---

How to obtain more information
Specific inquiries about this product and related statistics or services should be directed to:  Health Statistics Division,
Statistics Canada, Ottawa, Ontario, Canada, K1A 0T6 (telephone: (613) 951-1746).
For information on the wide range of data available from Statistics Canada, you can contact us by calling one of our toll-free
numbers. You can also contact us by e-mail or by visiting our Web site.
National inquiries line
1 800 263-1136
National telecommunications device for the hearing impaired
1 800 363-7629
Depository Services Program inquiries
1 800 700-1033
Fax line for Depository Services Program
1 800 889-9734
E-mail inquiries
infostats@statcan.ca
Web site
www.statcan.ca
Ordering and subscription information
This product, catalogue no. 82-003-XIE,  is publsihed quarterly in electronic format  at a price of CAN $17.00 per issue and
CDN $48.00 for a one-year subscription. To obtain single issues or to subscribe, visit our Web site at www.statcan.ca, and
select Our Products and Services.
This product is also available as a standard printed publication at a price of CAN $22.00 per issue and CDN $63.00 for a one-
year subscription. The following additional shipping charges apply for delivery outside Canada:
Single issue
Annual subscription
United States
CAN $ 6.00
CAN $24.00
Other countries
CAN $ 10.00
CAN $40.00
All prices exclude sales taxes.
The printed version of this publication can be ordered by
• Phone (Canada and United States)
1 800 267-6677
• Fax (Canada and United States)
1 877 287-4369
• E-mail
infostats@statcan.ca
• Mail
Statistics Canada
Finance Division
R.H. Coats Bldg., 6th Floor
120 Parkdale Avenue
Ottawa, ON K1A 0T6
• In person from authorized agents and bookstores.
When notifying us of a change in your address, please provide both old and new addresses.
Standards of service to the public
Statistics Canada is committed to serving its clients in a prompt, reliable and courteous manner and in the official language
of their choice. To this end, the Agency has developed standards of service that its employees observe in serving its clients.
To obtain a copy of these service standards, please contact Statistics Canada toll free at 1 800 263-1136.  The service
standards are also published on www.statcan.ca under About Statistics Canada > Providing services to Canadians.

---

Statistics Canada
Health Statistics Division
Health
Reports
t
Volume 17, Number 1
Published by authority of the Minister responsible for Statistics Canada
© Minister of Industry, 2005
All rights reserved. Use of this product is limited to the licensee and its employees. The product
cannot be reproduced and transmitted to any person or organization outside of the licensee’s
organization.
Reasonable rights of use of the content of this product are granted solely for personal, corporate or
public policy research, or educational purposes. This permission includes the use of the content in
analyses and the reporting of results and conclusions, including the citation of limited amounts of
supporting data extracted from the data product in these documents. These materials are solely for
non-commercial purposes. In such cases, the source of the data must be acknowledged as follows:
Source (or “Adapted from,” if appropriate): Statistics Canada, name of product, catalogue, volume
and issue numbers, reference period and page(s). Otherwise, users shall seek prior written
permission of Licensing Services, Marketing Division, Statistics Canada, Ottawa, Ontario, Canada,
K1A 0T6.
November 2005
Catalogue no. 82-003-XPE, Vol. 17, No. 1
ISSN 0840-6529
Catalogue no. 82-003-XIE, Vol. 17, No. 1
ISSN 1209-1367
Frequency: Quarterly
Ottawa
Note of Appreciation
Canada owes the success of its statistical system to a long-standing partnership between Statistics
Canada, the citizens of Canada, its businesses, governments and other institutions. Accurate and
timely statistical information could not be produced without their continued cooperation and
goodwill.

---

SYMBOLS
The following standard symbols are used in Statistics Canada
publications:
.
not available for any reference period
..
not available for specific reference period
…
not applicable
p
preliminary figures
r
revised figures
x
suppressed to meet the confidentiality requirements of the
Statistics Act
E
use with caution
F
too unreliable to be published
The paper used in this publication meets the minimum requirements of
American National Standard for Information Sciences – Permanence of
Paper for Printed Library Materials, ANSI Z39.48 – 1984.
∞

---

○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○
About Health Reports
Editor-in-Chief
Anik Lacroix
Health Reports is a quarterly journal produced
by the Health Statistics Division at Statistics
Senior Editor
Mary Sue Devereaux
Canada.  It is designed for a broad audience that
Editor
includes health professionals, researchers, policy-makers,
Barbara Riggs
educators, and students.  Its mission is to provide high quality,
Assistant Editor
Marc Saint-Laurent
relevant, and comprehensive information on the health status
Production Manager
of  the population and on the health care system.  The journal
Nicole Leduc
publishes articles of  wide interest that contain original and
Production and Composition
Agnes Jones
timely analyses of  health and vital statistics data.  The sources
Robert Pellarin
of  data are typically national or provincial/territorial
Administration
Céline Desfonds
administrative databases or surveys.
Data verification
Health Reports contains  Research Articles and Health
Julia Gal
Matters.  Research Articles present in-depth analysis, undergo
Associate Editors
Owen Adams
anonymous peer review, and are indexed in Index Medicus and
Arun Chockalingham
Elizabeth Lin
MEDLINE.  Health Matters are shorter, descriptive reports,
Nazeem Muhajarine
Yves Péron
largely based on survey and administrative data from Health
Georgia Roberts
Geoff Rowe
Statistics Division.
Eugene Vayda
For information on subscribing, see How to Order.  For other
information, contact the Editors, Health Reports, Health
Statistics Division, Statistics Canada, 3rd Floor, R.H. Coats
Building, Ottawa, Ontario, Canada K1A 0T6.  Telephone: (613)
951-1807; fax: (613) 951-4436.
Health Reports, Vol. 17, No. 1, November 2005
Statistics Canada, Catalogue 82-003

---

Requests to reprint
Citation recommendation
No part of  this publication may be reproduced
Health Reports has a unique Statistics Canada
without prior written permission from Statistics
catalogue number:  82-003.  The English paper
Canada.  To obtain this permission, an Application
version is 82-003-XPE; the electronic version is 82-
for Copyright Authorization must be submitted.  This
003-XIE.  The catalogue number facilitates storing
form is available from the Copyright Permission
and retrieving the journal in libraries, either on the
Officer, Marketing Division, Statistics Canada (fax:
shelf  or electronically.  Thus, we request that, when
613-951-1134).
citing a Health Reports article in other published
material, authors include our catalogue number.
Electronic version
Health Reports is also published as an electronic
Example:
product in PDF format.  Single issues may be
Parsons GF, Gentleman JF, Johnston KW. Gender differences
ordered (using Visa or MasterCard) from Statistics
in abdominal aortic aneurysm surgery. Health Reports (Statistics
Canada, Catalogue 82-003) 1997; 9(1): 9-18.
Canada’s Internet site, downloaded onto your
desktop and accessed with Adobe Acrobat Reader.
To order a recent issue of  Health Reports, visit our
site at http:// www.statcan.ca.  Select “English” from
the home page, then “Our products and services”
from the next page.  Select “Browse our Internet
publications (PDF or HTML),” “For sale,” and
“Health,” where you will find Health Reports
(Catalogue 82-003-XIE).
Health Reports, Vol. 17, No. 1, November 2005
Statistics Canada, Catalogue 82-003

