OUP user menu

★ Editor's Choice ★

Suicide and unintentional injury mortality among homeless people: a Danish nationwide register-based cohort study

Sandra Feodor Nilsson, Carsten Rygaard Hjorthøj, Annette Erlangsen, Merete Nordentoft
DOI: http://dx.doi.org/10.1093/eurpub/ckt025 50-56 First published online: 12 March 2013

Abstract

Background: Homeless people have elevated mortality, especially due to external causes. We aimed to examine suicide and unintentional injury mortality levels and identify predictors in the homeless population. Methods: A nationwide, register-based cohort study of homeless people aged 16 years and older was carried out using the Danish Homeless Register, 1999–2008. Results: In all, 32 010 homeless people (70.5% men) were observed. For men, the mortality rate was 174.4 [95% confidence interval (CI) = 150.6–198.1] per 100 000 person-years for suicide and 463.3 (95% CI = 424.6–502.0) for unintentional injury. For women, the corresponding rates were 111.4 (95% CI = 81.7–141.1) for suicide and 241.4 (95% CI = 197.6–285.1) for unintentional injury. Schizophrenia spectrum, affective, personality and substance use disorders were strongly associated with increased risk of suicide; the highest risk estimates were found for schizophrenia spectrum disorders among both men [hazard ratio (HR) = 3.1, 95% CI = 2.0–4.9] and women (HR = 15.5, 95% CI = 4.5–54.0). Alcohol and drug use disorders were predictors of death by unintentional injury for both men and women, whereas schizophrenia spectrum disorders and personality disorders were only significant predictors among men; the highest risk estimates were found for drug use disorders among men (HR = 2.2, 95% CI = 1.8–2.8) and women (HR = 3.1, 95% CI = 1.8–5.4). A history of psychiatric admission and emergency room contact were predictors for dying by suicide and unintentional injury. Conclusion: People in the homeless shelter population with a history of a psychiatric disorder constitute a high-risk group regarding the elevated suicide and unintentional injury mortality.

Introduction

Homeless people have elevated mortality levels.14 External causes of death such as suicide and unintentional injury are among the leading causes of death in this vulnerable population,3,5,6 and homeless people have between 7 and 19 times the risk of dying from these causes compared with the general population.1

Homelessness is associated with social deprivation, an unhealthy life style with elevated levels of psychiatric disorders, especially substance use disorders,1,7,8 somatic disorders,9 violence,10 injuries11 and unmet healthcare needs.12,13 In addition, homeless people are more frequently hospitalized at younger ages and more likely to be admitted for a psychiatric disorder than housed individuals.14,15 Substance use disorders can severely affect mental health; both cannabis and alcohol use are linked to a higher risk of mental disorders16 and psychotic outcomes.17 Previous studies indicate that the majority of days in psychiatric hospital care and the excess mortality in homeless people relate to substance use disorders.1,4,5,8 Furthermore, people with multiple or short stays in homeless shelters have been found to have increased risk of suicide.2 However, more information on the mechanisms involved in the excess mortality due to external causes among homeless people is needed to improve their conditions of life.18 Causes of death among homeless people have been studied,36,19 but little is known about predictors of cause-specific deaths in this population.2,20

Psychiatric disorders are associated with increased risk of death by suicide21,22 and unintentional injury.23 However, due to homeless people’s high-risk profile and excess mortality, the predictors or the predictive value of known predictors of death from suicide and unintentional injury may differ from what has been found in the general population.

Danish nationwide registration of people in contact with homeless shelters provides us with a unique opportunity to examine suicide and unintentional injury mortality levels and to identify predictors.

Methods

Study population

All persons aged 16 years and older with at least one contact with a Danish homeless shelter during the study period from 1 January 1999 to 31 December 2008 were included. A further inclusion criterion was a valid Civil Registration System (CRS) number that is assigned to all Danish residents. A person was defined as homeless from the date of the first stay in a Danish homeless shelter during the study period.

