BrainMeta'   Connectomics'  

Welcome Guest ( Log In | Register )

Reply to this topicStart new topic
> Alzheimer's Screening
post Nov 04, 2010, 07:46 PM
Post #1


Group: Basic Member
Posts: 1954
Joined: Oct 17, 2006
From: Bay area CA
Member No.: 5877

I am sure no one gives a shit about any of this, but I felt like posting nonetheless. Making the final touches before submitting for publication. Sorry, when I tried to paste the paper, all of the tables and charts got more or less destroyed. I will upload the paper somewhere so I can put a link.

Measuring Memory Performance in a Large Group Setting:
Test Characteristics

Peter J. Bayley, J. Wesson Ashford, Antony Guglielmone 


Memory impairment is often the most disabling feature of many pathological processes including neurodegenerative diseases, such as Alzheimer’s disease (AD), strokes, tumors, head trauma, hypoxia, hypoglycemia, certain vitamin deficiencies, and cardiac surgery (Newman et al, 2001, Mesulam, 2000 in Budson and Price, 2005, nejm). Despite the widespread understanding that memory impairments are a core symptom of such disorders, memory impairment often goes undiagnosed, and individuals often fail to recognize the significance of early cognitive symptoms of disease (Callahan et al., 1995, Ross et al, 1997, Sternberg et al., 2000, Valcour et al., 2000, Finkel et al., 2003, Ashford et al., 2007). Memory impairment associated with dementia is particularly common and epidemiological studies indicate that 5% to 15% of adults aged 70 and older exhibit signs of dementia much of which is undiagnosed (Hy et al, 2000, Neurology).

Several factors interfere with the detection of memory impairment associated with dementia, including a failure to screen, a failure to screen high-risk individuals, and a failure to use appropriate testing methods. With changes in the delivery of health care, physicians must work under strict time constraints, and many physicians do not routinely screen their patients for dementia (Lawrence et al., 2003). One solution to the failure to detect memory impairment as a result of dementia is to implement large-scale community memory screening. Numerous approaches have been advocated to screen for memory problems in the community (Ashford, 2008) and studies demonstrate that such programs can detect individuals who would otherwise remain undiagnosed (Lawrence et al., 2001, 2003; Crews et al., 2009). However, community screening programs present significant logistical problems. Currently available memory tests must be administered by a trained psychometrician in a one-to-one interaction in a confidential and quiet environment. Such tests are generally expensive to administer (especially to large numbers of individuals when specificity is low) and sometimes uncomfortable for the individuals taking the tests, leading to poor motivation for repeat testing.

Audience-based methods in which cognitive tests are administered to a large number of individuals simultaneously have not been widely developed as screening tools. However, such tests could be used to screen large groups of people for memory problems in order to identify high-risk individuals for further evaluation. Because memory impairment associated with dementia is commonly undiagnosed, a simple audience-based memory test designed to detect patients with early dementia would be particularly valuable.

A significant issue is how to assess the utility of an audience-based memory screening test to detect early dementia in older adults. Any population of older adults is likely to contain a diversity of memory impairments. However, AD is the most common form of dementia and is estimated to account for approximately two thirds of all dementia cases (Alzheimer’s Association, 2010). The initial symptom of AD is typically a prominent amnesia in which the core symptom is difficulty in encoding new information (Salmon and Bondi, 1999). Even when patients with early AD succeed in perceiving and immediately reproducing new information (i.e., repeating a series of words), many neuropsychological studies have shown that the encoded information is easily lost under conditions of delay, especially when there is interim distraction (Ashford & Jarvik, 1985; Teter & Ashford, 2002; Elias et al., 2000; Ashford et al., 1989; Ashford & Schmitt, 2001).

