- Email: [email protected]
Screening for pre-dementia Alzheimer’s disease
S88
Abstracts: Screening I / 1 (Suppl 1) (2005)
minute resting-state fMRI scans. AD and FTLD subjects were matched on age and mini-mental state examination scores. Healthy controls were slightly older than either the FTLD or AD patients. One AD subject could only complete a single 6-minute scan and one healthy control only had one usable scan. Preprocessing was done with SPM2. The smoothed, normalized images were subjected to independent component analysis. As in our previous study (Greicius et al., PNAS, 2004), an automated process was used to assign a goodness-of-fit score to each component reflecting how well the components matched a standard template of the default-mode network. The best-fit component was chosen as the default-mode component. Each subject’s highest best-fit score (one from each of their two scans) was used as a dependent variable in an ANOVA and as a candidate biomarker. Results: The ANOVA showed a main effect of diagnosis and post-hoc tests showed that AD scores were significantly less than those of healthy controls and the FTLD group (p ⬍ 0.05). Scores did not differ signficantly between the control and FTLD groups. Using a score of 2.7 as a cutoff, this method correctly classified 8/9 AD patients, 7/7 healthy controls, and 4/5 FTLD patients yielding a sensitivity of 89% in detecting AD and 100% specificity in distinguishing AD from healthy aging and 80% specificity in distinguishing AD from FTLD. Conclusions: These preliminary data suggest that resting-state fMRI merits wider-scale testing as a clinically useful biomarker of AD.
subjects and AD patients; (3) To explore the correlation between hypoperfusion and decline of cognitive functions. Methods: Twenty-one FTD patients (62⫾7 yrs), 24 AD patients (64 ⫾7 yrs), and 25 CN (62⫾7 yrs) subjects were studied with ASL-MRI and volumetric T1-weighted structural MRI at 1.5 Tesla. Group effects on perfusion and the correlation between perfusion and cognitive function were tested voxel-wise using Statistical Parametric Mapping (SPM2). Group classification by perfusion and GM density was tested by a logistic regression. FTD patients showed hypoperfusion in bilaterally frontal lobe and anterior cingulate than CN subjects and in right frontal lobe than AD patients. On the other hand, AD patients had hypoperfusion in bilaterally parietal lobe and precuneus and left posterior cingulate than FTD patients. Overall classification reached 67% with perfusion and 60% with GM density between FTD and CN, and 78% with perfusion and 62% with GM density between FTD and AD. Furthermore, perfusion achieved a significantly better classification than GM density for separating FTD from CN and AD. In addition, hypoperfusion was positively correlated with cognitive decline, including language, spatial memory and executive function. Conclusions: ASL-MRI in combination with structural MRI may help a differential diagnosis between FTD and AD. Furthermore, ASL-MRI may be more sensitive than structural MRI in detecting FTD and AD pathology and thus, could be useful for staging early dementia and monitoring disease progression.
SUNDAY, JUNE 19, 2005 ORAL SESSION O1-05 SCREENING I O1-05-01
STATISTICAL ISSUES FOR DEVELOPING ALZHEIMER SCREENING TESTS
J. W. Ashford1, Helena C. Kraemer2; 1Stanford / VA Alzheimer Center, Palo Alto, CA, USA; 2Stanford University, Palo Alto, CA, USA
O1-04-08
DIFFERENT PATTERNS OF HYPOPERFUSION IN FRONTOTEMPORAL DEMENTIA AND ALZHEIMER’S DISEASE BY ARTERIAL SPIN LABELING MRI
Antao Du1, Geon-Ho Jahng1, Bruce Miller2, Satoru Hayasaka1, Howie Rosen2, Joel H. Kramer2, Michael W. Weiner1, Norbert Schuff1; 1 VA Medical Center, San Francisco, San Francisco, CA, USA; 2 University of California, San Francisco, San Francisco, CA, USA Background: A differential clinical diagnosis between Frontotemporal dementia (FTD) and Alzheimer’s disease (AD) is sometimes difficult to make, especially at an early stage because of overlapping symptoms. Functional studies with PET and SPECT reported characteristic patterns of hypoperfusion in FTD and AD that might aid a differential diagnosis. However, PET and SPECT are not wide available and require injection of radioactive tracers. In contrast, arterial spin labeling MRI (ASL-MRI) provides an entirely non-invasive approach to measure cerebral perfusion. Objective(s): (1) To test if ASL-MRI detects different regional patterns of hypoperfusion between FTD and AD similar to results from PET and SPECT studies; 2) To determine the extent to which perfusion and gray matter (GM) loss classify FTD patients from cognately normal (CN)
Background: What difference does it make that a test is to be used for screening, rather than for discrimination or definitive diagnosis? What difference does it make that a test is to be used in a general medical practice rather than in a specialty clinic? Such considerations must affect the statistical evaluation of tests, but are frequently ignored. Objectives: First, the continuum of pathology which is the target of the screen must be defined. This leads to selecting a “gold standard” for comparison, and its fallibility must also be considered. Other issues considered in the statistical evaluation of such screening tests are the prevalence in the population, the fixed cost of the test, the cost of false positives, the cost of false negatives, and the benefit of true positives. Methods: Item Response Theory and classification algorithms will be discussed. Optimally effective items for screening should be selective. Costs and benefits need to be analyzed with respect to the incidence of dementia for the age of the subject being tested. Conclusions: A screening test for Alzheimer’s disease can be constructed and evaluated, but several issues must be addressed in the process. Further, iterative improvements of tests should be made. O1-05-02
