Roy É, Haley N, Leclerc P, Sochanski B, Boudreau J, Boivin J. Mortality in a Cohort of Street Youth in Montreal. JAMA. 2004;292(5):569-574. doi:10.1001/jama.292.5.569
Author Affiliations: Direction de Santé Publique de Montréal (Drs Roy and Haley, Mss Leclerc and Sochanski, and Mr Boudreau), and Joint Department of Epidemiology, Biostatistics, and Occupational Health (Drs Roy and Boivin) and Department of Family Medicine (Dr Haley), McGill University, Montreal, Quebec.
Context Many studies have shown a high prevalence of sexually transmitted diseases,
human immunodeficiency virus (HIV) infection, viral hepatitis, drug dependence,
and mental health problems among street youth. However, data on mortality
among these youth are sparse.
Objectives To estimate mortality rate among street youth in Montreal and to identify
causes of death and factors increasing the risk of death.
Design, Setting, and Population From January 1995 to September 2000, 1013 street youth 14 to 25 years
of age were recruited in a prospective cohort with semi-annual follow-ups.
Original study objectives were to determine the incidence and risk factors
for HIV infection in that population; however, several participants died during
the first months of follow-up, prompting investigators to add mortality to
the study objectives. Mortality data were obtained from the coroner's office
and the Institut de la Statistique du Québec.
Main Outcome Measures Mortality rate among participants and factors increasing the risk of
Results Twenty-six youth died during follow-up for a mortality rate of 921 per
100 000 person-years (95% confidence interval [CI], 602-1350); this represented
a standardized mortality ratio of 11.4. The observed causes of death were
as follows: suicide (13), overdose (8), unintentional injury (2), fulminant
hepatitis A (1), heart disease (1); 1 was unidentified. In multivariate Cox
regression analyses, HIV infection (adjusted hazard ratio [AHR] = 5.6; 95%
CI, 1.9-16.8), daily alcohol use in the last month (AHR = 3.2; 95% CI, 1.3-7.7),
homelessness in the last 6 months (AHR = 3.0; 95% CI, 1.1-7.6), drug injection
in the last 6 months (AHR = 2.7; 95% CI, 1.2-6.2), and male sex (AHR = 2.6;
95% CI, 0.9-7.7) were identified as independent predictors of mortality.
Conclusions Current heavy substance use and homelessness were factors associated
with death among street youth. HIV infection was also identified as an important
predictor of mortality; however, its role remains to be clarified. These findings
should be taken into account when developing interventions to prevent mortality
among street youth.
The characteristics of street youth vary depending on the social environment
where they live. In Canada and the United States, street youth are generally
25 years or younger, and approximately a third of them are girls. They are
homeless or, more frequently, they live in highly unstable residential conditions.
Different subcultures are represented among them, including punks, rappers,
skinheads, and others.1,2 Despite
some heterogeneity, these youth share many characteristics that jeopardize
their development and health; they are highly entrenched in the streets and
frequently engage in high-risk behaviors such as prostitution3 and
substance abuse, including injection drug use.4- 10 They
are increasingly recognized as a population at risk for a wide range of physical
and mental health problems. Many studies have examined their precarious living
conditions and risky sexual and substance use behaviors and the impact of
these factors on their health. These studies have brought to light their extreme
vulnerability in terms of sexually transmitted diseases, human immunodeficiency
virus (HIV) infection, viral hepatitis, drug dependence, and mental health
problems.6,7,9,11- 18 Other
studies have looked at the sexual abuse and physical violence that street
youth endure.19- 24 Data
on the ultimate consequence of these problems, the death of street youth,
Several studies have addressed mortality among homeless individuals.25- 33 Of
these, only 5 presented data for youth younger than either 25 or 30 years.
A study of residents in New York City homeless shelters showed a mortality
ratio of 2.7 for men 20 to 24 years of age.33 In
Toronto, Ontario, a ratio of 8.3 was measured among men 18 to 24 years old
using homeless shelters.27 In Boston, Mass,
the mortality ratio for males 18 to 24 years old having been in contact with
the Boston Health Care for the Homeless Program was 5.9; among females in
the same age category, this ratio was 11.8.29 In
Copenhagen, Denmark, a mortality ratio of 28.5 was observed among female users
of hostels for homeless people who were 15 to 24 years of age; for male users
of that age group, the mortality ratio was 13.3.25 Finally,
a mortality ratio of 37.3 was calculated for male rough sleepers (individuals
with no fixed address) in London who were 16 to 29 years of age.32 All
these studies clearly showed that homeless youth experienced mortality rates
much higher than their counterparts in the general population, although the
ratio varied greatly from one study site to another.
Two of the reviewed studies presented causes of death by age group.
In Boston, the principal cause of death among males 18 to 24 years old was
homicide, followed by "poisoning and injuries other than motor vehicle accident."
