in Sweden and Denmark (eTable 1 in Supplement 1).According to current Swedish and Danish regulations, the use of registry data for research purposes does not require informed consent from individuals held in these registries.Within each country, data were linked by the pseudonymized unique personal identification number assigned to all Swedish and Danish residents at birth or immigration.This study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guidelines.

Study Design and Population
The populations for this prospective cohort study were defined using the National Patient Registries were permitted as we considered these to be part of the first treatment episode).Purchases during the 3 months before cohort entry were permitted as we considered these to be part of the first treatment episode.Individuals entered the cohort on the date of their first contact for NAPD (2006-2013, inclusive) and were followed up for 5 years or until death or emigration.Participant race and ethnicity data were not included in this study because such data are not available in the Swedish and Danish national registries.

JAMA Network Open | Psychiatry
Hospitalization for Nonaffective Psychotic Disorders Among Refugee, Migrant, and Native-Born Adults

Determination of Refugee Status
Refugees included those whose grounds for residence in Sweden or Denmark as registered with immigration authorities was "refugee status" or "family reunification with a refugee."In line with previous studies, 26,27 as grounds of residence data were only available from 1993 in Denmark, we classified immigrants arriving from the major refugee-sending countries in the period 1986 to 1992 as refugees (eTable 1 in Supplement 1).All other individuals born abroad were classified as nonrefugee migrants; second-generation migrants were defined as individuals born in Sweden or Denmark with at least 1 parent who entered as a migrant.Those born in Sweden or Denmark with both parents born in these countries were classified as native born.

Psychiatric Hospitalization During Follow-Up
Our primary outcome was the total number of days spent hospitalized during the 5-year follow-up with a main diagnosis of NAPD (ICD-10 codes F20-F29) at discharge.Among those who had at least 1 hospital admission for NAPD during the 5-year follow-up, we also examined the total number of admissions and mean admission length (total number of hospital days/number of admissions).

Measurement of Covariates
Analyses were adjusted for factors (or proxies for these factors) shown to influence hospitalization in first-episode psychosis populations. 11These included sociodemographic variables (age, gender, family situation, type of residence region, household income at age 18 years, level of education, and unemployment), measures of work disability indexing functional impairment associated with psychiatric and/or somatic conditions (sickness absence, disability pension), and clinical factors (prior treatment for mental disorders, somatic conditions, and suicide attempts; prior psychotropic medication; and NAPD diagnosis at first contact).Covariates were determined using the Swedish and Danish registries detailed in eTable 1 in Supplement 1.For nonrefugee migrant and refugee groups, we additionally measured region of birth and duration of residence.

Statistical Analysis
Statistical analyses were performed from November 2022 to August 2023 in Sweden and Denmark separately using R software, versions 4.2.1 and 4.0.4(R Core Team 28 ), respectively.Population group differences in the primary outcome (total number of hospital days) were estimated using a hurdle model (glmmTMB package) suitable for analyzing count data with excess zeros. 29These 2-part models comprise a binary component and a truncated count component, 30 allowing for the simultaneous modeling of the likelihood of having any hospital days (any vs 0) and the number of days hospitalized among those with at least 1 day.In the present study, we used a logistic (yielding odds ratios [ORs], inversed for interpretability) and truncated negative binomial model (yielding incidence rate ratios [IRRs]), respectively, for these 2 components.Both components included the same covariates with log (follow-up time) used as an offset in the model; observations with missing covariates were excluded from adjusted analyses.For the primary outcome, subgroup analyses were performed to explore the association of region of birth and duration of residence (analyses were not adjusted for covariates due to small group sizes).Analyses of secondary outcomes were restricted to those hospitalized at least once during the follow-up period.A truncated negative binomial model was used to compare population groups on the total number of admissions during the 5-year follow-up (offset = log [follow-up time]).Mean admission length was analyzed using a generalized linear model with gamma distribution and log-link function.All P values were 2-sided, and P < .05 was considered significant.

