University of Southern California Dissertations and Theses

Education and intelligence test scores: Predictors of dementia?

(USC Thesis Other)

Education and intelligence test scores: Predictors of dementia?

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content INFORMATION TO USERS
This m an usc ript has been reproduced from the m ic r o f ilm master. UMI films
the text directly from the original or copy submitted. T h u s , som e thesis and
dissertation copies are in typewriter face, while others m a y be fro m any type of
com puter printer.
The quality of this reproduction is dependent upon the quality of the
copy submitted. B roke n or indistinct print colored or p o o r quality illustrations
and ph oto grap hs , print bleedthrough, substandard m a r g in s , and improper
alignment can adversely affect reproduction.
In the unlik ely event that the author did not send UMI a c o m p le te manu script
and there are m issin g pages, these will be noted. A ls o , if unauthorized
c opy righ t material had to be rem oved , a note will indicate th e deletion.
Oversize materials (e.g., m a p s, drawings, charts) are reproduced by
s ectio n ing the original, beginning at the upper left-hand c om er and continu ing
from left to right in equal sections with small overlaps.
ProQuest In form atio n and Learning
300 North Zeeb Road, A n n Arbor, Ml 48106-1346 U S A
800-521-0600
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
EDUCATION AND INTELLIGENCE TEST SCORES: PREDICTORS OF
DEMENTIA?
by
Kathleen H. Werte
A Dissertation Presented to the
FACULTY OF THE GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(Gerontology and Public Policy)
December, 2000
Copyright 2000 Kathleen H . Werle
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
UMI Number 3065551
Copyright 2000 by
Werle, Kathleen H.
All rights reserved.
U M T
UMI Microform 3065551
Copyright 2002 by ProQuest Information and Learning Company.
A D rights reserved. This microform edition is protected against
unauthorized copying under Tide 17, United States Code.
ProQuest Information and Learning Company
300 North Zeeb Road
P.O. Box 1346
Ann Arbor, Ml 48106-1346
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
UNIVERSITY OF SOUTHERN CALIFORNIA
THE GRADUATE SCHOOL
UNIVERSITY PARK
LOS ANGELES. CALIFORNIA 90007
This dissertation, written by
fo th le e n > JB ;> a W erle.............................................
under the direction of fc.e c Dissertation
Committee, and approved by all its members,
has been presented to and accepted by The
Graduate School, in partial fulfillm ent of re
quirements for the degree of
DOCTOR OF PHILOSOPHY
Dean o f Graduate Studies
Date P.
DISSERTATION COMMITTEE
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
ii
ACKNOWLEDGEMENTS
A number of people have contributed to my attainment of the
Doctor of Philosophy degree and to this dissertation. It gives me great
pleasure to acknowledge them here.
Dr. Elizabeth M. Zelinski, my committee Chair, provided the
inspiration for my thesis and the baseline data and idea for the meta
analysis. Dr. Robert Knight gave freely of his time to review this paper
at multiple points and provided me with insightful comments. Dr.
Elizabeth Graddy was a great listener and validated my understanding of
the results of the analysis, and gave me confidence in my ability to
succeed with this project
I am deeply indebted to my family, who sacrificed meals and
activities so that I could work on this paper. Jennifer will have completed
high school and most of her Master's degree requirements in the time I
have been in school, Ben and Jean have grown into wonderful young
adults, Mark joined our family, and my baby (Katie) is almost a teenager
now. I owe a special thanks to my husband, Jim, who has had to deal
with mood swings and idiosyncrasies that have gotten worse over the
last decade, and my mother, Paula Zielski, who believed that I could
succeed. I would also like to acknowledge my coworkers who have
suffered in this process, particularly Vera Nazarov, who created figures
from cursory drawings, and H. Clay Whitlow, for his support
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
TABLE OF CONTENTS
ACKNOWLEDGEMENTS......................................................................... ii
LIST OF TABLES.................................................................................... v
LIST OF FIGURES.................................................................................vfi
ABSTRACT........................................................................................... viii
CHAPTER I: INTRODUCTION.................................................................1
Background..........................................................................................1
Research Questions............................................................................. 2
Contribution to the Literature................................................................. 3
Organization of the Dissertation.............................................................3
CHAPTER II* 5
THEORETICAL FI^MEWORK AND REVIEW OF THE LITERATURE 5
Theoretical Framework..........................................................................5
Dementia............................................................................................. 9
Intelligence and Its Measurement......................................................... 12
Correlates of Intelligence Scores.......................................................... 17
Heredity...........................................................................................18
Age................................................................................................. 18
Education.........................................................................................22
Summary of Intelligence Correlates Literature.................................... 23
Correlates of Dementia........................................................................23
Heredity.......................................................................................... 24
Age................................................................................................. 24
Gender.............................................................................................25
Education.........................................................................................27
Premorbid Intelligence........................................................................ 43
CHAPTER III: RESEARCH DESIGN...................................................... 67
Sample.............................................................................................. 67
Sample Characteristics..................................................................... 7 1
Procedure.......................................................................................... 73
Dependent Variable............................................................................ 79
Independent Variables.........................................................................79
Intelligence Measures.......................................................................79
Methodology.......................................................................................84
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
iv
CHAPTER IV: RESULTS....................................................................... 9 1
CHAPTER V: DISCUSSION................................................................. 100
REFERENCES.................................................................................... 1 1 1
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
V
UST OF TABLES
Table 1: Years of Education and Age for Subjects With and Without
Dementia........................................................................................ 33
Table 2: Effect Sizes of Education on the Occurrence of Dementia and
Sample Characteristics..................................................................... 36
Table 3: Cumulative Effect Size for Education and Test o f Homogeneity... 37
Table 4: Correlations Between Effect Size and Potential Moderator Variables
.......................................................................................................40
Table 5: Regression of Potential Moderator Variables for Effect Size of
Education on Incident Dementia........................................................4 1
Table 6: Meta-analyses of Education Studies by Potential Moderator
Groupings....................................................................................... 42
Table 7: Comparison of Research Results of Studies Associating
Intelligence Scores and Dementia..................................................... 52
Table 8: Effect of Intelligence Measures on Dementia Grouped by Test.... 54
Table 9: Cumulative Effect Size for Intelligence Measures on Dementia and
Test of Homogeneity......................................................................... 55
Table 10: Study Codings for Potential Moderator Variables..................... 56
Table 11: Bivariate Correlation of Potential Moderator Variables With Effect
Size of Intelligence Scores on Incident Dementia) ............................. 58
Table 12: Regression of Potential Moderator Variables for Effect Size of
Intelligence on Incident Dementia......................................................60
Table 13: Effect of Measures of Fluid Intelligence on Dementia................6 1
Table 14: Effect of Crystallized Intelligence Measures on Dementia 62
Table 15: Effect of Short-Term Memory Abilities on Dementia................. 63
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
vi
Table 16: Meta-Analysis of Similarities Effect Sizes Grouped by Length of
Study.............................................................................................. 64
Table 17: Data Sources..........................................................................7 1
Table 18: Univariate Analysis of Variance Between Current Subjects and
Subjects Lost to Attrition on Demographic and STAMAT Variables 72
Table 19: Numbers of Subjects With Specific Dementia Diagnoses by Data
Source............................................................................................ 79
Table 20: Pearson’s Correlation Coefficients Between Dependent and
Independent Variables......................................................................86
Table 21: Pearson’s Correlation Coefficients Between Dependent and
Independent Variables Using a Matched Sample............................... 90
Table 22: Means and Standard Deviations for STAMAT Variables (Form OA)
.......................................................................................................94
Table 23: Logistic Regression Analysis for Variables Predicting Dementia
Using Both Medical Record and Death Certificate Data )....................96
Table 24: Univariate Analysis of Variance Between Males and Females on
Demographic and STAMAT Variables............................................... 97
Table 25: Logistic Regression Analysis for Variables Predicting Dementia
Using Death Certificate D ata........................................................... 99
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
UST OF FIGURES
Figure 1: Cognitive Reserve................................................................... 9
Figure 2: Sample History....................................................................... 68
Figure 3: Educational Attainment........................................................... 9 1
Figure 4: Recognition Vocabulary Score Distribution.............................. 93
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
ABSTRACT
viii
This study evaluated age, educational attainment length of follow up,
and scores on intelligence measures as predictors of dementia. Review and
meta-analysis of the literature indicated that education would not be a
predictor once the effect of age was controlled for. Scores on measures of
fluid intelligence were posited to be significant predictors, while scores on
measures of crystallized intelligence were not.
Scores on STAMAT intelligence measures of spatial ability and
reasoning (fluid intelligence components) and vocabulary (crystallized
intelligence component), administered in 1978, were available for 360
community-dwelling subjects age 63 and above from the Long Beach
Longitudinal Study. Medical record and death certificate data were collected
on 294 subjects to identify those who had subsequently been diagnosed with
a dementing condition; 29 cases were identified.
The logistic regression on the entire sample (those with death
certificate data and medical record data, N=294) was not significant (chi
square 9.89, ps.13), although length of study was predictive as an individual
variable. The logistic regression using only subjects for whom there were
death certificate data (N=245) was also not significant (chi square 11.07,
p=.09), but age and length of study were predictors of dementia. Years of
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
ix
education and scores on the STAMAT intelligence measures were not
significant predictors. A secondary analysis compared the dementia group
subjects on these variables with a control group matched on gender, age,
and source of data; this analysis was not significant The finding that
STAMAT measures of fluid intelligence were not risk factors for dementia
suggests that the STAMAT measures do not capture the speed effect or
visual/motor coordination that the WAIS measures do.
Consistent with the literature analyses, education and scores on the
STAMAT crystallized intelligence measure were not predictive of dementia
using either the entire sample or subsamples based on data source. Use of
death certificate data may have resulted in an artificially small number of
dementia cases, and a spurious disproportion of cases by gender are offered
as possible confounds. Sample subjects also scored higher on Recognition
Vocabulary, the single measure of crystallized intelligence.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
1
CHAPTER I: INTRODUCTION
Background
There is considerable interest and desire among investigators to
identify risk factors for Alzheimer’s disease and other dementias so that
protective factors can be enhanced and risk factors can be muted.
Knowledge of risk would have utility in targeting individuals most apt to
benefit from therapeutic or preventative interventions likely to be developed
in the near future.
The literature on risk factors for dementia suggests that education and
intelligence test scores obtained years before evidence of cognitive decline
might serve as reliable predictors of dementing illness, and that high
educational attainment and high intelligence levels may mitigate risk. A
number of studies have investigated this possibility, but with conflicting
outcomes. Variables related to education, intelligence test scores, and
dementia may obscure the relationship between these variables.
This study was designed to test the hypothesis that dementia can be
predicted based on an individual’s years of education and performance on
standardized psychometric measures of intelligence. Specifically it will
determine if lower educational attainment and lower cognitive ability as
measured using the Schaie-Thurstone Adult Mental Abilities Test (STAMAT)
are predictive of dementia diagnosis in subjects across as many as 19 years,
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
2
statistically controlling for variables that might influence the diagnosis of
dementia.
This study utilizes data collected on elderly subjects who were part of
a study to norm an intelligence instrument, conducted by K. W. Schaie in
1978. Data has been collected from medical records of consenting
individuals and from death certificates to determine which subjects have
subsequently been diagnosed with a dementing illnesses.
Significant findings would support the theory of cognitive reserve
(Blessed, Tomlinson, & Roth, 1968) which suggests that greater cognitive
ability protects individuals with degenerative brain changes related to normal
aging and brain pathology from reaching a threshold where diagnosis of
clinical dementia is more likely. While the literature suggests these findings,
this is a relatively new direction for research and additional exploration is still
needed in this area.
Research Questions
This study addresses the question of whether education and STAMAT
intelligence test scores are predictive of dementia, as has been suggested in
the literature. Specifically, it will look at the usefulness of education and
STAMAT measures of fluid and crystallized ability as predictors of dementia
diagnosis, having controlled for age and the time in years between testing
and diagnosis.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
3
Contribution to the Literature
This study will differ from other analyses in the distinction between
categories of intelligence measure (i.e., fluid and crystallized ability and
short- term memory) as predictors of dementia, and the consideration of the
time between testing and diagnosis of dementia. The STAMAT measures
used in this study have not been evaluated as correlates or predictors of
dementia in the literature. Sample subjects are part of the Long Beach
Longitudinal Study and are reasonably representative of the senior
population of the Los Angeles and Orange Counties, where most of the
subjects resided at the time of the original data collection. Length of follow-
up in this study is greater than in most studies reported, and certainly longer
than any studies with a comparable number of subjects.
Organization of the Dissertation
This dissertation will be presented in five chapters. The current chapter
introduces the study and defines its potential contributions to the literature.
The second chapter discusses the theoretical framework and the literature
that informs this study. Meta-analyses of the literature on the effect of
education and intelligence on dementia are presented. The hypotheses to
be tested conclude the second chapter. The third chapter describes the
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
research design and the methodology used in this study. Results of the
statistical analyses are presented in chapter four. The final chapter provide
summary and discussion of the results.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
5
CHAPTER II:
THEORETICAL FRAMEWORK AND REVIEW OF THE LITERATURE
A vast amount of research has been published on risk factors
associated with dementia Two of these, educational attainment and
intelligence, have been shown to reduce the risk for dementia in a number of
studies, but results reported in the literature are not consistent This chapter
will examine the body of literature that deals with these two variables as
predictors of dementia and factors that reduce risk. To put the literature into
perspective, this chapter will begin with an overview of the theoretical
framework that guides the investigation.
Second, the construct of dementia and its diagnosis will be reviewed.
Third, the construct of intelligence and its measurement will be reviewed and
studies that investigated a relationship between intelligence and dementia
will be discussed. Fourth, the literature on education as a risk factor of
dementia will be discussed. Last, the hypotheses to be tested in this study
will be outlined.
Theoretical Framework
The point at which the dementia diagnosis occurs has been referred to
as a threshold in a number of published works (Blessed, Tomlinson, &
Roth,1968; Mortimer, 1997; Schmand, Sinet, Geeriings, & Linenboom, 1997;
Snowden et al., 1996; Stem et al., 1995) This threshold is reported to be
buffered by cognitive reserve, the level of which varies among individuals.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
6
Blessed, Tomlinson, & Roth (1968) first used this terminology to report on the
association between qualitatively measured dementia and pathophysiological
changes in the brain on autopsy. They observed that some nondemented
individuals had a high plaque density on autopsy consistent with dementia
pathology. From this the authors concluded that a significant degree of
plaque formation could occur before exceeding the 'reserve capacity” of the
brain (p. 807), at which point intellectual impairment would be manifested.
This threshold more currently refers to a critical volume of functional
brain tissue (Snowden et al., 1996) that results from an increased number of
neurons and greater synaptic density acquired on the basis of stimulation
(Katzman, 1993), an acquired set of skills or repertoires that make the brain
more efficient or resilient (Gudand, 1981; Stem, Alexander, Prohovnik, &
Mayeux, 1992), and/or the amount of functioning brain tissue remaining at
any age (Mortimer, 1997). A related concept is that of plasticity, or
malleability, of brain tissue. Although factors that lead to neuron death have
been studied extensively, little is known about how these cells and tissues
react to and compensate for physiological insults. For example, cell loss has
been found to stimulate remaining neurons to generate new neural circuitries
in humans (Kami et al., 1995), and Swaab (1991) provided evidence from
both animal and human brain studies that stimulation of neurons prolongs
their functioning. In addition, Eriksson et al. (1998) recently reported that it is
possible for human adult brains to generate new neurons. Although it is not
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
7
yet known what factors result in greater individual cognitive reserve,
researchers are optimistic that reserve development is, to some degree, a
lifelong process rather than one that occurs only in youth, and that it can be
altered (Mortimer, 1997).
Because decremental brain changes occur with both normal aging
and pathology, Finch and Tanzi (1997) have argued that everyone will
develop Alzheimer's disease if they live long enough, although this view is
not universally accepted (Mueller et al., 1998). Greater cognitive reserve
could conceivably delay or even prevent the occurrence of dementia if an
individual dies prior to its manifestation (Brody, 1995). Equal brain
decrement in individuals with high and low thresholds may result in incident
dementia only in the individual with a low threshold, depending on survival.
The two variables of interest in this study have been associated with
an increased level of cognitive reserve: intelligence and educational
attainment A low level of intelligence might reduce cognitive reserve and
either 1 ) hasten the detection of dementia by affecting test scores, which
confounds diagnosis, or 2) promote earlier onset of symptoms, in which case
intelligence might serve as a true moderator of risk for dementia (Katzman,
1993; Mortimer, 1988). But Katzman also points out that the diagnostic
criteria for dementia in use over the last decade are fairly conservative.
Because it often takes several months and repeated assessments to meet
these criteria; it is unlikely that a single low intelligence test score would
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
8
result in false positive diagnosis of dementia because additional criteria
would be used for assessment (Katzman, 1993).
Mortimer (1988) speculated that the protective effects of education
might result through more efficient problem solving ability related to
instructional exercises, and increased mental activity throughout life
associated with more prestigious careers. This same argument might also be
applied to intelligence, which is highly correlated with education. Additional
variables associated with education and cognitive ability include heredity,
occupational status, income, health, activity level, and socioeconomic status,
any of which might affect an individual’s level of cognitive reserve (Albert et
al., 1995; Brayne & Calloway, 1990; Brody, 1997; Ceci & Williams, 1997;
Christensen et al., 1996; Jorm et al., 1998; Lindenberger & Baltes, 1994;
Mangione et al., 1993; Perlmutter & Nyquist, 1990; Swan, LaRue, Carmelli,
Reed, & Fabsitz, 1992). These factors influence the type of environment in
which an individual lives and works, and the type and amount of stimulation
provided in their environment
Figure 1 displays the concept of cognitive reserve depicted by
Mortimer (1988, p. 48), modified for this paper. The reserve buffer moves
the threshold for diagnosis so that an individual continues to function at an
acceptable level, and diagnosis (if the individual survives long enough for
decline to reach the threshold) is delayed.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
9
Figure 1. Cognitive Reserve
Threshold for Diagnosis
A Without Reserve Buffer
3
«
3
>
Reserve
Buffer
e
?
o
Threshold fa Diagnosis Q
With Reserve Buffer
Tim e
Whereas Mortimer depicted the diagnosis of dementia simply as the
point at which the cognitive threshold is crossed (point “A'), this figure adds a
reserve buffer that provides additional time between point “A,” the threshold
without a buffer, and point “B,” which allows for additional cognitive decline
before the criteria necessary for a diagnosis of dementia would be evident
The amount of reserve a buffer varies for each individual (Mortimer, 1997).
Dementia
Cognition can be thought of as a continuum of abilities (Huppert &
Beardsall, 1993; Von Dras & Blumenthal, 1992), some of which may decline
as a result of age-associated neural degeneration or disease. At some point
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
10
cognitive abilities can become sufficiently reduced that an individual displays
behavior indicative of serious decline in cognitive function, and incident
dementia or a more specific neuropathology may be diagnosed.
Dementia has been described as a clinical syndrome that involves
global deterioration of an individual's intellect, emotions, and cognitive
facilities without affecting level of consciousness (Roth, 1981). The loss of
intellectual function must be severe enough to interfere with social or
occupational functioning for diagnosis using current Diagnostic and Statistical
Manual of Mental Disorders criteria (DSMIV; American Psychiatric
Association, 1994). The diagnostic category of dementia encompasses
specific medical diagnoses including Alzheimer's disease (AD), multi-infarct
dementia, and about 70 other conditions (Office of Technology Assessment,
1987) which can be difficult to differentiate from each other. Whereas this
study focuses on dementias in general, much of the discussion will be
specific to AD because it is the most common form of dementia and the bulk
of recent literature has been specific to AD.
Memory loss is often the earliest sign of dementia, and is necessary
for diagnosis using DSM IV criteria (American Psychiatric Association, 1994).
Other cognitive abilities are usually impaired at diagnosis as well, and these
abilities usually decline over the remaining lifespan (Cunningham & Haman,
1992). Common diagnostic criteria such as the DSM-IV and the National
Institute of Neurological and Communicative Disorders and Stroke and the
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
11
Alzheimer's Disease and Related Disorders Association (NINCDS-ADRDA;
McKann et al., 1984) work group criteria are useful only after the individual
