Introduction
Homelessness, mental disorders, and substance use disorders are highly intertwined (Fazel, Geddes, & Kushel, Reference Fazel, Geddes and Kushel2014; Hodgson, Shelton, van den Bree, & Los, Reference Hodgson, Shelton, van den Bree and Los2013) and associated with negative outcomes, including victimization (Nilsson, Nordentoft, Fazel, & Laursen, Reference Nilsson, Nordentoft, Fazel and Laursen2020), imprisonment and recidivism (Nilsson, Nordentoft, Fazel, & Laursen, Reference Nilsson, Nordentoft, Fazel and Laursen2023), severe morbidity, and mortality (Aldridge et al., Reference Aldridge, Story, Hwang, Nordentoft, Luchenski, Hartwell and Hayward2018; Fine et al., Reference Fine, Dickins, Adams, De Las Nueces, Weinstock, Wright and Baggett2022; Gutwinski, Schreiter, Deutscher, & Fazel, Reference Gutwinski, Schreiter, Deutscher and Fazel2021; Momen et al., Reference Momen, Plana-Ripoll, Agerbo, Benros, Børglum, Christensen and McGrath2020; Nielsen, Hjorthøj, Erlangsen, & Nordentoft, Reference Nielsen, Hjorthøj, Erlangsen and Nordentoft2011; Nilsson et al., Reference Nilsson, Laursen, Osler, Hjorthøj, Benros, Ethelberg and Nordentoft2022). Experiencing multiple of these and other exclusionary situations has been linked to even higher excess mortality (Tweed, Leyland, Morrison, & Katikireddi, Reference Tweed, Leyland, Morrison and Katikireddi2022; Tweed et al., Reference Tweed, Thomson, Lewer, Sumpter, Kirolos, Southworth and Katikireddi2021). People experiencing homelessness have high rates of psychiatric disorders (Gutwinski et al., Reference Gutwinski, Schreiter, Deutscher and Fazel2021), even when examined in younger populations (Burke, Firmin, & Wilens, Reference Burke, Firmin and Wilens2022; Hodgson, Shelton, & van den Bree, Reference Hodgson, Shelton and van den Bree2014; Hodgson et al., Reference Hodgson, Shelton, van den Bree and Los2013; Liu, Koh, Hwang, & Wadhera, Reference Liu, Koh, Hwang and Wadhera2022; Smith-Grant, Kilmer, Brener, Robin, & Underwood, Reference Smith-Grant, Kilmer, Brener, Robin and Underwood2022). The most prevalent psychiatric disorders in people experiencing homelessness have been estimated to be substance use disorders, but high prevalence has been found also for major depression and schizophrenia (Gutwinski et al., Reference Gutwinski, Schreiter, Deutscher and Fazel2021).
Meta-analytic findings have identified psychiatric predictors of homelessness in high-income countries e.g., drug use disorders, psychotic disorders, and behavior problems (Nilsson, Nordentoft, & Hjorthoj, Reference Nilsson, Nordentoft and Hjorthoj2019b). In a systematic review of young people, studies pointed toward a bidirectional relationship between homelessness and psychiatric disorders (Hodgson et al., Reference Hodgson, Shelton, van den Bree and Los2013). High risk of homelessness has furthermore been found among psychiatric inpatients during the year after discharge from psychiatric departments (Nilsson, Laursen, Hjorthoj, & Nordentoft, Reference Nilsson, Laursen, Hjorthoj and Nordentoft2019a). The prevalence of homelessness in young people leaving out-of-home placement has been found to be higher in people with psychiatric disorders than in those without (Chikwava, O'Donnell, Ferrante, Pakpahan, & Cordier, Reference Chikwava, O'Donnell, Ferrante, Pakpahan and Cordier2022), and substance use disorders have been linked to first-time homelessness in a large US sample (Thompson, Wall, Greenstein, Grant, & Hasin, Reference Thompson, Wall, Greenstein, Grant and Hasin2013). However, studies often investigated associations in one direction or lifetime-ever associations and in selected study populations (Chikwava et al., Reference Chikwava, O'Donnell, Ferrante, Pakpahan and Cordier2022; Liu et al., Reference Liu, Koh, Hwang and Wadhera2022; Yoo et al., Reference Yoo, Krawczyk, Johns, McCormack, Rotrosen, Mijanovich and Doran2022). There has been a lack of population-based studies on the temporal relationship between homelessness and psychiatric disorders (Chikwava et al., Reference Chikwava, O'Donnell, Ferrante, Pakpahan and Cordier2022; Gutwinski et al., Reference Gutwinski, Schreiter, Deutscher and Fazel2021; Hodgson et al., Reference Hodgson, Shelton, van den Bree and Los2013). Clarification of this can assist in improving strategies to tackle health inequalities.