JAMA Network Open | Psychiatry
Hospitalization for Nonaffective Psychotic Disorders Among Refugee, Migrant, and Native-Born Adults

Total Hospital Days
Descriptive statistics for total hospital days attributable to NAPDs during the 5-year follow-up are provided in  Results for crude and adjusted hurdle models are provided in Table 2 (coefficients for all covariates provided in eTable 4 in Supplement 1).In the adjusted model, the odds of spending at least 1 day in hospital relative to their native-born peers were significantly higher among secondgeneration (OR, 1.17; 95% CI,  a Missing data Swedish cohort: household income at age 18 (n = 1).
c Measured during year of cohort entry.
d Measured on December 31 (Sweden) or September 30 (Denmark) in the calendar year before cohort entry.
e Measured in the 3 relative years (1080 days) before cohort entry date.
f NR to prevent determination of cells with counts less than 10 suppressed.
g Measured in the 6 months (180 days) before cohort entry date.Results of hurdle models applied to total hospital days for nonaffective psychotic disorders over the 5-year study period are shown by region of birth (any days [A] and, among those ever hospitalized, number of days [B]) and duration of residence (any days [C] and, among those ever hospitalized, number of days [D]) in Sweden and Denmark.These 2-component models estimate the odds of experiencing at least 1 day hospitalized (hurdle component − logistic model) and among those hospitalized, the number of days hospitalized (count component − truncated negative binomial model) for migrant groups relative to those who were native born (reference line).

Discussion
In this large population cohort study, we observed that all migrant groups (second generation, nonrefugee, and refugee) had increased levels of hospitalization for NAPDs compared with their native-born peers during the first 5 years of illness.Moreover, the odds of being hospitalized for NAPDs were most marked among nonrefugee migrants and refugees from Africa and those who were within 3 to 5 years of arrival.These patterns were consistent across Sweden and Denmark-2 countries that had markedly different models of early psychosis care and immigration and integration policies during the study period.
Our findings are consistent with those of previous studies that have observed an increased risk of hospitalization among nonrefugee migrants and refugees 14,15 and with Swedish studies showing that the risk is even greater among those who had more recently migrated and individuals who had migrated from Africa. 12,14Importantly, we extended these findings by showing this was also true of

in
Sweden and Denmark according to the International Statistical Classification of Diseases and Related Health Problems, Tenth Revision (ICD-10) 25 codes assigned at discharge or contact (for inpatient and outpatient treatments, respectively).We identified all individuals aged 18 to 35 years in Sweden and Denmark who received their first main diagnosis of NAPD (ICD-10 codes F20-F29) in inpatient or specialist outpatient care between January 1, 2006, and December 31, 2013.To reduce the risk of differential misclassification across population groups, all individuals were required to have resided in the host country (Sweden or Denmark) for at least 3 full calendar years before cohort entry and have no recorded inpatient or outpatient contacts with a main diagnosis of NAPD during the previous 3 years (1080 days).We additionally excluded individuals who had any recorded purchases of antipsychotic medication (Anatomic Therapeutic Chemical classification codes N05A, omitting lithium N05AN) in the 15 months before cohort entry except for the final 3 months (which days was substantially higher among all Danish groups (native born: 53.0 [15.0-139.0]days; second generation: 73.0 [19.0-180.5]days; nonrefugee: 62.0 [18.0-154.3]days; refugee: 83.5 [16.0-204.3]days) relative to their Swedish counterparts (native born: 31.0 [11.0-89.0]days; second generation: 38.0 [12.3-101.3]days; nonrefugee: 35.0 [12.0-91.0]days; refugee: 40.5 [13.0-115.8]days).

Figure .
Figure.Hospital Days for Nonaffective Psychotic Disorders by Region of Birth and Duration of Residence 18nmark over the past 2 decades18but are scarce in Sweden despite recent recommendations for their introduction.19Giventhat early intervention services aim to reduce DUP and improve long-term prognosis by providing prompt assessment and treatment for individuals with first-episode psychosis, JAMA Network Open.2023;6(10):e2336848. doi:10.1001/jamanetworkopen.2023.36848(Reprinted) October 6, 2023 2/14 Downloaded From: https://jamanetwork.com/ by a Copenhagen University Library User on 10/21/2023 intervention in psychosis services.These specialist outpatient services have become widespread in

