An Overview of the Canadian Study of Health and Aging
IAN MCDOWELL, GERRY HILL, AND JOAN LINDSAY
ABSTRACT. The Canadian Study of Health and Aging is a multicenter, population-based cohort
study of dementia with a sample of 10,263 participants aged 65 or over. Field work began in 1991,
and a follow-up study was undertaken in 1996-97. The present article describes the origins and
objectives of the study, provides an overview of its design, organization, and data collection
methods, and offers a brief summary of the main results.
Study Sampling in the Canadian Study of Health and Aging
IAN MCDOWELL, RICHARD AYLESWORTH, MARGARET STEWART, GERRY HILL, AND JOAN LINDSAY
ABSTRACT.
The Canadian Study of Health and Aging drew representative samples of people aged 65 or over from the community and institutions across Canada. The sample was designed to
provide regional and national prevalence estimates for dementia by age and sex. Thirty-six
sampling areas were used in a stratified cluster design with optimal allocation; sampling weights
were developed to provide population estimates. The sample included 9,008 people aged 65 or
over from the community, and 1,255 from institutions. This report describes the sampling
procedures, the methods used to recruit people to the study and participation rates, the characteristics
of the resulting sample, and the way in which sample weights should be used.
Data Collected in the Canadian Study of Health and Aging
IAN MCDOWELL, MARGARET STEWART, BETSY KRISTJANSSON, ELIZABETH SYKES, GERRY HILL, AND JOAN LINDSAY
ABSTRACT. The Canadian Study of Health and Aging collected data focusing on the epidemiology
of dementia, using interviews and questionnaires, clinical and neuropsychological examinations,
physical measurements and blood collection, and access to public records such as death
certificates, from people 65 and over in community (N = 9,008) institutional settings (N = 1,255).
The study produced 12 data sets, including community health interviews, clinical and neuropsychological
assessments, risk factor questionnaires, and caregiver interviews. This report describes
the data collection and processing procedures, summarizes the content of each data set, and
outlines the information collected in sufficient detail to permit its suitability for secondary
analyses to be scrutinized.
Study Organization in the Canadian Study of Health and Aging
IAN MCDOWELL, BARBARA HELLIWELL, ELIZABETH SYKES, GERRY HILL, AND JOAN LINDSAY
ABSTRACT. The Canadian Study of Health and Aging was a complex undertaking that faced
management challenges not encountered by smaller-scale projects. The study followed 10,263
elderly people in 18 study centers spanning six time zones; it was administered in two languages,
and over 70 investigators were involved. The data collected from each participant were not fixed,
but varied according to the results of earlier testing. The data could include a screening interview,
a self-completed risk factor questionnaire, an interview with a relative, a clinical examination,
neuropsychological testing, blood samples, and neuroimaging. This report describes the approach taken to organize the study, to track participants, and to
monitor adherence to the study protocol. It also describes the human organizational aspects,
including systems for staff training, for communicating among study centers, and for coordinating
the publication of results. The discussion proposes some guiding principles for administering
multicenter studies.
Correlates of Nonparticipation in the Canadian Study of Health and Aging
BARBARA HELLIWELL, RICHARD AYLESWORTH, IAN MCDOWELL, MONA
BAUMGARTEN, AND ELIZABETH SYKES
ABSTRACT. Correlates of nonparticipation in the community interview component of the
Canadian Study of Health and Aging and their impact on bias in the results were analyzed.
Characteristics of study subjects, their habitats, and encouragement techniques were analyzed to
identify correlates of variation in response rates across the 18 study centers. Refusal rates from 14%
to 41% varied by age, gender, city size, number of subjects and length of time for enrollment, and
method of approach. Cognitively impaired subjects had higher refusal rates which affected
prevalence estimates. At one study site, efforts to “convert” subjects who initially refused to
participate in the survey were successful with 26% of those who were recontacted.