Among females, also 18 to 24 years old, the 2 main causes of death, of equal
importance, were homicide and motor vehicle accident.29 The
leading identified cause of death among Toronto male shelter users 18 to 24
years of age was accident (other than being struck by a motor vehicle and
poisoning), followed equally by unintentional overdoses and suicide.27
These studies provided valuable information on mortality among young
homeless individuals. However, 3 of them were based on data extracted from
administrative databases and included little or no data on subjects.27,29,32 Even though this
allowed for the identification of causes of death, it precluded the analysis
of factors associated with death. Risk factor data were obtained in 2 studies.25,33 However, since the design of these
studies did not include any follow-up over time, analyses were based only
on baseline data. In addition, no analyses restricted to younger participants
In 1995, we initiated a prospective cohort study to determine the incidence
and risk factors for HIV infection among street youth in Montreal, Quebec.
In the first months following study inception, several participants died,
leading us to add the analysis of mortality to the study objectives.34 Our additional study objectives, in the context of
a longitudinal study design, were as follows: (1) to estimate the mortality
rate among street youth, (2) to determine the causes of death of these youth,
and (3) to identify factors that increase their risk of death.
The complete methodology of the study was described previously.18 Briefly, criteria for entry in the cohort were as
follows: (1) in the last year, having either regularly used the services of
community-based street youth agencies or been without a place to sleep more
than once; (2) being 14 to 25 years of age; (3) speaking English or French;
and (4) being able to provide informed consent and to complete an interviewer-administered
questionnaire. The street youth agencies operated drop-in centers, shelters,
and outreach vans; the range of services offered included food, short-term
housing, social services, and prevention interventions such as hepatitis B
vaccination and needle exchange.
Participants were recruited from January 24, 1995, to September 30,
2000. Study interviewers enrolled them through regular visits to all major
street youth agencies in Montreal. Participants were interviewed twice a year.
Detailed contact information was collected at each interview. Visiting community
organizations, leaving messages with friends and family, sending letters,
and contacting other organizations (such as Social Security, drug treatment
centers, probation offices, prisons, and youth rehabilitation centers) were
also used to follow up participants. At each visit, after giving informed
consent, they completed a 45-minute questionnaire, covering sociodemographic
characteristics, alcohol and drug use, and sexual behaviors, and provided
samples of gingival exudate for HIV antibody testing. Each visit was financially
compensated (Can $20). Ethical approval was provided by the institutional
review board of the Faculty of Medicine, McGill University, Montreal, Quebec.
Participants were eligible for follow-up until they reached 30 years
of age or until they reported, in 4 consecutive study questionnaires, not
using services from street youth organizations and not being homeless (ie,
sleeping outside or in a shelter or staying with friends out of necessity)
in Montreal; we thereafter refer to the latter participants as "no longer
Vital status of participants was verified throughout the study, whenever
we had information that a participant might have died. These ongoing verifications
were made with the coroner's office. For each confirmed death, a copy of the
coroner's report was obtained. Furthermore, at the end of the study, the vital
status of all participants was ascertained with the Institut de la Statistique
du Québec for the years 1995 through 2001; permission for access was
granted by the Commission d'Accès à l'Information (Quebec's
access to information commission).
The primary source of information on causes of death was the coroner's
reports. Whenever an autopsy was performed, a summary of the results was available
in the coroner's report. For cases not investigated by the coroner (n = 3),
we used the coded causes mentioned on the death certificate and provided by
the Institut de la Statistique du Québec. These causes were coded using
the International Classification of Diseases, Ninth Revision
(ICD-9), for deaths occurring in 1995 to 1999, or the International Classification of Diseases, 10th Revision (ICD-10), for
those occurring in 2000 and 2001. One participant died outside Canada. We
were informed of the death, and of its cause, by the youth's mother.
The follow-up period for every subject started at recruitment and ended
at the first of the following events: (1) death, (2) age 30 years, (3) being
no longer street-involved, or (4) 6 months after his/her last questionnaire.
Mortality rates were calculated overall and by subgroups defined according
to various characteristics. These rates were estimated using the person-time
method (number of deaths divided by person-years of follow-up); 95% confidence
intervals (CIs) were calculated using the Poisson distribution.
Standardized mortality ratios were calculated using the indirect method
of standardization by sex and age group; the comparison group was the general
population of the province of Quebec for 1996. Ninety-five percent CIs were
based on Byar's approximation.35
Predictors of mortality were identified using univariate and multivariate
Cox hazard regression analyses. Hazard ratios for potential predictors and
the corresponding 95% CIs were determined using univariate Cox regression.
All variables with P values ≤.10 in univariate
analyses were included in a multivariate Cox model using a backward procedure
(with likelihood ratio statistics). Those with P values
≤.05 were retained in the final model. The potential confounding effect
of variables excluded by the backward procedure was tested by adding them,
one at a time, in the final model. Finally, 2 × 2 interactions between
independent predictors were tested.
