Chronic Diseases in Canada Volume 28, Number 4, 2008

Chronic Diseases in Canada Volume 28, Number 4, 2008
Chronic Diseases
Volume 28, Number 4, 2008
in Canada
Chronic Diseases in Canada
a publication of the
Public Health Agency of Canada
Sylvie Stachenko
Principal Scientific Editor
(613) 946-3537
Robert A Spasoff
Associate Scientific
Risk factors for falling among community-dwelling seniors
using home-care services: An extended hazards model
with time-dependent covariates and multiple events
Howard Morrison
Deputy Scientific Editor
(613) 941-1286
Claire Infante-Rivard
Associate Scientific
Michelle Tracy
Managing Editor
(613) 954-0697
Elizabeth Kristjansson
Associate Scientific
Gender and the smoking behaviour of Ethiopian
immigrants in Toronto
Jacques Brisson
BS Leclerc, C Bégin, É Cadieux, L Goulet, N Leduc, M-J Kergoat,
and P Lebel
I Hyman, H Fenta, and S Noh
Ethnicity and mental health: Conceptualization,
definition and operationalization of ethnicity from a
Canadian context
DE Clarke, A Colantonio, AE Rhodes, and M Escobar
Association of comorbid mood disorders and chronic
illness with disability and quality of life in Ontario,
T Gadalla
Costs associated with mood and anxiety disorders, as
evaluated by telephone survey
SB Patten, JVA Williams, and C Mitton
Information for Authors: Special Call for Papers
CDIC Editorial Board
C Ineke Neutel
Neil E Collishaw
University of Ottawa
Institute on Care of
the Elderly
Physicians for a
Smoke-Free Canada
Kathryn Wilkins
James A Hanley
Health Statistics Division
Statistics Canada
Université Laval
McGill University
Clyde Hertzman
University of British
Chronic Diseases in Canada (CDIC) is a quarterly scientific
journal focussing on current evidence relevant to the
control and prevention of chronic (i.e. non-communicable)
diseases and injuries in Canada. Since 1980 the journal
has published a unique blend of peer-reviewed feature
articles by authors from the public and private sectors
and which may include research from such fields as
epidemiology, public/community health, biostatistics, the
behavioural sciences, and health services or economics.
Only feature articles are peer reviewed. Authors retain
responsibility for the content of their articles; the opinions
expressed are not necessarily those of the CDIC editorial
committee nor of the Public Health Agency of Canada.
Subscription is free upon request.
Chronic Diseases in Canada
Public Health Agency of Canada
130 Colonnade Road
Address Locator 6501G
Ottawa, Ontario K1A 0K9
Fax: (613) 941-3605
E-mail: [email protected]
Indexed in Index Medicus/MEDLINE
To promote and protect the health of Canadians through leadership, partnership,
innovation and action in public health.
— Public Health Agency of Canada
Published by authority of the Minister of Health.
© Her Majesty the Queen in Right of Canada, represented by the Minister of Health, 2008
ISSN 0228-8699
Également disponible en français sous le titre : Maladies chroniques au Canada
Risk factors for falling among community-dwelling
seniors using home-care services: An extended
hazards model with time-dependent covariates and
multiple events
BS Leclerc, MSc (1); C Bégin, MSc (1); É Cadieux, MSc (1); L Goulet, MD, PhD (2); N Leduc, PhD (2);
M-J Kergoat, MD (3); P Lebel, MD (3,4)
The identification of risk factors for falls in longitudinal studies becomes difficult because
of exposures that change during the follow-up and also because individual subjects may
experience an event more than once. These issues have been neglected and improper
statistical techniques have been used. The typical approaches have been to report the
proportion of fallers or the time to first fall. Both avoid the underlying assumption of
independence between events and discard pertinent data. We review the existing methods
and propose a Cox hazards extension. We exemplify it in the study of potential risk factors
associated with all falls in 959 seniors. Finally, we compare the results of the proposed
Wei, Lin, & Weissfeld (WLW) method with those of several other techniques. Stable
exposure variables measured at baseline and updated time-varying exposures include
socio-demographic characteristics, BMI, nutritional risk, alcohol consumption, home
hazards, gait and balance, and medications. Results demonstrate that the usual methods
of analyzing risk factors for falling are inappropriate, as they produce considerable
biases relative to the WLW model using time-dependent covariates. Results also show
that modeling for first events may be inefficient, given that the risk of occurrence varies
between falls.
Key words: Accidental falls, Cox model, elderly, environmental hazards, negative
binomial distribution, hazards model, regression analysis, survival
analysis, logistic models
Falls are common, recurrent problems with
serious consequences for elderly people
and the health care system.1 Evidence
of fall-risk factors has generally been
identified by prospective observational
designs. These studies may suffer from
problems similar to those found in cohort
studies of other issues, such as loss to
follow-up and variable follow-up time.
The identification of fall-risk factors
deals with additional problems such as
exposure changes during follow-up and
recurrent events in the same person.
These issues have been neglected and
inefficient statistical techniques have been
used. As a result, this may have distorted
the magnitude in estimates of particular
predictors or produced misleading results.
Moreover, this may have missed questions
of great clinical relevance.2-5
More than 15 years ago, Cumming, Kelsey,
and Nevitt6 advised that more attention be
paid to repeated measures regarding both
risk factors and rates for all falls. Despite
this, few researchers have challenged the
design of their studies and the analysis
of their data. Rather, they seem to have
been adversely affected, circumventing
the methodological complications by
discarding much relevant information.
The aim of the present paper is to raise
the awareness of researchers about some
epidemiological and statistical considerations. We review the statistical background
of methods of fall studies, introduce the
philosophical issues of time-dependent
covariates and multiple events, and discuss
the existing statistical techniques which
deal with them. We propose an extension
of the Cox proportional hazards traditional
model and use it in the identification of
potential risk factors associated with all
falls in elderly people living at home.
Finally, we compare the different results
obtained by various statistical methods.
Statistical background of
methods of fall studies
A variety of strategies has been used
to study the risk factors for recurrent
falls. Their analysis is complicated by
the within-subject correlation. In other
words, the occurrence of an event acts
on the risk of the next one. Failure to
account for dependence in the data leads
to the usual estimator of variance being
underestimated. This produces confidence
Author References
1 Direction de santé publique et d’évaluation, Agence de la santé et des services sociaux de Lanaudière, Joliette, Quebec
2 Groupe de recherche interdisciplinaire en santé, Université de Montréal, Montréal, Quebec
3 Centre de recherche, Institut universitaire de gériatrie de Montréal, Montréal, Quebec
4 Centre d’expertise sur la santé des personnes âgées et des aidants, Institut universitaire de gériatrie de Montréal, Montréal, Quebec
Correspondence: Bernard-Simon Leclerc, Service de surveillance, recherche et évaluation, Direction de santé publique et d’évaluation, Agence de la santé et des
services sociaux de Lanaudière, 245, rue du Curé-Majeau, Joliette, J6E 8S8, Tel: (450) 759-1157 extension 4324, Email: [email protected]
Chronic Diseases in Canada
Vol 28, No 4, 2008
intervals that are too narrow and a test of
significance too liberal (i.e. rejects the null
hypothesis too often).3,5,7
A summary of some of the discussed
methods is provided in Figure 1. A simplistic
approach to such problems involves
reporting the proportion of fallers (subjects
who fall at least once over an arbitrarily
defined period) or the time to a first fall.8
Both possibilities avoid the underlying
assumption of independent association
between multiple events. However, the use
of all available data for each individual
could be more efficient.4,8,9 The author of
a key paper has argued that the incidence
rate for falls was a public health priority6,
particularly for less robust elderly people.10
The challenge in analysing all falls arises
because some elderly are more prone to
recurrences than others; hence, they run a
higher risk of fall-related injury as opposed
to those who fall only once. The choice of
outcome, according to whether the focus
is on fallers or on the rate of falls, could
also affect the conclusion; i.e., knowing
whether a particular exposure constitutes a
risk factor. Stable over-time factors are more
likely to be related to the state of “being a
faller” than exposures that vary over time.6
participants, each of whom is observed over
three years and suffers three falls. One has
fallen once each year, another three times in
the first year, and the last three times in the
third year. The outcome variable ignores the
time of occurrence of these events.8 Thus, a
negative binomial modeling event rates may
not be the method of choice when the value
of important covariates or the likelihood of
event occurrence changes with the passing
of time.3 Greater efficiency and accuracy
can be obtained by modelling the lengths
of inter-episode intervals via time-to-event
techniques.9 Rather than focusing on the
numbers of cases, the time-to-event approach
considers the time between falls. If the
incidence rate is high, the intervals between
events will be short, and vice versa.3
In addition, measured risk factors of which
we want to evaluate the effects are usually
only fixed variables, defined at the initial
examination.2 They refer to the intrinsic
characteristics of the subjects (e.g., the
sex), the past exposures (e.g., prior
falls) or exposures present at baseline
(e.g., use of medication). Exposures that
occur after the starting point or vary over
time for an individual are not taken into
account. Examples, which can potentially
cause falls through short-term exposure
preceding the event, include environmental
hazards, alcohol consumption, and use of
medication. A great advantage of the timeto-event approach is its ability to handle
time-dependent covariates.3
Alternatives have been proposed for
dealing with multiple events. Among these
are the negative binomial regression, some
extensions to the Cox proportional hazards
model, and a modified logistic regression.
The dependent variable in the negative
binomial regression is the individual
event rate adjusted for the follow-up
time i.e., the number of falls for a person
divided by their specific follow-up time
(Figure 1).4,11 Since the negative binomial
distribution has one more parameter than
the Poisson, it naturally accommodates
for over-dispersion (i.e., the variance
typically exceeds the mean).8 Therefore,
this approach is robust for dependent
structure data, and suitable for frequent
and recurrent events.
The hazards models include the counting
process of Andersen & Gill16 (hereafter
referred to as AG), the conditional model of
Prentice, Williams, & Peterson17 (PWP), and
the marginal model of Wei, Lin, & Weissfeld18
(WLW). None of these approaches explicitly
models the dependence structure between
failure times. Instead, robust estimates of
variance are used to account for correlated
observations within subjects; i.e., the
so-called «variance-corrected» hazards
One problem using event rates is that the
likelihood of event occurrence must be
assumed to be constant through time within
one participant. A critical example could be to
consider the equivalent event rates for three
The distinction of the hazards methods can
be seen in terms of who is in the risk-set at
each failure.15,19 The AG rests on the strong
assumption that the risk of an event for a
given subject is unaffected by any earlier
Vol 28, No 4, 2008
events, unless a term that captures such
dependence (i.e., number of previous falls)
is included as a time-dependent covariate.3,7
In other words, the data for each subject
with multiple events could be described
as data for multiple subjects, where each
has delayed entry and is followed until the
next event (Figure 1). This model ignores
the order of the events; i.e., all falls are
indistinguishable, leaving each subject to
be “at risk” for an event as long as the
subject is still under observation at the
time the event occurs.3,7,8,13-15,19
The PWP is based on the idea that a subject
is not technically at risk for a later event until
all previous events have been experienced.
This is accomplished by stratifying data
by event order. Accordingly, the risk-set at
time t for the kth event is limited to those
subjects under study at t who have already
experienced k-1 events (not exemplified in
Figure 1).13-15,19 However, Robertson20 has
argued that the conditional assumption
of the order of events does not hold for
falls. As an illustration of her argument
(personal communication), let us speculate
that a person has slipped on water on
the kitchen floor without injury and, at
another time, has fallen on the pavement
outside. This has resulted in a hip fracture.
The person is at risk for both these falls
from the beginning of the study period;
i.e., the time at risk for the second fall on
the pavement does not start only after the
first fall in the kitchen has occurred.
The risk-set of the WLW marginal
approach includes all patients under
observation who have not yet experienced
the kth event. The time for each event
starts at the beginning of follow-up time
for each subject. Furthermore, each
subject is considered to be at risk for all
events, regardless of how many events
each subject has actually experienced. The
WLW does not impose any dependence
structure among the related failure times.
Thus, it ignores the ordering of events
but takes into account previous events by
situating each failure in an independent
stratum (Figure 1).4,7,8,13-15,19
The logistic regression analysis is the most
commonly used method in epidemiological
Chronic Diseases in Canada
research. D’Agostino et al.21 showed that
a so-called pooled logistic regression is
identical to the time-dependent covariate
Cox regression. This is what makes
the technique attractive to evaluate the
relationship of risk factors to disease
development. O’Loughlin22 applied such
an approach to the study of falls. The
theoretical basis for the use of this logistic
regression variant is well established when
the intervals between measurements of
risk factors are short, the probability
of an event within an interval small,
and the intercept for the pooled logistic
constant across intervals.21 The underlying
statistical requirements and the data setup
for the pooled logistic regression are very
close to those defined for the AG. Each of
the follow-up intervals for a single subject
is assumed to represent intervals from
different subjects. The method pools the
subjects at risk and the events developed
in each interval. The follow-up interview
number is included as a categorical
variable to test this assumption. Similarly,
the dependence between multiple falls
within the same individual is accounted for
by considering the occurrence of previous
falls as a predictor variable.23
However, the way in which the interval
observations are set up, as well as the
outcome variable of interest, differs in
both methods. The AG builds the intervals
according to the precise dates of events.
For example, the first interval will cover the
time span from entry into the study until
the time of the event, and the following
interval spans the time from the first event
to the next one, and so on (Figure 1).15
In contrast, the logistic regression uses
stable time points fixed by the researcher.
For example, an exam could be performed
at the same date each month to up-date
risk factors and to gather information
on falls that occurred in the interval of
observation (not exemplified in Figure
1).22,23 The analysis above is, in essence,
an investigation of fallers versus nonfallers in successive short intervals.22 Even
if, taken as a whole, the analysis allows
for more than one outcome to occur per
subject, less appreciated is the fact that it
Chronic Diseases in Canada
Schematic representation of statistical models for the study of risk factors for falls
(Modified from a figure published by Robertson, Campbell and Herbison8)
Hypothetical subject with follow-up of 12 days, falls on day 5 and 8. Let (0) represent no
fall and (1) a fall, xi a risk factor of subject i measured at time t, and ki its number of falls.
Then the baseline hazard is illustrated by λ0(t), the hazard for a fall for the ith subject by
λi and the hazard of the kth fall for this subject by λik. Person-time, pti is length of time at risk
for subject i, β’x denotes the effect size of factor x, p is the probability of event in exposed,
e, and unexposed, u, subjects.
Standard Cox regression. One data record covers entry until the 1st fall and discards any
information past that point. Total follow-up time is assigned to individual that never fell.
The dependent variable is time to first fall.
Andersen-Gill Cox regression. Three records cover entry until the 1st fall, from the 1st
until 2nd fall, from the last fall to the end of follow-up, the latter period being fall-free. The
dependent variable is time to each fall.
Marginal Wei, Lin & Weissfeld regression. Three records. Each fall as well as the final
fall-free period are treated in an independent stratum and time measured from entry. The
dependent variable is time to each fall.
Negative binomial regression. One record covers entry until the end of follow-up and
includes simply the total number of falls and follow-up time per subject. The dependent
variable is number of falls.
Logistic regression. One data record, which does not account for follow-up time and
ignores multiple falls by subject. The binary dependent variable is status of faller.
Vol 28, No 4, 2008
drops all additional falls that may occur in
each particular interval. It seems obvious
that if three falls per month are considered
equivalent to one fall for the same period,
the information translating the intensity of
short-term phenomena is lost.
The choice of one of these models must be
made starting from a priori ideas on the
types of relationships which exist between
the covariates and the risk of falling. In
negative binomial regression, AG, and
pooled logistic regression, no distinction
is made between the various events
that succeed one another. This restricts
the baseline hazard and the regression
coefficients do not vary according to the
rank of recurrence. A history of previous
falls is strongly recognized as a predictor of
subsequent falls.10,23 Intuitively, we would
expect a first fall to differ from the aetiology
of the subsequent falls. The predictors for
one fall that can occur by accident might
be different from those for recurrent falls
that can be associated with one’s health
condition.24,25 Hence, researchers and
practitioners may be interested in knowing
not only the overall covariate effects on
the risk of all failures, but also the specific
effects of independent variables for the
first, second, or subsequent events. The
binomial regression, AG, and pooled logistic
regression, contrary to the WLW, provide
no insights to answer such questions. In
accordance with the structure of the data
to be analyzed and the research question
to be answered, the WLW is expected to be
a naturally more appropriate method for
studying the risk factors of falls.
Subjects and procedures
Subjects were volunteers recruited between March 2002 and July 2005 to form
an open cohort that included communitydwelling persons, aged 65 years and
over and receiving home-care services.
People who could speak neither French
nor English, those not able to walk more
than six meters, and those with reduced
communication and cognition were
Vol 28, No 4, 2008
excluded. All subjects gave informed
consent. The study was approved by the
authorities of each participating centre.
This study is a part of a research project on
the evaluation of a multifaceted preventive
intervention.26 Participants were visited at
home, at entry and every six months, by
a trained physical rehabilitation therapist
in order to ascertain information about
potential risk factors. A fall was defined
as an event resulting in the subject
inadvertently coming to rest on the
ground, floor or other lower level (e.g.,
stairs). Excluded were sports-related falls.23
Subjects were asked about falls in the three
months preceding the initial interview and
were monitored for new falls by use of a
daily completed calendar and monthly
phone calls.
Material and social forms of an ecological
deprivation index were imputed to
participants, using postal codes to match
geographic areas of residence with
Canadian census data.27,28 Nutritional risk
screening was performed on a graded
13-point scale tool.29-32 Body weight was
self-reported and height was measured
using standard techniques. Gait, balance,
and mobility performance were assessed
by the Berg scale33-36 on a 56-point scale,
and by the Timed Up & Go test37,38 which
measures the overall time, in seconds,
to complete a series of functional tasks.
Subjects’ homes were assessed for 37
potential environmental hazards using
the Gill’s room-by-room assessment
form39,40 Housing types included: singlefamily house; apartment; row housing
or other unique entrance dwelling units;
private residential facilities for seniors;
other housing, including room in shared
accommodation. Data about the use of
benzodiazepines (yes/no) and number
of daily consumed prescribed drugs were
recorded directly from the containers. A
detailed history of alcohol consumption
was obtained through a questionnaire
developed by the Québec Institute of
Statistics.41,42 Responses were categorized
for both drinking in the preceding week
(yes/no) and usual drinking during the last
6 months (non-drinker, ≤ 2 times a month,
1-6 times a week, every day). Generally,
higher values of the measurements denoted
higher risk or impairment, except for the
Berg scale where the opposite was true.
Statistical analyses
Descriptive analyses were carried out
using SPSS® 13.0; regression analyses
using SAS® 9.1. The adjusted effects of
subject characteristics on the likelihood
of falling were investigated using three
survival-analysis techniques (conventional
Cox regression, AG extension, and WLW
extension), a negative binomial regression,
and a logistic regression.15,43
The dependent variable in all survival
analyses was time to fall for each participant
during the follow-up, measured in days.
Only cases with at least one month of
follow-up fall data were included. Subjects
were censored upon reaching 18 months
of follow-up (optional voluntary drop out),
end of study, or time of withdrawal for any
reason. Repeated falls were considered as
occurrences of the same type of indistinguishable events. Survival analyses were
performed with all covariates measured
on baseline only and with updated
covariates. Baseline covariates included
age, sex, number of falls in the three
months prior to study entry, type of
residence, and deprivation index. For the
time-varying covariates, including BMI,
nutritional risk, alcohol consumption,
home environmental hazards, gait and
balance, use of benzodiazepines and all
medications, the measurement closest to
the time preceding the fall was considered.2
Measurement of exposure to the middle of
the follow-up period was used in the case
of the people who did not fall. Thus we
tested the null hypothesis that the exposure
collected during the follow-up was not
associated to the risk of falling thereafter.2
No proportional hazards assumption was
required in Cox with time-dependent
covariates procedure, since the hazards
depended on time.2,43
Chronic Diseases in Canada
Adjusted relative risk estimates of factors for falls among the community-dwelling elderly,
according to different statistical regression methods
(all falls)
(first fall)
AG Coxa
(all falls)
(all falls)
with baseline covariatesd
Risk factor
Home hazards (nb)
BMI (kg/m2)
Berg score
Time Up & Go score
Age (yrs)
Residential facility housing
(first fall)
AG Coxb
(all falls)
(all falls)
with time-varying covariatese
One prior fall
Two or more prior fallsf
Significant (two-tailed): *p ≤ 0.05; **p ≤ 0.01; ***p ≤ 0.001; ****p ≤ 0.0001
Included number of previous falls during the follow-up as a time-dependent covariate to account for dependence between falls:
IRR =1.10****; b IRR =1.09****.
Subjects monitored less than 12 months who did not declare any falls (n = 221) were excluded, given that we could not define the status of faller.
All covariates measured on baseline only; e up-dated covariates included home hazards, BMI, Berg and Timed Up & Go scales.
History of falls in 3 months preceding initial interview.
The dependent variable in the logistic
regression was the state of being a faller
(subjects who fall at least once) over a
12-month period. The negative binomial
and logistic regressions were performed
with all covariates measured on baseline.
The statistical methods are summarized in
Figure 1. The linearity assumption of the
relationships was checked for continuous
predictor variables. All models were fit
using a stepwise-like process to retain
any variable in the presence of others
with a p-value ≤ 0.05. Robust sandwich
estimates of variance were used in the
survival-analysis, as well as the negative
binomial regression techniques, in order to
compensate for the lack of independence
between multiple falls.
The WLW approach estimated both
common and event-specific β for the
first five falls of each subject, as well as
the common β for all the observed falls.
The number of subjects at risk for a given
stratum, after the first fall, was made up of
all subjects who experienced a fall in the
preceding stratum minus those who were
Chronic Diseases in Canada
lost in the follow-up; n of subjects at risk
for a given pooled fall group was made
up of all subjects under observation in all
considered strata, as if subjects in each
stratum represented a different subject.
Each model was examined both with and
without past fall strata, as it could have
masked the effects of other variables of
Study subjects
Of the 959 persons who met the study
inclusion criteria, agreed to participate,
and received a home visit, 22 withdrew
without completely filling the baseline
assessments or before one month of
follow-up. Mean and median follow-up
times of the remaining 937 subjects were
488 and 458 days, respectively (range, 27
to 1330 days). Some 549 subjects (57.2%)
remained in the study at 12 months and
377 (39.3%) at 18 months. Respondents
were mainly women (75.7%). Mean age
(standard deviation) was 79.5 (6.7), of
which 76.4% were 75 years of age or older.
Thirty-nine percent (39.0%) experienced
at least one fall in the three months prior to
study entry and 14.9% had two or more.
Comparison of statistical methods
Table 1 summarizes the differences in
relative risks for falling obtained through
several statistical methods. Firstly, the
logistic regression (1) and time-to-first
fall using a standard Cox (3a) overlooked
the recurrence of falls and identified less
significant risk factors than did the negative
binomial (2), AG (4a), and WLW (5a), that
considered all the available information
(number between parentheses refers to
the concerned model in Table 1). Although
both logistic regression and standard Cox
identified the same risk factors, logistic
regression ignored the time of occurrence
of falling. This led to a conclusion of
higher magnitude of the related relative
risks, compared to standard Cox. The
values obtained by the logistic regression
were between 17.6% (1.47 vs 1.25) and
39.6% (3.28 vs 2.35) greater than those of
standard Cox.
Vol 28, No 4, 2008
Secondly, three methods - the negative
binomial regression (2), the AG (4a), and
WLW (5a) extensions of the Cox model
considered follow-up time, rate of all
falls, as well as dependence between falls,
using robust estimates of variance. WLW
revealed more significant fall-risk factors
than the other methods and accorded
less importance to the history of falls in
the three months preceding the initial
interview. Notably, the negative binomial
regression, in relation to the WLW,
exhibited a difference of 48.6% (3.15 vs
2.12) for the variable “two or more prior
falls”. The different emphasis given by
these three approaches to the dependence
among multiple event times explains the
difference in results. The negative binomial
regression does not integrate the length of
inter-fall intervals. The AG explicitly models
the impact of earlier falls on future events.
In this regard, the incidence rate ratio (IRR,
virtually equivalent to the so-called hazard
ratios) of 1.10 of the time-dependent term
“number of previous falls” modelled in the
AG (4a) indicates a 10% increase in hazard
for each unit increase in number of prior
falls. In contrast, WLW estimates separate
relationships for each fall and computes
the coefficients and the within-subject
correlation more directly than the AG,
thus providing efficient weighted average
estimates of effect (and variance).
Thirdly, results were compared for the models
both with and without time-dependent
covariables. The number of home hazards,
an exposure particularly likely to vary
during the follow-up, was not significantly
associated with falls in any of the models that
had only baseline covariates (1 to 5a). On the
contrary, the variable was always statistically
significant in the same models that controlled
variation of exposure throughout time (3b to
5b). All survival models with time-varying
covariables identified a greater number of
fall-risk factors than did the corresponding
technique with only baseline covariates (3b
vs 3a, 4b vs 4a, and 5b vs 5a), even when
estimates were calculated from the robust
variance. A more marked difference was
noted between techniques that modeled
only time to first fall and those that took
into consideration time to each fall. For the
marginal WLW model, inattention to timevarying covariables produced bias in various
directions. Lastly, results from the usual
methods of analysis of risk factors for falling
(1 and 3, in Table 1) produced considerable
biases relative to the WLW model using
time-dependent covariates (5b).
Risk factors for falls
The sample of 937 subjects reported 1,270
falls during a total of 457,283 persondays of observation, given that a same
person could report more than one event.
Among the subjects, 495 had no falls,
192 experienced one episode, and 250
had more than one. The consideration
of the first five falls gathered 90.0% of
the 442 fallers and 95.3% of the 937
individuals in the sample. Of all falls for
which information on consequences was
Adjusted and variance-corrected WLW incidence rate ratio by selected risk factors for falls among the
community-dwelling elderly, according to the fall rank or pooled fall group
Fst 5 falls
Fall rank number
Risk factor
na = 937
Falls (nb)
Home hazards (nb)
BMI (kg/m2)
Berg balance score
Benzodiazepine use
n = 429
n = 244
n = 140
n = 93
All falls
n = 1 843
n = 2 169
1 001
1 270
Alcohol use, past 6 months
≤ 2 times per month vs. other categories
Age (yrs)
Residential facility housing
Material deprivation index
Fourth vs. other quartiles
One fall prior initial interview
≥ 2 falls prior initial interviewc
Significant (two-tailed): *p ≤ 0.05; **p ≤ 0.01; ***p ≤ 0.001; ****p ≤ 0.0001.
n of subjects at risk for the considered fall stratum or pooled fall group.
The brackets show the variables had not reached statistical significance after “previous falls” were introduced.
History of falls in 3 months preceding initial interview.
Vol 28, No 4, 2008
Chronic Diseases in Canada
available, 44.4% resulted in injuries,
25.2% in activity limitations, 17.1% in
a medical consultation, and 5.6% in a
hospitalization. Altogether, 82.1% of falls
occurred in the subjects’ home.
Table 2 displays the adjusted associations
between the potential risk factors and the
incidence rate for specific and pooled falls.
The WLW marginal risk estimates for the
first fall stratum in Table 2 are precisely the
same as would occur if the analysis were
restricted to data on time to first fall using
a standard Cox model (column 3b, in Table
1). The only difference is that the p-values
presented in the former were calculated
from the robust rather than standard
(“naïve”) statistics. However, while the
estimates for the first fall stratum are
essentially equivalent in these two cases,
results for the other strata vary substantially
according to whether coefficients are
calculated from robust or from naïve
methods, providing some indication as
to the degree of dependence among the
events. Thus, male sex, residential facility
for seniors, number of home hazards, Berg
balance score, and age significantly and
independently predict time-to-first fall.
For example, the IRR = 1.45 found for
the residential facility for seniors indicated
that the subjects living in such places
experienced falls at a rate which was 45%
higher than for those living in any other
kind of housing. Similarly, the IRR of 1.12
for the home hazards indicated a 12%
increase in hazard for each unit increase in
number of items. However, since age has
an IRR of less than 1 (i.e. 0.98), increase in
age by one year led to decrease in hazard
by 2%.
Table 2 also compares the results when
distinct β were fit for each fall. Covariates
as age, home hazards, and Berg scale
show sustained and relatively constant
effects across the strata. Some others differ
both in the nature and magnitude of the
statistically significant variables, depending
on their position in the sequence. The
greatest differences in IRR appear in the
fifth episode. The entry, in the last step of
history of falling in the three months prior
to study entry, turned out to be highly
significant and did not alter either the
magnitude or significance of the IRR for
Chronic Diseases in Canada
the other variables already included in all
stratum models. The right-hand section
of Table 2 repeats the analysis under the
constraints of overall common β (weighted
average of the event-specific hazards),
both when falls beyond the fifth were not
applied (censored model) and when all fall
data were utilized (complete model). The
censored model identified seven variables,
three more than the time-to-first-fall
model (BMI, use of benzodiazepines, and
occasional alcohol consumption in the
past six months of follow-up) and one
more than the complete model (alcohol
consumption). However, these additional
variables were no longer significant in the
context of the contribution of all others,
once the history of falling was joined to the
censored model; furthermore, the use of
benzodiazepines and alcohol consumption
became insignificant in the complete set.