---

link to page 0 link to page 11 link to page 0 ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○
In This Issue
Research Articles
Insomnia…………………………………………………………………………………………………………………………. 9
Michael Tjepkema
In 2002, more than 13% of the adult population, an
estimated 3.3 million Canadians, had insomnia.  Insomnia
tended to afflict people with painful chronic conditions or mood/
anxiety disorders.  Insomnia was also associated with life stress,
obesity, frequent use of  alcohol or cannabis, being female and
low education.
Body mass and dependency ……………………………………………………………………………………… 27
Kathryn Wilkins and Margaret de Groh
At age 45 or older, underweight people and those in obese class
III (the highest level of obesity) were almost equally likely to be
dependent.  Women were at greater risk of  dependency than
were men, and for women, all classes of obesity were related to
being dependent.  Obesity was also predictive of subsequent
dependency.
Health Matters
Life expectancy …………………………………………………………………………………………………………… 43
Julie St-Arnaud, Marie P. Beaudet and Patricia Tully
• In 2002, average life expectancy at birth was 79.7 years:
77.2 years for men and 82.1 years for women.
• Between 1977 and 2002, the gap between male and female
life expectancy narrowed from 7.3 to 4.9 years.
• Mortality rates for leading causes of death dropped for both
sexes, except for lung cancer among women, which increased
sharply.
Health Reports, Vol. 17, No. 1, November 2005
Statistics Canada, Catalogue 82-003

---

link to page 0 link to page 0 link to page 0 ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○
Diseases of the circulatory system—Hospitalization and mortality ……………………… 49
Helen Johansen, Satha Thillaiampalam, Denis Nguyen and
Christie Sambell
• In 2001/02, more than 309,000 people were hospitalized
because of  diseases of  the circulatory system.
• Between 1994/95 and 2001/02, age-standardized
hospitalization rates for these diseases fell from 1,656 to
1,339 patients per 100,000 population aged 20 or older.
• Despite a decline in the mortality rate, diseases of the
circulatory system remained the leading cause of  death,
accounting for 34% of deaths in 2002.
Edentulism and denture use ……………………………………………………………………………………… 55
Wayne J. Millar and David Locker
• In 2003, an estimated 9% of Canadians aged 15 or older
reported that they had no natural teeth, down from 16% in
1990.
• At age 55 or older, women were more likely than men to be
edentate.
• About a quarter of the population aged 15 or older wore
full or partial dentures in 2003.
………………………………………………………………………………………… 61
How to Order
Information about Health Statistics Division’s products and
services, including prices
Health Reports, Vol. 17, No. 1, November 2005
Statistics Canada, Catalogue 82-003

---

Research Articles
In-depth research and analysis

---

---

27
Body mass and
dependency
○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○
Kathryn Wilkins and Margaret de Groh
Abstract
Objectives
The relationship between body mass index (BMI)
As efforts to understand the factors that influence
the quality of  life of  middle-aged and older adults
category and dependency in men and women aged 45 or
older is examined cross-sectionally and prospectively.
 gain importance,1 the influence of  weight on
Data sources
functional status has emerged as an area of  study in the
Data are from the 2003 Canadian Community Health
Survey and the 1994/95 through 2002/03 National
“epidemiology of  disability.”  Cross-sectional and
Population Health Survey, household populations.
Analytical techniques
prospective studies have indicated that individuals at the
Cross-sectional data were used to produce weighted
frequencies, cross-tabulations and multiple logistic
extremes of  the body mass index (BMI) ranges are far more
regression models to estimate the prevalence of
dependency and its relationship to BMI category.
likely to experience physical disability than are those in the
Associations between BMI and dependency two years
later were also explored.  Models were adjusted for
“normal” BMI category.2-5  However, findings about the
potential confounders.
Main results
role of  excess weight and other risk factors on functional
The prevalence of dependency was nearly the same
among those who were underweight as among those in
limitation are inconsistent.  Only about half of longitudinal
obese class III—the highest level of obesity.  Even when
the effects of potential confounders were controlled,
studies of  chronic conditions in relation to future functional
underweight and obese people faced higher odds of co-
existing dependency, compared with those in the normal
limitation have found a significant relationship with obesity,
BMI range.  Obesity was also predictive of subsequent
dependency.
although the variety of  measures and analytical approaches
used may account for the discrepant findings.6
Keywords
activities of daily living (ADL), instrumental activities of
This article provides a detailed examination of  the
daily living (IADL), body mass index, chronic illness,
independent living, longitudinal studies
association between BMI category and dependency (see
Authors
Analytical techniques, Data sources and Limitations).  Estimates
Kathryn Wilkins (613-951-1769; Kathryn.Wilkins@statcan.ca)
is with the Health Statistics Division at Statistics Canada,
are presented for the Canadian household population aged
Ottawa, Ontario, K1A 0T6.  Margaret de Groh (613-957-
1786; Margaret_de_Groh@phac-aspc.gc.ca) is with the
45 or older.  Because other studies have found women to
Chronic Disease Prevention Division of the Centre for
Chronic Disease Prevention and Control at the Public Health
be at consistently greater risk than men of  functional decline
Agency of Canada, Ottawa, Ontario, K1A 0K9.
over time, and that obesity seems to have a greater impact
on women,5,7-10 sex-specific analyses were conducted.
Health Reports, Vol. 17, No. 1, November 2005
Statistics Canada, Catalogue 82-003

---

28
Body mass and dependency
Analytical techniques
Cross-sectional analysis:  Based on data from the 2003 Canadian
(cycle 3 to 4) and 2000/01 to 2002/03 (cycle 4 to 5).  The first cycle
Community Health Survey (CCHS) (cycle 2.1), frequencies, cross-
in each two-cycle interval served as the baseline for the study of
tabulations and multiple logistic regression models were produced
incident dependency; each eligible respondent was thus considered
using data weighted to the 2003 Canadian population.  The bootstrap
as many as four times.  Respondents who were not dependent in
technique, which accounts for the design effects of the survey, was
the first cycle of the interval were re-assessed in the next cycle.  For
used to calculate variance.11-13  Statistical significance was established
example, respondents who reported no need for assistance in cycle
as p < 0.05.
1, but who were institutionalized or reported the need for assistance
Multiple logistic regression modelling was used to examine
in cycle 2, were categorized as having become dependent.  Similarly,
associations between BMI categories and dependency, while
respondents who had been categorized as not dependent in cycle 2
controlling for confounding factors.  Models were sex-specific.  In
were included for assessment of dependency status in cycle 3.
addition to BMI, variables entered into regression models were
Respondents categorized as dependent in cycle 2, and then not
selected based on findings from the literature and their availability in
dependent in cycle 3, were assessed again at cycle 4; respondents
the survey.  To distinguish variables having an indirect effect on
who were not dependent in cycle 4 were assessed at cycle 5.  Thus
dependency from those exerting a more direct influence, multiple
an individual respondent could potentially contribute two counts of
logistic regression models were fitted hierarchically.5  Variables were
incident dependency—one in cycle 2 or 3, and one in cycle 4 or 5.
entered sequentially into four models (Appendix Tables A and B), as
Multiple logistic regression analysis was used on the pooled set of
follows:
observations to model the odds of a new (within a two-year interval)
Model
Control variables
need for assistance or institutionalization in a long-term care facility,
   1
Age (continuous), education, main source of income,
relative to BMI category at baseline, while controlling for the effects
and living arrangements.
of other influences on this relationship.  Variables to control for
   2
Variables  in  Model  1  and BMI category, smoking
potential confounding included age, main source of income, level of
status, and leisure-time physical activity level.
education, living arrangements (alone or with others), chronic disease
   3
Variables in Models 1 and  2 and respiratory disease
(cancer, respiratory disease), smoking and level of leisure-time
and cancer.
physical activity.
   4
Variables in Models 1, 2, 3 and high blood pressure,
The literature on the effects of obesity advises caution in controlling
heart disease, diabetes and arthritis.
for risk factors or health problems that arise from obesity.  Inclusion
in multivariate models of conditions that may be intermediaries in
A variable for pain, based on sub-sample data (see Data sources),
the causal pathway from obesity to ADL/IADL dependency may mask
was also included in a separate model relating dependency to BMI
the full effects of obesity.14,15  Therefore, chronic conditions that were
level (see Table 3).
strongly positively related to BMI in preliminary analysis (high blood
Longitudinal analysis:  National Population Health Survey (NPHS)
pressure, diabetes, arthritis), or otherwise known to be related to
respondents were included in the analysis if they were at least 45
obesity (heart disease), were added to the multivariate models only
years old at the time of any of the cycle 1 to 4 survey interviews;
as a final step.
provided data on their weight and height and their ability to perform
All independent variables were based on data from cycles 1 through
personal and instrumental activities of daily living (ADL/IADL) in at
4.  For the models, the value of each independent variable was that
least one survey interview, and data on ADL/IADL in the subsequent
reported in the first of two consecutive interviews, and the value of
interview; and indicated at the time of the first of these interviews
the dependent variable (incident dependency) was that reported in
that they did not need the help of another person with these activities.
the second of these interviews.
The longitudinal analysis was conducted using pooling of repeated
Weighted data were used for all analyses.  Coefficients of variation
observations, combined with logistic regression.  Data on dependency
on estimates of proportion and differences between proportions and
were considered in two-cycle intervals (roughly corresponding to
odds ratios were calculated using the bootstrap technique, which
two-year periods, based on interview dates):  1994/95 to 1996/97
accounts for survey design effects.11-13
(cycle 1 to 2); 1996/97 to 1998/99 (cycle 2 to 3), 1998/99 to 2000/01
Health Reports, Vol. 17, No. 1, November 2005
Statistics Canada, Catalogue 82-003

