Schwindel-sprechstunde.de

Articles
Screening for dementia with the Memory
Impairment Screen
H. Buschke, MD; G. Kuslansky, PhD; M. Katz, MPH; W.F. Stewart, PhD; M.J. Sliwinski, PhD; H.M. Eckholdt, PhD; and R.B. Lipton, MD Article abstract—Objectives: To validate a sensitive and specific screening test for AD and other dementias, assess its
reliability and discriminative validity, and present normative data for its use in various applied settings. Background: To
improve discrimination in screening for AD and dementia, we developed the Memory Impairment Screen (MIS), a
4-minute, four-item, delayed free- and cued-recall test of memory impairment. The MIS uses controlled learning to ensure
attention, induce specific semantic processing, and optimize encoding specificity to improve detection of dementia. Meth-
ods: Equivalent forms of the MIS were given at the beginning and end of the testing session to assess alternate forms
reliability. Discriminative validity was assessed in a criterion sample of 483 aged individuals, 50 of whom had dementia
according to Diagnostic and Statistical Manual of Mental Disorders (3rd ed., revised) criteria. Results: The MIS had good
alternate forms reliability, high construct validity for memory impairment, and good discriminative validity in terms of
sensitivity, specificity, and positive predictive value. We present normative data for use in settings with different base
rates (prevalences) of AD and dementia. Conclusion: The MIS provides efficient, reliable, and valid screening for AD and
other dementias.
There are urgent scientific and public health reasons other metabolic causes, and infections.12,13 Effective to develop effective methods of screening for AD and recognition of the dementia syndrome is an impor- other dementing disorders, especially in the early tant step toward establishing and addressing a spe- clinical stages.1 Individuals with early disease may not consult physicians or may not mention their cog- Despite several decades of research, the 1996 US nitive complaints.2 They may be unaware of their Agency for Health Care Policy and Research Clinical decline or believe that their memory problems are Practice Guideline on recognition and initial assess- part of the normal aging process.3 Even if they report ment of Alzheimer’s disease and related dementias14 symptoms, dementia often goes undetected and un- indicates that there is no evidence to recommend one screening test over another. Because memory im- Dementia is a clinical syndrome with many reme- pairment is often the earliest feature of AD, a very diable causes. Two pharmacologic treatments for AD brief test for memory impairment with sufficient have been approved,5-7 and other pharmacologic sensitivity, specificity, and positive predictive value treatments are being developed.8,9 Early detection is (PPV) should serve as a good screening test for AD important because treatment may be more effective and other dementias with memory impairment.
early in disease when useful cognitive function can Very brief memory tests that assess (delayed) re- still be preserved.10,11 AD may soon meet criteria for call of three or four words are used in clinical prac- community-based “tertiary prevention” programs.1 tice to screen for memory impairment15 and are Other treatable causes of dementia include subdural included in mental status assessments.16-22 However, such very brief three- or four-word tests for memory See also page 224
From The Saul R. Korey Department of Neurology (Drs. Buschke and Kuslansky, M. Katz, and Drs. Sliwinski, Eckholdt, and Lipton), Einstein Aging Study(Drs. Buschke and Kuslansky, M. Katz, and Drs. Sliwinski, Eckholdt, and Lipton), Rose F. Kennedy Center for Mental Retardation and Human Development(Drs. Buschke, Kuslansky, Sliwinski, and Eckholdt), and Department of Epidemiology and Social Medicine (Dr. Lipton), Albert Einstein College of Medicine,Bronx, NY; Department of Epidemiology (Dr. Stewart), School of Public Health, The Johns Hopkins University, Baltimore, MD; and Innovative MedicalResearch (Drs. Stewart and Lipton), Towson, MD.
Supported by NIA grants AG03949 and AG12448, and NICHD grant HD-01799.
Presented in part at the annual meeting of the American Neurological Association; San Diego, CA; September 1997.
Received March 16, 1998. Accepted in final form September 12, 1998.
Address correspondence and reprint requests to Dr. Herman Buschke, Department of Neurology, Kennedy 912, Albert Einstein College of Medicine, Bronx,NY 10461.
Copyright 1999 by the American Academy of Neurology impairment have a high rate of false-negative errors Health Care Financing Administration (HCFA) lists for (low sensitivity) or false-positive errors (low specific- the area adjacent to our clinical research center. According ity) and may yield different results depending on the to the HCFA, their lists include 97% of individuals over words used.23-25 The 10-word Delayed Word Recall the age of 65. The eligibility criteria for inclusion in this Test26 discriminates between nondemented elderly study were age 65 and over, ambulatory, and adequate individuals and mild AD patients with high sensitiv- ability to understand and speak English. In comparison ity (89%) and specificity (98%) but takes longer to with the entire HCFA sample for the relevant segment of administer because the list is presented twice, and the Bronx, our study sample was slightly older but other- each time the individual must make up a sentence wise similar to the community in terms of sex (36% male), for each word. A very brief, well-standardized, accu- racial distribution (80.7% white, 16.2% African American, rate screening test for memory impairment and de- and 2.7% other), and education (16.6% had fewer than 9 mentia might find broader use in clinical practice.
