The study base comprised individuals who were 19–30 years old, residing in Sweden on December 31st, 2009 (n = 1,425,496). Individuals were excluded if they had incomplete information on reason for settlement in Sweden (n = 52,127). Moreover, non-refugee immigrants (n = 152,641) and those who had missing information on the year of immigration to Sweden (n = 388) were excluded. Additionally, those who came to Sweden when aged 18–30 years (n = 19,387) were not included, as the study focused on refugees who came to Sweden as minors. The final study population included 1,200,953 individuals, of whom 51,098 (4.25%) came to Sweden as refugee minors, i.e. below the age of 18. A flow chart illustrates the selection steps to reach the final study population (Supplementary Fig. 1).
The data sources used in this study were Swedish population registers that were linked by an anonymised personal identification number. From Statistics Sweden, the Longitudinal Integration Database for Health Insurance and Labour Market Studies (LISA) provided information on socio-demographics such as sex, age, type of residential area, family situation and educational level. LISA also provided information on LMM characteristics, including number of days with unemployment, number of net days with sickness absence and disability pension. STATIV (The Longitudinal Database for Integration Studies), also from Statistics Sweden, comprised information on reason for residence, i.e. refugee status. From the National Board of Health and Welfare, The National Patient Register (NPR) included information on date and diagnosis regarding in- and specialised outpatient healthcare, coded according to the International Classification of Diseases (ICD-10) . The Causes of Death Register, also from the National Board of Health and Welfare, provided information on date and cause of death .
The Swedish Migration Agency follows European Union regulations, Swedish legislation and the United Nations Convention relating to the status of refugees regarding granting residence permit in Sweden . Those individuals who were grouped as “refugees” had either of the following indicated for their reason for residing in Sweden: “refugee”, “in need of protection” or on “humanitarian grounds” .
Young refugees who arrived as minors (< 18 years) in Sweden and who had at least one accompanying parent at the time of arrival were considered as accompanied refugee minors . To identify refugees who were accompanied by at least one parent at the time of their arrival in Sweden, we used register data on their parents’ residency status and year of immigration. The young refugees were classified as ‘accompanied refugee minors’ if at least one of their parents’ has migrated to Sweden in the same year or any year before the young refugees arrived in Sweden. Those who did not meet these criteria were classified as ‘unaccompanied refugee minors’. Among the 51,098 young refugees in this study, 46,505 (91%) and 4593 (9%) were accompanied and unaccompanied refugee minors, respectively. All individuals born in Sweden were identified as “Swedish-born”.
The main outcome was hospitalisation due to suicide attempt. The secondary outcome was suicide. Suicide attempts were defined as having a diagnosis from hospital admissions during follow-up, coded with ICD-10 codes X60–X84 (self-harm) or Y10–Y34 (events of undetermined intent), retrieved from the NPR. The cases belonging to “undetermined intent” were included to reduce underreporting and to guarantee coherence regarding case ascertainment . Suicide was identified through the ICD-10 codes X60–X84 or Y10–Y34 in the Cause of Death Register.
Socio-demographic covariates included age, sex, educational level, type of residential area, and family situation. They were measured at baseline, January 31st, 2009. LMM factors included unemployment, sickness absence and disability pension, which were all measured during the baseline year (2009). Health-related factors were dichotomised and included mental disorders, with a history of any main or secondary diagnosis with any ICD-10 code ‘F’ in inpatient or specialised outpatient healthcare. Health-related factors also included history of (at least one) hospitalisation due to suicide attempt, and lastly, history of inpatient or specialised outpatient healthcare due to somatic illness from a main or secondary somatic diagnosis with any ICD-10 code (except ‘F’, ‘O’, ‘P’ and ‘Q’ codes). All health-related factors were measured during 2005–2009 and this time period was chosen to use the available data in the best possible way. All covariates were categorised, as seen in Table 1, with missing values shown as separate categories for covariates with missing information.
Labour market factors were also considered as possible confounders, with adjustments made for unemployment, sickness absence and disability pension. Refugees are more likely to be unemployed compared to the general population [19, 27]. Health-related factors were treated as possible confounders, with adjustments made for previous history of somatic disorders, previous history of mental disorders and previous history of hospitalisation for suicide attempt. History of mental disorders is more common among refugees compared to the general population , and rates for suicide attempt were reported to be lower among refugees compared to Swedish-born .
Descriptive statistics with frequencies and percentages were calculated for accompanied refugee minors, unaccompanied refugee minors and the Swedish-born population (Table 1). Cox proportional hazard regression models were used to estimate multivariate-adjusted hazard ratios (HR) with 95% confidence intervals (CI) for suicide attempt and suicide in accompanied and unaccompanied refugee minors compared to Swedish-born. Individuals were followed up from baseline, January 1st, 2010, until whichever of the following came first; death, outcome event, emigration or end of follow-up (December 31st, 2016). A Kaplan–Meier survival estimator was used to test the proportional hazard assumption, which was found to be met. Four regression models were applied to adjust for the covariates (Model 1 adjusting for age and sex; Model 2 additionally adjusting for other socio-demographic covariates; Model 3 adjusting for the covariates in model 2 and the LMM factors; Model 4 adjusting for the covariates in model 3 and the health-related factors.
An interaction test, i.e. partial likelihood ratio test was conducted to test the interaction between educational level and refugee status regarding risk of subsequent suicide attempt where educational level was entered as a categorised variable in the model (four categories: compulsory school (0–9 years), high school (10–12 years), college or university (> 12 years) and one category with missing information). Following the same method, interaction between history of mental disorder (Yes, No) and refugee status regarding risk of subsequent suicide attempt was also tested. All mentioned analyses were also performed for suicide as outcome measure. For all analyses a p value < 0.05 was considered statistically significant. All statistical analyses were performed using SAS 9.4 for Windows, besides the partial likelihood ratio test which was conducted using SPSS version 25.
A sensitivity analysis was performed where those with “in need of protection” or on “humanitarian grounds” were excluded from the refugee category. Another sensitivity analysis was conducted where cases of ‘undetermined intent’ were not included for comparison to the main analysis. Finally, we conducted a sensitivity analysis to investigate the influence of migration-related factors (country of birth and age at arrival) on the risk of suicidal behaviour for unaccompanied refugees compared with accompanied refugees (reference category). As these variables are not applicable for the Swedish-born, they were not included in this analysis as the reference group.
Ethical approval was obtained from the Regional Ethical Review Board, Karolinska Institutet, Stockholm, Sweden (Dnr: 2007/1762-31).