---

Body mass and dependency
29
Data sources
Canadian Community Health Survey (CCHS):   Cross-sectional
National Population Health Survey (NPHS):  Longitudinal analysis
analysis of data was based on cycle 2.1 of the CCHS, which was
was based on data from the first five NPHS cycles, 1994/95 through
conducted between January and December 2003.  The CCHS is a
2002/03.  Since 1994/95, the NPHS has collected information about
general health survey that collects cross-sectional information about
the health of the Canadian population every two years.  It covers
the health of Canadians every two years.  It covers the non-
household and institutional residents in all provinces and territories,
institutionalized household population aged 12 or older in all provinces
except persons on Indian reserves, on Canadian Forces bases and
and territories, except regular members of the Canadian Armed
in some remote areas.
Forces and residents of Indian reserves, Canadian Forces bases,
In 1994/95, 20,095 individuals were selected for the longitudinal
and some remote areas.  The overall response rate was 80.6%; the
panel.  Of these, 17,276 agreed to participate, for a response rate of
total sample size was 135,573, of whom 69,492 were 45 or older.
86.0%.  The response rates for subsequent cycles, based on these
The cross-sectional analyses (except those involving pain) were
17,276 respondents, were:  92.8% in cycle 2 (1996/97); 88.2% in
based on data from respondents in this sample.  Because of non-
cycle 3 (1998/99); 84.8% in cycle 4 (2000/01); and 80.6% in cycle 5
response to individual questionnaire items, the actual number of
(2002/03).
respondents used in each tabulation or model varied.  For example,
More detailed descriptions of the NPHS design, sample and
data for the following number of respondents were missing:  165 for
interview procedures can be found in published reports.17,18
dependency; 1,859 for body mass index (BMI); and 5,705 for main
The 2002/03 NPHS cycle 5 longitudinal “square” master file was
source of income.
used for this analysis.  This file contains records for all longitudinal
Questions on pain are part of the Health Utility Index (HUI).  In
respondents in the household component (n = 17,276) whether or
2003, the HUI was designated a “sub-sample” module; at the national
not they provided information for all five cycles (that is, individuals
level it was administered to a randomly selected subset of
selected for the longitudinal sample for whom information is available
respondents.  The health regions in Newfoundland and Labrador,
for cycle 1).  The longitudinal analysis in this study was based on
Prince Edward Island, Nova Scotia, New Brunswick and Québec
data for respondents meeting the following criteria: aged 45 or older
opted to have this module administered to all respondents in their
in cycle 1, 2, 3, or 4; not dependent (see Definitions) in at least one
provinces.  Of these respondents and the subset in the remaining
of these cycles and provided data on height and weight in that same
provinces and territories, 35,466 were 45 or older.  In this sub-sample,
cycle; and provided data on their dependency status in the following
data were missing on dependency (58), BMI (871) and main source
cycle.
of income (5,243).
Full descriptions of the CCHS and the National Population Health
A description of the CCHS methodology is available in a published
Survey are available on the Statistics Canada Web site @ http://
report.16
www.statcan.ca/english/sdds/0031t.htm.
The relationship between level of  BMI and co-
into account, including age, socio-economic status,
existing dependency was studied using cross-
living arrangements, health and behavioural risk
sectional data from the 2003 Canadian Community
factors, chronic diseases weakly or not related to
Health Survey (CCHS).  Then, the association
obesity, chronic conditions strongly related to
between BMI and subsequent dependency was
obesity, and chronic pain (see Definitions).
assessed with longitudinal data from the 1994/95
through 2002/03 National Population Health
Women at higher risk
Survey (NPHS).  For both analyses, several
For this analysis, CCHS respondents were
potentially confounding characteristics were taken
considered to be dependent if  they needed help with
Health Reports, Vol. 17, No. 1, November 2005
Statistics Canada, Catalogue 82-003

---

30
Body mass and dependency
Table 1
personal care activities such as washing and dressing
Percentage of people who were dependent, by task and sex,
and/or with other daily activities including
household population aged 45 or older, Canada, 2003
housework or meal preparation (see Definitions).  In
Men
Women
2003, women aged 45 or older were twice as likely
%
(16.8%) as their male counterparts (8.2%) to report
dependency (Chart 1, Table 1).  Women’s need for
Total (any task)
8.2
16.8*
help was especially pronounced for tasks involving
Everyday housework
5.9
12.3*
transportation or physical effort, notably running
Getting to appointments/
errands or getting to appointments and doing
 Running errands
5.6
12.1*
everyday housework.  Further analysis was
Preparing meals
3.6
5.2*
undertaken to investigate the extent to which these
Personal care (e.g., washing,
 dressing, eating)
2.6
3.3*
differences may have resulted from gender roles or
Moving about in house
1.8
2.3*
the older age distribution of  women.  Yet even at
Data source: 2003 Canadian Community Health Survey
younger ages, women were more likely than men to
* Significantly higher than estimate for men (p < 0.05)
report the need for help with meal preparation—
traditionally a more “female” task (data not shown).
The gap in dependency between the sexes persisted
Underweight, obesity linked to
even when controlling for the effects of  age, socio-
dependence
economic status, BMI, health and lifestyle risk
Adults at both BMI extremes—underweight and
factors, and chronic disease.  In fact, the odds of
obese class III—were significantly more likely than
dependency for women were twice the
those in the “normal” category to be dependent
corresponding odds for men (data not shown).
(Chart 2).  This “J-” or “U-shaped” relationship has
been noted in other reports.19-22  One-quarter of
Chart 1
Chart 2
Percentage of people who were dependent, by age group and
Percentage of people who were dependent, by body mass
sex, household population aged 45 or older, Canada, 2003
index and sex, household population aged 45 or older,
Canada, 2003
*
*
34
64
*
E
31
*
 Men
 Men
28
*
 Women
25
 Women
*
47
23
*20
*35
16
14
*
22
*
8
9
9
*
17
18
*
7
*
12
8
10
7
9
4
Underweight
Normal
Overweight
Obese I
Obese II
Obese III
Total, 45+
45-54
55-64
65-74
75-84
85+
(< 18.5)
(18.5-24.9)
(25.0-29.9)
(30.0-34.9)
(35.0-39.9)
(    40)
≥
Age group
Body mass index (BMI)
Data source: 2003 Canadian Community Health Survey
Data source: 2003 Canadian Community Health Survey
Note: Compared with estimates for 45-to-54 age group, all other age group
*Significantly different from estimate for same sex in “normal” BMI
estimates are statistically different within each sex (p < 0.05).
category (p < 0.05)
*Significantly higher than estimate for men (p < 0.05)
E Coefficient of variation 16.6% to 33.3% (interpret with caution)
Health Reports, Vol. 17, No. 1, November 2005
Statistics Canada, Catalogue 82-003