An age-sex interaction term tested in each
final model was not significant.
This article addresses the proper method of
examining falls and their determinants. No
statistical technique can reproduce human
behaviour exactly, and makeshift solutions
to time-varying exposures and recurrence
of events can lead to severe bias. To our
knowledge, the first and only example
where time-varying exposures and multiple
falls were ascertained simultaneously was
in a doctoral thesis deposited in 199122
and published later in a scientific review.23
However, substantial statistical progress
has appeared since then. In the current
issue, we discuss the various methods
for studying the exposure changes during
follow-up and recurrent events in the
same person. We further illustrate them
by identifying the risk factors for falls in
the elderly. We have concentrated on the
statistical/methodological aspects and
have mentioned the risk factor findings
only to the extent of showing different
results obtained by different analyses.
Methods that handle the aforementioned
data analytical features in a statistically
correct manner are now available in
commercial packages. They have been
addressed extensively in the statistical
literature, but not yet routinely applied and
reported for fall studies, as new advances in
the statistical world are often slow to reach
the clinical and public health fields.4 We
have presented throughout our paper our
arguments as to why the WLW approach is
expected to be an appropriate choice in the
context of our study. It provides a natural
framework for analysing time-varying
exposures and multiple events data using
minimal assumptions.2,44 Other authors
have reported that the WLW is robust
and performs quite well in many practical
The differences in the estimates obtained
through several statistical methods
analysing the risk factors for falling,
have been illustrated according to the
information provided. Results clearly
reveal that the usual methods, such as
binary outcome using a logistic regression
and time to first fall using a standard Cox,
produce considerable biases, as opposed to
the WLW model that uses time-dependent
covariates. In addition, modeling for first
events implicitly assumes that the first
event is representative of all events. Our
study denotes that this assumption is
questionable, more in the qualitative facet
of IRR estimates than in the quantitative.
Our results provide additional evidence
regarding the convenient choice of a
stratified model rather than a non-stratified
one, given that the risk of occurrence varies
substantially between occurrences. Mahé
and Chevret45 expect such possibilities
when the frequency of events per unit is
“small”, such as falls among communitydwelling elderly people.
Furthermore, our results are coherent with
earlier findings, although we are more
confident of the magnitude in estimates of
predictors. A few findings merit comment.
Number of home hazards and history of
falling are strong and consistent predictors
of falls, whatever their rank or pooling.
Prior overall falls increase the risk of
subsequent overall falls. This suggests that
if the causes of past falls - for which the
variable acts as a proxy - are not corrected,
the chances of sustaining further falls
due to the same causes are increased.23
The people living in a residential facility
for seniors are more at risk than others
to fall, possibly because the variable
Vol 28, No 4, 2008
acts as a surrogate measure of various
chronic conditions and poorer functional
autonomy. Similarly, younger people reveal
themselves to be at a higher risk of falling
compared to the older, probably because
of more vigorous lifestyle activities.
We further hope to eliminate any
misunderstanding about any incidence
measures reported in the research literature,
especially the dubious events per persontime relating the number of falls (single
in some subjects, multiple in others) to
the cumulative time of observation of all
subjects. It should not be confused with
the individual event rate adjusted for the
follow-up time that we discuss in our
paper, or with the incidence rate widely
used in epidemiology. In the events per
person-time measure, the numerator does
not express a number of subjects wherein
the event only occurs once, but rather a
number of events scattered among the
study subjects. Windeler and Lange46
have vigorously denounced this concept
because it has no exact interpretation on
an individual level. Hence, event rates
are the same (20 per 100 person-years)
whether 20 subjects are observed for 10
years and each suffers two falls, or 1000
subjects are observed for half a year and
100 (10%) of them have one fall each.
Having been introduced in the ‘80s and
still, unfortunately, sometimes reported
in peer-reviewed journals47-50, this concept
should be abandoned,46 as it impedes the
search for new approaches.
Happily, prospective design, frequent
contacts, repeated measures, and clinical
measurements performed by a therapist
limited information bias. Nonetheless,
some other exposures, such as nutrition
screening and alcohol use, were derived from
self-reports. Differential misclassification
could occur if the fact of a fall or recurrent
falls affected the accuracy with which the
individuals recalled relevant exposures
and subsequent outcomes. This would
exaggerate the magnitude of the effect on
the risk of falling.6 Also, the length of time
between a fall and the measure of follow-
Vol 28, No 4, 2008
up exposure obviously varied according to
the day on which the fall occurred. Hence,
an accurate assessment of exactly when
a change in exposure to time-dependent
covariates might have happened between
each six-month follow-up was not possible.
It would result in non-differential errors in
the measurement of exposures, thereby
diluting the observed relation. Another
potential for biased results might have
occurred because of dropouts, particularly
when the latter do not have the same rate
of outcome (risk of falling) as those who
continue in the study. With the exception of
people who refused the services and who
were less likely to fall, as opposed to the
active participants completing the study,
no other reason for loss to follow-up was
associated with the falls. Male sex, ageing,
residential facility for seniors, first quartile
deprivation index, lower Berg score, and
daily alcohol consumption at baseline
were associated with a significant shorter
duration in participation. As Campbell et
al.51 have already noted, those individuals
who are more frail and may be at greater
risk of falling are the ones most difficult
to involve and sustain in follow-up. This
would also lead to an underestimation of
the effects.
All the aforementioned considerations
lead us to believe that the results observed
in our study tend to be conservative. A
practical drawback of the WLW is the preprocessing effort and care required in the
dataset construction. The application of
this method depends on the completeness
of the reports of falls and knowledge of
calendar dates of falls. Future research
must make the transition from risk factors
for falling to community implementation
of interventions.
Finally, it would be useful to talk about
two substantive clinical findings that have
been deleted from the text. Firstly, the
degree to which balance and gait mediate
the relationship between medication and
the likelihood of falling was estimated.6 A
mediator is an intermediate variable that
occurs in the causal chain between an
exposure and an outcome. If a variable is
truly in the causal pathway, the association
between the latter two variables should
disappear upon adjustment for the
mediator.52 Adjustments for Berg balance
scale resulted in a maximum increase of 21%
in the effect of benzodiazepines, contrary to
the hypothesized reduction. Consequently,
the covariate does not act as a mediator or
as an appreciable confounder.
Secondly, falls leading to a medical consultation were examined as a secondary
outcome, hypothesized as a measure
of severity. For these cases, a variable
“previous falls” was included as a timedependent covariate. It was created to
consider whether a fall not resulting in a
medical consultation had been reported
in the three-month period preceding any
fall-related medical consultation. Adjusted
results identified the number of home
hazards (incidence rate ratio = 1.09), the
nutrition screening score (1.09), living
in a residential facility for seniors (1.67)
and fall history (1 prior fall = 0.59; ≥ 2
prior falls = 0.64) as significant and
independent predictors for all pooled
fall-related medical consultations. Having
fallen in the three months before each new
event under study was protective against
any fall for which people sought medical
The authors gratefully thank all older
clients and health care workers from
the community centres in Lanaudière
who participated in the study. We also
acknowledge the special contribution of
Josée Payette for her effort in preparing
the data files used in the analyses; Nancy
Leblanc, Julie Meloche, and Jean-François
Allaire from the Research Centre at the
Philippe-Pinel Institute of Montreal for
the statistical computations of regression
analyses; and Bruce Charles Bezeau for
the English revision of the manuscript.
The research was sponsored by the Agence
de la santé et des services sociaux de
Chronic Diseases in Canada
Ministère de la Santé et des Services sociaux
du Québec. La prévention des chutes dans un
continuum de services pour les aînés vivant
à domicile, Cadre de référence, Québec,
Direction générale de la santé publique, 2004,
61 p. Available from: URL http://msssa4.
nDocument (last accessed July 14, 2007).
Desquilbet L, Meyer L. [Time-dependent
covariates in the Cox proportional hazards
model. Theory and practice]. Rev Épidemiol
Santé Publique 2005;53(1):51-68. French.
Moulton LH, Dibley MJ. Multivariate timeto-event models for studies of recurrent
childhood diseases. Int J Epidemiol 1997;
Mahé C, Chevret S. Estimation of the
treatment effect in a clinical trial when
recurrent events define the endpoint. Stat
Med 1999;18(14):1821-9.
Glynn RJ, Buring JE. Counting recurrent
events in cancer research. J Natl Cancer
Inst 2001;93(7):488-9.
Cumming RG, Kelsey JL, Nevitt MC.
Methodologic issues in the study of frequent
and recurrent health problems. Falls in the
elderly. Ann Epidemiol 1990;1(1):49-56.
Goodman AC, Hankin JR, Kalist DE, Peng
Y, Spurr SJ. Estimating determinants of
multiple treatment episodes for substance
abusers. J Ment Health Policy Econ
Robertson MC, Campbell AJ, Herbison
P. Statistical analysis of efficacy in falls
prevention trials. J Gerontol A Biol Sci Med
Sci 2005;60(4):530-4.
Wang SJ, Winchell CJ, McCormick CG,
Nevius SE, O’Neill RT. Short of complete
abstinence: an analysis exploration of
multiple drinking episodes in alcoholism
treatment trials. Alcohol Clin Exp Res
Chronic Diseases in Canada
10. Hogan DB, MacDonald FA, Betts J, Bricker
S, Ebly EM, Delarue B, Fung TS, Harbidge
C, Hunter M, Maxwell CJ, Metcalf
B. A randomized controlled trial of a
community-based consultation service to
prevent falls. CMAJ 2001;165(5):537-43.
Glynn RJ, Buring JE. Ways of measuring
rates of recurrent events. BMJ 1996;
12. Lin DY. Cox regression analysis of
multivariate failure time data: the marginal
approach. Stat Med 1994;13(21):2233-47.
13. Finkelstein
Stamenovic E. Analysis of multiple failure
time data from an AIDS clinical trial. Stat
Med 1997;16(8):951-61.
14. Wei LJ, Glidden DV. An overview of
statistical methods for multiple failure
time data in clinical trials. Stat Med
15. Cleves M. How do I analyze multiple failuretime data using Stata?, Stata FAQ, 2002, 11
p. Available from: URL http://www.stata.
com/support/faqs/stat/stmfail.html (last
accessed July 14, 2007).
16. Andersen PK, Gill DR. Cox’s regression
model for counting processes. Ann Statist
17. Prentice RL, Williams BJ, Peterson AV.
On the regression analysis of multivariate failure time data. Biometrika
18. Wei LJ, Lin DY, Weissfeld L.
analysis of multivariate
marginal distributions. J Am
Stat Assoc
19. Lancar R. [Robust analysis methods for
multivariate survival times]. Rev Épidemiol
Santé Publique 1999;47(3):287-96. French
20. Roberston MC. Development of a falls
prevention programme for elderly people
: evaluation of efficacy, effectiveness, and
efficiency. A Ph.D. thesis submitted at
the University of Otago, Department of
Medical and Surgical Sciences, Dunedin,
New Zealand, 2001.
D’Agostino RB, Lee ML, Belanger AJ,
Cupples LA, Anderson K, Kannel WB.
Relation of pooled logistic regression to
time dependent Cox regression analysis:
the Framingham Heart Study. Stat Med
22. O’Loughlin J. The incidence of and risk
factors for falls and fall-related injury
among elderly persons living in the
community. A Ph.D. thesis at the McGill
University, Department of Epidemiology
and Biostatistics, Montreal, Québec,
Canada, 1991, 270 p.
23. O’Loughlin JL, Robitaille Y, Boivin JF,
Suissa S. Incidence of and risk factors
for falls and injurious falls among the
community-dwelling elderly. Am J
Epidemiol 1993;137(3):342-54.
24. Nevitt MC, Cummings SR, Kidd S, Black
D. Risk factors for recurrent nonsyncopal
falls. A prospective study. JAMA 1989;
25. Fletcher PC, Hirdes JP. Risk factors for
falling among community-based seniors
using home care services. J Gerontol A Biol
Sci Med Sci 2002;57(8):M504-10.
26. Bégin C. Projet-pilote régional de prévention
des chutes à domicile chez les personnes
âgées, Devis d’implantation dans les
CLSC, Saint-Charles-Borromée, Service de
prévention et de promotion, Direction de
santé publique, Régie régionale de la santé
et des services sociaux de Lanaudière,
2002, 120 p.
27. Pampalon R, Hamel D, Raymond G. Indice
de défavorisation pour l’étude de la santé et
du bien-être au Québec - Mise à jour 2001.
Institut national de santé publique du Québec,
2004, 11 p. Available from: URL http:// (last accessed
December 24, 2006).
Vol 28, No 4, 2008
28. Leclerc BS, Marquis G, Payette J. Tableau
de bord lanaudois sur la défavorisation.
Territoire de Lanaudière. Calibrage à
l’échelle de la région de Lanaudière,
Joliette. Agence de la santé et des services
sociaux de Lanaudière, Direction de
santé publique et d’évaluation, Service
de surveillance, recherche et évaluation,
2005, 87 p.
29. Payette H. Développement, validation et
évaluation d’un programme de dépistage
nutritionnel pour les personnes âgées
en perte d’autonomie vivant dans la
communauté, Sherbrooke, Centre de
recherche en gérontologie et gériatrie, Centre
d’expertise en gérontologie et gériatrie
inc., Institut universitaire de gériatrie de
Sherbrooke, non daté, pages multiples.
30. Payette H, Guigoz Y, Vellas BJ. Study
design for nutritional assessments in the
elderly », in Methods in Aging Research,
B.P. YU (ed), Boca Raton (Florida), CRC
Press LLC, 1999, p. 301-20.
Laporte M, Villalon L, Payette H. Simple
nutrition screening tools for healthcare
facilities: development and validity
assessment. Can J Diet Pract Res
32. Laporte M, Villalon L, Thibodeau J,
Payette H. Validity and reliability of simple
nutrition screening tools adapted to the
elderly population in healthcare facilities.
J Nutr Health Aging 2001;5(4):292-4.
33. Berg K. Balance and its measure in
the elderly: a review. Physiother Can
34. Berg KO, Maki BE, Williams JI, Holliday
PJ, Wood-Dauphinee SL. Clinical and
laboratory measures of postural balance
in an elderly population. Arch Phys Med
Rehabil 1992;73(11):1073-80.
35. Berg KO, Wood-Dauphinee SL, Williams JI,
Maki B. Measuring balance in the elderly:
validation of an instrument. Can J Public
Health 1992;83(suppl. 2):S7-11.
Vol 28, No 4, 2008
36. Berg K, Wood-Dauphinee S, Williams JI.
The Balance Scale: reliability assessment
with elderly residents and patients with
an acute stroke. Scand J Rehabil Med
37. Podsiadlo D, Richardson S. The timed «Up
& Go»: a test of basic functional mobility
for frail elderly persons. J Am Geriatr Soc
38. Lin MR, Hwang HF, Hu MH, Wu HD,
Wang YW, Huang FC. Psychometric
comparisons of the timed up and go,
one-leg stand, functional reach, and
Tinetti balance measures in communitydwelling older people. J Am Geriatr Soc
39. Gill TM, Williams CS, Robison JT,
Tinetti ME. A population-based study
of environmental hazards in the homes
of older persons. Am J Public Health
40. Gill TM, Williams CS, Tinetti ME.
Environmental hazards and the risk
of nonsyncopal falls in the homes of
community-living older persons. Med Care
41. Chevalier S, Lemoine O. Consommation
d’alcool, in Enquête sociale et de santé
1998, Québec, Institut de la statistique du
Québec, 2000, p. 117-33. (Collection La
santé et le bien-être)
42. Institut de la statistique du Québec Annexe
3. Questionnaire autoadministré (QAA) VI
– L’alcool, in Enquête sociale et de santé
1998, Québec, Institut de la statistique du
Québec, 2000, p. 15-7. (Collection La santé
et le bien-être)
45. Mahé C, Chevret S. Analysis of recurrent failure times data: should the
baseline hazard be stratified? Stat Med
46. Windeler J, Lange S. Events per
person year - a dubious concept. BMJ
47. Saari P, Heikkinen E, Sakari-Rantala R,
Rantanen T. Fall-related injuries among
initially 75- and 80-year old people during
a 10-year follow-up. Arch Gerontol Geriatr
48. Sambrook PN, Cameron ID, Chen JS,
Cumming RG, Lord SR, March LM, Schwarz
J, Seibel MJ, Simpson JM. Influence of
fall related factors and bone strength on
fracture risk in the frail elderly. Osteoporos
Int 2007;18(5):603-10.
49. Cook WL, Tomlinson G, Donaldson M,
Markowitz SN, Naglie G, Sobolev B, Jassal
SV. Falls and fall-related injuries in older
dialysis patients. Clin J Am Soc Nephrol
50. Mahoney JE, Palta M, Johnson J, Jalaluddin
M, Gray S, Park S, Sager M. Temporal
association between hospitalization and
rate of falls after discharge. Arch Intern
Med 2000;160(18):2788-95.
Campbell AJ, Robertson MC, Gardner MM,
Norton RN, Buchner DM. Falls prevention
over 2 years: a randomized controlled trial
in women 80 years and older. Age Ageing
52. Szklo M, Nieto FJ. Epidemiology: Beyond
the basics. Gaithersburg, MD: Aspen
Publications, 2000, 494 p.
43. Allison PD. Survival analysis using SAS:
A practical guide. Cary NC: SAS Institute
inc., 1995, 304 p.
44. Li QH, Lagakos SW.Use of the Wei-LinWeissfeld method for the analysis of a
recurring and a terminating event. Stat
Med 1997;16(8):925-40.
Chronic Diseases in Canada
Gender and the smoking behaviour of Ethiopian
immigrants in Toronto
I Hyman, PhD (1); H Fenta, PhD (2); S Noh, PhD (3)
The objective of this paper is to present descriptive data on gender and smoking among
Ethiopian immigrants in Toronto, Canada. The study used a cross-sectional epidemiological
survey design (N = 342). The main outcome measures identified subjects as current
(regular or occasional) smokers, daily smokers and former smokers. Overall, 20.8% of the
individuals in the sample were current smokers and 15.7% were daily smokers. Although
smoking rates (current and daily) were significantly higher among males compared to
females, nearly twice as many female as male daily smokers reported that they began
smoking post-migration (60.0% vs. 30.2%). Furthermore, 80.0% of female compared to
nearly 56% of male daily smokers reported that they were smoking more post-migration.
A significantly higher proportion of males compared to females were former daily smokers
(17.8% vs. 4.4%). These findings present a challenge for public health professionals in
terms of preventing the adoption of smoking among Ethiopian females and facilitating
smoking cessation among Ethiopian males. Correlates with current smoking suggested
that smoking prevention and cessation programs in newcomer immigrant communities
may benefit from incorporating social, economic and religious contexts of these newcomers’
lives from a gender-specific perspective.
Key words: smoking, immigration, gender
Smoking is a major risk factor for mortality
and morbidity.1,2 Several studies suggest
that smoking rates among immigrants to
Western countries have increased and this
is why many cancer rates among immigrants
are converging with those of the nativeborn population.3-5 Risk and protective
factors associated with smoking behaviour
are well documented in the literature. These
include age, gender, religiosity, level of
education, employment, stress and social
support.6-9 Among these, gender is a particularly prominent factor for immigrant
communities, with female immigrants
typically exhibiting significantly lower rates
of smoking than males.10-13,16 However, few
studies have examined smoking and correlates of smoking among recent immigrants
to Canada. This information is critical to
inform the development of smoking
prevention and cessation strategies targeting
newcomer communities. This study draws
on a community survey of Ethiopian
immigrants and refugees in Toronto,
Canada. The purpose of this paper is to
present descriptive data on the smoking
behaviour and on the risk and protective
factors associated with smoking in this
Literature Review
Post-migration changes in smoking
behaviour are well-documented3-5. Less
well documented are the determinants of
smoking in immigrant populations although
proposed models to explain these changes
include prolonged exposure to stressful
events, adverse circumstances such as
disadvantaged socio-economic status or
the loss of social networks, smoking
behaviour as a coping response to discrimination and poverty, and acculturative
changes in beliefs, values and norms about
It is further recognized that immigrants
are not a homogeneous group.15 Using
data from the 1996 National Population
Health Survey and the 2000-01 & 2002-03
Canadian Community Health Surveys,
McDonald16 found major differences in
smoking rates within Canadian immigrants
characterized by gender and length of stay
in the host country. For example, both
immigrant males and females showed an
increase in rates of smoking by length of
stay in Canada but after adjusting for
differences in demographic and socioeconomic characteristics, there was only
evidence of significant convergence for
male immigrants. Regional differences
were also observed. For example, among
non-English speaking immigrants born in
countries outside of Europe, immigrants
from the Middle East and Western Asia
were more likely to smoke compared to
immigrants from East Asian countries
(control group), while immigrants from
Southern Asia were less likely to smoke
compared to the control group. The logodds ratio for daily smoking for African
immigrants was not significantly different
from that of the control region of birth.
Gender is increasingly being recognized
as a determinant of immigrant women’s
health.17 Gender is known to influence
both settlement processes as well as
cognitive schemas about health.17,18 The
intersections of gender with minority
Author References
1 Department of Public Health Services, University of Toronto
2 The Ontario HIV Treatment Network
3 Department of Psychiatry, University of Toronto
Correspondence: Ilene Hyman, CERIS – The Ontario Metropolis Centre, 246 Bloor St. West, 7th Floor, Toronto, Ontario, Canada, M5S 1V4, Tel: (416) 946-0116,
Email: [email protected]
Chronic Diseases in Canada
Vol 28, No 4, 2008
status, income, employment and social
integration may impact directly on
exposure to stressful events and indirectly
on health.18- 20
Few studies have examined the smoking
behaviour or psychosocial and economic
determinants of smoking behaviour among
specific immigrant populations by gender.
It seems clear that to address identified
knowledge gaps, smoking research needs to
include: 1) studies of the prevalence and
correlates of smoking in specific immigrant
communities; and 2) intersectional studies
that examine smoking behaviour from both
a gender and a migration perspective.
Study Background
Since the mid-1970’s, Ethiopia has
experienced a major exodus of refugees. An
estimated 1.25 million Ethiopians fled to
neighbouring countries, such as Sudan,
Kenya, Djibouti and Yemen, and a relatively
smaller proportion of Ethiopians immigrated
to Europe and North America.21 Between
1974 and September 1998, over 13 000
Ethiopians migrated to Ontario.22 According
to the Ethiopian Association in Toronto, the
current Ethiopian population of Toronto
numbers from 45 000 to 50 000. In 1997, a
partnership was formed between the Culture,
Community and Health Studies Program of
the Centre for Addiction and Mental Health
and the Ethiopian Association in Toronto to
examine mental health, health, and access
issues experienced by the community. The
current paper draws on an epidemiological
study, Pathways and Barriers to Health Care
for Ethiopians in Toronto, conducted by the
authors, that collected extensive data on the
health and resettlement experiences of this
community. Ethics approval for this research
was obtained from the University of Toronto
Ethics Board.
Sample and Data. The study used a crosssectional epidemiological survey design. A
variety of strategies were used to recruit
subjects. Using a snowball technique, we
Vol 28, No 4, 2008
identified all possible Ethiopian ethnic,
religious, political and social organizations
in Toronto and obtained membership lists
from each organization. In addition, a list
of Ethiopian specific names was compiled
using the city telephone directory. Since
some Ethiopian Islamic names (e.g.
Mohammed, Osman, Fatuma) could not
be readily distinguished from nonEthiopian Islamic names, the Islamic
names from the telephone directory were
compiled separately and given to different
Ethiopian Muslim organizations and other
Ethiopian associations to identify those
who were of Ethiopian origin. We speculated that the Ethiopian organizations and
associations in Toronto could know some
of these Muslims although they might not
necessarily be included in their membership lists. The membership lists from the
different organizations, the Ethiopian
specific names and non-Ethiopian specific
Islamic names from the telephone directory
were combined to form a sampling frame.
The resulting sampling frame consisted of
4854 households. From this list, 400 households were selected using simple random
sampling method and one individual,
eighteen years or older, was selected from
each household. Ethiopians who resided in
Canada for less than 12 months were
excluded from the study. The interviews
were conducted by trained Ethiopian interviewers in Amharic. All interviewers
underwent extensive training including
interview skills training, procedures for
contacting potential subjects and general
information on immigration, settlement,
immigrant health and mental health. The
project was announced to the members of
the Ethiopian community through religious
organizations, community media and a
community newsletter. Between May 1999
and May 2000 a total of 342 individuals
completed the structured interview with an
overall response rate of 85%.
Measurement. Questions on smoking were
modelled on those appearing in national
and provincial health surveys. All respondents were asked to classify themselves as
current smokers (regular or occasional),
non-smokers and former daily smokers.
For current smoking status, respondents
who were smoking cigarettes at the time of
survey, regularly or occasionally, were
coded as 1.
Respondents who smoked cigarettes regularly or occasionally were asked whether
they smoked daily. Daily smokers were
then asked to provide information on the
age they began smoking and the number
of cigarettes smoked daily. Using information on age of arrival in Canada and the
age at which respondents started smoking
daily, we were able to determine the postimmigration onset of smoking – respondents who had started smoking daily in
Canada. In a separate question, daily
smokers were asked to indicate changes in
smoking habits since leaving Ethiopia
(smoking more now = 1, smoking less
now = 2, no change = 3). These two
variables were not mutually exclusive.
Former smoking status was determined for
respondents who used to smoke daily but
now classified themselves as non-smokers
or occasional smokers.
An array of potential risk and protective
factors were included in the analysis.
Marital status was coded as: currently
married (1), single, separated, divorced,
widowed (0). To determine levels of
religiosity respondents were asked to rate
the importance of their religious beliefs.
Responses were coded as: very important
or important (1), not so important or not
important (0). Level of education was
coded as: high school or less than high
school (1), college (2), and university
degree (3). Employment was coded as:
currently employed (1) and currently not
employed (0).
To assess post-migration stress, we used
the 14-item recent life events scale included
in the Quebec Health Survey (QHS), which
was derived from Paykel and colleagues.23
The QHS included events specifically
relevant to immigrant and minority communities (e.g. trouble because people
didn’t understand your language, trouble
with prejudice or discrimination). Stress
was computed as the count of events
Chronic Diseases in Canada
Description of the study of population by gender
experienced by the respondents during 12
months prior to the interview.
To assess exposure to pre-migration stressors, subjects were asked whether they
had experienced pre-migration traumatic
exposures to war and killing (coded as
yes=1, no=0), or whether they had been
interned in a refugee camp (yes=1,
The social support measure used was
adapted from Wolchik, et al.24 Three dimensions of social support were assessed –
advice and information, instrumental, and
emotional. For each dimension of support,
subjective ratings of satisfaction with
supports available were obtained on a scale
of 1 to 10. In this study, we used the index
score that combined the scores of the three
dimensions. The scale had internal consistency of 0.96 as measured by Cronbach’s
Data Analysis. Data analysis consisted of
descriptive analysis of the study population
in terms of socio-demographic characteristics and smoking behaviours. The proportion of the population considered to be
former smokers was calculated. Daily
smokers were described in terms of the
number of cigarettes smoked per day, time
of initiation of smoking (i.e. post-migration), and changes in smoking habits since
leaving Ethiopia.
Correlates of current smoking were examined using bivariate logistic regression
analysis. All analyses were gender-specific.
Multivariate analyses for females were not
conducted due to small cell sizes for female
smokers, a limitation of the present study.
Thus the results of this study do not identify
a model of smoking; rather, potential risk
and protective factors associated with
current smoking are identified that can be
tested in future studies.
Sample Description. Table 1 summarizes
data on the socio-demographic characteristics of the study sample by gender.
Approximately 60% of the sample was
Chronic Diseases in Canada
(n = 342)
(n = 203)
(n = 139)
Age in years (mean ± SD)***
35.4 ± 7.2
36.7 ± 7.3
33.3 ± 6.3
Marital status (married) — (%)
Ethiopian Orthodox
Religion — (%)
Roman Catholic
No religion
Religion is important to me
Religion is not important to me
High school or less
College education
University degree
Importance of religiosity — (%)*
Level of education — (%)***
Currently employed — (%)***
Length of stay in Canada (mean ± SD)
9.2 ± 4.5
9.5 ± 4.7
8.7 ± 4.2
Pre-migration trauma exposure — (%)**
Refugee camp internment experience — (%)
no stressful life events
1-2 stressful life events
≥ 3 stressful life events
8.3 ± 1.7
8.2 ± 1.8
8.5 ± 1.4
Number of post-migration life events — (%)
Satisfaction with social support (mean ± SD)
* p < 0.05; ** p < 0.01; *** p < 0.001
male. The respondents were predominantly
a young group with a mean age of 35.3
years, and female respondents were
significantly younger than males. The
majority of respondents were married
(55.4%) and members of the Ethiopian
Orthodox church (67.7%). A significantly
higher proportion of female respondents
considered religion to be very important
compared to male respondents (96.2% vs.
