Detection of Abnormal Memory Decline

in Mild Cases of Alzheimer's Disease

Using CERAD Neuropsychological Measures
Kathleen Welsh, PhD; Nelson Butters, PhD; James Hughes, MS; Richard Mohs, PhD; Albert Heyman, MD

\s=b\ The present study was designed to T^he Consortium to Establish a Reg- well known that learning or
determine which of the memory tasks istry for Alzheimer's Disease "acquisition" of new information is
included in the CERAD (Consortium to Es- (CERAD) was organized to develop severely impaired in early AD, recent
tablish a Registry for Alzheimer's Disease) brief standardized assessments of the work using standardized tests of
neuropsychological battery best differen- clinical and neuropsychological mani¬ memory35 and experimental para¬
tiate patients with early Alzheimer's dis- festations of Alzheimer's disease digms6 has emphasized the importance
ease from cognitively normal elderly con- (AD).1 The application of such stan¬ of the rate of forgetting as measured
trol subjects and also best distinguish be- dardized tests in multicenter studies of by delayed recall and savings scores.
tween the various levels of severity of the AD is expected to reduce the variety of Other studies have shown that intru¬
dementia process. A sample of CERAD pa- psychometric measures currently used sion errors7 and recognition memory
tients with Alzheimer's disease was strat- in a number of research laboratories, a performance8 may also be useful in
ified by disease severity into those with problem that often complicates com¬ making such differentiations.
mild, moderate, or severe dementia and parison of data. We examined various memory in¬
matched with control subjects for sex, The measures included in the dexes in AD subjects with mild, mod¬
age, and education. Using multivariate CERAD neuropsychological battery erate, and severe cognitive impair¬
procedures and cutting scores, the effi- characterize the primary cognitive ment and, also, in normal elderly con¬
cacy of each memory measure in distin- manifestations of AD, ie, memory dis¬ trol subjects whose data had been
guishing between these groups and con- turbance, language impairment, entered into the CERAD national
trol subjects was determined. The test for apraxia, and general intellectual dete¬ study.
delayed recall was found to be the best rioration. The CERAD battery as a
SUBJECTS AND METHODS
overall discriminatory measure. The other whole has been found to have substan¬
tests of memory, ie, immediate recall, tial test-retest reliability, cross-center Subjects
intrusion errors, and recognition memory, interrater reliability, and longitudinal To ensure that the groups would be
had poor overall discriminability. None of validity.2 It also has demonstrated dis¬ matched on the basis of age and education,
the CERAD memory measures were found criminative validity, ie, each of the in¬ subjects were selected for this study from
to be particularly powerful in staging the dividual psychometric tests differen¬ the total number of 549 individuals with AD
and 390 control subjects who had been en¬
severity of dementia. These findings sug- tiates normal elderly control subjects tered into the study from its onset in April
gest that tests for delayed recall may be from patients with mild to moderate 1987 to a cutoff period in July 1989. The data
particularly useful in the early detection of dementia. However, there is still need presented in this report are based on 147
Alzheimer's disease and should be consid- to establish the relative sensitivity and patients and 49 control subjects derived
ered in screening batteries for dementia in specificity of the individual test mea¬ from the pool who were recruited from the
community surveys. sures in distinguishing between pa¬ referral populations of the 19 participating
(Arch Neurol. 1991;48:278-281) tients with very mild dementia and institutions. The control subjects were ei¬
normal elderly individuals. Such infor¬ ther informants for the patients or com¬
mation would be important, for exam¬ munity volunteers. On entry into the study,
Accepted for publication August 27, 1990. informed consent was obtained from each
From the Department of Psychiatry and Bryan ple, in determining whether the in¬ subject and/or from a responsible person
Alzheimer's Disease Research Center (Dr Welsh) strument is effective in detecting early by procedures approved by the individual
and Department of Medicine, Division of Neurol- dementia in population surveys. center's Institutional Review Board.
ogy (Dr Heyman), Duke University Medical Cen- The present study is an initial anal¬
ter, Durham, NC; Psychology Service, Depart-
The patients included in the study met
ment of Veterans Affairs, Veterans Administra- ysis of the clinical sensitivity and spec¬ the diagnostic criteria for probable AD us¬
tion Medical Center, and Psychiatry Department, ificity of the verbal memory test, which ing the criteria adopted (with minor modi¬
University of California, San Diego (Dr Butters); is part of the tests included in the fications) from the National Institute of
Department of Biostatistics, University of Wash- CERAD battery. The specific issue ad¬ Neurological and Communicative Disor¬
ington, Seattle (Mr Hughes); and Department of ders and Stroke-Alzheimer's Disease and
Psychiatry, Mount Sinai School of Medicine, New dressed in this report is which of the
Related Disorders Association Work
York, NY (Dr Mohs). learning and memory measures in¬ Group.9 They were 50 years of age or older,
Presented, in part, at the Society for Neuro- cluded in the CERAD neuropsycholog¬ could speak and comprehend English, and
sciences meeting, Tucson, Ariz, November 1,1990. ical battery best differentiate patients
Reprint requests to Bryan Alzheimer's Disease had a caregiver/informant who could pro¬
Research Center, Box 3450, Duke University Med- with very early AD from normal eld¬ vide adequate history. The patients with
ical Center, Durham, NC 27710 (Dr Welsh). erly control subjects. Although it is severe cognitive impairment, as deter-