---

Body mass and dependency
31
underweight men (24.8%) and one-third of
BMI and dependency were somewhat weakened
underweight women (33.8%) aged 45 or older were
(Appendix Tables A and B, Model 4; Table 2).  In
dependent.  People with a BMI of  40 or more were
fact, only at the extremes of  the BMI categories,
categorized as obese class III; similar proportions
underweight and obese class III, did the odds of
of  men (28.2%) and women (30.7%) in this group
dependency among men remain significantly
were dependent.  For women, the relationship
elevated.
between obesity and dependency increased steadily
For women, the relationships between BMI and
with the level of  obesity.  The obesity–dependency
dependency were similar to those for men, but with
relationship was weaker for men; only those in the
class III category were more likely to be dependent
Table 2
than were men in the normal BMI range.
Adjusted odds ratios relating BMI category and other selected
characteristics to dependency, by sex, household population
Relationships persist
aged 45 or older, Canada, 2003
The relationships between BMI level and
Men
Women
dependency persisted when the potential influences
95%
95%
of  age, socio-economic status and health-related
Adjusted
confi- Adjusted
confi-
odds
dence
odds
dence
behaviours were taken into account simultaneously
ratio interval
ratio interval
(Appendix Tables A and B, Models 1 and 2).  At the
Age (continuous)
1.0* 1.0, 1.0
1.0* 1.0, 1.1
low end of  the BMI range, the odds of  dependency
Socio-economic factors
for underweight men and women were about twice
Less than secondary graduation 1.2 1.0, 1.3
1.0 0.9, 1.1
Secondary graduation or more†
1.0
…    
1.0
…    
those for each sex in the normal BMI category.
Main source of income is
 social assistance‡
1.9* 1.6, 2.3
1.3* 1.1, 1.4
Other research has attributed such a relationship to
Lives alone‡
1.5* 1.3, 1.7
0.9 0.9, 1.0
the likelihood of  underlying illness in underweight
Body mass index (BMI)
individuals.1  Although this study of  CCHS data took
Underweight (< 18.5)
2.0* 1.3, 3.2
1.9* 1.4, 2.6
Normal (18.5-24.9)‡
1.0
…    
1.0
…    
numerous chronic conditions into account, the
Overweight (25.0-29.9)
0.9 0.7, 1.0
1.0 0.9, 1.1
possibility of  underweight indicating frailty or
Obese class I (30.0-34.9)
1.1 0.9, 1.4
1.2* 1.0, 1.4
Obese class II (35.9-39.9)
1.0 0.7, 1.4
1.6* 1.2, 2.0
compromised health remains.
Obese class III (≥ 40)
3.6* 2.0, 6.6
2.3* 1.6, 3.2
As expected, controlling for the effects of cancer
Smoking status
Smoker
1.4* 1.1, 1.6
1.3* 1.2, 1.6
and respiratory disease (conditions not necessarily
Non-smoker†
1.0
…    
1.0
…    
related to BMI) had little effect on the strength of
Leisure-time physical
the association between BMI and dependency
 activity
Inactive
2.6* 2.1, 3.3
2.4* 2.1, 2.9
(Appendix Tables A and B, Model 3).
Moderate
1.2 0.9, 1.6
1.1 0.9, 1.3
The literature on the effects of  obesity advises
Active†
1.0
…    
1.0
…    
caution in controlling for risk factors or health
Chronic conditions
Respiratory disease‡
2.2* 1.8, 2.6
2.1* 1.8, 2.5
problems that arise from obesity.  Including
Cancer‡
2.1* 1.7, 2.7
2.2* 1.7, 2.9
conditions that may be intermediaries in the causal
Obesity-related
 chronic conditions
pathway from obesity to ADL/IADL dependency
High blood pressure‡
1.2* 1.0, 1.4
1.1 1.0, 1.2
in multivariate models may mask the full effects of
Heart disease‡
1.9* 1.6, 2.2
2.1* 1.8, 2.4
Diabetes‡
1.6* 1.3, 2.0
1.7* 1.4, 2.0
obesity.14,15  Therefore, chronic conditions that were
Arthritis‡
2.3* 2.0, 2.7
2.4* 2.1, 2.6
strongly positively related to BMI in preliminary
Data source:  2003 Canadian Community Health Survey
analysis, or that are otherwise known to be related
Notes:  Models are based on weighted data from records for 28,880 men and
37,783 women who provided information on mobility function.   Variables for
to obesity, were added to the multivariate models
"missing" BMI and source of income were included in the models to maximize
only as a final step.  When the potential effects of
sample size, but the odds ratios are not shown.  Because of rounding, some
odds ratios with 1.0 as the lower confidence interval are statistically significant.
several obesity-related conditions—high blood
† Reference category
pressure, heart disease, diabetes and arthritis—were
‡ Reference category is absence of condition; for example, reference category
for cancer is no reported diagnosis of cancer.
also taken into account, the associations between
* Significantly different from estimate for reference category (p < 0.05)
… Not applicable
Health Reports, Vol. 17, No. 1, November 2005
Statistics Canada, Catalogue 82-003

---

32
Body mass and dependency
Definitions
Data from the 2003 Canadian Community Health Survey (CCHS)
that had lasted, or was expected to last, at least six months.
were used to calculate prevalence estimates of dependency.
Respondents were read a list of conditions; cancer, respiratory
Respondents were categorized as dependent if they answered “yes”
disease, high blood pressure, heart disease, arthritis and diabetes
to at least one of the following questions:   “Because of any physical
were selected for this analysis.
condition or mental condition or health problem, do you need the
In middle-aged and older people, many of whom are retired and
help of another person with:
own their homes, level of income may not be a reliable indicator of
• preparing meals?”
socio-economic status.  In an effort to identify people of limited means,
• getting to appointments and running errands such as shopping
respondents were asked about their main source of income.
for groceries?”
Response categories were:  wages and salaries; income from self-
• doing everyday housework?”
employment; dividends and interest (for example, on bonds or
• personal care such as washing, dressing, eating or taking
savings); Employment Insurance; Workers’ Compensation; Canada
medication?”
or Québec Pension Plan benefits; retirement pension,
• moving about inside the house?”
superannuation and annuities; Old Age Security and Guaranteed
Respondents who answered “no” to all these questions were not
Income Supplement; Child Tax Benefit; provincial or municipal social
considered to be dependent.  Those with records that showed data
assistance or welfare; child support; alimony; and other (rental income
missing for all questions, or with “no” responses to some questions
or scholarships, for example).  Respondents who identified Canada
and missing responses to any other(s), were excluded from this
or Québec Pension Plan benefits, Old Age Security and Guaranteed
analysis.
Income Supplement, or provincial/municipal social assistance or
Data from the 1994/95 through 2002/03 National Population Health
welfare were categorized as receiving social assistance as their main
Survey (cycles 1 through 5, NPHS) were used to examine incident
income source.
dependency in relation to BMI category.  For the longitudinal analysis,
Two smoking status categories were defined:  smokers and non-
dependency was defined as a “yes” response to any of the preceding
smokers.  Smokers are those smoking either daily or occasionally
questions (with minor differences in wording, the same questions
and, in this case, “occasionally” includes only current occasional
were used by both the CCHS and the NPHS), or a respondent’s
smokers who used to smoke every day.  Non-smokers comprises
move from the household population into a long-term care facility.
people who had never smoked, plus occasional smokers who had
Body mass index (BMI) is a measure of weight adjusted for height,
never smoked every day, as well as  former smokers (daily or
calculated by dividing weight in kilograms by height in metres
occasional smokers who had quit smoking altogether).
squared.  BMI categories were defined using the standards adopted
Level of leisure-time physical activity was based on calculations
by Health Canada:
that took into account the reported frequency and duration of a
• underweight:  < 18.5
respondent’s leisure-time physical activities in the three months
• normal:  18.5 to 24.9
before the survey, and the estimated metabolic energy demand of
• overweight:  25.0 to 29.9
each activity.24,25  Leisure was classified as active (3.0 or more
• obese class I:  30.0 to 34.9
kilocalories per kilogram per day), moderately active (1.5 to 2.9 kcal/
• obese class II:  35.0 to 39.9
kg/day), or inactive (below 1.5 kcal/kg/day).
• obese class III:  ≥ 40.0.23
Living arrangements were defined dichotomously as living alone
Height and weight were self-reported by CCHS and NPHS
or with others.
respondents.
The presence of pain was established based on a “no” response
The presence of a chronic condition was established by asking
to the question, “Are you usually free of pain or discomfort?”
respondents if a doctor had told them that they had a chronic disease
an important exception.  Although including
obesity (Appendix Table B, Model 4; Table 2).  Thus
obesity-related chronic diseases weakened the odds
any degree of  obesity appears to make its own
ratios slightly, BMI remained significantly related to
contribution to dependency for women, aside from
dependency for women in all three categories of
the influences of  obesity-related conditions.
Health Reports, Vol. 17, No. 1, November 2005
Statistics Canada, Catalogue 82-003