We previously demonstrated that the Double These participants received a neuropsychological test battery as well as medical, epidemiologic, social, and be- Memory Test (DMT) has high discriminative validity havioral questions every 18 months. A neurologic evalua- for AD and related dementias in comparison with tion was performed on those participants who had Blessed standard memory tests that do not optimize encod- Information Memory and Concentration (BIMC)16 test er- ing specificity.27,28 The advantage of the DMT, rela- ror scores of greater than 8, a total recall score of 44 or less tive to conventional memory tests, is greatest for the on the FCSRT,29,31 or who were suspected of having demen- patients with mild disease; the DMT has a sensitiv- tia based on self-report, informant report, or observations ity of 93% and a specificity of 99% for the diagnosis of the testers. Additional diagnostic testing, ordered by of mild dementia.28 Although these unusually high the neurologist, included neuroimaging, blood tests, and sensitivities and specificities suggest that the DMT other evaluations as needed. A diagnosis of dementia was may be useful in detecting early dementia, it is not made according to Diagnostic and Statistical Manual of suitable as a screening test. It has 128 items, must Mental Disorders (3rd ed., revised) (DSM-III-R) criteria,32 be administered by a trained tester in two parts, and and a diagnosis of AD was based on National Institute of requires approximately 20 minutes to complete. The Neurological and Communicative Disorders and Stroke– multi-trial Free and Cued Selective Reminding Test Alzheimer’s Disease and Related Disorders Assocation (FCSRT),29 a forerunner of the DMT, also uses con- (NINCDS-ADRDA) criteria.33 Of the 50 participants with trolled learning and cued recall but, because of its DSM-III-R dementia (yielding a sample base rate of length, is also not suitable for screening.
10.4%), 39 met criteria for AD. The other 433 participants The Memory Impairment Screen (MIS) is a brief, four-item delayed free- and cued-recall memory im- The participants received the MIS as part pairment test that also uses controlled learning and of a neuropsychological test battery that included the cued recall to optimize encoding specificity. Con- Wechsler Adult Intelligence Scale–Revised,34 the Logical trolled learning29 requires the individual to search Memory I and Verbal Paired Associates I subtests from the for and identify a to-be-remembered item in response Wechsler Memory Scale–Revised,35 the FCSRT,29,31,36 the to its category cue. The category cues are also used to BIMC test,16 and the Zung Depression Scale.37 All partici- elicit retrieval by cued recall. Using the same cues at pants scoring in the depressed range on the Zung Depres- acquisition and retrieval coordinates acquisition and sion Scale (Ն50) were clinically assessed by a study retrieval, optimizing encoding specificity because psychiatrist using the Structured Clinical Interview for the “specific encoding operations performed on what is DSM-III-R.38 Alternate forms of the MIS were adminis- perceived determine what is stored and what is tered at the beginning and the end of the neuropsycholog- stored determines what retrieval cues are effective in ical test session in counterbalanced order. Participants providing access to what is stored.”30 We postulated were tested individually and told before presentation of the that improving retrieval by these procedures would test items that they should remember the items so thatthey could recall them later. Each participant was pre- provide a screening test with improved discrimina- tive validity. To test the potential of the MIS as a items to be recalled printed in 24-point uppercase letters.
screening test for AD and dementia, we assessed its Each item belonged to a different category. The individual reliability, construct validity with respect to a stan- was asked to read the items aloud and then asked to iden- dard memory test, and discriminative validity as a tify and name each item (e.g., “potato”) when the tester said its category cue (e.g., “vegetable”). The sheet was thenremoved. After a nonsemantic interference task (repeated Methods.
counting from 1 to 20 and back) lasting approximately 2 to consent as specified by the Committee on Clinical Investi- 3 minutes, the individual was asked for free recall of the gations at Albert Einstein College of Medicine. Partici- four items in any order. The category cues were then pre- pants were 483 community-residing volunteers, 50 of sented to elicit cued recall of only those items that were whom had dementia, who participated in the Einstein Ag- not retrieved by free recall. The number of items retrieved ing Study, a longitudinal study of dementia and normal by free recall and the number retrieved by cued recall were aging. The sample included 286 community volunteers re- cruited from senior centers and physician practices. In ad- The MIS score is calculated as [2 ϫ (free re- dition, 197 individuals were systematically sampled from call)] ϩ [cued recall]. Simple free recall provides a single January (2 of 2) 1999
score—the number of items retrieved by free recall. Cued Table 1 Demographic and neuropsychological characteristics of
recall in the MIS provides a second score—the number of items retrieved by cued recall. If cued recall of all items were tested, a combined free- and cued-recall score would be obtained by adding the total cued-recall score to thetotal free-recall score. However, in the MIS, cued recall is tested only for those items that were not retrieved by free recall. Because items retrieved by free recall also are re- trieved by cued recall, the equivalent of a total cued-recallscore can be obtained by adding free recall (i.e., those items that would also have been retrieved by cued recall) to the actual cued recall obtained by testing only those items that were not retrieved by free recall. Adding thisequivalent total cued-recall score to the free-recall score results in an MIS score equal to [2 ϫ (free recall)] plus (obtained) cued recall. Adding cued recall to free recall alsoincreases the range of scores (0 to 8), which may improve Means (SD) are presented except for sample size and sex.