Downloaded From: http://archneur.jamanetwork.com/ by a Michigan State University User on 06/17/2015

 mined by scores greater than 24 on the Ori¬
Table 1.—Demographic Information and MMSE Scores for Control Subjects and
entation-Concentration-Memory test10 or
less than 10 on the Mini-Mental State Ex¬ Groups With AD *

amination (MMSE),11 were excluded. Con¬

Groups With AD
trol subjects met the same inclusion crite¬ Control
ria as the patients but were free of cognitive Subjects Mild Moderate Severe
impairment at entry. Variables (n =
49) (n =
49) (n =
49) (n -
49)
The pool of 549 subjects was stratified MMSE score
into three groups based on the severity of Mean 28.9 25.0 21.0
overall cognitive impairment, as defined by SD 1.3 2.9
the MMSE scores. The mild group com¬
Sex, No.
prised those with MMSE scores of 24 or Male 25 25 25 25
greater; the moderate group had scores be¬ Female
tween 19 and 24, whereas the severe group
Age, y
scored between 10 and 19. This procedure Mean 71.1 71.1
resulted in 84 cases of mild AD, 206 cases of SD 5.6 5.6
moderate impairment, and 255 cases of se¬
Education, y
vere impairment. Mean
A comparison of demographic variables
SD 2.9 2.5 3.0
revealed that the mean age of patients was
MMSE indicates Mini-Mental State Examination; and AD, Alzheimer's disease.
significantly higher than that of control
subjects (71.1 years vs 68.2 years; SD 6.8=

and 7.6, respectively; < .0001) and that

the average education of the patients was
significantly lower than that of the control Table 2. Mean Performance Levels of the Control Subjects and the Groups With AD*
subjects (13.1 years vs 14.0 years; SD 3.0
—