88.5%). More than two-third of the
respondents (67.8%) had some postsecondary education and the majority
(78.1%) were currently employed. Compared to males, females were significantly
less likely to have a university education
and be employed (p < 0.001). The average
length of stay in Canada was 9.2 years (9.5
years for males and 8.7 years for females).
A significantly higher proportion of males
than females experienced pre-migration
trauma (26.71% vs. 11.94%) and refugee
camp internment (12.6% vs. 7.2%),
although in the latter case the difference
was not significant. Approximately half of
male and female participants had
experienced at least one stressful life event.
Males and females were similar in their
reported levels of satisfaction with social
Descriptive Data on Smoking. Table 2
describes the study population by smoking
status and gender. Overall, 20.8% of the
sample were current smokers (used
cigarettes regularly or occasionally). The
current smoking rate was 2.7 times higher
among male respondents (n = 56, 27.7%)
compared to female respondents (n = 14,
10.4%). The gender differential in current
Vol 28, No 4, 2008
Description of the study of population by smoking status and gender
(n = 342)
(n = 203)
(n = 139)
Former smoker — (%)***
Daily smoker — (%)***
Daily smoking variables
(n = 53)
(n = 43)
(n = 10)
12.0 (6.7)
12.0 (6.7)
11.8 (6.7)
Initiation of daily smoking post-migration — (%)
Increase in daily smoking post-migration — (%)
Smoking variables
Current smoker (regular or occasional) — (%)***
Number of cigarettes smoked daily — Mean (SD)
*** p < 0.001; a p < 0.1
smoking rates was significant (p < 0.001).
There was also a significant gender
difference in the proportion of the population
who were former smokers (17.8% males,
4.4% females, p < 0.001).
Among the 53 daily smokers, 21.3%
(n = 43) were male and 7.4% (n = 10)
were female (p < 0.001). The mean number
of cigarettes smoked daily was similar for
males and females (12.0 and 11.8,
respectively). Approximately 35.8% of daily
smokers (n = 53) started smoking postmigration and gender differences were
apparent; only 30.2% of males but 60.0%
of females were non-smokers before they
left Ethiopia, and began smoking after they
left the homeland (p < 0.1). Gender differences were also apparent among the
respondents to the question on changes in
smoking habits since leaving Ethiopia.
Approximately 56% of male smokers and
80.0% of female smokers reported an
increase in smoking post-migration; this
difference did not reach statistical
Correlates of Current Smoking. Bivariate
associations of current smoking (regular or
occasional) with potential risk and protective factors were assessed using logistic
regression. Results are shown in Table 3.
Among male respondents, current smokers
(compared to non-current smokers), were
less likely to report religion as important or
very important to them and more likely to
have experienced pre-migration trauma,
and to have reported higher levels of
satisfaction with social support. The same
associations were not observed among
Ethiopian females. Among Ethiopian
females, current smokers (compared to
non-current smokers) were less likely to be
currently married and more likely to have
spent more years in Canada and to have
experienced post-migration life events. We
found different sets of correlates for males
and females. Multiple logistic regression
analysis of factors associated with current
smoking in men confirmed the results of
the bivariate analysis among Ethiopian
males. This analysis was not performed for
Ethiopian females due to small sample
sizes (data available upon request).
Our study findings highlight dramatic
gender differentials in the smoking
behaviour of Ethiopian males and females,
which persist post-migration. Striking
differences were also observed in the
Bivariate logistic regression of factors associated with regular/occasional smoking by gender
Age (years)
(0.97, 1.05)
(0.94, 1.11)
Currently married (= 1)
(0.33, 1.16)
(0.05, 0.77)
Religion very important (=1)
(0.08, 0.51)
High school or less (=1)
(0.44, 2.42)
(0.15, 13.11)
College (=1)
(0.74, 3.17)
(0.16, 12.66)
University (reference)
Employment (1 = employed)
(0.32, 1.95)
(0.43, 4.99)
Length of stay in Canada
(0.96, 1.10)
(1.02, 1.33)
Exposed to pre-migration trauma (=1)
(1.10, 3.99)
(0.06, 4.46)
Refugee camp internment (=1)
(0.77, 5.07)
(0.11, 9.68)
Number of post-migration life events
(0.88, 1.32)
(0.99, 1.97)
Satisfaction with social support
(1.01, 1.49)
(0.68, 1.60)
OR = Odds Ratio
CI = 95% confidence interval of the estimated OR
na = No estimate due to empty cells
* p < 0.05; *** p < 0.001
Vol 28, No 4, 2008
Chronic Diseases in Canada
correlates of current smoking by gender.
Among Ethiopian males, factors associated
with regular or occasional smoking
included pre-migration trauma, religiosity,
and satisfaction with social support.
Female smokers (compared to nonsmokers) were more likely to be married,
have spent more years in Canada and to
have experienced a higher number of
stressful life events.
The results of this study are consistent
with gender differentials in smoking rates
observed in the 2003 Global Youth Tobacco
Survey among Ethiopian students (grades
9-11), which reported smoking rates of
15.2% and 5.7% for boys and girls,
respectively.25 They are also consistent
with other studies conducted among
native-born and foreign-born populations
in the United States that suggest that
immigrant women are less likely to smoke
than their male immigrant and female
native-born counterparts.9-12 Although
McDonald16 did not find significant
increases in the log-odds of smoking
among immigrant women by length of
stay, similar trends were observed for
immigrant women. It is also possible that
his findings may have masked countryspecific variations in immigrant smoking
This study is the first study to collect data
on the smoking behaviour of Ethiopian
immigrants in Canada. Although current
smoking rates in Ethiopia are not available,
studies conducted in the 1980s showed
that 38% of males and 3.4% of females
were smokers,26 suggesting that male
Ethiopian immigrants may have had a
healthier lifestyle (e.g. lower smoking
rate) than Ethiopians in their home
country, but females appear to exhibit the
opposite pattern. Still, compared to
smoking rates among males and females
in the Toronto population as a whole, the
smoking rate for Ethiopian males in
Toronto (27.7%) was slightly higher
(25.6%), whereas the rate for Ethiopian
females in Toronto (10.4%) was considerably lower (17.8%).27
The data presented in this paper confirm
other studies that suggest that immigrant
behaviours, such as consumption of a high
Chronic Diseases in Canada
fat diet, smoking and alcohol abuse,
increase over time to resemble those of the
majority culture.13,28-29 Only 44.2% of maleand 20.0% of female- daily smokers
reported “no change” in smoking since
migration. The vast majority of female
daily smokers (80%) and over half of male
daily smokers reported that they were
smoking more since they left Ethiopia.
Furthermore, 18.3% of Ethiopian males
but only 4.4% of Ethiopian females could
be classified as “former smokers.” In the
case of male Ethiopians, this is a positive
finding, suggesting that the daily smoking
rate for Ethiopian males is declining and
converging with that of the Toronto male
population as a whole. On the other hand,
the prevalence of current smoking among
females in the Ethiopian community was
initially much lower than that among
females in the Toronto population as a
whole, but may be increasing.
Among Ethiopian males, factors associated
with regular or occasional smoking
included religiosity and satisfaction with
social support. Religiosity had an inverse
effect on smoking behaviour (i.e. highly
religious people were less likely to smoke).
Other studies have found strong religious
beliefs and cultural retention to be
associated with maintaining healthenhancing behaviour such as a traditional
high-fibre diet, non-smoking and non-use
of alcohol.15,30-31 While social support is
usually associated with positive health
behaviours,4,32-33 in the Ethiopian community, with regards to smoking, the
opposite pattern was observed. This may
be because many Ethiopian males,
particularly without close family in
Canada, socialise in smoking environments
such as cafes and bars. The same may not
be true for women because of social norms
restricting women’s freedom of movement
and gender role expectations.21,34
Pre-migration traumatic experiences were
associated with an increased risk of
smoking, but only among males. This
finding was consistent with other research
suggesting strong relationships between
traumatic stress and smoking behaviour in
non-immigrant populations.35-36 It is possible that this association was not observed
among females because only a small subset
of women experienced
trauma and smoked.
Despite the finding that Ethiopian males
experienced more stressful life events
compared to Ethiopian females, postmigration stressful life events were only
associated with regular or occasional
smoking in the latter group. The acculturative stress model proposes that immigrants
turn to artificial support when they find
the experience of immigration stressful
and disorienting. Alternatively, it has been
suggested that alcohol and substance use
are “learned” as a part of an assimilation
process.37 Our findings suggest that male
and female Ethiopian immigrants may use
different coping behaviours in response to
stress. For example, Ethiopian women may
have less opportunity to consume alcohol
and drugs and/or these behaviours may be
considered less culturally acceptable compared with smoking.
Our finding that females who resided in
Canada longer were more likely to smoke
is not surprising. A study of pregnant
Southeast Asian immigrant females in
Montreal found that women who were in
Canada longer were less likely to report
adequate social support and more likely to
report stressful life experiences than
newcomer immigrant women.38 Similarly
newcomers’ strategies for coping with
stress may also change with time spent in
the host country.
The present study was the first community
health survey of Ethiopian immigrants in
North America. Among its strengths are
the attempts made to develop a comprehensive sampling frame and the excellent
response rate. However the study had
certain limitations that must be acknowledged. Firstly, potential candidates were
excluded if they had no telephone, stable
address or membership status in Ethiopian
organisations. Secondly, the low prevalence
rate of smoking among Ethiopian females
in the study population limited statistical
power and prevented the identification of
interaction terms in a multivariate logistic
regression analysis.
These findings present an obvious challenge for public health professionals and
Vol 28, No 4, 2008
the Ethiopian community; that is, how to
prevent the adoption of smoking among
Ethiopian females and how to facilitate
smoking cessation among Ethiopian males.
Feminist researchers increasingly view
smoking and substance abuse not simply
as a negative lifestyle choice but as a
rational response to real pressures associated with gender and class inequity.39-40
New immigrant women face multiple and
more barriers to maintaining or changing
health behaviour, compared to women in
the general population.42-43 Study findings
suggest that smoking prevention and
cessation programs in the Ethiopian
community must acknowledge and address
the gender, social and religious contexts of
these newcomers’ lives. These findings are
equally relevant to other newcomer
communities in Canada who may
experience similar contextual and genderspecific issues.
Acevedo M. The role of acculturation in
explaining ethnic differences in the prenatal
health-risk behaviours, mental health, and
parenting beliefs of Mexican American and
European American at-risk women. Child
Abuse Neglect. 2000;24:111-27.
Acevedo-Garcia D, Pan J, Jun HJ, Osypuk
TL, Emmons KM. The effect of immigrant
generation on smoking. Soc Sci Med.
Ng E, Wilkins R, Gendron F, Berthelot JM.
Dynamics of Immigrants’ Health in Canada:
Evidence from the National Population
Health Survey. 2005. URL: http://www.
Public Health Agency of Canada. Toward a
Healthy Future: Second Report on the
Health of Canadians. 1999. URL: http://
Support from our community partner, the
Ethiopian Association in Toronto. Funding
from the Centre of Excellence for Immigration and Settlement (CERIS) and Heritage
Canada. The authors acknowledge the
contribution of Dr. Lorraine Greaves,
Executive Director, British Columbia Centre
of Excellence for Women’s Health who
reviewed the final manuscript.
Baron-Epel O, Haviv-Messika A, Tamir D,
Nitzan-Kaluski D, Green M. Multiethnic
differences in smoking in Israel: Pooled
analysis from three national surveys. Eur J
Public Health. 2004;14:384-89.
Maxwell AE, Bernaards CA, McCarthy
WJ. Smoking prevalence and correlates
among Chinese- and Filipino-American
adults: Findings from the 2001 California
Health Interview Survey. Prev Med.
Heart and Stroke Foundation of Canada.
The Changing Face of Heart Disease and
Stroke in Canada 2000. Ottawa: Heart and
Stroke Foundation of Canada, 1999.
Cancer Care Ontario. Cancer Incidence
and Mortality in Ontario, 1964-1996.
Toronto: Cancer Care Ontario, 2000.
Kliewer EV, Smith KR. Ovarian cancer
mortality among immigrants in Australia
and Canada. Cancer Epidemiol Biomarkers
Prev. 1995;4:453-58.
Vol 28, No 4, 2008
10. Wilkinson LM, Spitz MR, Strom SS, et al.
Effects of nativity, age at migration, and
acculturation on smoking among adult
Houston residents of Mexican descent. Am
J Public Health. 2005;95:1043-49.
Taylor KL, Kerner JF, Gold KF, Mandelblatt
JS. Ever vs. never smoking among an
urban, multiethnic sample of Haitian-,
Caribbean-, and US-born blacks. Prev Med.
12. King G, Polednak AP, Bendel R, Hovey D.
Cigarette smoking among native and
foreign-born African Americans. Ann
Epidemiol. 1999;9:236-44.
13. Chen J, Ng E, Wilkins R. The health of
Canada’s immigrants in 1994-5. Health
Rep. 1996;7(4):33-45.
14. Albrido-Lanza AF, Armbrister AN, Florez
KR, Aguirre AN. Toward a theory-driven
model of acculturation in public health
research. Am J Public Health. 2006;96:13421346.
15. Hyman I. Immigration and Health. Working
Paper 01-05. Health Policy Working Paper
Series. Ottawa: Health Canada, September,
2001. URL:
16. McDonald JT. The health behaviours of
immigrants and native-born people in
Canada. Working Paper No 01-06. Halifax/
Moncton: Atlantic Metropolis Centre, 2006.
17. Thurston WE and Vissandjee B. An
ecological model for understanding culture
as a determinant of women’s health.
Critical Public Health. 2005;15:229-242.
18. Walters V. The social context of women’s
health. BMC Women’s Health. 2004; 4: S2.
19. Im E & Yang K. Theories on immigrant
women’s health. Health Care for Women
International. 2006;6:666-681.
20. Agnew V. Gender, Migration and Citizenship
Resources Project, Part II: A Literature
Review and Bibliography on Health.
Toronto: Centre for Feminist Research,
York University, 2002.
McSpadden L, Moussa H. I have a name:
The gender dynamics in asylum and in
resettlement of Ethiopian and Eritrean
refugees in North America. J Refugee
Studies. 1993;6:203-25.
Chronic Diseases in Canada
22. George U, Mwarigha MA. Consultation on
Settlement Programming for African
Newcomers: Final report. Toronto:
University of Toronto: Centre for Applied
Social Research, 1999.
23. Paykel ES, Mayers JK, Diendelt MN,
Klerman GL, Lindenthal JJ, Pepper MP.
Life events and depression: A controlled
study. Arch Gen Psychiat. 1969;25:340-47.
24. Wolchik SA, Beals J, Sandler IN. Mapping
children’s support networks: Conceptual
and methodological issues. In: Belle D, ed.
Children’s Social Networks and Social
Supports. New York: John Wiley & Sons,
25. Centres for Disease Control. 2003. Ethiopia
– Addis Ababa Global Youth Tobacco
Survey (GYTS). URL: http://www.afro.
26. World Health Organization. Tobacco or
Health: A Global Status Report, Country
Profiles by Region. 1997. URL: http://
27. Toronto Health Profiles. 2001. URL: http://
28. Hyman I. Immigrant and visible minority
women. In: Stewart DE, Cheung A, Ferris
LE, Hyman I, Cohen M, Williams IJ, eds.
Ontario Women’s Health Status Report.
Toronto: Ontario Women’s Health Council,
Chronic Diseases in Canada
29. Statistics Canada. How Healthy are
Canadians? Annual report. Ottawa:
Statistics Canada, 1999. Cat. No. 82-003SIE.
37. Rebhun LA. Substance use among
immigrants to the United States. In: Loue
S, ed. Handbook of Immigrant Health.
New York: Plenum Press, 1998:493-519.
30. Marmot MG, Syme SL. Acculturation and
coronary heart disease in JapaneseAmericans. Am J Epidemiol. 1976;
38. Hyman I, Dussault G. Negative consequences
of acculturation: Low birthweight in a
population of pregnant immigrant women.
Can J Public Health. 2000;91:357-61.
Scribner R, Dwyer JH. Acculturation and
low birthweight among Latinos in the
Hispanic HANES. Am J Public Health.
39. Adrian M, Lundy C, Eliany M, eds.
Women’s use of alcohol, tobacco and other
drugs in Canada. Toronto: Addiction
Research Foundation, 1996.
32. Sabogal F, Marin G, Otero-Sabogal R,
Marin BV, Pérez-Stable EJ. Hispanic
familism and acculturation: What changes
and what doesn’t? Hispanic J Behav Sci.
40. Horne T. Women and Tobacco. A Framework for Action. Ottawa: Health Canada,
33. Suarez L. Pap smear and mammogram
screening in Mexican-American women:
The effects of acculturation. Am J Public
Health. 1994;84:742-46.
34. Matsuoka A, Sorenson J. Ghosts and
Shadows: Construction of Identity and
Community in an African Diaspora. Toronto:
University of Toronto Press, 2001:13-14.
35. Vlahov D, Galea S, Ahern J, et al. Consumption of cigarettes, alcohol, and marijuana
among New York City residents six months
after the September 11 terrorist attacks. Am
J Drug Alcohol Ab. 2004;30:385-407.
41. Gomberg ESL and Nirenberg TD.
Antecedents and Consequences. In
Gomberg ESL and Nirenberg TD eds.
Women and Substance Abuse. New Jersey:
Ablex Publishing Corporation, 1993.
42. Oxman-Martinez J, Abdool SN and
Loiselle-Leonard M. Immigration, women
and health in Canada. Can J Public Health.
2000; 91:394-395.
43. Vissandjee B, Weinfeld M, Dupéré S, et al.
Sex, gender, ethnicity and access to health
care services: Research and policy challenges for immigrant women in Canada.
JIMI. 2001;2(1).
36. op den Velde W, Aarts PGH, Falger PRJ, et
al. Alcohol use, cigarette consumption and
chronic post-traumatic stress disorder.
Alcohol Alcoholism. 2002;37:355-61.
Vol 28, No 4, 2008
Ethnicity and mental health: Conceptualization,
definition and operationalization of ethnicity from a
Canadian context
DE Clarke, PhD (1-4); A Colantonio, PhD (3-6); AE Rhodes, PhD (2,3,7-9); M Escobar, PhD (3)
The current study provides a critical review of Canadian studies on ethnicity and mental
health with respect to the definition, conceptualization and operationalization of ethnicity.
It provides a discussion on the methodological issues related to these factors and their
implications to guide future research and enable comparability of results across studies.
Sociological Abstracts, PsycINFO, MEDLINE and CINAHL were used to identify relevant
Canadian articles published between January 1980 and December 2004. The review
highlights a number of key issues for future researchers to consider such as the need for:
1) clear rationales as to why ethnicity is important to their outcome of interest; 2) clarity
on the definition of ethnicity, which affects its conceptualization and operationalization;
3) a theoretically driven conceptualization of ethnicity, which should be related to the
research question of interest; and 4) clear rationales for the decisions made regarding the
data source used, the operationalization of ethnicity, and the ethnic categories included
in their studies.
Key words: ethnicity, ethnic origin, culture, race, mental health
Canada’s immigrant population originates
from all over the globe with increasing
numbers from Africa, Asia, the Caribbean,
and the Middle East (i.e. visible minorities).1 Its culturally diverse Aboriginal
population adds to the ethnic mix.2 Canada
prides itself on being a multicultural society
by acknowledging the right of every person
to identify with his/her cultural background
while partaking in the Canadian way of
life.3-6 This was advocated in its 1971
legislated policy on multiculturalism, which
emphasized fair treatment of everyone,
regardless of race, colour or ethnicity;
particularly in terms of educational and
occupational opportunities.3-6 Evidence of
systemic inequalities, including access to
educational and employment opportunities,
housing, health and mental health care, are
still evident across ethnic groups,7,8 which
can impact the mental health of the population.2,9 For instance, Aboriginal peoples continue to have poorer mental health compared
to the general population and, along with
visible minorities, have ongoing difficulties
accessing culturally sensitive mental health
care.9 Better understanding of the relationship
between ethnicity and mental health is
compelling and highly relevant for policymakers and mental health practitioners in
the Canadian context.
Canadian literature on ethnicity and
mental health is quite limited despite its
relevance. Much of the existing studies in
this area were conducted in the United
States (US)10-15 and the United Kingdom
(UK),16-19 which has uncertain applicability
to Canada. The political, social and
economic ramifications associated with
ethnicity or ethnic identity, likely differ
from country to country. Differences in
the countries’ ethnic compositions, histories of immigration policies and racism/
slavery and ethnic or racial categorizations
hinder cross-country comparisons. The
history of slavery and segregation has a
great deal of meaning for ethno-racial
groupings in the US.2 Furthermore, the US
is seen as a society that assimilates immigrants by the “melting pot” phenomenon,
in which the immigrants are expected to
adapt to the American way of life rather
than retaining their culture.3 Canada has
not had the same history of racial
segregation and is viewed as a mosaic in
which immigrants are encouraged to
retain their unique cultural background
while partaking in the greater Canadian
society.3,6,20 Also, Latin Americans comprise a larger component of the US
immigrant population than Canada’s.2,3
The time of arrival and ethnic composition
of the immigrant population in the UK and
Canada also differ. Up to 1962, citizens of
previously colonized countries such as
Jamaica and India (countries that remained
within the Commonwealth) were granted
open access and actively recruited for
Author References
1 Department of Mental Health, Johns Hopkins Bloomberg School of Public Health, Baltimore, Maryland
2 Department of Psychiatry, University of Toronto, Toronto, Ontario
3 Graduate Department of Public Health Sciences, University of Toronto
4 Toronto Rehabilitation Institute, Toronto, Ontario
5 Department of Occupational Science and Occupational Therapy, University of Toronto
6 Department of Rehabilitation Sciences, University of Toronto
7 Suicide Studies Unit, St. Michael’s Hospital, Toronto, Ontario
8 Inner City Health Research Units, St. Michael’s Hospital
9 Institute for Clinical Evaluative Sciences, Sunnybrook & Women’s College Hospital, Toronto, Ontario
Correspondence: Diana E Clarke, Department of Mental Health, Johns Hopkins Bloomberg School of Public Health, Room 802, 624 N Broadway, Baltimore,
MD 21205, Tel : (410) 955-0416, Email: [email protected]
Vol 28, No 4, 2008
Chronic Diseases in Canada
immigration to the UK.21 During this time,
Canada’s immigrant population came
primarily from Europe, the UK and the
US.22,23 Post-1971, when the UK changed
its immigration act, the majority of its
immigrant population originated from
Europe and South Africa,21 while the
majority of Canada’s immigrants came
from Asia, Africa, the Caribbean, Latin
America and the Middle East.23-26 These
points indicate that results based on UK
and/or US data might not be applicable to
With respect to the operationalization of
ethnicity, variations exist across the
literature on ethnicity and mental health.
Lack of a clear definition of ethnicity and
ongoing debate regarding how the variable
should be conceptualized might account
for this.27,28 Other factors such as data
source used, feasibility, time period of data
collection, region of study and sample size
also affect the operationalization of the
variable. To date, there is no seminal
Canadian paper that discusses the methodological issues related to the definition,
conceptualization and operationalization
of ethnicity and their implications, to guide
future research and to enable comparability
across studies. This paper provides a
critical review of the original empirical
Canadian studies on ethnicity and mental
health with emphasis on these issues. A
general overview of the definition and
conceptualization of ethnicity is first provided to guide the review process.
Quantitative studies involving populationbased data with more than one ethnic
group are highlighted.
The bibliographic databases Sociological
Abstracts, PsycINFO, MEDLINE and
CINAHL were used to identify relevant
original empirical studies that were
quantitative in nature for this review with
the application of Group 1 and 2 search
terms outlined in Figure 1. Studies were
included if they: 1) concerned ethnicity
and mental health; 2) involved data on
Chronic Diseases in Canada
Canadians; 3) involved more than one
ethnic group; 4) were published between
January 1980 and December 2004; and 5)
were quantitative in nature.
General review articles and chapters on the
theories and definition of ethnicity, not
specific to Canada, were used to provide an
overview on the definition and conceptualization of ethnicity. As well, they guided
the critique of the identified studies on
ethnicity and mental health reviewed
herein. Articles on ethnicity and immigration
policies specific to Canada were used to
provide historical context to facilitate
understanding of the ethnic composition of
the Canadian population over time and its
impact on the studies reviewed.
After all duplicates were removed, 49
empirical quantitative studies that involved
two or more ethnic groups were identified
and included in this review (Figure 1) and
summarized in Appendix 1. Twenty-seven
studies2,5,29-53 were based on population or
community surveys with secondary data
analysis, 12 were smaller studies using nonclinical samples54-65 and 10 were smaller
studies based on clinical/specialized
samples.66-75 The year and region of
publication, sample size, ethnic groups
included and critique regarding the
definition of ethnicity are outlined in
Appendix 1.
1. Definition
The lack of consensus on the definition of
ethnicity20,76-80 was reflected in the 49
included Canadian studies on ethnicity and
mental health,2,5,29-75 which failed to
explicitly define the variable. There are two
major perspectives on how ethnicity
emerges: the primordialist and constructivist
views.20,76,81 Traditional primordialists view
ethnicity as “an ascribed status, given at
birth, that is more or less fixed and
permanent.”20,76,81-85 Accordingly, the individual’s identity includes the biological,
cultural, political and economic conditions
of the group into which s/he is born, be it
dominant over, or dominated by, other
groups.20,78,81,84 A softer primordialist view
stresses the social and non-biological basis
of ethnicity, acknowledging that ethnic
identity can also be socially constructed
and based on “the circumstances at hand”
thereby being “situational not biological”
and “flexible not fixed.”81 The constructivists
view ethnicity as “a social construction
with ecological and social factors being its
key determinants.”12,24-26 Kaufman85stated
that distinction between the two perspectives
is artificial, so the definition of ethnicity
should include a synthesis of both. Hence,
the underlying theme is that ethnicity
involves sharing of a common culture,
which may be based on a combination of
factors such as language, religion, national
identity, customs, social and/or political
position within a country’s social system.81
2. Conceptualization
There are two major perspectives on how
ethnicity is conceptualized: 1) ethnic
identity which refers to self-identification
with particular cultural group/s,20,76,80,81 or
2) ethnic origin which refers to classification
due to the ethnic or cultural groups to
which the individual’s ancestors belong.20
Specific cultural traits such as language,
surnames, or region of birth were used as
proxies for ethnic origin in some studies.
From a Canadian perspective, the country’s
policy on multiculturalism, which has
guided government policies since 1971 and
advocated for individuals to retain their
ethno-cultural background,8 dictates that
ethnicity comprised “ethno-cultural particularism” and adherence to Canadian
“values.”76 This is reflected in the conceptualization of ethnicity across national
surveys, which forces participants to
identify with their ancestral background
regardless of time in Canada. As well, this
policy accounts for the designation of all
non-Aboriginal persons who originate
Vol 28, No 4, 2008
Vol 28, No 4, 2008
Chronic Diseases in Canada
Results of the search strategies used across bibliographic databases to identify Canadian literature on ethnicity and mental health
from Africa, Asia, the Caribbean, Latin
America and the Middle East, “who are
non-Caucasian in race or non-white in
colour” as visible minorities.1,3,6
Conceptualization based on Ethnic
Ethnic identity can be internal or external.20,78
Internal ethnic identity refers to the
individual’s self-identification with specific
ethnic group/s.78,86,87 This conceptualization
reflects the individual’s level of adherence
to and identification with the values,
customs and ideologies of a particular
culture,87 which can be internally or
externally driven.78,86-89 Individuals may
choose to identify with particular ethnic
group/s because of their belief in the
customs, norms and ideologies and the
meaning such identification gives to their
life (i.e. internally driven). However,
circumstances in the individuals’ external
environments, such as discrimination based
on ethnicity and the feeling of being
different from the ethnic majority, may
cause some to self-identify with particular
ethnic groups because of the sense of
solidarity and belonging they may garner
from such groups (i.e. externally
Since external ethnic identity relates to
categorization by others based on the
person’s nationality or ethnic origin and
does not reflect the individual’s adherence
to and/or identification with specific
culture/s,15 it is discussed in the context of
ethnic origin in the ensuing section. Such
external assignment of ethnicities to
individuals by the dominant society might
conflict with their subscription and
adherence to particular ethnic groups and
is potentially distressing. This was
illustrated by Mahtani6 in her interview
with Canadian-born visible minority
mixed-racial women who often feel forced
by the ethnic majority to identify their
ancestral background when asked about
their ethnic identity (i.e. by the question,
“but where are you from?”).