BIMC ϭ Blessed Information Memory Concentration test; WAIS- R ϭ Wechsler Adult Intelligence Scale–Revised; MIS ϭ Memory nate forms reliability using the intraclass correlation. In- ternal consistency was determined using Cronbach’scoefficient alpha.
Construct validity of the MIS for detecting memory im- dementia group was slightly older and had slightly less pairment was assessed using the chance corrected kappa education, although the gender composition of the two statistic to measure the concordance between memory im- groups was similar. The dementia group scored approxi- pairment on the MIS and impairment as defined by perfor- mately one-half SD unit higher ( p Ͻ 0.001) on the Zung mance on the FCSRT. The MIS is intended to screen for Depression Scale than did the nondementia group. Zung memory impairment and is not intended as an index of Depression Scale scores were not correlated with the MIS memory ability. The FCSRT cut-score was selected based scores for the demented (Spearman ϭ 0.09, p ϭ 0.53) or for on its ability to distinguish individuals with and without the nondemented groups (Spearman ϭ 0.002, p ϭ 0.96).
Based on clinical interviews by a psychiatrist of individu- Discriminative validity of the MIS was assessed by cal- als with Zung Depression Scale scores of Ն50, seven par- culating the sensitivity and specificity of the MIS for de- ticipants met DSM-III-R criteria for major depressive tecting dementia and for detecting AD for various MIS disorder, six of whom were in the nondemented group.
cut-scores. The base rate of dementia in our sample was Alternate forms reliability for the MIS was 10.4% (50 of 483). Of the 50 demented participants, 39 assessed by administering one of two forms to a subset (78%) met criteria for possible or probable AD.
(n ϭ 429) of clinically diagnosed demented and nonde- Although normative data are often presented in the mented participants at the beginning and end of a neuro- form of percentiles or means and SDs, we present norms in psychological evaluation. The intraclass correlation was the form of the probability of dementia (or AD) given dif- high, 0.69, indicating strong concordance between the two ferent MIS cut-scores. To calculate the probability of de- forms of the MIS and a high level of alternate forms reli- mentia, one must know the test sensitivity and specificity ability. Internal consistency also was good for both forms— at each cut-score as well as the base rate of dementia.39,40 the coefficient alpha was 0.67 for both forms.
Because the base rate of dementia varies with setting as well as age, we present normative data that cover a broad a screening test for memory impairment as determined by range of dementia base rates. Normative data are pre- the 16-item multi-trial FCSRT.31 Memory impairment is sented for various dementia base rates as the probability identified by the FCSRT by a total free- plus cued-recall of dementia or AD (i.e., the PPV) using each MIS score as a score over all three trials of 44 or less. Classification of memory impairment by the MIS (as defined by a score of 4 Probability of dementia or AD was calculated according or less) was strongly associated with memory impairment to Bayes’ theorem.39 The formula for PPV yields the pro- as defined by the gold standard FCSRT (kappa ϭ 0.62).
portion of individuals who screen positive at a given cut- score who actually have dementia (for a specific prevalence tive validity of the MIS first as a screening test for demen- rate). Note that the PPV also represents the probability of tia in general and then specifically for clinically diagnosed dementia for individuals with scores falling at or below a AD. The trade-off between sensitivity and specificity of the given score on the MIS in a sample with a given base rate MIS as a screening test for dementia and AD as the cut- of dementia. PPV is a useful index of the efficiency of a score varies is shown by the receiver operating character- screening test in different applied settings that have differ- istic (ROC) curves41,42 in figure 1 (all dementia, including ent base rates of dementia. The negative predictive value AD, versus nondemented) and figure 2 (only AD versus (NPV) is the proportion of individuals who screen negative nondemented). The area under the ROC curve is regarded as a useful index of a diagnostic test’s performance. Thearea under the ROC curve for discrimination of dementia Results.