for each group; < .0001). Since age and Groups With AD
Memory Control
education frequently modify outcome on Measures Subjects Mild Moderate Severe
cognitive testing, it was anticipated that Learning
the age and education differences in our Trial 1 4.8 (1.4) 2.8(1.4) 2.2 (1.6) 1.0(1.1)
groups might bias the findings and dispro¬ Trial 2 7.0(1.5) 4.2(1.5) 3.4 (1.6) 2.2 (1.3)
portionately favor the performance of the Trial 3 7.9(1.6) 4.7 (1.8) 3.9(1.7) 2.3(1.6)
control subjects.
Delay recall 6.8(1.9) 1.8 (1.8) 0.7 (1.5) 0.5 (1.0)
To control for these factors, the smallest
Savings, 85.6 (19.3) 35.8 (30.9) 16.7 (32.8) 16.3 (28.2)
group (mild AD cases) was computer
matched to individuals in each of the other Recognition
groups on the basis of age (within 2 years), Rec-Yes 9.7 (0.6) 8.2 (2.4) 7.7 (2.3) 6.9 (2.6)
education, and gender. Matching on these Rec-No 9.7 (1.5) 7.7 (2.7) 7.5 (2.6) 6.4 (3.0)
three variables resulted in four groups of 49 Intrusions
individuals (25 men, 24 women) each, as Trial 1 0.3 (0.5) 0.6 (1.0) 0.6 (1.5) 0.4 (0.8)
seen in Table 1. Trial 2 0.2 (0.4) 0.4 (0.8) 0.4 (1.2) 0.5 (1.9)
0.1 (0.2) 0.3 (0.7) 0.6(1.1) 0.7 (1.0)
Neuropsychological Assessment Delay 0.4 (0.7) 0.9 (0.9) 0.9 (1.4) 0.8 (1.6)

The CERAD neuropsychological battery2 'Values in parentheses are standard deviations. Higher scores on the intrusion measures and low scores
on all other memory measures represent poorer performance levels. AD indicates Alzheimer's disease; Rec-Yes,
includes a measure of category fluency, a
true-positive recognition responses; and Rec-No, true-negative recognition responses.
modified Boston Naming Procedure; the
MMSE; a word-list learning, memory, and
recognition procedure; and an assessment on the study list, is series of memory measures. Assessment
of constructional praxis. As mentioned ear¬ sponses of words not a
also recorded. of acquisition was measured by the number
lier, the tests were chosen to measure the The delayed recall procedure tests mem¬ of words learned on each of the three
principal cognitive manifestations of AD
and to have a range of difficulty sufficient to ory for the 10-item word list after a delay of learning trials (hereafter abbreviated Tl,
characterize patients with the different 5 to 8 minutes. To prevent rehearsal of the T2, or T3) and by the number of intrusion
list, an intervening (constructional praxis) errors committed on each of the learning
stages of disease.
This study examined the performance on task is administered between the last learn¬ trials.
three learning and memory procedures: ing trial and the spontaneous uncued recall The memory measures included the num¬
word-list learning, word-list delayed recall, of the word list. The number correctly ber of words recalled correctly on the delay
and word-list delayed recognition. The recalled on the delayed recall trial (maxi¬ trial (Td) and three measures of recognition
mum score of 10) and the number of intru¬ memory: total number correct; correct rec¬
word-list learning task consists of an im¬
mediate free recall procedure for a 10-item sion errors are recorded. The delayed word- ognition of target items (ie, correct "yes"
word list assessed over three separate list recognition procedure tests the sub¬ responses, hereafter abbreviated Rec-Yes),
learning trials. On the first trial, 10 printed ject's recognition of the 10-item word list and correct validation of distractors (ie,
words are presented at a rate of one every (targets) from among 10 distractor items. correct "no" responses, hereafter abbrevi¬
2 seconds; the subject is instructed to read Words are presented one by one, and the ated Rec-No). Savings scores defined as per¬
each word aloud to ensure adequate atten¬ subject must indicate whether each item cent of words correctly recalled on Td when
tion to the stimuli. Immediately following had or had not been on the previously compared with the last learning trial [(Td/
presentation, the subject is asked to recall learned list of 10 words by responding "yes" T3) X 100 Savings(%)] were also calcu¬
=

as many of the words as possible. Two ad¬

for targets and "no" if the item (the dis- lated. Only data from subjects scoring two
ditional learning trials immediately follow tractor) was not on the list. words or more correct on the last learning
trial were included in the savings measures.
using the same procedure. The items on the Measures
list are randomized across the three learn¬ Statistical Analyses
ing trials in a standardized manner. The To assess the relative impairment of
number of words recalled correctly on each learning and memory functions in AD, the The CERAD Data Management Center
trial (maximum 10) is scored. In addition, performances of the three AD groups and carried out statistical calculations using
the number of intrusion errors, ie, re- the normal control group were compared on the statistical analysis system (SAS, Ver-