In the Cox analyses and for estimation of the mortality rates, all independent
variables other than sex were treated as time-dependent, either irreversible
or transient. Time-dependent irreversible predictors were measured at each
interview, but their value could change only once, from absence of the factor
to presence of it. These predictors included being older than 18 years, HIV
infection (based on the test performed at each questionnaire), homosexual
activities (with regular or casual partners), and sexual abuse.
For time-dependent transient predictors, exposure was measured at each
interview and their value could vary from questionnaire to questionnaire.
Predictors corresponding to the preceding 6 months were homelessness, drug
injection, and survival sex (defined as the exchange of sex for money, drugs,
or something else). Predictors assessed for the preceding month were daily
alcohol use and use of more than 2 categories of drugs (9 categories were
considered: marijuana, cocaine/crack, heroin, speedball [cocaine and heroin
combined], amphetamines, hallucinogens [mushrooms, LSD (lysergic acid diethylamide),
PCP (phencyclidine)], solvents, medications used for nonmedical reasons, and
other drugs). Statistical analyses were performed with SPSS for Windows (release
From January 24, 1995, to September 30, 2000, 1013 youth were recruited
in the cohort. Approximately 12% of offers to participate were refused. Participants
completed from 1 to 11 questionnaires (average of 5.3 questionnaires per participant;
87.2% completed at least 1 follow-up questionnaire after their recruitment
questionnaire). During follow-up, 145 participants reached 30 years of age
or were no longer street-involved. Overall, participants cumulated 2822 person-years
of follow-up, for an average follow-up of 33.4 months per participant.
At study entry, the mean age of participants was 19.9 years. As indicated
in Table 1, two thirds were boys,
most were born in Canada, 80% had been homeless in the 6 months prior to study
entry, and a quarter had ever been involved in survival sex. Substance use
was high among them with the majority of youth having ever used cannabis,
hallucinogens, and cocaine or crack at entry. Almost half of the participants
had ever injected drugs. Fourteen participants were HIV-infected at study
entry, for a prevalence of 1.4%; 16 incident HIV infections were observed
during the study period.36
Twenty-six of the 1013 participants died during follow-up. Three additional
deaths occurred outside the follow-up period: 1 participant died 30 months
after his last interview (he had refused to continue participation) and 2
died 16 months after their last interview (1 had been excluded from the cohort
since he was no longer street-involved, and 1 had been lost to follow-up).
These 3 deaths were excluded from all subsequent analyses.
Causes of death are listed in Table
2. Suicide and drug overdoses were the 2 leading causes of death.
Among male participants, suicide was the main cause of death; among female
participants, the main cause of death was drug overdose. Five youth died before
reaching 20 years of age: 3 of a drug overdose, 1 of unintentional injury,
and 1 of suicide. Of the 13 participants who committed suicide, 9 died by
hanging, 2 jumped from a bridge, and 2 jumped or ran in front of a moving
vehicle. One cause of death was unidentified. All overdose deaths involved
illicit drugs, and none of them were classified as intentional by the coroner.
The 26 deaths observed during follow-up represented a mortality rate
of 921 per 100 000 person-years (95% CI, 602-1350). Among male participants,
the mortality rate was 1148 (95% CI, 720-1739) and among female participants,
442 (95% CI, 120-1131). The standardized mortality ratio was 11.4 (95% CI,
7.4-16.7); the estimates were 11.1 for males (95% CI, 6.9-16.8) and 13.5 for
females (95% CI, 3.6-34.5).
Mortality rates by participant characteristic and results of the univariate
and multivariate Cox regression analyses are presented in Table 3. The independent predictors of mortality identified in the
final model were HIV infection, daily alcohol use (last month), homelessness
(last 6 months), drug injection (last 6 months), and male sex. None of the
variables excluded by the backward procedure had a confounding effect, and
no interactions were detected between variables retained in the final model.
To our knowledge, this study represents the first prospective cohort
study on mortality among street youth. It includes longitudinal follow-up
and analyses based on time-dependent factors. The mortality rate of 921 per
100 000 person-years observed in our study participants is extremely
high, exceeding 11 times the rate observed among youth in the general population.
Similar ratios were reported in young homeless in Toronto and Boston,27,29 while higher ratios were reported
in Denmark and England25,32 and
lower ones in New York City.33
The 2 main causes of death in our study were suicide and drug overdose.