Chronic Diseases in Canada
A major weakness with self-identification
of ethnicity is that it is based on the
individual’s perception, which may change
from situation to situation and over time
and likely produce different results
depending on the time and research
question being investigated.27,28 This
possibly accounted for the lack of
conceptualization of ethnicity based on
ethnic identity across studies, with the
exception of the respective studies of De
Wit et al.,40 De Wit and Beneteau,41,42 and
Feldman et al.43 De Wit et al.40-42 conducted
secondary data analysis on the 1990
Ontario Health Survey in which respondents
were asked to select up to four ethnic
categories that matched their “ethnic or
cultural identity.” De Wit and Beneteau41,42
further utilized information on the primary
language that respondents used at home to
help “better identify ethnic groups.” If
respondents identified French or Frenchplus other ethnic groups but did not select
French as their primary language they
were reclassified as Anglophones41,42
indicating externally driven classification
of ethnic identity. This externally driven
classification of ethnic identity might have
been in conflict with the individual’s true
ethnic identity. If level of adherence to
particular cultural values, customs and
ideologies are important in the individual’s
perception of, and subsequent likelihood
to acknowledge, discuss or report a mental
health condition, and this is a key issue
being investigated, then the use of ethnic
identity would provide more appropriate
information. Feldman et al.43 used the
item, “Please write down the term that
best describes the ethnic character of your
everyday home environment,” which
implies an ethnic identity conceptualization
of the variable.
Conceptualization based on Ethnic
Ethnic origin refers to categorization of
individuals into ethnic group/s based on
ancestral origins.78 The question “To which
ethnic or cultural group/s do your ancestors
belong?” has been used to capture ethnic
origin.90,91,84,95 This method of conceptualization is reportedly more objective and
produces well-defined ethnic groups that
are representative of the population, but
fails to capture level of adherence with
cultural ideologies or customs.27,92,93
Conceptualization of ethnicity based on
ethnic origin was observed in some
Canadian studies on ethnicity and mental
health, particularly those that used national
or regional survey data,2,5,31,32,35,46-48,52
including the National Population Health
Survey (NPHS),90 Canadian Community
Longitudinal Survey of Children and Youth
(NLSCY),94 National Survey of Giving,
Volunteering and Participating (NSGVP)95
and the 1971 Aging in Manitoba (AIM)
survey.31 Each survey inquired about the
ethnic or cultural group/s of the individuals’
ancestors and provided a list of ethnic
categories to enable multiple choices.90,91,94,95
This acknowledged the existence of multiethnic groups, which are typical of
Canada,8 and improved the reliability of
the groups identified.27 Reflective of the
emphasis of the country’s multiculturalism
policy, survey respondents who identified
themselves as “Canadian” were required
to select their ancestral origin even if they
were Canadian-born.
Specific Cultural Traits as Proxies for
Ethnic Origin
One might assume that studies utilizing
data from surveys in which ethnicity was
based on ethnic origin would conceptualize
ethnicity accordingly; however, this was
not the case. Ali5 and Ma47 used country of
birth in thinking about ethnicity despite
utilization of the CCHS91 and the NLSCY94
data respectively, both with specific
ethnicity questions. The rationale for using
country of birth rather than responses to
the ethnicity questions was not provided
but likely due to the use of public-use data.
Public-use versions of Canadian national
surveys are readily available to researchers
Vol 28, No 4, 2008
but are stripped of unique identifying
information and lack the detailed information on ethnicity and culture available in
the restricted-use versions.
Penning32 utilized the Social Change in
Canada Survey (SCCS) data, in which
country of birth was used to determine
ethnicity. Ascertainment of ethnicity based
on country of birth in the study pinpoints
an inherent limitation with conducting
secondary data analyses where one has to
make do with whatever variables are
available. With this in mind, it is still
important to mention that with increasing
global industrialization and migration it
has become inappropriate to use country/
region of birth as the sole indicator of
ethnicity. Individuals might have been
born in one country to parents from
another country, spent only a few years in
that country and returned with their
parents to the countries of their ancestral
background thus choosing to identify with
their ancestral culture. Therefore, country/
region of birth, also utilized by Ali,5 Barnes
et al.,35 Ma,47 Cohen and MacLean,51
Rousseau et al.53,61 and Aubert et al.,65
would inaccurately classify such individuals. Most countries are comprised of
many ethnic groups and most ethnic
groups can be found to have many different
countries of birth. Country of birth as a
proxy for ethnic origins fails to capture
such complexities and likely misclassifies
some people. This does not negate the use
of country/region of birth as a proxy for
ethnicity but it underlines the need for
detailed discussions of the inherent limitations of its use and implications for the
interpretation of subsequent results.
Language and surname were proxies for
ethnic origin in some studies37,54-59 but are
susceptible to misclassification of some
individuals. The utilization of surnames to
identify ethnic groups by Dion and
Giordano57 in their investigation of ethnic
difference in depression in university
students in Toronto, failed to correctly
classify non-white/non-Caucasian individuals with non-ethnic last names. If
specific ethnic groups were more likely to
have individuals with non-ethnic surnames, for example Blacks from the
Caribbean and South Asians of Christian
Vol 28, No 4, 2008
descent,57 and these individuals were more
or less likely to have depression, their
exclusion would have biased the results
obtained. Bagley58 also used surnames and
language, but in combination with an
unspecified “ethnicity” question to conceptualize ethnicity. Surnames were
initially used to identify Chinese elderly
persons living in Calgary, Alberta, with
follow-up phone calls by the study recruiters
who identified eligible Cantonese-speaking
individuals and inquired about their
ethnicity. This combination of factors
overcame the shortcomings of using
surnames alone.
Walter37 and Dion59 used language as a
proxy for ethnicity, which might have
accurately classified some ethnic groups
but not others (e.g. Chinese for individuals
who specified Chinese dialect). The
English-speaking population is ethnically
heterogeneous, with English being the
primary language for many countries, so
treating the group as homogenous is
erroneous.59 Also, although French and
Spanish are primary languages for France
and Spain respectively, they are also the
primary languages for some countries in
the Caribbean and Africa. The assumption
that all individuals who endorsed these
languages are of European descent would
be inaccurate.59 Different ethnic groups
may speak the same language or there may
be language differences within ethnic
groups, which would not be captured with
language as the sole proxy for ethnic
scale. Ethnicity refers to mutual cultural
characteristics such as religion, language,
customs, and ancestry,20,76,78 but race refers
to common physical characteristics.28,65,78,86101
Winker,28 Williams77 and LaVeist97
stressed the importance of differentiating
race from ethnicity or culture. Race is a
poor proxy for ethnicity despite overlapping
features28,97 and is questionable as its sole
indicator in respective studies by Fry and
Grover54 and Devins et al.74
Wu et al.2 used ethnic origin and racial
background to create ethno-racial groups.
This enabled the identification of visible
minority groups2 and acknowledged the
argument that power and status differences
also exist across racial groups.77 This is
important since “the dominant or minority
status of the group mirrors its position
within the stratification system of the
larger society,”101 which in turn affects
access to social, political, and economic
resources.3,4,101 The authors2 explained that
their ethno-racial groups reflected the
social stratification of the Canadian society
(i.e. the vertical mosaic),2-4,7,8 which
provided a context that facilitated interpretation of results obtained. The vertical
mosaic refers to the hypothesis that ethnic
groups are differentially integrated in the
larger Canadian society based on histories
of immigration policies that were linked to
changing industrial and employment
demands over time.2-4,7,8 This differential
integration into the Canadian society
affected the groups’ socioeconomic
status,2-4 which is significantly associated
with mental health outcomes.
Ethnicity, Race or Ethnicity plus Race
In examining the studies by Weekes, et
al.71 and Cohen and McLean,51 respectively,
it became apparent that categorization was
based on race, despite the use of the term
“ethnicity”. Also, the researchers51,71
implied that race and culture were the
same concepts despite numerous theoretical articles indicating otherwise.28,76-79,85-99
To illustrate, Cohen and McLean51 utilized
data from the 1999 General Social Survey
(GSS) in which respondents were asked
about their “cultural or racial backgrounds”96 and Weekes and colleagues71
expressed interest in examining the
“cultural sensitivity” of their outcome
The identified studies appeared to have
conceptualized ethnicity on the basis of
ethnic origin by using ethnicity questions,
cultural traits as proxies for ethnic origin,
or a combination of ethnic origin and racial
background. For some studies, although
the term “ethnicity” was used it was
unclear how it was conceptualized and ascertained.30,33,34,54,55,70,75 The intrinsic weaknesses of using proxy measures such as
country of birth, language or surnames
underline the need for ethnicity to be
ascertained using rigorously tested and
validated questions. Questions aimed at
Chronic Diseases in Canada
capturing ethnic identity or ethnic origin
are unable to tell which aspect of ethnicity
the individual brings forward in response
during an interview and whether this
might be affected by the interviewer and/
or by the specific question being asked.76
This highlights the need for clear and
theoretically driven rationales for studying
ethnicity across mental health outcomes.
3. Operationalization
The operationalization of ethnicity is
important for interpretation of study
results and enabling cross-study comparison. However, variations existed in the
operationalization of ethnicity across
studies. Even studies that utilized the
same data sources and conceptualized
ethnicity similarly operationalized the
variable differently. Factors that influenced
this included diversity of the study
population, time period and region of data
collection, sample size restrictions and
purpose of the study. The effects of these
issues are discussed in the ensuing
Time Period of Data Collection and
Time period of data collection and study
influenced the operationalization of
ethnicity across studies. Time period of
data collection is related to changes in the
immigration policy over time, which were
influenced by economic changes that
dictated labour shortages and the need for
immigrants to fill specific employment
immigration policy was based on national
origin and gave preference to immigrants
from Britain, Europe and the US to fill
Hence, immigrants from these regions
represented approximately 95% of
Canada’s immigrant population up to
198622,25 and studies using data collected
at this time had little or no representation
of other ethnic groups.29,31,32,35,66,67
The 1967 change to a universal point
system was the force behind the increased
ethnic diversity of the Canadian population,
with European immigrants comprising
Chronic Diseases in Canada
57% of immigrants in 1970 but only 20%
in 1996.22,105 Changes in the ethnic
composition of the immigrant population
occurred gradually with small numbers
from Africa, Asia, the Caribbean, Latin
America, and the Middle East initially
significantly over time.22,24,25,102,105,106 This
was reflected in the 3 to 5% representation
of these immigrants in Canada up to the
1980s compared to their 73% representation
between 1996 and 2001.107 Variations in the
proportions of these immigrant groups
relative to each other and to immigrants
from Britain, the US and Europe were
reflected in studies conducted with
ethnicity data collected over the years.
Earlier studies had little or no visible
minority groups29,31,35,66,67 but their numbers
increased in later studies.2,5,35,38,43,46-49,53,61-63
Whether these groups were kept as
separate categories depended on the
outcome of interest and the nature of the
study (i.e. secondary data analysis using
public-use data that lacked detailed
ethnicity information). In the investigation
of rare outcomes, such as some mental
health conditions, small sample sizes for
specific visible minority ethnic groups
resulted in them being collapsed into a
single category, which likely failed to
capture the individuals’ perception of their
ethnic identity. This also limited the
interpretation of the results of such
Region of Data Collection and Study
Variations in settlement patterns of
different ethnic groups in Canada resulted
in different ethnic compositions across
geographic regions. This affected the
operationalization of ethnicity as it relates
to the region of data collection and study.
Settlement patterns of immigrant groups
have always been related to the location of
employment opportunities, which were
dictated by economic changes over
time.78,102 Earlier immigrants from Britain,
Europe and the US106 came to Canada in
response to the rapid expansion of the
Canadian West with integration of the
region with the domestic and world
markets and international demand for
immigrants settled in rural areas of Western
Canada to fill the demands for laborers
and farmers.66,102 Studies conducted in
Western Canada (e.g. Manitoba,31,35,70
Alberta66 and the Northwest Territories67)
were likely to have significant representation
of British, European, North American and
Aboriginal ethnic groups but almost no
representation of visible minority groups.
Later immigrants from Asia, Africa, the
Caribbean, Latin America and the Middle
East were more likely to settle in Montreal,
Toronto and Vancouver102,108 because of
their perception of better employment
opportunities that corresponded to their
skills and educational background.102
Supportive data showed that only 58% of
recent immigrants settled in these areas in
1981 but the percentage rose to 78% by
2001.102 Recent studies based on data that
includes these three regions had varying
ethnicities. 2,5,32,43,44,46-51,53-55,57-63,65,72,74,75
Reliance on public-use data prevented
examination of very specific ethnic
categories because of the inaccessibility of
such detailed information.
The Purpose of the Research Study
As seen in the respective studies of Barnes
et al.,35 Lavallee and Bourgault, 45 Seltzer
and Langford,67 and others,29,30,34,37,50,51,58,61,64
there is flexibility in whether
broad or specific categories are used when
operationalizing ethnicity in descriptive or
enumerative studies as long as they reflect
the ethnic composition of the population
under investigation.20 However, the use of
specific ethnic categories is more
informative.20 For analytic studies, operationalization of ethnicity needs to be
theoretically driven to provide a framework
for the analyses and interpretation of the
results.20 Operationalization of ethnicity in
the analytic studies reviewed in this
paper2,5,31,32,36-43,45-49,52,58,59,75 appeared to be
based on the social stratification system in
Canada, though not always explicitly
Studies interested in examining the mental
health status or outcome in specific ethnic
groups tended to be clear in their selection
and ascertainment of those groups but
often aggregated all other ethnic
Vol 28, No 4, 2008
groups.30,31,33,34,36,38-42,44,45,51,54,61,67-71 For example,
Fry and Grover54 were interested in
examining mental health outcomes in
Asians compared to Caucasians and
therefore only included these groups.
Liban and Smart,30 Tonkin,34 Beiser et al.,39
Borzecki et al.,68 Tcheng-Laroche and
Prince33 and Dewit and Beneteau,35,42 in
their studies, were mainly interested in the
mental health of Native Indians30,34,39,68 and
therefore selected these groups specifically
while aggregating all other ethnicities in
the comparison groups.
Havens and Chappell, in investigating the
effects of age, sex and ethnicity on mental
health in Manitoba, included North
American, British, French, Polish/Russian/
Ukrainian and ‘Other European’ ethnic
groups.31 These groups were reflective of
the time period and region of study. North
Americans, British and French were among
the earlier immigrants admitted to Canada
to fill occupational demands based on the
country’s immigration policy at that time.
The ‘Other European’ group, which
included immigrants from Germany,
Norway, Denmark, Sweden, Iceland, the
Netherlands and Belgium,31 was granted
entry into Canada over the Polish/Russian/
Ukrainian when the demand for more
immigrants arose.22,106 After World War II
and prior to the change to the universal
immigrants from Poland, Russia, the
Ukraine and other Eastern European
countries were admitted to Canada to help
survivors of the Nazi Holocaust and to fill
specific occupational roles.22,106 Therefore,
the ethnic groups included in the study31
were incorporated into the social hierarchy
of Canada at different levels and different
points in time, emphasizing the segregation
of the Canadian society along ethnic
lines.22,104 Aboriginal peoples also contributed to the ethnic diversity of Manitoba’s
population, specifically its rural regions,109
but were excluded from the study to
“reduce cultural bias”.31 This statement
implied that cultural aspects of ethnicity
were important in mental health
functioning and needed to be kept constant
Vol 28, No 4, 2008
across groups. It further implied that either
the ethnic categories included in the study
were culturally similar to each other yet
different from Aboriginal peoples or that
the individual ethnic groups that comprised
each broader ethnic category were
culturally similar but the individual ethnic
groups that comprised the broader
Aboriginal category were too culturally
dissimilar to be combined.
Penning32 included similar ethnic groups
as Havens and Chappell31 in examining the
same hypothesis with psychosocial wellbeing as the outcome and using a nationally
representative sample of Canadians aged
30 and over. The groups also mirrored the
social stratification of the Canadian society.
Havens and Chappell,31 but not Penning,32
found ethnic differences in mental health
functioning with a “triple jeopardy” effect
of age, sex and ethnicity. Methodological
issues related to differences in the definition
and ascertainment of the mental health
outcomes, the conceptualization (e.g. use
of country of birth32 versus a specific
ethnicity question31) and operationalization
of ethnicity, the ethnic and age composition
of the study populations, and the regions
of study likely played a role in the
discrepancy observed.
Since Penning32 used country of birth to
ascertain ethnicity and mentioned no
excluded ethnic groups, it is assumed that
Aboriginal peoples and any Canadian-born
visible minority individuals, though
minimal, were included into the Canadian
group. This would make the Canadian
group heterogeneous and different from
those included in the North American
group in Havens and Chappell.31 Also,
since Penning32 utilized data from a
national survey, the ethnic category ‘Other’
probably comprised participants who were
members of Canada’s visible minority
groups whereas these groups were
excluded by Haven and Chappell31 because
of small sample size. Penning,32 separated
Canadians and Americans while combining
immigrants from France, Germany,
Norway, Denmark, Sweden, Iceland, the
Netherlands and Belgium into a Northern
European group. These factors affected the
ability to compare the results of the two
Understanding the mental health of
Canada’s immigrant population in terms
of depression and alcohol dependency was
the aim of the study by Ali.5 Ethnic
differences were examined as a secondary
objective in the immigrant group only.5
Using region of birth, the ethnic groups
examined (i.e. US/Mexico, South and
Central America and the Caribbean,
Europe, Africa and Asia) mirrored the
variation in time of entry into Canada by
the different immigrant groups and their
differential incorporation in the social
hierarchy of the country. This provided a
framework for analyses and interpretation
of the results. Disaggregation of only the
immigrant group based on ethnicity
implied that either the investigation of
ethnic variation in mental health was only
important for immigrant groups and not
Canadian-born or all Canadian-born
individuals had similar experiences that
potentially affected their mental health.
Since the Canadian-born group probably
had many first generation Canadians who
likely had similar experiences as their
immigrant parents, these assumptions
were likely inaccurate. Given the report of
poor mental health in Canada’s Aboriginal
population,8 combining them into the
Canadian-born group might not be
appropriate. The use of region of birth has
numerous drawbacks that could have
potentially biased the results obtained but
was likely used because of the lack of
ethnicity-related information in the publicuse version of the CCHS Cycle 1.1 dataset,
which is assumed to be the data source
used by Ali.5
In examining behavioural and emotional
problems in immigrant versus nonimmigrant children in Canada, Ma47
considered ethnicity relevant but only for
the immigrant group. Immigrant children
were disaggregated into ethnic groups
based on region of birth, including those
from the US, Europe, Asia and other
regions47 possibly due to the use of the
Chronic Diseases in Canada
public-use version of the NLSCY.94 These
ethnic categories differed from Ali5 despite
the same method of conceptualizing
ethnicity. The studies tested different
hypotheses, and were interested in different
populations and different outcomes, which
affected the ability to compare results
across studies.
Beiser et al.46 utilized NLSCY94 data to
examine familial poverty and emotional
and behavioural problems in immigrant
children versus Canadian-born and included
ethnicity as a control variable. Unlike Ma,47
ethnicity was examined in the entire
sample46 thereby eliminating the ambiguities
observed in interpreting the results of the
former studies. Although the study utilized
a data source in which various ethnic
categories were available, ethnicity was
operationalized by using four broad
categories including White/European,
Asian, Black and Other.46 This implied an
interest in, or expectation of, an effect based
on visible minority status. The NLSCY
included Aboriginal peoples living offreserves,94 and since there was no explicit
indication of their exclusion, it is assumed
that they were either combined in the
‘Other’ category or included in the ‘White/
European’ category. The appropriateness of
the inclusion of Aboriginal children into
either category is questionable given major
cultural and social differences.
The NSGVP95 was utilized by Mata48 to
investigate satisfaction with life in Canada
across ethnic groups and to test the
hypothesis that any variations could be
explained by ethnic differences in
socioeconomic status (SES). Nineteen
mutually exclusive ethnic categories were
derived for the study with sample size
limitations being integral in the collapsing
of different ethnic groups into broader
categories (e.g. ‘Black’, ‘South Asian’,
‘Italian’ and ‘Portuguese’ ethnic groups
collapsed into ‘Black/South Asian’ and
‘Italian/Portuguese’).48 The distinction
between certain other ethnic categories
was unclear. For instance, the categories
‘European’, ‘Italian/Portuguese’, ‘German’,
‘French’ ‘Ukrainian’, ‘Polish’, etc., all of
definite European origin were used without
the provision of a clear rationale. This
combination of specific and broad ethnic
Chronic Diseases in Canada
categories without rationale limited the
interpretation and comparability of the
results. It would have been beneficial if the
ethnic categories were guided by a
theoretical framework so as to provide
more meaning to the results and decrease
the observed ambiguity.20
Wu et al.2 tested the hypothesis that ethnic
differences in depression could be explained
by variations in SES and access to social
support resources (1996 NPHS90 data). The
included ‘ethno-racial’ groups were ‘East/
Southeast Asian’, ‘Chinese’, ‘South Asian’,
‘Aboriginal’, ‘Black’, ‘Arabic/West Asian’,
‘Latin American’, ‘Jewish’, ‘French’,
‘English’, ‘Other Whites’, and ‘mixed
racial’, were “representative of the social
stratification of the Canadian society” and
reflected differences in SES and access to
social support resources.2 The partial
disaggregation of the Asian group into
‘East/Southeast Asian’ and ‘Chinese’ was
likely due to the overwhelming presence of
‘Chinese’ (approximately 56%) in the
group, which would have obscured the
interpretation of any association with
depression. A clear rationale for this partial
disaggregation was not provided but left to
the readers to assume. Lack of sufficient
sample size prevented the disaggregation of
other ethnic groups including Blacks and
South Asians.2
Wang and El-Guebaly52 used a ‘White’
operationalizing ethnicity despite the
utilization of the 1996 NPHS90 data. This
was likely due to the use of the public-use
version of the survey data, though not
explicitly stated. This dichotomy, although
handy in separating visible minorities from
the ethnic majority, has been criticized for
its inadequacy in giving a clear view of
ethnic differences in mental health.2,27,104
Refinement is needed to account for
cultural and/or ecological differences
within such broad ethnic groups,104 which
might include cultural biases in reporting
mental health issues. Although the specific
ethnicities within the broad ‘non-White’
group are distinct from each other in
languages, histories, customs and social
mobility,2 discrimination in educational
and occupational opportunities based on
visible minority status2,22,103,109 lends some
support to combining the groups into this
category. The use of such a broad ethnic
category is appropriate if the researchers
were asserting that it is the common
experience of being ‘non-White’ that
impacts mental health above and beyond
the effect of cultural differences in
individuals’ perceptions of, and attitudes
towards, mental health problems, but this
was not stated.98
Dion and Giordano included ‘AngloCeltics’,
European’, ‘East European’, ‘East Asian’,
and ‘South Asian’ in their study.57 These
ethnic groups were said to parallel the
ethnic composition of the University of
Toronto population in 198857 and reflected
variations in immigration experiences,
entrance status and adaptation to life in
Canada with respect to parental conflicts,
and the likelihood to perceive economic
discrimination.57 According to the study,
‘South Asians’ and ‘Southern Europeans’
were more likely to suffer depression than
Europeans’ because of higher likelihood of
parental conflict and perceived economic
discrimination.57 Surnames were used to
misclassified ‘Blacks’ and ‘South Asians’
of Christian background. Since ‘South
Asian’ was an ethnic group of interest in
the study,57 this misclassification would
have likely affected the results obtained by
diluting the measure of effect and affected
the generalizability of the results to all
‘South Asians’ in Toronto. The researchers57
explained that since an earlier study
including 300 students at the same
university found less than 3% indicating
West Indian heritage, the potential
misclassification by surnames created little
or no bias to their results. This inaccurately
implied that among ‘South Asians’ only
those of West Indian heritage had nonethnic last names. India’s Christian
population accounted for about 2.3% of its
total population according to its 1991 and
2001 censuses.110
In summary, the operationalization of
ethnicity in the existing Canadian studies
differed depending on how the variable
Vol 28, No 4, 2008
was conceptualized, on which data sources
were used, on time period and region of
data collection, the purpose of the study,
and whether the study involved secondary
data analysis using public-use versions of
national surveys. Some studies tried to
disaggregate ethnic categories but were
faced with sample size limitations that led
to collapsing of distinct ethnic groups
while keeping others disaggregated, with
no clear rationale. Sample size limitations
were found across national survey datasets,
particularly for the visible minority ethnic
groups, thus underlining the need for
future population surveys aimed at
providing information on the health and/
or mental health of the Canadian population
to over-sample these groups. This would
enable the examination of mental health
differences across visible minority groups,
which is important for the planning of
mental health programs to serve Canada’s
ethnically diverse population.
conceptualization and operationalization
of ethnicity across Canadian studies on
ethnicity and mental health uncovered a
number of key issues that are highlighted
a) Ethnicity is a complex and abstract
term for which a single and generally
accepted definition has not been
derived. However, the underlying theme
is that it involves the sharing of a
common culture. Across studies,
ethnicity was not defined a priori,
which was likely due to lack of
consensus on its definition. Therefore,
as suggested by reviewers from the
US28,77,79 and UK,27,92,93 clear rationales
as to why ethnicity is important to the
outcome of interest are necessary.
b) Examination of the relationship between
ethnicity and mental health needs to be
encouraged with better infrastructure
involving improved funding opportuni-
Vol 28, No 4, 2008
ties, given the relevance to the Canadian
context. Secondary data analysis is
valuable in such research efforts.
Therefore, making detailed information
on ethnicity more readily available to
researchers while maintaining survey
participants’ confidentiality is vital.
c) Some researchers seemed to have
utilized public-use versions of survey
data, which hindered their ability to
disaggregate ethnic groups due to lack
of such detailed information in these
files. Their use need to be stated
explicitly and their inherent limitations
d) Ethnicity can be conceptualized based
on ethnic origin, ethnic identity, or a
combination of ethnic origin and race.
Each method has strengths and weakness. Its conceptualization should be
theoretically driven and related to the
research question of interest. The
ethnicity based on ethnic origin,
whether through the use of a specific
questions or proxy measures such as
language, surname or country of birth.
Countries of birth, language and
surnames are prone to misclassification
of certain ethnic groups and biased
results if rate of misclassification
differed across outcome groups. This
does not negate their use as proxies for
ethnicity; however, there is a need for
researchers to clearly outline their
inherent limitations and the implications
for the results obtained.
e) Ethnicity based on ethnic origin can be
more stable if participants are given a
list of ethnic categories with the option
of choosing multiple categories.27,93 This
list should be based on preliminary
fieldwork to identify common ethnic
categories. An open-ended ‘other’ option
should be included so as not to restrict
the individuals’ choices. The list should
be appended to the research results or
made available upon request to enable
study replication and to enable readers
to ascertain the representativeness of the
ethnic categories to the general
f) Operationalization of ethnicity varied
from very broad to very specific ethnic
categories even if ethnicity was conceptualized similarly or the same data
source was used. These variations were
related to the lack of a clear definition
of ethnicity, differences in the time
period, region of study and data
collection, the purpose of the studies
and the utilization of public-use data.
Researchers are encouraged to provide
a clear outline of the decisions made
regarding the data source used (particularly in secondary data analysis), the
operationalization of ethnicity, and the
categories included in their studies.
This will facilitate the interpretation of
the results and attempts to replicate the
research findings.
g) Differences in the outcomes of interest,
how such outcomes were measured
and variability in the relevant variables
controlled for were additional factors
that affected the ability to compare
results across studies. Therefore,
researchers should clearly define their
outcomes of interest and summarize
the strengths and weaknesses of the
measures used to ascertain outcomes
and other relevant variables.
Diana E. Clarke is a Research Associate in
the Department of Mental Health at Johns
Hopkins Bloomberg School of Public
Health and Fellow in Population Health
Studies in the Department of Psychiatry at
University of Toronto. Dr. Clarke is
supported by a Canadian Institute for
Health Research Post-doctoral Fellowship
Award (Grant # 200602MFE-159564115967), the Toronto Rehabilitation
Institute Foundation and in part by the
Population Health Fellowship Award from
the Department of Psychiatry at the
Chronic Diseases in Canada
University of Toronto. This study was
funded in part by the Ministry of Health
and Long-term Care and the Ontario
Mental Health Foundation.
Statistics Canada. 2001 Census Visible
Minority and Population Group User Guide.
Wu Z, Noh S, Kaspar V, Schimmele CM.
Race, Ethnicity, and Depression in
Canadian Society. J Health Social Behav.
Esses VM, Gardner RC. Multiculturalism in
Canada: Context and Current Status.
Canadian Journal of Behavioral Science.
Li PS. Social Inclusion of Visible Minorities
and Newcomers: The Articulation of
“Race” and “Racial” Difference in Canadian
Society. Prepared for 2003 Social Inclusion
Conference. What Do We Know and Where
Do We Go: Building a Social Inclusion
Research Agenda. Ottawa, March 27-28,
2003. URL:
Ali J. Mental Health of Canada’s
Immigrants. Supplement of Health Reports.