has experienced significant decline in cognitive functioning (Morris & Rubin,
1991); the date of onset is difficult to determine precisely as a result
Small et al. (1995) reported significantly reduced cerebral parietal
metabolism, a common finding in demented persons, among a small group of
at-risk individuals with mild memory complaints. This suggests that brain
changes associated with dementia may occur well before the diagnosis of
brain pathology. A preliminary report on these and additional subjects
followed longitudinally showed that reduced regional cerebral metabolism
has a high specificity (43 of 46 subjects with low metabolism progressed to
develop dementia) and negative predictive value (3 of 34 subjects who had
negative scans developed dementia) over an average three-year follow-up
period (Silverman et al., 1999).
Clinical diagnosis of dementia requires a complete history and
physical examination and laboratory studies to rule out physical and
psychological problems that can cause memory impairment or other
presenting symptoms. Assessment is required to determine the individual's
current functional status, and in some cases the primary physician may need
to refer a client for neuropsychological evaluation for definitive diagnosis
(Mortimer, 1997; Robinson, 1998). Diagnosticians are required to rely on
clinical judgment in addition to objective data because recommended cutoff
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
12
scores on psychometric tests are not fully reliable and may be confounded by
age, educational attainment, language, and other factors (Lezak, 1995;
Morris & Rubin, 1991).
Because of the high cost of clinical identification of dementia in a large
population of at-risk individuals, the most feasible way to accurately identify
dementia cases is to set up a two-stage design that first screens subjects for
cognitive impairment and then adds clinical testing to separate cases from
non-cases (Jorm, 1990). Some researchers, however, screen for cognitive
impairment but do not follow up with clinical assessment of individuals who
score in the demented range. Low education, sensory deficits, health
problems, psychiatric disorders, developmental disability, and day-to-day
variability may result in screening scores that lead to inaccurate classification
of some participants as demented (false positives) (Mortimer & Graves,
1993). For this reason, studies will only be reported here if clinical evaluation
beyond an initial screening for subjects was conducted, using established
criteria.
Intelligence and Its Measurement
Intelligence can be conceptualized as the ability to learn and deal with
novel information, and reflects those abilities necessary for survival and
valued by the dominant culture (Gardner, 1986; Neisser et al., 1996). The
psychometric approach to measurement of these abilities developed as a
means of predicting an individual’s ability to succeed in educational
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
13
endeavors and as a way of learning about individual and group differences
(Cattell, 1987).
There is some debate as to what intelligence tests, or IQ tests as they
are commonly referred to, actually measure. In his review of the literature,
Schaie (1996) noted that for adult samples, IQ measures correlate with
success in academic and vocational endeavors. Conflicting reports in the
literature about the correlation between intellectual assessment and
everyday functioning (Diehl, Willis, & Schaie, 1995; Salthouse, 1990) raise
the issue of ecological (or external) validity of such tests. For example, Lave,
Murtaugh, & de la Rocha (1984) demonstrated that female shoppers in
California could easily perform mathematical calculations to determine a
comparative value of grocery store products that they were unable to perform
when assessed with a pencil and paper test In addition, substantial gains in
IQ between subsequent generations in numerous countries, attributed to
environmental factors, suggest improved problem-solving ability rather than a
true increase in intelligence (Flynn, 1999).
Walters and Gardner (1986) argue that most intelligence tests only
measure the capacity to answer paper and pencil test questions of a logical
or a verbal nature and overlook important contextual facets of intelligence
and real life application to novel situations. Most authors would concede that
there are additional cognitive abilities such as wisdom, social cognition,
judgment, and creativity that are not measured by standardized tests of
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
14
intelligence in use today (Cattell, 1987; Gardner, 1983; Salthouse, 1991;
Wechsler, 1981a).
Reed and Jensen (1993) observed that both choice reaction time and
visual processing speed are strongly correlated with intelligence scores. This
suggests that IQ tests can be useful to assess neurological integrity and
functioning. The views of the authors cited on this topic are not necessarily
contradictory, but rather provide a realistic appraisal of what IQ tests scores
provide: data that indicate the degree of neurological integration of brain
functions but do not capture the sum of cognitive abilities. In addition, IQ
scores are useful in the prediction of outcomes that rely on specific cognitive
abilities.
Intelligence test scores may be influenced by a wide variety of factors
including sensory deficits, cohort membership, memory ability, health status,
medication effects, discomfort, low motivation, financial difficulties,
misunderstanding of test instructions, fatigue, and anxiety (American
Psychological Association Report, 1998). Sample bias poses yet another
validity issue. Samples used to estimate average adult scores may include
individuals with predinical dementia which would lower group norms
(Sliwinski, Upton, Buschke, & Stewart, 1996), and sample bias related to
willingness to volunteer for such testing may also exist (Salthouse, 1991;
Schaie, 1996).
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
15
It is generally agreed that IQ measures are reliable indicators of real
world cognitive abilities. Performance on the Wachsler Adult Intelligence
Scale is subject to day to day variability yet it remains highly correlated with
other indices of intelligence, and test - retest reliability remains consistently
high (r=< 90) over extended periods of time (Wechsler, 1981b).
Psychometric measures of intelligence enable comparison of individuals and
groups with others on the specific cognitive abilities assessed. Systematic
sampling and testing of subjects has provided much information about
individual and group differences (Neisser et al., 1996). To facilitate
comparisons deviation quotients are used, where tests are standardized to
have a mean score of 100 and a standard deviation of 15 (Salthouse, 1991;
Wechsler, 1981a). Standardized test batteries in use today often include
measures of verbal and mathematical ability, reasoning, spatial orientation,
speed, construction, and memory abilities (Lezak, 1995). Individuals’
intelligence subtest scores generally correlate positively with each other,
indicating that the different abilities are related. Yet individuals will vary in
their performance across tests of separate abilities (Neisser et al., 1996;
Walters & Gardner, 1986).
intelligence tests have been developed with and without benefit of
formal theory as to what they measure. Historically test items were grouped
on the basis of strong correlation with each other in the belief that the same
construct was being measured. Each construct was then intuitively named
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
16
by the researcher. A number of abilities have been identified through this
process which in sum capture what psychologists believe to measure the
essence of human intelligence (Cattell, 1987). Several theories regarding
the structure of intelligence have used factor analysis of test items as the
scientific basis of evidence. One widely used theory in geropsychology is
based on the assumption of eight main grouping factors including Cattail's
fluid and crystallized intelligences plus six additional intelligence domains. In
this view, fluid intelligence is perceived to be primarily biological or innate
and includes abilities such as general reasoning, figural relations, concept
formation, and symbol classification (Horn & Noll, 1997). This largely
inherited component of intelligence has been shown to be vulnerable to
decrement associated with biological aging and diseases affecting the brain
(Gatz, Lowe, Berg, Mortimer, & Pedersen, 1994; Schaie, 1989). Fluid
intelligence scores are the strongest correlate of everyday problem-solving
performance observed in older adults (Diehl, Willis, & Schaie, 1995; Haysiip
& Maloy, 1992; W illis & Schaie, 1986).
Crystallized intelligence is perceived to be affected by environment,
education, and life experience, and it depends to some degree on fluid
intelligence for development Comprised of abilities such as verbal
comprehension, semantic relations, and general information, crystallized
intelligence is assumed to be modifiable well into adulthood. Other
intelligence domains include short-term apprehension and retrieval abilities,
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
17
long-term memory and retrieval abilities, visual processing and spatial
orientation abilities, abilities of listening and hearing, processing speed, and
abilities of quantitative thinking (Horn & Noll, 1997).
Psychometric measures provide a means of assessing the biological
and neural integrity of an individual and indicate gross functioning of the
central nervous system (Cattell, 1987; Eysenck, 1995; Reed and Jensen,
1993); specific intelligence profiles can point to areas of focal damage to the
brain (Lezak, 1995). In addition to their usefulness in neurodiagnostics,
intelligence tests have been used to measure change over time for
individuals and groups and as an outcome measure in studies testing the
effects of interventions designed to modify intelligence (Schaie, 1996).
Correlates of Intelligence Scores
Researchers have uncovered a number of factors that correlate with
abilities measured by intelligence tests. These factors include heredity, age,
education, environment, social dass, parental sodoeconomic status,
occupational status, job performance, income, gender, health, activity level,
and nutritional status (Baron & Cerella, 1993; Brody, 1997; Ced & Williams,
1997; Christensen et al., 1997; Lindenberger & Baltes, 1994; Manton,
Siegler, & Woodbury, 1986; Perimutter & Nyquist, 1990); many of these
variables also correlate with each other. The relationship between heredity,
age, education, and scores on measures of fluid and crystallized ability
measured by standardized intelligence tests will be discussed next
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
18
Heredity
Studies of twins, particularly those raised in different environments,
have consistently shown a high level of heritability for intelligence. Heredity
is estimated to account for 52% of the variance in composite intelligence
scores, 55% of the variance in verbal ability scores, 32% of the variance in
spatial ability scores, and 62% of the variance in processing speed
(McCleam et al., 1997). A summary of 5 studies conducted both in Europe
and the United States reported heritability of .68 to .78 in monozygous twins
reared apart (McGue, Bouchard, lacono, & Lykken, 1993). It should be
noted, however, that strong genetic predisposition does not negate the
influence of environment on the development of intelligence.
Age
Intellectual decline associated with advanced age has been
demonstrated in both cross-sectional and longitudinal research at the group
level (Bray & Howard, 1983; Cunningham & Owens, 1983; Schaie, 1996).
Cognitive decline is not a universal phenomenon, however. Intraindividual
differences are observed, with some elderly persons showing modest
increments in performance over time rather than decrements (Cunningham &
Owens, 1983; Schaie, 1996). Analyses of data from Schaie’s Longitudinal
Study on Aging using the Primary Mental Abilities tests, for example,
demonstrate that reasoning, spatial orientation, perceptual speed, and verbal
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
19
memory decrease slightly but reliably with age, but number and verbal ability
remain stable through much of old age (Schaie, 1996).
There is considerable evidence that much of the age-associated
variance in intellectual performance may be related to differences in
processing resources such as attention, perceptual speed and memory
(Craik, 1986; Salthouse, 1985; Salthouse & Czaja, 2000). The relationship
here appears to be reciprocal: scores on reasoning and vocabulary
measures also predict memory performance (Zelinski, Gilewski, & Schaie,
1993; Zelinski & Stewart, 1998). Reduced speed has consistently been
shown to have an adverse effect on cognitive processes and abilities (Baron
& Cereila, 1993; Hertzog, 1989; Salthouse, 1985), and speed measures
share much of the age-related variance in cognitive ability scores
(Undenberger, Mayr, & Kliegl, 1993; Nettelbeck & Rabbitt, 1992). Further
investigation has shown that vision and hearing deficits account for all of the
negative age differences in speed, indicating that sensory ability is a more
powerful predictor of negative age differences in cognition than the other
processing resources it affects (Undenberger & Bates, 1994).
Recent studies have partialled the effects of perceptual speed, visual
acuity, and working memory on cognitive ability among adults aged 18 and
over, including elderly subjects (Salthouse, Hancock, Meinz, & Hambrick,
1996). Three separate experiments were conducted to determine the unique
and combined variance explained for corrected near visual acuity and age on
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
20
perceptual speed. A strong inverse relationship was found between age and
corrected visual acuity in each of the three samples (r squared = 38 to .51)
using the same vision measure Undenberger and Baltes had used, even
though the Salthouse et al. (1996) samples included young adults. With
education and self-reported health status statistically controlled for in these
carefully conducted studies, commonality analysis indicates that much of the
effect of age on cognitive functions is moderated by a common physiologic
variable that reflects primitive functioning of the central nervous system, of
which corrected vision is one component Although this common
hypothetical variable has not been identified, it appears to account for much
of the decline in cognitive functioning associated with aging.
In a discussion of the theoretical underpinnings of intellectual declines
associated with aging, Salthouse (1991) offered several thoughts based on
his own and other research. First, there is evidence that similar declines on
cognitive performance measures are seen with aging regardless of whether
tests are administered timed or untimed. Second, differences in intelligence
test scores associated with age are small compared with the full range of
individual differences. And third, older adults who agree to participate in
cognitive testing tend to have a higher education and socioeconomic status
than the general population, which could reduce the estimation of true age
differences.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
21
Most people do not have baseline measures against which
intelligence changes can be compared. Normal aging effects would be
evidenced by gradual changes in measured intelligence. Larger or more
rapid decremental changes in IQ scores are more likely to be indicative of
neuropathology, and reflect the severity o f disease.
In conclusion, age is inversely correlated with intelligence scores in
late life. Measured decrement reflects a decline in physiological functioning
of an individual’s processing resources. An overarching variable such as
uncompensated neuronal loss or uncompensated damage that limits or
reduces available processing resources might provide a physiologic basis for
deficits seen in both normal and pathological brain aging. This physiologic
variable could affect multiple processing resources and lead to cognitive
decline.
Intelligence scores are discrete measures of the degree of neural
integrity of the brain, and they correlate strongly with indexes of success in
our society. IQ scores correlate strongly with speed and other processing
measures, and the relationship between sensory processing and IQ scores
has been demonstrated to be a hierarchical one, with sensory processing
mediating the effect of age on IQ scores (Salthouse & Czaja, 2000). This
suggests that both measures of intelligence and sensory processing might be
useful in the prediction of dementia. Intelligence measures are of interest
here, however, because there are a number of longitudinal studies that have
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
22
administered IQ tests to subjects a number of years ago, and data are
available for analysis now.
Education
Another major correlate of intelligence scores is the number of years
of formal education. Wechsler (1958) reported correlations of 0.349 to 0.714
between education and WAIS subtest scores with subjects ranging to age
54, with no observable trend based on the age of the group. Matarazzo and
Herman (1984) found similar results with the WAIS-R. In a sample of
subjects aged 70-103 with varied educational attainment, Undenberger and
Baltes (1994) found that education explained 12.4% of the variance in
intelligence scores when tested alone, but only 3.9% of the variance after
partialling out the predictive power of age, vision, and hearing. They suggest
that among older adults the effect of education is shared with sensory ability,
and that education alone contributes little unique variance in cognitive ability.
The relationship between intelligence and education appears to be
bidirectional; there is no temporal ordering that would substantiate primacy of
intelligence over education (Ceci, 1991; Ced & Williams, 1997). Both
intelligence and education may reflect genetically mediated cognitive
differences (Pedersen, Reynolds, & Gatz, 1996). It has been hypothesized
that individuals with higher IQs may find education more rewarding than
individuals with lower IQs and remain in school longer as a result (Rehberg &
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
23
Rosenthal, 1978), and IQ measures may be used to determine eligibility for
enrollment in programs of higher learning.
Summary of Intelligence Correlates Literature
Psychometric measures of intelligence allow us to observe
intraindividual differences on various facets of intelligence. Scales used to
measure intelligence can be classified as measures of fluid, crystallized, or
other domains of intelligence, and even more specifically in terms of the
constructs they measure.
There are declines on measures of intelligence during later life at the
aggregate level, but considerable intraindividual variation exists. Age and
educational attainment must be considered in any effort to study the unique
contribution of intelligence test scores on any dependent variable, such as
dementia, because of their strong correlation with intelligence scores. At this
point the literature on heredity, age, and education as it relates to the risk for
and prediction of dementia will be explored.
Correlates of Dementia
A number of variables have been identified through epidemiologic
studies that correlate significantly with incident dementia, albeit
inconsistently. These variables include heredity, age, gender, education,
premorbid intelligence, general health status, immunologic and metabolic
abnormalities, viral processes, estrogen replacement, comorbid psychiatric
conditions, head trauma, exposure to organic solvents, and low
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
24
socioeconomic status, to name the more commonly observed correlates;
many of these variables also correlate with each other (Amaducci, Lippi, &
Fratiglioni, 1988; Craft etal., 1999; Leissring, Sugarman, & LaFeria, 1998;
Meyer et al., 1998; Mortimer, 1988; Sramek & Cutler, 1999). The
relationship between heredity, age, education, and dementia will be
discussed in the following paragraphs.
Heredity
Dementia and AD in particular are highly heritable. Family history is
the single contributory factor most consistently identified after age (Gatz,
Lowe, Berg, Mortimer, & Pedersen, 1994; Mortimer, 1990), yet it is possible
that entire families are exposed to common environmental risk factors (Gatz
et al., 1997). Heritability variance is estimated at .74 for AD and .43 for
dementia in general; the remaining variance is explained by other factors
(Gatz et al., 1997). It is likely that environmental factors plus genetic
susceptibility are important to accelerating or retarding disease expression.
In the chronic disease model commonly applied to dementia there can be
both initiating factors and promoting factors that control the onset and
trajectory of disease (Katzman, 1993).
Age
Review of the literature indicates that age is dearly the most important
risk factor for dementia. The rate of dementia occurrence has been found to
rise exponentially with age. For example, inddence rates for moderate and
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
25
severe dementia in the Framingham study were .7% for ages 65 to 69,
2.7% for ages 70 to 74, 5.1% for ages 75 to 79, 8.1% for ages 80 to 84, and
11.8% for ages 85 to 89 (Bachman et al., 1993). Another study, conducted
in East Boston, reported the prevalence of AO alone (not including other
dementias) to be 3% for community residents aged 64 to 74,18.7% for
those 75 to 84, 47.2% for those over 85 years of age, and 10.3% for all
subjects aged 64 and over based on clinical evaluation of subjects (Evans et
al., 1989). Similar rates are reported for the Baltimore Longitudinal Study
sample (Sayetta, 1966). The difference in rates across studies is most likely
the result of methodological differences in the identification of cases (Evans
etal.,1989).
Gender
A number of studies have found a higher rate of AD in women than
men. Jacobs et al. (1995) found that the risk ratio for AD associated with
female gender was 2.5. Aronson et al. (1990) and Rocca, Amaducci, and
Schoenberg (1986) also observed rates of dementia that were more than
twice as high for females as males. Sulkava et al. (1985) reported a higher
prevalence of primary dementia in women (including AD) compared to men,
but men had a higher prevalence of secondary dementias (those due to a
known health problem, such as alcoholism). The overall rates for subjects in
this study age 65 and over were 6.9% for women, and 6.2% for men.
Munoz, Ganapathy, Eliasziw, & Hachinski (2000) found the incidence of AD
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
26
to be 32% higher for women than men. The incidence of AD, however,
appears to be reduced in older women who receive estrogen supplemention
(Paganini-Hall, Buckwalter, Logan, & Henderson, 1993).
While the majority of studies report a greater risk associated with
dementia for females than males, the Framingham study found no difference
in the prevalence of dementia by gender. In addition, a reanalysis of data
from the Rochester, Minnesota epidemiologic study (Rocca, Cha, Waring, &
Kokmen, 1998) determined that the incidence rates for men and women
were similar when examined using age-specific rates. Hall, Gao, Unverzagt,
& Hendrie (2000) found no gender difference in the rate of AD among African
Americans in rural Indianapolis. Age-related increases in cerebrospinal fluid
volume, indicative of brain atrophy, are far greater in men than women,
particularly from frontotemporal, parietal, and pariefo-ocdpital regions of the
brain (Coffee et al., 1998). Additionally, a meta-analysis of dementia
prevalence studies reported between 1945 and 1985 and adjusted for age,
Jorm, Korten, and Henderson (1987) showed that the prevalence of AD
higher for females than males, but the difference was not significant when
age was controlled for.
It is unclear from the literature if gender differences in rate of incident
dementia actually exist Much of the observed risk for females may be
explained by greater longevity, particularly in studies that fail to adjust for
survival effects (Rocca, Cha, Waring, & Kokmen, 1998). In addition, elderly
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
27
women are more likely than men to live alone, and so may be diagnosed with
dementia more readily because they lack assistance in everyday activities,
and come to the attention of a physician as a result This theory is evidenced
by the finding that Alzheimer’s disease patients had a much greater level of
functional impairment on diagnosis if they were living with a son or daughter
than alone or with a spouse (Stanwick et al., 1999).
Education
As a covariate of dementia, education has a less consistent
association with dementia in the literature than age. It is, however, the most
consistent predictor of cognitive status in the elderly (e.g., Crum, Anthony,
Bassett, & Folstein, 1993; Evans et al., 1989; Fratiglioni et al., 1993;
Shichita, Hatano, Ohashi, & Matuzaki, 1986; White et al., 1994). For
example, Pedersen et al. (1996) reported that between 20 and 30% of the
variance in Mini Mental State Exam scores is shared with education and
cognitive ability and that the heritable component of intelligence mediates
both educational attainment and MMSE scores.
Education is correlated with social class, race, ethnic membership,
immigration status, coping ability, social activity, work and life satisfaction,
health, job complexity, and interpersonal skills (Stem et al., 1995), any or all
of which may provide some protection against cognitive decline. In addition,
environmental complexity may promote maintenance of cognitive functioning
throughout the lifespan (Schooler, 1984). Education and/or covariates of
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
28
education may exert a protective effect by providing some level of cognitive
reserve (Canadian Study of Health and Aging, 1994; Stem et al., 1992).
Education may reduce the risk for dementia by 1) creating a screening test
bias; 2) serving as a proxy for any number of putative risk factors such as
good early nutrition, reduced alcohol consumption, or reduced occupational
exposure to neurotoxins such as solvents or pesticides; or 3) increased
cerebral capacity and neuronal reserve (Friedland, 1993; Mortimer & Graves,
1993).
There is evidence of a relationship between education and dementia
in the literature. For example, Stem et al. (1994) reported the relative risk
(RR) for dementia associated with low education was 2.02 (95% confidence
interval (Cl) 1.33 to 3.06) in a longitudinal study conducted in New York City
with 593 subjects, 106 of whom were diagnosed with dementia. The mean
education for subjects in the low education group was 5.3 + 2.6 years,
compared to 13.0 + 2.8 years for the high education group. This study also
found low lifetime occupational attainment associated with greater risk, and
the combination of low education/low occupational attainment to have the
greatest relative risk (RR=2.87,95% Cl 1.32 to 3.63). In addition to the