A Recommended Method for Obtaining the Age at Onset of Dementia From the CSHA Database
FABRICE ROUAH AND CHRISTINA WOLFSON
ABSTRACT. In studies of dementia, the age at onset (AAO) of the disease is often described
without indicating how it was obtained. We used the “CAMDEX algorithm,” an ad hoc procedure,
to compute the AAO of dementia from the CSHA database. An AAO could be calculated for 983
of 1,132 subjects with dementia. A similar procedure (the “clinical algorithm”) was used to
calculate a second AAO, which was compared to that obtained by the first algorithm. The
CAMDEX and clinical algorithms produced mean AAOs of dementia of 80.5 years (SD = 9.2 years,
n = 983) and 79.6 years (SD = 9.7 years, n = 829), respectively. The sample correlation coefficient
between the CAMDEX and clinical algorithms was .899 while the intraclass correlation coefficient,
ICC(2,1), was .898. This method could prove useful for researchers using the CSHA data who need
an AAO for those subjects with dementia.
Reliable Individual Change Scores on the 3MS in Older Persons With Dementia: Results From the Canadian Study of
Health and Aging
JOSÉ ANDRÉS CORREA, ANNE PERRAULT, AND CHRISTINA WOLFSON
ABSTRACT. We examined the degree of interrater agreement on the Modified Mini-Mental State Examination (3MS), administered both in the home and at the clinical examination, to determine the boundaries of reliable individual changes for 257 community-dwelling older persons who received a diagnosis of dementia at CSHA-1. Individual score differences were approximately normally distributed (mean of differences 0.2; SD 8.0; 95% confidence interval -16 to 16). The intraclass correlation coefficient was 0.85. Except for the language of testing, there was no relationship between score differences and the determinants investigated (i.e., age, education, type and severity of dementia). This study provides evidence that, in a time frame compatible with no change in cognition, the discrepancy between repeat 3MS scores can be as large as ± 16. These limits represent the range of variability consistent with no change and should be considered when interpreting individual change scores.
Measuring Psychological Well-Being in the Canadian Study of Health and Aging
PHILIPPA J. CLARKE, VICTOR W. MARSHALL, CAROL D. RYFF, AND BLAIR WHEATON
ABSTRACT. The Canadian Study of Health and Aging (CHSA) provided an opportunity to examine the positive aspects of aging. CHSA-2 included the 18-item Ryff multidimensional measure of well-being, which taps six core theoretical dimensions of positive psychological functioning. The measure was administered to 4,960 seniors without severe cognitive impairment
or dementia at CSHA-2. Intercorrelations across scales were generally low. At the same time, the internal consistency reliability of each of the 6 subscales was not found to be high. Confirmatory factor analyses provide support for a 6-factor model, although some items demonstrate poor factor loadings. The well-being measures in CSHA-2 provide an opportunity to examine
broad, descriptive patterns of well-being in Canadian seniors.
Imputation of Missing Dates of Death or Institutionalization for Time-to-Event Analyses in the Canadian Study of Health and Aging
MARIE-FRANCE DUBOIS AND RÉJEAN HÉBERT
ABSTRACT. Data from the Canadian Study of Health and Aging (CSHA) allow investigators to study patterns and predictors of mortality and institutional placement in a well characterized, population-based cohort of elderly Canadians. However, it is impossible to study the timing of these events if the date of occurrence is missing. This technical article describes a procedure for imputing missing dates of death or institutionalization. The first step consists in identifying and correcting dates that are inconsistent with other available dates on which we know the event has or has not occurred. A missing date for an event is then replaced by the middle of a range of plausible dates for its occurrence. This constitutes a valuable addition to the CSHA data since it precludes the loss of information that results from discarding subjects with missing occurrence dates in time-to-event analyses.