Measures

Data sources

A longitudinal cohort design was applied using information from the Homeless Register,24 the CRS,25 the Psychiatric Central Register26 and the Cause of Death Register.27 The CRS number enabled individual-level linkage.25

In Denmark, all residents in homeless shelters are registered with their CRS number in the Homeless Register, which is maintained by the National Social Appeals Board.24 The CRS was used to obtain information on emigration and loss to follow-up. Data on causes of death were obtained from the Cause of Death Register.27 Information on admissions to psychiatric hospitals since 1969 as well as psychiatric outpatient and emergency room contacts from 1995 was obtained from the Psychiatric Central Register.26

Before 1994 psychiatric disorders were registered according to 8th revision of the International Classification of Diseases (ICD-8),28 and thereafter according to the 10th revision ICD-10.29 Individuals were classified into following groups of disorders: schizophrenia spectrum disorders, substance use disorders, affective disorders and personality disorders as described earlier.1

Outcomes of interest

The outcomes of interest were death by suicide (ICD-10: X60–X84, Y87.0) and unintentional injury (ICD-10: V01–V89, V90–V99, W00–X59, Y40–Y86, Y87.2, Y88–Y89). Deaths by events of undetermined intent were not included (ICD-10: Y10–Y34).27

Predictors of interest

We examined following potential predictors: history of psychiatric contact (any psychiatric contact where the patient has been diagnosed with a psychiatric disorder), schizophrenia spectrum disorders, alcohol use disorders, drug use disorders, affective disorders, personality disorders, specific types of registered substance use disorders (alcohol, opioids, cannabis, sedatives, cocaine, other drugs), type of psychiatric contacts (ever inpatient, outpatient, emergency room only) and psychiatric inpatient days (measured within the 5 years before the first registered homeless shelter contact) and number of contacts to homeless shelters within the first calendar year of contact (1, 2–5, >5).

Statistical analysis

Mortality rates (MRs) were calculated by sex and cause per 100 000 person-years as averages over the period 1999–2008.

Adjusted Cox proportional hazard regression models were constructed. People were considered at risk of dying by suicide and unintentional injury from their first homeless shelter contact and until death, emigration, loss to follow-up or end of study, whichever came first. The time unit was number of days. Initially, we tested if gender interacted with psychiatric contact with regard to death by suicide and unintentional injury. The regression analysis included 22 508 men and 9439 women, as 63 individuals were excluded from the analyses because of negative follow-up time (i.e. if a person was registered as dead before the registration in some of the other registries). In all, 802 (2.5%) individuals emigrated and 186 (0.6%) individuals were lost during follow-up. The mean follow-up time was 1926 days (SD = 1109) for men and 1875 days (SD = 1088) for women. The analyses were adjusted for age at first homeless shelter contact recorded in the Homeless Register, country of origin and main source of income. All analyses were carried out using SPSS (version 18).

Permission to use the data was obtained from the Danish Data Protection Agency (2009-41-4021). Register-based studies in Denmark do not require permission from ethics committees.

Results

Characteristics of the study population

From January 1999 to December 2008, 32 158 individuals were registered in the Homeless Register with a valid CRS number. Of those, 148 children below 16 years of age were excluded. The study population consisted of 32 010 (70.5% men) persons. A majority of 14 104 (62.5%) men and 5507 (58.2%) women had a record of a psychiatric diagnosis. The most frequent diagnosis was substance use disorders. Among the men, 8714 (38.6%) had a history of an alcohol use disorder and 5446 (24.1%) had a history of drug use disorder; for women the corresponding numbers were 2813 (29.8%) and 1744 (18.4%). A total of 4161 (13.0%) persons died during the 167 190 person-years of follow-up.

Mortality rates

The annual suicide MR was 174.4 [95% confidence interval (CI) = 150.6–198.1] and 111.4 (95% CI = 81.7–141.1) per 100 000 persons for men and women, respectively. The annual MR for deaths by unintentional injury was 463.3 (95% CI = 424.6–502.0) per 100 000 men and 241.4 (95% CI = 197.6–285.1) per 100 000 women.