studies that report mean education level by group membership (dementia, no
dementia) there are also a large number that report significant differences
having coded education as a categorical variable and not reported group
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
29
education means and standard deviations (e.g., La Rue & Jarvick, 1987;
Unverzagt, Hui, Fallow, Hall, & Hendrie, 1996).
Support for education as a risk moderator is also found in the medical
literature. Positron emission tomographic scanning has been used to
measure cerebral regional blood flow in three groups of patients matched for
severity level of AO but varying educational levels. Whole cortex mean flows
within groups were comparable, but parietotemporal perfusion deficit, a
finding specific to AD, was greatest in the group with the highest level of
education. This suggests that the pathology was most advanced in the
subjects with high education, but the clinical expression of AD severity was
no worse than in their lesser educated peers (Stem et al., 1992). These
authors conclude that education or other related cognitive variables provide a
reserve that inhibits clinical expression of AD until the disease is more
advanced. A more recent study reported by the same researchers
demonstrated that greater intellectual, interpersonal, and physical job
demands were related to greater parietal blood flow and lower educational
level in AD subjects. (Stem et al., 1995).
In contrast to the findings in most of the literature, one study found
that education significantly reduced the age at which dementia was
diagnosed (Moritz & Petitti, 1993). In this sample, well-educated subjects
had lower severity of dementia than lesser educated subjects but were
diagnosed at a younger age. The authors concluded that higher education
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
30
leads to jobs in which the loss of cognitive abilities would be more
observable.
Not all studies have shown a differential risk for dementia attributable
to education, however (e.g., Graves et al., 1990; Heyman et al., 1984). For
example, a study of 50 subjects with dementia compared with 223
unmatched, undiagnosed subjects, reported by Prince, Cullen, and Mann
(1994), found that limited education was significantly associated with low
cognitive scores (odds ratio (OR)= 2.12, p<.05) but not with the diagnosis of
dementia. Other studies that found no significant difference in education
between sample groups with and without dementia are reported by
Christensen, Maltby, Jorm, Creasy, and Broe (1992), Crystal et al. (1996),
Howieson et al. (1997), Ganguli, Dodge, Chen, Belle, & DeKosky (2000),
Kaye at al. (1997), Morris and Rubin (1991), and Rubin et al. (1998).
To more closely examine the effect of education on incident dementia,
a meta-analysis of the literature was performed. Meta-analysis is a way of
bringing clarity to a disparate body of literature that cannot be achieved by
examining aggregated studies in a simple vote-counting manner to determine
if the majority of studies support a hypothesis. It allows for nonparametric
statistical comparison and combining of studies by examining the size of the
difference between groups in a single study across multiple samples in a
standardized metric, and offers a means of reducing the likelihood of errors
associated with low statistical power related to small sample sizes in
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
31
individual studies because it pools the data from multiple studies (Hedges &
Olkin, 1985).
The means and standard deviations for education by group
membership (dementia, no dementia) for each of the research studies
identified in the literature are provided in Table 1. The literature search
included online databases in psychology, medicine, health, and sociology,
and was followed by a review of aging and psychology journals from the past
five years. Only published studies using samples from the United States are
included in Table 1, which provides reported means and standard deviations
for education for both a dementia and no dementia group that were not
matched samples. In a number of studies a common control group was
compared to more than one dementia group distinguished in the report by
dementia severity. There is considerable diversity in the samples used in
these studies on both age and ethnic background. These studies indude
samples of Latino (Harwood et al., 1999; Tang et al., 1998) and African
American subjects (Callahan et al, 1996; Tang et al., 1998). They also
induded very young and very dd samples. Two studies were exduded from
the estimation of cumulative effect size and homogeneity because their
variance in education was extremely low (Flicker, Ferris, & Reisberg, 1993;
Grober, Lipton, Hall, & Crystal, 2000). The total number of subjects in the
education analysis induded 1,758 partidpants with dementia and 4,022
partidpants with no dementia (total N = 5,780).
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
32
In addition to significant diversity of samples, a potential confound in
many of these studies is that subjects who were diagnosed with incident
dementia were significantly older than subjects with no dementia In a
sample of subjects with dementia, Koss et al. (1996) showed that age and
educational attainment were significantly and inversely related, with older
subjects having significantly less education. Perhaps more important in
these studies, there is a likelihood that some of the younger subjects in the
no dementia group may yet develop incident dementia by the time they are
the same age as the dementia group subjects. To more closely examine the
relationship between education and age in the studies reported here, mean
age of subjects in both dementia and no dementia groups is provided in
Table 1 as well as educational attainment This simple analysis shows that
in the majority of studies dementia group subjects were on average older
than no dementia group subjects. Severity of dementia, another potential
confound, is also reported in Table 1 .
Meta-analysis uses standard deviation units to calculate an unbiased
estimator of effect size for each study as well as the mean effect size across
studies. An effect size (d) was calculated for each study by dividing the
difference in group mean scores between subjects who developed dementia
and those who did not by their pooled standard deviation, for each measure.
The effect size is a measure of the mean difference between groups scaled
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Table 1: Years of Education and Age for Subjects With and Without Dementia
Study Dem entia Grou p
N Education Mean Severity
S S D )
N o Dementia Group
N Education Mean
rsD )
Significant
Grou p
DMOrenca
D em en tia
G ro u p
AoeM ean
N o Dementia
Grou p
A ge
Difference
Swears)
Callahan et al, 1996 65 8.3 (5.6) mixed 181 10.6(2.7)
•
(SD)
80.9 (13.7)
(3D)
73.6 (5.4) 7.3
Callahan et al, 1996 48 8.7 (6.2) mixed 181 10.6 (2.7)
•
81.9(13.2) 73.6 (5.4) 8.3
Crystal et al., 1996 8 10.4 (3.9) mixed 9 9.3(4)
•
not reported not reported
Crystal et al., 1996 7 9.7 (3.5) mixed 9 9.3(4) not repotted not reported
Flicker et al., 1993 18 12.8 (.04) 1 50 15.1 (.05)
•
69.8 (1.2) 71.9 (.9) •2.1
FHckeretal., 1993 22 12.7 (.06) 3 50 15.1 (.05)
0
72.2 (1.2) 71.9 (.9) .3
Groberat al., 2000 32 11.6 (.53) mixed 232 12.2 (.20) 80.9 (1.1) 76.9 (.41) 4.0
Hall etal., 2000 43 7.8 (3.7) mixed 180 9.2 (3.2)
•
82.7 (6.4) 76.0 (7.1) 6.7
Harwood et al., 1999 392 12.1 (3.4) mixed 202 13.8(2.9)
•
79.9 (6.9) 75.7 (5.9) 4.2
Harwood et al., 1999 188 9.9 (4.9) mixed 84 10.8 (5.2) 78.0 (8.1) 71.5(4.7) 4.5
Haxby etal.,1990 1 1 16.3 (2.1) 1 60 16.6 (2.1) 63(8) 62.5 (10) .5
Haxby etal., 1990 13 13.8 (2.6) 2 60 16.6 (2.1)
•
68(9) 62.5 (10) 5.5
Haxby etal.,1990 8 14.4 (3.5) 3 60 16.6 (2.1)
*
66.0(7) 62.5(10) 3.5
Howieson et al., 1997 18 14.6 (3.5) mixed 116 14.4 (3.1) 89.6 (5.4) 85.4 (3.0) 4.2
Jacobs, et al. 1995 41 7.5 (4.1) mixed 402 11 (4.6)
•
79.4 (8.6) 72.7 (6.6) 6.7
Kaye et al., 1997 12 14.6 (3.8) mixed 12 13.6 (3.4) 90.4 (5.2) 86.8 (1.9) 3.6
Locascio et al., 1995 56 14.0 (3.4) 1 60 15.1 (3.1) 70.7 (8.5) 69.8 (11.2) .9
Locascioetal., 1995 39 12.3 (3.1) 2 60 15.1 (3.1)
•
70.4 (8.0) 69.8 (11.2) .6
fK.05
Reproduced with permission o f th e copyright owner. Further reproduction prohibited without permission.
Table 1 (Continued): Years of Education and Age for Subjects With and Without Dementia
Stu dy Dem en tia Grou p N o D o m o n tia Group Significant D om en tia N o Demantia A g o
N Education Mean Severtty N Education Mean Group Gro u p Group Difference
(SD) (SD) Difference Age Mean Age Mean (years)
______________________________________________________________ (S D ) ( S D )__________
Locascio et al., 1995 18 11.7(4.1) 3 60 15.1 (3.1)
ft
72.3 (9.6) 69.8 (11.2) 2.5
Manly et al., 2000 50 6.7 (4.3) mixed 93 9.1 (4.1)
ft
78.2 (7.5) 73.9 (5.8) 4.3
Morris etal., 1991 10 13.9(3.7) 1 4 13.8 (3.5) 78.9 (5.0) 81.6(6.1) -2.7
Osteiweil et al., 1994 21 7.3 (4.6) mixed 141 9.9 (3.4)
ft
not reported not reported
Rubin et al., 1993 61 12.5 (3.4) 1 122 13.6 (3.2)
ft
73.2 (6.7) 71.2(8.2) 2.0
Rubin et al., 1993 108 12 (3.5) 1 122 13.6 (3.2)
ft
71.3 (7.3) 71.2(8.2) .1
Rubin etal., 1998 27 13.4 (3.4) mixed 55 13.2 (3.2) 73.6 (4.9) 70.7 (4.6) 2.9
Stem etal., 1994 106 7.1 (4.7) mixed 487 10.1 (4.5)
ft
79.0 (8.3) 72.9 (7.0) 6.1
Storandt et al., 1995 50 13.1 (3.7) 1 101 14.4 (3.6)
ft
76.4 (8.9) 79.2 (9.4) -2.8
Storandt et al., 1995 62 12.8 (3.2) 1 101 14.4 (3.6)
ft
75.6 (9.2) 79.2 (9.4) -3.6
Tang etal., 1998 221 6.2 (4.5) mixed 858 8.6 (4.4)
ft
78.0 (6.5) 75.3 (5.8) 2.7
Tierney et al., 1996 21 13.0(3.0) mixed 78 14.3 (3.2) 74.4(7.1) 71.5(7.8) 2.9
Wilson et at., 1999 95 6.4 (4.1) mixed 196 9.6 (2.6)
ft
80.2 (6.5) 73.6(4.7) 6.6
p<.05
35
so that even studies using different but comparable measures can be
compared and aggregated (Hedges & Olkin, 1985).
Results of the meta-analysis are shown in Table 2. This table
shows the effect sizes for each study comparison; where there are more
than one comparisons attributed to the same author the control group is
redundant but the dementia group subjects are different for each
comparison. Hedges and Olkin’s (1985) adjustment procedure for small
sample sizes was applied to the calculation of each effect size to reduce
small sample bias. Confidence intervals were calculated for each effect
size at the p< .05 level, and the effect sizes are significant if the
confidence interval does not include zero.
Potential moderators are also shown in Table 2 coded as dummy
variables. The dementia group is coded 1 as older than the no dementia
group if a significant difference in age between groups was reported; the
remaining studies are coded 0. Dementia group age was coded 1 for
age means that fell between 65 and 74 years, 2 for 75 to 84 years, and 3
for 85 or more years. Severity is coded 1 for mild or very mild, 2 for
moderate, 3 for severe, or mixed.
The effect sizes for education as a risk factor for dementia varied
greatly between studies. Some of the study comparisons had small
positive effect sizes but the majority had moderately large negative
effect sizes. Negative effect sizes indicate that subjects in the dementia
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Table 2: Effect Sizes of Education on the Occurrence of Dementia and
Sample Characteristics
Study
Bib*#
E V V K V
Size(d)
9 6 % Confidence
Limits
L o w e r Upper
Dem en tia
G rou p
Older
DMIMIhI
G ro u p
Dementia
Severity
Callahan et al, 1996 -.62 -.91 -.33 1 t mixed
Callahan et al, 1996 -.61 -.94 -.30 1 2 mixed
Crystal et al., 1996 .27 -.69 1.22 NR* NR mixed
Crystal et al.. 1996 .10 -.89 1.09 NR NR mixed
Hall etal., 2000 -.42 -.78 -.09 1 2 mixed
Haiwood et al., 1999 -.52 -.70 -.35 1 2 mixed
Ha (wood et al.. 1999 -.18 -.44 .07 1 2 mixed
Haxby etal.,1990 -.14 -.78 .50 1 1
Haxby et al., 1990 -1.26 -1.9 -.63 1 1 2
Haxby etal.,1990 -.95 -1.7 -.20 1 1 3
Hovw eson et al.. 1997 .06 -.43 .56 1 3 mixed
Jacobs, etal. 1995 -.77 -1.1 -.45 1 2 mixed
Kaye et al., 1997 .27 -.46 1.01 1 3 mixed
Locasdo et al, 1995 -.34 -.70 .03 0 1 1
Locasdo et al.. 1995 -.90 -1.3 -.47 0 1 2
Locasdo et al., 1995 -1.0 -1.6 -.46 0 1 3
Manly etal., 2000 -.56 -.91 -.21 1 2 mixed
Morris et al., 1991 .03 -1.1 1.19 0 1 1
Ostenweil et al., 1994 -.72 -12 -26 NR NR mixed
Rubin etal., 1993 -.34 -.64 -.03 0 1 1
Rubin etal., 1993 -.48 -.74 -21 0 1 1
Rubin etal., 1998 .06 -.40 .52 1 1 mixed
Stem et al., 1994 -.66 -.87 -.45 1 2 mixed
Storandt et al., 1995 -.37 -.71 -.02 0 2 1
Storandt et aL, 1995 -.45 -.77 -.13 0 2 1
Tang etal., 1998 -.54 -.69 -.39 1 2 mixed
Tierney et al., 1996 -.41 -.89 .08 0 1 mixed
Wilson etal.. 1999
*1 .0 1
-1.27
-.75 ..1_ 2 . .. m ix e d
• Not reported
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
37
group had lower mean educational attainment than subjects in the no
dementia group.
An average effect size was calculated for the entire group of studies to
determine the common effect measured by all the studies, which is the
observed d+ statistic. Then a test of homogeneity was performed to
determine if the group of comparisons share the same effect size.
Homogeneity of studies was determined by calculating the Q statistic, which
is based on the sum of the weighted sums of squares. The Q statistic was
examined relative to the critical value for chi square with the k-1 degrees of
freedom. When the Q statistic exceeds the critical value the hypothesis that
the effect size is equal across studies is rejected, and an attempt is made to
identify moderating variables or differences between studies that explain why
the effect size is not consistent (Hedges & Olkin, 1985). Results of the meta
analysis for all of the comparisons without the Flicker et al. (1993) and
Grober et al. (2000) studies are provided in Table 3, below.
Table 3: Cumulative Effect Size for Education and Test of Homogeneity
N Mean (d+) 95%Cl fo rd * Homogeneity (Q)
Lower Upper
28 -.53 -.57 -.49 64.46*
* Significance indicates rejection of the hypotheses of homogeneity at p<.05
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
38
The combined mean estimator, d+, is significant for the effect of years
of education on incident dementia, but the studies lack homogeneity. When
the value of Q is greater than the critical value of chi square, one or more
variables mediate the observed effect
As was reported earlier, age has been strongly associated with
dementia in the majority of epidemiologic studies. Its relationship to
education as a risk moderator for dementia has at least two possible
explanations. First education is negatively correlated with age (Koss et a!.,
1996), as average educational attainment has increased incrementally with
each cohort over the past decade. Therefore one would expect that older
subjects, at greater risk for dementia, would have lower educational levels
than younger subjects. This effect of age on educational attainment is likely
to be small (if it exists at all) in the studies reported here because the age
difference between dementia and no dementia group subjects spans less
than a generation in all studies. In addition, the effect of age and difference
in age can be statistically controlled for in a regression analysis.
Second, where the subjects in the no dementia group are younger
than subjects in the dementia group, the no dementia group subjects remain
at risk for dementia as long as they live. This right censoring effect, where
some of the no dementia group subjects may be diagnosed with dementia at
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
39
a later date, can skew the results of studies that compare the two groups on
any number of variables. This was demonstrated with repeated studies of
the Bronx Aging Study samples on intelligence, where repeated longitudinal
assessments found significant effects for intelligence on later analysis that
were not present on analysis several years earlier (Sliwinski et al., 1996), as
more of the sample subjects developed dementia. Sixty percent of the
comparisons identified in Table 2 reported that dementia group subjects were
significantly older than no dementia group subjects.
To determine the impact of potential moderator variables for which
data were reported in these studies, correlations between the variables were
examined. The correlation matrix for characteristics for the 28 comparisons
is presented in Table 4. "Dementia group older* refers to the difference in
age between groups, or relative age, shown in Table 2. Correlations were
calculated based on mean age in years and difference in age between
groups in years; dementia severity was coded 1 for mild or very mild
dementia, 2 for moderate dementia, and 3 for severe dementia. Mixed
severity was coded 1 as a separate variable. Nonsignificant effect sizes
were replaced with zeroes in this analysis.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
40
Table 4: Correlations Between Effect Size for Education and Potential
M oderator V ariables_____________ _______________________________
1 2 3 4
Effect 1.00
Size (28)
P =
Dementia -.07 1.00
Group Age (25) (13)
p= .75
P =
Relative -.25 .49 1.00
Age (25) (25) (13)
p= .23 p= .01
P =
Severity of .49 .36 .58 1.00
Dementia
(11) (11) (11) (11)
p= 12 p=28 p=06
p=
Mixed -.23 .65 .76
Severity (28) (25) (25)
(11)
P= 25 p=.00 p=.00
p=
As Table 4 shows, there are no significant correlations with effect size.
Dementia group subject age was correlated with the difference in age
between subject groupings, and also with severity of dementia Subjects in
groups where discrete dementia severity was provided (39% of studies) were
younger than subjects in the mixed dementia groups. In addition, mixed
dementia grouping was significantly correlated with relative age of the
groups.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
41
Separate regression analyses were next conducted with effect size as
the dependent variable. Severity, age of the dementia group, and relative
age entered simultaneously in the first regression; the regression was not
significant (F=3.39, P=.08). A second regression was performed with mixed
severity (coded 1 for yes, 0 for no), age of the dementia group, and relative
age entered simultaneously. The ordinary least squares regression analysis,
presented in Table 5, shows that age of the dementia group subjects and
relative age were both significant at the p< 006 level (F=12.35, p= 002).
Table 5: Regression of Potential Moderator Variables for Effect Size of
Education on Incident Dementia
Variable a B
3 § 8
t Significance
o ft
Dementia Group Age .04 .73 .10 3.79 .005
Relative Age -.70 -.85 .16 -4.43 .002
Mixed Severity .10 .12 .19 .53 .613
Constant -3.25 .70
*
it
< 5 >
The combined mean estimator and corresponding Q statistic were
next calculated for groups of studies to determine if the heterogeneity
between studies would be eliminated, which would confirm the effect of
moderator variables. All of the studies presented in Table 2 (N=28) were
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
42
included in these analyses based on their characteristics. Results are
provided in Table 6, below.
Table 6: Meta-analyses of Education Studies by Potential Moderator
Groupings___________________________________________
N d+
9 5 % Cl ford+
Lower llDDer Q Statistic
Dementia group age 65-74 1 1 -.48 -.52 -.40 23.2*
Dementia group older 3 -.51 -.54 -.49 12.6*
Dementia group not older 8 -.48 -.52 -.41 10.6
Dementia group age 75-84 13 -.66 -.70 -.60 9.33
Dementia group age 85 and over 2 .13 .09 .17 .22
Dementia group older
than no dementia group
15 -.55 -.59 -.51 47.3*
Dementia group age 75-84 10 -.58 -.62 -.53 23.3
No significant age difference
between groups
10 -.46 -.50 -.40 10.9
* Significance indicates rejection of the hypotheses of homogeneity at p=.05
As Table 6 indicates, each of the effect sizes was significant when examined
by moderator variable grouping. Moderator variables accounted for 76% of
the variance in the effect of education on dementia, and homogeneity of
groupings was achieved in all but the youngest subject groups. This
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
43
suggests that education does not have as strong an effect on incident
dementia as has been reported in the literature, where studies have
generally not controlled for subject age, because age has been shown here
to significantly moderate the effect o f education on dementia.
In addition to the moderators identified in the meta-analyses there are
a number of additional variables that could confound the effect of education
on incident dementia. Quality of education, socioeconomic status, income,
occupation, home or work environment, health status and the presence of
chronic diseases, lifelong learning and mental stimulation each could
contribute to the effect of education on incident dementia. The samples
included in the education meta-analysis represent diverse populations,
characteristics of which could also moderate effect sizes.
In sum, the effect of education on dementia is significant with
increased education associated with lower incidence of dementia, but the
effect is significantly moderated by age. The literature therefore indicates
that education will not be a predictor of dementia when age is statistically
controlled for.
Premorbid Intelligence
Premorbid intelligence has been explored as a potential correlate of
dementia with mixed results. By the time a diagnosis of dementing illness is
made intellectual declines have presumably occurred, and premorbid
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
44
intellectual ability cannot be measured directly. Tests used to determine
premorbid ability based on theories of retained or "hold" (versus lost or "don't
hold") abilities have not been particularly useful in persons with AO (Storandt,
Stone, & La Barge, 1995). Longitudinal study of at-risk groups is therefore
most appropriate to test a relationship between intelligence and dementia.
The first to report a relationship between intelligence and dementia,
La Rue and Jarvik (1987) studied aging twins to identify differences between
those who subsequently became demented (n=26) and those who did not
(n=24). Subjects clinically diagnosed with mild, moderate, and severe
dementia had scored lower on measures of intelligence twenty years earlier
than nondemented subjects, after effects for age and education were
partialled out Significant group differences were reported for the Wechsler-
Bellevue Vocabulary, Similarities, Digit Symbol, Digits Forward, and Block
Design tests (p< 02). Dementia group subjects had also scored lower on
Digits Backward and the Tapping test of psychomotor speed, but group
differences were not significant perhaps due to the low statistical power
afforded by such a small sample.
In a similar study, Jacobs et al. (1995) found that the WAIS-R
Similarities test was predictive of AD among 443 subjects who were free of
dementia at the onset of the study and followed prospectively for 1 to 4 years
(mean 2.05 years ± 0.8), controlling for age and educational attainment.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
45
Forty-one subjects were diagnosed with A O at follow up. Using Cox
proportional hazard modeling, the adjusted risk ratio for developing AO was
0.8 for baseline scores on the Similarities test (95% Cl 0.6 to 0.9), which
measures abstraction ability. In other words, higher scores had a protective
function. Verbal fluency was also measured at baseline, but was not
predictive of dementia. Subjects who developed AD in this study were
significantly older and had less education than the no dementia group
subjects, but these variables were not statistically controlled for.
Katzman et al. (1989) followed 434 community volunteers who were
determined to be free of dementia at the start of the study, to identify factors
associated with the subsequent development of dementia. Subjects
completed the WAIS Vocabulary and Similarities tests as well as a mental
status exam annually; subjects who scored low or declined in score were
evaluated clinically. The intelligence measures were not predictive for the 56
subjects who went on to develop dementia during the 5-year follow-up, but
their initial mental status score was predictive.
In a later study using the same Bronx Aging Study sample, Masur,
Sliwinski, Upton, Blau, and Crystal (1994) reported on a group of 317
participants followed prospectively over 4 to 1 1 years to again determine if
psychometric tests could predict dementia. Subjects ranged in age from 75
to 85 and were free of dementia at baseline and at 6 months from testing; 64
were ultimately diagnosed with dementia during the follow-up period. Each
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
46
subject completed the Information, Vocabulary, Similarities, Digit Span,
Digit
Symbol, and Object Assembly WAIS subtests, Verbal Fluency test, and the
Raven Colored Matrices test Missing data were replaced by inserting the
mean value for each group (dementia, no dementia) so that subjects would
be retained in the study. The percent of missing data was not reported; this
could have affected the validity of the results, particularly if it was a large
amount of data. Using a stepwise multiple logistic regression procedure,
significant predictive effects were reported for the verbal fluency test (odds
ratio (OR) 1.47,95% Cl 1.02 to 2.10, p=0.007) and the Digit Symbol test
scores (OR 1.05,95% Cl 1.01 to 1.09, p=0.012); group means for the
various measures were not provided. Lower scores were associated with
greater risk for dementia. Age, gender, and level of education were not
significant when added after the intelligence measures, perhaps because test
performance is correlated with age or there is limited variability on these
variables. Predictions based on logistic regression calculations for individual
subjects determined that the overall sensitivity and specificity of the model
was 50% and 94%, respectively. The model was far better at predicting no
dementia than a dementia diagnosis, which the authors suggest is useful for
clinical work with aging adults.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
47
An interesting facet of this study was a subsequent analysis of length
of time between testing and an outcome of dementia. The model classified
95.8% of subjects correctly when diagnosis was made less than three years
after testing, and misclassified 99.2% of subjects when diagnosis was made
more than six years after testing. This suggests that the length of time
between testing and diagnosis of dementia may be a mediating factor in the
use of intelligence measures for the prediction of dementia, although this
finding is incongruent with studies that have predicted dementia over longer
periods of time.
An extension of the previous studies is reported by Sliwinski et al.
(1996), who examined baseline composite WAJS Verbal (VIQ) and
Performance (PIQ) scores for a group of participants in the Bronx Aging
Study who were free from clinical dementia at the beginning of the study and
followed longitudinally for 4 to 8 years. The Performance Scale includes the
W A J S Digit Symbol, Picture Completion, Block Design, Picture Arrangement,
and Object Assembly tests. The WAIS Verbal Scale includes
Comprehension, Arithmetic, Similarities, Digit Span, and Vocabulary tests
(Lezak, 1995). Ninety of 342 subjects were clinically diagnosed with