Estimating Antemortem Cognitive Status of Deceased Subjects in a Longitudinal Study of Dementia
MARGARET STEWART, IAN MCDOWELL, GERRY HILL,AND RICHARD AYLESWORTH
ABSTRACT. There was a five-year delay between the two waves of the Canadian Study of Health and Aging during which 2,982 participants died. Their cognitive status before death should be taken into account in estimating the incidence of dementia in the cohort. Information concerning antemortem cognitive status was available from death certificates and from an interview with a close relative of the decedent at the CSHA-2 follow-up. The interview included a direct question
on whether the person had been diagnosed with dementia and questions covering cognitive signs and symptoms from which we formed an algorithm to predict probability of dementia. These sources of information were validated using a small sample of study participants who died within five months of undergoing the CSHA clinical examination. Sensitivity of the death certificate and the question regarding diagnosis of dementia was low (33% and 44%), although their specificity was very high. Accordingly, we combined these with the predictive algorithm to form an overall estimate of the probability of antemortem dementia. This raised the sensitivity to 82% (specificity
93%).
Defining Parkinsonism in the Canadian Study of Health and Aging
SUSAN L. MITCHELL AND KENNETH ROCKWOOD
ABSTRACT. This study sought an operational definition of parkinsonism in elderly people (n = 2,914) who underwent a clinical examination in the Canadian Study of Health and Aging (CSHA). Parkinsonism was defined as having two of the following features: (1) bradykinesia of face or limbs, (2) resting tremor, (3) rigidity, and (4) abnormality of gait and posture. The association of parkinsonism with other parkinsonian-related features (prior diagnosis of Parkinson’s disease, use of drugs with extrapyramidal side effects, and use of antiparkinsonian medications) and variables not expected to be related to parkinsonism (stroke and Hachinski score > 5) was determined. Parkinsonism was identified in 337 people (11.6%). It was significantly more likely with other parkinsonian-related characteristics, and was not associated with a history of stroke, but was slightly higher among those subjects with a Hachinski score > 5. Posture and gait abnormalities were significantly associated with other parkinsonian-related variables, but were also more common among subjects with stroke-related features. When the gait and posture disturbance category was excluded as a parkinsonian sign, the narrower definition was more specific but less sensitive in detecting cases with a clinical diagnosis of Parkinson’s disease. Despite limitations, the approach presented in this article is a valid method to operationalize parkinsonism from the dataset.
Assessing Hypertension in the Canadian Study of Health and Aging
HEATHER S. DAVIS, HEATHER R. MERRY, CHRIS MACKNIGHT,AND KENNETH ROCKWOOD
ABSTRACT. We investigated the self-report hypertension variables in the CSHA, recorded in
the screening questionnaire and the Self-Administered Risk Factor (SARF) questionnaire. The
two questions showed high agreement (phi coefficient 0.83). Each was modestly but significantly
associated with other simultaneous reports of heart disease and stroke, and with
subsequent mortality. Only the SARF asked questions about treatment; controlling for treatment
effects, five-year survival was longest among those with no hypertension and no treatment
(mean survival time 1,645 days; 95% CI 1,632 to 1,658), and shortest for those with no reported
hypertension who were receiving ”antihypertensive” medications presumably prescribed for
other cardiovascular disease (mean survival time 1,496 days; 95% CI 1,457 to 1,535). The SARF
questions incorporating high blood pressure and treatment appear preferable to assess the risks
associated with hypertension.
Development and Validation of an Indicator of Support for Community-Residing Older Canadians
BETSY KRISTJANSSON, KRISTA BREITHAUPT, AND IAN MCDOWELL
ABSTRACT. Lack of social support is an important risk factor for disability, psychiatric illness, cognitive impairment, institutionalization, and mortality. Social networks are also important for the caregiving and emotional support that elderly people need to allow them to function well in the community. This article details the development and validation of an index of the instrumental support available to older community residents in the Canadian Study of Health and Aging
(CSHA). Preliminary item review, internal consistency, and exploratory factor analysis were carried out on a random half of the sample. The second half of the sample was used for crossvalidation;internal consistency, exploratory factor analysis, and item response theory analysis were carried out. The final scale had six items; alpha internal consistency was 0.76 and IRT reliability was 0.85. A one-factor solution was most easily interpretable. IRT analyses showed that
the scale was homogeneous and that most items were highly discriminating. The instrumental support scale also had a high correlation with size of social network; it was related to marital status and gender, and predicted institutionalization between the two phases of the study.