Predictors of death by suicide and unintentional injury

Gender interacted significantly with a history of psychiatric contact both on risk of dying from suicide (P = 0.04) and unintentional injury (P = 0.09). Further analyses were, thus, conducted separately for men and women to elucidate potential gender differences. Since analyses presented in the tables were performed separately for men and women, they did not allow for comparison of gender effects on mortality. In a separate model, we estimated the direct effect of gender on suicide and unintentional injury mortality and found that men had higher risk of dying from suicide than women [hazard ratio (HR) = 1.5, 95% CI = 1.1–2.0] and unintentional injury (HR = 1.8, 95% CI = 1.5–2.2) (not shown in table).

In terms of sociodemographic covariates, we found that men aged 25–34 years had increased risk of dying by unintentional injury compared with those aged 16–24 years, whereas the age groups 25–34, 35–44 and 45–54 were associated with increased risk in women (table 1). People from low- and middle-income countries had reduced risk of dying by suicide (not significant in women) and unintentional injury compared with those from Denmark. Furthermore, men with income from employment had reduced risk of dying by unintentional injury.

View this table:
Table 1

Multiple Cox regression analyses of sociodemographic and homelessness predictors for the outcomes, death by suicide and unintentional injury, among homeless men (N = 22 508) and women (N = 9439) during 1999–2008

VariablesSuicideUnintentional injury
N (%)aHR95% CIPN (%)aHR95% CIP
Men
 Age at first homeless shelter contactb
        16–2419 (0.7)1.00.945 (1.7)1.0<0.0001
        25–3448 (0.9)1.20.7–2.0179 (3.4)1.81.3–2.5
        35–4465 (0.9)1.20.7–2.1170 (2.5)1.30.9–1.8
        45–5453 (1.0)1.40.8–2.3115 (2.2)1.20.8–1.7
        55–6417 (0.8)1.20.6–2.331 (1.4)0.80.5–1.4
        ≥654 (0.9)1.70.6–5.010 (2.4)1.40.7–2.9
 Country of originb,c
        Denmark188 (1.0)1.00.02498 (2.7)1.0<0.0001
        Nordic countries6 (1.3)1.40.6–3.217 (3.6)1.61.0–2.6
        EU5 (1.3)1.40.6–3.58 (2.1)0.90.4–1.8
        Other high-income countries0 (0.0)0.02 (3.1)1.30.3–5.1
        Low- and middle-income countries7 (0.2)0.30.1–0.625 (0.8)0.30.2–0.5
 Main source of incomeb
        Cash benefit69 (0.8)1.00.2208 (2.5)1.00.001
        Pension46 (1.0)0.80.5–1.2153 (3.2)1.00.8–1.3
        Income from employment15 (0.8)1.00.6–1.818 (1.0)0.40.3–0.7
        Other45 (1.2)1.20.9–1.877 (2.0)0.80.6–1.0
        Unknown31 (0.8)0.80.5–1.294 (2.5)0.90.7–1.1
 Number of homeless shelter contactsd
        1119 (0.9)1.01.0238 (1.8)1.0<0.0001
        2–570 (0.9)1.00.7–1.5233 (3.0)1.41.1–1.7
        >517 (1.0)1.10.6–1.979 (4.8)2.11.6–2.8
Women
 Age at first homeless shelter contactb
        16–248 (0.5)1.00.65 (0.3)1.00.09
        25–348 (0.3)0.60.2–1.732 (1.2)3.91.5–10.0
        35–4421 (0.8)1.20.5–2.933 (1.3)2.81.1–7.3
        45–5413 (0.8)0.90.3–2.336 (2.1)3.51.3–9.1
        55–643 (0.4)0.50.1–2.29 (1.3)2.20.7–6.9
        ≥651 (0.9)1.20.1–9.82 (1.7)2.80.5–14.9
 Country of originb,c
        Denmark50 (0.8)1.00.7109 (1.6)1.00.009
        Nordic countries1 (0.4)0.50.1–3.82 (0.8)0.50.1–1.9
        EU0 (0.0)0.03 (1.9)1.60.5–5.1
        Other high-income countries0 (0.0)0.01 (4.0)2.30.3–16.3
        Low- and middle-income countries3 (0.1)0.40.1–1.42 (0.1)0.10.0–0.4
 Main source of incomeb
        Cash benefit17 (0.4)1.00.427 (0.7)1.00.001
        Pension19 (1.1)1.50.7–3.253 (3.1)2.51.5–4.2
        Income from employment1 (0.1)0.30.0–2.43 (0.4)0.50.1–1.6
        Other10 (0.6)1.40.6–3.018 (1.1)1.40.7–2.5
        Unknown7 (0.5)0.90.4–2.216 (1.2)1.20.7–2.3
 Number of homeless shelter contactsd
        133 (0.5)1.00.457 (0.8)1.00.01
        2–516 (0.7)1.20.6–2.552 (2.2)1.91.2–2.9
        >55 (1.6)2.10.7–6.38 (2.6)1.90.9–4.1
  • aN of total number of individuals in each category dying by specific cause of death during follow-up.