dementia during the follow-up period. Subjects who developed dementia
were slightly older (80.5 ± 3.1 years) at baseline than those who did not (78.9
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
48
±2.9 years). Baseline mean V 1 Q scores were 99.8 (±14.8) for subjects who
developed dementia, and 112.8 (+15.9) for those who did not (p<.01).
Similarly, mean PIQ scores were 94.6 (±13.5) for those who developed
dementia, and 106 (±11.9) for those who did not (p<.01). Subtest means
and standard deviations by group were provided by Dr. Sliwinski in a
personal communication dated August 4,1997. Sliwinski and colleagues’
interpretation of the data is that inclusion of subjects with preclinical dementia
biases the norming of psychometric measures downward. Another
interpretation is that a greater number of subjects will develop dementia if a
longitudinal study is continued for a longer follow-up period, based on
analysis of the three studies over time. This has important implications for
longitudinal studies designed to distinguish characteristics of subjects likely
to develop dementia, suggesting that subjects should be followed until death,
and that some additional number might have developed dementia had they
not died from some other cause.
Linn et al. (1995) followed Framingham study subjects who were free
of dementia for 13 years to determine if subjects who developed dementia
during this interval differed from those who did not on a variety of intelligence
measures at baseline. Subjects who were ultimately diagnosed with probable
AD were significantly older than subjects who remained free of AO, but there
were no significant differences on gender or educational attainment between
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
49
groups. Digit Span (summed Digits Forward and Digits Backward) and
Digits Backward were significantly (Afferent between groups, but mean
scores were higher for subjects who developed incident dementia than no
dementia group subjects on each of these measures, which the authors
suggest might could
be due to compensation for other deficits early in AD. The analysis was
conducted again deleting data for subjects who developed dementia during
the first 6 years of the study on the premise that they may have already had
a predinical dementia; the outcome was essentially unchanged. WAIS
Similarities scores were significantly higher for subjects who did not develop
dementia when compared singularly with t-tests, but not when examined with
education effects and other neuropsychological variables partialled in a
regression model. It should be noted that the proportion of subjects who
developed dementia in this study is roughly half that reported in the other
studies reported in the literature, even though subjects were not younger.
Devanand, Folz, Gorlyn, Moeller, and Stem (1997) followed a group of
75 memory clinic participants diagnosed initially with questionable dementia
for a period of one to five years. At follow-up the 3 1 subjects diagnosed with
incident dementia were found to have performed significantly worse on
baseline WAIS-R Performance IQ, Picture Arrangement, Digit Symbol, and
Block Design tests than subjects determined not to have a true dementia
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
50
syndrome. There was no significant difference between the groups on Full
Scale IQ, Verbal IQ, Similarities, Letter Fluency, or Category Fluency tests.
Howieson et al. (1997) studied the oldest group of subjects reported in
this literature (80 to 106 years) for up to five years to identify differences
between subjects who developed dementia and those who did not The rate
of dementia, almost 35%, was considerably higher than in the other studies
reported as a result of subject age, and subjects diagnosed with dementia
were on average 4 years older than those who were not. The dementia
group subjects had a slightly higher mean level of education than subjects in
the no dementia group. Subjects who developed incident dementia also had
lower baseline scores on WAIS-R Block Design and Picture Completion
tests, but no significant differences on Digits Forward, Digits Backward, or
Vocabulary measures, compared to control group subjects.
Table 7 presents a summary of major findings of each of the studies
reported above. There is clearly a lack of consistency both across studies
and within certain studies on different measures.
Each of these studies followed subjects initially free of dementia over
time, and compared the initial intelligence scores of subjects who were later
diagnosed with dementia with intelligence scores of subjects who remained
free of dementia. In addition to published data, intelligence test means and
standard deviations were referred to but not provided in two studies.
Relevant data was solicited and provided by Dr. Sliwinski for the Bronx Aging
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
5 1
Study sample, and from Dr. Jacobs; means and standard deviations provided
via personal communication were used in the meta-analyses. The Katzman
et al. (1989) sample data were not included in the analysis because the
same sample was used in later studies (Masur et al., 1994; Sliwinski et al.,
1996).
Study characteristics and potential moderator variables are identified
in Table 8. Several of the studies reported age by percentage of subjects
who completed a particular grade level. In these cases, mean educational
attainment was calculated from data provided.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Reproduced with permission o f th e copyright owner. Further reproduction prohibited without permission.
Table 7: Comparison of Research Results of Studies Associating Intelligence Scores and Dementia
Author Mean Age and Years of Education No. With Length of Measures**
Dementia Group No Dementia Group Dementia/ Study
__________________________________Total N o .__________________________________
Sample
Characteristics
Devanand et al.
1997
68.6
14.2*
63.0
14.2*
31/75 1-5 years WAIS-R Full Scale IQ Patients from
(mean 2.5) Verbal Scale Score (VIQ) a m em ory
Performance Scale Score (PIQ)** disorder
Similarities clinic
Picture Arrangement
Digit S ym bo l
B lo c k Design**
Word Fluency**
Howieso n et al., 89.6
1997 14.6
85.4
14.4
16/46 1-5 years Digits Fo rw ard
(mean 2.8) Digits Backward
Vocabulary
Picture Completion**
B lock Design**
Oregon Bra in
A ging Study
Jacobs et al.,
1995
79.4
7.46
72.7
10.95
41/443 1-4 years Similarities* North Manhattan
(mean 2.05) Letter and Category Fluency A ging Project
Katzman et al.
1989
80.8 79.8
Not reported; see Masur
56/434 1-5 years WAIS Vocabulary
WAIS Similarities
Bronx Aging
Study
La Rue & Jarvik 84.3*
1987 8.2*$
84.3*
8.2*$
26/50 20 years Similarities**
Digit Symbol**
Vocabulary **
B lo c k Design**
Digits Forward**
Digits Backwards
Twin study
o i
M
‘ De m en tia a n d n o d em en ti a g r o u p c o m b i n e d ‘S ign ific an t d iff ere nc es b e tw ee n g r o u p s at p<.05 $ Calculated f r o m data p r o v i d e d in p u b li c a ti o n
Reproduced with permission o f th e copyright owner. Further reproduction prohibited without permission.
Table 7 (Continued): Comparison of Research Results of Studies Associating Intelligence Scores and Dementia
Author Mean Age and Years of Education No. With Length of Measures** Sample
Dementia Gro up No Dementia Group Dementia/ Study Characteristics
_____________________________________ Total No.________________________________________________
L in n et al., 1995 75.9
11.83$
72.3
12.0$
55/1045 13 years Similarities
Digit Span***
Digits Forward
Digits Backward*
Framingham Study
Masuretai.,1994
Sli winski et al.
1996
79.9
8.16$
80.5
8.38$
78.7
10.19$
78.9
10.01$
64/317 4-11 years
90/342 4-8 years
Information
Vocabulary
Similarities
Digit Symbol**
Verbal Fluency**
Raven Colored Matrices
Digit Span
Bronx Aging Study
Similarities**
Vocabulary**
Information**
B lo ck Design**
Digit Symbol**
Digit Span**
Verbal Scale Score (VIQ)'
Performance Scale Score
B ro nx A ging Study
(PIQ)*
* Dementia and no dementia group combined "Significant differences between groups at p<.05
"•Significant in opposite direction expected (AD score>non AD score) $ Calculated from data provided in publication
8
The effect size for each of the intelligence measures is provided in
Table 8. The majority of measures have a moderate, significant negative
effect size, with higher test scores associated with a lower rate of dementia
Table 8: Effect of In te llig e n t M o a a ra s on Dementia Grouped b y Test
95% Cl
Measure Studv d Lower Unoer
Similarities La Rue & Jarvik (1987) -.93 -1.36 -.51
Devanand et al. (1997) -.35 -.71 .01
Jacobs etal. (1995) -.86 -1.17 -.55
Linn et al. (1995) -.30 -.56 -.03
Sliwinski (1997) -.58 -.88 -.29
Vocabulary Howieson et al. (1997) -.08 -.57 .42
La Rue & Jarvik (1987) -.79 -1.20 -.38
Sliwinski (1997) -.71 -1.01 -.41
Information Sliwinski (1997) -.89 -1.20 -.59
Block Design Devanand et al. (1997) -.77 -1.16 -.38
Howieson et al. (1997) -.87 -1.40 -.33
La Rue & Jarvik (1987 -.46 -.88 -.04
Sliwinski (1997) -.72 -1.02 -.42
Digit Symbol Devanand et al. (1997) -.73 -1.11 -.34
La Rue & Jarvik (1987) -.70 -1.12 -.29
Sliwinski (1997) -1.03 -1.34 -.72
Picture Devanand et al. (1997) -.90 -1.30 -.50
Arrangement
Picture Completion Howieson et al. (1997) -.72 -1.24 -.20
Verbal Fluency Devanand et al. (1997) -.07 -.43 .28
Digits Forward Howieson et al. (1997) -.55 -1.06 -.04
La Rue & Jarvik (1987) -.11 -.47 .26
Linn et al. (1995) .23 -.03 .50
Digits Backward Howieson et al. (1997) -.09 -.58 .40
La Rue & Jarvik (1987) -.19 -.56 .17
Linn et al. (1995) .36 .10 .63
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
55
The cumulative effect size was calculated for all of the intelligence
measures. As shown in Table 9, the cumulative effect size is significant, but
aggregation of measures resulted in significant heterogeneity.
Table 9: Cumulative Effect Size for Intelligence Measures on Dementia and
Test of Homogeneity________________________________________
Mean (d+) 95% Cl for d+ Homogeneity (Q)
Lower Upper
25 -.47 -.04 -.90 134.0*
* S ig n if i c a n c e i n d i c a t e s r e j e c t i o n of the h y p o t h e s e s of h o m o g e n e i t y at p < .0 5
Moderator variables are believed to mediate the cumulative effect
size. These moderators may indude the age and educational attainment of
subjects, length of study, and the type of construct each test measures (e.g.
fluid or crystallized intelligence, short-term memory). Mean age and
education and length of each of the studies are identified in Table 7, and are
shown coded for groups of studies in Table 10. Additional analyses were
designed to be conducted using these codes to compare groups of studies.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Table 10: Study Codings for Potential Moderator Variables
Age*
Variables
Education** Length*** Classification1
Devanand et al., 1997 1 3 1
Block Design 1
Digit Symbol 1
Picture Arrangement 1
Similarities 2
Howieson et al., 1997 3 3 2
Block Design 1
Picture Completion 1
Vocabulary 2
Digits Forward 3
Digits Backward 3
Jacobs etal., 1995 2 1 1
Similarities 2
LaRue and Jarvik, 1987 2 2 3
Block Design
1
Digit Symbol 1
Similarities 2
Vocabulary 2
Digits Forward 3
Digits Backward 3
Linn etal., 1995 2 2 3
Similarities 2
Digits Forward
3
Digits Backward 3
Sliwinski et al., 1996 2 2 1
Block Design
1
Digit Symbol
1
Similarities
2
Vocabulary
2
Infor m a ti o n -----------
2
* M e a n age of d em en tia g r o u p s u b j e c t s 0=<65,1=85-74,2=75-84,3=85 a n d o v e r
** M e a n e d u c a t i o n of d e m e n t ia g r o u p s u b j e c t s 1=1-8 y e a r s , 2=9-12 y e a rs , 3=>13 y e a rs
*** L e n g t h of s tu d y 1*1-3 y e a r s , 2*4-11 ye a rs , 3* >12 y e a r s
**** l*Fluid in te llig en c e , 2*a ys ta lK ze d in te llig e n c e , 3 *s h o rt- te n n m e m o r y
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
57
A bivariate correlation analysis was performed to determine if any of
the potential moderator variables were significantly correlated with effect size
of intelligence measures on incident dementia. The correlation matrix is
presented in Table 11, below. Effect size was correlated with both fluid
intelligence measures and short term memory, but in opposite directions. A
greater negative effect size was associated with lower scores on fluid
intelligence measures, but higher scores on short term memory measures.
Effect size was also inversely correlated with the difference in education in
years between the dementia and no dementia groups, but not with
educational level of the dementia group, age, or difference in ages of the
dementia and no dementia groups. Inverse relationships are also observed
between fluid and crystallized intelligence measures, fluid intelligence and
short term memory measures, and crystallized intelligence and short term
memory measures.
The difference in education between groups is inversely related to the
difference in age between groups. Both education and the difference in age
between groups are inversely related to the length of time between 1978 and
diagnosis of dementia or death. It should be pointed out that the same age
and education were attributed to both (dementia and no dementia) groups in
two studies (25% of measures) where age and educational attainment data
were not provided by group.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
58
Table 11: Bivariate Correlation of Potential Moderator Variables With
Effect Size of Intelligence Scores on Incident Dementia (N-24)
1 2 3 4 5 6 7 8
1. Effect 1.00
Size
P = -
2. Dementia .05
Group Age p=. 8 1
3. Dementia .25
Group
in
CM
£
Education
4. Age .00
Difference p=99
5. Education -.42
Difference p=04
6. Length of .21
Study p= 3 1
7. Fluid -.52
Intelligence p=.01
8. Crystallized -.09
Intelligence p=. 69
9. Short Term .68
Memory p=.00
1.00
P=
-.05 1.00
p= 82 p=.
-.48 .67 1.00
p=.02 p=.00
P =
-.04 -.56 .09
p=.86 p=.01 p=.66
.15 -.58 -.80
p=.49 p=.00 p=.00
-.14 .16 .10
p=.51 p=.46 p=.64
-.04 -.28 -.01
p=.85 p=.18 p=.95
.20 .14 -.10
p— .34 p— .52 p=.65
1.00
P = -
-.32
p=.12
1.00
P =
-.10
p=.64
-.18
p=.40
1.00
P = -
.38
p=.07
-.03
p=.88
-.60
p=.00
1.00
P =
-.31
p=.14
.24
p=.26
-.45 -.45
p=.03 p=.03
Potential moderator variables were next regressed on effect size
using an ordinary least squares regression to see if there was a multivariate
explanation; results are provided in Table 12 below (F=6.97, p= 001).
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
59
Intelligence types were entered into the equation separately; only short
term memory was significantly related to effect size. It is interesting that short
term memory has an inverse effect with the effect size; higher short term
memory scores are associated with smaller effect sizes. Fluid intelligence
attained a near significant t, but crystallized intelligence was not even added
to the equation because it's Beta was zero. The difference in years of age
and education were entered into the equation because both variables had
significant correlations with other variables; difference in years of education
between groups was significant for effect size but the difference in the age of
the groups was not The difference in years of education and age, however,
may not be valid for 25% of studies that do not provide this data by group,
but rather, the same number was assigned to both groups. The remaining
variables were not added to the regression in order to keep the number of
variables entered into the equation low.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
60
Table 12: Regression of Potential Moderator Variables for Effect Size of
Intelligence on Incident Dementia
Variable & B t Significance o ft
Intelligence Measure Type
Short Term Memory .34 .40 .15 2.57 .036
Fluid Intelligence -.25 -.34 .13 -1.97 .063
Difference in Education -.12 -.37 .05 -2.3 .032
Difference in Age .03 .17 .02 1.13 .279
Constant -.50 .13
R* = .51
Because short term memory measure scores significantly moderated the
effect of intelligence measures on incident dementia, meta-analyses were
performed on measures grouped according to the type of intelligence each
test measures based on Horn and Cattel’s (Horn and Noll, 1997) Gf-Gc
theory and grouped by relative educational attainment difference. Table 13
shows the meta-analysis for measures of fluid intelligence. The effect sizes
for these performance measures are all significant and fairly large. In
addition, Block Design and Digit Symbol are both homogeneous for effect, as
is the composite of all measures of fluid intelligence. Another measure of
fluid intelligence, Raven's Colored Progressive Matrices was not significantly
different between groups in Masur et al.'s (1994) study, but means and
standard deviations were not provided for this measure.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Table 1 3 : Effect of Fluid Intelligence Measures on Dementia
6 1
Measure Study d
i
d+
95% Cl
Lower Upper Q
Block D e s ig n Devanand et al. (1997)
Howieson et al. (1997)
La Rue & Jarvik (1987)
Sliwinski (1997)
-.77
-.87
-.46
-.72
-.72
-1.16 -.38
-1.40 -.33
-.88 -.04
-1.02 -.42
-.91 -.52 1.76
Digit S y m b o l Devanand et al. (1997)
La Rue & Jarvik (1987)
Sliwinski (1997)
-.73
-.70
-1.03
-.94
-1.11 -.34
-1.12 -.29
-1.34 -.72
-1.15 -.73 2.18
Picture Arrangement Devanand et al. (1997) -.90 -1.30 -.50
Picture Com pletio n Howi eson et al. (1997) -.72 -1.24 -.20
All Measures of Fluid Intelligence (N=9) -.82 -.95 -.70 5.56
Table 14, below, shows the meta-analysis of verbal tests as measures
of crystallized intelligence. Each of the crystallized intelligence measures had
significant negative effect sizes with the exception of Devanand et al.’s
(1997) Similarities measure and Howieson et al.’s (1997) Vocabulary
measure. When analyzed as a group, Vocabulary was homogeneous for
effect, but the Similarities measures and the composite effect of all measures
of crystallized intelligence were moderated by some other variabies(s).
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
62
Table 14: Effect of Crystallized Intelligence Measures o n Dementia
Measure Study d d+ Lower Upper Q
Similarities La Rue & Jarvik (1987)
Devanand etal. (1997)
Jacobs etal. (1 9 9 5 )
Linn etal. (1 9 9 5 )
Sliwinski (1 9 9 7 )
-.93
-.35
-.86
-.30
-.58
-.50
-1.36
-.7 1
-1.17
-.56
-.88
-.64
-.51
.0 1
-.55
-.03
-29
-.36 11.62*
Vocabulary Howieson et al. (1997)
La Rue & Jarvik (1987)
Sliwinski (1 9 9 7 )
-.08
-.79
-.71
-.66
-.57
-120
-1 .01
-.87
42
-.38
-.41
-.44 5.77
Information Sliwinski (1 9 9 7 ) -.89 -1.20 -.59
All m easures of crystallized intelligence
(N =9)
-.61 -1.00 .13 21.14*
* Significant heterogeneity at p= .05
Table 15, below, shows that the effect sizes for measures of short
term memory ability are nonsignificant with the exception of Howieson et al.'s
(1997) Digits Forward and Linnet al.’s Digits Backward. Both Digits
Backward and Forward include studies with both positive and negative effect
sizes, so it is not surprising that the cumulative effect size for Digits Forward
is not significant and for Digits Backward is significant but small and
heterogeneous.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Table 15: Effect of Short-Term Memory Abilities on Dementia
63
Measure Study d d+
95% Cl
Lower Upper Q
Digits Forward Howieson et al. (1997)
L a R u e & Jarvik (1987)
Linnetal. (1995)
-.55
-.1 1
.2 3
.18
-1.06
-.47
.03
-.02
-.04
26
.54
.38 7.78*
Digits Backward Howie son et al. (1997)
L a R u e & Jarvik (1987)
L inn et al. (1995)
-.09
-.19
.3 6
. 3 1
-.58
-.56
.10
.12
.40
.17
.63
. 5 1 6.73*
All measures of short term memory (N=6) .16 .02 .3015.19*
* S ig n if ic a n t h e t e r o g e n e i t y at p=.05
To determine if moderator variables would reduce or eliminate
heterogeneity for the Similarities measure, meta-analysis was next
conducted on this measure by length of study. As shown in Table 16,
homogeneity was achieved for the group of studies of less than four years
duration, but not for studies conducted over more than 12 years, of which
there are only two. Meta-analyses were not conducted on short-term
memory measures to examine moderator variables because of the small
number of studies in this group. The studies could not be analyzed by the
difference in education between groups because there was only one
measure with a difference in years and two measures without a difference in
years.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
64
Table 16: Meta-analysis of Similarities Effect Sizes Grouped By Length of
Study
Length of Study Number of Studies d+
95% Cl
Lower Upper Q
1-3 years 3 -.71 -.89 -.52 4.52
Over 12 years 2 -.34 -.56 -.11 6.12*
* S ig n if i c a n t h e t e r o g e n e i t y at p=.05
Based on these analysis, short term memory measures were
significant moderators of the effect size of intelligence on dementia, but
would not be expected to directly predict incident dementia. The cumulative
effect size for Digits Forward was not significant, and the effect size for Digits
Backward was not homogeneous, and proved to be moderated by the length
of the study. Fluid intelligence measures were not significant predictors of
effect size in the regression analysis, but their effect sizes were generally
large and significant, and their cumulative effect size was large, significant
and homogeneous. Fluid intelligence measures should also be predictors of
dementia. Crystallized intelligence measures were not significant in the
regression analysis, and one measure was significant and homogeneous
(vocabulary) but there were only three studies included in the meta-analysis,
and the other measure (Similarities) was significant but moderated by the
length of the study, with homogeneity achieved only for the group of studies
that were three or less years in length.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
65
Data available for subjects in the current study indude scores for
the STAMAT Recognition Vocabulary, Letter and Word Sets, and Figure and
Object Rotation measures. STAMAT measures of intelligence will be
described in detail in the next chapter. These measures indude four
measures of fluid intelligence and a single measure of crystallized
intelligence, which will be used to test the hypotheses of this dissertation.
As discussed in previous sections of this paper, age remains the
greatest and most consistent risk factor for dementia. As demonstrated in
the meta-analysis, education does not have as strong an effect on inddent
dementia as has been reported in the literature, where studies have
generally not controlled for subjed age, because age has been shown here
to significantly moderate the effect of education on dementia. In addition,
fluid intelligence measures would be expeded to predid inddent dementia,
but short term memory measures and crystallized intelligence measures
would not Based on findings from the literature review and synthesis
presented here, this study seeks to test the following hypotheses, stated in
the affirmative:
1) Age of subjects w ill be positively assodated with risk for inddent
dementia,
2) Lower educational attainment will be assodated with greater risk for
inddent dementia,
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
3) STAMAT Object and Figure Rotation and Letter and Word Sets scores,
measures of fluid intelligence, will be inversely associated with the risk for
dementia, and
4) STAMAT Recognition Vocabulary scores, a measure of crystallized
intelligence, will be inversely associated with risk for dementia.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
67
CHAPTER H I: RESEARCH DESIGN
Sample
The original participants were recruited from a health maintenance
organization based in Orange County, California by K . Werner Schaie and
colleagues in 1978. Of nearly 3000 members stratified by age who were
invited to participate in normative testing of the STAMAT, approximately
equal numbers of men (250) and women (258) participated. Of the 508
subjects between the ages of 55 and 89 who participated in the original
study, 349 identified their socioeconomic (SES) status in 1978, which was
then ranked by the researchers on a scale of zero (unskilled) to 9
(professional). The mean SES rating was 5.01 (SD=1.85). Of the original
sample, 493 had valid intelligence scores on the STAMAT Recognition
Vocabulary test. Figure Rotation test, Object Rotation test, Word Sets, and
Letter Sets, in addition to memory measures. Subjects had been randomly
assigned to take each of the two spatial measures (Figure Rotation and
Object Rotation) and the two reasoning measures (Word sets and Letter
Sets) in a counterbalanced presentation order (Zeiinski et al., 1993). There
were 360 subjects between the ages of 63 and 89 who were tested in 1978
who were the focus of this study. Age 63 was selected as the lower age
parameter because it was the age at testing of youngest known subject in the
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
68
current study to develop dementia. The history of this sample is shown in
Figure 2, below.
Figure 2: Sample History
Timeline Sample Subjects
1978
1999
Subjects Not Included
508 onginal
subjects
493 with valid
STAMAT
scores
15 without valid
STAMAT Scores
u
360 age 63 and
above
133 age 62 and
below
. a
294 subjects
with medical
record or death
certificate data
66 subjects
without follow-
up data
Subjects from the original sample were sent a letter explaining the
current study and a consent form for release of medical information, as well
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
69
as a postage-paid, addressed return envelope in 1994. The HMO permitted
access to computerized membership data to check addresses and next of kin
listing. Regional telephone books were also used to locate subjects. The
Health Care Finance Administration and the National Death Index conducted
a Social Security Number match and provided addresses or confirmed death
of subjects in 1994. A second and third mailing or telephone call was made
to subjects believed to be at a particular address (i.e., correspondence was
not returned as undeliverable) who had not responded in an effort to include
as many of the original subjects in the current study as possible. One
hundred ninety-five consents to examine medical records were obtained from
subjects, their guardians, or their next of kin. In addition, 16 subjects or their
next of kin declined to give consent
The HMO destroys records of members who have died seven years
after their death; this resulted in a loss of potential data on 15 subjects for
whom consent was obtained. Medical record data collected on 75 subjects
were analyzed in this study. The death certificate search was updated in
1999 using the National Death Index Plus, which provided cause of death
information on matched subjects who died prior to 1998. This search
updated records for subjects whose medical records had been reviewed but
who died since the initial search. Death certificate data were collected on
245 subjects, including 3 1 for whom medical record data was available.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
70
In addition to medical record and death certificate data, a third source
of data was used. Twenty-six of the original 1978 subjects consented to
participate in this study but their medical records were not available. Five of
these participated in a third assessment as part of a longitudinal memory
study by Dr. Elizabeth M. Zelinski in 1994, and were determined to be