Mahtani M. Interrogating the hyphennation: Canadian multicultural policy and
“mixed race” identities. In Hier SP and
Bolaria BS, eds. Identity and Belonging:
Rethinking race and ethnicity in Canadian
society. Toronto, Canada: Canadian
Scholars’ Press Inc, 2006;Chapter 2:31-42.
Porter J. The Vertical Mosaic: An Analysis
of Social Class and Power in Canada.
Toronto: University of Toronto Press,
Lian JZ, Matthews DR. Does the Vertical
Mosaic Still Exist? Ethnicity and Income in
Canada, 1991. Can Rev Sociol Anthropol.
Chronic Diseases in Canada
Kirmayer LJ, Minas H. The future of
cultural psychiatry: An international
perspective. Can J Psychiatry. 2000;45:438446.
10. Centers for Disease Control and Prevention.
Suicide among Black youths – United
States, 1980-1995. MMWR Morb Mortal
Wkly Rep. 1998;47:106-193.
19. McKenzie K, van Os J, Samele C, van Horn
E, Tattan T, Murrau R. Suicide and
attempted suicide among people of
Caribbean origin with psychosis living in
the UK. Br J Psychiatry. 2003;183:40-44.
20. Isajiw WW. Definitions
Ethnicity. 1974;1:111-124.
Shiang J. Does culture make a difference?
Racial/ethnic patterns of completed suicide
in San Francisco, CA 1987-1996 and clinical
implications. Suicide Life Threat Behav.
12. Joe S, & Kaplan MS. Suicide Among African
American Men. Suicide Life Threat Behav.
13. Oquendo MA, Ellis SP, Greenwald S,
Malone KM, Weissman MM, Mann JJ.
Ethnic and sex differences in suicide rates
relative to major depression in the United
States. Am J Psychiatry. 2001;158:16521658.
14. Price JH, Dake JA, Kuchareswki R. Assets
as predictors of suicide attempts in African
American inner-city youths. Am J Health
Behav. 2001;25:367-375.
15. O’Donnell L, O’Donnell C, Merrit Wardlaw
D, Stueve A. Risk and resiliency factors
influencing suicidality among African
American and Latino youth. Am J
Community Psychol. 2004;33:37-49.
16. Neeleman J, Mak V, Wessely S. Suicide by
age, ethnic group, coroners’ verdicts and
country of birth: A three year survey in
17. Neeleman J, Wessely S. Ethnic minority
suicide: A small area geographical study in
South London. Psychol Med. 1999;29:429436.
18. Bhugra D, Desai M, Baldwin D. Attempted
suicide in west London: Rates across ethnic
communities. Psychol Med, 1999;29:11251130.
Bell, BD. The Performance of Immigrants
in the United Kingdom: Evidence from the
GHS. The Economic Journal. 1997;107:333344.
22. Green AG, Green DA. The Economic Goals
of Canada’s Immigration Policy, Past and
University of British Columbia. 1996.
23. Li PS. Race and ethnicity. In Li PS (Ed)
Race and Ethnic Relations in Canada,
Second Edition. Oxford University Press,
Canada. 1999;3-20.
24. Beaujot R. Immigration and Canadian
Demographics: State of the Research.
Citizenship and Immigration Canada.
1998. URL:
25. Citizenship and Immigration Canada.
Citizenship and Immigration Statistics,
1966-1996. 2004. URL: http://epe.lac-bac.
26. Belanger A, Malenfant EC. Population
projections of visible minority groups,
Canada, provinces and regions 2001-2017.
Ottawa: Statistics Canada, Demography
Division. 2005. Cat. No. 91-541-XIE.
27. Aspinall PJ. Operationalising the collection
of ethnicity data in studies of the sociology
of health and illness. Sociol Health Illn.
28. Winker MA. Measuring Race and Ethnicity:
Why and How? JAMA. 2004;292:16121614.
Vol 28, No 4, 2008
29. Murphy HBM. European cultural offshoots
in the New World: Differences in their
mental hospitalization patterns. Part II:
German, Dutch, and Scandinavian
influences. Archiv fur Psychiatrie und
Nervenkrankheiten (Archives of Psychiatry
and Neurological Sciences). 1980;228:161174.
30. Liban CB, Smart RG. Drinking and drug
use among Ontario Indian students. Drug
Alcohol Depend. 1982;9:161-171.
Havens B, Chappell NL. Triple Jeopardy:
Age, Sex, and Etnicity. Canadian Ethnic
Studies. 1983;15:119-129.
32. Penning MJ. Multiple Jeopardy: Age, Sex,
and Ethnic Variations. Canadian Ethnic
Studies. 1983;15:81-105.
33. Tcheng-Laroche F, Prince R. Separated and
divorced women compared with married
controls: Selected life satisfaction, stress
and health indices from a community
survey. Soc Sci Med. 1983;17:95-105.
34. Tonkin RS. Suicide methods in British
Columbian adolescents. J Adolesc Health
Care. 1984;5:172-177.
35. Barnes GE, Currie RF, Segall A. Symptoms
of Depression in a Canadian Urban Sample.
Can J Psychiatry. 1988;33:386-392.
36. Sack WH, Beiser M, Phillips N, BakerBrown G. Co-morbid symptoms of
depression and conduct disorder in First
Nations children: Some findings from the
Flower of Two Soils Project. Cult Med
Psychiatry. 1993;16:471-486.
37. Walters, V. Stress, Anxiety and Depression:
Women’s Accounts of their Health
Problems. Soc Sci Med. 1993;36:393-402.
38. Beiser M, Cargo M, Woodbury MA. A
comparison of psychiatric disorder in
different cultures: Depressive typologies in
Southeast Asian refugees and resident
Canadians. Int J Methods Psychiatr Res.
39. Beiser M, Sack W, Manson SM, Redshirt R,
Dion R. Mental and the academic
performance of First Nations and majorityculture children. Am J Orthopsychiatry.
40. De Wit ML, Embree BG, De Wit DJ.
Determinants of the risk and timing of
alcohol and illicit drug use onset among
natives and non-natives: Similarities and
differences in family attachment. Soc Biol.
41. DeWit DJ, Beneteau B. Predictors of the
prevalence of alcohol use and related
problems among Francophones and
Anglophones in the province of Ontario,
Canada. Journal of the Study of Alcoholism.
42. DeWit DJ, Beneteau B. Predictors of the
prevalence of tobacco use among
Francophones and Anglophones in the
province of Ontario. Health Educ Res.
43. Feldman L, Harvey B, Holowaty P, Shortt
L. Alcohol use beliefs and behaviors among
high school students. J Adolesc Health.
44. Beiser M, Dion R, Gotowiec A. The
hyperactivity symptoms among Native and
Non-Native elementary school children. J
Abnorm Child Psychol. 2000;28:425-437.
45. Lavallee C, Bourgault C. The health of
Cree, Inuit and Southern Quebec Women:
Similarities and Differences. Can J Public
Health. 2000; 91:212-216.
46. Beiser M, Hou F, Hyman I, Tousignant M.
Poverty, Family Process, and the Mental
Health of Immigrant Children in Canada.
Am J Public Health. 2002;92:220-227.
47. Ma X. The First Ten Years in Canada: a
Multi-Level Assessment of Behavioral and
Emotional Problems of Immigrant Children.
Canadian Public Policy. 2002;28:395-418.
48. Mata F. A Look at Life Satisfaction and
Ethnicity in Canada. Canadian Ethnic
Studies. 2002;34: 51-63.
Vol 28, No 4, 2008
49. Wu Z, Hart R. Social and health factors
associated with support among elderly
immigrants in Canada. Research on Aging.
50. Blackstock C, Trocme N, Bennett M. Child
Aboriginal and Non-Aboriginal families in
Canada. Violence Against Women.
Cohen MM, MacLean H. Violence against
Canadian Women. BMC Women’s Health.
52. Wang JW, El-Guebaly N. Sociodemographic
Factors Associated with Comorbid Major
Dependence in the General Population.
Can J Psychiatry. 2004;49:37-44.
53. Rousseau C, Drapeau A. Premigration
exposure to political violence among
independent immigrants and its association
with emotional distress. J Nerv Ment Dis.
54. Fry PS, Grover SC. Cognitive Appraisals of
Life Stress and Depression in the Elderly: A
Cross-Cultural Comparison of Asians and
Caucasians. International Journal of
Psychology. 1982;17:437-454.
55. Dyal, James A; Chan, Carolina. Stress and
distress: A study of Hong Kong Chinese
and Euro-Canadian students. Journal of
Cross-Cultural Psychology. 1985;16(4):447466.
56. Blandford AA, Chappell NL. Subjective
well-being among Native and non-Native
elderly persons: Do differences exist? Can
J Aging. 1990;9(4):386-399.
57. Dion KL, Giordano C. Ethnicity and Sex as
Correlates of Depression Symptoms in a
Canadian University Sample. Int J Soc
Psychiatry. 1990;36:30-41.
58. Bagley CR. Mental health and social
adjustment of elderly Chinese immigrants
in Canada. Canada’s Mental Health.
Chronic Diseases in Canada
59. Dion KL. Ethnolinguistic Correlates of
Alexithymia: Toward a Cultural Perspective.
J Psychosom Res. 1996;41:531-539.
60. Heine SJ, Lehman DR. Culture, SelfDiscrepancies, and Self-Satisfaction. Pers
Soc Psychol Bull. 1999;25:915-925.
61. Rousseau C, Mekki-Berrada A, Moreau S.
Trauma and extended separation from
family among Latin American and African
62. Howard LA, Ahluwalia JS, Lin SK, Sellers
EM, Tyndale RF. CYP2E1*1D regulatory
polymorphism: association with alcohol
and nicotine dependence. Pharmacogenetics. 2003;13:321-328.
63. Tweed RG, White K, Lehman DR. Culture,
stress and coping: Internally- and
externally-targeted control strategies of
European Canadians, East Asian Canadians
and Japanese. Journal of Cross-Cultural
Psychology. 2004;35:652-668.
70. Norton GR, Rockman GE, Malan J, Cox BJ,
Hewitt PL. Panic attacks, chemical abuse
and suicidal ideation: A comparison of
71. Weekes JR, Morison SJ, Millson WA, Fettig
DM. A comparison of Native, Métis, and
Caucasian offender profiles on the MCMI.
Canadian Journal of Behavioural Science.
72. Pawliuk N, Grizenko N, Chan-Yip A,
Gantous P, Mathew J, Nguyen D.
Acculturation style and psychological
functioning in children of immigrants. Am
J Orthopsychiatry. 1996;66:111-121.
73. Zapf PA, Roesch R, Hart SD. An examination
of the relationship of homelessness to
mental disorder, criminal behaviour, and
health care in a pretrial jail population.
Can J Psychiatry. 1996;41:435-440.
64. Khanlou N. Influences on adolescent selfesteem
secondary schools. Public Health Nurs.
74. Devins GM, Edworthy SM, Aramis Lupus
State Models Research Group. Illness
intrusiveness explains race-related qualityof-life differences among women with
systemic lupus erythematosus. Lupus.
65. Aubert P, Daigle MS, Daigle JG. Cultural
Traits and Immigration: Hostility and
Suicidality in Chinese Canadian Students.
Trancultural psychiatry. 2004;41: 514-532.
75. Hodelet N. Psychosis and offending in
British Columbia: characteristics of a
secure hospital population. Crim Behav
Ment Health. 2001;11163-172.
66. Bland RC, Orn H. Schizophrenia:
Sociocultural Factors. Can J Psychiatry.
76. Troper H, Weinfeld M. Ethnicity, Politics
and Public Policy: Case studies in Canadian
Diversity. Toronto: University of Toronto
Press, 1999:3-25.
67. Seltzer A, Langford A. Forensic psychiatric
assessments in the Northwest Territories.
Can J Psychiatry. 1984;29:665-668.
68. Borzecki M, Wormith JS, Black WH. An
examination of differences between native
and non-native psychiatric offenders on
the MMPI. Canadian Journal of Behavioural
Sciences. 1988;20:287-301.
69. Chandrasena R, Beddage V, Fernando
MLD. Suicide among immigrant psychiatric
patients in Canada. Br J Psychiatry.
Chronic Diseases in Canada
77. Williams DR. Race and Health: Basic
Questions, Emerging Directions. Annual
Review of Epidemiology. 1997;7:322-333.
78. Isajiw WW. Understanding diversity:
Ethnicity and race in the Canadian Context.
Toronto: Thompson Educational Publishing
Inc., 1999:17-36.
79. Comstock RD, Castillo EM, Lindsay SP.
Four-Year Review of the Use of Race and
Ethnicity in Epidemiologic and Public
Health Research. Am J Epidemiol.
80. Trimble JE, Dickson R. Ethnic Identity. In
Fisher CB and Lerner RM, eds. Applied
developmental science: An encyclopedia
of research, policies, and programs.
Thousand Oaks: Sage. In press.
81. Allahar A. The social construction of
primordial identities. In Hier SP and Bolaria
BS, eds. Identity and Belonging: Rethinking
race and ethnicity in Canadian society.
Toronto: Canadian Scholars’ Press Inc.,
2006;Chapter 2:31-42.
82. Connor W. Ethnonationalism: The Quest
for Understanding. Princeton: Princeton
University Press, 1994:89-117.
83. Smith AD. The ethnic origins of nations.
New York: Basil Blackwell, 1986:1-20.
84. Van Den Berghe PL. Ethnicity and the
sociobiology debate. In Rex J and Mason,
D, eds. Theories of Race and Ethnic
Relations. Cambridge University Press,
85. Kaufman SJ. Ethnic fears and ethnic war in
Karabagh. Centre for Strategic and
International Studies. 1998. URL: www.
86. Cheung YW. Approaches to ethnicity:
Clearing roadblocks in the study of
ethnicity and substance abuse. International
Journal of Addictions. 1993;28(12):12091226.
87. Phinney J. Ethnic identity and acculturation.
In Chun K, Organista PB, and Marin G,
eds. Acculturation: Advances in theory,
measurement, and applied research.
Washington, DC: American Psychological
Association, 2003:63-81
88. Tajfel H. Social identity and intergroup
relations. Cambridge: Cambridge University
Press, 1982:207-240.
89. Weinreich P, Saunderson W, eds. Analyzing
identity: Cross-cultural, societal and
clinical contexts. New York: Routledge,
2003:7-110 & 115-170.
Vol 28, No 4, 2008
90. Statistics Canada. National Population
Health Survey, Cycle 2. 1996. URL: http://
n&db=IMDB&dbg=f&adm=8&dis=2 .
91. Statistics Canada. Canadian Community
Health Survey, Cycle 1.1. 2000. URL:
health/ .
92. Nazroo JY. Genetic, cultural or socioeconomic vulnerability? Explaining ethnic
inequalities in health. Sociol Health Illn.
93. Nazroo JY. The structuring of ethnic
inequalities in health: economic position,
racial discrimination and racism. Am J
Public Health. 2003;93(2):277-84.
94. Statistics Canada. National Longitudinal
Survey for Child and Youth, Cycle 1. 1994.
bg=f&adm=8&dis=2 .
95. Statistics Canada. National Survey of
Giving, Volunteering and Participating,
Cycle 1. 1997. URL: http://www.statcan.
&db=imdb&dbg=f&adm=8&dis=2 .
96. Statistics Canada. General Social Survey –
Victimization. 1999. URL: http://www.
Vol 28, No 4, 2008
97. LaVeist TA. Beyond Dummy Variables and
Sample Selection: What Health Services
Researchers Ought to Know about Race as
a variable. Health Serv Res. 1994;29:1-16.
104. Breton R, Roseborough H. Ethnic
differences in status. In Blishen BR et al.,
eds. Canadian Society. Toronto: Macmillan,
98. Yinger
Strength? Source of Conflict? New York:
State University of New York Press, 1994:167 & 167-198.
105. Thompson EN. Immigrant Occupational
Skill Outcomes and the Role of Region-OfOrigin-Specific Human Capital-September
2000. Ottawa: Human Resources Development Canada. 2000. Cat. No. W-00-8E.
99. Stolley PD. Race in Epidemiology. Int J
Health Serv. 1999;29:905-909.
100. Chaturvedi
epidemiological determinant-crudely racist
or crucially important? Int J Epidemiol.
101. James CE. Race, ethnicity and cultural
identity. In Hier SP and Bolaria BS, eds.
Identity and Belonging: Rethinking race
and ethnicity in Canadian society. Toronto,
Canada: Canadian Scholars’ Press Inc.,
2006;Chapter 3:43-55.
102. Di Biase S, Bauder H. Immigrants in
Ontario: Linking Spatial Settlement
Patterns and Labour Force Characteristics.
Department of Geography, University of
Guelph. 2004. URL:
Settlement%20in%20Ontario_report.pdf .
103. Ali J, Grabb E. Ethnic origin, class origin
and educational attainment in Canada:
Further evidence on the mosaic thesis.
Journal of Canadian Studies. 1998;32:
106. Wallace M. “Planning Amidst Diversity:
The Challenges of Multiculturalism in
Urban and Suburban Greater Toronto.”
PhD Thesis. University of Waterloo, 1999.
107. Beshiri, R. Immigrants in rural Canada:
2001 update. Rural and Small Town Canada
Analysis Bulletin. 2004;5:1-27. Cat. No.
108. Schellenberg G. Trends and Conditions in
Census Metropolitan Areas: Immigrants in
Canada’s Census Metropolitan Areas.
Statistics Canada. 2004. Cat. No. 89-613MIE - No. 003. URL: http://www.statcan.
89-613-MIE2004003.pdf .
109. Manitoba and Northern Affairs. Aboriginal
People in Manitoba 2000. URL: http://
html .
110. Office of the Registrar General, India.
Proportion and growth rate of population
by religious communities - India : 19612001. URL:
religiondata/statement.pdf .
Chronic Diseases in Canada
Quantitative empiric studies on ethnicity and mental health conducted in Canada
(or with Canadian data) that included two or more ethnic groups
N = Sample size of study, A = Age range of study individuals, S = Source(s) of data, D = Data type, DA = Data analyses, T = Type of study,
DV = Dependent Variable, IV = Independent Variable
Author(s), year of
publication & region
of study
Study Sample
(sample size, data source, age
group included and study design)
Ethnic groups included
Outcome Examined
Studies based on population-based or large community-based surveys/samples (n = 27)
1. Murphy, 1980
2. Liban & Smart, 1982
(Ontario, Canada)30
3. Havens & Chappell,
1983 (Manitoba,
4. Penning, 1983
5. Tcheng-Laroche &
Prince, 1983 (Montreal,
Quebec, Canada)33
6. Tonkin, 1984 (British
Columbia, Canada)34
7. Barnes et al., 1988
(Winnipeg, Manitoba,
DV: Mental hospitalizations
1. A clear definition of ethnicity not given.
N = not given
A = 15+
S = Dominion Bureau of Statistics reports on
first admissions to Canadian Mental Hospitals
in 1961 and the 1961 Population Census.
D = Cross-sectional
DA = Standardized Morbidity Ratio (SMR)
T = Descriptive
German Origin;
Dutch Origin;
Scandinavian Origin;
N =128
A = 10 to 20 (i.e. Grade 7 to 13)
S = 1979 Survey of Alcohol and Drug Use
among Ontario students
D = Cross-sectional
DA = Descriptive and Chi-square analyses
T = Descriptive
Native Indian [64];
Non-Native Indian [64]
N = 3647
A = 65+
S = Aging in Manitoba Survey (1971)
D = Cross-sectional
T = Analytic
North American [370]; British
French [216]; Polish/Russian/
Ukrainian [685];
Other European (German,
Norwegian, Danish, Swedish,
Icelandic, Dutch, Belgian)
DV: Perceived well-being,
perceived health status,
mental health functioning
N = 2253
A = 30+
S = Social Change in Canada Survey (1977)
D = Cross-sectional
T = Analytic
Canadian [1720]; American
[57]; British [159]; North
European (France, Germany,
Austria, Scandinavia,
Netherlands) [96]; South
European (Greece, Portugal,
Spain, Italy) [65]; East
European (Russia, Hungary,
Poland) [96]; Others [62]
DV: Perceived psychological
N = 128
A = not given
S = Community survey of a representative
sample of separated or divorced mothers in
D = Cross-sectional
DA = Chi-square, ANOVA
T = Analytic
Francophone [62];
Anglophone [66]
DV: Psychosocial stress per the 1. A clear definition of ethnicity not given.
Langner Scale, self-esteem per 2. Study interested in examining “cultural
the RSE and life satisfaction
N = 122
A = < 20 years old
S = Vital statistics data (with follow-up review
of all deaths reported to the provincial Chief
Coroner’s office 1978/9)
D = Cross-sectional
DA = Descriptive and chi-square analyses
T = Descriptive
Native Indian [33];
Non-Natives [89]
N = 524
A = 18 to 80
S = 1983 Winnipeg Area Study
D = Cross-sectional
DA = ANOVA, X2, & multiple classification
T = Exploratory
DV: Depression (CES-D)
English [84];
East European [62];
West European [82]; Canadian
Other [97]
Chronic Diseases in Canada
2. Ethnicity based on “country from which the
individual’s ancestors on the male side
came when they settled in Canada”,
therefore an ethnic origin conceptualization.
3. Ethnic groups reflective of the immigration
pattern in Canada during the time period
(i.e. 1960’s)
DV: Frequency and problems
with alcohol and drug use
1. A clear definition of ethnicity not given.
2. Ethnicity based on cultural background but
how ascertained not specified.
3. Although cultural background in the survey
included a breakdown into English
Canadian, French Canadian, Asian, Native
Indian and other, the non-Native Indian
categories were aggregated for the study
based on matching.
1. The ethnic groups in the study not
representative of the whole of Canada but
of Manitoba specifically.
2. Ethnic groups reflective of the immigration
pattern in the province and of the time
3. Ethnic origin based on ethnicity question in
the survey
1. A clear definition of ethnicity not given.
2. Some ethnic categories not clearly
explained for example the ethnic group
referred to as “Other”.
3. Ethnic origin based on country of birth.
3. Study included French- and
English-Canadians in Montreal. It is unclear
how being French and/or English-Canadian
were ascertained.
DV: Suicides and psychiatric
1. A clear definition of ethnicity not given.
2. Ethnicity used as a covariate but how it was
ascertained not indicated because it was
abstracted from the coroners’ records.
3. The rationale for the inclusion of ethnicity
as a covariate not explained.
1. Ethnicity examined as a predictor of
2. Ethnic groups reflective of the ethnic
composition region of study.
3. Ethnic origin based on country of birth.
Vol 28, No 4, 2008
Author(s), year of
publication & region
of study
Study Sample
(sample size, data source, age
group included and study design)
Ethnic groups included
Outcome Examined
8. Sack et al., 1993
(Canada & US)36
N = 1115
A = 7 to 9 (Grades 2 and 4)
S = Flower of Two Soils Project
D = Prospective and longitudinal (3 year followup)
DA = Correlation analyses, ANOVA, chi-square
T = Analytic
First Nation children from the
Plains (South Dakota, US),
Northern Woodlands
(Manitoba, Canada), Desert
(New Mexico, US) and Coastal
(British Columbia, Canada)
compared to a sample of
non-Native children at each
DV: Depressive symptoms
using new measures of
psycho-pathology and mental
health (the SOS)
1. No explicit definition given for ethnicity.
9. Walters, 1993
(Hamilton, Ontario,
N = 356
A = 21+
S = A stratified random sample of women in
D = Cross-sectional
DA = Chi-square
T = Descriptive
English-speaking vs. other
DV: Stress, anxiety and
1. Ethnic origin based on primary language.
10. Beiser et al., 1994
(Vancouver, British
Columbia, Canada)38
N = 1667
S = 1348 refugees from the Refugee
Resettlement Project and an area-probability
sample of 319 Vancouver residents matched to
refugees on age and sex
D = Cross-sectional
DA = Grade of Membership analysis (GOM: a
multivariate clustering technique)
T = Analytic
Southeast Asians [1348];
Resident Canadians [319];
Southeast Asian group
disaggregated: Chinese [755];
Vietnamese/Laotian [593]
11. Beiser et al., 1998
(Canada & US)39
N = 1708
A = 7 to 9 (Grades 2 and 4)
S = Flower of Two Soils Project
D = Prospective and longitudinal
DA = Principal component factor analysis on
the psychopathology measure, correlation
analyses, ANOVA, chi-square analyses
T = Analytic
First Nation children from the
Plains (South Dakota, US),
Northern Woodlands
(Manitoba, Canada), Desert
(New Mexico, US) and Coastal
(British Columbia, Canada)
[1251] compared to a sample
of non-Native children at each
site [457]
DV: Depressive symptoms
using new measures of
psychopathology and mental
health (the SOS)
1. No explicit definition given for ethnicity.
12. DeWit et al., 1999
(Ontario, Canada)40
N = 4531
A = 19+
S = Native Ontario Community Survey and the
Mental Health Supplement of the Ontario
Health Survey.
D = Cross-sectional (but age at onset
information used to look at incidence)
DA = Descriptive and chi-square and survival
T = Analytic
Native Indian [876];
Non-Natives [3655]
DV: Alcohol drug use and
1. No explicit definition given for ethnicity.
13. DeWit and
Beneteau, 1999
(Ontario, Canada)41
N = 5150
A = 16+
S = 1990 Ontario Health Survey
D = Cross-sectional but age at onset
information allowed survival analyses
DA = Chi-square, survival and logistic
regression analyses
T = Analytic
Anglophone [4023];
Francophone [1127]
DV: Alcohol consumption (i.e. 1. No explicit definition of ethnicity given.
frequency and volume);
2. Ethnicity based on a combination of ethnic
alcohol-related problems (i.e.
identity and primary language used at
driving under the influence,
family conflicts, work conflicts,
sought help for drinking,
hospitalization for drinking,
and/or arrested for drunk
14. DeWit and
Beneteau, 1999
(Ontario, Canada)42
N = 5150
A = 16+
S = 1990 Ontario Health Survey
D = Cross-sectional but age at onset
information allowed survival analyses
DA = Chi-square, survival and logistic
regression analyses
T = Analytic
Anglophone [4023];
Francophone [1127]
DV: Daily tobacco
consumption (i.e. frequency
and volume)
1. No explicit definition of ethnicity given.
15. Feldman et al., 1999
(Toronto, Ontario,
N = 1236
A = Grade 9 to 13 students
S = 1994 Survey of Grade 9 to 13 students in
the Borough of East York, Toronto
D = Cross-sectional
DA = descriptive analyses, stratified analyses
and multiple logistic regression analyses
T = Analytic
Canadian [379];
European [277];
Asian [314];
Other [140];
Not stated [126]
DV: Alcohol use beliefs and
1. No definition given for ethnicity.
Vol 28, No 4, 2008
2. Study specifically interested in cultural
differences between the various First Nation
groups across North America, hence it is
implied that ethnicity is based on culture.
2. No definition for ethnicity.
3. Language stated as a possible proxy for
DV: Psychiatric disorders
including depression, anxiety
and somatization per the
CES-D, DIS and the Senegal
Health Scales
1. No explicit definition given for ethnicity.
2. Researchers interested in
“psychopathological expression among
different ethno-cultural groups”, therefore
ethnicity based on culture.
3. Ethnicity based on where the individual
emigrated from.
2. Study specifically interested in cultural
differences between the various First Nation
groups across North America.
2. Study specifically interested in cultural
differences between Native Indians and
non-Natives in Canada.
2. Ethnicity based on a combination of ethnic
identity and primary language used at
2. “Please write down the term that best
describes the ethnic character of your
everyday home environment” was used to
ascertain ethnicity indicating an ethnic
identity conceptualization.
Chronic Diseases in Canada
Author(s), year of
publication & region
of study
Study Sample
(sample size, data source, age
group included and study design)
Ethnic groups included
Outcome Examined
First Nation children from the
Plains (South Dakota, US),
Northern Woodlands
(Manitoba, Canada), Desert
(New Mexico, US) and Coastal
(British Columbia, Canada)
[1555]; compared to a sample
non-Native children at each
site [489].
DV: Attention Deficit/
Hyper-activity Disorder
(ADHD) per DSM symptom
criterion but items measured
by the TIF, the CAP and the
SOS scales which contained
items drawn from the CBCL,
the CPTRS and the DIS for
1. No explicit definition given for ethnicity.
17. Lavallee & Bourgault, N = 27 130 women
2000 (Canada)45
A = 15+
S = 1991 Cree Health Survey, 1992 Inuit Health
Survey & 1992-93 Quebec Health and Social
D = Cross-sectional
DA = Weighted Frequency distribution,
Chi-square and ANOVA analyses
T = Descriptive
Cree [1999];
Inuit [1597];
Southern Quebecers [23 564]
DV: Alcohol consumption,
illicit drug use, psychological
distress, lifetime suicidal
1. Researchers interested in the mental health
of Cree, Inuit and Southern Quebec women,
so no mention or definition of ethnicity.