SCREENING FOR PRE-DEMENTIA ALZHEIMER’S DISEASE
Herman Buschke; Albert Einstein College of Medicine, New York, NY, USA Background: If the criterion for memory impairment is low recall, memory impairment can only be detected when recall is low, and earlier memory impairment when memory is declining but is still in the normal range cannot be detected. Additional indicators of memory impairment are needed to detect early pre-dementia Alzheimer’s disease when declining memory is still in the normal range and treatment may be more effective. Objective(s): Screen for pre-dementia AD when declining memory is still in the normal range. Methods: Controlled learning and cued recall of two coordinated lists was used to to detect earlier memory impairment in a community-dwelling sample of
Abstracts: Screening I / 1 (Suppl 1) (2005) aged by lower recall of the 2nd list even when recall of the 1st list was still within normal limits. The 1st list included 16 items each from a different category. The 2nd list included 16 different items from the same categories. Recall of the 2nd list was assessed as percent of each person’s own 1st list recall. All aged with dementia had 2nd list recall that was 50 % or less than their own 1st list recall. Only about half of the aged with MCI had such low 2nd list recall. About 13% of the aged with normal 1st list recall had such low 2nd list recall. Aged with 2nd list recall that was 50% or less of their 1st list recall had lower scores on other cognitive tests and worse Blessed IMC mental status scores. Conclusions: Memory impairment shown by low 2nd list recall in only some but not in all aged with MCI should improve MCI drug trials by discriminating aged with memory impairment likely to be due to AD from those with life-long low memory. Memory impairment shown by low 2nd list recall in aged who still have normal normal 1st list recall should make possible earlier drug trials and treatment to preserve cognitive function by detecting memory impairment when declining memory is still in the normal range. O1-05-03
IMPLEMENTATION OF UNIVERSAL COGNITIVE SCREENING IN PRIMARY CARE: INITIAL RESULTS
Soo Borson1, James M. Scanlan1, Jeffrey Hummel1,2, Kathy Gibbs3; 1 University of Washington, Seattle, WA, USA; 2UW Physicians Network, Seattle, WA, USA; 3University of Washington Physicians Network, Seattle, WA, USA Background: Primary care doctors are well positioned for early detection of dementia, but rarely screen their patients and miss many, especially, milder cases. Time, complexity, and workload are important barriers to early detection of dementia in primary care offices. Simple, short cognitive screens that can be given by non-physicians could improve initial detection rates if adopted in the primary care setting. Objective: To test the feasibility of implementing cognitive screening of older adults in two primary care practices, using a simple 2-minute screen, the Mini-Cog, developed for this use. Methods: Medical assistants in two primary care clinics were trained to administer the Mini-Cog to all older adults during a routine primary care visit and to inform the physician of results. Two clinics in the same network were used for ‘usual practice’ comparison. Outcomes were percent of eligible patients screened and percents screening positive for cognitive impairment, diagnosed with dementia, prescribed any dementia medication, or referred for specialized dementia assessment. Changes in the intervention clinics were compared with changes occurring in the control clinics. Results: Medical assistants screened 70% of eligible patients during the intervention period (n⫽524, mean age 75); patient acceptance was very high. 26 MAs were trained in the screening procedure, achieving 96% concordance with expert raters in Mini-Cog scoring. Sixteen percent of screened patients scored in the impaired range (Mini-Cog 0-2), consistent with epidemiological data, but dementia diagnoses were made in only 8% of the population. Relative to the baseline pre-intervention period and to control clinics, the screening program resulted in significant increases in dementia diagnoses and specialist referrals, and a less robust increase in treatment rates. Conclusions: Simple cognitive screening can be reliably implemented in primary care with a high rate of patient acceptance. With minimal training and feedback, medical assistants screened patients at a high level of competence. While the impact of screening on physician recognition of dementia was positive, optimal rates of diagnosis and treatment will require additional interventions in primary care. O1-05-04
WHAT IS THE BEST DEMENTIA SCREENING INSTRUMENT FOR GENERAL PRACTITIONERS TO USE?