cognitively intact at that time. STAMAT data on these 5 subjects were
included in this study, and although there was no medical information
available for them, their data were analyzed with the data of nondemented
subjects with 1994 coded as the last year of follow-up.
This provides a sample of 294 subjects for this study, 29 of whom
were ultimately diagnosed with dementia. This means that 9.86% of study
subjects were identified as having developed dementia during the follow-up
period. The correlation between source of data for diagnosis (death
certificate or medical record) is -.73 (p=.000); there are only five dementia
group subjects for whom both medical record and death certificate data are
available. Table 17 shows the source of data for subjects included in this
study.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
71
Table 17: Data Sources
All Subjects
Medical Record Data 75
Death Certificate Data 245
Subjects With Both Medical Records and Death Certificates < 3 1 >
Subjects From 1994 Data Collection 5
Total 294
Subjects with Dementia
Medical Record Data 10
Death Certificate Data 24
Subjects With Both Medical Records and Death Certificates <5>
Total 29
Sample Characteristics
Differences in characteristics between the current sample and original
sample subject not included in the present study were examined using
Analysis of Variance (ANOVA). ANOVA was used to evaluate the mean
difference between groups on age and education. Multivariate Analysis of
Variance (MANOVA) was next used to examine the mean differences
between groups of STAMAT test scores because it statistically controls for
correlations between variables under investigation (Bray & Maxwell, 1985).
As Table 18 shows, the subjects lost to attrition were significantly
older than subjects for whom study data is available. Those lost to follow-up
had scored consistently higher on the STAMAT variables than the subjects in
the current study sample as well, but only one of the univariate analyses
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
72
were statistically significant The MANOVA was not significant for
differences between groups (F=1.2, p=31).
The gender distribution for the current sample is not significantly
different from the original study sample but it is significantly different than the
subjects who were not located for follow-up (F=4.57, p= 033), which included
23 males (34.8%) and 43 females (65.2%).
Table 18: Univariate Analysis of Variance Between Current Subjects and
Subjects Lost to Attrition on Demographic and STAMAT Variables
Variable Current Study Subjects Original Subjects N ot
(N=294) Included (N -66)
Mean (SD) Mean (SD) ANOVA Results
Age 72.49 years (5.72) 69.86 years (4.98) F= 11.86 p= 001
Education 12.22years(2.87) 12.71 years(2.83) F=1.59 p=208
Univariate MANOVA
Results
STAMAT
Recognition
Vocabulary
37.39 (10.79) 39.14 (9.83) F=1.46 p=.228
Figure
Rotation
15.60 (7.89) 17.70 (9.27) F=3.55 p=.06
Object
Rotation
20.34 (10.71) 23.76 (11.73) F=5.29 p=.022
Letter Sets 8.96 (5.33) 10.06 (6.13) F=2.16 p=. 143
Word Sets 9.43 (5-02) 10.33 (5.04) F=1.74 p=.188
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
73
Procedure
Medical record data were collected at the HMO regional warehouse. A
standardized data collection form was used to record data on each medical
diagnosis indicative of a chronic health condition and its earliest date of
diagnosis. Data on the neuropsychological workup for dementia and
psychiatric illness were recorded on a separate form. Medical records for
subjects with a dementia diagnosis were determined by the researcher to
include an adequate medical workup prior to a diagnosis of dementia. The
researcher is a Registered Nurse who has expertise in medical terminology
and documentation as well as assessment and identification of dementia.
When a subject identified a health care provider who was not affiliated
with the HMO on the consent form, that provider was sent a short letter
describing the study, a copy of the signed consent form, and a simple
medical history survey. The physician survey solicited data on the medical
diagnoses and year onset or permission for the researcher to review the
records at the provider site. The survey sent to physicians was short and
simple to fill out, and a self-addressed, stamped envelope was provided to
improve the return rate. Of the 34 physician surveys mailed out, 10 were
returned completed, four copies of medical records were mailed to the
researcher, and 3 physicians allowed the researcher to review records at a
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
74
private medical office. The remaining 17 were either not returned or were
returned as undeliverable. A second mailing to physician offices netted no
additional subject data
Medical record data were collected on 75 subjects. Ten medical
records were reviewed separately by a clinical psychologist who consulted
on this study to determine reliability of the data extracted. Interrater reliability
was calculated to be .95.
Death certificates were reviewed at county records facilities in Los
Angeles, Orange County, and Riverside. San Bernardino County had no
public viewing facility but records identified as a likely match were purchased.
The 1999 National Death Index Plus provided cause of death
information for subjects identified as a likely match with study subjects.
Criteria for inclusion of death certificate or cause of death information
included a match on at least two of the following: year of birth, Social
Security Number (available for some but not all subjects), place of residence,
or next of kin where known from medical record or National Death Index
data. Exact date of birth was used in place of year where the date was
available from the medical record. Death certificate data interrater reliability
was determined by having a master's prepared Registered Nurse
independently code data from 10 death certificates from copies of the
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
certificates and 10 from the data collection sheet used by the primary
researcher interrater reliability was calculated at 1.0.
Last, consenting subjects for whom no data was available but who
participated in longitudinal retesting for a study by Dr. Elizabeth Zelinski in
1994 were included in the current study. These 5 subjects were deemed
cognitively intact by the researchers testing them at that time. These
subjects were identified separately in the data file, with last entry date of
1/1/94 given because they were tested at some time in 1994; they were
coded simply as not having dementia
Individual level data were made available by Dr. Zelinski for each of
the STAMAT measures, age, income, and educational attainment. Medical
record and death certificate data, collected without knowledge of intelligence
test scores, and information about the dementia status of the 5 subjects on
whom medical history was not available, was merged with the original data.
The major threat to validity in this study is that death certificate data is
the predominant basis for dividing subjects into the dementia or no dementia
groups. Health records are a documentation of historic events, and validity
depends on both the accuracy and completeness of recorded data, as well
as the abstraction process, which is subject to systematic and random error
at multiple points. The validity of data extracted from records has been
relegated to the expertise of the researcher or extractor and is generally
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
76
more valid if a standardized data collection format is used and the data
extractor skilled or trained in the methods of collecting data (Linet, Harlow,
McLaughlin, & McCaffrey, 1989), as it was in this study. The reliability of
data has been relegated to evaluation of interrater reliability in many studies,
which is important but does little to ascertain the reliability of the data in the
record itself.
Aaronson and Burman (1994) identified six factors that affect the
validity of medical records: 1 ) clinical competence of the medical
practitioners; 2) patient cooperation and compliance; 3) type of provider; 4)
setting of care; 5) situational factors; and 6) type of data. Clinical
competence is difficult if not impossible for the researcher to determine.
Many HMOs have developed standardized criteria and decision pathways for
management of costly patient problems, which improves the validity of
diagnosis and documentation in the medical record (Eddy, 1996).
The data collected for this study consisted of documented chronic
health problems and their date of diagnosis if known. Because the data
reflect health problems for which a patient would see a physician repeatedly,
it is believed that these conditions were documented a high percentage of
the time. The Problem Oriented Medical Record system was used
consistently in these records; it provides a list of health problems and dates
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
77
of onset for each patient so that a physician who did not know the patient
would be aware of conditions the patient had been treated for. In addition,
this HMO had received a three-year NCQA (National Council on Quality
Assurance) certification in 1995, an exceptional achievement, for the quality
indicators observed during their review. NCQA is a voluntary certification
designed for HMOs, sought and used as a marketing tool because it
represents a standard of excellence. Additionally, the HMO received no
State Department o f Corporations citations for violations during the period
from 1978 to its merger with another HMO in the mid-1990s (personal
communication, Janet Newport, July 8,1997).
The use of death certificate data, particularly cause of death and
secondary or contributing causes of death, is a cause for concern in this
study. A substantial number of studies have demonstrated that incident
dementia is underdocumented on death records for people with known
dementing conditions (Beard et al., 1996; Bums, Jacoby, Luthert, & Levy,
1990; Lanska, 1998; Macera, Sun, Yeager, & Brandes, 1992; Martyn &
Pippard, 1988; Olichney, Hofsetter, Galasko, Thai, & Katzman, 1995). Beard
et al. (1996) found that only 11.9% of 917 Mayo Clinic patients with
Alzheimer's disease who died between 1960 and 1984 had a dementia
diagnosis listed on their death certificate. Macera et al. (1992) noted that a
dementia diagnosis was recorded on 23% of death certificates for 450
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
78
participants in an Alzheimer's registry who died between 1988 and 1990.
The rates have improved in recent studies in the United States, where an
estimated 24% o f documented dementia cases are still unreported on death
certificates (Olichney et al., 1995).
There have been concerted efforts by the federal government over the
last two decades directed at improving the validity of death certificates
because they are a source of data used for policy making. Most states
(including California) had adopted a nationally standardized certificate of
death by 1997, which provides space for the physician to list contributing
causes of death, but compliance in the use of the form varies (Magrane,
Gilliand, & King, 1997). Physicians historically have not been instructed on
the completion o f these forms (Magrane, Gilliand, & King, 1997; Messite &
Stellman, 1996) and may not recognize the utility of thoroughness in this
task.
The overall rate of dementia in this sample is 9.86%, including the 5
subjects without medical record or death certificate data. The rates
separated by source of data are 13.3% for subjects with medical record data
and 9.8% for those with death certificate data. It is believed that the
documentation o f dementia on the death certificates reflects valid cases, but
cases of dementia might have been omitted as contributing or comorbid
conditions.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
79
Dependent Variable
The dependent variable is evidence of clinical diagnosis of dementia
in the medical record or death certificate. All forms of chronic dementia were
combined because of the very small number of specific dementia diagnoses.
The number of subjects identified with each dementia diagnosis is provided
in Table 1 9 .
Table 19: Numbers of Subjects With Specific Dementia Diagnoses by Data
Source
Diaanosis Medical Record Death Certificate Total*
Alzheimer's Disease 5 7 9
Arteriosclerotic Cerebrovascular Disease 1 5 6
Multi-infarct Dementia 0 1 1
Organic Brain Syndrome 0 4 4
Senile Dementia 4 3 6
Dementia, not specified
_ Q
4 3
Total 10 24 29
*Nonredundant cases
Independent Variables
Intelligence Measures
The STAMAT (Schaie-Thurstone Adult Mental Abilities Test) is a
modification of Thurstone and Thurstone's (1949) Primary Mental Abilities
(PMA) Test which was derived from a factor analysis of approximately 60
other measures of intelligence (Schaie, 1979). Although the PMA was
designed for use with adults of all ages, Schaie’s interest in older adults led
to the identification of specific issues of concern in intelligence test
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
80
administration in the Seattle Longitudinal Study on Aging. The STAMAT
Older Adult ( Form OA) version was adapted to improve the validity of scores
for older adults. Five subtests of the PMA measure what are believed to be
different facets of intelligence, and can be summed to approximate
intellectual ability and educational aptitude (Schaie, 1979). Two additional
tests were created for the STAMAT Form OA which are parallel measures of
the Object Rotation and Letter Sets scales; both were administered to this
sample. STAMAT administration to elderly subjects is facilitated by using
large type disposable booklets that do not require transcription to computer
scan forms. Although administered as timed tests, as the PMA tests are,
STAMAT test scores may be affected less by subjects’ perceptual speed
than the PMA because answers are recorded directly in the test booklet
(Zelinski et al., 1993). Indeed, the STAMAT has been shown to reduce age
differences attributable to perceptual and motor speed declines (Hertzog,
1989). Parallel measures of the two of the tests were designed to increase
the meaningfulness of the exercises by using stimulus items that are familiar
to adults. Cautiousness of older adults is addressed by providing information
to the test taker on when it is appropriate to guess unknown answers and
when wrong answers will result in a penalty (Schaie, 1985).
In addition, Popkin, Schaie, and Krauss (1983) found that the order in which
the parallel tests were administered affected test scores. Specifically,
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
subjects who took the Figure Rotation test after the Object Rotation test
had higher scores on Figure Rotation than subjects who took this test first
Subjects in this study were assigned alternate testing order on a random
basis, but there is no documentation of the order of testing for subjects who
were included in this study.
Each of the tests administered to this sample will be discussed
separately here. Two additional STAMAT measures, Number and Word
Fluency, were not administered to this sample.
The Recognition Vocabulary test assesses understanding of words
in a multiple choice test format Thurstone and Thurstone (1949) described
their original test equivalent the PMA Verbal Meaning test, as the ability to
understand verbally expressed ideas with a speed component This ability is
used in activities in which information is obtained by reading or hearing
words. This multiple-choice test requires a subject to select a synonym for
the word targeted in each test item. The test is comprised of 50 items, and
the difficulty level increases throughout the test Four minutes are allowed
for test completion. This test is scored simply as the number of correct
answers, with a maximum score of 50 (Schaie, 1985). Recognition
Vocabulary is believed to be a measure of crystallized ability (Schaie, 1989).
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
82
The Figure Rotation test measures the ability to identify figures when
they are rotated in the same two-dimensional plane. This can be a difficult
concept to grasp and initially some people will select the mirror image, which
is incorrect To increase validity, the test-taker is exposed to sample
questions and the correct answers to make certain the expectation is dear.
Figure Rotation uses a multiple choice format with 6 answers to choose from.
Like its PMA equivalent, the Space test there is a 5-minute time limit for test
completion. The score is determined by subtracting the number of incorrect
responses from the correct responses. There are 20 items; multiple answers
on each item are correct The maximum possible score is 54 (Schaie, 1985).
Figure Rotation test measures spatial orientation, a fluid ability (Schaie,
1989).
The Object Rotation test is a parallel form of the Figure Rotation test
The major difference is that test items are line drawings of familiar objects
rather than unfamiliar geometric figures. Like Figure Rotation, it has a 5-
minute time lim it there are 20 questions, and the score is determined by
subtracting the number of incorrect responses from the correct responses.
The maximum possible score is 30 (Schaie, 1985).
The Letter Sate test is a purported measure of inductive reasoning
ability, equivalent to the PMA Reasoning test Subjects solve ordering
problems in which they dedde which letter comes next for example in the
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
83
item abxcdxefxghx, and circle the correct letter from several choices. Six
minutes are allowed for the completion of this test Subjects must identify a
pattern or rule in the presentation order to be able to identify the letter that
would logically follow. The number of correct responses is recorded, with a
maximum score of 30 possible. There is no penalty for guessing (Schaie,
1985).
The Word Seta test is the parallel form of Letter Sets, added as part
of Form OA. This test presents ordered words rather than letters, which add
to the meaningfulness of the stimuli for older adults. For example, days of
the week or months of the year are presented in a pattern that must be
discerned to be able to identify the day or month that would come next The
test-taker circles the correct answer in the answer booklet. Like the Letter
Sets test, there are 30 items scored by summing correct answers, and the
subject has 6 minutes to complete the test Both Letter Sets and Word Sets
are believed to be reasoning measures reflecting fluid abilities (Schaie,
1989).
Validity and reliability of the PMA and STAMAT scales have been
reported in several studies. For example, Schaie (1965) reported that the
PMA variables on which the STAMAT is based maintain their factorial
separation, as evidenced by correlations which were consistently less than .5
over 21 years of testing. Likewise, Hertzog (1989) found factorial stability for
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
84
the STAMAT Form A variables, which did not include the parallel tests,
over time. Test-retest reliability coefficients for the five scales used in this
study were each reported at .80 to .86 (Schaie, 1985). Gardner (1986)
observed, however, that some of the Form OA scales were so brief that they
yielded relatively large standard errors. And as previously mentioned, the
STAMAT has been shown to reduce the observed age difference between
young-old and old-old subjects to about half that seen with the PMA (Popkin
et al., 1983), and so is likely to be a valid assessment tool used with older
adults.
The STAMAT Manual (Schaie, 1985) provides correlations for the
variables in the STAMAT Form OA: for adults of all ages Figure Rotation
and Object rotation were correlated at r=.78 and Letter Sets and Word Sets
were correlated at r=.85 (p values are not reported). The development of the
Object Rotation scale was an effort to improve content validity, in that the
exercise more closely approximates a task that might be necessary in
independent living. Studies that demonstrate the ecological validity of the
STAMAT related to functional competence, however, have not been
published.
Methodology
The parallel forms of the spatial orientation and reasoning measures
are believed to capture the same constructs as the original PMA tests. The
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
85
correlation between the parallel spatial orientation tests in the current study is
.71 (p= 000), so these measures were combined to reduce the number of
variables being entered in the analyses. The correlation between the two
reasoning measures is .76 (p=.000), therefore these measures were
combined as well. Pearson’s correlation coefficients for variables are
presented in Table 20, below.
The relationship between gender and dementia was significant when
examined with a chi square analysis; 21 females developed dementia
compared to 8 males in this sample. Gender was inversely associated with
the Combined Rotation Score variable, a measure of fluid intelligence; men
significantly outperformed women on this measure. Female gender and the
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Reproduced with permission o f th e copyright owner. Further reproduction prohibited without permission.
Table 20: Pearson's Correlation Coefficients Between Dependent and Independent Variables
1 2 3 4 5 6 7
Dementia
1.00
P =
Age .07
p=.23
1.00
P = -
Gender
(female)
*
.00
p=96
1.00
P =
Education -.03
p=,57
-.14
p*.01
-.04
p=.49
1.00
p=
Combined
Rotation Score
-.01
p = .8 1
-.23
p=.00
-.24
p=.00
.12
p=.03
1.00
P =
Combined Word
and Letter Sets
.03
ps.59
-.30
p -.O O
.10
p=.08
.27
p=.00
.36
p=.00
1.00
P =
Recognition
Vocabulary
.00
p=.99
-.17
p -.O O
.04
p-48
.29
p=.00
.24
p=.00
.56 1.00
P =
Years after
Testing
.13
p=.03
-.24
p=.00
. 2 1
p=00
.04
p=51
.04
p=.53
.13
ps.02
.14
p-,02
a= chi square (1,29)^5.83, p=.016
8
87
number of years between testing and dementia outcome were significantly
correlated, but years after testing was also correlated with age of the
subjects; each of these correlations are small. The relationship between
gender and years between testing and outcome shows that women I ived
longer beyond testing than men in this sample, but that demented males
were older than demented females at diagnosis.
Age and each of the STAMAT measure scores were inversely related;
greater age was correlated with less education and lower test scores. Age
was also correlated with the number of years between testing and outcome,
as would be expected. Higher educational attainment was assodated with
higher scores on each of the STAMAT variables, and each of the STAMAT
variables correlated with each other as well.
The correlations obtained suggest that a regression analysis would
not demonstrate significant capability of the STAMAT variables for predicting
dementia, although gender and years between testing and outcome would be
significant predictors.
Logistic regression analysis was used to obtain parameter estimates
for each of the independent variables of interest with a diagnosis of dementia
as the dichotomous outcome variable. This method was selected to test the
hypotheses in this study because it provides an analysis of the relationship
between several exogenous variables and a dichotomous dependent
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
88
variable based on identification of those variables that explain the greatest
amount of variance. It allows the program to select the order in which
variables are added based on their contribution to the likelihood ratio chi
square. In addition, logistic regression remains reasonably robust even
when the variables are not normally distributed (Norusis, 1993).
Missing STAMAT variable scores were imputed from ordinary least
squares regression coefficients for age, educational attainment, and
STAMAT scores for the remaining tests; 3.7% of all STAMAT scores were
imputed. Regression models were run separately using 1) all subjects, 2)
subjects for whom medical record data are available, and 3) the subjects for
whom there is only death certificate information. The SPSS program used to
compute the statistics selected the variables that had the greatest effect on
outcome variable from among the variables selected: age, educational
attainment, years between testing and outcome, Recognition Vocabulary,
combined Word and Letter Sets, and combined Object and Figure Rotation.
The likelihood chi square was used to determine if the model correctly
predicted a dementia diagnosis significantly better than the intercept alone.
Results of logistic regression analyses yield parameter estimates
(odds ratios) which, if significant, allow for prediction of an individual's odds,
log-odds, and probability for a dementia diagnosis based on knowledge of
that person's scores on each of the variables. These parameter estimates
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
89
define the marginal effect of each variable on the probability of a dementia
diagnosis as in other general linear models. The probability for a given
individual is determined by summing the constant plus the sum of each
coefficient times the value of the variable. The odds are then calculated by
dividing the probability of no event by the probability that the event will occur
(Knoke & Bohmstedt, 1994).
The coefficient (b) represents the effect of the variable as a function of
one of the two values on the dependent variable, which in this study are 0
(no dementia diagnosis) or 1 (dementia diagnosis); it is constrained to fall
between 0 and 1 (Knoke & Bohmstedt, 1994). Results of these analyses will
be provided in Chapter IV.
A second analysis was performed to compare subjects who developed
dementia with a group who did not, matched on gender, age, and source of
data (medical records or death certificate). The correlation matrix for
variables using the matched sample is presented in Table 21, below. In this
sample, none of the variables were correlated with incident dementia. Age
was inversely correlated with gender (males were older than females), with
scores on each of the STAMAT variables, and years between testing and
outcome. A Multivariate Analysis of Variance was computed to compare the
groups on the STAMAT variables, but there was no difference between
groups (F=1.36, p=.25).
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Reproduced with permission o f th e copyright owner. Further reproduction prohibited without permission.
Table 21: Pearson's Correlation Coefficients Between Dependent and Independent Variables Using Matched
Sample
1 2 3 4 5 6 7
1 . Dementia 1.00
P =
2. Age
-.00
p=98
1.00
P =
3. Gender
*
-28
p=.03
1.00
P =
4. Education .02
p=. 88
-.14
ps.29
. 0 1
p=91
1.00
P =
5 . Combined
Rotation Score
.09
p=. 5 1
-.4 1
p=.00
-.08
p=.54
.23
ps.09
1.00
P =
6. Combined W ord
and Letter Sets
.05
p =.7 1
-.52
p=.00
. 3 1
p=.02
23
p=.09
.46
p=,00
1.00
P =
7. Recognition
Vocabulary
-.12
p=.38
-.34
p = .0 1
.16
p=.24
.38
p=.00
.37
p=,00
.46
p=.00
1,00
p=
8. Years after
Testing
. 2 1
p=.1 1
-.38
p=,00
.36
p=00
.20
p=.13
.16
p=.22
.47
p=.00
.46
p -.O O
a= chi square (1,58)=11.65, p= 0006
8
CHAPTER IV: RESULTS
9 1
This chapter will present the results of analyses designed to test the
hypotheses of this study, which were presented at the end of chapter two.