18. Ali, 2002 (Canada)5
N = 92 379
A = 15 to 75
S = Canadian Community Health Survey, Cycle
D = Cross-sectional
DA = Multiple logistic regression analysis
T = Analytic
US/Mexico [952]; S. America,
C. America, Caribbean [2273];
Europe [7749]; Africa [1139];
Asia [6314]
DV: Depression & alcohol
1. Ethnic origin based on region of birth.
N = 13 349
A = 4 to 11
S = National Longitudinal Survey of Children &
Youth (NLSCY/1994-1995)
D = Cross-sectional
DA = Multiple logistic regression analyses
T = Analytic
Immigrants [684]; Born in
Canada to immigrant parents
[2573]; Canadian-born to
Canadian-born parents [10
Ethnicity then examined
(Asian, Black, other, White/
DV: Emotional and
behavioural problems
1. Ethnicity examined as a covariate.
20. Ma, 2002 (Canada)47 N = 2304
A = 7 to 11
S = NLSCY 1994/5
D = Cross-sectional
DA = Factor analysis and mixed level modeling
T = Analytic
Immigrant Children [182];
Non-immigrant Children
[2122]. Ethnicity then
examined as a covariate (US,
Europe, Asia, other regions).
DV: Conduct disorder, indirect 1. Ethnic origin based on region of birth.
aggression, property offences, 2. Ethnicity examined as a covariate.
hyperactive behaviour,
3. No definition given for ethnicity.
pro-social disorder, emotional
disorder & a composite
behavioural /emotional
disorder index
21. Mata, 2002
N = 17 109
A = 15+
S = National Survey of Giving, Volunteering and
D = Cross-sectional
DA = Multiple regression analyses
T = Analytic
DV: Life satisfaction (4-pt
19 Single/multiple ethnic
categories (Canadian, French, Likert scale)
British, German, Ukranian,
Polish, Italian/Portuguese,
Dutch, Chinese, Black/South
Asian, Aboriginal, Canadian &
French, Canadian & British,
Canadian & Other, British &
French, French & Other,
European, Rest of singles, Rest
of multiples)
1. Ethnic origin based on the ethnicity
question in the survey.
N = 3009 elderly immigrants
A = 65+
S = National Population Health Survey (NPHS,
D = Cross-sectional component
DA = Descriptive and multiple linear regression
T = Analytic
Chinese/South Asian [274];
Other [2735]
1. Ethnicity based on the ethnic origin
question in the survey.
N=70 538
S= NPHS, 1996
D=Cross-sectional component
DA: Descriptive and multiple linear regression
T = Analytic
East & Southeast Asian [624];
Chinese [800]; South Asian
[809]; Aboriginal [975];
Black [788]; Arabic & West
Asian [325]; Latin American
[176]; Jewish [197]; French
English [9281]; “Other” Whites
[50 294]; Mixed race [689]
16. Beiser et al., 2000
(Canada & US)44
19. Beiser et al., 2002
22. Wu & Hart, 2002
23. Wu et al., 2003
N = 2044
A = 7 to 9 (Grades 2 and 4)
S = Flower of Two Soils Project and the School
Option for Native Children Study
D = Prospective and longitudinal
DA = Principal component factor analysis on
the psychopathology measure, correlation
analyses, ANOVA, chi-square analyses
T = Analytic
Chronic Diseases in Canada
2. Study specifically interested in cultural
differences between the various First Nation
groups across North America, hence it is
implied that ethnicity is based on culture.
2. The group ‘Southern Quebec’ likely
included multiple ethnic groups.
2. Although immigrants broken down into
regions migrated from, the Canadian-born
group wasn’t, which limited comparisons
based on ethnicity.
2. Ethnic origin based on ethnicity question.
3. The “Others” race/ethnicity category not
clearly defined.
DV: Emotional problems and
psychological distress (CIDI)
2. Mutually exclusive ethnic groupings driven
in part by sample size issues.
3. Rationale for some ethnic groupings
4. No definition of ethnicity.
2. The combination of Chinese and all South
Asians indicate that ethnic origin based on
large geo-cultural region.
DV: Depression (CIDI)
1. Ethnicity based on the ethnic origin
question in the survey and in combination
with race used to create ethno-racial
2. Offered definitions of race and ethnicity a
Vol 28, No 4, 2008
Author(s), year of
publication & region
of study
24. Blackstock et al.,
2004 (Canada)50
Study Sample
(sample size, data source, age
group included and study design)
Ethnic groups included
Aboriginal [614];
N =3159
White [2114];
A = childhood (range not given)
S = 1998 Canadian Incidence Study of Reported Other Visible Minority [431]
Child Abuse and Neglect (CIS-98)
D = Cross-sectional
DA = Chi-square and ANOVA analyses
T = Descriptive
Outcome Examined
DV: Frequency of Child
Maltreatment (physical and
sexual abuse and neglect);
psychosocial problems
1. No explicit definition of ethnicity given.
2. Ethno-racial classification determined by
ethno-racial status of one or both biological
3. Conceptualized based on ethnic origin.
25. Cohen & Maclean,
2004 (Canada)51
N = 26 000
A = 15+
S = All females from the 1999 General Social
D = Cross-sectional
DA = z-test with p < 0.05 significance level
T = Descriptive
Visible minority vs. non-visible DV: Physical, sexual, financial
or emotional abuse;
Aboriginal vs. non-Aboriginal medication use for anxiety,
depression or insomnia in
those abused.
1. Ethnicity based on country of birth.
26. Wang & El-Guebaly,
2004 (Canada)52
N = 72 940
A = 12+
S = NPHS (1996/7)
D = Cross-sectional
DA = Multiple logistic regression analyses
T = Analytic
White [67 802];
Non-White [5138]
1. Ethnicity examined as a covariate.
DV: Major depressive episode
(MDE), alcohol dependence
(AD) and mental health
service use
2. No definition given for visible minority.
2. No description of the groups included in the
non-white category.
3. Non-white population likely included
Aboriginals, a group with great likelihood
of MDE and AD, which likely explains the
higher risk observed in non-immigrant
4. Race only categorization despite use of the
term “ethnicity”.
DV: Emotional distress (i.e.
depression and anxiety) per
the SCL-25 based on the
Hopkins Symptom Checklist
1. Explicit definition of ethnicity not given.
27. Rousseau &
Drapeau, 2004
(Montreal, Quebec,
N = 1871
A = 15 to 87
S = Quebec Cultural Communities Survey
(survey of recent immigrants living in the
metropolitan Montreal area who landed in
Canada between 1988 and 1997)
D = Cross-sectional
DA = Chi-square, t-test, ANOVA
T = Descriptive
28. Fry & Grover, 1982
(Canada & USA)54
N = 320
A = 65 to 80
S = Random sample drawn form professional
clubs, community associations, recreation
centres for elderly, social welfare agencies and
private homes.
D = Cross-sectional
T = Analytic
Asian-Indian [160]; Caucasian
DV: Depression (per the BDI), 1. Definition of ethnicity not given.
life stress [per the Life Event
2. How ethnicity ascertained unclear.
Inventory], cognitive appraisal
3. Seem to be a race only categorization.
and locus of control
29. Dyal & Chan, 1985
(Waterloo, Ontario,
N = 251
A = 17 to 29
S = Samples of convenience (i.e. students in
undergraduate courses and some volunteers)
D = Cross-sectional
T = Analytic
Euro-Canadian [112];
Hong Kong Chinese [100];
Chinese immigrants to
Canada [39]
1. No explicit definition given for ethnicity.
DV: Stressful life events per
the Problems with Living
2. Implication that ethnicity based on culture
Adjustment scale (developed
since the study interested in cross-cultural
for the study); distress per the
DSS based on Langer 22-item
3. Unclear whether the Euro-Canadian groups
scale of impaired functioning;
actually identified themselves as such or if
12- items from the DDS; worry
this categorization based on the
per the worry scale of the
researchers’ observations.
30. Blandford &
Chappell, 1990
(Winnipeg, Manitoba,
N = 390
A = 50 and over
S = Survey of Natives in Winnipeg in 1981
D = Cross-sectional
DA = Chi-square and logistic regression
T = Analytic
Natives [193]
Non-Natives [197]
DV: Satisfaction with life;
loneliness per the UCLA
Loneliness Scale
1. Ethnicity not specifically defined.
31. Dion & Giordano,
1990 (Toronto, Ontario,
N = 352
A = 1st-year university students – mean age of
20.32 years (1988)
S = A sample of undergraduate students in an
introductory psychology course
D = Cross-sectional
DA = ANOVA, X2, and multivariate log-linear
T = Analytic
Anglo-Celtic [165]; North
European [22];
South European [79]; East
European [36];
East Asians [25];
South Asians [25]
IV: Sex and ethnicity
DV: Depression (total & item
scores for the BDI)
Note: indicated in the article
that equal numbers of each
ethnic group were selected
from the registry to represent
the target population [n = 750
* 4 = 3000] but the eligible
population was 1871 with no
breakdown of the numbers in
each ethnic group
2. Researchers interested in recent immigrants
in the Montreal area who were born in one
of four geocultural areas (China, Hong
Kong, Taiwan and Macao [Chinese], Haiti
[Haitian], North Africa and the Middle East
[Arabs] and Latin America [Hispanics].
Therefore, ethnicity conceptualized based
on region of birth.
Smaller studies with primary data collection – non-clinical sample (n = 12)
Vol 28, No 4, 2008
2. Conceptualized based on ethnic identity but
no indication of how ethnic identity was
Ethnic origin based on surnames, with the
aid of a number of dictionaries of
surnames/family names.
2. This limited the ability to identify black
individuals with West Indian ethno-cultural
background and South Asians of Christian
Chronic Diseases in Canada
Author(s), year of
publication & region
of study
Study Sample
(sample size, data source, age
group included and study design)
Ethnic groups included
Outcome Examined
Canadian born of European
descent [100];
Chinese immigrants long
established in Canada [50];
Chinese immigrants newly
arrived in Canada [50];
Chinese in Hong Kong [100]
DV: Physical and mental
health as measured by the
GHQ; loneliness per the UCLA
Loneliness Scale; quality of
life & acculturation; & global
satisfaction with life
1. Explicit definition given for ethnicity.
33. Dion, 1996 (Toronto, N = 950
A = ?18-21: 1st-year university students
S = Convenience sample
D = Cross-sectional
T = Exploratory
Language as indicator of
ethnic origin
DV: Alexithymia (TAS-20) and
three under-lying factors
(DIF= difficulty identifying
feelings, DDF= difficulty
describing feelings, EOT=
externally oriented thinking)
1. Language identified by Dion as a possible
proxy of ethnicity (i.e. ethnic origin).
34. Heine and Lehman,
1999 (Canada &
Japanese [161];
Asian-Canadian [151];
Euro-Canadian [90]
DV: Personality traits: 20
items to capture the
individuals’ ratings of their
actual & ideal self and what
they thought described the
average student traits;
difference between actual &
ideal self, and importance of
traits to success in one’s
country were also assessed;
depression (ZSDS)
1. No explicit definition given for ethnicity; it
appeared to be based on culture.
DV: Emotional profile of
subjects based on the
SCL-90R; Post-traumatic stress
disorder per the DSM-IV.
1. Ethnic origin based on geo-cultural region
of birth.
32. Bagley, 1993
(Calgary, Alberta,
Canada & Kowloop,
Hong Kong, China)58
N = 300
A = 60 to 74 years old
S = The 100 Canadian-dwelling elderly
Cantonese-speaking Chinese sample was
recruited by randomly selecting Chinese
surnames from the telephone directory with an
initial call by a Cantonese-speaking individual
who established the presence of an elderly
person (aged 60 to 74) and then inquired about
and established the ethnic origin of the
individual. Stratified sampling of young to
elderly in Kowloop, Hong Kong to match the
age and sex profile of Canadian-dwelling
Cantonese-speaking elderly Chinese group.
D = Cross-sectional
DA = Comparison of the standardized scores
T = Descriptive
N = 402
A = 18 to 25?
S = Survey of university students
D = Cross-sectional
DA = ANOVA with post hoc tests (Tukey’s HSD)
T = Analytic
35. Rousseau et al.,
2001 (Montréal,
Latin American [60];
N = 113
African American [53]
A = 20 to 65
S = Sample of refugees who sought help from
community organizations that provide services
to refugees/immigrants in the Montreal area
D = Qualitative & cross-sectional quantitative
design combined
DA = Descriptive analyses and Spearman rank
correlation and t-tests for comparative analyses
T = Descriptive
36. Howard et al, 2003
(Toronto, Ontario,
Canada; USA; Taipei
City, Taiwan)62
N = 3030
A = not given
S = All participants from Canada are volunteers
recruited through newspaper and flyer
advertisements in Toronto; Taiwanese recruited
at the Taipei City Psychiatric Center; African
American smokers recruited at the University of
Kansas Medical Center (Clinical and community
D = Association Analysis (Cross-sectional)
DA = ANOVA; T-test, Mann-Whitney U-test and
Chi-square analyses
T = Analytic
From Canada:
Indo-Asian [48];
Chinese [210];
Japanese [128];
African Canadian [58];
Native Indian [228];
Caucasian [1734]
N = 123 (Study 1)
415 (Study 2)
A = 18 to 47
S = Undergraduate students from the University
of British Columbia, Canada and from
Ritsumeikan University (Study 1) or Kurume
University (Study 2), Japan
D = Cross-sectional
T = Analytic
Study 1:
Western European Canadian
East Asian Canadian [57];
South Asian/Mixed descent
Canadian [18];
Japanese [26]
37. Tweed et al., 2004
(British Columbia,
Canada and Japan)63
Chronic Diseases in Canada
2. Different techniques seem to have been
used to establish ethnicity across groups.
Surnames, language and unknown question
for Canadian dwelling Chinese; country of
birth for Hong Kong Chinese and unknown
method for Euro-Canadians.
2. No definition for ethnicity.
2. Method of sample selection unclear.
3. SES and/or SS hypotheses not tested.
2. Explicit definition of ethnicity not given.
DV: CYP2E1*1D allele; alcohol 1. Ethnicity not explicitly defined.
2. Ethnic background based on the
individual’s grandparents therefore an
ethnic origin conceptualization.
Taiwanese [420];
African American [204]
DV: Stressful and negative life
events; coping skills per the
WCCL and Japanese coping
1. Ethnicity not explicitly defined but based on
2. Conceptualized based on descent, therefore
ethnic origin.
Study 2:
European Canadian [68];
East-Asian Canadian [106];
Japanese [241]
Vol 28, No 4, 2008
Author(s), year of
publication & region
of study
38. Khanlou, 2004
(Hamilton, Ontario,
Study Sample
(sample size, data source, age
group included and study design)
N = 550
A = Grades 9 to 13
S = Survey of students at 4 secondary schools
in the Hamilton Wentworth region in 1998
D = Cross-sectional
DA = Chi-square and ANOVA analyses
T = Descriptive
Ethnic groups included
Information given for the
three most frequently
occurring ethnic/cultural
background of the mothers
and fathers of the students.
Outcome Examined
DV: Self-esteem per the RSE
and the CSE scales
N = 170
A = 18+
S = Convenience sample of students
D = Cross-sectional
T = Analytic
Other Canadian [81]; Chinese
Canadian [89]
2. Ethnicity based on the “parents’ original
ethnic or cultural background” indicating an
ethnic origin conceptualization.
3. Specific ethnic groupings were used with
some having very small sample size. As a
result comparison only done on the top
three ethnic categorizations.
Based on mother:
Italian [48];
Portuguese [43];
Irish [35];
English [35]
4. The study indicated that close to 30% of
the sample reported 2 or more ethnic
backgrounds indicating a “mixed” ethnic
Based on Father:
Italian [69];
Portuguese [42];
Canadian [38]
39. Aubert et al., 2004
1. No definition given for ethnicity.
DV: Hostility (HDHQ); suicide
probability (SPS); lifetime
aggressive behaviour incl.
suicide attempt; suicidal
thoughts and deliberate self
1. Ethnicity examined as a covariate
2. It is unclear if “Other Canadian” group
include ethnicity other than
3. Ethnic origin based on country of birth.
Smaller studies with primary data collection – clinical/specialized sample (n = 10)
DV: Schizophrenia
1. Ethnicity examined as a covariate.
N = 43 (33 non-immigrants & 10 immigrants)
A = 15 to 49
S = 1963 First episode hospitalization data
D = Retrospective follow-up on a cohort of
persons with schizophrenia compared to the
general population
DA = Descriptive and Chi-square analyses
T = Descriptive
10 immigrants classified as
“other European” [6], &
Eastern European [4]
41. Seltzer & Langford,
1984 (Northwest
Territories, Canada)67
N = 85
A = 15 to 25
S = Sample of convenience (i.e. all persons
referred by the courts or legal counsel to the
Department of Psychiatry at the regional
hospital in the calendar year 1981)
D = Cross-sectional
DA = Descriptive
T = Descriptive
Inuit [41];
Métis/Dene [27];
Caucasian [17]
1. No explicit definition given for ethnicity.
DV: DSM-III psychiatric
diagnosis and type of criminal 2. The Native group broken down into Inuit
offences committed
and Metis, indicating cultural distinctions,
but the Caucasian group not disaggregated.
42. Borzecki et al., 1988
N = 275 (all males)
A = 28.41 = mean age at admission
S = successive 1st admission to the facility
between January 1978 and September 1982
D = Cross-sectional
DA = ANOVA, ANCOVA, and Pearson’s
T = Analytic
Natives (i.e. Inuit, Métis, and
Indians) [57];
Non-Natives [218]
DV: Psychological profile per
the MMPI
43. Chandrasena et al.,
1991 (Ontario,
N = 117
A = range not given
S = All suicides in 3 psychiatric facilities in
Ontario between 1967 and 1990
D = Cross-sectional
DA = Descriptive Analyses
T = Descriptive
Canadian-born [94];
Foreign-born [23]
DV: Suicide
44. Norton et al., 1995
(Manitoba, Canada)70
N = 80
S = Attendees at the Alcoholism Foundation of
D = Cross-sectional
DA = Chi-square
T = Descriptive
Native-Canadians [37];
Anglo-Canadians [43]
1. No explicit definition of ethnicity given.
DV: Suicidal ideation (per
NIMH Epidemiologic
2. Unclear how ethnicity was ascertained.
Catchment Area survey), panic
(PAQ), chemical abuse (the
depression (BDI)
45. Weekes et al., 1995
(Ontario, Canada)71
N = 301
A = 18 to 59
S = Sample of convenience: Adult males
incarcerated in a medium security federal
D = Cross-sectional
DA = MANOVA, T-TEST, correlation and
principal component analyses
T = Analytic
Caucasian [203];
Native [59];
Métis [39]
DV: Psychopathology per the
40. Bland & Orn, 1981
(Alberta, Canada)66
Vol 28, No 4, 2008
2. The term “ethnic group” used but no
definition of what it means.
3. Ethnic origin based on place of birth.
1. No explicit definition of ethnicity given.
2. Ethnicity based on ancestry (ethnic origin), at
least for the native sample, while all
non-natives, despite ancestral heterogeneity
are grouped together.
1. Ethnicity mentioned but not defined.
2. Canadian-born versus foreign-born indicate
a country of birth conceptualization but not
explicitly stated.
1. No explicit definition of ethnicity
2. The terms ethnicity, “cultural group”, and
“racial identification” used in the article but
categorization based on the individuals
self-report of “racial identification”.
Chronic Diseases in Canada
Author(s), year of
publication & region
of study
Study Sample
(sample size, data source, age
group included and study design)
Ethnic groups included
Outcome Examined
DV: Revised CBCL; Depression 1. No explicit definition given for ethnicity
Self-rating Scale; the What I
2. There were 34 parents and 48 children so
Think and Feel Scale; the
unclear of the ethnicity breakdown of the
Children’s Psycho-somatic
Symptom Checklist; and the
3. Since ethnicity for parents given but not for
Hare Self-Esteem Scale.
children, it appears that ethnicity based on
region/place of parent’s birth.
N = 34 multiethnic families (one parent
included) with a total of 48 children (of which 8
born in Asia and 3 in India. No other mention
of the ethnic breakdown of the children).
A = age range of parents not given but children
ranged in age from 6.5 to 17 years old
S = convenience sample of the parents and
their children seen in a Pediatric Clinic in
Montreal, Canada
D = Cross-sectional
T = Analytic
Ethnic breakdown of the
parents given but not for the
47. Zapf et al., 1996
(Vancouver, British
Columbia, Canada)73
N = 790
A = not specified
S = Adult males randomly selected from
Vancouver Pretrial Service Centre between
August 1, 1989 and July 31, 1990
D = Cross-sectional
DA = Chi-square and Pearson Correlation
T = Descriptive
Although the term “ethnicity”
used, there was no
breakdown of the ethnic
DV: Mental disorder per the
BPRS and the Diagnostic
48. Devins et al., 2000
(Canada & US)74
N = 405
A = not specified
S = The Arthritis, Rheumatism and Aging
Medical Information System Lupus Project
D = Longitudinal
DA = Principal component and path analyses
T = Analytic
White [335];
Black [40];
Asian [30]
(all female)
1. The Black and Asian groups comprised of
DV: Psychosocial well-being
mostly (i.e. 97% & 83%) individuals from
(ABS); learned helplessness
the US, whereas almost equal proportion of
(RAI); emotional distress
whites from the US and Canada).
(CES-D); musculo-skeletal pain
(HAQ); overall psycho-social
2. Racial classification rather than ethnicity.
49. Hodelet, 2001
(British Columbia,
N = 175
A = 19 to 75
S = Secure Forensic Psychiatry Hospital, all case
records for patients in hospital between
December 1, 1998 and February 28, 1999
D = Cross-sectional
DA = Chi-square and ANOVA analyses
T = Descriptive
White [153];
Native American [26];
Oriental/East Asian [11];
South Asian [5];
Black [1]
DV: Type of offence; type of
psychiatric diagnosis;
psychosis; psychotic drive
46. Pawliuk et al., 1996
(Montreal, Canada)72
Chronic Diseases in Canada
Asian [26];
European [5];
Indian/S. American/Middle
Eastern [3]
1. Ethnicity stated as a covariate.
2. No definition given for ethnicity.
3. No indication of how ethnicity
conceptualized or operationalized.
1. Explicit definition of ethnicity not given.
2. “Ethnic origin” indicated but unclear how
this was ascertained in the individuals’
medical records.
Vol 28, No 4, 2008
Association of comorbid mood disorders and chronic
illness with disability and quality of life in Ontario,
T Gadalla, PhD (1)
Mood disorders are more prevalent in individuals with chronic physical illness compared
to individuals with no such illness. These disorders amplify the disability associated with
the physical condition and adversely affect its course, thus contributing to occupational
impairment, disruption in interpersonal and family relationships, poor health and
suicide. This study used data collected in the Canadian Community Health Survey, cycle
3.1 (2005) to examine factors associated with comorbid mood disorders and to assess
their association with the quality of life of individuals living in Ontario. Results indicate
that individuals with chronic fatigue syndrome, fibromyalgia, bowel disorder or stomach
or intestinal ulcers had the highest rates of mood disorders. The odds of having a comorbid
mood disorder were higher among women, the single, those living in poverty, the Canadian
born and those between 30 and 69 years of age. The presence of comorbid mood disorders
was significantly associated with short-term disability, requiring help with instrumental
daily activities and suicidal ideation. Health care providers are urged to proactively screen
chronically ill patients for mood disorders, particularly among the subgroups found to
have elevated risk for these disorders.
Key words: Ontario, mood disorders, chronic diseases, quality of life, short-term
Mood disorders (major depressive disorder,
bipolar disorder, mania or dysthymia) are
the most prevalent of all mental disorders.
One in 7 adults (13.4%) living in Canada
reported symptoms that met the criteria
for a mood disorder at some point in their
lives and about 5.3% of the Canadian
population aged 15 years and over were
identified as suffering from a mood
disorder in 2002.1 The burden of depression
to individuals and societies is such that
the World Health Organization has
projected that by the year 2020, unipolar
major depression will be the leading cause
of disability-adjusted life years (DALYs)
after cardiovascular disease.2 Evidence
exists that the prevalence of mood
disorders in individuals with chronic
physical illness is noticeably higher
compared to individuals with no such
illness. According to the 2002 Mental
Health and Well-being Survey, an individual with a chronic physical condition
was twice as likely as an individual
without such a condition to have a mood
disorder.3 Depressive disorders often
accompany chronic illnesses such as heart
disease, stroke, Parkinson’s disease,
cancer and HIV/AIDS (Evans et al, 2005).4
For example, reported prevalence rates of
depression range from 17% to 27% in
patients with cardiac disease, from 22%
to 29% in patients with cancer and from
9% to 26% in patients with diabetes.4
Fuller-Thomson and Sulman5 reported
that among Canadians who had inflammatory bowel disease in 2002, 16.3%
suffered from depression.
High prevalence of mood disorders among
individuals with chronic physical conditions represents a significant burden to
individuals and society. At the individual
level, they can lead to occupational impairment, disruption in interpersonal and
family relationships, poor health and
suicide.6 Existing research suggests that
major depression interacts with physical
illness to amplify the disability associated
with many physical conditions as well as
adversely affect the course of physical
illnesses.4 Individuals with physical health
problems often experience anxiety or
depression, which affects their response
to the treatment of their physical illness.
On the other hand, individuals with
mental illness can develop physical
symptoms and illnesses, such as weight
loss and biological disturbances associated
with eating disorders. Most of the research
done in this field focuses on the impact of
depression on quality of life after adjusting
for the severity of physical illness (e.g.
Vali and Walkup7 and Ades, et al.8) rather
than comparing the impact of comorbid
mental disorders in individuals with and
without physical illness. These studies
usually find that depression adversely
affect patients’ quality of life after adjusting
for severity of physical illness. Ferketich,
et al.9 analyzed data collected in a National
longitudinal study and found that depressed men had a 71% greater risk of
developing heart disease and were 2.34
times more likely than non-depressed men
Author References
1 Faculty of Social Work at the University of Toronto
Correspondence: Tahany Gadalla, Assistant Professor, Faculty of Social Work/University of Toronto, 246 Bloor Street West, Toronto, Ontario, Canada M5S 1A1,
Tel: (416) 946-0623, Email: [email protected]
Vol 28, No 4, 2008
Chronic Diseases in Canada
to die from this condition. Simon, et al.10
compared SF-36 subscale ratings on a
clinical sample of depressed patients with
and without chronic physical illness and
concluded that depressive disorders and
chronic physical illness produced differing
patterns of impairment, and that comorbid
depressive disorders created a substantial
burden of additional functional impairment.
Katon11 estimated medical costs for patients
with major depression to be almost 50%
higher than the costs of chronic physical
illness alone. The associations between
other mood disorders such as bipolar
disorder or dysthymia and physical illness
and their effect on the quality of life of the
physically ill have not been examined.
At the national level, comorbid mood
disorders can adversely affect the economy
through reduced productivity and higher
health care costs. Several studies have
investigated the impact of comorbid
depressive disorders on the cost of health
services using patients with specific physical
conditions such as arthritis.7 In a study of
the economic burden of mental health in
Canada, Stephens and Joubert12 estimated
work-related productivity losses due to
mood disorders alone to be $4.5 billion
annually. Little is known about the impact
of comorbid mood disorders on individuals’
productivity, daily functioning and quality
of life in the Canadian population. Further,
the cost to society of the comorbidity of
mental disorders and physical conditions
requires further investigation.
The widely reported under-treatment of
mood disorders magnifies and reinforces
their adverse impact on the lives of the
physically ill and society as a whole.
Although effective interventions for the
treatment of mood disorders are available,
most individuals who suffer from them fail
to consult health professionals. Strakes, et
al.13 reported that only 40% of people with
probable depression in Atlantic Canada
consulted a general practitioner or a mental
health specialist about their condition.
Despite a universal coverage of physician
and hospital services in Canada, only
56.4% of Canadian women identified as
having at least one major depressive
episode (MDE) in 2002 reported accessing
Chronic Diseases in Canada
health care resources in the 12 months
prior to the interview.14 Among Canadians
with inflammatory bowel disease who
were identified as having major depression,
only 40% were using antidepressants, and
between one third and one half were not
consulting any mental health professionals.5 Identification of the sociodemographic characteristics of physically
ill individuals at high risk of having mood
disorders provides health care providers
with essential knowledge for targeting
high risk individuals leading to early
diagnosis and intervention.
Most studies that explored the comorbidity
of mental disorders and physical conditions
have used clinical samples, which may
lead to biased results. Thus, researchers
have highlighted the role of populationbased studies in determining the extent
and nature of comorbidity between mental
disorders and physical illness and the
impact of such comorbidity on the afflicted
individuals and the society at large. To
date, only a few studies which have
examined the comorbidity between mental
disorders and physical conditions have
been conducted in community based
samples. In addition, most of these studies
used self-reported diagnosis of physical
conditions or self-reported symptoms of
mental disorders. The present investigation
was based on the most recent data available
on a representative sample of individuals
living in the province of Ontario, Canada.