  • bThe Cox regression model included four variables: a history of psychiatric contact (defined as any diagnosis of a psychiatric disorder), age at first homeless shelter contact listed in the Homeless Register, country of origin and main source of income (defined as main source of income at last contact in the Homeless Register).

  • cThis variable is defined by following categories: ‘Denmark’, ‘Nordic countries’ including Sweden, Finland, Iceland, Norway and the Faroe Islands, ‘EU’ including Belgium, Bulgaria, Cyprus, Estonia, France, Greece, Ireland, Italy, Latvia, Lithuania, Luxembourg, Malta, Netherlands, Poland, Portugal, Romania, Slovakia, Slovenia, Spain, UK, Czech Republic, Germany, Hungary and Austria, ‘other high-income countries’ including USA, Australia, Canada, New Zealand, Switzerland, Liechtenstein, Monaco, Andorra, San Marino and Vatican City and ‘low- and middle-income countries’ including all other countries.

  • dMeasured within the first year of homeless shelter contact recorded in the Homeless Register. The model includes only individuals who were alive at least 1 year after the first contact to a homeless shelter (n = 20 316 for men; n = 8499 for women). The model is adjusted for all other covariates shown in the table and a history of psychiatric contact.

A dose–response effect was found between an increasing number of shelter visits and death by unintentional injury among men; in women, between two and five homeless shelter contacts were associated with an elevated risk of dying compared with a single contact.

The adjusted regression models revealed an elevated risk of dying by suicide and unintentional injury for individuals with a history of schizophrenia spectrum disorders, alcohol use disorders, drug use disorders, affective disorders and personality disorders when compared with no psychiatric contact (table 2). Especially, homeless women with psychiatric disorders displayed high risk of dying by suicide. The most pronounced HRs were found for schizophrenia spectrum disorders in both men (HR = 3.1, 95% CI = 2.0–4.9) and women (HR = 15.5, 95% CI = 4.5–54.0) compared with no psychiatric contact. All psychiatric disorders were predictors of death by unintentional injury among men; however, the estimate for affective disorders was only borderline significant. Women with a history of either alcohol or drug use disorders were linked to an increased risk of dying by unintentional injury when compared with no psychiatric contact. Personality disorders had only a borderline significant effect on unintentional injury in women. For unintentional injury, the highest risk estimates were found for drug use disorders in both men (HR = 2.2, 95% CI = 1.8–2.8) and women (HR = 3.1, 95% CI = 1.8–5.4).