We hypothesized that psychiatric disorders and homelessness would be strongly associated in both directions (Gutwinski et al., Reference Gutwinski, Schreiter, Deutscher and Fazel2021; Nilsson et al., Reference Nilsson, Laursen, Hjorthoj and Nordentoft2019a, Reference Nilsson, Nordentoft and Hjorthoj2019b), especially substance abuse, schizophrenia spectrum disorders, and behavioral disorders (Fazel et al., Reference Fazel, Geddes and Kushel2014; Gutwinski et al., Reference Gutwinski, Schreiter, Deutscher and Fazel2021; Hodgson et al., Reference Hodgson, Shelton, van den Bree and Los2013; Nilsson et al., Reference Nilsson, Laursen, Hjorthoj and Nordentoft2019a, Reference Nilsson, Nordentoft and Hjorthoj2019b). The objective was to explore the intertwined nature of psychiatric disorders and homelessness using nationwide registers.
Methods
Study design and participants
We conducted a nationwide cohort study including all individuals living in Denmark at least one day during the follow-up period from 1 January 2002, until 31 December 2021, and born between 1 January 1984, and 31 December 2006. A three-year wash-out period (1999–2001) was used to reduce problems of prevalent cases included as incident cases. Everyone was at the earliest included from their 15th birthday.
We retrieved the study population from the Danish Civil Registration System, which contains data on birth date, sex, vital status, and country of origin (Pedersen, Reference Pedersen2011). A personal identification number is part of this register and makes optimal linkage between registers possible.
The study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guidelines.
Data
We studied psychiatric disorders, i.e., any psychiatric disorder and individual mental disorders and substance use disorders, and homelessness as both exposures and outcomes.
Psychiatric disorders
We measured any psychiatric disorder as any diagnosis in the International Classification of Diseases version ten (ICD-10) F-chapter (Division of Mental Health World Health Organization, 1994) for the period from 1994–2021 with corresponding ICD-8 codes for the time prior to 1994. Psychiatric diagnostic information was based on inpatient, outpatient, and emergency room contacts including public and private health care. Definitions of the individual psychiatric disorders examined can be found in online Supplemental Table S1. The diagnoses were defined according to (ICD-10). The information on psychiatric disorders was retrieved from the Danish Psychiatric Central Research Register (1967–1994) (Mors, Perto, & Mortensen, Reference Mors, Perto and Mortensen2011) and the Danish National Patient Register (1995–2021) (Schmidt et al., Reference Schmidt, Schmidt, Sandegaard, Ehrenstein, Pedersen and Sørensen2015) having information on inpatient contacts since the establishment of the registers and outpatient and emergency room contacts since 1995. First psychiatric diagnosis was used to define the beginning of being exposed with a psychiatric disorder.
Homelessness
Homelessness was measured by first contact to a homeless shelter under the Consolidation Act on Social Services, Section 110 (Danish Ministry for Children and Social Affairs, 2018). Individuals with at least one night accommodated in these shelters were defined as experiencing homelessness. Danish municipalities are obliged to offer temporary stays in these shelters for people experiencing homelessness, if an individual does not have a place to live or is unable to live at home and has personal problems that require extra support. The Danish Homeless Register contains information on all homeless shelter stays from 1 January 1999, to 31 December 2021 (Benjaminsen, Reference Benjaminsen2015; Nielsen et al., Reference Nielsen, Hjorthøj, Erlangsen and Nordentoft2011; Statistics Denmark, 2020). According to FEANTSA's definition of homelessness, our definition covered primarily ‘People living in accommodation for the homeless’ (FEANTSA, 2005), and not individuals living rough or those living temporarily in conventional housing with family or friends who never used the homeless shelters. However, as previously described during a long follow-up period a considerable part of the rough sleepers are likely to use the shelters (Benjaminsen, Reference Benjaminsen2015).