Demographic and neuropsychological character- (including AD) is 0.94. The area under the ROC curve for istics of the study sample are summarized in table 1. The discrimination of AD is even larger, 0.97. These figures January (2 of 2) 1999
Table 2 Dementia sensitivity and specificity of each MIS score
with corresponding probabilities of dementia (PPV) at different
base rates
Figure 1. Receiver operating characteristics of the Mem-ory Impairment Screen as a screening test for dementia MIS ϭ Memory Impairment Screen; PPV ϭ positive predictive graphically indicate that the specificity of the MIS is con- (see Discussion), a cut-score of 4 provides a high level of sistently high for a large range of sensitivities. The influence sensitivity (0.80), specificity (0.96), and PPV (Ͼ0.69) for all of the sensitivity-specificity trade-off on the usefulness of the base rates except the lowest at 5%. Because the base rate MIS as a screening test for AD and dementia is examined in of dementia in the current sample is approximately 10%, the following section on normative data.
approximately 70% of study participants who scored 4 or less had some type of dementia. The NPV reflects the specificity, and PPV (i.e., probability of dementia or AD) probability that an individual who screens negative does for each cut-score on the MIS for different base rates of not have disease. A cut-score of 4 provides excellent NPVs dementia. Performance at or below the cut-score is taken at low base rates of dementia. For example, at a base rate as evidence of dementia. Sensitivity measures the propor- of 5% the NPV is 0.99; at a base rate of 10% the NPV is tion of those with dementia who are correctly identified as 0.98; at a base rate of 20% the NPV is 0.95. Thus one can demented. Specificity measures the proportion of those be reasonably sure that those who screen negative do not without dementia who are correctly identified as nonde- mented. PPV is the proportion of those who meet a given The results are similar when the analysis is restricted cut-score (i.e., screen positive) and actually have dementia.
to cases of AD and controls (see table 3). A MIS cut-score of Table 2 displays the sensitivity and specificity for de- 4 had a sensitivity of 0.87 for AD, which is slightly higher tecting any type of dementia using each level of MIS score than the sensitivity for all dementia. This differential as a cutoff. Although the selection of an optimal cut-score sensitivity is expected because (early) memory impairment will depend on the intended clinical or research application Table 3 AD sensitivity and specificity of each MIS score with
corresponding probabilities of AD (PPV) at different base rates
Figure 2. Receiver operating characteristics of the Mem- MIS ϭ Memory Impairment Screen; PPV ϭ positive predictive ory Impairment Screen as a screening test for AD. January (2 of 2) 1999
is the hallmark of AD but not of all other (non-AD) using the FCSRT as the gold standard; the FCSRT has an empirically based cut-score for dementia- To examine how the MIS performed for different levels related memory impairment.31 Classifications by the of dementia severity, we assessed discriminative validity MIS and by the longer FCSRT were quite similar, separately for demented individuals who fell at or below 13 supporting the view that the MIS measures the same (mild) and at or above 14 (moderate) on the BIMC test.
The sensitivity for mildly demented individuals with any type of dementia at an MIS cut-score of 4 is 0.69; for of the MIS for dementia and AD was assessed in a moderately demented individuals the sensitivity is 0.92.
large sample of individuals with and without clini- When analyses are restricted to only AD dementia, the cally diagnosed dementia. In clinical practice, when sensitivity for mildly demented individuals is 0.79, and for older patients present with memory or cognitive moderately demented individuals the sensitivity is 0.95.
complaints, the first issue is to diagnose or exclude The demographic variables age, education, and gender as well as their interactions (with each other and with the dementia of any sort (e.g., AD, vascular dementia, MIS) were added to a logistic regression model as covari- dementia with Lewy bodies). If dementia is present, ates to determine their effect on predictive accuracy. None a more specific etiologic workup is essential. Thus, it of these demographic variables or interaction terms was is clinically important for a screening instrument to significant at the 0.05 level. When the Zung Depression identify dementia syndromes with diverse etiologies.
Scale score and its interaction were added to the logistic Although we present normative data separately for regression model, they were not significant at the 0.05 detecting all dementia (with memory impairment) or level. The ethnicity of the 17 false positives was examined only AD, we believe that the discriminative validity to determine the possible influence of ethnic background data for detecting all dementia is most relevant for on classification by the MIS. Of the 17 nondemented indi- the majority of clinical and research applications.
viduals who failed the MIS (cut-score ϭ 4), 2 (11.7%) were Discriminative validity of the MIS is influenced by African Americans, which is less than expected given the dementia severity. In moderate dementia, sensitivity sample composition of 16.2% African Americans. Thus, in- is 0.92 for all dementia and 0.95 for AD. Although dividuals of minority backgrounds were no more likely to sensitivity for mild dementia is lower, it is reason- be incorrectly identified as memory impaired by the MIS ably high— 0.69 for all dementia and 0.79 for AD.