Use of the Chronic Disease Score to Measure Comorbidity in the Canadian Study of Health and Aging
CHRIS MACKNIGHT AND KENNETH ROCKWOOD
ABSTRACT. Most older adults have multiple chronic diseases. Consideration of these conditions can improve the performance of statistical models in epidemiological analyses. The Chronic Disease Score (CDS) is a measure of comorbidity derived from medication usage, which may have some advantages over measures derived from other sources. The calculation of the CDS from data contained in the Canadian Study of Health and Aging (CSHA) is described. This measure can be used to estimate comorbidity within the CSHA database.
Effects of Screening Errors and Differential Mortality on the Estimation of the Incidence of Dementia in
the Canadian Study of Health and Aging
GERRY HILL, IAN MACNEILL, RICHARD AYLESWORTH, IAN MCDOWELL, WILLIAM FORBES, AND JEAN KOZAK
ABSTRACT. The Canadian Study of Health and Aging produced an estimate of the incidence of dementia among elderly Canadians by following up, after 5 years, the undemented found in an initial prevalence survey. Initial and follow-up estimates could be biased by false-negative error in the screening tool used for subjects living in the community, and by erroneous classification of subjects who died in the interim. Here, we use a deterministic model to quantify those possible biases. We conclude that, using the estimates of the errors from control samples, the incidence among community subjects would be overestimated by 15%, and the incidence among the institutional subjects would be underestimated by 37%. The overall incidence would be underestimated by 14%. Most of the bias can be attributed to inaccuracies in the classification of deaths.
Linkage of the Canadian Study of Health and Aging to Provincial Administrative Health Care Databases in Nova Scotia
ALEXANDRA M. YIP, GEORGE KEPHART, AND KENNETH ROCKWOOD
ABSTRACT. The Canadian Study of Health and Aging (CSHA) was a cohort study that included 528 Nova Scotian community-dwelling participants. Linkage of CSHA and provincial Medical Services Insurance (MSI) data enabled examination of health care utilization in this subsample. This article discusses methodological and ethical issues of database linkage and explores variation in the use of health services by demographic variables and health status. Utilization over 24 months following baseline was extracted from MSI’s physician claims, hospital discharge abstracts, and Pharmacare
claims databases. Twenty-nine subjects refused consent for access to their MSI file; health card numbers for three others could not be retrieved. A significant difference in healthcare use by age and self-rated health was revealed. Linkage of population-based data with provincial administrative health care databases has the potential to guide health care planning and resource allocation. This process must include steps to ensure protection of confidentiality. Standard practices for linkage consent and routine follow-up should be adopted. The Canadian Study of Health and Aging (CSHA) began in 1991-92 to explore dementia, frailty, and adverse health outcomes (Canadian Study of Health and Aging Working Group, 1994). The original CSHA proposal included linkage to provincial administrative health care databases by the individual CSHA study centers to enhance information on health care utilization and outcomes of study participants. In Nova Scotia, the Medical Services Insurance (MSI) administration, which drew the sampling frame for the original CSHA, did not retain the list of corresponding health card numbers. Furthermore, consent for this access was not asked of participants at the time of the first interview. The objectives of this study reported here were to examine the feasibility and ethical considerations of linking data from the CSHA to MSI utilization data, and to explore variation in health services use by demographic and
health status characteristics in the Nova Scotia community cohort.