Statistical analyses
Absolute risks
First, we calculated cumulative incidences using the Aalen–Johansen estimator considering emigration and death as competing risks (Aalen & Johansen, Reference Aalen and Johansen1978). The probability of incident sheltered homelessness following any psychiatric disorder and the probability of any psychiatric disorder following first homeless shelter contact was studied. These analyses were compared to the probability in a sex- and age-matched comparison group from the general population consisting of five individuals per case.
Relative risks
Second, we estimated incidence rate ratios (IRRs) which give an estimate of how much more a later condition/circumstance of interest occurred in individuals with a prior condition/circumstance of interest, compared to individuals without the prior condition/circumstance. Separate analyses were conducted for each psychiatric disorder studied. IRRs and 95% confidence interval (CI) were obtained from the Poisson Regression, which approximates a Cox regression model. Adjusting covariates were chosen a priori based on previous studies (Nielsen et al., Reference Nielsen, Hjorthøj, Erlangsen and Nordentoft2011; Nilsson et al., Reference Nilsson, Laursen, Hjorthoj and Nordentoft2019a). We adjusted for sex (male, female), age (one-year groups), calendar time (years), and country of origin (Denmark, other Western countries, and non-Western countries) in all regression analyses. We also conducted these analyses by sex. Piecewise constant intensity for the 1-year age-groups, which is assumed in the Poisson model, was met (Andersen & Keiding, Reference Andersen and Keiding2002). Exposures were handled as time-dependent covariates. An individual counted as exposed from first date of exposure or from start of follow-up for those previously exposed and onward.
In the analysis of (1) the association between psychiatric disorders and the incidence of sheltered homelessness and (2) the association between first homeless shelter contact and the incident of psychiatric disorders, individuals were followed from 1 January 2002, or from their 15th birthday, whichever came last, and until outcome, emigration, death, or end of follow-up on 31 December 2021.
Time-patterns
Third, we studied the risk of sheltered homelessness and psychiatric disorders according to the time in years from exposure to outcome by defining a covariate according to a hierarchical pattern: less than one year, one to two years, two to five years, or five or more years.
Sensitivity analyses
Information on treatment for substance use disorders from the following registers: Registry of Drug Abusers Undergoing Treatment, the National Register on Treatment with Heroin and Methadone, and the National Registry of Alcohol Treatment were included in a sensitivity analysis to use as broad a data basis as possible (Nilsson et al., Reference Nilsson, Laursen, Osler, Hjorthøj, Benros, Ethelberg and Nordentoft2022). In another sensitivity analysis, the year prior to first exposure was included in the exposure definition i.e., people were at risk of outcome one year prior to their first diagnosis or homeless shelter contact.
The statistical analyses were performed using SAS (version 9.4.).
Results
Psychiatric disorders and subsequent homelessness
From 1 January 2002, to 31 December 2021, 1530325 individuals [747 484 females (48.8%) and 782 841 males (51.2%)] aged 15–38 years were included in the study population and accounted for 16 787 562 person-years at risk with 11 433 (0.8%) cases of sheltered homelessness with an IR of 0.7 per 1000 person-years. At the end of follow-up, the median age was 26.1 years (5th percentile: 16.2; 95th: 36.6). Higher IR of sheltered homelessness was found for people with any psychiatric disorder compared with those with no psychiatric disorder (2.7 v. 0.3 per 1000 person-years) (Table 1a).
Study characteristics according to the incidence rate of (a) sheltered homelessness and (b) any psychiatric disorder
a Total numbers for the entire study population: incident homeless shelter cases: 11 433; 16 787 562 person-years, and IR of 0.7 (95% CI 0.7–0.7) per 1000 person-years.
b Total numbers for the entire study population: incident psychiatric disorder cases based on inpatient, outpatient, and emergency room contacts: 210 730; 14 131 060 person-years, and IR of 14.9 (95% CI 14.9–15.0) per 1000 person-years.