Downloaded From: http://archneur.jamanetwork.com/ by a Michigan State University User on 06/17/2015

 Table 3.—The Efficacy of Test Measures in Correctly Classifying Subjects Based on
percentage of patients (sensitivity) and
control subjects (specificity) classified cor¬
Cutting Scores (+2 SDs From the Control Mean)* rectly by the discriminant procedure.
Groups With AD, % (No.] RESULTS
Memory Control
Measures Subjects Mild Moderate Severe Mean scores for each of the learning
Learning and memory measures were computed
Trial 1 96 (47) 41 (20) 67 (33) 92 (45)
for the control subjects and for the pa¬
Trial 2 94 (46) 49 (24) 74 (36) 96 (47)
tients with AD (Table 2). Analyses of
Trial 3 98 (48) 41 (20) 59 (29) 90 (44)
these data disclosed a highly signifi¬
Delay recall 94 (46) 86 (42) 96 (47) 96 (47)t
cant group difference on each test
Savings 96 (47) 62 (29) 93(41) 82 (28)t
Recognition (P < .001) with the patients perform¬
Rec-Yes 96 (47) 39 (19) 53 (26) 73 (35) ing more poorly than the matched
Rec-No 98 (48) 25 (12) 32 (16) 48 (23) control subjects.
Intrusions Evaluation of sensitivity and speci¬
Trial 1 96(47) 14(7) 16(8) 14(7) ficity at a set cutoff score was used to
Trial 2 78(38) 33 (16) 27 (13) 33(16) determine the efficacy of each memory
94 (46) 20 (10) 35 (17) 39 (19) measure in differentiating the patients
Delay 90 (44) 27 (13) 26 (12) 18(9) with mild dementia from the control
*The percentage and number of subjects correctly classified are indicated for each memory measure. Cut¬ subjects, the mildly demented from
ting scores were selected 2 SDs from the control mean and rounded up to nearest whole integer. By deliberately those moderately demented, and the
setting the cutoff at this level, classification of controls will be, by definition, approximately 95% correct.
tThe best overall measure is the delayed recall measure. Savings scores also differentiated well between
moderately demented from the pa¬
groups, but they were not as powerful as the delayed recall measure. The differing capacities of the two mea¬
tients with more severe dementia. By
sures: delayed recall and savings scores, appear due to restricting the analysis of savings to individuals who analyzing the frequency distribution
successfully learned two words or more by the learning trial.3 By this procedure, the data from the most severely of scores, the percentage of patients
amnesic individuals are not included in the savings measures and the sizes of the groups with Alzheimer's dis¬
and control subjects correctly classi¬
ease (AD) are reduced from 49 subjects to 47, 45, and 34 subjects for the mild, moderate, and severe groups,
respectively. fied by a 2-SD cutoff score was deter¬
mined (Table 3). On the basis of this
procedure, it was apparent that not all
measures were equally effective in dis¬
Table 4.—Efficacy of the Memory Measures in Determining Group Assignment Based criminating between the older-aged
on Stepwise Discriminant Function Analysis Results control subjects and the groups with
Comparables
dementia. The best overall discrimina¬
Control subjects vs group
tor was found to be delayed recall. This
with mild AD*t measure correctly classified 96% of the
Delay .588 132.7 <.001 control subjects, 86% of the patients
Rec-No .026 2.48 with mild dementia, and 96% of both
Group with mild AD vs group the moderately and severely demented