Prior to conducting these analyses, however, exploratory analyses were
conducted, including examination of the frequency and distributions of each
variable. Each of the STAMAT variables except Recognition Vocabulary,
age, education, and years after testing was normally distributed as evidenced
by a non-significant K-S statistic.
Educational attainment for this sample is depicted in Figure 3, below.
Mean educational attainment for this sample was 12.2 (S.D. 2.9) years, with
56% having completed high school, and 25% having completed more than
n
i i m ii n
2 9 4 $ 9 7 • 0 10 11 12 13 14 15 tS 17
EDUCATIONAL ATTAINMENT
Figure 3: Educational Attainment
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
92
two years of college. The mean educational attainment for all people in the
U.S. who were age 55 and above in 1978 was 11.0 (U.S. Department of
Commerce, 1979). This sample had a somewhat higher mean level of
education than the population this sample was designed to represent In
addition, the effect size of education on incident dementia in this study was
nonsignificant (d=-.11,95% C.l. -.49, .27).
Recognition Vocabulary scores were not normally distributed for this
sample. Out of a range of zero to 50 possible, the mean score was 37.39
(S.D. 10.79). As Figure 4 below indicates, a large number of subjects scored
perfect or near perfect scores, indicating that a ceiling effect occurred.
Recognition Vocabulary scores may be significantly higher in this
sample than in the general population, where scores would be expected to
be normally distributed.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
4A flO
Ricognllloii Vosibulary Scorn
Figure 4: Recognition Vocabulary Score Distribution
The mean STAMAT scores of the participants are compared with the
normed scores provided in the STAMAT manual in Table 22 below. This
current sample was part of the sample used to norm the STAMAT version
OA in 1978.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
94
Table 22: Means and Standard Deviations for STAMAT Variables (Form OA)
Measure Sample Mean (sd) Normal Group Mean (sd)
Recognition
Vocabulary
37.39 (10.79) 30.13 (9.46)
Object Rotation 20.34 (10.71) 20.71 (9.89)
Figure Rotation 15.60 (7.89) 16.45 (13.82)
Letter Sets 8.96 (5.33) 8.90 (5.27)
Word Sets 9.43 (5.02) 9.06 (4.54)
The current sample, with a mean age of 72.5 years, scored dose to
the age 71-77 norms reported by Schaie (1985) in the Schaie-Thurstone
Adult Mental Abilities Test Manual for all measures except Recognition
Vocabulary. It is conceivable that the scores of this sample biased the norms
upward on this measure. The correlation between the Recognition
Vocabulary scores and education was significant at .29. This suggests that
approximately 8% of the variance in Recognition Vocabulary scores may be
due to educational attainment in this sample. It is not possible to identify
other reasons why the scores are higher for this sample than the rest of the
population tested with these measures.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
95
Effect sizes were calculated for the STAMAT variables in this study.
The effect size for Recognition Vocabulary was .003 (C.l. -.38, .38); the effect
size for the combined Word Sets and Letter Sets variable was .25 (C.l. -.13,
.63); and the effect size for the combined Figure Rotation and Object
Rotation variable was -.04 (C.l. -.42, .39). Unlike the majority of other
studies, none of the intelligence measure variables had a significant effect on
incident dementia.
The results of the logistic regression analyses are shown in Tables 23
and 25, below. These tables list the estimated coefficient, represented as
the log odds, and the odds ratio for each variable. Each variable, including
the intercept or constant, has one degree of freedom. The odds ratio
represents the actual change in the odds with one unit change in the
variable. For example, the odds for the development of dementia for each
additional year between testing and outcome is increased by a factor of .12
based on the data in this particular model.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
96
Table 23: Logistic Regression Analysis for Variables Predicting Dementia
Using Both Medical Record and Death Certificate Data (N=294)
Coefficient Odds Ratio
Age 07 1.07
Education -.05 .95
Recognition Vocabulary -.01 .99
Spatial Orientation -.00 1.00
Reasoning .03 1.03
Years After Testing .12 1.12*
Constant -8.74
Likelihood ratio chi square 9.89
Degrees of freedom 6
p-value .13
* p = . 0 1
The model comprised of all subjects who were followed up for this
study was not significant, and only years after testing was significant A
second regression on this same group of subjects was performed, this time
using 1-3 years between testing and outcome as a dummy variable, with 1-3
years coded 1 and all other lengths coded 0; the model was not significant
(chi square 5.48, p= 48 with 6 degrees of freedom). The models predicted
no cases of dementia, so all 29 subjects with dementia would have been
misdassified.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
97
The mean age of subjects in this sample in 1978 was 72.5, and
dementia subjects were only slightly older than no dementia subjects (mean
ages 73.4 and 72.3, respectively). The incidence of dementia for this sample
is 9.86%, however it is 5.5% for males and 14.1 % for females. There are
some additional differences between males and females in this sample,
which show up on univahate analysis of variance between males and
females in this sample, results of which are shown in Table 24, below.
Table 24: Univariate Analysis of Variance Between Males and Females on
Demographic and STAMAT Variables
Variable Male Study Subjects
(N=145)
Mean (SD)
Females Study Subjects
(N= 149)
Mean (SD) ANOVA Results
Age 72.50 (5.76) 72.47 (5.7)
F Significance of F
.002 .96
Education 12.34 (4.94) 12.11 (2.69) .474 .49
STAMAT
Recognition
Vocabulary
36.94 (10.68) 37.83 (10.93) .496 .48
Rotation 40.18 (16.76) 31.82 (16.75) 18.29 .00
Word and
Letter Sets
17.41 (9.19) 19.36 (10.15) 2.97 .09
Follow-up
(years)
9.69 (4.94) 12.11 (2.69) 13.53 .00
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
98
As Table 24 shows, male and female subjects are equally matched on
all variables except for the spatial orientation measure (combined Object
Rotation and Figure Rotation), on which males scored higher, and years of
follow-up, which was longer for females. Another logistic regression was
performed adding gender in addition to the variables reported above; the
model was not significant (chi square 13.91, p=..53 with 7 degrees of
freedom).
The regression model results using medical record data only included
data on 75 subjects, 10 of whom were identified as having a dementia
diagnosis. The incidence of dementia for this subsample is 13.3%. None of
the variables in this model were a significant predictor of dementia diagnosis
at the p s 05 level when entered in a stepwise fashion.
The results of the regression model for subjects with death certificate
data is presented in Table 25. This subsample consisted of 245 subjects, 24
of whom were identified as having a dementia diagnosis. The incidence of
dementia for this subsample is 9.8%. Unlike the model using all subjects,
this model showed the number of years between testing and follow-up to be
significantly associated with incident dementia. Age and gender were not
significant predictors. Additional variables were not entered into the stepwise
regression because they increased the log likelihood by less than .01
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
99
Table 25: Logistic Regression Analysis for Variables Predicting Dementia
Using Death Certificate Data (N=245)__________________________
Coefficient Odds Ratio
Age .09 1.09*
Education -.05 .95
Recognition Vocabulary -.02 .98
Spatial Orientation .0 1 1 .0 1
Reasoning .04 1.04
Years after testing .14 1.15**
Intercept -9.69
Likelihood ratio chi square 11.07
Degrees of freedom 6
p-value .09
*p<05 **p=. 007
percent This model also predicted no cases of dementia, so that all 24
subjects with dementia would have been misdassifed.
The logistic regression was next run on subjects with death certificate
data using a dummy variable for years between testing and outcome (1-3
years coded 1, ail others 0); the model was not significant (chi square 5.79,
p=.45 with 6 degrees of freedom). Data for this group of subjects was
regressed again including gender as a variable; the model was not significant
(chi square 13.92, p=.053 with 7 degrees of freedom).
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
100
CHAPTER V: DISCUSSION
Each of the hypotheses tested will be discussed in order. This
chapter will conclude with a summary of information gained as a result of this
study, and suggestions for further research.
The first hypothesis was that the age of subjects would be positively
associated with risk for incident dementia. This hypothesis was only partially
supported in this study. The logistic regression on the full sample for whom
follow-up data are available did not find age to be significantly associated
with greater risk for dementia, but the model using only data from subjects
with death certificate outcome data showed that each additional year of life
increased risk by a factor of .09 (p=.047). Greater age had been correlated
with lower educational attainment and lower scores on each of the STAMAT
variables, as would be expected based on the literature review, and also with
fewer years of follow-up. Age was not, however, correlated with incident
dementia.
While age was a significant predictor in the regression, using only
subjects with death certificate data, it was not predictive in the full sample.
There are several possible explanations why age did not predict dementia in
the full sample; age attained a near significant B (p=.055) however. Years
between testing and outcome achieved significance in both regressions; for
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
101
each additional year the risk increases by a factor of .12 to .15. Age and
length of study were correlated (r= 13, p= 03).
Another reason that age might not have been a significant risk factor
for dementia is that the full sample had a lower rate of dementia (10%) than
the sample using medical records data (13.3%), suggesting that as many as
1 1 cases of dementia might have been missed, applying the rate of the
medical records sample to the death certificate sample. This difference
would be significant in a sample with only 29 cases of incident dementia.
The same rate of dementia, however, was observed in the sample using
death certificate data only, and age was a significant predictor in this sample.
The literature clearly indicates that age is the most important risk
factor for dementia; the rate of dementia occurrence has been shown to rise
exponentially with age (Bachman et al., 1993; Evans et al., 1989; Sayetta,
1986). It is likely that the effect of age was confounded in this study by length
of follow up or some other sample irregularity.
The second hypothesis predicted that lower educational attainment
would be a significant risk factor for incident dementia This hypothesis was
not supported. Education was not correlated with incident dementia, and did
not change the likelihood chi square significantly in the regression models.
Education was not significantly correlated with STAMAT Word Sets, Letter
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
102
Sets, Object Rotation or Figure Rotation in this study, but it was significantly
correlated with Recognition Vocabulary (r=. 37, p=. 001). As discussed
earlier, Recognition Vocabulary scores are believed to be much higher in this
sample than in the normal population.
The mean educational attainment in this sample is slightly higher than
the population of adults age 63 and over in California. Educational
attainment of males and females in this study is comparable, however. The
effect size for education is not significant (d=-.11, C.l. -.49, .27), but this is
not inconsistent with the literature which, when subjected to meta-analysis,
showed that age accounted for much of the effect of education on incident
dementia.
The third hypothesis was that STAMAT spatial orientation and
reasoning measure scores, indicative of fluid intelligence, would be inversely
associated with the risk for dementia This hypothesis was not supported by
the analyses. There were no correlations between combined rotation
measures and combined word and letter sets and incident dementia, and the
regression analyses demonstrated no significant effect on incident dementia.
Scores on these particular measures were within the normal range for
similar-age subjects. Female subjects, however, had significantly lower
scores on the combined rotation scores than males, and the rate of dementia
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
103
was significantly higher in females. While female subjects were not older
than male subjects in 1978, their average follow-up period was significantly
longer (see Table 24; F=13.53, p=.000), so that at the time of diagnosis or
last follow-up they would have been older than male subjects. Longer
survival may have contributed to the fact that subjects who developed
dementia were overwhelmingly female (72.4%), at a ratio larger than many
other published studies, which may have influenced the results. The study
included very similar numbers of male and female subjects both initially and
for in the sample for whom follow up data are available, and there were
similar numbers of subjects with death certificate data who were male
(N=125) and female (N=120). Of the 75 subjects for whom there is medical
record data, however, there are only 28 males (37.3%) but 47 (62.7%)
females. This may reflect a tendency of women in this study to seek medical
attention at a greater frequency than men, with a subsequently higher rate of
diagnosis of dementing conditions. In addition, there could be a bias,
unconscious or deliberate, in either the diagnosis or the notation of dementia
diagnosis on death certificates and medical records.
The literature analysis is strongly supportive of the hypothesis that
higher fluid intelligence scores would be associated with reduced risk for
dementia. Meta-analysis of intelligence measures had indicated that
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
104
measures of fluid intelligence would have the greatest effect on incident
dementia; effect sizes ranged from -.72 to -.94 and were homogeneous.
None of the studies reported used STAMAT measures of intelligence,
however. These instruments may have diminished age effects that are seen
using WAIS measures. WAIS tests reported in the literature may measure
different constructs than the STAMAT spatial orientation and reasoning tests.
For example, Digit Symt)Ol is reported to rely heavily on visual and motor
coordination and speed (Belsky, 1990), which is not measured with the
Figure and Object Rotation measures or the Word or Letter Sets measures.
The WAIS performance measures also require the handling of materials,
which is not the case with the STAMAT tests. The WAIS measures are all
speed dependent, while the STAMAT test format eliminates some of the
difficulties older adults have transferring answers onto an answer sheet, and
therefore reduces age effects. Effect sizes calculated for each of these
combined measures were not significant
The final hypothesis predicted that STAMAT Recognition Vocabulary
scores, a measure of crystallized intelligence, would be inversely associated
with risk for dementia. This hypothesis was not supported. There was no
correlation between Recognition Vocabulary and incident dementia, and as
shown in Figure 4 and Table 22, the Recognition Vocabulary scores are
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
105
considerably higher than would be expected in a normal distribution. Scores
were essentially the same for males and females in this sample.
The literature on crystallized intelligence measures as a predictor of
dementia is mixed, but the cumulative effect size for all measures of
crystallized intelligence was not significant The meta-analysis supported the
null hypothesis, and the results of the current study are consistent with that
finding: the effect size for Recognition Vocabulary in this study was not
significant
The current sample, as mentioned previously, scored on average 24%
higher on Recognition Vocabulary than the score reported by Schaie (1985)
as the norm for subjects age 7 1 to 77. It is possible that the high scores on
Recognition Vocabulary prevented a true effect on incident dementia from
being demonstrated, or that scores on this specific measure are not
associated with risk for dementia. In either case, there was no effect of risk
reduction for Recognition Vocabulary scores in this study.
There are several possible reasons why the hypotheses were not
supported in this study. To begin with, the absolute number of cases of
dementia identified is small. This may have resulted in having too little
statistical power for the analyses to be significant even if there were true
effects.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
106
A second confound is that the use of death certificate data may have
resulted in a lower observed rate of dementia as a result of missed cases
that were not reported on the death certificate, which could have affected the
outcomes of this study. This is likely in that 13.3% of subjects for whom
medical record data are available were identified as having incident
dementia, compared to 9.86% of subjects who had death certificate data. If
the 13.3% rate held for the entire sample there would have been almost 1 1
additional cases of incident dementia identified. The 9.86% rate is not
inconsistent with rates reported in the literature for other longitudinal
samples, however (Bachman et al., 1993; Evans et al., 1989; Sayetta, 1986).
A third confound is the gender distribution of dementia cases in this
study. The gender distribution of the Long Beach Longitudinal Study and the
sample for whom there is follow-up data were evenly distributed by gender,
but 72.4% of subjects with incident dementia were female. While females
have been shown to have a higher rate of dementia in some studies, the
three to one gender ratio is higher than would be expected.
A fourth confound is that this study used intelligence measures that
had not been used in previous research on the association between
intelligence and dementia. The STAMAT variables are theoretically based,
but may capture different constructs than the WAIS measures for which
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
107
means and standard deviations were reported in the literature. The WAIS
performance measures included as measures of fluid intelligence are
strongly influenced by processing speed, which may not be true of the
STAMAT measures. In addition, the parallel form of the STAMAT Figure
Rotation and Letter Sets measures (i.e., Object Rotation and Word Sets)
may have reduced the influence of processing speed that the original
STAMAT measures captured.
In spite of the potential confounds in this study, the design is sound
and has some advantages over studies reported in the literature. The follow
up period in this study is longer than the majority of studies reported in the
literature. This allowed for following subjects until all but 16% were dead,
reducing the effect of right censoring described previously. In addition, the
longitudinal design allowed for subjects to serve as their own controls,
whereas some of the studies reported in the literature used cross sectional
samples and then matched them with a control group.
In summary, the literature review and meta-analyses led to the
development of hypotheses which predicted that age and fluid intelligence
measures would be predictive of incident dementia, and these hypotheses
were not supported in this study. Age effects were most likely obscured by
the inclusion of length of follow up in the regression analyses, but the finding
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
108
that STAMAT measures of fluid intelligence were not risk factors for
dementia could be valid results which conflict with the literature because the
STAMAT measures do not capture the speed effect or visual/motor
coordination that the WAIS measures do.
Consistent with the literature analyses, education and scores on the
STAMAT crystallized intelligence measure were not predictive of dementia
using either the entire sample or subsamples based on data source. The
use of death certificate data may have resulted in an artificially small number
of dementia cases, and a spurious disproportion of cased by gender are
offered as possible confounds in this study. Sample subjects also scored
higher on Recognition Vocabulary, the single measure of crystallized
intelligence.
There are a number of questions raised by this study which can only
be answered with additional research. Specifically, the risk for dementia
associated with fluid intelligence measures was not observed in this study,
suggesting that perhaps the studies reported in the literature have measured
a construct other than fluid intelligence.
To improve the validity of future studies, longitudinal samples should
be followed until all subjects have died to eliminate the potential bias from
including subjects in the no dementia group who will eventually develop
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
109
dementia. Samples selected for general purposes rather than for studies of
intelligence or memory would be preferential because there would be
reduced risk of bias based on self selection on variables of interest In
addition, medical records data should be favored over death certificate data if
available because they may provide greater sensitivity in the identification of
cases of incident dementia.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
110
REFERENCES
Aaronson, L S., & Burman, M . E. (1994). Focus on psychometrics:
Use of health records in research: Reliability and validity issues. Research in
Nursing and Health. 17. 76-73.
Aronson, M. K., Ooi, W. L, Morgenstren, H., Hafner, A., Masur, D ,
Crystal, H., Frishman, W. H., Fisher, D., & Katzman, R. (1990). Wbmen,
myocardial infarction, and dementia in the very old. Neurology. 40.1102-
1106.
Albert, M. S., Jones, K., Savage, C. R, Berkman, L., Seeman, T.,
Blazer, D., & Rowe, J. W. (1995). Predictors of cognitive change in older
persons: McArthur Studies of successful aging. Psychology and Aging. 10
(4), 578-589.
Amaducci, L. A., Lippi, A., & Fratiglioni, L. (1988). What risk factors
are known? In A. S. Henderson & J. H . Henderson (Eds.), Etiology of
dementia of the Alzheimer's type. New York: John Wiley & Sons, pp29-37.
American Psychological Association. (1994). Diagnostic and Statistical
Manual of Mental Disorders: DSM-IV. Fourth Edition.
American Psychological Association Report (1998). Guidelines for the
evaluation of dementia and age-related cognitive decline. American
Psychologist 53 (12), 1298-1303.
Bachman, D. L., Wolf, P. A., Linn, R. T., Knoefel, J. E., Cobb, J. L,
Belanger, A. J., White, L. R., & D’Agostino, R. B. (1993). Incidence of
dementia and probable Alzheimer's disease in a general population: The
Framingham study. Neurology. 43.515-519.
Baron, A. & Cerella, J. (1993). Laboratory tests of the disuse account
of cognitive decline. In J. Cerella, J. Rybash, W. Hoyer, & M. L. Commons
(Eds.) Adult information processing: Limits on loss. San Diego: Academic
Press, pp. 175-203.
Beard, C. M., Kokmen, E., Sigler, C., Smilt, G. E., Petterson, T., &
O'Brien, P. C. (1996). Cause of death in Alzheimer's disease. American
Epidemiological Report. 6 (3), 195-200.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
1 1 1
Belsky, J. K. (1990). The Psychology of Aoing Theory. Research, and
Interventions. Pacific Grove: Brooks/Cole.
Blessed, G., Tomlinson, B . E., & Roth, M. (1968). The association
between quantitative measures of dementia and of senile change in the
cerebral gray matter of elderly subjects. British Journal of Psychiatry. 114.
797-811.
Bray, D. W., & Howard, A. (1983). The AT&T longitudinal studies of
managers. In K. W. Schaie (Ed.), Longitudinal studies of adult psychological
development New York: Guilford Press.
Brayne, C., & Calloway, P . (1990). The association of education and
socioeconomic status with the Mini Mental State Examination and the clinical
diagnosis of dementia in elderly people. Aae and Ageing. 19.91-96.
Brody, J. A (1995). Postponement as prevention in aging. In R. N.
Butler & J. A Brody (Eds.), Delaying the onset of late-life dysfunction. New
York: Springer Publishing Company.
Brody, N. (1997). Intelligence, schooling, and society. American
Psychologist 5£ (10), 1046-1050.
Bums, A , Jacoby, R., Luthert, P., & Levy, R. (1990). Cause of death
in Alzheimer's disease. Aoe and Aoeino. 19(5). 341-344.
Callahan, C. M., Hall, K. S., Hui, S. L , Musick, B. S., Unverzagt, F.
W., & Hendrie, H. C. (1996). Relationship of age, education, and occupation
with dementia among a community-based sample of African-Americans.
Archives of Neurology. 54.134-140.
Canadian Study of Health and Aging. (1994). Risk factors for
Alzheimer’s disease in Canada. Neurology. 4 4 .2073-2080.
Cattell, R. B. (1987). Intelligence: Its structure, growth and action.
New York: North-Hoiland.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
112
Ceci, S. J. (1991). How much does schooling influence general
intelligence and its cognitive components? A reassessment of the evidence.
Developmental Psychology. 27 (5), 703-722.
Ceci, S. J. & Williams, W. M. (1997). Schooling, intelligence, and
income. The American Psychologist 52 (10), 1051-1058.
Christensen, H., & Mackinnon, A (1992). Wechsler Intelligence Scale
profiles in Alzheimer type dementia and healthy ageing. International Journal
of Geriatric Psychiatry. 7,241-246.
Christensen H., Maltby, N , Jorm, A F., Creasy, H., & Broe, G. A
(1992). Cholinergic ‘blockade’ as a model of the cognitive deficits in
Alzheimer's disease. Brain. 115.1681-1699.
Christensen, K., Korten, A , Jorm, A F., Henderson, A S., Scott, R.,
Jacomb, P . A , & Mackinnon, A J. (1997). Education and decline in
cognitive performance: Compensatory but not protective. International
Journal of Geriatric Psychiatry. 1 g (3), 323-330.
Christensen, K., Korten, A , Jorm, A F., Henderson, A S., Scott, R., &
Mackinnon, A J. (1996). Activity levels and cognitive functioning in an elderly
community sample. Aoe and Aoeino. 25 (1), 72-80.
Coffee, C. E., Lucke, J. F., Saxton, J. A , Ratcliff, G., Unites, L J..
Billig, B., & Bryan, R. N. (1998). Sex differences in brain aging: A
quantitative magnetic resonance imaging stuyd. Archives of Neurology. 55.
169-179.
Craik, F. I. M. (1986). Afunctional account of age differences in
memory. In F. Klix & H. Hagendorf (Eds.), Human memory and cognitive
capabilities. Amsterdam: Elsevier, pp. 409-422.
Craft, S., Asthana, S., Schellenberg, G., Cherrier, M., Baker, L. D.,
Newcomer, J., Plymate, S., Latendresse, S., Petrova, A , Raskind, M.,
Peskind, E., Lofgreen, C., & Grimwood, K. (1999). Insulin metabolism in
Alzheimer’s disease differs according to Apolipoprotein E genotype and
gender. Neuroendocrinoloav. 70 (2). 146-152.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
113