Henry Brodaty1, Lee-Fay Low2, Louisa Gibson3, Kim Burns3; 1 University of New South Wales, Randwick, New South Wales, Australia; 2Australian National University, Canberra, ACT, Australia; 3 Prince of Wales Hospital, Randwick, New South Wales, Australia Background: The delay of 8-32 months between symptom onset and a diagnosis of dementia suggests that greater attention to diagnosing demen-
S89
tia in its mild stages is necessary. Earlier identification may enable earlier treatment and empower individuals to plan for their future sooner. General practitioners (GPs), although best situated to detect dementia early, display limited capacity to achieve this. Only 26-39% of GPs screen for dementia routinely, and it is estimated that routine screening could double the number of cases GPs identify. General practitioners report the lengthy nature and inadequacies of current screening tools such as the Mini-mental state examination (MMSE) as reasons for their failure to screen. Objectives: We reviewed existing dementia screening tools with a view to informing and recommending suitable instruments to GPs based on their performance and practicability for general practice. Methods: Pre-Medline, Medline, PsycINFO, and the Cochrane Library databases were searched, with 16 instruments included in the final review. The General Practitioner Assessment of Cognition (GPCOG), Mini-Cog, and Memory Impairment Screen (MIS) were chosen as most suitable for routine screening. The GPCOG, Mini-Cog, and MIS were all validated in community, population, or general practice samples, are easy to administer, and have administration times under five minutes. These instruments have negative predictive validities and misclassification rates at least as good if not better than the MMSE, but are quicker and easier to perform. Conclusion: It is recommend that GPs adopt routine screening using the GPCOG, Mini-Cog, or MIS when patients are over the age of 75 years or when cognitive impairment is suspected.
O1-05-05
A BRIEF ALZHEIMER’S SCREEN FOR USE IN CLINICAL PRACTICE
Frederick A. Schmitt1, Marta S. Mendiondo1, Richard J. Kryscio1, John W. Ashford2; 1University of Kentucky, Lexington, KY, USA; 2 Stanford/VA Alzheimer Center, Palo Alto, CA, USA Background: With advances in the treatment of Alzheimer’s disease (AD), clinical focus has shifted to early patient identification. Therefore screening procedures have gained renewed popularity in clinical practice. Recall and fluency tests have been able to distinguish normal individuals from early AD patients. We evaluated a Brief Alzheimer Screen (BAS) derived from item response procedures in a longitudinal cohort of normal elderly, persons with MCI, and mild AD. Objective(s): Evaluate a brief mental status test for general practitioners to screen for MCI and AD. Methods: We previously reported on the BAS using the cognitive assessments from 406 normal subjects and 342 mild AD patients in the CERAD dataset and then applied it to a second validation sample. Logistic regression was used to derive estimates of sensitivity and specificity. The resulting logistic model for discriminating between mild AD and controls included: recall of 3 words, number of animals named in 30 seconds, date, and spelling of WORLD backwards with sensitivity and specificity over 99% and 87%, respectively. The derived logistical model was then applied to a larger sample (N⫽971) of participants followed in our AD Center’s longitudinal cohorts. Three groups were compared using the BAS: normal, MCI, and mild AD defined by a CDR of 0.5 or 1. Analyses showed excellent discrimination between these clinically diagnosed groups (see Table) with significant separation using the BAS weighted score. GROUP (N)
Mean
Median
SEM
Normal (318) MCI (178) Mild AD (475)
33.60 30.62 18.58
33.27 30.75 17.42
0.69 0.28 0.32
Conclusions: The use of item response approaches can help identify items that can improve discrimination between clinical groups when developing screening measures. In addition, simple screening approaches should be designed and evaluated in order to improve detection and promote early treatment in busy medical practice settings where at risk older adults receive their care.