The number of participants with non-Alzheimer de-mentia is relatively small; further experience with Discussion.
memory impairment, dementia, and AD that is brief, The sensitivity and specificity of the MIS com- easy to use, reliable, and valid. It is based on our pares favorably with results in other samples for previous research demonstrating that memory tests longer screening tests for dementia such as the with high encoding specificity, such as the DMT and BIMC test16 and the Mini-Mental State Examination the FCSRT, have higher discriminative validity for (MMSE).17 Because the BIMC test was used by clini- AD and for dementia in general.27-29,31,36 The MIS cians in this study to assign diagnoses, we cannot uses category cues during acquisition (learning) to compare the MIS and the BIMC test obtained in this ensure attention, induce semantic processing, and study and must rely on analysis of the MMSE and optimize encoding specificity, and during retrieval to the BIMC test in other studies. Recent analyses of benefit from encoding specificity, increase retrieval, neuropsychological tests for diagnosis of dementia by and improve discrimination. Because the MIS is a Wilder et al.42 and Meiran et al.40 reported the dis- screening test, individuals who fail it are referred for criminative validity of the MMSE and the BIMC test definitive diagnostic assessment. Individuals who for detecting AD. When Wilder et al.42 selected cut- pass the test may be rescreened in the future de- scores for these tests so that sensitivity is high pending upon the context of screening. In this sec- (0.90), their specificities are unimpressive— 0.44 for tion we summarize the psychometric properties of the MMSE and 0.60 for the BIMC test. Data from the MIS, illustrate how the normative data might be Meiran et al.40 indicate that when MMSE sensitivity used in particular clinical and research applications, is 0.91, MMSE specificity is only 0.25. The specificity of MIS for detecting AD is 0.96 when sensitivity is We demonstrated that the MIS has high alternate set to 0.87 (see table 3). These data suggest that the forms reliability, indicating that the two forms are MIS may be a more sensitive screen for AD than similar and that individuals perform in a highly re- these commonly used mental status examinations.
peatable fashion. The high reliability scores show Because the sensitivity and specificity of the MIS in that the alternate forms of the MIS can be used for this study are compared with the sensitivity and repeated administration. The availability of equiva- specificity of the MMSE and the BIMC test in other lent alternate forms is an advantage in applications studies, these comparisons must be interpreted with that require serial screening. There was a high de- caution; the populations used in different studies gree of internal consistency for the two forms of the may have been different. In particular, the influence MIS (Cronbach’s alpha ϭ 0.67 for both).
of education and cultural background on the relative performance of the MIS, the MMSE, and the BIMC as a screen for memory impairment was evaluated test requires further study.43,44 Although the MMSE January (2 of 2) 1999
and the BIMC test are influenced by education, at ity (0.91). How accurately this cut-score classifies least in this sample education did not affect the dis- patients will depend on the dementia base rate in the clinical setting. In a clinic-based sample of indi- viduals with memory complaints we might reason- performance on cognitive tests and enters into the ably expect a dementia base rate of 20%. At that differential diagnosis of dementia.38 In the current base rate, a MIS cut-score of 5 corresponds to a 0.72 study, Zung Depression Scale scores were not corre- probability of dementia. This means that 72% of lated with MIS scores for the demented or for the those meeting a MIS cut-score of 5 will have demen- nondemented groups. Only seven participants met tia (i.e., PPV ϭ 0.72), and the sensitivity of 0.86 DSM-III-R criteria for major depression, and of these means that only 14% of all the demented patients six were in the nondemented group. Of the 17 false- will be missed by the screen. In this setting the NPV positive instances in the study (MIS of 4 or less, but of 0.96 means that of those who screen negative, only not clinically demented), only one had depression.
4% would prove to have dementia on a definitive We do not have adequate data to draw firm conclu- sions about the influence of moderate or severe de- In contrast to a purely clinical application, re- pression on MIS performance, because of small search applications may require higher specificity to numbers of depressed individuals. However, there is efficiently identify potentially diseased individuals.
little evidence to suggest that depression influenced When the base rate of dementia is low, screening can MIS performance in our study results.
be used to identify individuals likely to have demen- tia on definitive evaluation. For example, if the base mative data needed to select cut-scores for the MIS.
rate of dementia is 10% in the target population, a The PPVs, which reflect the probability of dementia researcher might select from table 2 a MIS cut-score for individuals falling at or below a MIS cut-score for of 3, which has a very high specificity (0.98). At that a given base rate, provide an estimate of how likely specificity the PPV is 0.78, which means that 78% of those who screen positive will prove to have disease those who screened positive will meet criteria for upon definitive assessment. We provide the following dementia. Unlike the clinician, whose choice of cut- examples to illustrate how to use the MIS or any score will often be driven by the need to maximize screening test to classify individuals. A clinician sensitivity, the researcher may require high specific- could be confident about classifying individuals scor- ity to yield a high PPV. But even with this more ing 4 or less on the MIS as demented because the restrictive criterion, the NPV of 0.97 means that only probability of being demented is 0.84 (at a base rate 3% of those who screen negative would prove to have of 20%; see table 2). However, if the base rate were dementia with a definitive assessment. Depending only 5%, a clinician could not make such a confident on the context, individuals who initially screen nega- classification because the probability of dementia is tive may be rescreened to identify individuals missed only 0.52. In general, clinicians will need to adopt lower (i.e., more stringent) cut-scores to maintain a Limitations and future directions. given level of confidence for samples with lower base results apply only to use of the MIS as a screening test for detection of dementia-related memory im- To select an optimal cut-score, a clinician or re- pairment, dementia, and AD in community-residing searcher needs to consider the base rate of the de- adults aged 66 to 97 years. Cut-scores selected from mentia in the screened population as well as the tables 2 and 3 may not have the same level of dis- clinical or research goals of the screening program.