with moderate
Delay .091 8.86 <004
patients. Although the patients with
AD committed more intrusion errors
Group with moderate AD vs
than did the control subjects, the in¬
group with severe AD§
Rec-Yes .065 5.23 trusion measures had poor overall dis-
Intrusions .052 4.09 <.047 criminability. Recognition memory
*AD indicates Alzheimer's disease. also proved to be a poor discriminating
fin this comparison, the delay measure was the single best discriminator between the groups. When this score measure.
was combined with the true-negative recognition responses (Rec-No), the discriminant formula correctly clas¬ The most effective combination of
sified 91% of the subjects, separating 89% of the mildly demented patients from 92% of the control subjects.
this comparison, only the delay recall measure entered into the linear model. This measure correctly clas¬ discriminating tests was determined
sified 65% of the subjects, separating 47% of the mildly demented patients from 84% of the moderately demented by submitting the data to stepwise
patients. linear discriminant function analysis
§ln this comparison, the true-positive recognition responses (Rec-Yes) and intrusion errors entered into the
equation, correctly classifying 66% of the subjects: 73% of the moderate group and 59% of the severe group.
(Table 4). All tests were included in the
procedure. This procedure, again,
identified the delay recall score as the
Sion 5, SAS Institute, Cary, NC, 1981). The trol subjects). These scores were subse¬
single best discriminator between the
results were analyzed in several ways. To quently used to identify the number of sub¬ patients with mild dementia and the
control subjects. When combined with
provide an overall view of group differ¬ jects scoring in the adequate or defective the correct "no" score (Rec-No) on the
ences, mean scores on each memory mea¬ range on each memory measure. Sensitivity
sure were computed for the normal group and specificity were determined through a recognition task, the discriminant
and AD groups. Differences between groups comparison of performance on the different function formula correctly classified
were assessed through a one-way ANOVA. memory measures. 91% of the subjects, separating 89% of
When significant differences arose, post hoc In addition to the above analyses, the ac¬ the mildly demented patients from
comparisons were made with individual t curacy of the test scores in predicting 92% of the control subjects.
tests. whether a person was a normal control In the comparison of the mildly and
Once the distributions for each of the subject or a patient with mild, moderate, or moderately demented patients, the de¬
memory measures were established, cut¬ severe AD was determined by means of
ting scores were then determined, so that stepwise linear discriminant function anal¬ layed recall measure was the only
measure entering into the linear
impaired performance on each task was de¬ ysis. This procedure identifies the combina¬
fined as scores falling 2 SDs below the con¬ tion of tests that best discriminates the pa¬ model. This measure correctly classi¬
trol mean (ie, performance levels exceeded tients from the control subjects. This fied 65% of the subjects, separating
by approximately 95% of those of the con- method also allows the computation of the 47% of the mildly impaired patients