Crum, R. M., Anthony, J. C., Bassett, S. S., & Folstein, M. F. (1993).
Population-based norms for the Mini-Mental State Examination by age and
educational level. Journal of the American Medical Association. 269.2386-
2391.
Crystal, H. A., Dickson, D., Sliwinski, M., Masur, D., Blau, A., & Lipton,
R. B. (1996). Associations of status and change measures of
neuropsychological function with pathologic changes in elderly, originally
nondemented subjects. Archives of Neurology 53.82-87.
Cunningham, W. R., & Haman, K. L (1992). Intellectual functioning
and mental health. In J. E. Birren, R. B. Sloane, & G. D . Cohen (Eds.),
Handbook of mental health and aoino (pp.339-3541. San Diego: Academic
Press, Inc.
Cunningham, W. R., & Owens, W. A. (1983). The Iowa State study of
the development of intellectual abilities. In K. W. Schaie (Ed.), Longitudinal
studies of adult psychological development. New York: Guilford Press.
Devanand, D. P., Folz, M., Goriyn, M., Moeller, J. R., & Stem, Y.
(1997). Questionable dementia: Clinical course and predictors of outcome.
Journal of the American Geriatrics Society. 45. 321-328.
Diehl, M., Willis, S. L., & Schaie, W. K. (1995). Everyday problem
solving in older adults: Observational assessment and cognitive correlates.
Psychology and Aoino. 10 (3), 478-491.
Eddy, D. M. (1996). Clinical decision making: From theory to practice.
Journal of the American Medical Association. 275 (8), 650-658.
Eriksson, P. S., Perfileiva, E., Bjork-Eriksson, T , Albom, A., Nordborg,
C., Peterson, D. A., & Gage, F. H. (1998). Neurogenesis in the adult
hippocampus. Nature Medicine. 4.(11) 1313-1317.
Evans, D. A., Funkenstein, H., Albert, M. S., Schorr, P. A., Cook, N.
R., Chown, M. J., Hebert, L E., Hennekens, C. H., & Taylor, J. 0. (1989).
Prevalence of Alzheimer's disease in a community population of older
persons. Journal of the American Medical Association. 262. 2551-2556.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
114
Eysenck, H. (1995). Can we study intelligence using the experimental
method? Intelligence. 20.217-228.
Finch, C. E., & Tanzi, R. E. (1997). Genetics of aging. Science. 278,
407-411.
Flicker, C., Ferris, S., & Reisberg, R. (1993). A two-year longitudinal
study of cognitive function in normal aging and Alzheimer's disease. Journal
of Geriatric Psychiatry and Neurology. 6, 84-96.
Flynn, J. R. (1999). Searching for justice: The discovery of IQ gains
overtime. American Psychologist. 5-20.
Fratiglioni, L., Jorm, A., Grot, M., Viitanen, M., Holmen, K., Ahlbom,
A., & Winblad, B. (1993). Predicting dementia from the Mini-Mental State
Examination in an elderly population: The role of education. Journal of
Clinical Epidemiology. 46,281-287.
Friedland, R. P. (1993). Epidemiology, education, and the ecology of
Alzheimer’s disease. Neurology. 4 3 .113-120.
Ganguli, M., Dodge, H. H , Chen, P., Belle, S., & DeKosky, S. T.
(2000). Ten-year incidence of dementia in a rural elderly U.S. community
project The MoVIES Project. Neurology. 5 4.1109-1116.
Gardner, E. J. (1986). Schaie-Thurstone Adult Mental Abilities Test
In O. K. Buros (Ed.), The Tenth Mental Measurements Yearbook. Highland
Park, NJ: Griffin Press, 718-720.
Gardner, H. (1983). Frames of mind: The theory of multiple
intelligences. New York: Basic Books.
Gardner, H. (1986). The waning of intelligence tests. In R. J.
Sternberg & D. K. Detterman (Eds.), What is intelligence? New Jersey:
Ablex Publishing Corporation, p. 73-76.
Gatz, M., Lowe, B., Berg, S., Mortimer, J., & Pedersen, N. (1994).
Dementia: Not just a search for the gene. The Gerontologist. 34 (2), 251-255.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
115
Gatz, M., Pedersen, N., Berg, S., Johansson, B., Johansson, K.,
Mortimer, J. A., Posner, S. F., Viitanen, M., Winblad, B., & Ahlbom, A.
(1997). Heritability for Alzheimer's disease: The study of dementia in
Swedish twins. Journal of Gerontoioov: Medical Sciences. 52A. (2), M117-
MI 25.
Graves, A. B , White, E., Koepsell, T. D.t Reifler, B. V., van Belle, G.,
Larson, E . B., & Raskind, M. (1990). A case-control study of Alzheimer’s
disease. Annals of Neuroloov. 28. 766-774.
Grober, E., Lipton, R. B., Hall, C., & Crystal, H. (2000). Memory
impairment on free and cued selective reminding predicts dementia.
Neuroloov. 54, 827-832.
Guriand, B. J., (1981 ).The borderlands of dementia: The influence of
sociocultural characteristics on rates of dementia occurring in the senium. In
N. E. Miller & G. D. Cohen (Eds.), Clinical aspects of Alzheimer's disease
and senile dementia. New York: Raven Press, pp. 61-80.
Hall, K. S., Gao, S., Unverzagt, F. W., & Hendrie, H. C. (2000). Low
education and childhood rural residence: Risk for Alzheimer's disease in
African Americans. Neuroloov. 54.95-99.
Harwood, D. G., Barker, W. W., Loewenstein, D. A., Ownby, R. L., S t
George-Hyslop, P., Mullan, M., & Duara, R . (1999). A cross-ethnic analysis
of risk factors for AD in white Hispanics and white non-Hispanics. Neuroloov.
52, 551-556.
Haxby, J. V., Grady, C., Koss, E., Howlitz, B , Heston, L., Schapiro,
M., Friedland, R . P., & Rapoport, S. I. (1990). Longitudinal study of cerebral
metabolic asymmetries and associated neuropsychological patterns in early
dementia of the Alzheimer type. Archives of Neuroloov. 47 (7), 753-760.
Hayslip, B , & Maloy, R. M. (1992). The interface between
psychometric abilities and everyday cognitive functioning. In R. L . West & J.
D. Sinnot (Eds.), Everyday memory and aoino. (pp. 190-198). New York:
Springer-Verlag.
Hedges, L. V., & Olkin, I. (1985). Statistical methods for meta-
analvsis. Orlando, FI: Academic Press.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
116
Hertzog, C. (1989). Influences of cognitive slowing on age differences
in intelligence. Developmental Psychology. 25 (4), 636-651.
Heyman, A., Wilkinson, W. E., Stafford, J. A., Helms, M. J., Signon, A.
H., & Weinberg, T. (1984). Alzheimer's disease: a study of epidemiological
aspects. Annals of Neuroloov. 15.335-341.
Horn, J. L., & Noll, J. (1997). Human cognitive capabilities: Gf-Gc
Theory. In D. P. Flanagan, J. L Genshaft, & P. L. Harrison (Eds.),
Contemporary intellectual assessment: Theories, tests, and issues. New
York: Guilford Press, pp. 53-91.
Howieson, D. B., Dame, A , Camiciolo, R., Sexton, G., Payami, H., &
Kaye, J. A. (1997). Cognitive markers preceding Alzheimer's dementia in the
healthy oldest old. Journal of the American Geriatrics Society. 45.584-589.
Huppert, F. A., & Beardsall, L. (1993). Prospective memory
impairment as an early indicator of dementia. Journal of Clinical and
Experimental NeuroosvcholOQV. 15 (5), 805-821.
Jacobs, D. M., Sano, M., Dooneief, G., Marder, K., Bell, K . L., & Stem,
Y. (1995). Neuropsychological detection and characterization of predinical
Alzheimer's disease. Neuroloov. 45.957-962.
Jorm, A. (1990). The eoidemioloov of Alzheimer's disease and related
disorders. New York: Chapman and Hall.
Jorm, A., Korten, A , & Henderson, A (1987) The prevalence of
dementia: A quantitative integration of the literature. Acta Psvchiatrica
Scandinavica. 76. 465-479.
Jorm, A F., Rodgers, B., Henderson, A S., Korten, A E., Jacomb, P .
A , Christensen, H., & Mackinnon, A (1998). Occupational type as a
predictor of cognitive decline and dementia in old age. Aoe and Aoeino. 27,
477-483.
Kami, A , Meyer, G., Jezzard, P., Adams, M. M., Turner, R., &
Ungerieider, L. G. (1995). Functional MRI evidence for adult motor cortex
plasticity during motor skill learning. Nature. 377.155-158.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
117
Katzman, R. (1993). Education and the prevalence of dementia and
Alzheimer's disease. Neuroloov. 43.13-20.
Katzman, R., Aronson, M., Fuld, P., Kawas, C., Brown, T.,
Morgenstem, H., Frishman, W., Gidez, L , Eder, H., & Ooi, W. L. (1989).
Development of dementing illness in an 80-year-old volunteer cohort Annals
of Neuroloov. 25 (4), 317-324.
Kaye, J. A., Swihart, T., Howieson, D., Dame, A., Moore, M. M.,
Kamos, T., Camicioli, R., Ball, M., Oken, B., & Sexton, G. (1997). Volume
loss of the hippocampus and temporal lobe in healthy elderly persons
destined to develop dementia. Neuroloov. 4 8 .1297-1304.
Khoke, D. & Bohmstedt, G . W. (1994). Statistics for social data
analysis. (3rd ed.). Itasca, Illinois: F. E. Peacock Publishers, Inc.
Koss, E., Edland, S., Fillenbaum, G., Mohs, R., Clark, C., Galasko, D ,
& Morris, J. C. (1996). Clinical and neuropsychological differences between
patients with earlier and later onset of Alzheimer's disease: A CERAD
analysis, part XII. Neuroloov. 46.136-141.
La Rue, A., & Jarvik, L. F. (1987). Cognitive function and prediction of
dementia in old age. International Journal of Aaino and Human
Development £5 (2), 79-89.
Lanska, D. J. (1998). Dementia mortality in the United States: Results
of the 1986 National Mortality Followback Survey. Neurology. 50.362-367.
Launer, L. J., Andersen, K., Dewey, M. E., Letenneur, L, Ott, A.,
Amaducci, L A., Brayne, C., Copeland, J. R. M., Dartiques, J. F , Kragh-
Sorensen, P., Lobo, A , Martinez-Lage, J. M., Stijnen, T., Hofman, A , & the
EURODEM Incidence Research Group and Work Groups. (1999). Rates and
risk factors for dementia and Alzheimer's disease: Results from EURODEM
pooled analyses. Neurology. 5g. 78-84.
Lave, J., Murtaugh, M., & de la Rocha, O. (1984). The dialectic of
arithmetic in grocery shopping. In B. Rogaff & J. Lave (Eds.), Everyday
cognition: Its development in social context Cambridge, MA Harvard
University Press, Pp. 67-94.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
118
Leissring, M. A., Sugarman, M. C., & LaFerla, F. M . (1998). Herpes
simplex virus infections and Alzheimer’s disease: Implications for drug
treatment and immunotherapy. Drugs and Aoino. 13.193-198.
Lezak, M. D. (1995). Neuropsychological Assessment New York:
Oxford University Press.
Lindenberger, U., & Baltes, P . B. (1994). Sensory functioning and
intelligence in old age: A strong connection. Psychology and Aoino. 9 (3),
339-355.
Lindenberger, U., Mayr, U., & Kliegl, R. (1993). Speed and intelligence
in old age. Psychology and Aoino. 8 (2), 207-220.
Linet, M. S., Harlow, S. D., McLaughlin, J. K., & McCaffrey, L. 0.
(1989). A comparison of interview data and medical records for previous
medical conditions and surgery. Journal of Clinical Epidemioloov. 42 (12),
1207-1213.
Linn, R. T., Wolf, P. A , Bachman, D. L., Knoefel, J. E., Cobb, J. L,
Belanger, A J., Kaplan, E. F , & D ’Agostino, R. B. (1995). The “preclinical
phase" of probable Alzheimer’s disease. Archives of Neuroloov. 52.485-490.
Locascio, J. J., Growden, J. H., & Corkin, S. (1995) Cognitive test
performance in detecting, staging, and tracking Alzheimer’s disease.
Archives of Neuroloov. 52. 1087-1099.
Macera, C. A , Sun, R., Yeager, K. K., & Brandes, D. (1992).
Sensitivity and specificity of death certificate diagnoses for dementing
illnesses, 1988-1990. Journal of the American Geriatrics Society. 40. 479-
481.
Magrane, B. P., Gilliand, M . G. P., & King, D. E. (1997). Certification
of death by family physicians. American Family Physician. 56 (5), 1433-
1438.
Mangione, C. M., Seddon, J. M., Cook, E. F., Krug, J. H., Sahagian,
C. R., Campion, E. W., & Glynn, R . J. (1993). Correlates of cognitive function
scores in elderly outpatients. Journal of the American Geriatrics Society. 41.
491-497.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
119
Manly, J. J., Merchant, C. A., Jacobs, 0. M., Small, S. A., Bell, K.,
Ferin, M., & Mayeux, R. (2000). Endogenous estrogen levels and
Alzheimer's disease among postmenopausal women. Neuroloov. 54, 833-
837.
Manton, K. G., Siegler, I. C., & Woodbury, M. A. (1986). Patterns of
intellectual development in later life. Journal of Gerontology. 41 (4), 486-499.
Martyn, C. N., & Pippard, E. C. (1988). Usefulness of mortality data
in determining the geography and time trends of dementia. Journal of
Epidemiology and Community Health. 42, 134-137.
Masur, D. M., Sliwinski, M., Lipton, R. B., Blau, A. D., & Crystal, H. A.
(1994). Neuropsychological prediction of dementia and the absence of
dementia in healthy elderly persons. Neuroloov. 4 4 .1427-1432.
Matarazzo, J. D., & Herman, D. 0. (1984). Relationship of education
and IQ in the WAIS-R standardized sample. Journal of Consulting and
Clinical Psychology. 52. 631-634.
McClearn, G. E., Johansson, B., Berg, S., Pedersen, N. L , Ahern, F.,
Petrill, S. A., & Ptomin, R. (1997). Substantial genetic influence on cognitive
abilities in twins 80 or more years old. Science. 276 (5318V 1560-1563.
McGrew, K. (1997). Analysis of the major intelligence batteries
according to a proposed comprehensive Gf-Gc framework. InD. P.
Flanagan, J. L Genshaft, & P. L. Harrison (Eds.), Contemporary intellectual
assessment Theories, tests, and issues. New York: Guilford Press, pp. 151-
182.
McGue, M., Bouchard, T. J., lacano, W. G., & Lykken, D. T. (1993).
Behavioral genetics of cognitive ability: A life-span perspective. In R. Plomin
& G. E . McCleam (Eds.), Nature, nurture. & psychology (pp. 59-76).
Washington, DC: American Psychological Association.
McKhann, G., Drachman, D., Folstein, M., Katzman, R., Price, D., &
Stadlan, E. M. (1984). Clinical diagnosis of Alzheimer's disease: Report of
the NINCDS-ADRDA Work Group under the auspices of Department of
Health and Human Services Task Force on Alzheimer's Disease. Neurology.
34, 939-944.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
120
Messite, J., & Stelland, S . D . (1996). Accuracy of death certificate
completion: The need for formalized physician training. Journal of the
American Medical Association. 275 (10) 794-796.
Meyer, J. S., Terayama, Y , Konno, S., Akiyama, H., Margishvili, G.
M.t & Mortel, K. F. (1998). Risk factors for cerebral degenerative changes
and dementia. European Neuroloov. 39 (S1), 7-16.
Moritz, D. J., & Pettiti, 0. B . (1993). Association of education with
reported age of onset and severity of Alzhiemer's disease at presentation:
Implications for the use of clinical samples. American Journal of
Epidemiology. 137. 456-462.
Morris, J. C., & Rubin, E . H . (1991). Clinical diagnosis and course of
Alzheimer's disease. The psychiatric clinics of North America. 14 (2), 223-
236.
Mortimer, J. A. (1988). Do psychosocial risk factors contribute to
Alzheimer's disease? In A. S . Henderson & J. H. Henderson (Eds.), Etiology
of dementia of Alzheimer's type. New York: John Wiley & Sons, pp.39-52.
Mortimer, J. A. (1990). Epidemiology of dementia: Cross-cultural
comparisons. In R. J. Wurtman, S. Corkin, & J. H. Grwodon (Eds.),
Alzheimer’s disease. Advances in neuroloov. 51. New York: Raven Press,
pp. 27-33.
Mortimer, J. A. (1997). Brain reserve and the clinical expression of
Alzheimer's disease. Geriatrics. 52 (suppl.2), S50-S53.
Mortimer, J. A., & Graves, A. B. (1993). Education and other
socioeconomic determinants of dementia and Alzheimer's disease.
Neuroloov. 43. S39-S44.
Mueller, E. A., Moore, M . M., Kerr, D. R . C., Sexton, G., Camicioli, R.
M., Howieson, D. B , Quinn, J. F., & Kaye, J. A. (1998). Brain volume
preserved in healthy elderly through the eleventh decade. Neuroloov. 51.
1555-1562.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
12 1
Munoz, D. G., Ganapathy, G. R-, Eliasziw, M., & Hachnski, V. (2000).
Educational attainment and socioeconomic status of patients with autopsy-
confirmed Alzheimer disease. Archives of Neuroloov. 57.85-89.
Neisser, U., Boodoo, G., Bouchard, T. J., Boykin, A. W., Brody, N.,
Cece, S. J., Halpem, 0. F., Loehlin, J. C., Pertoff, R.t Sternberg, R. J., &
Urbina, S. (1996). Intelligence: Known and unknowns. American
Psychologist 51 (2), 77-101.
Nettelbeck, T., & Rabbitt, P. M. (1992). Aging, cognitive performance,
and mental speed. Intelligence. 16,189-205.
Norusis, M. J. (1993). SPSS for Windows Advanced Statistics
Release 6.0. Chicago: SPSS, Inc.
Office of Technology Assessment (1987). Losing a million minds:
Confronting the tragedy of Alzheimer’s disease and other dementias.
Washington, DC: U.S. Government Printing Office. (GPO Publication
Number 052-003-01059-3).
Olichney, J. M., Hofsetter, C. R., Galasko, D, Thai, L J., & Katzman,
R. (1995). Death certificate reporting of dementia and mortality in a
Alzheimer's Disease Research Center cohort Journal of the American
Geriatrics Society. 43. 890-893.
Osterweil, D., Mulford, P., Syndulko, K., & Martin, M. (1994). Cognitive
function in old and very old residents of a residential facility: Relationship to
age, education, and dementia. Journal of the American Geriatrics Society. 42
(7), 766-773.
Pedersen, N. L., Reynolds, C. A., & Gatz, M. (1996). Sources of
covariation among Mini-Mental State Examination scores, education, and
cognitive abilities. Journal of Gerontology: Psychological Sciences. 51 (2).
P55-P63.
Perimutter, M., & Nyquist, L. (1990). Relationships between self-
reported physical and mental health and intelligence performance across
adulthood. Journal of Gerontology. 45 (4), P145-155.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
122
Popkin, S. J., Schaie, K. W., & Krauss, I. K. (1963). Age-fair
assessment of psychometric intelligence. Educational Gerontology. 9,47-55.
Prince, M., Cullen, M., & Mann, A. (1994). Risk factors for Alzheimer's
disease and dementia: A case-control study based on the MRC elderly
hypertension trial. Neuroloov. 44.97-104.
Reed, T. E., & Jensen, A. R. (1993). Choice reaction time and visual
pathway conduction velocity both correlate with intelligence but appear not to
correlate with each other Implications for information processing.
Intelligence. 17. 191-203.
Rehberg, R. A., & Rosenthal, E. R. (1978). Class and merit in the
American high school. New York: Longman.
Robinson, B. (1998). Diagnosis of irreversible dementia: How
extensive the evaluation? Geriatrics. 53.49-50.
Rocca, W. A., Amaducci, LA..& Schoenberg, B . S. (1986).
Epidemiology of clinically diagnosed Alzheimer's disease. Annals of
Neuroloov. 19- 415-424.
Rocca, W. A., Cha, R . H., Waring, S. C , & Kbkmen, E. (1998).
Incidence of dementia and Alzheimer's disease: A reanalysis of data from
Rochester, Minnesota, 1975-1984. American Journal of Epidemiology. 148.
51-62.
Roth, M. (1981). The diagnosis of dementia in late and middle life. In
J. Mortimer & L Schuman (Eds.), The epidemiology of dementia. New York:
Oxford University Press.
Rubin, E. H , Kinscherf, D. A., & Morris, J. C. (1993). Psychopathology
in younger versus older persons with very mild dementia of the Alzheimer
type. American Journal of Psychiatry. 150 (41.639-642.
Rubin, E. H., Storandt, M., Miller, P., Kinscherf, D. A., Grant, E. A.,
Morris, J. C., & Berg, L. (1998). A prospective study of cognitive function and
onset of dementia in cognitive health elders. Archives of Neurology. 55.395-
401.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
123
Saithouse, T. A. (1985). Speed of behavior and its implication for
cognition. In J. E. Birren & K. W. Schaie (Eds.), Handbook of the psychology
of aoino (2nd ed.). New York: Van Nostrand Reinhold. pp. 400-426.
Saithouse, T. A. (1990). Cognitive competence and expertise in
aging. In J. E. Birren & K. W. Schaie (Eds.), Handbook of the psychology o f
aoino (3rd ed., pp. 310-319). San Diego: Academic Press, Inc.
Saithouse, T. A. (1991). Theoretical perspectives on cognitive aoino.
Hillsdale, NJ: Lawrence Erlbaum Associates Publishers.
Saithouse, T. A. & Czaja, S. J. (2000). Structural constraints on
process explanations in cognitive aging. Psychology and Aging. 15 (1) 44-
55.
Saithouse, T. A., Hancock, H . E., Meinz, E. J., & Hambrick, D. Z.
(1996). Interrelations of age, visual acuity, and cognitive functioning.
Journal of Gerontology. Psychological Sciences. 51 (6), 317-330.
Sayetta, R. B. (1966). Rates of senile dementia - Alzheimer’s type in
the Baltimore Longitudinal study. Journal of Chronic Disease. 4,271-286.
Schaie, K. W. (1979). The primary mental abilities in adulthood: An
exploration in the development of psychometric intelligence. In P. Baltes &
O. Brim (Eds.), Life-span development and behavior (Vol. 2). San Francisco:
Academic Press, pp. 67-115.
Schaie, K. W. (1985). The Seattle Longitudinal Study: A twenty-one
year investigation of psychometric intelligence. In K. W. Schaie (Ed.),
Longitudinal studies of adult psychological development New York: Guilford
Press.
Schaie, K. W. (1989). The hazards of cognitive aging. The
Gerontologist 29.484-493.
Schaie, K. W. (1996). Intellectual development in adulthood. New
York: Cambridge University Press.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
124
Schaie, K. W., & Hertzog, C. (1986). Toward a comprehensive model
of adult intellectual development Contributions of the Seattle longitudinal
study. In R. Sternberg (Ed.), Advances in human intelligence. Hillsdale. NJ:
Lawrence Erlbaum Associates, pp70-118 .
Schmand, B., Smit, J. H., Geerlings, M. I., & Lindenboom, J. (1997).
The effects of intelligence and education on the development of dementia. A
test of the brain reserve hypothesis. Psychological Medians. 2 7 .1337-1344.
Schooler, C. (1984). Psychological effects of complex environments
during the life span: A review and theory. Intelligence. 8,259-281.
Shichita, K., Hatano, S., Ohashi, H., & Matuzaki, T. (1986). Memory
changes in the Benton Visual Retention Test between ages 10 and 75.
Journal of Gerontology. 41. 385-386.
Silverman, D. H., Chang, C. Y., Cummings, J. L., Kowell, A P.,
Czemin, J., Small, G. W., Gambhir, S . S., Hoh, C. K., & Phelps, M. E. (1999).
Prognostic value of regional brain metabolism in evaluation of dementia
Comparison with long-term clinical outcome. Journal of Nudear Medicine.
4 Q (Supplement), 7 1 P.
Sliwinski, M., Lipton, R. B , Buschke, H., & Stewart, W. (1996). The
effects of predinical dementia on estimates of normal cognitive functioning in
aging. Journal of Gerontology. 5 1 (4 1 . P217.
Small, G. W., Mazziotta, J. C , Collins, M. C., Baxter, L. R., Phelps, M.
E., Mandelkem, M. A , Kaplan, A , La Rue, A , Adamson, C. F., Chang, L ,
Guze, B. H., Confer, E. H., Saunders, A M., Haines, J. L , Pericak, M. A , &
Roses, A D. (1995). Apolipoprotein E Type 4 Allele and cerebral glucose
metabolism in relatives at risk for familial Alzheimer disease. Journal of the
American Medical Association. 273.942-947.
Snowden, D. A , Kemper, S. J., Mortimer, J. A , Greiner, L. H.,
Wekstein, D. R., & Markesbery, W. R . (1996). Linguistic ability in early life
and Alzheimer's disease in late life. American Medical Association Journal.
275. 528-532.
Sramek, J. J., & Cutler, N. R. (1999) Recent developments in the drug
treatment of Alzheimer's disease. Drugs and Aoino. 14.359-373.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
125
Stanwick, G. R , Coehn, R. F., Maguire, C. P., Kirby, M., Walsh, J. B.,
O’Neill, 0., Coakley, D., & Lawlor, B. A. (1999). The association betweem
demographic factors, disease severity and the duration of symptoms at
clinical presentation in elderly people with dementia. Aoe and Ageing. 28.(3),
295-307.
Stem, Y., Alexander, G. E., Prohovnik, I., & Mayeux, R. (1992).
Inverse relationship between education and parietotemporal perfusion deficit
in Alzheimer's disease. Annals of Neuroloov. 32 (3), 371-375.
Stem, Y., Alexander, G. E., Prohovnik, I., Stricks, L., Link, B., Lennon,
M. C., & Mayeux, R. (1995). Relationship between lifetime occupation and
parietal flow; Implications for a reserve against Alzheimer's disease
pathology. Neuroloov. 45.55-60.
Stem, Y., Gurland, B., Tatemichi, T. X , Tang, M . X , Wilder, D., &
Mayeux, R. (1994). Influence of education and occupation on the incidence
of Alzheimer's disease. American Medical Association Journal. 271 (13),
1004-1010.
Storandt, M., Stone, X, & LaBarge, E. (1995). Deficits in reading
performance in very mild dementia of the Alzheimer Type. Neuropsychology.
9(2), 174-176.
Sulkava, R., Wikstrom, J., Aromaa, A., Raitasalo, R., Lehtinen, V.,
Lahtela, X , & Palo, J. (1985) Prevalence of severe dementia in Finland.
Neuroloov. 35,1025-1029.
Swaab, D. F. (1991). Brain aging and Alzheimer’s disease, “wear and
tear* versus “use it or lose if. Neurobioloov of AQino. 12.317-324.
Swan, G. E., LaRue, A , Carmelli, D., Reed, T. E., & Fabsitz, R . R .
(1992). Decline in cognitive performance in aging twins: Heritability and
biobehavioral predictors from the National Heart, Lung, and Blood Institute
Twin Study. Archives of Neuroloov. 49.476-481.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
126
Tang, M., Stem, Y., Marder, K., Bell, K., Gurfand, B., Lantigua, R.,
Andrews, H., Feng, L, Tycko, B., & Mayeux, R. (1998). The APOE- e4 allele
and the risk of Alzheimer's disease among African Americans, whites, and
hispanics. Journal of the American Medical Association. 279 (10V 751-755
Thurstone, L L., & Thurstone, T. G. (1949). Examiner Manual for the
SRA Primary Mental Abilities Test (Form 10-14). Chicago: Science Research
Associates.
Tierney, M. C., Szalai, J. P., Snow, W. G., Fisher, R. H., Tsuda, T.,
Chi, H., McLachlan, D. R., & S t George-Hyslop, P. H. (1996). A prospective
study of the clinical utility of APOE genotype in patients with memory
impairment. Neuroloov. 46.149-154.
U. S. Department of Commerce, Bureau of the Census (1979).
Statistical Abstract of the United States. 100* Edition.
Unverzagt, F. W., Hui, S. L , Farlow, M . R., Hall, K. S., & Hendrie, H .
C. (1998). Cognitive decline and education in mild dementia. Neuroloov. 50.
181-185.
Von Dras, D . D., & Blumenthal, H. T. (1992). Dementia of the aged:
Disease or atypical - accelerated aging? Biopathological and psychological
perspectives. Journal of the American Geriatrics Society. 40. 285-294.
Walters, J. M., & Gardner, H. (1986). The theory of multiple
intelligences: Some issues and answers. In R. J. Sternberg & R. K. Wagner
(Eds.), Practical intellioence-Nature and origins of competence in the
everyday world fpp. 236-268). New York: Cambridge University Press.
Wechsler, D. (1958). Measurement of adult intelligence. 4th Edition.
Baltimore: Williams and Wilkins.
Wechsler, D. (1981). The psychometric tradition: Developing the
Wechsler Adult Intelligence Scale. Contemporary Educational Psvcholoov.6
82-85.
Wechsler, D. (1981). Wechsler Adult Intelligence Scale - Revised.
New York: The Psychological Corporation.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
127
White, L , Katzman, R., Losonczy, K, SaKve, M., Wallace, R.,
Berkman, L , Taylor, J., Fillenbaum, G., & Havlik, R. (1994). Association of
education with incidence of cognitive impairment in three established
populations for epidemiologic studies of the elderly. Journal of Clinical
Epidemioloq v . 47 (4), 363-374.
Willis, S. L , & Schaie, W. K. (1986). Practical intelligence in later
adulthood. In R. J. Sternberg & R. K. Wagner (Eds.), Practical intellioence-
Nature and origins of competence in the everyday world (dp. 236-268). New
York: Cambridge University Press.
Wilson, R. S., Beckett, L. A., Bennett, L. A., Albert, M. S., & Evans, D .
A (1999). Change in cognitive function in older persons from a community
population. Archives of Neurology. 56.1274-1279.
Zelinski, E. M , Gilewski, M. J., & Schaie, K. W. (1993). Individual
differences in cross-sectional and 3-year longitudinal memory performance
across the adult life span. Psychology and Aoino. 8 (2), 176-187.
Zelinski, E. M , & Stewart, S. T. (1996). Individual differences in 16-
year memory changes. Psychology and Aoino. 13 (4). 622-630.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Frequency and risk factors of poststroke dementia