The process of memory encoding can be tested in several different ways. However, recognition memory tests are especially suitable for this purpose as they provide the target stimuli within the framework of the test. For this reason, poor performance on a test of recognition memory provides strong evidence for an underlying encoding impairment, raising the possibility of an emerging Alzheimer’s process (Lowndes and Savage, 2007). In contrast to individuals with early AD, healthy adults can quickly and accurately encode massive amounts of new information. For example, landmark studies in the 1960s and 1970s demonstrated that after viewing thousands of scenes for a few seconds each, healthy individuals perform well above chance on tests of recognition memory (Shepard, 1967; Standing 1973). Taken together with the encoding deficits found in early AD, these results suggest that recognition memory tests which are well within the memory capacity of healthy individuals should provide an effective screen for early AD. It is important to note that even though recognition memory has a huge capacity in healthy adults, it is nonetheless vulnerable to age-associated cognitive decline and increased age is associated with lower levels of performance (Schacter et al., 1992; Grady et al., 1995).
The aim of the current study was to measure AD-related memory performance in an audience population. For this purpose a test was designed to detect the type of memory problem most specific to early AD using a format that was developed to assess memory during a formal presentation to groups of interested individuals (Ashford, 2008). The test was designed to be interesting to maintain audience attention. The current study sought to characterize this repeat detection task and evaluate age-related changes in recognition memory as measured by this task in order to determine normal performance ranges. Test performance was expected to decrease with age due to a combination of normal age-associated memory impairment and specific memory problems possibly related to dementia.


The test was administered to over 1050 subjects between July, 2007 and June, 2008 at 26 sites (community events, senior citizen centers, retirement living communities, etc. in San Francisco Bay Area). The audiences ranged in size from 9 to 142 individuals (average group size = 39; SD 34; range 9-142). Of the subjects who provided test forms that were appropriately marked, there were 940 adults who ranged in age from 40 to 97 years old. Of these 940 participants, 868 individuals provided three specific demographic items of information: age, education and gender. Thus 868 individuals were included in the data analysis and had a mean age of 75.9 years (SD 11.4; range 40.0 – 97.6) and mean education of 16.1 years (SD 2.52; range 6–21). For this group, 68.7% were female and 86.6% reported being “white”. Participants were divided into six sub-groups according to age as shown in Table 1. Education level declined by 1.3 years from 16.9 to 15.6 from the youngest age-group to the oldest age group, though the variation did not reach statistical significance (F(5, 867) = 1.93, p > .05). All age groups contained more females than males (see Table 1). The groups varied significantly in the number of males to females (χ2 (N = 868) = 12.9, p = .02).


An audience-based memory test was developed considering the concept of recognition of a large number of easily remembered images. A “multiple-N-back” (or repeat detection) format was used with numerous complex visual stimuli. Generally the images were of discrete objects, though similar objects and difficult to name objects were used to avoid strict reliance on verbal cues and to provide a challenge and maintain the interest of the subjects (the assortment of images was developed over several years). This approach also reduced the ceiling and floor effects. Audience testing procedures are unusual in cognitive neuroscience research, so a primary aim of the study was to provide proof of concept that a recognition memory test can be successfully administered to a large number of individuals simultaneously.
Twenty-five color images (digital camera) of manmade items were selected from a wide range of pictures. From these 25 items, a 50-item recognition memory test was constructed in the following way. The 25 items were first arranged in a random sequence, with repeated images interspersed. Fourteen of the items were first-time repeats and were inserted among the initial presentations of the test items. Eleven of the items were shown for a third time, making recognition easier for subjects with impaired memories, providing more learning regarding a particular stimulus set, and allowing a comparison of first repeat recognition with second repeat recognition. The order was arranged such that there was an average inter-repetition-interval from the initial presentation to the first repeat of 7.93 items, ranging from 2 to 25 intervening items. The eleven items that were second repeats were inserted into the test with an average inter-repetition interval of 21.1, ranging from 10 to 36 items between the second and third presentations. The eleven un-repeated test items served as foils.

The 50 items were numbered in sequence (1-50) with a large numeral in the top left hand corner and transferred to a PowerPoint presentation. A second series of ten items was constructed using similar color images and was used as a practice test before the full test was given (5 images, 3 repeated once, 2 repeated a second time). The need for such a practice test had become obvious during pilot work, which indicated that about 10% of audience members could not follow the verbal instructions.

Participants were provided with a single sheet of paper. Demographic information was collected on one side of the page (age, education, race) and the other side was used as an answer sheet for the recognition memory testing. The answer sheet had columns of numbers corresponding to the 10 slides of the practice-test and the 50 slides of the full test. A single circle was adjacent to each number on which a subject could indicate their response by filling in the circle, and the sheet was organized so that it could be scanned for data entry.