Absolute risks of sheltered homelessness
After ten years, 3.0% (95% CI 2.9–3.1) of individuals with any psychiatric disorder compared to 0.8% (95% CI 0.7–0.8) in the age- and sex-matched comparison group from the general population had at least one homeless shelter contact (Fig. 1a and online Supplemental Table S2a).
Probability of (a) sheltered homelessness and of (b) any psychiatric disorder. The figure shows the probability of (a) sheltered homelessness after first psychiatric disorder compared with sex- and age-matched individuals from the general population. The figure also shows the probability of (b) any psychiatric disorder after first homeless shelter contact compared with sex- and age-matched individuals from the general population.
Relative risks of sheltered homelessness
Any psychiatric disorder was associated with 9.2-times increased IR of sheltered homelessness compared with no psychiatric disorder (95% CI 8.8–9.6) after adjustment for sex, age, calendar time, and country of origin (Fig. 2). Each psychiatric disorder was associated with elevated IRs of sheltered homelessness compared with not having that disorder; highest for drug use disorder (IRR 20.6, 95% CI 19.7–21.4). Schizophrenia spectrum disorder and personality disorder were associated with around 10-times increased IRs of homelessness compared with not having these disorders. Also, individual drug use disorders were associated with increased IRs of homelessness compared with not having these disorders (online Supplemental Table S3).
Associations of psychiatric disorders and sheltered homelessness. The figure shows the incidence rate ratios of (a) sheltered homelessness by any and individual psychiatric disorders compared with absence of the specific disorders after adjustment for sex, age, calendar time, and country of origin. The figure also shows the incidence rate ratios of (b) any and individual psychiatric disorders by experiences of sheltered homelessness compared with individuals without homeless shelter contact after adjustment for sex, age, calendar time, and country of origin.
Sex-differences in the risk of sheltered homelessness
Females with any psychiatric disorder had higher IRR of sheltered homelessness (10.1, 95% CI 9.5–10.9) than their male counterparts (8.9, 95% CI 8.5–9.3) when compared with those without a psychiatric disorder. Higher IRR was also found in females than in males for most of the individual disorders including the drug use disorders (e.g. opioids, cannabis, and cocaine) (online Supplemental Tables S5, S6).
Time-patterns for the risk of sheltered homelessness
Figure 3 shows a highly increased IRR of sheltered homelessness during the first year after any psychiatric disorder compared with individuals with no psychiatric disorder (IRR 17.3, 95% CI 16.2–18.5), but high risk was also found during the following years. A 7.8-times (95% CI 7.4–8.1) increased IRR was found for sheltered homelessness even five or more years after first psychiatric disorder compared with no psychiatric disorder (see online Supplementary information). First year after first diagnosis was associated with the highest IRR of sheltered homelessness, also for most individual disorders (Fig. 4). Any drug use disorder compared with no drug use disorder was associated with an IRR of sheltered homelessness of 38.8 (95% CI 36.3–41.6) during first year after the first diagnosis (online Supplemental Table S11).
The association between any psychiatric disorder and sheltered homelessness by time of exposure. The figure shows the incidence rate ratio of sheltered homelessness from first psychiatric disorder and onwards compared with no psychiatric disorder. The figure also shows the incidence rate ratio of any psychiatric disorder from first homeless shelter contact and onwards compared with no homeless shelter contact. Estimates are adjusted for sex, age, calendar time, and country of origin.
Incidence rate ratios of sheltered homelessness and individual psychiatric disorders by time of exposure. The figure shows the incidence rate ratio of sheltered homelessness from first individual psychiatric disorder and onward compared with absence of that disorder. The figure also shows the incidence rate ratio of the individual psychiatric disorders from first homeless shelter contact and onwards compared with no homeless shelter contact. Estimates are adjusted for sex, age, calendar time, and country of origin. The y-axis for drug use disorder is up to IRR of 50 instead of 30 as this one differed substantially from the others.
Homelessness and subsequent psychiatric disorders
During the study period, 1 406 410 individuals [697 341 females (49.6%); 709 069 males (50.4%)] aged 15–38 years were included in the study population and accounted for 14 131 060 person-years at risk of any psychiatric disorder. In total, 210 730 individuals had a psychiatric disorder during follow-up (IR 14.9 per 1000 person-years). At the end of follow-up, the median age was 24.5 years (5th percentile: 15.9; 95th percentile: 36.3). Higher IR for any psychiatric disorder was found for people experiencing sheltered homelessness compared with not having these experiences of homelessness (76.0 v. 14.8 per 1000 person-years) (Table 1b).