Disability and Frailty Among Elderly Canadians: A Comparison of Six Surveys
THE CANADIAN STUDY OF HEALTH AND AGING WORKING GROUP*
ABSTRACT. The first wave of the Canadian Study of Health and Aging (CSHA) constituted a large
health survey of a representative sample of elderly Canadians. Other Canadian surveys from the
same era provided equivalent figures, and the present report compares the results of 6 surveys on
a variety of health indicators. Agreement was close on self-reported chronic health conditions,
adequate for several indicators of functional limitation, but was lower for overall self-ratings of
the impact of health problems on day-to-day life. Using the CSHA data to compare alternative
operational definitions of frailty, a definition based on ADL limitations appeared to offer an
underestimate; addition of IADL questions or cognitive limitations provided figures that appeared
more plausible. Survey estimates of chronic health conditions appear consistent, as are
estimates of certain ADL disabilities. Care must be taken with interpreting more subjective
reports, while prevalence of frailty varies considerably according to the definition used.
Estimating the Prevalence of Dementia in Elderly People: A Comparison of the Canadian Study of
Health and Aging and National Population Health Survey Approaches
VINCE SALAZAR THOMAS, SULTAN DARVESH, CHRIS MACKNIGHT, AND KENNETH ROCKWOOD
ABSTRACT. The Canadian Study of Health and Aging (CSHA) and the National Population Health Survey (NPHS) collected data on the prevalence of dementia in differing fashions. The CSHA used a two-stage method with objective testing and expert judgment, and the NPHS used self-report and proxy data. The present report compares estimates of prevalence and the methodology for ascertainment in the two surveys. The more detailed approach of the CSHA offers the more valid means of estimating prevalence and providing data on subtypes, and can be used in natural history studies. The NPHS measures, including a self/proxy report of diagnosed dementia and a derived cognitive measure, are not sufficiently valid for useful inferences to be made. However, the NPHS method can be improved through supplementation with data on functional
disability, providing age group-specific point estimates closer to the CSHA’s estimates of cognitive impairment and dementia from the community sample. Future waves of the NPHS may wish to include objective cognitive function measures as a cost-efficient and more accurate method of estimating the prevalence of the dementia syndrome without attempting to estimate the prevalence of particular causes of that syndrome.
Reliability and Validity of Questions About Exercise in the Canadian Study of Health and Aging
HEATHER S. DAVIS, KATHLEEN MACPHERSON, HEATHER R. MERRY, CAROLYN WENTZEL, AND KENNETH ROCKWOOD
ABSTRACT. Regular exercise in elderly people has beneficial health effects. We examined exercise frequency and intensity from the Canadian Study of Health and Aging Risk Factor Questionnaire (RFQ). The reliability and validity of these two questions individually, and when combined to form a scale, are reported. Agreement between the self-administered RFQ and an interviewer-administered Add-on Study was examined using intraclass correlations, which were 0.80 for frequency (95% CI 0.77-0.82, p < .001) and 0.75 for intensity (95% CI 0.71-0.78, p = .012). Individuals reporting high levels of exercise frequency, intensity, and a combination of the two showed a smaller proportion of adverse health markers than those reporting no regular exercise. Predictive validity assessed by Cox proportional hazards modeling of mortality
showed that the high and moderate levels of frequency, intensity, and combined exercise groups differed significantly (all p < .001) from the no exercise group. We have found that these exercise questions, though simple, appear reliable and valid. The finding that even comparatively crude exercise questions can demonstrate an important relationship to death suggests that the signal for exercise is a strong one, and future studies should seek to better examine mechanisms by
which exercise benefit is conferred.
Cognitive Impairment, No Dementia: Concepts and Issues
HOLLY A. TUOKKO, ROBERT J. FRERICHS, AND BETSY KRISTJANSSON
ABSTRACT. This article reviews the concept of mild cognitive impairment in groups of people whose cognitive impairment does not warrant a diagnosis of dementia (cognitive impairment, no dementia; CIND). Problems with the application of existing sets of criteria to the Canadian Study of Health and Aging (CSHA) data sets are addressed and a procedure for identifying a subgroup presumed “at risk” for developing dementia is presented. Application of an informant’s report of
changes in cognitive functioning and neuropsychologists’ ratings of mild to severe deficits in any of eight cognitive domains results in approximately half of the CIND cases being identified as “at risk.” The rationale for the collection of specific information related to CIND in CSHA-2 is provided. A minority of people identified with CIND at CSHA-2 showed only memory impairment, and most demonstrated cognitive loss over the preceding five-year interval. This article
provides a conceptual basis for procedures to identify people with cognitive impairment most likely to decline to dementia.