View this table:
Table 2

Multiple Cox regression analyses of psychiatric predictors for death by suicide and unintentional injury among homeless men (N = 22 508) and women (N = 9439) during 1999–2008

VariablesSuicideUnintentional injury
N (%)aHRb95% CIPN (%)aHRb95% CIP
Men
    Psychiatric contactc170 (1.2)2.51.7–3.6<0.0001422 (3.0)1.61.3–2.0<0.0001
 Specific psychiatric diagnosesd
    Schizophrenia spectrum disorders43 (1.3)3.12.0–4.9<0.0001109 (3.4)1.81.4–2.3<0.0001
    Alcohol use disorders112 (1.3)2.51.7–3.7<0.0001273 (3.1)1.61.3–2.0<0.0001
    Drug use disorders69 (1.3)2.71.8–4.1<0.0001241 (4.4)2.21.8–2.8<0.0001
    Affective disorders56 (1.5)3.02.0–4.6<0.000196 (2.5)1.41.0–1.80.02
    Personality disorders54 (1.2)2.31.5–3.6<0.0001158 (3.4)1.71.3–2.1<0.0001
Women
    Psychiatric contactc51 (0.9)8.32.5–27.30.00197 (1.8)1.81.1–2.90.03
 Specific psychiatric diagnosesd
    Schizophrenia spectrum disorders20 (1.8)15.54.5–54.0<0.000118 (1.7)1.50.8–2.80.3
    Alcohol use disorders29 (1.0)8.22.4–28.70.00170 (2.5)2.21.3–3.70.004
    Drug use disorders18 (1.0)8.32.3–29.40.00162 (3.6)3.11.8–5.4<0.0001
    Affective disorders25 (1.2)10.73.1–37.0<0.000136 (1.7)1.60.9–2.90.09
    Personality disorders31 (1.4)11.93.5–40.7<0.000146 (2.1)1.81.0–3.10.04
  • aN of total number of individuals in each category dying by specific cause of death during follow-up.

  • bSeparate models were conducted and adjusted for age at first homeless shelter contact listed in the Homeless Register, country of origin and main source of income (defined as main source of income at last contact in the Homeless Register).

  • cThe HR of homeless persons with a history of psychiatric contact (defined as any diagnosis of a psychiatric disorder) was calculated relative to homeless persons with no records of previous psychiatric contact (reference category).

  • dThe HR of homeless persons with a history of specific psychiatric disorders was calculated relative to homeless persons with no records of previous psychiatric contact (reference category). The diagnosis-specific models are additionally adjusted for a history of other psychiatric disorders.

A history of a substance use disorder with either alcohol or opioids was a significant predictor of suicide for men (table 3). For both men and women, any substance use disorder (cannabis was only borderline significant in men, P = 0.10) was associated with elevated risks of dying by unintentional injury compared with those with no history of a substance use disorder.

View this table:
Table 3

Multiple Cox regression analyses of specific types of substance use disorders and the outcomes, death by suicide and unintentional injury, among homeless men (N = 22 508) and women (N = 9439) during 1999–2008

VariablesSuicideUnintentional injury
N (%)aHRb95% CIPN (%)aHRb95% CIP
Men
    Alcohol112 (1.3)1.71.2–2.40.002273 (3.1)1.61.3–2.0<0.0001
    Opioids25 (1.5)2.01.2–3.40.00598 (5.8)2.62.0–3.4<0.0001
    Cannabis23 (1.0)1.30.8–2.20.370 (3.0)1.31.0–1.80.1
    Sedatives12 (1.1)1.50.8–2.90.251 (4.9)2.31.6–3.2<0.0001
    Cocaine3 (0.7)1.00.3–3.41.019 (4.7)2.21.3–3.60.002
    Other drugs50 (1.3)1.91.2–2.90.003191 (5.1)2.62.0–3.3<0.0001
Women
    Alcohol29 (1.0)1.30.6–2.50.570 (2.5)3.11.9–5.1<0.0001
    Opioids4 (0.7)0.80.2–2.40.620 (3.6)4.12.2–7.7<0.0001
    Cannabis5 (1.0)0.90.3–2.60.813 (2.6)3.51.7–7.20.001
    Sedatives8 (1.3)1.30.5–3.30.628 (4.6)5.93.2–10.9<0.0001
    Cocaine0 (0.0)4 (3.9)6.32.1–18.90.001
    Other drugs12 (1.1)1.10.5–2.50.846 (4.3)5.83.4–9.9<0.0001
  • aN of total number of individuals in each category dying by specific cause during follow-up.