All sites adhered to a standard format, which began with a 20-minute introductory talk, with slides about Alzheimer’s disease and the signs of dementia. As part of the talk, all participants were offered a memory test, and the audience was told that participating in the memory test was optional, but that individual test scores would be provided at the end of the presentation. A statement outlining the subjects’ rights was provided to all audience members on a written page and reviewed on a slide (Protocol approved by Stanford University Institutional Review Board; no identifying information was collected, and therefore written consent was not required). The same 10-item practice test and 50-item memory test were used at all sites (2 individuals acknowledged taking the test before, but were not identified).

The repeat-detection test was presented by projecting test items onto a screen using a laptop computer and projector. No effort was made to assess visual acuity of audience members or to assure adequate visibility from all parts of the room. However, the slides were generally easily seen from all vantage points of every room in which the test was administered.
The memory test was given in a repeat-detection format. Participants were told that they would see a series of 50 pictures one at a time for 5s per image (no inter-image interval). They were instructed to look at each picture carefully and any time they thought an image was repeated they should note the image number shown in the top left hand corner and mark the circle corresponding to that number on their answer sheet. No response was required if they thought an image was not repeated (i.e., novel). The 10-item practice test was given first. The presenter then addressed any questions relating to the test procedure, and then the full 50-slide test was given (250s). After the test, the participants handed their papers to the rater to be scored. A rater scored each participant’s answer sheet, after which the scores were returned anonymously to each participant. If scores indicated a high probability of memory problems, a notation was made on the anonymous score sheet encouraging the subject to visit their primary care clinician for further evaluation (note that about 50% of such positive screens might be expected to accept such a referral – Boustani et al., 2005).

Data analysis

Results from the recognition test were analyzed for number of two types of errors, missed items (recognition failures) and false positives. Correct recognitions (hits = total repeated items - # missed items) and false positives were used to determine a signal detection parameter, discriminability score (d′) (Green & Swets, 1966). A standard correction was sometimes necessary when calculating d′ values, as the percent hits and the percent false positives were sometimes 100% or 0%. Following MacMillan and Creelman, we converted 0% to 1/(2N)% and 100% to 1 − 1/(2N)% where N = the number of trials. As described above, eleven target items were presented three times during the recognition memory test in order to minimize the possibility that participants would base their responses on how they had responded to similar items that had been presented earlier in the test. The scores for the test were based on the responses to the first occurrence of these 11 targets, the responses to the three items shown once, and on the response to the 11 foils.
A three-way univariate analysis of variance (ANOVA) was used to examine the effect of age, education, and gender on errors and d’ scores. In order to determine the effects of age on test performance, participants were divided into six groups as shown in Table 1. To determine the effects of education on test performance, participants were divided into five groups corresponding with the major divisions of attainment in the U.S. educational system represented by the audiences [i.e., < 12 years (high school), 13-15 years (some college), 16 years (college completion), 17-19 years (masters degree), and 20-21 years(advanced degree)]. Significant effects were investigated using the Tukey Studentized Range / HSD post hoc test procedure to identify homogenous subgroups.