Abstracts: Screening I / 1 (Suppl 1) (2005)
minute resting-state fMRI scans. AD and FTLD subjects were matched on age and mini-mental state examination scores. Healthy controls were slightly older than either the FTLD or AD patients. One AD subject could only complete a single 6-minute scan and one healthy control only had one usable scan. Preprocessing was done with SPM2. The smoothed, normalized images were subjected to independent component analysis. As in our previous study (Greicius et al., PNAS, 2004), an automated process was used to assign a goodness-of-fit score to each component reflecting how well the components matched a standard template of the default-mode network. The best-fit component was chosen as the default-mode component. Each subject’s highest best-fit score (one from each of their two scans) was used as a dependent variable in an ANOVA and as a candidate biomarker. Results: The ANOVA showed a main effect of diagnosis and post-hoc tests showed that AD scores were significantly less than those of healthy controls and the FTLD group (p ⬍ 0.05). Scores did not differ signficantly between the control and FTLD groups. Using a score of 2.7 as a cutoff, this method correctly classified 8/9 AD patients, 7/7 healthy controls, and 4/5 FTLD patients yielding a sensitivity of 89% in detecting AD and 100% specificity in distinguishing AD from healthy aging and 80% specificity in distinguishing AD from FTLD. Conclusions: These preliminary data suggest that resting-state fMRI merits wider-scale testing as a clinically useful biomarker of AD.
subjects and AD patients; (3) To explore the correlation between hypoperfusion and decline of cognitive functions. Methods: Twenty-one FTD patients (62⫾7 yrs), 24 AD patients (64 ⫾7 yrs), and 25 CN (62⫾7 yrs) subjects were studied with ASL-MRI and volumetric T1-weighted structural MRI at 1.5 Tesla. Group effects on perfusion and the correlation between perfusion and cognitive function were tested voxel-wise using Statistical Parametric Mapping (SPM2). Group classification by perfusion and GM density was tested by a logistic regression. FTD patients showed hypoperfusion in bilaterally frontal lobe and anterior cingulate than CN subjects and in right frontal lobe than AD patients. On the other hand, AD patients had hypoperfusion in bilaterally parietal lobe and precuneus and left posterior cingulate than FTD patients. Overall classification reached 67% with perfusion and 60% with GM density between FTD and CN, and 78% with perfusion and 62% with GM density between FTD and AD. Furthermore, perfusion achieved a significantly better classification than GM density for separating FTD from CN and AD. In addition, hypoperfusion was positively correlated with cognitive decline, including language, spatial memory and executive function. Conclusions: ASL-MRI in combination with structural MRI may help a differential diagnosis between FTD and AD. Furthermore, ASL-MRI may be more sensitive than structural MRI in detecting FTD and AD pathology and thus, could be useful for staging early dementia and monitoring disease progression.
SUNDAY, JUNE 19, 2005 ORAL SESSION O1-05 SCREENING I O1-05-01
STATISTICAL ISSUES FOR DEVELOPING ALZHEIMER SCREENING TESTS
J. W. Ashford1, Helena C. Kraemer2; 1Stanford / VA Alzheimer Center, Palo Alto, CA, USA; 2Stanford University, Palo Alto, CA, USA
O1-04-08
DIFFERENT PATTERNS OF HYPOPERFUSION IN FRONTOTEMPORAL DEMENTIA AND ALZHEIMER’S DISEASE BY ARTERIAL SPIN LABELING MRI
Antao Du1, Geon-Ho Jahng1, Bruce Miller2, Satoru Hayasaka1, Howie Rosen2, Joel H. Kramer2, Michael W. Weiner1, Norbert Schuff1; 1 VA Medical Center, San Francisco, San Francisco, CA, USA; 2 University of California, San Francisco, San Francisco, CA, USA Background: A differential clinical diagnosis between Frontotemporal dementia (FTD) and Alzheimer’s disease (AD) is sometimes difficult to make, especially at an early stage because of overlapping symptoms. Functional studies with PET and SPECT reported characteristic patterns of hypoperfusion in FTD and AD that might aid a differential diagnosis. However, PET and SPECT are not wide available and require injection of radioactive tracers. In contrast, arterial spin labeling MRI (ASL-MRI) provides an entirely non-invasive approach to measure cerebral perfusion. Objective(s): (1) To test if ASL-MRI detects different regional patterns of hypoperfusion between FTD and AD similar to results from PET and SPECT studies; 2) To determine the extent to which perfusion and gray matter (GM) loss classify FTD patients from cognately normal (CN)
Background: What difference does it make that a test is to be used for screening, rather than for discrimination or definitive diagnosis? What difference does it make that a test is to be used in a general medical practice rather than in a specialty clinic? Such considerations must affect the statistical evaluation of tests, but are frequently ignored. Objectives: First, the continuum of pathology which is the target of the screen must be defined. This leads to selecting a “gold standard” for comparison, and its fallibility must also be considered. Other issues considered in the statistical evaluation of such screening tests are the prevalence in the population, the fixed cost of the test, the cost of false positives, the cost of false negatives, and the benefit of true positives. Methods: Item Response Theory and classification algorithms will be discussed. Optimally effective items for screening should be selective. Costs and benefits need to be analyzed with respect to the incidence of dementia for the age of the subject being tested. Conclusions: A screening test for Alzheimer’s disease can be constructed and evaluated, but several issues must be addressed in the process. Further, iterative improvements of tests should be made. O1-05-02
