The Clinical Neuropsychologist

ISSN: 1385-4046 (Print) 1744-4144 (Online) Journal homepage: http://www.tandfonline.com/loi/ntcn20

How Do Scores on the ADAS-Cog, MMSE, and CDR-

SOB Correspond?

Steve Balsis, Jared F. Benge, Deborah A. Lowe, Lisa Geraci & Rachelle S.
Doody

To cite this article: Steve Balsis, Jared F. Benge, Deborah A. Lowe, Lisa Geraci & Rachelle S.
Doody (2015): How Do Scores on the ADAS-Cog, MMSE, and CDR-SOB Correspond?, The Clinical
Neuropsychologist, DOI: 10.1080/13854046.2015.1119312

To link to this article: http://dx.doi.org/10.1080/13854046.2015.1119312

Published online: 30 Nov 2015.

Submit your article to this journal

Article views: 43

View related articles

View Crossmark data

Full Terms & Conditions of access and use can be found at

http://www.tandfonline.com/action/journalInformation?journalCode=ntcn20

Download by: [Orta Dogu Teknik Universitesi] Date: 15 January 2016, At: 04:07
 The Clinical Neuropsychologist, 2015
http://dx.doi.org/10.1080/13854046.2015.1119312

How Do Scores on the ADAS-Cog, MMSE, and CDR-SOB

Correspond?

Steve Balsis1, Jared F. Benge2, Deborah A. Lowe1, Lisa Geraci1,

and Rachelle S. Doody3
1
Department of Psychology, Texas A & M University, College Station, TX, USA
2
Department of Neurology, Baylor Scott & White Health, Temple, TX, USA
3
Department of Neurology, Alzheimer’s Disease and Memory Disorders Center, Baylor College
Downloaded by [Orta Dogu Teknik Universitesi] at 04:07 15 January 2016

of Medicine, Houston, TX, USA

Objective: Clinicians and researchers who measure cognitive dysfunction often use the
Alzheimer’s Disease Assessment Scale—Cognitive Subscale (ADAS-Cog), the Mini-Mental
State Examination (MMSE), or the Clinical Dementia Rating scale (CDR-SOB). But, the use of
different measures can make it difficult to compare data across patients or studies. What is
needed is a simple chart that shows how scores on these three important measures correspond to
each other. Methods: Using data from 1709 participants from the Alzheimer’s Disease
Neuroimaging Initiative and item response theory-based statistics, we analyzed how scores on
each measure, the ADAS-Cog, the MMSE, and the CDR-SOB, correspond. Results: Results
indicated multiple inflections in CDR-SOB and ADAS-Cog scores within a given MMSE score,
suggesting that the CDR-SOB and ADAS-Cog are more precise in measuring the severity of
cognitive dysfunction than the MMSE. Conclusions: This study shows how scores on these
three popular measures of cognitive dysfunction correspond to each other, which is very useful
information for both researchers and clinicians.

Keyword: ADAS; Alzheimer’s disease; CDR; Dementia; IRT; MCI; MMSE; Psychometric;
Neuropsychological assessment.

The past several decades have seen significant growth in Alzheimer’s disease
(AD) research. Several research and clinical priorities have emerged from this research,
including maximizing the potential from clinical trial data and evaluating/identifying
individuals in the earliest stages of the disease (Mitchell et al., 2012). The measurement
of cognition is a critical step for identifying the early stages of the disease, and across
studies and centers, various scales are employed to this end. Some of the most com-
monly used instruments to assess AD-related cognitive decline including global decline
are the Alzheimer’s Disease Assessment Scale—Cognitive Subscale (ADAS-Cog;
Rosen, Mohs, & Davis, 1984), which is frequently used in pharmaceutical trials, the
Mini-Mental State Examination (MMSE; Folstein, Folstein, & McHugh, 1975), which

Address correspondence to: Steve Balsis, Ph.D., Associate Professor, Department of Psychology, Texas
A&M University, Mailstop 4235, College Station, TX 77843, USA. E-mail: [email protected]
Data used in preparation of this article were obtained from the Alzheimer’s Disease Neuroimaging Ini-
tiative (ADNI) database (adni.loni.usc.edu). As such, the investigators within the ADNI contributed to the
design and implementation of ADNI and/or provided data but did not participate in analysis or writing of this
report. A complete listing of ADNI investigators can be found at: http://adni.loni.usc.edu/wpcontent/uploads/
how_to_apply/ADNI_Acknowledgement_List.pdf
(Received 31 August 2015; accepted 9 November 2015)