The 3-way ANOVA revealed significant main effects of age [F(5,811) = 12.97, p< .001] and education [F(4,811) = 5.46, p< .001] on recognition memory test scores. No significant effects were found for gender (F=.62). No significant interactions were found between the three factors age, education, and gender (all F’s<1.2, all p’s >.05). For post-hoc test analysis, homogenous subsets were examined, six age groups by decade from 40 to 99 and the 5 education groups described above. Additional separate analyses for age were done for education over 12 years since there was no significant education effect noted in this range.
Results indicated that the recognition memory test was sensitive to age and was more difficult for older adults than younger adults (Figure 1a). Test discriminability gradually declined numerically with increasing age. Age-associated effects were investigated further by examining the error rates: the missed items and the false positive rates that contributed to the overall d’ scores. Increased age was associated with significant increases in both the miss rate [F(5,867) = 14.10, p <.001] and the false alarm rate [F (5, 867) = 13.96, p <.001] (Figure 1b).
Although the six age groups did not differ significantly in number of years of education (see Participant section), it was noted that the oldest group also had numerically the lowest average level of education. In order to verify that the effect of age on test performance was not confounded by educational level, participants in the group with the lowest level of education (i.e., < 12 yrs of education, n=82) were excluded in a secondary analysis. Results were essentially the same as when all participants were included. These data strongly indicate that test discriminability declined significantly with increasing age [F(5,785) = 26.20, p<.001]. Again, increased age was associated with significant increases in both the miss rate [F(5,785) = 11.48, p <.001] and the false alarm rate [F (5, 785) = 13.30, p <.001].
The post-hoc subsets were further analyzed with the post-hoc Tukey tests, which is automatically corrected for multiple comparisons. This analysis did not support a significant decline in discriminability between age groups from 40 to 89 but the group older than 90 was significantly worse. Further, the miss rate was not statistically different within the groups with age range 40-89 yrs, but also showed a significant increase in the age group 90-99 yrs. False alarm rates showed a similar pattern and were homogenous within each of the following age ranges: 40-69 yrs, 50-79 yrs, 60-89 yrs, and 90-99 yrs. When the individuals with education less than or equal to 12 years were removed, the post-hoc tests showed that test performance was similar across the age ranges 40-59 yrs, 50-69 yrs, 60-79 yrs, 70-89 yrs, and 90-99 yrs. The miss rates showed the same pattern as for the previous analysis and were not statistically different within the age range 40-89 yrs, but showed a significant increase relative to the age range 90-99 yrs. False positive rates showed a similar pattern both to the previous analysis and to miss rates and were homogenous within each of the following age ranges: 40-69 yrs, 50-79 yrs, 60-89 yrs, and 80-99 yrs.
The effect of education on test performance is shown in Figure 2. Test performance was lower for those with education levels of 12 years or less relative to those with more education, however, performance reached a plateau after 12 years of education above which no significant improvement in performance was seen. Post-hoc tests showed that test performance was homogenous within two educational ranges: <12 yrs of education and 13-21yrs of education (the same pattern of results was found when participants were grouped into four groups rather than five groups on the basis of years of education). A one-way ANOVA confirmed that the mean age did not vary significantly across the five education groups [F(4,867) = 2.15, p> .05]. However, the lowest educational group (<12 yrs) was also numerically the oldest (79.5 yrs old vs. group mean of 76.4 yrs old for those with over 12 yrs of education), suggesting that levels of education vary systematically with age, and the poorer performance of the lower education group may actually be due to an age effect. In order to demonstrate that the effects of education on test score were associated with low levels education (i.e., <12 yrs of education), the analysis was repeated using only individuals having more than 12 years of education. This ANOVA showed that when individuals with low education were excluded, no significant effect of education on test performance were found [F(3, 785)= 1.65, p>.05].
Due to the repeat-detection format of the test, participants were required to hold items in memory for a variable delay. The inter-repetition-interval ranged from 2 to 25. This delay could disrupt recognition performance in two ways. First, as the number of intervening items increased, the time delay between the first and subsequent presentations of the same item could reduce recognition. Second, as other test items were presented during the time delay, interference could build up across the delay. To explore these effects, a regression analysis was performed between the number of intervening items and percent correct. No significant relationship was found between the number of intervening items and recognition performance (R = .12, F(1,7) = .10, p>.05) (Figure 3). As shown in Figure 3, the inter-repetition-interval had little overall effect on recognition and performance was maintained at a high level across repeated items (average = 89%). Another issue related to the repeat-detection format is that when test items are repeated multiple times, each subsequent presentation could serve as a retrieval cue to reactivate and strengthen the memory representation of the information stored during earlier study (S. Thios, P. D'Agostino, J. Verb. Learn. Verb. Beh. 15, 529, 1976). In the current test, eleven items were shown three times, and recognition performance would be expected to increase across these presentations. Indeed, this effect was observed and is shown in Figure 4. A paired t-test compared the mean percent correct between the first and second repetitions and showed that this difference (91.6% vs. 95.5% correct) was statistically significant (t(867) = -10.30, p < .005).


Table 1. Demographic characteristics of participants.

