The Validity of the CANS-MCI II (2005 – 2014)

The Validity of the CANS-MCI II (2005 – 2014)

Emory Hill, PhD and Jo Anne Laboff, MSW

Scores on the CANS-MCI were compared with the results of full neuropsychological examinations that were blind to the CANS-MCI results. Identical analyses were performed using full independent neuropsychological evaluation classifications on the 74 subjects who returned a year later.


To determine the ability of the Computer-Administered Neuropsychological Screen for Mild Cognitive Impairment (CANS-MCI) as an accurate screen for mild cognitive impairment (MCI), scores on the CANS-MCI were compared with the results of a full neuropsychological examinations that were blind to the CANS-MCI results. Logistic regression models were used to predict the dichotomous outcomes of MCI vs. normal cognitive functioning (as determined by the neuropsychological exam). Because education can affect scores on measures of cognitive impairment[1,2] samples were separated into individuals with a high school degree or less (N=26) and those with schooling beyond high school (N=57). Gender and age were included in the model. Receiver operating characteristic (ROC) analyses were performed to calculate the sensitivity (the proportion of persons who have MCI that are defined as having MCI) and specificity (proportion of persons who have normal cognitive functioning that are defined as having normal functioning) of the CANS-MCI.

The regression model statistics were very strong (Table 1) indicating a good fit of the data to the model. The CANS-MCI has extremely high levels of sensitivity and specificity (100%) in classifying those with an education up to a high school degree. The optimum sensitivity and specificity for those with 13+ years of education is lower (100%; 85%) but still excellent (Table 2). These findings indicate the CANS-MCI can be a useful screening measure to determine if a person needs to be assessed for cognitive impairment.

Table 1:  CANS-MCI Logistic Regression 3

Education X2 Nagelkerke R2 Predicted Classification % Correct
Less Than 13 Years 35.4 1.0 100
13 or More Years 50.3 .79 84.2

 

Table 2:  CANS-MCI ROC Analyses

Education Area Under Curve % Sensitivity % Specificity
Less Than 13 Years 1.0 100 100
13 or More Years .96 100 84.8

The same analyses were performed using full independent neuropsychological evaluation classifications on the 74 subjects who returned a year later. Despite small numbers of subjects to date, these data indicate that the overall probability that a full neuropsychological evaluation will indicate MCI can be effectively predicted by the CANS-MCI a year earlier. The regression model statistics for the 1-year follow-up evaluations were strong but limited by small sample size. The algorithms correctly classified 85% of participants with a high school degree or less (Chi-square = 11.7; Nagelkerke pseudo-R2 =.63 ) and 80% of those with at least some college (Chi-square = 31.4; Nagelkerke pseudo-R2 =.59) indicating a good fit of the data to the model (Table 3). The CANS-MCI has good levels of sensitivity and specificity in classifying those with an education up to a high school degree. The optimum sensitivity and specificity for those with 13+ years of education is lower but still excellent.

Table 3: CANS-MCI Logistic Regression Analysis (1-year follow-up) 4

Education X2 Nagelkerke R2 Predicted Classification % Correct
Less Than 13 Years 11.7 .63 85
13 or More Years 31.4 .59 19.6

Table 4: CANS-MCI ROC Curve Analysis (1-year follow-up) 4

Education Area Under Curve % Sensitivity % Specificity
Less Than 13 Years .917 92.9 83.3
13 or More Years .888 83.9 73.9

ROC curve analyses on the two educational levels revealed that cut-points lead to sensitivities/specificities of .93/.83 (<=12 yrs) and .84/.74 (13+ yrs). Areas under the curve were high: .917 for <= 12 yrs education and .888 for 13+ yrs (Table 4). Of course, there is the possibility that relative to an external standard the CANS-MCI is more accurate than the independent full neuropsychological evaluations. That appears to be the case in the study by Scanlan et al (2012) which found better agreement between the CANS-MCI measures and well known standard tests. They concluded that “an optimal cognitive screen would be expected to detect all of the patients indicated as demented by the DRS2 and those judged to have MCI by the WMS1/WMS2. By those criteria, the CANS-MCI (typically completed in under 35 minutes) appears as sensitive to dementia and MCI detection as a full neuropsychological examination (typically completed in 2.5 hours).” [5]

Mattis and Wechsler Memory Scale Immediate/Delayed

(Percent of subjects found impaired on Mattis Dementia Rating scale or Wechsler Memory Scale (n>168)

Neuropsychological Exam CANS-MCI Memory alone 3-factor CANS-MCI
Mattis Impaired (Memory & Initiation) 89% 93% 100%
Wechsler Impaired (WMS1 & WMS2) 79% 80% 95%

Independent longitudinal research in progress[6] has produced preliminary data confirming the ability of the CANS-MCI to detect the earliest signs of serious decline as determined by imaging. In the first 48 cases analyzed, the CANS-MCI memory factor score and executive functioning factor score showed significant declines in subjects who had initially tested as completely normal but subsequently showed declines in cognitive ability, but would still not be officially classified as MCI or demented by UDS criteria. A subset of subjects (N=21), were also assessed with volumetric MRI, revealing that the CANS-MCI memory factor score was positively related to hippocampal brain volume (right hippocampus: r=.43, p<.05; left hippocampus: r=.38, p<.08) [6]

The CANS-MCI is meant to be used longitudinally, comparing each person to his or her own previous performance. This allows for even more precise predictive detection even in those people who are still above average and would otherwise not seem worthy of immediate medical attention. If new treatments evolve that slow, stop or reverse the progression of cognitive decline toward dementia, the earliest possible prediction of decline will become critical.

Footnotes

  1. Gifford DR, Cummings JL. Evaluating dementia screening tests: Methodological standards to rate their performance. Neurology 1999;52:224-227.

  2. Lorentz WI, Scanlan JM, Borson S. Brief screening tests for dementia. Canadian Journal of Psychiatry 2002;47(8):723-732.

  3. Jane B. Tornatore, PhD, Emory Hill, PhD, Jo A. Laboff & Brian Fogel 
Validity of Mild Cognitive Impairment Touch Screen Tests: The CANS-MCI Study. 
International Psychogeriatric Association 11th Congress, Chicago, August, 2003.

  4. Jane B. Tornatore, PhD, Emory Hill, PhD, Jo A. Laboff, MSW, Brian Fogel 
One year Follow-Up Analyses of Scoring Algorithms for a Mild Cognitive Impairment Screen: The CANS-MCI Study, Washington, D.C., June, 2005.

  5. James Scanlan, PhD, Jo Laboff, MSW, Emory Hill, PhD Self-reported memory fails to substitute for objective memory measures. Alzheimer’s & Dementia 2012; 8 (4): S780.

  6. Frederick A Schmitt, PhD Department of Neurology, Alzheimer’s Disease Research Center, University of Kentucky Medical Center, unpublished presentation at Screen, Inc. Annual Science Meeting, January, 2013.

Emory Hill, Ph.D.
Founder of Screen Inc., Dr. Hill has a PhD in Clinical Psychology, State University of New York at Buffalo. Later he completed an Informatics Fellowship (post-PhD) at the VA where he studied interface design, multimedia programming, user resistance, evaluation of adaptations to new medical record systems, and the implementation of automated medical records. A trained psychologist and psychometric specialist, Emory was in private practice for nearly 20 years. Before that, he served as an Assistant Professor of Psychology at SUNY, Brockport, NY.