crimination in specific clinical applications because For example, a clinician might chose a cut-score to the distribution of MIS scores for demented and non- maximize sensitivity so that impaired individuals demented older adults in primary care settings may are not missed by the screen, whereas a researcher not be equivalent to that observed in a sample of conducting clinical trials might wish to maximize the community volunteers. Although the current results PPV so that only those with a very high probability are encouraging, additional field studies are required of dementia screen positive and are enrolled in the to verify the validity of the MIS in screening for memory impairment and dementia in primary care settings. An important direction for future research of the normative data presented in tables 2 and 3 would be to develop MIS norms for use in specific will depend on the clinical or research goals of the screening program. For a clinical application in An additional concern is that the clinical diagnosis which early diagnosis is important and follow-up of dementia is a fallible gold standard, although we testing for definitive diagnosis is safe and inexpen- assigned diagnoses with great care using contempo- sive, optimal sensitivity is important. In this case, a rary research criteria. In this sample, individuals missed diagnosis would be undesirable given the lost with preclinical dementia may not have crossed the opportunity for treatment of AD or other types of threshold required for diagnosis. If such individuals dementia. Accordingly, a clinician might use the have memory impairment on the MIS, they appear data in table 2 to select a MIS cut-score of 5, which as false positives in our assessment of discriminative has a reasonably high sensitivity (0.86) and specific- validity. Of the 17 individuals considered false posi- January (2 of 2) 1999
tives (MIS score of 4 or less but not clinically de- 13. Tatemichi T, Sacktor N, Mayeux R. Dementia associated with mented), 12 had a Clinical Dementia Rating45 score cerebrovascular disease, other degenerative diseases, andmetabolic disorders. In: Terry R, Katzman R, Bick KE, eds.
of 0.5 (i.e., questionable dementia) and four devel- Alzheimer’s disease. New York: Raven Press, 1994:123–166.
oped DSM-III-R dementia by the next wave of 14. U.S. Agency for Health Care Policy and Research. Recognition follow-up 18 months later. Although we do not have and initial assessment of Alzheimer’s disease and related de- adequate data to comment on the discriminative va- mentias. Clinical Practice Guideline No. 19. Silver Springs,MD: U.S. Agency for Health Care Policy and Research, 1996.
lidity of the MIS for preclinical AD, it appears that 15. Petersen RC. Memory assessment at the bedside. In: Yanagi- the preclinical group may account for a substantial hara T, Petersen R, eds. Memory disorders. New York: Marcel proportion of the false positives, arguably leading to 16. Blessed G, Tomlinson R, Roth M. The association between quantitative measures of dementia and of senile changes in One final and important limitation of the MIS is the cerebral gray matter of elderly subjects. Br J Psychiatry that it is a screening test and should not be used in isolation to make clinical diagnoses of dementia. By 17. Folstein M, Folstein S, McHugh P. A practical method for grading the cognitive state of patients for the clinician. J Psy- identifying individuals with a very high (or very low) likelihood of dementia, a good screening test can play 18. Strub R, Black F. The mental status examination in neurol- an important part in the diagnostic process. No screening test should replace a comprehensive clini- 19. Teng E, Chui H. The modified Mini-Mental State (3MS) exam- ination. J Clin Psychiatry 1987;48:314 –317.
cal evaluation. The MIS is intended to identify for 20. Kokmen E, Naessens J, Offord K. A short test of mental clinical evaluation individuals who are likely to have status: description and preliminary results. Mayo Clin Proc 21. Weintraub S, Mesulam M. Mental state assessment of young and elderly adults in behavioral neurology. In: Mesulam M, Acknowledgment
ed. Principles of behavioral neurology. Philadelphia: FA The authors thank Anna Sapozhnik Miller, Betty Forro, Charlotte Magnotta, Maria Luisa Giraldi, and Mary Joan Sebastian for 22. Kokmen E, Smith G, Petersen R, Tangalos E, Ivnik R. The their assistance with data collection, and Drs. Randall Berliner short test of mental status: correlations with standardized and Howard Crystal for assistance with clinical evaluations and psychometric testing. Arch Neurol 1991;48:725–728.