  • bSeparate adjusted models examine the HR of homeless people with a history of specific types of substance use disorders compared with those not registered with any substance use disorder (reference category). Models are adjusted for age at first homeless shelter contact listed in the Homeless Register, country of origin, main source of income (defined as main source of income at last contact in the Homeless Register), a history of abuse of all other types of drugs, schizophrenia spectrum disorders, affective disorders and personality disorders.

We also found that psychiatric inpatient contact during the 5 years prior to first being registered in the Homeless Register was linked to an increased risk of dying by suicide and unintentional injury when compared with individuals with no psychiatric contact in this period. Women with emergency room contact as the only type of psychiatric contact were at increased risk of dying by both suicide and unintentional injury compared with those with no psychiatric contact. For men, psychiatric emergency room contact was only a predictor for death by unintentional injury; not suicide. However, men with psychiatric outpatient contact also had increased risk of death by unintentional injury and of the same range as for emergency room contact (table 4). In comparison to people with no psychiatric contact, men and women who had spent either 3–10 days or more than 10 days as a psychiatric inpatient had elevated risk of dying from suicide and unintentional injury. Women were also at increased risk of dying by suicide (only borderline significant) and unintentional injury if they had spent just 1–2 days as a psychiatric inpatient.

View this table:
Table 4

Age-adjusted Cox regression analyses of psychiatric contact type and the outcomes, death by suicide and unintentional injury, among homeless men (N = 22 508) and women (N = 9439) during 1999–2008a

VariablesSuicideUnintentional injury
N (%)bHRc95% CIN (%)bHRc95% CI
Men
 Types of psychiatric contactsd,e
        No contact91 (0.7)1.0267 (1.9)1.0
        Ever inpatient87 (1.6)2.51.8–3.3195 (3.6)1.91.6–2.3
        Outpatient12 (0.8)1.30.7–2.341 (2.7)1.51.1–2.0
        Emergency room only16 (0.9)1.50.9–2.647 (2.8)1.61.1–2.1
 Psychiatric inpatient daysd
        0119 (0.7)1.0355 (2.1)1.0
        1–28 (1.1)1.50.7–3.114 (1.9)0.90.5–1.5
        3–1029 (1.8)2.61.7–3.953 (3.3)1.61.2–2.1
        >1050 (1.6)2.31.7–3.2128 (4.1)2.01.6–2.4
Women
 Types of psychiatric contactsd,e
        No contact10 (0.2)1.039 (0.7)1.0
        Ever inpatient33 (1.4)8.34.0–17.060 (2.6)3.62.4–5.4
        Outpatient4 (0.5)3.11.0–10.08 (1.0)1.70.8–3.6
        Emergency room only7 (1.1)6.52.5–17.310 (1.5)2.21.1–4.5
 Psychiatric inpatient daysd
        021 (0.3)1.057 (0.8)1.0
        1–23 (1.0)3.41.0–11.68 (2.7)3.31.6–7.0
        3–105 (0.9)2.81.1–7.612 (2.1)2.41.3–4.5
        >1025 (1.8)5.83.2–10.540 (2.8)3.22.1–4.8
  • aThe overall P-value was significant for all variables displayed in the table (P = 0.0001).

  • bN of total number of individuals in each category dying by specific cause during follow-up.