© 2015 Taylor & Francis

 2 STEVE BALSIS ET AL.

is frequently used by clinicians and researchers interested in cognitive aging, and the
Clinical Dementia Rating scale (CDR-SOB; Hughes, Berg, Danziger, Coben, & Martin,
1982; Morris, 1992; O’Bryant et al., 2008), which is commonly used in clinical trials
and practice for rating severity, including in mild and prodromal stages of disease
(Aisen et al., 2011).
While each of these scales has a number of strengths and weaknesses, the fact
that different measures are used across different research centers, studies, and settings
creates limitations for theory and practice. For researchers, the variability in assessments
used across studies limits the ability to compare results across studies or combine data
sets to increase statistical power. For clinicians, the variability in assessments creates
imprecision in assessing change over time or the effectiveness of various interventions.
Downloaded by [Orta Dogu Teknik Universitesi] at 04:07 15 January 2016

Modern psychometric techniques can help provide statistical bridges between

these various dementia staging measures. Item Response Theory (IRT), originally devel-
oped in educational assessment, is one such technique. The process underlying IRT is
mathematically complex, but conceptually straightforward. The first step is to verify the
assumption that all of the measures assess the same general underlying construct (see
Hambleton, Swaminathan, & Rogers, 1991 for an overview of IRT methodologies). For
the purposes of this article, this construct is cognitive dysfunction. The second step is
to assign each participant a precise value of cognitive dysfunction based on available
clinical data. In other words, because the ADAS-Cog (Rosen et al., 1984), MMSE
(Folstein et al., 1975), and CDR-SOB (Hughes et al., 1982; Morris, 1992; O’Bryant
et al., 2008) measures cognitive dysfunction, the observed scores on each measure can
be used to estimate the underlying severity of cognitive dysfunction. This estimation is
a new score known as theta. Once theta is calculated for each participant derived from
multiple indicators, the raw scores on these measures can be mapped onto the ranges of
theta for which they serve as indicators. Thus, similar ranges of underlying cognitive
dysfunction can be “cross-walked” to different observed scores among the measures
with a high degree of reliability (see Wang, Isenor, & Graybeal, 2011). Cross-walking
in this regard refers to the process of equating a raw score on one measure with a raw
score on a different measure. This article outlines our methodology and results and
offers a usable table that shows how raw scores on these three instruments correspond
in the relatively early stages of AD as defined more concretely in the participants
section.

MATERIALS AND METHODS

Data used in the preparation of this article were obtained from the Alzheimer’s
Disease Neuroimaging Initiative (ADNI) database (adni.loni.usc.edu). The ADNI was
launched in 2003 as a public–private partnership. The initial goal of ADNI was to
recruit 800 participants, but ADNI has been followed by two other initiatives, ADNI-
GO and ADNI-2. To date these three protocols have recruited over 1500 adults, ages
55–90, to participate in the research, consisting of older people who are cognitively
healthy, people with early or late MCI, and people with early AD. Demographic infor-
mation and clinical data used for this study were downloaded from the ADNI data
repository (adni.loni.usc.edu) on 28 May 2014.
 ADAS-COG, MMSE AND CDR-SOB 3

Participants
The analysis for this study used baseline data from 1709 participants (768 female
and 941 male) enrolled across all three ADNI phases. Participants were an average of
73.70 years old (SD = 7.18) and highly educated (M = 15.93, SD = 2.85 years), and the
majority identified their race as White (n = 1579, 92.4%). Other races represented
include Black or African American (n = 75, 4.4%), Asian (n = 29, 1.7%), American
Indian or Alaskan Native (n = 3, .2%), and Native Hawaiian or Other Pacific Islander
(n = 2, .1%); 18 participants (1.1%) reported that they were more than one race, and 3
participants (.2%) did not report their race. Ten participants reported their ethnicity as
Hispanic or Latino (.6%); 1643 (96.1%) indicated that they were not Hispanic or
Latino, while 56 participants (3.3%) had unknown ethnicity. Baseline diagnoses
Downloaded by [Orta Dogu Teknik Universitesi] at 04:07 15 January 2016