23. Guilmette T, Tshoh J, Malcolm C. Orientation and three-word recall in predicting memory: age effects and false-negativeerrors. Neuropsychiatry Neuropsychol Behav Neurol 1995;8: Disclosure
The Albert Einstein College of Medicine owns the copyright and 24. Cullum C, Thompson L, Smernoff E. Three-word recall as a makes this test available as a service to the research community measure of memory. J Clin Exp Neuropsychol 1993;15:321– but licenses the test for commercial use.
25. White T, Bauer R, Bowers D, Crosson B, Kessler H. Recall of three words after five minutes: its relationship to performance References
on neuropsychological memory tests. Appl Neuropsychol 1995;2:130 –138.
1. Wilcock G, Ashworth D, Langfield J. Detecting patients with Alzheimer’s disease suitable for drug treatment. Br J Gen 26. Knopman DS, Ryberg S. A verbal memory test with high predictive accuracy for dementia of the Alzheimer type. Arch 2. Ross G, Abbott R, Petrovitch H, et al. Frequency and charac- teristics of silent dementia among elderly Japanese-American 27. Buschke H, Sliwinski M, Kuslansky G, Lipton RB. Aging, encoding specificity and memory changes in the Double Mem- 3. Kotler-Cope S, Camp C. Anosognosia in Alzheimer’s disease.
ory Test. J Int Neuropsychol Soc 1995;1:483– 493.
Alzheimer Dis Assoc Disord 1995;9:52–56.
28. Buschke H, Sliwinski M, Kuslansky G, Lipton RB. Diagnosis 4. Callahan C, Hendrie HC, Tierney WM. Documentation and of early dementia by the Double Memory Test: encoding spec- evaluation of cognitive impairment in elderly primary care ificity improves diagnostic sensitivity and specificity. Neurol- patients. Ann Intern Med 1995;122:422– 429.
5. Farlow M, Gracon S, Hershey L, Lewis K, Sadowsky C, Dolan- 29. Buschke H. Cued recall in amnesia. J Clin Neuropsychol Ureno J. A controlled trial of tacrine in patients with probable Alzheimer’s disease. JAMA 1992;268:2523–2529.
30. Tulving E, Thomson D. Encoding specificity and retrieval pro- 6. Raskind M, Sadowsky C, Sigmund W, Beitler P, Auster S.
cesses in episodic memory. Psychol Rev 1973;80:352–372.
Effects of tacrine on language, praxis, and noncognitive be- 31. Grober E, Buschke H, Crystal H, Bang S, Dresner R. Screen- havioral problems in Alzheimer’s disease. Arch Neurol 1997; ing for dementia by memory testing. Neurology 1988;38:900 – 7. Gelmacher D. Donepezil (Aricept) therapy for Alzheimer’s dis- 32. American Psychiatric Association. Diagnostic and statistical ease. Compr Ther 1997;23:492– 493.
manual of mental disorders. 3rd ed, revised. Washington, DC: 8. Schordert M. Alzheimer’s disease: fundamental and therapeu- American Psychiatric Association, 1987.
tic aspects. Experientia 1995;51:99 –105.
33. McKhann G, Drachman D, Folstein M, Katzman R, Price D, 9. Van Reekum R, Black S, Conn D, Clarke D. Cognition- Stadlan E. Clinical diagnosis of Alzheimer’s disease: report of enhancing drugs in dementia: a guide to the near future. Can the NINCDS-ADRDA Work Group under the auspices of De- J Psychiatry 1997;42(suppl 1):35s–50s.
partment of Health and Human Services Task Force on Alz- 10. Khachatorian Z, Phelps C, Buckholtz N. The prospect of devel- heimer’s Disease. Neurology 1984;34:939 –944.
oping treatments for Alzheimer’s disease. In: Terry R, Katz- 34. Wechsler D. Wechsler Adult Intelligence Scale–Revised. New man R, Bick K, eds. Alzheimer’s disease. New York: Raven York: The Psychological Corporation, 1981.
35. Wechsler D. Wechsler Memory Scale–Revised. San Antonio, 11. Gauthier S, Thal L, Rosser M. The future diagnosis and treat- TX: The Psychological Corporation, 1987.
ment of Alzheimer’s disease. In: Gauthier S, ed. Clinical diag- 36. Buschke H. Criteria for identification of memory deficits: im- nosis and management of Alzheimer’s disease. London: plications for the design of memory tests. In: Gorfein D, Hoff- man RR, eds. Memory and learning: the Ebbinghaus 12. Siu A. Screening for dementia and investigating its causes.
Centennial Conference. Hillsdale, NJ: Lawrence Erlbaum As- January (2 of 2) 1999
37. Zung W. A self-rating depression scale. Arch Gen Psychiatry receiver operating characteristic (ROC) curve. Radiology 1982; 38. Spitzer R, Williams J, Gibbon M, First M. Structured Clinical 42. Wilder D, Cross P, Chen J, et al. Operating characteristics of Interview for DSM-III-R. Nonpatient Edition (SCID-NP, ver- brief screens for dementia in a multicultural population. Am J sion 1). Washington, DC: American Psychiatric Press, 1990.