---

Body mass and dependency
33
Effects of pain compounded by obesity
obese class III would be dependent were over 6
Previous research has indicated that the likelihood
times higher than the odds for men in the normal
of pain increases with BMI,26 and that pain is
BMI range when pain was taken into account.
strongly related to physical function.5,27  For men,
The odds ratio for pain was also high, and strongly
when pain was accounted for (see Methods) together
significant.  Notably, with pain in the model, the
with age, source of  income and educational
relationship between underweight and dependency
attainment, living arrangements, health-related
was no longer statistically significant.  When an
behaviours and chronic conditions unrelated to
interaction term for obesity (classes I to III
obesity, the odds that those in obese classes II and
combined) and pain was included in the model, its
III would be dependent were still significantly
odds ratio was also significantly elevated (2.3; 95%
elevated (Table 3).  In fact, the odds that men in
confidence interval = 1.1 – 4.7), indicating that for
men with chronic pain, obesity compounds the
Table 3
probability of  dependency (data not shown).
Adjusted odds ratios relating BMI category and selected
The results for women were similar (Table 3).  For,
characteristics to dependency, controlling for pain, by sex,
women, however, obesity at all three levels remained
household population aged 45 or older, Canada, 2003
significantly related to dependency when pain was
Men
Women
taken into account.  The interaction term for obesity
95%
95%
and pain was not significant for women (data not
Adjusted
confi- Adjusted
confi-
odds
dence
odds
dence
shown).
ratio interval
ratio interval
Age (continuous)
1.1* 1.1, 1.1
1.1* 1.1, 1.1
Obesity predictive of dependency
Socio-economic factors
Findings from respondents followed over time
Less than secondary graduation 0.7* 0.5, 0.9
0.6* 0.5, 0.7
Secondary graduation or more†
1.0
…    
1.0
…    
differed somewhat from those based on the 2003
Main source of income is
data (partly because of  a smaller sample size
 social assistance‡
1.4* 1.0, 1.8
1.2 1.0, 1.4
Lives alone‡
1.2 0.9, 1.6
0.7* 0.6, 0.8
resulting in less statistical power).  According to
Body mass index (BMI)
longitudinal data from the NPHS, before controlling
Underweight (< 18.5)
1.3 0.6, 3.2
1.5 1.0, 2.4
for obesity-related chronic conditions, only men in
Normal (18.5-24.9)‡
1.0
…    
1.0
…    
Overweight (25.0-29.9)
1.3 1.0, 1.7
1.0 0.8, 1.3
obese class I at the outset of  a two-year period had
Obese class I (30.0-34.9)
1.2 0.8, 1.7
1.4* 1.1, 1.7
Obese class II (35.9-39.9)
2.2* 1.1, 4.6
2.2* 1.5, 3.1
significantly elevated odds of  dependency by the
Obese class III (≥ 40)
6.2* 2.8,13.5
4.2* 2.5, 7.2
end (Table 4).  Because of  the small sample size in
Smoking status
the category, the odds ratio for men in obese class
Smoker
1.3 1.0, 1.7
1.2 1.0, 1.5
Non-smoker†
1.0
…    
1.0
…    
III fell just short of  significance (p = 0.051).  When
Leisure-time physical
controlling for well-known obesity-related
 activity
conditions (heart disease, high blood pressure,
Inactive
3.9* 2.6, 5.9
3.5* 2.6, 4.7
Moderate
1.9* 1.2, 3.0
1.5* 1.0, 2.1
diabetes, arthritis) and pain, underweight was the
Active†
1.0
…    
1.0
…    
only BMI category to remain predictive of
Chronic conditions
Respiratory disease‡
3.1* 2.2, 4.4
2.0* 1.5, 2.6
dependency among men.
Cancer‡
2.4* 1.4, 4.1
2.0* 1.3, 3.0
The relationship between obesity and subsequent
Pain‡
4.1* 3.1, 5.4
3.4* 2.9, 4.1
dependency was more pronounced for women
Data source:  2003 Canadian Community Health Survey
Notes: Based on records for 14,882 men and 19,502 women.   Variables for
(Table 5).  Before the obesity-related conditions were
"missing" BMI, leisure-time activity level and main source of income were
taken into account, underweight women, as well as
included in the models to maximize sample size, but the odds ratios are not
shown.  Because of rounding, some odds ratios with 1.0 as the lower confidence
those in obese classes II and III, had significantly
interval are statistically significant.
higher odds of  becoming dependent over the next
† Reference category
† Reference category
two years, compared with women in the normal BMI
‡ Reference category is absence of condition; for example, reference category
range.  But when the obesity-related chronic
for pain is no reported chronic pain.
* Significantly different from estimate for reference category (p < 0.05)
… Not applicable
Health Reports, Vol. 17, No. 1, November 2005
Statistics Canada, Catalogue 82-003