Evaluating Screening Tests for Dementia and Cognitive Impairment in a Heterogeneous Population in the Presence of Verification Bias
ALAN DONALD AND LINDA VAN TIL
This article reviews two potentially serious sources of error in the evaluation of screening
tests, namely, verification bias and the influence of demographic covariates. It demonstrates how to deal with these problems statistically. Verification bias arises when not all subjects receive a definitive diagnosis following a screening test. If only a small proportion of those who screen negative are sent for diagnosis, the calculated test sensitivity is an overestimate and the calculated specificity an underestimate. The methodology outlined in this article may be extended to psychological and
medical screening tests in general.
Measurement of the Influence of the Physical Environment on Adverse Health Outcomes: Technical Report From the
Canadian Study of Health and Aging
CAROLYN WENTZEL, HEATHER ROSE, AND KENNETH ROCKWOOD
ABSTRACT. A paucity of information exists to characterize the relationship between the health status of elderly people and their physical environment. The Canadian Study of Health and Aging (CSHA) is a multicenter study of the distribution of dementia among community-dwelling and institutionalized Canadians aged 65 years and older. The study also provides the opportunity to examine issues such as the physical environment which may be related to the health of elderly people. Six items were used to assess the cleanliness, neatness, and maintenance of the inside and outside of the homes of 8,134 community-dwelling individuals. Data were also obtained to evaluate cognition, physical health, and functional capacity. Five years after the original survey,information pertaining to subsequent institutionalization and/or mortality was obtained. A significant relationship was found between classification of physical environment and the outcomes of institutionalization and mortality. The likelihood of both adverse outcomes was notably higher for individuals living in a “less than ideally maintained environment” compared to an “ideally maintained environment.” Limitations of the six items used to assess the physical environment and ways in which to improve the sensitivity of the items, consequently avoiding
measurement bias, are discussed.
Contribution of Self-Reported Health Ratings to Predicting Frailty, Institutionalization, and Death Over a 5-Year Period
GLORIA M. GUTMAN, ANNETTE STARK, ALAN DONALD, AND B. LYNN BEATTIE
ABSTRACT. Cross-sectional data from Phase 1 of the Canadian Study of Health and Aging was used to examine the relationship between two self-report health measures: “How would you say your health is these days?”(HEALTH) and “How much do your health troubles stand in the way of your doing the things you want to do?”(TROUBLE). The contribution of these measures to
predictive models for institutionalization and mortality is examined, using linked data from Phases 1 and 2. Their relationship to a proposed frailty measure is also examined. At CSHA-1, a majority of respondents perceived that they were in good health and did not feel that their health problems interfered with their preferred activities. At all frailty levels, a majority of both males and females rated their health as “very good” or “pretty good.” As frailty increased, health problems increasingly interfered with normal activities. Logistic regression of the longitudinal data indicated
that, despite their correlation, HEALTH and TROUBLE cannot act as proxies for each other. They appear to predict independently; adding one to the other significantly improved prediction of institutionalization and mortality.
The Canadian Study of Health and Aging: Organizational Lessons From a National, Multicenter, Epidemiologic Study
KENNETH ROCKWOOD, CHRISTINA WOLFSON, AND IAN MCDOWELL
The Canadian Study of Health and Aging was a large, multidisciplinary, national core study—with a number of “add-on” investigations—of the epidemiology of dementia and the health of older people. This structure was a fiscally prudent way to balance between mandated and investigator-initiated inquiry. In hindsight, several important features of the study would be repeated. Future studies might profitably consider a longer funding period for analysis, and a more strategic approach to in-depth, supplementary
studies.