  • cSeparate adjusted models examine the HR of homeless people with different types of psychiatric contact and number of psychiatric inpatient days compared with those with no psychiatric contact (reference category). Models are adjusted for age at first homeless shelter contact listed in the Homeless Register.

  • dMeasured during the 5 years before first contact in the Homeless Register.

  • eThis variable was constructed using a hierarchical order: ‘ever inpatient’ includes all people with at least one psychiatric inpatient contact, ‘outpatient’ includes people with at least one psychiatric outpatient contact and no inpatient contact and ‘emergency room only’ includes all people with psychiatric emergency room contact but no psychiatric inpatient or outpatient contact.

Discussion

To our knowledge, this nationwide study is the first to examine predictors of deaths by suicide and unintentional injury among homeless men and women. The results revealed elevated MRs for suicide and unintentional injury in the Danish homeless shelter population, especially for men. Strong associations were identified between suicide and schizophrenia spectrum disorders, alcohol use disorders, drug use disorders, affective disorders and personality disorders. For suicide, the risk estimates were particularly elevated among the women with psychiatric disorders compared with those with no psychiatric contact. For both genders, the highest risk estimates were found for schizophrenia spectrum disorders. With regard to deaths from unintentional injury, substance use disorders were predictors for both men and women, whereas schizophrenia spectrum disorders and personality disorders were only significant predictors among men. For unintentional injury mortality, the highest risk estimates were found for drug use disorders. A history of psychiatric inpatient admission and psychiatric emergency room contact were also important predictors for dying by suicide and unintentional injury in the homeless population.

The elevated suicide and unintentional injury mortality levels confirm previous findings among homeless people.6,19 Our MRs were higher for unintentional injury than for suicide. This suggests that homeless people are an important high-risk group, especially with regard to unintentional injury. The strong association between substance use disorders and risk of dying by unintentional injury suggests that these disorders explain the excess deaths in this group, which is supported by previous studies.1,35,20 Thus, increased efforts to reduce substance use disorders in the homeless shelter population would likely influence the unintentional injury mortality.

We found strong associations between psychiatric disorders and suicide in accordance with findings for the general population in Denmark.22,30 It is well known that substance use disorders alone and in combination with other psychiatric disorders significantly increase the risk of suicide.22 We did not find significant differences between the risk estimates of different psychiatric disorders and did not examine the risk associated with comorbid substance abuse. However, one might assume that homeless people with, for example, both an affective disorder and a substance use disorder have higher risk of suicide than those with only one psychiatric disorder. In the general population, the suicide risks linked to psychiatric admission were much higher than our estimates for the homeless people.31 Different measures in the two analyses are likely to explain differences in risk estimates in the two populations. However, it could be that a larger proportion of homeless people with no record of a psychiatric contact have somatic disorders and/or untreated psychiatric disorders than of the general population.

The difference in risk of dying by suicide between those with and without psychiatric contacts was greater for women than men in the homeless population. This is in line with findings for the general population.31,32 It is possible that homeless women with psychiatric disorders are more inclined to seek treatment for their problems than men. Thus, the explanation of the gender difference can be that the reference group of homeless men, i.e. those with no psychiatric contact, may have a higher proportion of untreated psychiatric disorders compared with the corresponding group among the women. Another possible explanation is that homeless women have a worse mental health status than the homeless men.8

Homeless women with psychiatric emergency room contact had a higher risk of dying by suicide and unintentional injury compared with those with no psychiatric contact. This could reflect that these women had mental health problems but were not really engaged in treatment. In addition, an elevated risk of death by unintentional injury was found for homeless men with psychiatric emergency room contact. Our findings suggest that homeless people with a psychiatric admission or emergency room contact constitute important high-risk groups. Homeless patients are more likely than their non-homeless counterparts to use these types of psychiatric hospitalization and less likely to use psychiatric outpatient services.15 It has been suggested that approaches to socio-medical integration in the emergency departments are important to promote the health among homeless people.33 Also, high rates of suicide ideation, deliberate self-harm and suicide attempts have been found in homeless people in contact with the mental healthcare services.34,35