This investigation had two objectives.
First, it examined the relationship between
socio-demographic characteristics and
mood disorders in Ontarians with chronic
physical conditions. Second, it assessed
short-term disability, limitations in activities
of daily living and suicidal ideation in
individuals with comorbid mood disorders
compared with those with chronic physical
illness only.
This research was based on a subset of the
data collected by Statistics Canada in cycle
3.1 of the Canadian Community Health
Survey and are available for public use.15
The survey was conducted in 2005 and
employed a multistage stratified cluster
probability sampling in which dwelling
was the final sampling unit. The sample
was stratified by province, and urban
versus rural regions, within province. The
survey sample represented approximately
98% of the Canadian population aged 12
or older who resided in private dwellings
in the ten provinces and the three
territories. Persons living on Indian
Reserves, residents of institutions, fulltime members of the Canadian Armed
Forces and residents of certain remote
regions were excluded from the survey.
Fifty percent of the respondents were interviewed face-to-face using the computerassisted personal interviewing (CAPI)
method and 50% were interviewed by
telephone using the computer-assisted
telephone interviewing (CATI) method.
Cycle 3.1 of the CCHS included a list of
common questions to be used in all
provinces and a number of lists of optional
questions for provinces to select among
them. This was done to allow each province
to select questions related to their particular
needs and priorities. Consequently, not all
variables are available for all provinces.
Data collected from Ontario participants
included all the variables required for the
present analysis. Hence, Ontario data were
used in this research.
Participants were asked the following
question, “Now I’d like to ask about certain
chronic health conditions which you may
have. We are interested in long-term
conditions, which are expected to last or
have already lasted 6 months or more and
that have been diagnosed by a health
professional.” The list of chronic physical
illnesses included food allergies, other
allergies, asthma, fibromyalgia, arthritis or
rheumatism, back problems, high blood
pressure, migraine headaches, chronic
bronchitis, emphysema, pulmonary disease,
diabetes, epilepsy, cancer, heart disease,
bowel disorder, cataracts, glaucoma, thyroid
condition, chronic fatigue syndrome, mood
disorders (such as major depressive
disorder, bipolar disorder, mania or
dysthymia), and anxiety disorders (such
Vol 28, No 4, 2008
as phobia, obsessive-compulsive disorder
or panic disorder).
Two variables were used to measure shortterm disability: the number of disability
days during the two weeks prior to the
interview and whether the participants
required help with their usual daily
activities. Survey participants were asked
about the number of days they stayed in
bed for all or most of the day due to illness
or emotional/mental health during the
two weeks prior to the interview. They
were also asked a series of questions about
whether they needed help with instrumental activities of daily living such as
preparing meals, shopping for groceries
and other necessities, getting to appointments, doing everyday housework,
personal care or moving about inside their
home because of a long-term health
condition. A long-term health condition
was defined as a condition that is expected
to last or has lasted 6 months or more. In
addition, survey participants who live in
Ontario were asked whether they had
considered suicide in the 12 months prior
to the interview.
Socio-demographic characteristics and
health indicators used in this research
included gender, age group (12 to 29 years,
30 to 49 years, 50 to 69 years, 70 years or
older), marital status (married/common
law versus divorced/separated/widower/
never married), immigration status (immigrant versus Canadian born), education
level (less than secondary school degree,
secondary school graduate, some post
secondary education and post secondary
graduate) and income level (low 30%,
middle 40% and upper 30% of the income
distribution in Ontario). The income level
variable was derived by Statistics Canada
to measure the participant’s household
income relative to the household incomes
of all participants living in the province
and having the same household size.15
First, an adjusted ratio of the participant’s
total household income to the low-income
cutoff corresponding to their household
size and community size was calculated.
The adjusted ratios were then divided into
deciles, i.e. 10 categories including approximately the same percentage of residents.
Vol 28, No 4, 2008
Data analyses
Prevalence of comorbid mood disorders
among individuals with chronic physical
conditions was calculated. Expected probabilities of random co-occurrence of
chronic physical conditions and mood
disorders were calculated and compared
with the observed probabilities. Chi-square
tests were used to assess bivariate relationships between prevalence of comorbid
mood disorders and socio-demographic
characteristics in individuals with chronic
physical illness as well as bivariate relationships between mood disorders and quality
of life. Logistic regression analysis was
used to identify socio-demographic factors
associated with a high risk of having mood
disorders in individuals with chronic
physical illness. Sampling weights were
rescaled and used in all analyses. Rescaling
the weights to have an average of one has
two advantages. First, it takes into account
the unequal probabilities of selection of
survey participants and adjusts the sample
results to the demographic composition of
Ontario population so that the results
represent the population of Ontario and
not just the sample itself. Second, it keeps
the total sample size unchanged to guard
against inflating the sample size for
hypothesis testing.
Table 1 shows the prevalence of chronic
physical illness and mood disorders by
gender. Based on data presented in this
table, the prevalence of mood disorders
among men with chronic physical illness
was 6.5% compared with 1.9% among
men with no such illness. The prevalence
of mood disorders among women with
chronic physical illness was 10.5% compared with 3.8% among women with no
chronic physical illness. Using the sampling
weights published by Statistics Canada, it
can be estimated that approximately
2 557 000 men and 3 170 000 women of
Ontario residents were living with chronic
physical illness in 2005. Among them,
approximately 165 000 men and 334 000
women had been diagnosed with at least
one mood disorder.
As shown in Table 1, 10 105 of all men
(49.3%) were diagnosed by a health
professional as having a chronic physical
illness and 4.1% of all men were diagnosed
as having a mood disorder. If these conditions were independent, the probability
that they co-occur by chance alone would
be 2.02% (49.3% x 4.1%). However, the
observed probability of their co-occurrence
was 3.18% indicating an association bet-
Prevalence of chronic physical illness and mood disorders by gender,
Ontario 2005 (sample size = 41 701)
Chronic physical condition
Number (%)
Mood disorders
652 (6.5%)
198 (1.9%)
850 (4.1%)
9 453 (93.5%)
10 191 (98.1%)
19 644 (95.9%)
10 105
10 389
20 494
1 319 (10.5%)
334 (3.8%)
1 653 (7.8%)
11 209 (89.5%)
8 345 (96.2%)
19 554 (92.2%)
12 528
8 679
21 207
Number (%)
1 971 (8.7%)
532 (2.8%)
2 503 (6.0%)
20 662 (91.3%)
18 536 (97.2%)
39 198 (94.0%)
22 633
19 068
41 701
Chronic Diseases in Canada
ween the two conditions. Similarly, the
observed probability of chronic physical
illness and mood disorders in women was
found to be 6.22% compared with a
probability of 4.61% of the two conditions
occurring by chance alone.
The prevalence rates of comorbid mood
disorders among individuals with different
demographic and health characteristics are
presented in Table 2. Higher prevalence of
mood disorders were found among women,
individuals 30 to 49 years of age, lower
income groups, Canadian-born, and individuals who were divorced, separated,
widowed or never married. The highest
proportions of individuals with mood
disorders were found among individuals
who were diagnosed with chronic fatigue
syndrome, fibromyalgia, bowel disorder,
stomach or intestinal ulcers, chronic
bronchitis or those suffering from the
effects of a stroke.
Bivariate chi-square tests indicated that
the prevalence of comorbid mood disorders
was significantly associated with the
individual’s gender, age, marital status,
immigration status, and education and
income levels. Hence, these variables were
used in a logistic regression analysis to
predict presence of mood disorders in the
physically ill. Logistic regression analysis
results shown in Table 3 indicate that the
odds of having a mood disorder were
higher among women, the single, those
living in poverty, the Canadian-born and
those between 30 and 69 years of age.
Chronically ill woman were 65% more
likely than chronically ill men to have
mood disorders. The odds of suffering
from mood disorders among chronically ill
individuals 30 to 49 years of age were
almost three times higher than the odds
for individuals 70 years or older. Those
who were 50 to 69 years old had more than
twice the odds of suffering from mood
disorders than individuals 70 years of age
or older. The odds of having mood
disorders for those in the lower 30% of
income distribution were twice the odds of
high income individuals. Respondents
who were divorced, separated, widowed
or never married were 50% more likely to
have mood disorders than those with
partners. Education level was not a
significant predictor of mood disorders.
Table 4 includes a comparison of quality of
life measures in individuals with chronic
physical conditions with and without
mood disorders. Data presented in this
table show that 41.7% of men with mood
disorders had disability days in the two
weeks prior to the interview, compared
with only 17.7% without mood disorders
(p < 0.0005). Forty-four percent of women
with mood disorders reported disability
days compared with 22.0% without mood
disorders (p < 0.0005). Thirty-six percent
Prevalence of mood disorders in individuals with chronic conditions by socio-demographic characteristics,
Ontario 2005 (sample size = 22 633)
Total Number
% With mood
10 105
12 528
3 721
30 to 49
7 512
50 to 69
7 699
Married/common law
% With mood
Chronic condition:
Age group, years:
12 to 29
Total Number
Food allergies
2 108
Other allergies
7 636
3 325
7 140
3 701
Other back problems
8 089
14 803
High blood pressure
6 365
7 811
Migraine headaches
4 721
1 009
Education level:
Less than second. school
4 975
2 014
Second. school grad.
3 670
1 621
11 571
Some post-secondary
Post-secondary graduate
Heart disease
6 421
Stomach/intest. ulcers
15 500
Effects of smoke
6 127
7 727
5 490
Thyroid condition
Income within Ontario:
1 296
Bowel disorder
1 672
1 861
Chronic fatigue syndrome
Chronic Diseases in Canada
1 997
2 183
Vol 28, No 4, 2008
Socio-demographic factors associated with comorbid mood disorders
among individuals with chronic physical illness, Ontario 2005
(sample size = 19 213)
Odds ratio (95% CI)
1.65 (1.49, 1.84)
< 0.0005
12 to 29
1.42 (1.14, 1.76)
< 0.0005
30 to 49
2.89 (2.40, 3.49)
< 0.0005
50 to 69
2.31 (1.90, 2.79)
< 0.0005
Age group, years:
70+ (reference)
1.51 (1.35, 1.70)
Married/common law (reference)
< 0.0005
0.64 (0.57, 0.73)
Canadian-born (reference)
< 0.0005
Income within Ontario:
1.96 (1.70, 2.25)
1.38 (1.20, 1.58)
< 0.0005
High (reference)
of chronically ill men and 43.7% of chronically ill women with mood disorders
reported needing help in activities of their
daily living such as preparing meals,
getting to appointments, doing housework,
personal care or moving about inside their
home. In contrast, only 15.2% of men and
25.1% of women with no comorbid mood
disorders reported experiencing such
limitations. Data in Table 4 also show that
14.9% of chronically ill men with mood
disorders reported having suicidal thoughts
in the 12 months prior to the interview
compared with only 1% of those without a
mood disorder (p < 0.0005). In women,
the percentages were 11.7% in those with
a mood disorder compared to 0.7% without
a mood disorder (p < 0.0005). Using the
sampling weights published by statistics
Canada, it can be estimated that among
Ontarians with comorbid mood disorders
and chronic physical illness in 2005,
approximately 24 000 men and 39 000
women had suicidal thoughts.
The rates of mood disorders in chronically
ill individuals living in Ontario in 2005
ranged between 7.9% in those with
cataracts and 37.2% in those with chronic
fatigue syndrome compared with 2.8% in
individuals with no chronic physical conditions. The prevalence of mood disorders in
chronically ill men was more than threefold than in men with no chronic illness.
In women with one or more chronic
physical condition, it was almost threefold than in non chronically ill women.
Prevalence of mood disorders were highest
among individuals who had been diagnosed as having fibromyalgia, bowel
disorder, stomach or intestinal ulcers,
chronic bronchitis or those suffering from
the effects of a stroke. Rates of mood
disorders in individuals with heart disease,
cancer and diabetes observed in this study
were lower than those reported in the
literature.4 This is not surprising since this
is a community sample that included those
individuals with cancer in remission.
Individuals with comorbid mood disorders
were more likely to be female, middle
aged, living in poverty, without a partner
and born in Canada. These characteristics
are similar to those of individuals with
depression in the general population.16
Findings of this investigation showed a
highly significant impact of the presence of
Relationship between comorbid mood disorders and quality of life of individuals with
chronic physical conditions by gender, Ontario 2005 (sample size = 22 633)
Mood disorder
number (%)
number (%)
Chi-square (df = 1)
Disability days in last 2 weeks ≥ 1 day
272 (41.7%)
1 671 (17.7%)
< 0.0005
Need help with daily activities
234 (35.9%)
1 438 (15.2%)
< 0.0005
97 (14.9%)
95 (1.0%)
< 0.0005
9 453
Disability days in last 2 weeks ≥ 1 day
575 (43.6%)
2 469 (22.0%)
< 0.0005
Need help with daily activities
576 (43.7%)
2 814 (25.1%)
< 0.0005
Suicidal thoughts
154 (11.7%)
83 (0.7%)
< 0.0005
1 319
11 209
Suicidal thoughts
Total – Men
Total – Women
Vol 28, No 4, 2008
Chronic Diseases in Canada
mood disorders on all measures of quality
of life for the chronically ill. The presence
of mood disorders was associated with
more than double the risk of short-term
disability for men and almost double the
risk for women. It was also associated with
double the proportions of chronically ill
men and women requiring help with their
daily living activities. In addition, the
presence of comorbid mood disorders was
associated with an increase in the proportion of those having suicidal ideation –
15-fold in men and 17-fold in women.
A bidirectional relationship between mood
disorders and chronic physical illness has
been proposed that may explain their high
comorbidity.4 A diagnosis of a disabling
physical illness and the associated decline
in physical health may cause enough
distress to trigger a depressive episode in
vulnerable persons. On the other hand,
research is discovering that depression
itself may act as a risk factor for a variety
of chronic illnesses. Goodman and
Whitaker (2002)17 noted that patients with
major depression had higher rates of
unhealthy behaviors such as smoking and
overeating, which may lead to higher
incidence of diabetes and heart disease.
Depression has also been shown to be an
independent risk factor for type 2 diabetes
mellitus.18 Perretta, et al.19 suggested that
depression and mania may act as risk
factors for HIV infection by promoting
high-risk behaviors. Large population
studies indicate that depressed mood or
stressful life events may increase the risk
of cancer.4 In addition, the presence of
depression can hinder compliance to treatment and weaken cognitive function, thus,
adversely affecting patients’ prognosis and
increasing their morbidity.
Ample research has focused on trying to
understand the mechanisms of the relationship between depression and cardiovascular
disease. In addition to promoting unhealthy
behaviors and noncompliance with cardiac
rehabilitation and medical regimens, some
of the possible biologic mechanisms that
may explain the increased risk associated
with depression in heart disease patients
Chronic Diseases in Canada
are that depressed patients have decreased
heart rate variability,20 increased platelet
aggregation21 and higher levels of inflammatory risk markers.22 Other mood disorders
may also act as risk factors for physical illnesses and/or adversely affect their prognosis; however, the underlying mechanisms
are not as extensively researched as those
of depression.
This study has a number of limitations.
The identification of individuals as having
mood disorders and/or chronic physical
illness was not done by clinicians. The
survey did not include individuals living in
nursing homes, mental institutions or
chronic care hospitals, thus, the data
underestimate the prevalence of both
mood disorders and chronic illnesses.
Participants were not asked about each
mood disorder separately. Instead, major
depressive disorder, bipolar disorder,
mania or dysthymia were combined in one
question. Consequently, the prevalence
and impact of each of these disorders on
quality of life cannot be inferred. Further,
the cross-sectional nature of the data
precluded an examination of the temporal
sequence of onset of mood disorders and
chronic physical conditions. Given these
limitations, the present study determined
the prevalence and correlates of mood
disorders among individuals with physical
chronic illness using the most up-to-date
data available on a representative sample
of Canadians living in Ontario. In addition,
survey respondents were asked to only
report mood disorders and physical conditions that had been diagnosed by a health
professional. To the authors’ knowledge,
this is the first study to examine the
associations between comorbid mood
disorders and short-term disability, limitations of daily activities and suicidal
ideation in this population.
Results of this study indicate that detecting
and treating depression is as important as
treating the physical illness for maintaining
quality of life and helping the individual
cope with and manage the physical illness.
The identification of subgroups of physically ill individuals who are at high risk of
suffering from mood disorders provide
clinicians with important knowledge for
targeting these vulnerable groups for the
purpose of early diagnosis and intervention. Health care providers are urged to
proactively screen chronically ill patients
for mood disorders, particularly among the
subgroups found to have elevated risk for
these disorders. Based on a survey done by
the Depression and Bipolar Support
Alliance,23 it was concluded that most
primary care physicians lacked knowledge
about mood disorders in general and
bipolar disorder in particular. Findings of
this study stressed the importance of
between mental and physical health and of
including the treatment of the patient’s
mood disorders as an integral part of the
treatment plan of the chronic condition.
It is also imperative to encourage individuals with physical illness to communicate
their psychological symptoms to their
health care providers and not to accept
such symptoms as natural consequence to
their physical condition. Barriers to seeking
help for mental health, whether perceived
or real, should be recognized and removed.
The collection of data on cultural and
social factors, such as beliefs about stigma
and shame in association with mental
illness and trust in health care providers, is
essential for developing innovative and
creative strategies for promoting mental
health. Culturally appropriate psychoeducational programs could help Canadians
learn to recognize the symptoms of mood
disorders and be aware of the importance
and potential benefits of treatment. Such
programs could promote self-identification and lead to early diagnosis and
This research was based on data collected
and made available by Statistics Canada.
The views and opinions expressed do not
represent the views of Statistics Canada.
Vol 28, No 4, 2008
Statistics Canada. Canadian Community
Health Survey, Mental Health and Well
Being. Public-Use Microdata Documentation.
Ottawa: Statistics Canada, 2002.
Ades P, Savage P, Tischler M, Poehlman E,
Dee J, Niggel J. Determinants of disability
in older coronary patients. Am Heart J.
Ferketich A, Schwartzbaum J, Frid D,
Moeschberger M. Depression as an antecedent to heart disease among women and
men in the WHANES I study. National
Health and Nutrition Survey. Arch Intern
Med. 2000;160:1261-1268.
Michaud CM, Murray CJ, Bloom BR.
Burden of disease: Implications for future
research. JAMA. 2001;285:535-539.
Public Health Agency of Canada. The
Human Face of Mental Health and Mental
Illness in Canada. Ottawa: Public Health
Agency of Canada, 2006. Cat. No. HP519/2006E.
Evans DL, Charney DS, Lewis L, Golden
RN, Gorman JM, Krishnan KR, Nemeroff
CB, Bremner JD, Carney RM, et al. Mood
disorders in the medically ill: Scientific
review and recommendations. Biol
Psychiatry. 2005;58:175-189.
Fuller-Thomson E, Sulman J. Depression
and inflammatory bowel disease: findings
from two nationally representative
Canadian surveys. Inflamm Bowel Dis.
Olfson M, Klerman GL. Depressive symptoms and mental health service utilization
in a community sample. Soc Psychiatry
Psychiatr Epidemiol. 1992;27(4):161-167.
Vali F, Walkup J. Combined medical and
psychological symptoms: Impact on
disability and health care utilization of
patients with arthritis. Med Care. 1998;
Vol 28, No 4, 2008
10. Simon G, Von Kroff M, Rutter C, Paterson
D. Treatment process and outcomes for
managed care patients receiving new antidepressant prescriptions from psychiatrists
and primary care physicians. Arch Gen
Psychiatry. 2001;58:395-401.
Katon WJ. Clinical and health services
relationships between major depression,
depressive symptoms and general medical
illness. Biol Psychiatry. 2003;54:216-226.
12. Stephens T & Joubert N. The economic
burden of mental health problems in Canada.
Chronic Dis Can. 2001;22(1):18-23.
13. Starkes J, Poulin C, Kisely S. Unmet need
for the treatment of depression in Atlantic
Canada. Can J Psychiatry. 2005;50:580-90.
14. Gadalla TM. Comparison of users and nonusers of mental health services among
depressed women: A National Study.
Women&Health. 2008;47(1):1-19.
15. Statistics Canada. Canadian Community
Health Survey, Cycle 3.1, 2005. Public-Use
Microdata Documentation. Ottawa: Statistics
Canada, 2006.
16. Patten SB. Incidence of major depression
in Canada. Can Med Assoc J. 2000;
17. Goodman E, Whitaker R. A prospective
study of the role of depression in the
development and persistence of adult
obesity. Pediatrics. 2002;110:497-504.
18. Kawakami N, Takatsuka N, Shimizu H,
Ishibashi H. Depressive symptoms and
occurrence of type 2 diabetes among
Japanese men. Diabetes. 1999;22:1071-1076.
19. Perretta P, Akiskal HS, Nisita C, Lorenzetta
C, Zaccagnini E, Della Santa M, Cassano
GB. The high prevalence of bipolar II and
associated cyclothymic and hyperthymic
temperaments in HIV-patients. J Affect
Disord. 1998;50:215-224.
20. Gormen JM, Sloan RP. Heart rate variability
in depressive and anxiety disorders. Am
Heart J. 2000;140(4 suppl):77-83.
Whyte E, Pollock B, Wagner W, Mulsant B,
Ferrel R, Mazumdor S, Reynolds C 3rd.
Influence of serotonin-transporter-linked
promoter region polymorphism on the
platelet activation in geriatric depression.
Am J Psychiatry. 2001;158:2074-2076.
22. Miller GE, Stetler CA, Carney RM,
Freedland KE, Banks WA. Clinical
depression and inflammatory risk markers
for coronary artery disease. Am J Cardiol.
23. Depression and Bipolar Support Alliance
(2002). General Public Survey (Report).
Chicago, IL: DBSA.
Chronic Diseases in Canada
Costs associated with mood and anxiety disorders, as
evaluated by telephone survey
SB Patten, MD, PhD (1); JVA Williams, MSc (1); C Mitton, MSc (2)
Costing studies are central to health policy decisions. Available costing estimates for mood
and anxiety disorders in Canada may, however, be out of date. In this study, we estimated
a set of direct health care costs using data collected in a provincial telephone survey of
mood and anxiety disorders in Alberta. The survey used random digit dialing to reach a
sample of 3394 household residents aged 18 to 64. A telephone interview included items
assessing costs without reference to whether these were incurred by the respondent,
government or a health plan. The survey interview also included the Mini Neuropsychiatric
Diagnostic Interview (MINI). Costs for antidepressant medications appear to have
increased since the last available estimates were published. Surprisingly, most medication
costs for antidepressants were incurred by respondents without an identified disorder.
Also, an unexpectedly large proportion of medication costs were for psychotropic
medications other than antidepressants and anxiolytic-sedative-hypnotics. These results
suggest that major changes have occurred in the costs associated with antidepressant
treatment. Available cost-of-illness data may be outdated, and some assumptions made
by previous studies may now be invalid.
Key words: Cross-sectional studies, Depressive Disorders, Costs and Cost Analysis
Depressive disorders present an important
challenge to population health.1 According
to the Global Burden of Disease Project,
unipolar major depression is the 4th leading
contributor to disease burden world-wide,
ranking second in developed countries
such as Canada.2 The impact of this
condition on population health relates
partially to its high prevalence: approximately 5% of Canadians experience an
episode of major depression in any given
year.3 However, its impact is magnified by
a peak prevalence in the 25-44 year age
group4 which is a critical period for
education, establishment of relationships
and economic productivity. Major depression is also a recurrent condition, and one
that is frequently associated with psychiatric and medical comorbidity.5,6 All of
these factors tend to magnify its impact on
population health. Finally, major depression has an impact on mortality.7
Around the world, a variety of costing
studies for depressive disorders have been
carried out. Typically, the goal of these
studies has been to estimate overall costs,
including direct and indirect treatment costs.
Population-based costing studies typically
use data from a variety of sources such as
government reports, administrative data and
national surveys. This is because data
collected from any particular source (e.g. a
national survey) are unlikely to be
comprehensive enough for detailed costing.
Recently, Thomas and Morris8 published
cost-of-illness estimates for depression in
England. Their study integrated data from
a large number of sources and provides
what has been considered the best
available estimates for Europe.9 However,
the approach taken to assess some costing
elements was crude. For example, in the
assessment of medication costs, treated
prevalence was estimated using a formula
that began with administrative data for
physician visits. The estimated number of
people making visits for depression was
multiplied by the proportion of antidepressant medications prescribed nationally in
specific antidepressant categories, and in
turn, by the average costs of prescriptions
in those categories. This approach assumes
that the total cost can be conceptualized as
the treated prevalence multiplied by the
cost per treated person. The approach
seems reasonable on the surface, but does
not consider that antidepressant treatment
is often continued long after the remission
of depressive episodes.10
Stephens and Joubert11 estimated the total
burden of major depression and distress
in Canada in 1998, also using data from a
variety of sources. Their overall cost
estimate was $14.4 billion. Their analysis
was based partially on data from the
National Population Health Survey (NPHS)
and partially on data collected by the 1998
Economic Burden of Illness in Canada
(EBIC) project (http://www.phac-aspc.
All of the data on medication costs came
from EBIC. In turn, the EBIC estimates
were taken from a cost tracking database
at the Canadian Institute for Health Information (CIHI) containing cost estimates
from a private-sector company: IMS Health
Canada. This approach is useful because
IMS Health Canada databases can identify
both the total volume of medication sales
and the proportion of treatment recom-
Author References
1 University of Calgary, Department of Community Health Sciences, Calgary, Alberta
2 Health Studies, University of British Columbia, Kelowna, British Columbia
Correspondence: Scott B. Patten, University of Calgary, Department of Community Health Sciences, 3330 Hospital Drive NW, Calgary, Alberta T2N 4N1, Tel:
(403) 220-8752, Email: [email protected]
Chronic Diseases in Canada
Vol 28, No 4, 2008
mendations for specific indications.
According to EBIC, $108 million was spent
on drugs for anxiety states, and $252
million, for depressive disorders. However,
this analysis is now 10 years old, and some
of its estimates are nearly 15 years old.
The ensuing period of time has seen
increasing recognition of major depression
as a health problem, likely diminishing
stigma related to the condition and the
emergence of new pharmacological and
non-pharmacological treatments.
In 2003, sponsored by the Alberta Mental
Health Board, a provincial mental health
survey was conducted in Alberta, called
the Alberta Mental Health Survey.12 In
2004 another initiative, sponsored by
Alberta Health and Wellness and
Administered by the Institute of Health
Economics, called the Alberta Depression
Initiative, was established. As a component
of this latter initiative, a second provincial
telephone survey focusing on mental
health was carried out in the following
year, this time with an emphasis on
depression. Some of the survey results are
relevant to future costing studies for mood
and/or anxiety disorders in Canada and
are summarized in this report.
Survey Data Collection
Alberta is a Canadian province with a population of 3.3 million residents, dispersed
over an area of 661 190 km2. Telephone
survey methods were the most feasible
strategy for obtaining an approximately
representative sample in this geographically
dispersed population. Approximately 2/3 of
the Alberta population resides in two cites:
Edmonton and Calgary. The sampling procedure was stratified so that 1/3 of the
sample would come from each of these
cities, with the balance coming from
remaining rural areas. Data collection was
restricted to household residents between
the ages of 18 and 64 with a residential
telephone line.
Data collection was carried out by the
population survey unit of the Quality,
Safety and Health Information (QSHI)
Vol 28, No 4, 2008
portfolio in the Calgary Health Region
( during 2005.
A listing of provincial, residential telephone
numbers was initially sampled and the last
digit was substituted with a random integer
in order to increase coverage of unlisted
numbers13,14 and to avoid bias that might
be introduced if households with unlisted
numbers differed from those with listed
numbers. When a household was reached,
a pseudo-random procedure, the “last
birthday method” was used to randomly
select a single subject from the household.
As many as nine call-backs were made in
an effort to reach all sampled individuals.
These calls were distributed over working
hours, evening and weekends.
The Mini Neuropsychiatric Interview
(MINI), a brief structured diagnostic interview,15,16 was used as a diagnostic indicator
for Major Depression (MD) and a set of
other common mood and anxiety disorders.
For major depressive episodes, past-14-day
prevalence was assessed (essentially, the
point prevalence for this disorder), for
dysthymia, past-2-year prevalence was
assessed. For panic disorder, agoraphobia,
generalized anxiety disorder and social
phobia, current prevalence was also assessed.
Since it has been shown that differences
between survey instruments often relate to
the role of “clinical significance” probes in
the scoring algorithms,17 we incorporated an
interference item asking respondents
whether their psychiatric symptoms interfered with their life. Episodes were considered
clinically significant if respondents reported
“a lot” of interference.
A pharmacoepidemiology module and a
module designed to identify non-pharmacological treatment were also included in the
interview. The pharmacoepidemiology
module operated with a cyclical item-flow
structure, initially asking about medications
taken for the treatment of broadly defined
but relevant symptoms (“Do you currently
take any prescription medications for
anxiety, depression, stress, energy levels,
sleeping, pain management, fibromyalgia
or migraine headaches?”), and then
looping through each reported medication
with a series of items inquiring about the
number and size of tablets, reasons for use
of the medication and duration of use.