Absolute risks of any psychiatric disorder
After ten years of follow-up, 47.1% (95% CI 45.3–48.0) of the individuals with at least one homeless shelter contact compared to 11.1% (95% CI 10.7–11.5) in the age- and sex-matched comparison group from the general population had at least one psychiatric disorder (online Supplemental Table S2b).
Relative risks of psychiatric disorders
The adjusted IRR of any psychiatric disorder in people experiencing sheltered homelessness compared with those without shelter contact was 7.0 (95% CI 6.7–7.4) (Fig. 2). Especially high risk was found for drug use disorder (IRR 15.3, 95% CI 14.6–16.1) and schizophrenia (IRR 14.7, 95% CI 13.7–15.9) for individuals experiencing sheltered homelessness compared with not using homeless shelters. However, more than 9-times increased IRs of e.g., behavioral disorder, alcohol use disorder, personality disorder, and organic disorder were linked to sheltered homelessness. Sheltered homelessness was furthermore associated with increased IRR of individual drug use disorders compared with not using homeless shelters (online Supplemental Tables S7).
Sex-differences in the risk of psychiatric disorders
Males experiencing sheltered homelessness had a higher IRR of any psychiatric disorder (7.9, 95% CI 7.5–8.4) than females with sheltered homelessness (IRR 5.4, 95% CI 5.0–5.9) compared to males and females without homeless shelter use. Males experiencing homelessness had higher IRR of drug use disorder, schizophrenia spectrum disorder, personality disorder, and behavioral disorder than their female counterparts when compared with those without homeless shelter use. However, females experiencing sheltered homelessness compared with those without these experiences had higher IRRs of several individual drug use disorders than males (online Supplemental Tables S9, S10).
Time-patterns in the risk of psychiatric disorders
During the year after first homeless shelter contact, the IRR of any psychiatric disorder associated with homelessness compared with no homeless shelter contact was 13.3 (95% CI 12.3–14.4) (Fig. 3) with increased risk also during the following years (see online Supplementary information). This pattern was also found for most individual psychiatric disorders (Fig. 4). Sheltered homelessness compared with no homeless shelter use was most strongly associated with increased rates of drug use disorder (IRR 29.8, 95% CI 27.6–32.3), schizophrenia spectrum disorder (IRR 22.0, 95% CI 19.8–24.4), and behavioral disorder (IRR 23.1, 95% CI 20.9–25.5) during first year after first homeless shelter contact. Even higher IRRs were found for the individual drug use disorders (online Supplemental Table S12).
Sensitivity analyses
Inclusion of information from the substance use treatment registers did not change the results (online Supplemental Table S3, S4 and S7, S8).
When including the year prior to exposure as a window of exposure, high IRRs of both outcomes were found. For most disorders, the rate of sheltered homelessness or psychiatric disorder was higher during the year prior to first psychiatric disorder or first homeless shelter contact (i.e. exposure) than during the period after first year from exposure. Risk patterns varied by disorder (online Supplemental Figs S13, S14 and Table S15, S16).
Discussion
Based on population-based data of 11 433 people who had at least one homeless shelter contact and 210 730 people who had a psychiatric disorder over 20 years, strong associations between psychiatric disorders and homelessness in individuals aged 15–38 years were found. Ten years after first psychiatric disorder, 3.0% experienced sheltered homelessness. Among people experiencing sheltered homelessness, 47.1% had a psychiatric disorder within ten years from their first homeless shelter contact. Increased adjusted relative risks were found in both directions of the association, in males and females, and for all individual psychiatric disorders studied, especially during the first year after first psychiatric disorder and first homeless shelter contact, respectively, but with strong associations several years after. Drug use disorder and schizophrenia spectrum disorder were most strongly associated with sheltered homelessness.