Table 1 .
We identified 7733 individuals in Sweden (4222 native born [54.6%], 1648 second-generation migrants [21.3%], 861 nonrefugee migrants [11.1%], and 1002 refugees [13.0%]).Mean (SD) age for the total Swedish cohort was 26.0 (5.1) years; 4919 were male (63.6%) and 2814 were female There were some notable differences across population groups.In Sweden and Denmark, it was less common for both native-born and second- refugee, 50 of 541 (9.2%).Native-born individuals were also least likely to be residing in a city at cohort entry (Sweden: 1702 of 4222 [40.3%];Denmark: 2519 of 6367 [39.6%]) and to have lived in a low-income household at age 18 years (Sweden: 476 of 4222 [11.1%];Denmark: 1246 of 6367 [19.6%]).Treatment for other mental disorders in the 3 years before cohort entry and use of any psychotropic medication in the previous 6 months was most common among native-born individuals in both countries.With regard to country-level differences, schizophrenia (ICD-10: F20) was the most common diagnosis at cohort entry in Denmark (3129 of 8747 [35.8%]), whereas those with other psychotic disorders (ICD-10: F24-F29) formed the majority in Sweden (3622 of 7733 [46.8%]).Moreover, a higher proportion of individuals in Sweden than in Denmark received their first NAPD diagnosis in inpatient settings (4052 of 7733 [52.4%] vs 3061 of 8747 [35.0%], respectively).Data on region of birth and duration of residence are provided in eTable 3 in Supplement 1. Europe was the most common region of birth for Swedish nonrefugees (304 of 861 [35.3%]) and refugees (342 of 1002 [34.2%]) and Danish nonrefugees (374 of 904 [41.4%]), whereas most Danish refugees were born in West Asia (210 of 541 [38.8%]).Most refugees and nonrefugee migrants had resided in their host country for 11 or more years; those with a short duration of residency (3-5 years) were the minority across all groups, ranging from 3.7% (20 of 541 Danish refugees) to 15.6% (128 of 861 Swedish nonrefugee migrants).

Table 2 .
For all groups, the proportion of individuals who spent at least 1 day in hospital

Table 1 .
Sociodemographic and Clinical Characteristics of the Swedish and Danish Study Cohorts at Baseline (continued) Abbreviations: NAPD, nonaffective psychotic disorder; NR, not reported.

Table 2 .
Hospital Days for Nonaffective Psychotic Disorders During the First 5 Years of Illness Among Native-Born Individuals and Second-Generation, Nonrefugee, and Refugee Migrants in Sweden and Denmark a age, sex, family situation, region of residence, household income at age 18 years, education level, unemployment, sickness absence, disability pension, previous treatment for any mental disorder, previous treatment for somatic conditions, recent psychotropic medication purchases, psychotic disorder diagnosis at first contact, and calendar year at cohort entry.Hospitalization for Nonaffective Psychotic Disorders Among Refugee, Migrant, and Native-Born Adults a Adjusted models include JAMA Network Open.2023;6(10):e2336848. doi:10.1001/jamanetworkopen.2023.36848(Reprinted) October 6, 2023 8/14 Downloaded From: https://jamanetwork.com/ by a Copenhagen University Library User on 10/21/2023 No. of days in hospital (truncated negative binomial component) by region of birth

Table 3 .
Number and Mean Duration of Hospitalizations for Nonaffective Psychotic Disorders During the First 5 Years of Illness Among Native-Born Individuals and Second-Generation, Nonrefugee, and Refugee Migrants in Sweden and Denmark Group No. of hospital admissions (truncated negative binomial model) a Mean admission length (GLM with log-transformed response) a Variable Definitions and Data Sources eTable 2. Follow-up in the Swedish and Danish Study Cohorts During the 5-Year Study Period eTable 3. Region of Birth and Duration of Residence for Nonrefugee Migrant and Refugee Populations eTable 4. Days Spent in Hospital for Nonaffective Psychotic Disorders During the First 5 Years of Illness Among Native-Born Individuals and Second-Generation, Nonrefugee, and Refugee Migrants in Sweden and Denmark: Results of Fully Adjusted Hurdle Models With Coefficients Provided for All Covariates eTable 5. Days Spent in Psychiatric Hospital for Nonaffective Psychotic Disorders During the First 5 Years of Illness Among Native-Born Individuals and Second-Generation, Nonrefugee, and Refugee Migrants in Sweden and Denmark by Region of Birth and Duration of Residence eTable 6. Number and Mean Duration of Hospitalizations for Nonaffective Psychotic Disorders During the First 5 Years of Illness (for Those Admitted at Least Once) Among Native-Born Individuals and Second-Generation, Nonrefugee, and Refugee Migrants in Sweden and Denmark: Results of Fully Adjusted Models With Coefficients Provided for All Covariates