It is plausible that some drug overdoses were in fact suicides. Several studies
have shown that overdose is often the method chosen by drug users to commit
suicide, and that nonintentional and intentional overdoses cannot always be
differentiated.37,38 The overlap
between suicide and overdose may be especially important for females. Male
street youth, as those in the general population, are probably more inclined
to choose irreversible means, such as hanging, to kill themselves; these deaths
can easily be identified as suicide. However, female street youth, as females
in the general population, might tend to choose less violent methods, such
as overdoses; therefore, these deaths are less likely to be recognized as
The causes of death for males in our study were similar to those observed
by Hwang in Toronto; the 3 main causes were the same even though the order
differed.27 The picture is, however, very different
from what was seen in Boston, where the leading cause of death for both males
and females was homicide.29 For females in
Boston, motor vehicle accident was another leading cause of death; in our
study, no female participants died from such a cause. The difference between
our results and those of the Boston study regarding death by homicide is not
surprising given that the overall homicide rate is 3.8 times higher in the
United States than in Canada.40
The leading cause of death among Montreal male street youth is similar
to the leading cause in the general population.41,42 In
2000, the leading cause of death among male Quebec residents 15 to 19 years
old and 20 to 24 years old was suicide (38.2% and 36.2% of all deaths in these
groups), followed by motor vehicle accident (32.6% and 29.2%). However, among
female Quebec residents in the same age groups, the leading cause of death
was motor vehicle accident (45.5% and 31.0%) followed by suicide (15.6% and
Being HIV-infected was the strongest independent predictor of death
during follow-up. This association might be due to the fact that HIV infection
was a direct cause of death. Four of the deceased participants were HIV-infected,
including 1 youth who died from hepatitis A and who was also infected with
hepatitis C virus. This death was clearly related to his HIV status. For the
3 other cases, the relationship is less clear: 2 died of a drug overdose and
1 of a noninfectious heart condition.
Another important predictor of death was being recently homeless. Hwang43 noted that the literature is not clear about the
strength of the association between homelessness and mortality. Part of the
problem resides in the difficulty of assessing exposure to homelessness. Shelter
use is often equated with homelessness, which can lead to misclassification
because periods of life on the street are then not considered. In our study,
both types of episodes were included since homelessness was defined as sleeping
outside or in a shelter or staying with friends out of necessity. In addition,
treating homelessness as a time-dependent variable resulted in a more valid
estimation of the association between homelessness and death. Therefore, we
can conclude that the risk of death does increase during episodes of homelessness.
Given the importance of suicide as a direct cause of death, it is possible
that the causal effect of homelessness on mortality is mediated by other factors
such as despair during episodes of homelessness.
Two other independent predictors of mortality among street youth were
daily alcohol use and drug injection. The association between substance misuse
and death has been reported by several authors.25,30,31,33 This
finding is coherent with the causes of death that we observed. Eight of the
26 deaths were due to drug overdose. In addition, at least 1 other death was
drug-related (the hepatitis A death), not to mention the cases of suicide
and unintentional injury that occurred when the youth were intoxicated (6
of the 8 for whom toxicological analyses were conducted). Substance use, which
is very frequent among street youth, without doubt carries a high risk of
mortality and is responsible for numerous deaths in this young population.
Another identified predictor of mortality was sex. As found in the general
population, male sex is associated with an increased risk of death among street
youth. However, other authors found that this survival advantage of females
did not necessarily hold among young homeless individuals.28 The
question of the difference between street-involved men and women regarding
death remains open.
Potential limitations must be taken into account when interpreting our
results. First, possible misclassification of exposures is always a concern.
In our analysis, we chose to end follow-up for a given participant no later
than 6 months after his/her last completed questionnaire. Because of that,
we consider that potential misclassification was significantly reduced; participants
are less likely to have changed exposure category within 6-month intervals.
However, the potential misclassification bias related to self-reported behaviors
remains. Some youth may have been reluctant to report behaviors that were
less socially acceptable. We tried to reduce this to a minimum by giving repeated
assurances of confidentiality. Second, we may have underestimated the mortality
rate due to deaths occurring outside the province of Quebec. However, we think
that the number of such deaths should be very small. Even though these deaths
would not have been detected by official sources, we would generally have
been informed of them by close family members or friends when tracing participants
for their follow-up interviews, as indeed happened for 1 participant.
Given the large diversity of our recruitment sites, we are confident
that our findings are generalizable to the population of street youth using
street youth services in Montreal. However, these results may not be applicable
to those not accessing services; these youth may be further marginalized and
at greater risk of death, or, conversely, represent relatively well-organized
youth, and at lower risk. Regarding the generalizability of our results to
street youth in other Canadian and US cities, the significant similarities
in their situations should make our findings applicable to street youth in
most major urban centers in these countries. Nevertheless, they may not be
applicable to street youth from developing countries who are facing a totally
different social environment.
In conclusion, mortality is high in the street youth population. Treatment
of addiction and mental health problems should represent public health priorities
to prevent deaths in these young people. Additional studies providing a better
understanding of the role of other factors such as HIV infection and homelessness
are also needed to support the development of appropriate health and social
services for these vulnerable youth.