The strong associations between psychiatric disorders and homelessness support previous findings of an intertwined relationship (Gutwinski et al., Reference Gutwinski, Schreiter, Deutscher and Fazel2021; Nilsson et al., Reference Nilsson, Laursen, Hjorthoj and Nordentoft2019a, Reference Nilsson, Nordentoft and Hjorthoj2019b; Thompson et al., Reference Thompson, Wall, Greenstein, Grant and Hasin2013), but also adds important information e.g., about the timing from first psychiatric diagnosis and from first homeless shelter contact. Previous studies have shown that substance use disorders predict psychotic problems in people experiencing homelessness (Jones et al., Reference Jones, Gicas, Seyedin, Willi, Leonova, Vila-Rodriguez and Honer2020). Additionally, multiple v. fewer of psychiatric and social exposures have been linked to higher risk of mortality (Tweed et al., Reference Tweed, Leyland, Morrison and Katikireddi2022, Reference Tweed, Thomson, Lewer, Sumpter, Kirolos, Southworth and Katikireddi2021). The compiled evidence supports the importance of early intervention to avoid the accumulation of psychiatric and social problems.
Females were to a higher degree than males at risk of homelessness after first psychiatric disorder. This supports that females experiencing homelessness constitute a group of highly vulnerable individuals with high rates of psychiatric morbidity and other service needs (Arnos & Acevedo, Reference Arnos and Acevedo2023; Milaney, Williams, Lockerbie, Dutton, & Hyshka, Reference Milaney, Williams, Lockerbie, Dutton and Hyshka2020; Winetrobe et al., Reference Winetrobe, Wenzel, Rhoades, Henwood, Rice and Harris2017). Social factors might contribute more to the risk of homelessness in males or more undiagnosed psychiatric disorders predating their homelessness. However, females experiencing homelessness have been found to be at higher risks of violent victimization (Nilsson et al., Reference Nilsson, Nordentoft, Fazel and Laursen2020) and adverse childhood trauma than males (Milaney et al., Reference Milaney, Williams, Lockerbie, Dutton and Hyshka2020). Meta-analytic findings showed that around a quarter of young females in developed regions reported abuse-related reasons for street involvement, which was more than in males (Embleton, Lee, Gunn, Ayuku, & Braitstein, Reference Embleton, Lee, Gunn, Ayuku and Braitstein2016).
Males experiencing homelessness had higher IRRs of psychiatric disorder than females. However, a noticeable finding was that females experiencing sheltered homelessness had higher IRRs of several individual drug use disorders than males, although it for any drug use disorder was lower.
This study confirms that psychiatric disorders are important predictors of becoming homeless (Chikwava et al., Reference Chikwava, O'Donnell, Ferrante, Pakpahan and Cordier2022; Nilsson et al., Reference Nilsson, Laursen, Hjorthoj and Nordentoft2019a, Reference Nilsson, Nordentoft and Hjorthoj2019b; Thompson et al., Reference Thompson, Wall, Greenstein, Grant and Hasin2013). More knowledge of effective interventions is needed focusing on the prevention of social marginalization in psychiatrically vulnerable individuals (e.g. prior to and during discharge from psychiatric admission, leaving out-of-home placement, and release from prison). First health-care contact could also be important for implementing prevention strategies.
Among persons with a first-time homeless shelter contact, but without prior hospital-based psychiatric disorder, approximately 15% will receive a psychiatric diagnosis within the next year. Furthermore, the increased risks of psychiatric disorders remained even after two years of follow-up. This leads to important research questions: are some of these previously undiagnosed psychiatric disorders predating the homelessness episode that emerge over time due to severity, or was the period before homelessness associated with emergent symptoms and it was later that the illness became clear? There is a need for studies clarifying this to show whether our findings indicate problems of undetected diagnoses, or we should be better to recognize specific symptoms and risk states earlier to reduce the risk of homelessness. Irrespective of the cause, the first homeless shelter contact becomes an effective time window to launch a preventive strategy to mitigate or prevent psychiatric disorders. Housing first interventions have been found to be effective e.g., in a Canadian setting to improve housing stability, mental functioning and hospital use after 4-year follow-up in people experiencing homelessness (Loubière et al., Reference Loubière, Lemoine, Boucekine, Boyer, Girard, Tinland and Auquier2022). However, there are substantial gaps in the knowledge of effective intervention strategies.