Downloaded From: http://archneur.jamanetwork.com/ by a Michigan State University User on 06/17/2015

 from 84% of the moderately impaired tween the more severely demented General Hospital, Boston); L Bierer, MD (Mount
Sinai Medicai Center, New York, NY); L. Harrell,
individuals. groups, although the accuracy of cor¬
In the later stages of disease, the rect group placement is limited (66% PhD, MD (University of Alabama, Birmingham);
W. C. Wiederholt, MD, and D. Salmon, PhD (Uni¬
moderate and severe comparison, the accuracy). Despite this limitation, the versity of California at San Diego, La Jolla); S.
combination of correct "yes" recogni¬ data suggest that for staging of de¬ DeKosky, MD, and F. Schmitt, PhD (University of
tion (Rec-Yes), and intrusion errors mentia, attention to the commission of Kentucky Medicai Center, Lexington); G. Ratcliff,
DPhil (University of Pittsburgh [Pa] Medicai
appeared to be the best—albeit not intrusions and to recognition memory School); P. Tariot, MD, and C. Podgorski, PhD
very powerful—discriminators be¬ is likely to be more informative at the (University of Rochester [NY] Medicai Center); V.
tween groups. The overall accuracy of later stages of AD. Other memory Henderson, MD, and D. Freed, PhD (University of
the combined measures in discrimi¬ Southern California, Los Angeles); M. Weiner,
measures, including delayed recall
MD, and J. Horn, PhD (University of Texas
nating between the groups was 66%, that is important at the early stages of Southwestern Medical Center, Dallas); E. Larson,
separating 73% of the moderately im¬ dementia, do not differentiate between MD, L. Teri, PhD, and Z. Haycox (University of
paired individuals from 59% of the se¬ the later stages of the disease. Washington, Seattle); L. Schut, MD, and J. Mor¬
verely demented patients. It should be cautioned that the cur¬ timer, PhD (Veterans Administration Medical
rent study compared patients who had Center, Minneapolis, Minn); and J. Morris, MD,
COMMENT and E. LaBarge, MEd (Washington University
a diagnosis of AD with normal, high- School of Medicine, St Louis, Mo) for their assis¬
The major finding of this study per¬ functioning control subjects. It has not tance in the neurological and neuropsychological
tains to the ability of the CERAD ver¬ determined the predictive validity of evaluations of the patients and control subjects
described in this article.
bal memory test to discriminate be¬ the memory tasks for detecting cogni¬ Members of the CERAD Steering Committee
tween patients with very mild AD and tive changes in an open community not listed previously are Leonard Berg, MD, and
normal elderly control subjects. Our population in whom the degree of Mokhtar Gado, MD (Washington University
School of Medicine, St Louis, Mo), and Patricia
results suggest that in patients with memory impairment may be very sub¬
subtle cognitive impairment caused by tle. Nonetheless, our results in pa¬ Davis, MD, and Suzanne Mirra, MD (Emory Uni¬
versity School of Medicine, Atlanta, Ga).
AD, predictive accuracy can be en¬ tients with very mild AD indicate that We appreciate the helpful comments provided
hanced by focusing on delayed recall the delayed recall measure is a robust by Gerda Fillenbaum, PhD. Fleta Ware prepared
measures. Other measures, including discriminator of early impairment and the manuscript and Kathleen Keating assisted
with the preparation of the tables.
learning (immediate recall), recogni¬ should be considered in screening for
tion memory, and intrusions, are also dementia in population surveys. References
affected by AD, but they are not as In conclusion, this study of the
CERAD verbal memory task suggests 1. Morris JC, Mohs R, Rogers H, et al. CERAD
powerful in terms of differentiating clinical and neuropsychological assessment of
the patients with very mild AD from that it may be a useful clinical tool for Alzheimer's disease. PsychopharmacolBull. 1988;
the control subjects. The data suggest the early detection of AD in this pop¬ 24:641-651.
that emphasis on these latter mea¬ ulation. However, all memory indexes 2. Morris JC, Heyman A, Mohs RC, et al. The
sures in the early stages of dementia were not of equal value. Delayed recall Consortium to Establish a Registry for Alzhei-
mer's Disease (CERAD), I: clinical and neuropsy-
caused by AD is more likely to lead to proved to be the most sensitive and se¬ chological assessment of Alzheimer's disease.