Geriatr Psychiatry 1995;3:96 –107.
39. Elwood R. Clinical discriminations and neuropsychological 43. Katzman R. Education and the prevalence of dementia and tests: an appeal to Bayes’ theorem. Clin Neuropsychologist Alzheimer’s disease. Neurology 1993;43:12–20.
44. Kittner SJ, White LR, Farmer ME, et al. Methodological is- 40. Meiran N, Stuss D, Guzman A, Lafleche G, Willmer J. Diag- sues in screening for dementia: the problem of education ad- nosis of dementia: methods for interpretation of scores of 5 justment. J Chron Dis 1986;39:163–170.
neuropsychological tests. Arch Neurol 1996;53:1043–1054.
45. Morris JC. The Clinical Dementia Rating (CDR): current ver- 41. Hanley J, McNeil B. The meaning and use of the area under a sion and scoring rules. Neurology 1993;43:2412–2414.
Tacrine use in nursing homes
Implications for prescribing new cholinesterase inhibitors
D.R. Gifford, MD, MPH; K.L. Lapane, PhD; G. Gambassi, MD; F. Landi, MD; V. Mor, PhD; and R. Bernabei, MD, for the SAGE Study Group* Article abstract—Objective: To describe the use of tacrine in nursing home residents using data from a clinically based
resident assessment instrument used by all US nursing homes. Methods: Data were from the Systematic Assessment of
Geriatric Drug Use via Epidemiology (SAGE) database, a population-based data set with information on 329,520 patients
admitted to all Medicare/Medicaid certified nursing homes in four US states (Maine, Mississippi, New York, and South
Dakota) from 1992 through 1995. The SAGE database combines information from the Minimum Data Set (MDS) and the
On-Line Survey and Certification Automated Record. We identified all residents receiving tacrine and up to five control
residents per case matched on state, date of tacrine use, cognitive function, and dementia diagnosis. Results: A total of
1,640 (0.5%) nursing home residents received tacrine at least once. Only 38% of these residents had a diagnosis of AD
documented on the MDS; regardless of dementia diagnosis, 25% had severe cognitive impairment, 35% were severely
dependent in activities of daily living (ADL), and 17% had both severe cognitive and ADL impairment. Only 8% achieved
a therapeutic dose of at least 120 mg/d. After adjusting for confounding variables, wandering and being physically abusive
were the strongest predictors of tacrine use. Conclusions: A minority of nursing home residents received tacrine. Of those
who did, a significant proportion were unlikely to benefit from its use because of their level of cognitive and ADL
impairment, or because low doses were used. As new medications become available for dementia, MDS data can be used
by nursing homes to monitor the use of these therapies.
The first cholinesterase inhibitor, tacrine, was ap- tory setting, patients with a confirmed diagnosis of proved in 1993 for the treatment of AD in the United AD who receive tacrine exhibit, on average, statisti- States.1 A second cholinesterase inhibitor, donepezil, cally significant improvement on cognitive function was approved in 1997.2 According to current guide- tests.6-9 Data from these trials and others suggest lines,3,4 cholinesterase inhibitors are only indicated that tacrine also may affect other outcomes,7,10-12 but for patients with a diagnosis of AD who have mild to moderate dementia (e.g., Mini-Mental State Exami- nation [MMSE] score of 10 to 26).1,2,5 In the ambula- Little is known about the pharmacologic manage- See also page 227
*See the Appendix on page 243 for a list of members of the SAGE Study Group.
From the Departments of Medicine (Dr. Gifford) and Community Health (Drs. Gifford, Lapane, Gambassi, and Mor) and the Center for Gerontology andHealth Care Research (Drs. Gifford, Lapane, Gambassi, and Mor), Brown University, Providence, RI; and Istituto di Medicina Interna e Geriatria (Drs.
Gambassi, Landi, and Bernabei), Universita` Cattolica del Sacro Cuore, Rome, Italy.
This study was conducted while D.R.G. was a Pfizer/American Geriatric Society Postdoctoral Fellow and SURDNA fellow at the Brown Center forGerontology and Health Care Research. The study was also supported in part by the “Do Good Nursing Homes Achieve Good Outcomes?” project (NationalInstitute on Aging #AG11624) and by the “Changing Roles in Nursing Homes” project (#500-89-0046; US Department of Health and Human Services, HealthCare Financing Administration, to University of Michigan with subcontract to Brown University).
Received May 18, 1998. Accepted in final form October 24, 1998.
Address correspondence and reprint requests to Dr. David R. Gifford, Center for Gerontology and Health Care Research, Brown University, Box G-B 222,Providence, RI 02912; e-mail: [email protected] Copyright 1999 by the American Academy of Neurology

Source: http://www.schwindel-sprechstunde.de/mis.pdf