SCREENING FOR PRE-DEMENTIA ALZHEIMER’S DISEASE
Herman Buschke; Albert Einstein College of Medicine, New York, NY, USA Background: If the criterion for memory impairment is low recall, memory impairment can only be detected when recall is low, and earlier memory impairment when memory is declining but is still in the normal range cannot be detected. Additional indicators of memory impairment are needed to detect early pre-dementia Alzheimer’s disease when declining memory is still in the normal range and treatment may be more effective. Objective(s): Screen for pre-dementia AD when declining memory is still in the normal range. Methods: Controlled learning and cued recall of two coordinated lists was used to to detect earlier memory impairment in a community-dwelling sample of
Abstracts: Screening I / 1 (Suppl 1) (2005) aged by lower recall of the 2nd list even when recall of the 1st list was still within normal limits. The 1st list included 16 items each from a different category. The 2nd list included 16 different items from the same categories. Recall of the 2nd list was assessed as percent of each person’s own 1st list recall. All aged with dementia had 2nd list recall that was 50 % or less than their own 1st list recall. Only about half of the aged with MCI had such low 2nd list recall. About 13% of the aged with normal 1st list recall had such low 2nd list recall. Aged with 2nd list recall that was 50% or less of their 1st list recall had lower scores on other cognitive tests and worse Blessed IMC mental status scores. Conclusions: Memory impairment shown by low 2nd list recall in only some but not in all aged with MCI should improve MCI drug trials by discriminating aged with memory impairment likely to be due to AD from those with life-long low memory. Memory impairment shown by low 2nd list recall in aged who still have normal normal 1st list recall should make possible earlier drug trials and treatment to preserve cognitive function by detecting memory impairment when declining memory is still in the normal range. O1-05-03
IMPLEMENTATION OF UNIVERSAL COGNITIVE SCREENING IN PRIMARY CARE: INITIAL RESULTS
Soo Borson1, James M. Scanlan1, Jeffrey Hummel1,2, Kathy Gibbs3; 1 University of Washington, Seattle, WA, USA; 2UW Physicians Network, Seattle, WA, USA; 3University of Washington Physicians Network, Seattle, WA, USA Background: Primary care doctors are well positioned for early detection of dementia, but rarely screen their patients and miss many, especially, milder cases. Time, complexity, and workload are important barriers to early detection of dementia in primary care offices. Simple, short cognitive screens that can be given by non-physicians could improve initial detection rates if adopted in the primary care setting. Objective: To test the feasibility of implementing cognitive screening of older adults in two primary care practices, using a simple 2-minute screen, the Mini-Cog, developed for this use. Methods: Medical assistants in two primary care clinics were trained to administer the Mini-Cog to all older adults during a routine primary care visit and to inform the physician of results. Two clinics in the same network were used for ‘usual practice’ comparison. Outcomes were percent of eligible patients screened and percents screening positive for cognitive impairment, diagnosed with dementia, prescribed any dementia medication, or referred for specialized dementia assessment. Changes in the intervention clinics were compared with changes occurring in the control clinics. Results: Medical assistants screened 70% of eligible patients during the intervention period (n⫽524, mean age 75); patient acceptance was very high. 26 MAs were trained in the screening procedure, achieving 96% concordance with expert raters in Mini-Cog scoring. Sixteen percent of screened patients scored in the impaired range (Mini-Cog 0-2), consistent with epidemiological data, but dementia diagnoses were made in only 8% of the population. Relative to the baseline pre-intervention period and to control clinics, the screening program resulted in significant increases in dementia diagnoses and specialist referrals, and a less robust increase in treatment rates. Conclusions: Simple cognitive screening can be reliably implemented in primary care with a high rate of patient acceptance. With minimal training and feedback, medical assistants screened patients at a high level of competence. While the impact of screening on physician recognition of dementia was positive, optimal rates of diagnosis and treatment will require additional interventions in primary care. O1-05-04
WHAT IS THE BEST DEMENTIA SCREENING INSTRUMENT FOR GENERAL PRACTITIONERS TO USE?