---

34
Body mass and dependency
Limitations
The analyses are based on self-reported data.  Because overweight
successive survey cycles were included in the analysis, even though
and obese people tend to underreport their weight,28 the results may
they may have been subsequently lost to follow-up.  This approach
have been somewhat distorted by non-random misclassification.  Such
also allowed those who became eligible for analysis (for example,
bias would weaken the observed strength of the association between
reached age 45) sometime after the first interview cycle to be included.
body mass index (BMI) category and dependency.
To assess the effects of non-response on the results, the weighted
BMI is based on height and weight, and does not take factors such as
proportions of non-respondents were compared for a few selected
percentage of body fat or waist circumference into account.  As well,
variables (age group, sex, BMI).  No significant differences in the
the same BMI cutpoints are applied to both sexes, even though body
proportions of non-respondents emerged by sex or the six BMI
fat mass is higher in women than men for the same BMI category.29
categories.  By age group, the proportion of non-respondents was slightly
Previous studies in which BMI categories were defined according to the
higher among those aged 45 to 64 (7.8%) than among seniors (6.3%).
distribution in a specific population cannot be generalized to represent
Non-respondents (weighted data), by selected variables, household
other populations.2,3,6  This point is especially relevant in view of the
component, National Population Health Survey, 1994/95 to 2002/03
rapid increase of obesity among Canadian adults.30
No information is available on weight loss due to illness.  Also, although
it is reasonable to assume that underweight is an indicator of frailty and
Non-respondents
compromised health, this analysis may not have adequately controlled
%
for these factors.
Age group
The potential for selection bias due to respondent attrition is
45 to 64
7.8*
problematic in longitudinal research.  Selective loss to follow-up, or failure
65 or older
6.3
to collect information from respondents who were in poorer health or
Sex
whose health deteriorated rapidly between survey cycles, may have
Men
7.4
weakened the observed relationship between obesity and the onset of
Women
7.2
physical dysfunction.  The analysis was based on respondents aged 45
Body mass index (BMI)
or older for whom requisite data were available over the first five NPHS
Underweight (< 18.5)
7.2
cycles.  From one survey cycle to the next, respondents were lost from
Normal (18.5 to 24.9)
7.7
the analysis for reasons such as refusal to participate, death, item non-
Overweight (25.0 to 29.9)
6.9
response, or relocation out of the country.  From the pooled total of
Obese class I (30.0 to 34.9)
6.3
21,390 respondents assessed in the “baseline” cycles, 1,417 (6.6%)
Obese class II (35.0 to 39.9)
5.6
did not respond in the follow-up cycle.
Obese class III (≥ 40)
5.8
Respondents and non-respondents (unweighted sample), household
* Significantly higher than proportion of non-respondents for 65-or-older
population aged 45 or older, by two-cycle interval, National Population
age group (p < 0.05)
Health Survey, 1994/95 to 2002/03
The survey weights were those applied to the cycle 1 (1994/95) data
Number
according to the response status at that time; the weights were not
(percentage) of
inflated to account for subsequent non-response.  This could have biased
respondents
at baseline
the estimates if continuers in the longitudinal panel differed from non-
Number of
Number of
who became
respondents according to characteristics considered in the analysis.
respondents respondents non-respondents
No inference of causality or temporal ordering is possible from analyses
at baseline at follow-up
next cycle
based on the CCHS, because the data are cross-sectional.  Although
the NPHS longitudinal data were used to establish the chronological
1994/95 to 1996/97
sequence between independent and dependent variables, causality (of
(Cycle 1 to 2)
5,547
5,247
300 (5.4%)
dependency by obesity) cannot be inferred.  The associations observed
1996/97 to 1998/99
may result from factors not considered in this analysis.
(Cycle 2 to 3)
5,388
5,097
291 (5.4%)
The dependent variable, that is, the need for help from another person
1998/99 to 2000/01
with selected instrumental and personal activities of daily living, was
(Cycle 3 to 4)
5,241
4,875
366 (7.0%)
based on self-report and was not validated against objective criteria or
2000/01 to 2002/03
by direct observation.  Variation in unmeasured subjective factors, such
(Cycle 4 to 5)
5,214
4,754
460 (8.8%)
as readiness to admit a need for assistance, likely explains some of the
Total
21,390
19,973
1,417 (6.6%)
observed differences in responses.
Loss to follow-up from the longitudinal panel was minimized in two
Assessment of chronic diseases was made by asking respondents
ways.  Instead of being excluded from the analysis, people who entered
about conditions that had been diagnosed by a health practitioner and
long-term care facilities were categorized as having become ADL-/IADL-
that had lasted, or were expected to last, six months or more.  No clinical
dependent.  Also, data from respondents were considered in two-year
validation of these self-reported conditions was carried out.
intervals.  Therefore, those who were interviewed in at least two
Health Reports, Vol. 17, No. 1, November 2005
Statistics Canada, Catalogue 82-003

---

Body mass and dependency
35
Table 4
Table 5
Adjusted odds ratios relating BMI category and selected
Adjusted odds ratios relating BMI category and selected
characteristics to subsequent dependency, male household
characteristics to subsequent dependency, female household
population aged 45 or older, Canada, 2003
population aged 45 or older, Canada, 2003
Model 1
Model 2
Model 1
Model 2
95%
95%
95%
95%
Adjusted
confi- Adjusted
confi-
Adjusted
confi- Adjusted
confi-
odds
dence
odds
dence
odds
dence
odds
dence
ratio interval
ratio interval
ratio interval
ratio interval
Age (continuous)
1.1* 1.1, 1.1
1.1* 1.1, 1.1
Age (continuous)
1.1* 1.1, 1.1
1.1* 1.1, 1.1
Socio-economic factors
Socio-economic factors
Less than secondary graduation 1.1 0.9, 1.4
1.1 0.9, 1.5
Less than secondary graduation 1.1 0.9, 1.4
1.1 0.9, 1.3
Secondary graduation or more†
1.0
…    
1.0
…    
Secondary graduation or more†
1.0
…    
1.0
…    
Main source of income is
Main source of income is
 social assistance‡
1.5* 1.2, 1.9
1.4* 1.1, 1.7
 social assistance‡
1.2 1.0, 1.5
1.0 0.8, 1.3
Lives alone‡
1.1 0.8, 1.4
1.1 0.9, 1.5
Lives alone‡
0.9 0.8, 1.1
0.9 0.8, 1.1
Body mass index (BMI)
Body mass index (BMI)
Underweight (< 18.5)
2.5 1.0, 6.5
2.5* 1.0, 6.1
Underweight (< 18.5)
1.9* 1.3, 2.9
2.2* 1.4, 3.3
Normal (18.5-24.9)‡
1.0
…    
1.0
…    
Normal (18.5-24.9)‡
1.0
…    
1.0
…    
Overweight (25.0-29.9)
1.0 0.7, 1.3
0.9 0.7, 1.2
Overweight (25.0-29.9)
1.2 1.0, 1.5
1.0 0.8, 1.3
Obese class I (30.0-34.9)
1.5* 1.0, 2.1
1.2 0.8, 1.7
Obese class I (30.0-34.9)
1.2 0.9, 1.6
1.0 0.7, 1.3
Obese class II (35.9-39.9)
0.8 0.3, 2.3
0.6 0.2, 1.7
Obese class II (35.9-39.9)
2.0* 1.3, 3.0
1.5 1.0, 2.2
Obese class III (≥ 40)
2.9 1.0, 8.2
1.9 0.6, 5.8
Obese class III (≥ 40)
3.0* 1.4, 6.5
2.6* 1.2, 5.5
Smoking status
Smoking status
Smoker
1.4* 1.0, 2.0
1.4* 1.0, 2.0
Smoker
1.6* 1.2, 2.0
1.6* 1.3, 2.0
Non-smoker†
1.0
…    
1.0
…    
Non-smoker†
1.0
…    
1.0
…    
Leisure-time physical
Leisure-time physical
 activity level
 activity level
Inactive
1.3 1.0, 1.6
1.3* 1.0, 1.6
Inactive
1.6* 1.3, 1.9
1.4* 1.2, 1.7
Moderate/Active†
1.0
…    
1.0
…    
Moderate/Active†
1.0
…    
1.0
…    
Chronic conditions
Chronic conditions
Respiratory disease‡
3.2* 2.1, 4.9
2.6* 1.7, 4.1
Respiratory disease‡
1.7* 1.1, 2.5
1.3 0.9, 1.9
Cancer‡
0.9 0.5, 1.6
0.7 0.3, 1.4
Cancer‡
1.4 0.8, 2.2
1.3 0.8, 2.1
Obesity-related chronic
Obesity-related chronic
 conditions
 conditions
High blood pressure‡
1.4* 1.0, 1.8
High blood pressure‡
1.0 0.8, 1.2
Heart disease‡
1.7* 1.2, 2.4
Heart disease‡
1.7* 1.3, 2.3
Diabetes‡
1.9* 1.4, 2.7
Diabetes‡
2.3* 1.6, 3.2
Arthritis‡
1.1 0.8, 1.4
Arthritis‡
1.4* 1.2, 1.7
Pain‡
2.5* 1.8, 3.3
Pain‡
2.5* 2.0, 3.1
Data source:  1994/95 to 2002/03 National Population Health Survey
Data source:  1994/95 to 2002/03 National Population Health Survey
Notes:  Models 1 and 2 are based on 8,993 and 8,966 records, respectively.
Notes:  Models 1 and 2 are based on 10,882 and 10,857 records, respectively.
Because of rounding, some odds ratios with 1.0 as the lower confidence interval
† Reference category
are statistically significant.
‡ Reference category is absence of condition; for example, reference category
† Reference category
for cancer is no reported diagnosis of cancer.
‡ Reference category is absence of condition; for example, reference category
* Significantly different from estimate for reference category (p < 0.05)
for cancer is no reported diagnosis of cancer.
… Not applicable
* Significantly different from estimate for reference category (p < 0.05)
… Not applicable
conditions and pain were introduced in the model,
obesity-related diseases appear to be associated more
the odds ratios remained significantly elevated only
directly with dependency than is obesity.  By
for underweight and obese class III women.
contrast, the relationship between obesity and
The longitudinal relationship between BMI and
dependency persisted for women even when the
dependency and a stronger effect of  BMI in
effects of  obesity-related diseases and pain were
predicting disability in women than men are
taken into account.
consistent with other research.3,5, 7-10,31  Among men,
Health Reports, Vol. 17, No. 1, November 2005
Statistics Canada, Catalogue 82-003