The strong bidirectional relationship between sheltered homelessness and psychiatric disorders might also be explained by shared risk factors. Such factors could be adverse childhood experiences and parental factors (Fazel et al., Reference Fazel, Geddes and Kushel2014; Hodgson et al., Reference Hodgson, Shelton, van den Bree and Los2013; Nilsson, Laursen, Hjorthoj, Thorup, & Nordentoft, Reference Nilsson, Laursen, Hjorthoj, Thorup and Nordentoft2017; Nilsson et al., Reference Nilsson, Laursen, Hjorthoj and Nordentoft2019a, Reference Nilsson, Nordentoft and Hjorthoj2019b). Future studies of these associations are needed. A meta-analysis of reported reasons for street involvement found that psychosocial factors, any type of abuse, poverty, and family conflict were frequently reported reasons (Embleton et al., Reference Embleton, Lee, Gunn, Ayuku and Braitstein2016).
Strengths and limitations
Our study has important strengths. Using the Danish nationwide register-data we were to our knowledge for the first time able to study the two-way association between psychiatric disorders and sheltered homelessness. Additionally, we were able to study individuals during a 20-year period with complete follow-up information, sex-specific associations, and include a high level of details.
The study also has limitations. First, due to the administrative data, we were not able to study real incidences of psychiatric disorders and homelessness. Our data on psychiatric disorders were based on inpatient, outpatient, and emergency room contacts and not from the general practitioners. We had no information on the specific treatment received. There will be some uncertainty associated with the date of onset of the disorder, and our data are expected to capture the time of first severe diagnosis requiring treatment in the hospital system. However, diagnoses of severe mental illnesses such as schizophrenia and bipolar disorder are most often received in the secondary sector and most of the register-based psychiatric diagnoses have been found to be valid (Munk-Jorgensen & Ostergaard, Reference Munk-Jorgensen and Ostergaard2011). Furthermore, information on undiagnosed disorders and unsheltered homelessness is not available. We only had data on homeless shelter contacts since 1999, but we used a three-year washout period to limit problems of mixing prevalent and incident homeless cases. Furthermore, due to the lack of register-data prior to 1999, we were only able to study a young cohort with complete follow-up. Furthermore, the results may not be generalizable to homeless youth below 18 years, who may to a higher degree choose other options for accommodations. We did not study the contribution of migration status in the current study, but this is an area for future research.
Conclusions
This study confirms the strong associations between homelessness and psychiatric disorder and quantifies the high probabilities that having one of the conditions gives for subsequently having the other. Individuals with psychiatric disorders are at high risk of homelessness. Individuals experiencing sheltered homelessness are also at high risk of either developing psychiatric disorders or of receiving psychiatric diagnoses that had already emerged. The year after first psychiatric disorder or first homeless shelter contact constitutes a high-risk period for adverse outcomes and an opportunity for intervention. Health and social care professionals should be aware of the high risks of accumulated psychiatric and social problems from adolescence and early adulthood and onwards. Prevention efforts to reduce these accumlulated risks are needed.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/S0033291723002428
Acknowledgements
This study was funded by a grant from Lundbeckfonden to SFN (grant number F-61171-23-50). SF is funded by the Wellcome Trust (#202836/Z/16/Z) and Oxford Health NIHR Biomedical Research Council.
Author's contributions
SFN obtained funding for the study. All authors designed the study. SFN and TML had full access to all data in the study and verify the underlying data. SFN analyzed the data with supervision from TML. SFN and TML take responsibility for the integrity of the data and the accuracy of the data analyses. All authors interpreted the data. SFN drafted the manuscript. All authors critically revised the manuscript.
Competing interest
None.
Ethical standards
The study was approved by the Danish Data Protection Agency, and data access was agreed by Statistics Denmark and the Danish Health Data Authority. Approval by the Ethics Committee and written informed consent is not required for register-based projects. All data were de-identified and not recognizable at an individual level.
Data statement
The data that support the findings of this study are available from Statistics Denmark. The data access requires the completion of a detailed application form from the Danish Data Protection Agency, the Danish National Board of Health and Statistics Denmark. For more information on accessing the data, see https://www.dst.dk/en.