Henry Brodaty1, Lee-Fay Low2, Louisa Gibson3, Kim Burns3; 1 University of New South Wales, Randwick, New South Wales, Australia; 2Australian National University, Canberra, ACT, Australia; 3 Prince of Wales Hospital, Randwick, New South Wales, Australia Background: The delay of 8-32 months between symptom onset and a diagnosis of dementia suggests that greater attention to diagnosing demen-
S89
tia in its mild stages is necessary. Earlier identification may enable earlier treatment and empower individuals to plan for their future sooner. General practitioners (GPs), although best situated to detect dementia early, display limited capacity to achieve this. Only 26-39% of GPs screen for dementia routinely, and it is estimated that routine screening could double the number of cases GPs identify. General practitioners report the lengthy nature and inadequacies of current screening tools such as the Mini-mental state examination (MMSE) as reasons for their failure to screen. Objectives: We reviewed existing dementia screening tools with a view to informing and recommending suitable instruments to GPs based on their performance and practicability for general practice. Methods: Pre-Medline, Medline, PsycINFO, and the Cochrane Library databases were searched, with 16 instruments included in the final review. The General Practitioner Assessment of Cognition (GPCOG), Mini-Cog, and Memory Impairment Screen (MIS) were chosen as most suitable for routine screening. The GPCOG, Mini-Cog, and MIS were all validated in community, population, or general practice samples, are easy to administer, and have administration times under five minutes. These instruments have negative predictive validities and misclassification rates at least as good if not better than the MMSE, but are quicker and easier to perform. Conclusion: It is recommend that GPs adopt routine screening using the GPCOG, Mini-Cog, or MIS when patients are over the age of 75 years or when cognitive impairment is suspected.
O1-05-05
A BRIEF ALZHEIMER’S SCREEN FOR USE IN CLINICAL PRACTICE
Frederick A. Schmitt1, Marta S. Mendiondo1, Richard J. Kryscio1, John W. Ashford2; 1University of Kentucky, Lexington, KY, USA; 2 Stanford/VA Alzheimer Center, Palo Alto, CA, USA Background: With advances in the treatment of Alzheimer’s disease (AD), clinical focus has shifted to early patient identification. Therefore screening procedures have gained renewed popularity in clinical practice. Recall and fluency tests have been able to distinguish normal individuals from early AD patients. We evaluated a Brief Alzheimer Screen (BAS) derived from item response procedures in a longitudinal cohort of normal elderly, persons with MCI, and mild AD. Objective(s): Evaluate a brief mental status test for general practitioners to screen for MCI and AD. Methods: We previously reported on the BAS using the cognitive assessments from 406 normal subjects and 342 mild AD patients in the CERAD dataset and then applied it to a second validation sample. Logistic regression was used to derive estimates of sensitivity and specificity. The resulting logistic model for discriminating between mild AD and controls included: recall of 3 words, number of animals named in 30 seconds, date, and spelling of WORLD backwards with sensitivity and specificity over 99% and 87%, respectively. The derived logistical model was then applied to a larger sample (N⫽971) of participants followed in our AD Center’s longitudinal cohorts. Three groups were compared using the BAS: normal, MCI, and mild AD defined by a CDR of 0.5 or 1. Analyses showed excellent discrimination between these clinically diagnosed groups (see Table) with significant separation using the BAS weighted score. GROUP (N)
Mean
Median
SEM
Normal (318) MCI (178) Mild AD (475)
33.60 30.62 18.58
33.27 30.75 17.42
0.69 0.28 0.32
Conclusions: The use of item response approaches can help identify items that can improve discrimination between clinical groups when developing screening measures. In addition, simple screening approaches should be designed and evaluated in order to improve detection and promote early treatment in busy medical practice settings where at risk older adults receive their care.