FUNCTIONAL LOSS AND COGNITIVE DECLINE
M639
function could be predicted. There has been considerable
controversy over whether such a hierarchy can be con-
structed (6,14–18). Whereas some have confirmed this no-
tion (18,19), others have refuted it (6,15,17), and others
have shown that the hierarchy is only one of many ap-
proaches that are possible (20). Other studies (6,14,16,17)
have examined the relationship of specific IADLs (e.g.,
telephone use, finances, shopping, cooking, and cleaning)
within the hierarchy of basic functional activities. Several
groups have proposed measurement scales that combine
ADLs and IADLs into a single hierarchical structure, but
these groups have achieved varying degrees of success, pos-
sibly due to the different patient populations studied. For
example, a distinct hierarchy is difficult to determine using
study populations in which the prevalence of physical dis-
ability and medical illness vary because these will confound
the effect of cognition on function (6,14,17,18).
The Canadian Study of Health and Aging (CSHA) I mea-
sured both the cognitive and functional status of a represen-
tative sample of elderly persons at baseline. Five years later,
a second study (CSHA II) remeasured these variables. This
provided a unique opportunity to estimate the cognitive sta-
tus of persons at the time of the loss of independence of spe-
cific ADLs and IADLs. Within the cohort of community-
dwelling CHSA participants, the objectives of this study
were to (i) examine the relationship between the loss of spe-
cific functional activities and cognitive status during the
time period these losses occurred, (ii) compare the cognitive
status of participants who had loss of specific functional ac-
tivities with those who did not, and (iii) determine whether
there is a predictable hierarchical scale of functional loss as-
sociated with declining cognitive status.
of education. Prospective data regarding the cognitive status
of the participants were collected using the Modified Mini-
Mental State Examination (3MS) (21). The 3MS cognitive
screen was derived from the Mini-Mental State Examina-
tion (MMSE). The 3MS screen includes four additional
items and a score up to 100 points, which improves its abil-
ity to discriminate between those with and without demen-
tia. However, with a cut-off score of 77 (out of 100), the
3MS cognitive screen has a better sensitivity and specificity
than the MMSE in identifying persons with dementia (22).
The functional status of participants was followed using the
Older Americans Resources and Services (OARS) question-
naire (23). The OARS consists of 14 items pertaining to
level of independence in both ADL and IADL (see Appen-
dix). The ADLs were eating, dressing, grooming, walking,
transferring in and out of bed, taking a bath or shower, and
going to the bathroom. The IADLs were telephone use,
transportation out of walking distance, shopping, preparing
meals, doing housework, and taking medication. Each func-
tional item can be categorized as independent, partially de-
pendent, or completely dependent. The CSHA modified the
OARS by substituting a single question regarding the “use
of the bathroom” for two original OARS questions regard-
ing continence.
Statistical Analysis
Only persons who were independent in a given ADL/
IADL at baseline were included in the analyses. This cohort
was then classified into the following two groups: those
who remained independent at 5-year follow-up or those
who became partially or completely dependent at 5-year
follow-up.
In the first analysis, the objective was to describe the
level of cognition associated with a loss of independence in
a specific ADL/IADL. Therefore, subjects who were inde-
pendent at baseline but who became dependent (partially or
completely) in a given functional activity were identified.
We assumed that functional loss occurred at the mid-point
of the 5-year period. Therefore, the best estimate of their
cognitive status at the time of this loss was the mean of their
baseline and 5-year 3MS scores. Standard deviations (SD)
and 95% confidence intervals (CIs) were calculated for each
mean 3MS score.
In the second analysis, the objective was to compare the
change in cognitive status between persons who remained
independent in a specific ADL/IADL task and those who
did not. Thus, for specific functional activities, the mean
change in the 3MS score for each group was calculated by
subtracting the mean 3MS score at baseline from that at 5
years. Analysis of covariance (ANCOVA) was used to com-
pare the mean change 3MS score (dependent variable) be-
tween those who did and those who did not lose indepen-
dence in a specific ADL/IADL (independent variable) while
adjusting for mean baseline 3MS score in each cohort (co-
variate). We felt it was important to adjust for baseline cog-
nitive status because this may affect the degree to which
3MS scores could change over time. Means, SD, and p val-
ues were generated from these analyses. All data were ana-
lyzed using SPSS software (SPSS Inc., Chicago, IL).
METHODS
Study Population
The study population was drawn from the CSHA I and II
studies, which prospectively followed up a randomly se-
lected, representative sample of 10,263 elderly Canadians
for 5 years. A detailed description of CSHA methods has
been reported previously (1). In summary, the CSHA was a
multi-center study of the epidemiology of dementia, health,
and disability among Canadians aged 65 years and over. It
included elderly persons living both in the community and
institutions. Only the community-dwelling sample (n ꢀ
9008) was included in our analysis, which was stratified by
age (65–74 years, 75–84 years, and 85 years and older),
with over-sampling of the older cohorts. Study nurses con-
ducted face-to-face interviews with subjects and/or proxies
in the subjects’ own environment with demographic, medi-
cal, social, cognitive, and functional information systemati-
cally collected.
Definition of Variables
During face-to-face interviews, measurements of baseline
clinical data, including cognitive and functional status, were
obtained. Approximately 5 years later, repeat data collec-
tion was performed on available study participants. Demo-
graphic variables collected included gender, age, and level