Individuals with more than one stay in the homeless shelters within the first year of contact had elevated risks of dying by unintentional injury compared with only one stay. People with multiple homeless shelter contacts might represent a group living in a permanent homelessness situation that is likely to be associated with social deprivation, substance use problems and a stressful and unsecure lifestyle.36,37 Thus, this group is probably more vulnerable with respect to poisoning, traffic accidents and falls compared with those with only one homeless shelter stay. However, it is possible that people with only one stay during the first year in homelessness are actually permanently homeless.

Our study has several strengths. First, the Homeless Register provides access to a relatively unselected sample of the homeless shelter population. In addition, it is a strength that we were able to examine homeless people in almost all ages including both young people and adults. However, as previous studies primarily have studied selected groups of homeless people, our results might be difficult to compare directly with those. Using individual-level, nationwide and prospectively collected register data, we were able to investigate specific external causes of death for both genders and calculate risk estimates with great power. Additionally, we had access to complete information on psychiatric hospitalization and death.

This study also has limitations. The study population did not include homeless people with no contact to the homeless shelters. These people are likely to have higher morbidity and less likely to use health services than the sheltered homeless people.38 Furthermore, we had no information on untreated psychiatric disorders or outpatient treatment for substance use disorders. Thus, underestimation of morbidity and mortality is expected. Information on housing status during follow-up was not taken into account, which means some individuals probably ceased to be homeless during the study. This may lead to conservative estimates. Additionally, the design of this study did not allow for direct comparisons with the general population. Future studies should aim to include a relevant control group, e.g. the general population or non-homeless people with psychiatric disorders. Another limitation is that some suicides and unintentional deaths might have been recorded as events of undetermined intent, which were not included in this study. Particularly in homeless people with high prevalence of comorbidity, determining the exact cause of death can be difficult. Additionally, we were unable to control for some potential confounders, e.g. criminality,39 childhood maltreatment40 and physical diseases.9 Furthermore, due to limited periods of coverage by registers, we cannot establish whether homelessness preceded first psychiatric contact or vice versa. Finally, in some of the diagnosis categories or homeless shelter contact groups, the risk estimates were based on only a few cases. This may reduce the strength of the associations.

Our findings can probably be generalized to homeless people from 16 years of age living under similar conditions in other welfare countries with a social safety net. Homeless people are a vulnerable high-risk group due to their elevated suicide and unintentional injury mortality levels. The findings have clinical and political implications showing strong associations between psychiatric disorders and suicide and unintentional injury mortality in the homeless shelter population. Especially, the women with psychiatric disorders seem to be a vulnerable group that deserve special attention with regard to suicide prevention. Additionally, it should be noticed that people with more than one stay in the homeless shelters, and especially men, constitute a high risk group with regard to death by unintentional injury. This study underlines the need to focus more on people in the homeless shelter population with psychiatric disorders when aiming to reduce the elevated levels of suicide and unintentional injury mortality.

Funding

This work was supported by the Danish Council for Independent Research (10-080986) and the Faculty of Health and Medical Sciences, University of Copenhagen.

Conflicts of interest: None declared.

Key points

  • This large follow-up study confirms that homeless people have high-mortality levels due to suicide and unintentional injury.

  • A history of schizophrenia spectrum disorders, alcohol use disorders, drug use disorders, affective disorders and personality disorders is strongly associated with increased risk of suicide for homeless men and women.

  • Alcohol and drug use disorders are predictors of death by unintentional injury for both homeless men and women, whereas schizophrenia spectrum disorders and personality disorders are only significant predictors among men.

  • A history of psychiatric admission and psychiatric emergency room contact are important predictors of death by suicide and unintentional injury in the homeless population.

  • Homeless people, especially men, with multiple contacts to the homeless shelters constitute a high-risk group with regard to death by unintentional injury.

References

View Abstract