---

36
Body mass and dependency
Concluding remarks
In Canada today, one of  the most troubling public
The findings of  this study indicate that obesity is
health dilemmas is the rising prevalence of
associated with co-existing dependency in middle-
overweight and obesity, now affecting the majority
aged and older Canadians.  This relationship
of  middle-aged and older adults.30  Despite cultural
persisted even when controlling for potentially
norms that stigmatize excess weight, along with
confounding factors such as socio-economic status,
ample evidence of  its adverse health effects, the
living arrangements and level of  physical activity,
proportion of  Canadian adults who are obese has
as well as chronic disease and pain.  The results
risen considerably over the past few decades.30
suggest that, in addition to its associations with pain
Loss of  independence is a dire consequence of
and disease, obesity independently contributes to
obesity.  Caring for people who need assistance with
dependency.
basic activities of  daily living usually falls first to
Also important is the association between
family members or friends.  When these sources of
underweight and dependency.  Both men and
help are unavailable, formal home care services may
women who were categorized as underweight had
be sought.  Dependency is also strongly predictive
strikingly higher odds of  dependency when
of  eventual institutionalization.7  In view of  recent
compared with their counterparts in the normal BMI
rapid increases in the proportion of  people who
range.
are obese, coupled with the aging of  the population,
Longitudinal data from the National Population
the burden on informal caretakers and the health
Health Survey were used to establish the order of
care system  can be expected to increase in the near
events between obesity and dependency.  Obesity
future. 
was found to be predictive of  future dependency in
men and women aged 45 or older.
 8 Katz DA, McHorney CA, Atkinson RL. Impact of  obesity
○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○
References
on health-related quality of  life in patients with chronic
illness. Journal of  General Internal Medicine 2000; 15: 789-96.
 1 Verbrugge LM, Gates DM, Ike RW. Risk factors for disability
 9 Yan LL, Daviglus ML, Liu K, et al. BMI and health-related
among US adults with arthritis. Journal of  Clinical Epidemiology
quality of  life in adults 65 years and older. Obesity Research
1991; 44(2): 167-82.
2004; 12(1): 69-76.
 2 Han TS, Tijhuis MAR, Lean MEJ, et al. Quality of  life in
10 Diehr P, Newman AB, Jackson SA, et al. Weight-modification
relation to overweight and body fat distribution.  The American
trials in older adults: what should the outcome measure be?
Journal of  Public Health 1998; 88(12): 1814-20.
Current Controlled Trials in Cardiovascular Medicine 2002; 3(1):1.
Published online January 7, 2002. http://
 3 Launer LJ, Harris T, Rumple C, et al. Body mass index, weight
www.pubmedcentral.gov/articlerender.fcgi?tool=docline
change, and risk of  mobility disability in middle-aged and
&pubmedid=11985775
older women. The epidemiologic follow-up study of
NHANES I. Journal of  the American Medical Association 1994;
11 Rao JNK, Wu CFJ, Yue K. Some recent work on resampling
271(14): 1093-8.
methods for complex surveys. Survey Methodology (Statistics
Canada, Catalogue 12-001) 1992; 18(2): 209-17.
 4 LaCroix AZ, Guralnik JM, Berkman LF, et al. Maintaining
mobility in late life: II. Smoking, alcohol consumption,
12 Rust KF, Rao JNK. Variance estimation for complex surveys
physical activity, and body mass index. American Journal of
using replication techniques. Statistical Methods in Medical
Epidemiology 1993; 137: 858-69.
Research 1996; 5: 281-310.
 5 Clark DO, Stump TE, Wolinsky FD. Predictors of  onset of
13 Yeo D, Mantel H, Liu TP. Bootstrap variance estimation for
and recovery from mobility difficulty among adults aged 51-
the National Population Health Survey. American Statistical
61 years. American Journal of  Epidemiology 1998; 148(1): 63-71.
Association: Proceedings of  the Annual Meeting of  the American
Statistical Association, Survey Research Methods Section. Baltimore:
 6 Boult C, Kane RL, Louis TA, et al. Chronic conditions that
August 1999.
lead to functional limitation in the elderly.  Journal of
Gerontology: Medical Sciences 1994, 49, M28-M36.
14 Manson JE, Stampfer MJ, Hennekens CH, et al. Body weight
and longevity: a reassessment. Journal of  the American Medical
 7 Beckett LA, Brock DB, Lemke JH, et al. Analysis of  change
Association (JAMA) 1987; 257: 353-8.
in self-reported physical function among older persons in
four population studies. American Journal of  Epidemiology 1996;
15 Allison DB, Fontaine KR, Manson JE, et al. Annual deaths
143(8): 766-78.
attributable to obesity in the United States. Journal of  the
American Medical Association 1999; 282: 1530-8.
Health Reports, Vol. 17, No. 1, November 2005
Statistics Canada, Catalogue 82-003

---

Body mass and dependency
37
16 Béland Y. Canadian Community Health Sur vey—
24 Statistics Canada. Appendix F. Derived variables. National
methodological overview. Health Reports (Statistics Canada,
Population Health Survey: 1994-95 Public Use Microdata File
Catalogue 82-003) 2002; 13(3): 9-14.
(Catalogue 82F0001XCB) Ottawa: Minister of  Industry,
1995.
17 Swain L, Catlin G, Beaudet MP. The National Population
Health Survey—its longitudinal nature. Health Reports
25 Stephens T, Craig CI, Ferris BF. Adult physical activity in
(Statistics Canada, Catalogue 82-003) 1999; 10(4): 69-82.
Canada: findings from the Canada Fitness Survey. Canadian
Journal of  Public Health 1986; 77(4): 285-90.
18 Tambay J-L, Catlin G. Sample design of  the National
Population Health Survey. Health Reports (Statistics Canada,
26 Andersen RE, Crespo CJ, Bartlett SJ, et al. Relationship
Catalogue 82-003) 1995; 7(1): 29-38.
between body weight gain and significant knee, hip, and back
pain in older Americans. Obesity Research 2003; 11(10):
19 Heo M, Allison DB, Faith MS, et al. Obesity and quality of
1159-62.
life: mediating effects of  pain and comorbidities. Obesity
Research 2003; 11(2): 209-16.
27 Trakas K, Oh PI, Singh S, et al. The health status of  obese
individuals in Canada. International Journal of  Obesity 2001;
20 Månsson N-O, Eriksson K-F, Israelsson B, et al. Body mass
25: 662-8.
index and disability pension in middle-aged men—non-linear
relations. International Journal of  Epidemiology 1996; 25(1): 80-
28 Rowland MI. Self-reported weight and height. American
5.
Journal of  Clinical Nutrition 1990; 52: 1125-33.
21 Okoro CA, Hootman JM, Strine TW, et al. Disability, arthritis,
29 Ross R, Shaw KD, Rissanen J, et al. Sex differences in lean
and body weight among adults 45 years and older. Obesity
and adipose tissue distribution by magnetic resonance
Research 2004; 12(5): 854-61.
imaging: anthropometric relationships. American Journal of
Clinical Nutrition 1994; 59: 1277-85.
22 Lee I-M, Manson JE, Hennekens CH, et al. Body weight
and mortality: A 27-year follow-up of  middle-aged men.
30 Tjepkema M. Adult obesity in Canada: Measured height and
Journal of  the American Medical Association 1993; 270: 2823-8.
weight. Statistics Canada; 2005. http://www.statcan.ca/
english/research/82-620-MIE/2005001/articles/adults/
23 Health Canada. Canadian Guidelines for Body Weight Classification
aobesity.htm
in Adults (Catalogue H49-179) Ottawa: Health Canada, 2003.
(Canadian guidelines are available online at: http://www.hc-
31 Rissanen A, Heliövaara M, Knekt P, et al. Risk of  disability
sc.gc.ca/hpfb-dgpsa/onpp-bppn/cg_bwc_introduction
and mortality due to overweight in a Finnish population.
_e.html.)
British Medical Journal 1990; 301: 835-7.
Health Reports, Vol. 17, No. 1, November 2005
Statistics Canada, Catalogue 82-003