false-negative errors than is attention lective measure. The memory test does Neurology. 1989;39:1159-1165.
to the delayed recall measure for not appear to have much practical 3. Moss MB, Albert MS, Butters N, et al. Dif-
which sensitivity and specificity are value in the later staging of the illness, ferential patterns of memory loss among patients
in which there are significant memory with Alzheimer's disease, Huntington's disease,
both high. Thus, for detection of the and alcoholic Korsakoff syndrome. Arch Neurol.
early stages of AD, the delayed recall impairments. Other cognitive mea¬ 1986;43:239-243.
measure (which is an index of forget¬ sures from the CERAD battery or an 4. Knopman DS, Ryberg S. A verbal memory
evaluation of global cognitive function test with high predictive accuracy for dementia of
ting) appears to be most useful. These the Alzheimer type. Arch Neurol. 1989;46:141-145.
findings are consistent with other re¬ based on scores derived from the entire
5. Troster AI, Jacobs D, Butters N, et al. Dif-
ports that stressed the importance of CERAD battery may prove more use¬ ferentiating Alzheimer's disease from Hunting-
rate of forgetting in distinguishing AD ful for the staging of such patients. ton's disease with the Wechsler Memory Scale
from other dementing disorders,3'5 as Revised. Clin Geriatr Med. 1989;5:611-632.
This study was supported by the -National
6. Hart RP, Kwentos JA, Harkins SW, et al.
well as from normal control cases.6 Institute on Aging, Bethesda, Md, grants
Rate of forgetting in mild Alzheimer's-type de-
AG05128 and AG06790 to Duke University Med¬
Delayed recall proved useful in de¬ ical Center, Durham, NC, and by funds from the mentia. Brain Cogn. 1988;7:31-38.
tecting early evidence of AD; however, Medical Research Service of the Veterans Admin¬
7. Butters N, Granholm E, Salmon DP, et al.
because AD cases "bottom out" so istration, Washington, DC, and by the National Episodic and semantic memory: a comparison of
amnesic and demented patients. J Clin Exp Neu-
quickly on the measure, it had little Institute on Aging grants AG08204 and AS05131
to the University of California at San Diego. ropsychol. 1987;9:479-497.
practical value in staging this disor¬ We would like to acknowledge the various 8. Branconnier RJ, Cole OJ, Spera KF, et al.
der. All memory measures discrimi¬ CERAD investigators who contributed to this Recall and recognition as diagnostic indices of
nate clearly between patients with work. These investigators include W. Strittmat¬ malignant memory loss in senile dementia: a
moderate dementia and elderly control ter, MD, and F. Pirozzolo, PhD (Baylor University Bayesian analysis. Exp Aging Res. 1982;8:189-193.
9. McKhann G, Drachman D, Folstein M, et al.
subjects, and also between the severely College of Medicine, Houston, Tex); J. Blass, MD, Clinical diagnosis of Alzheimer's disease: report
demented patients and control sub¬ PhD, and P. Brady (Burke Rehabilitation Center/ of the NINCDS-ADRDA Work Group under the
Cornell University Medical College, White Plains,
jects. However, none of the measures NY); P. Whitehouse, MD, and M. Patterson, PhD auspices of Department of Health and Human
Services Task Force on Alzheimer's Disease. Neu-
was particularly powerful in discrimi¬ (Case Western Reserve University, Cleveland,
Ohio); R. Mayeux, MD, and Y. Stern, PhD (Co¬ rology. 1984;34:939-944.
nating between the moderately and lumbia University, New York, NY); N. Earl, MD 10. Katzman R, Brown T, Fuld P, et al. Valida-
severely demented cases. The results (Duke University Medical Center, Durham, NC);
tion of a short orientation-memory concentration
test of cognitive impairment. Am J Psychiatry.
from the linear discriminant function H. Karp, MD, C. Hill, PhD, and B. Faherty, RN,
1983;140:734-739.
analyses indicate that, of the various MSN (Emory University School of Medicine, At¬
11. Folstein JF, Folstein SE, McHugh PR.
memory measures, recognition mem¬ lanta, Ga); C. Clark, MD (Graduate Hospital, 'Mini-Mental State': a practical method for grad-
Philadelphia, Pa); B. Rovner, MD, and J. Brandt,
ory and intrusion errors have the PhD (Johns Hopkins University, Baltimore, Md); ing the cognitive state of patients for the clinician.
J Psychiatr Res. 1975;12:189-198.
greatest value in discriminating be- J. Growdon, MD, and E. Hoffman (Massachusetts

Downloaded From: http://archneur.jamanetwork.com/ by a Michigan State University User on 06/17/2015