PDF

Assessment of prognostic comorbidity in hospital outcomes research: Is there a role for outpatient pharmacy data?

PDF

Heart disease among middle -aged and elderly persons in the United States: Trends and a multistate model

PDF

Hospital conversions: The California experience

PDF

Gender and the bump: An investigation of the reminiscence effect in the Long Beach Longitudinal Study

PDF

Course completion and older adult students: The effects of self-efficacy, self -regulation, and motivation

PDF

Cross -cultural examination of mental health measures in dementia caregivers: Assessment of the Center for Epidemiologic Studies -Depression Scale and the Zarit Burden Inventory

PDF

Control and dynamics during horizontal impulse generation

PDF

Factors influencing the interpretation of novel words as adjectives in 4-year-old Spanish -speaking children

PDF

A methodology to identify high -risk patients with diabetes in the California Medicaid populations (Medi -Cal)

PDF

Biopharmaceutical strategic alliances: Interorganizational dynamics and factors influencing FDA regulatory outcomes

PDF

Globalization and the decline of the welfare state in less developed countries

PDF

Evaluation of the Care Advocate Program: Bridging managed care and home community -based services

PDF

Dementia caregiving and ethnicity: African American caregivers and the sociocultural stress and coping model

PDF

Effects of cultural values and behaviors in sociocultural stress and coping among Korean-American and Korean caregivers

PDF

Experimental evaluation of a new intervention designed to promote dementia caregivers' acceptance and empathic responding towards care -recipients

PDF

Attachment, accommodation, and the outcomes of romantic relationships

PDF

Behavioral and neural investigations of perceptual affect

PDF

Business political influence on agency regulatory decision making: The efficacy of corporate political activity in Taiwan's CATV license awarding

PDF

A foundation for general-purpose natural language generation: Sentence realization using probabilistic models of language

Asset Metadata

Creator Werle, Kathleen H. (author)

Core Title Education and intelligence test scores: Predictors of dementia?

School Graduate School

Degree Doctor of Philosophy

Degree Program Gerontology and Public Policy

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag Gerontology,OAI-PMH Harvest,psychology, cognitive

Language English

Contributor Digitized by ProQuest (provenance)

Advisor Zelinski, Elizabeth M. (committee chair), Graddy, Elizabeth (committee member), Knight, Robert (committee member)

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c16-184211

Unique identifier UC11334923

Identifier 3065551.pdf (filename),usctheses-c16-184211 (legacy record id)

Legacy Identifier 3065551.pdf

Dmrecord 184211

Document Type Dissertation

Rights Werle, Kathleen H.

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the au...

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus, Los Angeles, California 90089, USA