Respondents were prompted to report
information directly from their pill bottle
labels to ensure accuracy of this information.
Dosage was determined by combining
information about the size(s) and number
of relevant tablets or capsules taken,
including pro re nata (prn) schedules. The
dosages were recorded in milligrams per
day. The survey interview also included
items evaluating the frequency of use of
health services. These items were typically
paired, with an initial item asking about use
of the service and a second quantifying the
frequency of use. For example, one item
asked: “In the past 6 months have you
consulted with a specialist physician other
than a general practitioner, family doctor,
emergency room physician or psychiatrist
about your health?” If the answer was “yes”
this was followed by “How many times?”
Similar items assessed visits with family
physicians, psychiatrists, psychologists and
social workers, as well as emergency and
walk-in clinic attendance, radiological
procedures, and hospitalizations.
Costing Procedures
We obtained costs per milligram for each
medication from the provincial drug plan
costing guide (
pdf). In situations where various dosage
forms had different per milligram costs, we
calculated the cost for the lowest milligram
dosage for each drug. The costs associated
with generic preparations were used when
a generic preparation was available, since
pharmacists can substitute the lowest cost
alternative when filling a prescription.
When a person was taking more than one
drug from a particular class, the costs were
estimated in the same way for each drug,
and the costs of individual drugs were
added together. For ease of interpretation,
the daily costs were multiplied by 365 to
project an annual medication cost. This
approach seemed reasonable since 80% of
survey respondents who reported taking
antidepressants in the survey reported
taking them for longer than one year. The
same procedure was followed for
anxiolytic-sedative-hypnotic and antipsychotic medications.
Chronic Diseases in Canada
For utilization of professional services, fee
codes for the various professional groups
were used. For physicians, fee codes from
the Alberta Health Care Insurance Plan
(AHCIP) were used. For family physicians,
a basic code for an office visit not requiring
a complete history and evaluation was used
(fee code 03.03A). Emergency room visits
were handled separately, using an estimated
cost per single visit at twice the fee for a
visit to a family physician. For specialist
physicians, the fee code for a single referred
consultation visit for 15 minutes was used
(fee code 03.03FA). Fees associated with 45
minutes of clinical management time were
used for psychiatrists (skill level code
08.19A) and for psychologists, fee schedules
published by the Provincial Psychological
Association were used.
It was not feasible to itemize radiological
procedures and laboratory tests reported
by survey participants. Instead, representative mid-point costing estimates were
identified for common or typical procedures
within each category. In this study these
were: a chest X-ray, basic CT of the head
and a basic MRI of the head. The per
procedure costing estimates used in the
study came from a costing guide produced
by the Calgary Health Region. This guide
is intended to ensure cost recovery in
research projects, including materials and
supplies and personnel-related costs.
Similarly, mean per diem costs associated
with hospitalizations from across the
province used estimates developed by the
Calgary Health Region for administrative
purposes. The documents from the Calgary
Health Region were considered to be a
reasonable source since the Region
includes the city of Calgary, but also rural
areas stretching southwest of the city to
the border with British Columbia.
For each category, a cost projection to the
total population was made. The weighted
prevalence for the mood and/or anxiety
disorder diagnostic category was multiplied
by the total provincial population in the
studied age range, the proportion using
the specific procedure, treatment or service
and by the mean cost associated with it to
Chronic Diseases in Canada
produce a final cost estimate. Sampling
weights were calculated as the inverse of
an estimated selection probability within
age and sex categories in each of the three
sampling regions for the provincial population. The resulting weight was further
adjusted by multiplying it by the ratio of
the number of eligible residents in each
household divided by the number of telephone lines going into the household. The
weights were treated as relative weights in
the analysis.
In total, 18 113 telephone numbers were
called. More than half of these were
disqualified from the sample for the
following reasons: 1663 households had
no eligible residents, 329 were blocked
calls, 846 reached only answering machines
or voice mailboxes (i.e. no person could be
contacted with call-backs), 1747 were
business lines, 3333 were not in service,
1385 reached fax machines, 378 reached
non-English-speaking households, 845
were never answered during the call-back
protocols, 11 met with hostile interruptions
and 79 were disqualified for miscellaneous
reasons. There were 7497 calls that successfully reached eligible households. At
the household level, there were 3443
refusals (45.9%). Of the 4054 households
from which respondents could be selected,
there were 635 individual refusals (15.7%).
Of the 3419 consenting respondents, interviews were completed in all but 25 (0.7%),
so that 3394 interviews were completed.
After checks for data completeness and
accuracy, 45 records were removed from
the data set because of concerns about
data quality. The final analysis included
data collected from the remaining 3345
individuals (82.5% of individuals who
were invited to participate and provided
adequate data). If the response rate is
calculated using the number of eligible
households in the denominator, however,
it is 44.6%. Using the Marketing Research
and Intelligence Association (MRIA) method
of calculation (http://www.tpsgc-pwgsc., 7628
of the 18 113 can be classified as “out of
scope” leading to a response rate of 31.9%.
There were 168 survey participants with a
mood disorder (current major depression or
dysthymia) according to the MINI, leading
to a prevalence estimate of 4.6%. This
estimate is consistent with existing literature. The 30-day prevalence of major
depression in the Canadian Community
Health Survey, Mental Health and Wellbeing
(CCHS 1.2) was 1.8%.4 The CCHS 1.2 did
not assess dysthymic disorder, but a
systematic review by Waraich estimated a
past-year prevalence of approximately 2%.18
One hundred and ninety two respondents
had one or more anxiety disorders, leading
to a prevalence estimate of 6.3%. This is
slightly higher than the CCHS 1.2 past-year
estimate of anxiety disorder prevalence
(4.7%),19 but the CCHS 1.2 did not assess
generalized anxiety disorder, which has an
annual population prevalence of 3.1% in
the US.20 Most respondents with disorders
had comorbid mood and anxiety disorders.
Only 2.2% had a mood or anxiety disorder
without having both. 4.1% had comorbid
mood and anxiety disorders. Since the MINI
is a brief screening instrument, its ability to
distinguish between mood and anxiety
disorders is questionable. The analysis
therefore focussed on two groups: respondents with a mood or anxiety disorder (but
not both) and respondents with a comorbid
mood and anxiety disorder.
With respect to medication use at the time
of the survey, 7.4% of the population
reported taking an antidepressant, 3.1%
reported taking a sedative hypnotic medication and 1.5% took another medication
potentially related to the management of
mood disorders: mood stabilizers, psychostimulants or atypical antipsychotic medications. Although the latter group of
medications are pharmacologically distinct,
they were grouped together in the subsequent analysis because there were insufficient numbers to examine these categories
separately. It should also be noted that the
question stems indicated that the medications of interest were those taken for
psychotropic purposes, and this category
Vol 28, No 4, 2008
would not include, for example, anticonvulsant mood stabilizers taken for the treatment of epilepsy.
Of the 2.2% of the population with noncomorbid mood or anxiety disorders, 20.4%
were taking one or more antidepressant
medications. Of those with comorbid mood
and anxiety disorders, 44.4% were taking
one or more antidepressant medications. In
the group with no detected mood or anxiety
disorders, 5.6% were taking an antidepressant. The estimated mean annual cost for
antidepressants in each of these groups is
summarized in Table 1. At the time of the
survey, there were 2 105 167 provincial
residents in the 18 to 64 age range. The
average cost per person multiplied by the
estimated prevalence and number of people
in the province in each diagnostic category
led to a total cost estimate of $124 million
for antidepressants. This included $7 million
for people with a noncomorbid mood or
anxiety disorder, $35 million for people with
comorbid disorders and $83 million for those
with no MINI-detected disorder.
Although the frequency with which antidepressants were taken by respondents
without detected mood or anxiety disorders
was relatively low at 5.6%, and while the
projected annual per person cost in this
group was slightly lower than the group
with MINI-detected disorders (see Table
1), the overall cost in this group was
highest because the number of people in
this category was the highest. This suggests
that most of the costs associated with
antidepressant treatment now occur in the
continuation and maintenance phases10 of
treatment since these respondents no
longer met symptomatic criteria for mood
or anxiety disorders. The result is in itself
not necessarily surprising since acute
treatment with antidepressants usually
takes 6 to 8 weeks, whereas maintenance
treatment lasting one year or longer is
generally recommended, and indefinite
treatment may be indicated in those with
highly recurrent disorders.10 There are
however, other possible explanations for
this result, as discussed below.
A similar pattern was seen for sedativehypnotic medications (see Table 1). In
people with mood or anxiety disorders, the
Vol 28, No 4, 2008
frequency of use of these medications was
much higher than that of respondents
without these disorders. However, most
people taking these medications did not
have a mood or anxiety disorder. Hence,
most of the costs occurred in the no-disorder
group. The estimated total cost of sedativehypnotic use in the population was $8
million, much lower than the cost of
A variety of other medications were reported
by the survey respondents: lithium (n = 7),
carbamazepine (n = 3), valproate (n = 5),
gabapentin (n = 9), lamotrigine (n = 1),
topiramide (n = 20), dexamphetamine
(n = 6), chlorpromazine (n = 1), olanzapine (n = 10), quetiapine (n = 8), clozapine
(n = 2) and risperidone (n = 20). Ninetytwo respondents reported taking one or
more of these medications. Unfortunately,
only 69 of these respondents (75%) were
willing or able to provide dosage information. Considered as a group, an estimated
1.5% of the total population was taking one
or more medications from this category.
The frequency of use in respondents with
non-comorbid mood or anxiety disorders
was similar to that of the total population
(see Table 1), but the frequency was much
higher in the group with comorbid disorders
(16.4%). Approximately 1% of the population without a mood or anxiety disorder
was taking one of these medications –
approximately consistent with the prevalence of psychotic disorders in the
population. The projected total cost was
$38 million, $25 million of which was
accounted for by the group with no detected
disorder. These results suggest, however,
that a sizable proportion of treatment costs
for mood and anxiety disorders may now
occur outside of the pharmaceutical classes
traditionally believed to be most important
in the treatment of mood and anxiety
disorders: antidepressant medications and
sedative-hypnotic medications.
Table 2 presents the costs associated with
utilization of services provided by mental
health professionals. The frequency with
which the services of psychiatrists and
psychologists were utilized was much
higher in people with mood and anxiety
disorders, especially in the comorbid
category. Having a mood or anxiety
disorder was associated with a three-fold
increase in these costs, whereas a more
than eight-fold increase was seen in the
comorbid group. However, most of the
costs for psychologists occurred in the
group with no MINI-defined disorder,
suggesting that these professionals may
often be engaged in treatment activities
not specifically directed at these disorders,
e.g. marital therapy. Table 3 presents the
costs associated with service utilization for
family physicians and specialist physicians
other than psychiatrists. The respondents
with mood and anxiety disorders were
more likely to see a physician, and to have
more visits. Both factors contributed to
higher per-respondent costs in the group
with mood and anxiety disorders.
Costing data for diagnostic tests are presented in Table 4. In the case of radiological
procedures, the frequency of testing (and
hence, per-subject costs according to the
procedures employed in this study) did not
differ by disorder category. However, the
proportion of subjects in the various subgroups having these tests was higher in
subjects with mood and anxiety disorders.
As shown in Table 5, the costs associated
with emergency room visits and hospitalizations were also elevated in respondents
having comorbid mood and anxiety
Many of the results presented here are
consistent with an existing literature of
costing studies for depressive disorders.
They illustrate that costs associated with
medication use and utilization of services
are increased in people with these disorders. As expected, this was true both for
measures of mental health care and for
general medical care. An important interpretive issue is that mood disorders are
strongly associated with long-term medical
conditions,21-23 such that these increased
costs may reflect an indirect effect of nonpsychiatric conditions. When one considers
that approximately 10% of the Canadian
population live in Alberta, the results
suggest a substantial increase in the costs
of pharmacotherapy for depression since
the publication of the 1998 EBIC. In the
Chronic Diseases in Canada
EBIC, the total national costs for medications for mood and anxiety disorders was
only twice as high as that found in the
province of Alberta alone. In other words,
if one assumes that 10% of the $360
million estimated by the EBIC was spent
in Alberta, the provincial estimate would
be approximately $36 million; however,
the current study indicates that the cost of
antidepressant medications in Alberta may
be between 3 and 4 times higher than that.
The increase in cost is not unexpected, as
both the number of retail prescriptions for
medications and the average cost per
prescription are increasing in Canada.24
Although antidepressants and sedativehypnotic medications were the most
frequently taken medications, an appreciable
number of respondents reported taking
other medications such as atypical antipsychotics. These medications now appear
to make a substantial contribution to total
treatment costs. Future analyses should
include costs for atypical antipsychotics,
stimulants and mood stabilizers.
Another interesting finding is the high
proportion of costs in the no-disorder
group. It should be emphasized that the
MINI would not detect treated disorders
that are in remission, so in some proportion
of instances these costs may represent
appropriate use of the medications. However, it is likely that some of the medication
use in the no-disorder group is inappropriate (i.e. no disorder indicating treatment
is present). Symptoms of depression and
anxiety can occur in response to losses,
threats and stressors. The Canadian health
system is not well-structured for delivering
the brief psychological interventions that
would generally be appropriate in such
cases. In Canada, primary mental health
care is usually funded on a fee-for-service
model. Primary care practices in Canada
typically see large volumes of patients for
brief visits. Antidepressants may sometimes be prescribed inappropriately due to
a lack of time or expertise on the part of
primary care physicians. Similarly, the use
of sedative-hypnotic and atypical antipsychotic medications in the comorbid
group may partially reflect poor access to
non-pharmacologic treatments such as
sleep hygiene and relaxation techniques.
Finally, some proportion of medication use
in this group may be due to treatment for
conditions other than mood and anxiety
disorders. For example, antidepressants
are frequently used for chronic pain and
migraine prophylaxis.
The results may have important implications for future costing studies. For
example, the UK study discussed above8
used billing data to identify treated prevalence, but people on long-term maintenance
treatment may not have visits coded
specifically for a mood or anxiety disorder.
In fact, any approach that begins with
prevalence data is likely to miss a large
proportion of medication costs, since the
treated disorders may be in remission and
may also be missed in symptom-based
prevalence measures used in surveys.
There are several limitations associated
with this study. First, the study focused on
psychotropic medication use. Since some
chronic medical conditions are associated
with mood and anxiety disorders, it is
likely that increased costs associated with
non-psychiatric procedures and treatments
would occur in the subjects with mood
and anxiety disorders. It was not considered
feasible to record all medications being
Annual and six-month costs associated with medication use, by disorder category
Current comorbid
mood and anxiety
No current mood or
anxiety disorder
Prevalence 2.2%
Prevalence 4.1%
Prevalence 93.7%
% Utilizing
Mean cost/person
Provincial total cost
Current mood or
anxiety disorder
Provincial total cost
mood stabilizers,
% Utilizing
Mean cost/person
Provincial total cost
% Utilizing
Mean cost/person
* numbers too small to support estimation.
Chronic Diseases in Canada
Vol 28, No 4, 2008
Utilization of services provided by mental health specialists, by disorder category
(during the 6 months prior to the survey)
Current mood
or anxiety
mood and
No current
mood or
% Utilizing
Mean number of visits
Mean cost/person
Provincial total cost
% Utilizing
Mean number of visits
taken by the subjects during the telephone
interview, however. Second, the recording
of costs was certainly not exhaustive. In
particular, indirect costs were not assessed.
Whereas various categories of cost were
estimated in respondents with and without
evidence of a disorder, it does not follow
that the disorders are necessarily the cause
for the costs. For example, we looked at
the costs associated with hospitalization
for any reason. Since chronic illnesses are
associated with mood disorders, the hospital
costs may more closely reflect these medical
comorbidities than the depression itself.
Mechanisms linking depression to health
care costs are likely to be complex. For
example, a negative effect of depression on
self-management of chronic conditions
could lead to an increased frequency of
hospital admissions. It is doubtful whether
such costs can confidently be attributed to
a specific diagnostic category.
Additional limitations pertain to the use of
telephone survey methods. Declining
response rates to telephone surveys in
recent decades have heightened concerns
about vulnerability of such studies to bias
(see recent report by Public Works and
Government Services Canada at http:// Because brevity is needed in telephone survey interviews, the assessment
of utilization was necessarily crude. However, telephone-based data collection did
allow the collection of data from a large
Vol 28, No 4, 2008
Mean cost/person
Provincial total cost
sample dispersed across a large geographical region, and this included more detailed
information about medication use than
has generally been available in prior
costing studies. However, response rates
were less than ideal, raising questions
about the validity of the estimates. Also, the
interviewers encountered some difficulty
with the collection of detailed information
about dosages from respondents who were
taking multiple medications or medications
that involve variable dosages or prn schedules. A high frequency of missing data,
for example, in the case of medications
other than antidepressants and sedativehypnotics may limit the validity of those
Telephone survey methods are unlikely to
be sufficient, in themselves, as a source of
data for cost-of-illness studies. However,
as the results of this study show, data
collected in telephone surveys can help to
inform the planning and interpretation of
comprehensive cost-of-illness studies. The
focus of this study was restricted to direct
costs associated with mood and anxiety
disorders. A detailed cost of illness study
would also need to include indirect costs
such as diminished productivity due to the
effects of these disorders. Another limitation of the current study is the lack of
assessment of substance dependence and
abuse. Since this was not measured, it was
not possible to determine whether the
frequency of sedative-hypnotic use, in
particular, may have been related to abuse
or dependence on these substances.
Even though many of the cost estimates in
the current study were based on approximations, these approximations were made in
the same way across disorder categories,
which should facilitate valid comparisons.
However, the comparisons may be misleading if the assumptions employed are
variably valid across disorder groups. For
example, if hospital admissions for mood
Costs associated with service utilization, family and specialist physicians
(not including psychiatrists), by disorder category
(during the 6 months prior to the survey)
% Utilizing
Mean number of visits
Mean cost/person
Provincial total cost
% Utilizing
Mean number of visits
Mean cost/person
Provincial total cost
Current mood
or anxiety
mood and
No current
mood or
Chronic Diseases in Canada
Diagnostic testing, by disorder category (during the 6 months prior to the survey)
Current mood
or anxiety
mood and
No current
mood or
% Utilizing
Mean number of tests
Mean cost/person
Total cost
CT scan
% Utilizing
Mean number of tests
% Utilizing
Mean number of tests
Mean cost/person
Total cost
Mean cost/person
Total cost
Lab tests
% Utilizing
Mean number of tests
Mean cost/person
Total cost
and anxiety disorders are less expensive
than hospitalizations for other reasons,
bias could be introduced.
In Canada, future costing studies will need
to maximize the use of available data
about the frequency of maintenance treatment and adopt a broad view of drug costs
in order to obtain valid cost estimates. The
economic “landscape” associated with
mood and anxiety disorders appears to be
evolving over time and future costing
studies will need to accommodate these
new realities.
This project was funded by the Institute of
Health Economics through the Alberta
Depression Initiative. Dr. Patten is supported
by a Health Scholar award from the Alberta
Heritage Foundation for Medical Research
and is a Fellow with the Institute of Health
Economics. Dr. Mitton is supported by the
Michael Smith Foundation for Health
Research and holds a Canada Research
Chair in Health Care Priority Setting.
Emergency room visits and hospitalizations, by disorder status (during the 6 month prior to the survey)
Emergency room visits
Current mood or
anxiety disorder
Current comorbid mood
and anxiety disorders
No current mood or
anxiety disorder
Prevalence 2.2%
Prevalence 4.1%
Prevalence 93.7%
% with a visit
Mean number of visits
Mean cost/person
Total cost
% with a hospitalization
Mean number of hospital days
Mean cost/person
Total cost
Chronic Diseases in Canada
Vol 28, No 4, 2008
Insel TR, Fenton WS. Psychiatric epidemiology. It’s not just about counting anymore.
Arch Gen Psychiatry. 2005;62:590-3.
Ayuso-Mateos JL. Global burden of unipolar depressive disorders in the year 2000.
Global Burden of Disease Draft 28-05-03.
World Health Organization Global Program
on Evidence for Health Policy. 2003.
Statistics Canada. Canadian Community
Health Survey Mental Health and Well-being.
Ottawa: Statistics Canada, 2004 Mar 9.
Patten SB, Wang JL, Williams JVA, Currie
SR, Beck CA, Maxwell CJ, et al. Descriptive
epidemiology of major depression in
Canada. Can J Psychiatry. 2006;51:84-90.
Blazer DG, Kessler RC, McGonagle KA,
Swartz MS. The prevalence and distribution
of Major Depression in a national community sample: the National Comorbidity
Survey. Am J Psychiatry. 1994;151:979-86.
Kessler RC, Berglund P, Demler O, Jin R,
Koretz D, Merikangas KR, et al. The
epidemiology of major depressive disorder:
results from the National Comorbidity
Survey Replication (NCS-R). JAMA. 2003;
Wulsin LR, Vaillant GE, Wells VE. A
systematic review of the mortality of
depression. Psychosom Med. 1999;61:6-17.
Thomas CM, Morris S. Cost of depression
among adults in England in 2000. Br J
Psychiatry. 2003;183:514-9.
Andlin-Sobocki P, Wittchen H-U. Cost of
affective disorders in Europe. Eur J Neurol.
2005;12(Suppl. 1):34-8.
Vol 28, No 4, 2008
10. Kennedy SH, Lam RW, Cohen NL, Ravindran
AV, and the CANMAT Depression Work
Group. Clinical guidelines for the treatment
of depressive disorders. IV. Medications and
other biological treatments. Can J Psychiatry.
2001;46(Suppl 1):38S-58S.
Stephens T, Dulberg C, Joubert N. Mental
health of the Canadian population: a comprehensive analysis. Chronic Dis Can.
12. Patten SB, Adair CE, Williams JVA, Brant
R, Wang JL, Casebeer A, et al. Assessment
of mental health and illness by telephone
survey: Experience with an Alberta Mental
Health Survey. Chronic Dis Can. 2006;
18. Waraich PS, Goldner EM, Somers JM, Hsu L.
Prevalence and incidence studies of mood
disorders: a systematic review of the literature. Can J Psychiatry. 2004;49:124-38.
19. Kessler RC, Chiu WT, Demler O, Merikangas
KR, Walters EE. Prevalence, severity, and
comorbidity of 12-month DSM-IV disorders
in the National Comorbidity Survey Replication. Arch Gen Psychiatry. 2005 Jun;
20. Statistics Canada. Canadian Community
Health Survey: Mental health and wellbeing. The Daily, Wednesday, September
3, 2003. URL:
13. Mullet GM. The efficacy of plus-one dialing:
Self reported status. American Statistical
Association: Proceedings of the Section on
Survey Research Methods 1982;575-6.
Patten SB, Beck CA, Kassam A, Williams
JVA, Barbui C, Metz LM. Long-term medical
conditions and major depression: strength
of association in the general population.
Can J Psychiatry. 2005;50:195-202.
14. Potthoff RF. Telephone sampling in epidemiological research: to reap the benefits,
avoid the pitfalls. Am J Epidemiol. 1994;
22. Gagnon LM, Patten SB. Major depression
and its association with long-term
medical conditions. Can J Psychiatry.
15. Lecubier Y, Sheehan DV, Weiller E, Amorim
P, Bonora I, Sheehan KH, et al. The MINI
International Neuropsychiatric Interview
(MINI). A short diagnostic structured interview: reliability and validity according to the
CIDI. Eur Psychiatry. 1997;12:224-31.
23. Wells KB, Golding JM, Burnam MA.
Psychiatric disorder in a sample of the
general population with and without
chronic medical conditions. Am J
Psychiatry. 1988;145(8):976-81.
16. Sheehan DV, Lecubier Y, Sheehan KH,
Janavs J, Weiller E, Keskiner A, et al. The
validity of the Mini International Neuropsychiatric Interview (MINI) according to
the SCID-P and its validity. Eur Psychiatry.
24. Pharmaceutical trends: Top products,
classes and companies sourced from IMS
data. URL: http://www.imshealthcanada.
17. Narrow WE, Rae DS, Robins LN, Regier
DA. Revised prevalence estimates of mental
disorders in the United States: using a
clinical significance criterion to reconcile 2
surveys’ estimates. Arch Gen Psychiatry.
Chronic Diseases in Canada
Information for Authors
Chronic Diseases in Canada (CDIC) is a
quarterly scientific journal focussing on
the prevention
and control of noncommunicable diseases and injuries in
Canada. Its feature articles are peer
reviewed. The content of articles may
include research from such fields as
epidemiology, public/community health,
biostatistics, the behavioural sciences,
and health services or economics. CDIC
endeavours to foster communication on
chronic diseases and injuries among
public health practitioners, epidemiologists
and researchers, health policy planners
and health educators. Submissions are
selected based on scientific quality, public
health relevance, clarity, conciseness and
technical accuracy. Although CDIC is a
publication of the Public Health Agency
of Canada, contributions are welcomed
from both the public and private sectors.
Authors retain responsibility for the
contents of their papers, and opinions
expressed are not necessarily those of the
CDIC editorial committee nor of the Public
Health Agency of Canada.
Article Types
Peer-reviewed Feature Article: Maximum
4,000 words for main text body (excluding
abstract, tables, figures, references) in the
form of original research, surveillance reports,
meta-analyses or methodological papers.
Status Report: Describe ongoing national
programs, studies or information systems
bearing on Canadian public health (maximum
3,000 words). Abstract not required.
Workshop/Conference Report: Summarize significant, recently held events relating
to national public health (maximum 1,200
words). Abstract not required.
Cross-Canada Forum: For authors to
present or exchange information and
opinions on regional or national
surveillance findings, programs under
development or public health policy
initiatives (maximum 3,000 words).
Abstract not required.
Letter to the Editor: Comments on
articles recently published in CDIC will be
considered for publication (maximum
500 words). Abstract not required.
Book/Software Review: Usually solicited
by the editors (500B1,300 words), but
requests to review are welcomed. Abstract
not required.
Submitting Manuscripts
Submit manuscripts to the Editor-in-Chief,
Chronic Diseases in Canada, Public Health
Agency of Canada 130 Colonnade Road,
Address Locator: 6501G, Ottawa, Ontario
K1A 0K9, e-mail: [email protected]
Since CDIC adheres in general (section on
illustrations not applicable) to the
“Uniform Requirements for Manuscripts Submitted to Biomedical
Journals” as approved by the International
Committee of Medical Journal Editors,
Chronic Diseases in Canada:
Research. Intervention. Prevention.
The Public Health Agency’s peer-reviewed journal.
Subscribe today! Submit your manuscripts!
Special: Call for papers on chronic disease intervention research.
[email protected]
Chronic Diseases in Canada
Vol 28, No 4, 2008
authors should refer to this document for
complete details before submitting a
manuscript to CDIC (see <www.icmje.
Checklist for Submitting
Cover letter: Signed by all authors, stating
that all have seen and approved the final
manuscript and have met the authorship
including a full statement regarding any
prior or duplicate publication or submission
for publication.
First title page: Concise title; full names
of all authors and institutional affiliations;
name, postal and e-mail addresses, telephone and fax numbers for corresponding
author; separate word counts for abstract
and text.
Second title page: Title only; start page
numbering here as page 1.
Vol 28, No 4, 2008
Abstract: Unstructured (one paragraph,
no headings), maximum 175 words (100
for short reports); include 3B8 key words
(preferably from the Medical Subject
Headings (MeSH) of Index Medicus).
Text: Double-spaced, 1 inch (25 mm)
margins, 12 point font size.
Acknowledgements: Include disclosure of
financial and material support in acknowledgements; if anyone is credited in acknowledgements with substantive scientific
contributions, authors should state in cover
letter that they have obtained written
References: In AVancouver [email protected] (consult
a recent CDIC issue for examples);
numbered in superscript in the order cited
in text, tables and figures; listing up to six
authors (first three and Aet [email protected] if more);
without any automatic reference numbering
feature used in word processing; any
unpublished observations/data or personal
communications used (discouraged) to be
cited in the text in parentheses (authors
responsible for obtaining written permission); authors are responsible for verifying
accuracy of references.
Tables and Figures: Send vector graphics
only. Each on a separate page and in
electronic file(s) separate from the text
(not imported into the text body); as selfexplanatory and succinct as possible; not
too numerous; numbered in the order that
they are mentioned in the text; explanatory
material for tables in footnotes, identified
by lower-case superscript letters in alphabetical order; figures limited to graphs or
flow charts/templates (no photographs),
with software used specified and titles/
footnotes on a separate page.
Number of copies: If submitting by mail,
one complete copy, including tables and
figures; one copy of any related supplementary material, and a copy of the
manuscript on diskette. If submitting by
e-mail to [email protected],
please fax or mail the covering letter to the
address on the inside front cover.
Chronic Diseases in Canada
Was this manual useful for you? yes no
Thank you for your participation!

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project

Download PDF