Age group 40-49
60-69 70-79 80-89 90-99
N 29 68 135 239 359 38
Gender (% F) 76 82 76 65 66 63
M (SD) M (SD) M (SD) M (SD) M (SD) M (SD)
Age (yrs) 44.8 3.1 55.3 2.8 65.8 2.9 75.8 2.9 84.4 2.7 92.1 1.6
Education (yrs) 16.9 2.0 15.9 2.3 16.4 2.3 15.9 2.7 16.1 2.5 15.6 2.6


Alzheimer's Association. 2010 Alzheimer's Disease Facts and Figures. Alzheimer's & Dementia: The Journal of the Alzheimer's Association. March 2010; 6(2):158-194.

Ashford JW, Borson S, O'Hara R, Dash P, Frank L, Robert P, Shankle WR, Tierney MC, Brodaty H, Schmitt FA, Kraemer HC, Buschke H. Fillit H. Should older adults be screened for dementia? It is important to screen for evidence of dementia. Alzheimer?s & Dementia. (2007) April;3:75-80.

Implementing a Screening and Diagnosis Program for Dementia in Primary Care
Malaz Boustani, MD, MPH,1,2,3 Christopher M. Callahan, JGIM, 2005

Raber J, Huang Y, Ashford JW. ApoE genotype accounts for the vast majority of AD risk and AD pathology. Neurobiol Aging 2004;25:641–650. [PubMed: 15172743]

Callahan CM, Hendrie HC, Tierney WM. Documentation and evaluation of cognitive impairment in elderly primary care patients. Ann Intern Med 1995;122:422–429. [PubMed: 7856990]

Ross GW, Abbott RD, Petrovitch H, Masaki KH, Murdaugh C, Trockman C, et al. Frequency and characteristics of silent dementia among elderly Japanese-American men: the Honolulu-Asia Aging Study. JAMA 1997;277:800–805. [PubMed: 9052709]

Sternberg SA, Wolfson C, Baumgarten M. Undetected dementia in community-dwelling older people: the Canadian Study of Health and Aging. J Am Geriatr Soc 2000;48:1430–1434. [PubMed: 11083319]

Valcour VG, Masaki KH, Curb JD, Blanchette PL. The detection of dementia in the primary care
setting. Arch Intern Med 2000;160:2964–2968. [PubMed: 11041904]

Finkel SI. Cognitive screening in the primary care setting: the role of physicians at the first point of entry. Geriatrics 2003;58:43–44. [PubMed: 12813873] Freund B. Office-based evaluation of the older driver. J Am Geriatr Soc 2006;54:1943–1944. [PubMed: 17198503]



Fig. 1 A. The relationship between discriminability performance (d′) and age in 868 individuals on a multiple-n-back recognition memory test. Numbers inside the bars indicate the group n. The bars show the mean discriminability score for each age group and brackets show SEM. Participants studied a series of 50 objects and indicated when an item was repeated. The test consisted of 50 trials: the 14 targets (some of which appeared 2 or 3 times) and 11 unique foils. B. The mean miss rate and false alarm rate on the multiple-n-back recognition memory test for each age group. The brackets show SEM.

Fig. 2 The relationship between Discriminability performance (d′) and education in 868 individuals on a multiple-n-back recognition memory test. Numbers inside the bars indicate the group n. Participants studied a series of 50 objects and indicated when an item was repeated. The test consisted of 50 trials: the 14 targets (some of which appeared 2 or 3 times) and 11 unique foils.

Fig. 3 The relationship between the number of intervening items (between initial and first repeat presentations) and percent correct in 868 individuals on a multiple-n-back recognition memory test. The bars show the mean discriminability score for each age group and brackets show SEM.

Fig. 4 The relationship between percent correct and item repetition in 868 individuals on a multiple-n-back recognition memory test. Eleven items were shown three times during the test, and recognition performance increased between the first and second repetition. A paired t-test demonstrated that the difference was significant (p < .005).
User is offlineProfile CardPM
Go to the top of the page
+Quote Post

Reply to this topicStart new topic
1 User(s) are reading this topic (1 Guests and 0 Anonymous Users)
0 Members:


Lo-Fi Version Time is now: 22nd October 2018 - 05:22 AM

Home     |     About     |    Research     |    Forum     |    Feedback  

Copyright © BrainMeta. All rights reserved.
Terms of Use  |  Last Modified Tue Jan 17 2006 12:39 am

BrainMeta is supported by donors of The Neurological Foundation. Donate today to help support us.