---

38
Body mass and dependency
Appendix
Table A
Adjusted odds ratios relating BMI level and selected characteristics to dependency, male household population aged 45 or older,
Canada, 2003
Model 1
Model 2
Model 3
Model 4
Adjusted
95%
Adjusted
95%
Adjusted
95%
Adjusted
95%
odds confidence
odds confidence
odds confidence
odds confidence
ratio
interval
ratio
interval
ratio
interval
ratio
interval
Age (continuous)
1.1*
1.0, 1.1
1.1*
1.0, 1.1
1.0*
1.0, 1.1
1.0*
1.0, 1.0
Socio-economic factors
Less than secondary graduation
1.5*
1.3, 1.7
1.2*
1.0, 1.4
1.2*
1.0,1 .4
1.2
1.0, 1.3
Secondary graduation or more†
1.0
…
1.0
…
1.0
…
1.0
…
Main source of income is social assistance‡
2.1*
1.8, 2.4
2.2*
1.9, 2.6
2.1*
1.8, 2.5
1.9*
1.6, 2.3
Lives alone‡
1.2*
1.1, 1.4
1.4*
1.2, 1.6
1.4*
1.3, 1.7
1.5*
1.3, 1.7
Body mass index (BMI)
Underweight (< 18.5)
2.2*
1.4, 3.4
2.0*
1.2, 3.1
2.0*
1.3, 3.2
Normal (18.5-24.9)†
1.0
…
1.0
…
1.0
…
Overweight (25.0-29.9)
1.0
0.8, 1.2
1.0
0.9, 1.2
0.9
0.7, 1.0
Obese class I (30.0-34.9)
1.4*
1.1, 1.8
1.4*
1.1, 1.8
1.1
0.9, 1.4
Obese class II (35.9-39.9)
1.6*
1.1, 2.2
1.5*
1.1, 2.1
1.0
0.7, 1.4
Obese class III (≥ 40)
6.3*
3.0,13.0
6.3*
3.0,13.1
3.6*
2.0, 6.6
Smoking status
Smoker
1.4*
1.2, 1.7
1.4*
1.1, 1.6
1.4*
1.1, 1.6
Non-smoker†
1.0
…
1.0
…
1.0
…
Leisure-time physical activity
Inactive
2.8*
2.2, 3.5
2.7*
2.2, 3.4
2.6*
2.1, 3.3
Moderate
1.2
0.9, 1.6
1.2
0.9, 1.6
1.2
0.9, 1.6
Active†
1.0
…
1.0
…
1.0
…
Chronic conditions
Respiratory disease‡
2.5*
2.1, 3.0
2.2*
1.8, 2.6
Cancer‡
2.4*
1.9, 3.0
2.1*
1.7, 2.7
Obesity-related chronic conditions
High blood pressure‡
1.2*
1.0, 1.4
Heart disease‡
1.9*
1.6, 2.2
Diabetes‡
1.6*
1.3, 2.0
Arthritis‡
2.3*
2.0, 2.7
Data source:  2003 Canadian Community Health Survey
Notes: Based on weighted data from records of 29,313 (Model 1), 29,112 (Model 2), 29,059 (Model 3), and 28,880 (Model 4) male respondents who provided
information on mobility function.   Variables for "missing" BMI and main source of income were included in the models to maximize sample size, but the odds ratios are
not shown.  Because of rounding, some odds ratios with 1.0 as the lower confidence interval are statistically significant.
† Reference category
† Reference category
‡ Reference category is absence of condition; for example, reference category for cancer is no reported diagnosis of cancer.
* Significantly different from estimate for reference category (p < 0.05)
… Not applicable
Health Reports, Vol. 17, No. 1, November 2005
Statistics Canada, Catalogue 82-003

---

Body mass and dependency
39
Table B
Adjusted odds ratios relating BMI level and selected characteristics to dependency, female household population aged 45 or older,
Canada, 2003
Model 1
Model 2
Model 3
Model 4
Adjusted
95%
Adjusted
95%
Adjusted
95%
Adjusted
95%
odds confidence
odds confidence
odds confidence
odds confidence
ratio
interval
ratio
interval
ratio
interval
ratio
interval
Age (continuous)
1.1*
1.1, 1.1
1.1*
1.1, 1.1
1.1*
1.1, 1.1
1.0*
1.0, 1.1
Socio-economic factors
Less than secodnary graduation
1.3*
1.2, 1.4
1.1
1.0, 1.2
1.1
0.9, 1.2
1.0
0.9, 1.1
Secondary graduation or more†
1.0
…
1.0
…
1.0
…
1.0
…
Main source of income is social assistance‡
1.3*
1.2, 1.5
1.4*
1.3, 1.6
1.4*
1.2, 1.6
1.3*
1.1, 1.4
Lives alone‡
0.9*
0.8, 0.9
1.0
0.9, 1.1
0.9
0.9, 1.0
0.9
0.9, 1.0
Body mass index (BMI)
Underweight (< 18.5)
2.0*
1.5, 2.6
1.8*
1.3, 2.4
1.9*
1.4, 2.6
Normal (18.5-24.9)†
1.0
…
1.0
…
1.0
…
Overweight (25.0-29.9)
1.1
1.0, 1.2
1.1
1.0, 1.2
1.0
0.9, 1.1
Obese class I (30.0-34.9)
1.6*
1.4, 1.8
1.5*
1.3, 1.8
1.2*
1.0, 1.4
Obese class II (35.9-39.9)
2.2*
1.8, 2.7
2.1*
1.7, 2.5
1.6*
1.2, 2.0
Obese class III (≥ 40)
3.6*
2.6, 4.9
3.4*
2.4, 4.6
2.3*
1.6, 3.2
Smoking status
Smoker
1.4*
1.2, 1.6
1.3*
1.2, 1.5
1.3*
1.2, 1.6
Non-smoker†
1.0
…
1.0
…
1.0
…
Leisure-time physical activity
Inactive
2.6*
2.2, 3.1
2.6*
2.2, 3.0
2.4*
2.1, 2.9
Moderate
1.2
1.0, 1.4
1.1
0.9, 1.4
1.1
0.9, 1.3
Active†
1.0
…
1.0
…
1.0
…
Chronic conditions
Respiratory disease‡
2.6*
2.2, 3.0
2.1*
1.8, 2.5
Cancer‡
2.3*
1.8, 2.9
2.2*
1.7, 2.9
Obesity-related chronic conditions
High blood pressure‡
1.1
1.0, 1.2
Heart disease‡
2.1*
1.8, 2.4
Diabetes‡
1.7*
1.4, 2.0
Arthritis‡
2.4*
2.1, 2.6
Data source:  2003 Canadian Community Health Survey
Notes: Based on records of 38,242 (Model 1), 38,035 (Model 2), 37,968 (Model 3), and 37,783 (Model 4) female respondents who provided information on mobility
function.   Variables for "missing" BMI and main source of income were included in the models to maximize sample size, but the odds ratios are not shown.  Because
of rounding, some odds ratios with 1.0 as the lower confidence interval are statistically significant.
† Reference category
† Reference category
‡ Reference category is absence of condition; for example, reference category for cancer is no reported diagnosis of cancer.
* Significantly different from estimate for reference category (p < 0.05)
… Not applicable
Health Reports, Vol. 17, No. 1, November 2005
Statistics Canada, Catalogue 82-003

---