represented a range of cognitive impairment: 416 (24.3%) were cognitively normal

(CN), 106 (6.2%) had subjective memory complaints, 308 (18.0%) had early MCI, 561
(32.8%) had late MCI, and 318 (18.6%) had very mild AD through AD as defined just
below.
According to ADNI protocol, CN participants serve as the controls and show no
signs of MCI, dementia, or significant depression. CN participants have normal cogni-
tion (defined as CDR = 0, MMSE between 24 and 30, and WMS-R Logical Memory II
subscale score above education-adjusted cutoffs), and neither the participants nor their
study partners report memory concerns. Beginning in ADNI2, participants with signifi-
cant memory concerns (SMC) were enrolled. SMC participants have normal cognition
(defined the same as for CN participants), but they show significant concerns about
memory (either reported by the participant, study partner, or clinician), as measured
using the Cognitive Change Index, a self-report scale. MCI participants have been
enrolled since ADNI1, but in ADNI-GO and ADNI-2, MCI participants, either those
with early MCI or late MCI, are people who have a subjective memory concern and
significant amnestic dysfunction (defined by CDR = .5 plus an abnormal score on the
WMS-R Logical Memory II subscale). However, MCI subjects have sufficiently pre-
served functional abilities and global cognition (MMSE score between 24 and 30), such
that they do not meet criteria for AD. The determination of EMCI versus LMCI is
based on the severity of impairment on the WMS-R Logical Memory II subscale, a
measure of delayed recall for auditory information presented in story format. The ADNI
distinguishes between early MCI and late MCI based on the WMS Logical Memory
score alone, and this distinction should be taken with appropriate caution as it is based
on only one score.
Participants were diagnosed with AD if they met National Institute of
Neurological and Communicative Disorders and Stroke/Alzheimer’s Disease and
Related Disorders Association (NINCDS/ADRDA) criteria for probable AD (McKhann
et al., 1984). At the time of the ADNI diagnosis, AD participants demonstrated deficits
in their global cognition, abnormal memory functioning (based on scores on the
WMS-R Logical Memory II subscale), and showed significant concerns with memory
(reported by the participant, study partner, or clinician). Finally, participants were
excluded if they had a history of significant neurologic disease (including multi-infarct
dementia and subdural hematoma). Detailed information about the ADNI inclusion and
exclusion criteria can be found online, http://www.adniinfo.org/scientists/ADNIStudy
Procedures.aspx.
 4 STEVE BALSIS ET AL.

Measures
The ADAS-Cog (Rosen et al., 1984) is a cognitive assessment with raw scores
that range from 0 to 70, where higher scores indicate greater cognitive dysfunction. The
MMSE (Folstein et al., 1975) is a brief screening instrument for cognitive dysfunction
with raw scores that range from 0 to 30, where lower scores indicate greater cognitive
dysfunction. The CDR-SOB (Hughes et al., 1982; Morris, 1992; O’Bryant et al., 2008)
is a clinician-rated staging method for cognitive dysfunction and functional ability. It
requires an individual severity rating on each of six domains (or boxes): memory, orien-
tation, judgment and problem-solving, community affairs, and activities of daily living.
Under the traditional scoring method, an algorithm is used to produce global scores that
indicate five stages of cognitive dysfunction: 0 = no dementia, .5 = questionable demen-
Downloaded by [Orta Dogu Teknik Universitesi] at 04:07 15 January 2016

tia, 1 = mild dementia, 2 = moderate dementia, and 3 = severe dementia (Morris, 1992).
Alternatively, the Sum of Boxes scoring approach (CDR-SOB) yields scores from 0 to
18 by summing the CDR domain scores (see O’Bryant et al., 2008 for a description of
how to stage using CDR-SOB). We used CDR-SOB in the current study to give a more
fine grained assessment of the correspondence between the CDR measure and the
ADAS-Cog and MMSE. In the present sample, participants had the following scores
(MMSE: M = 27.24, SD = 2.60; ADAS-Cog: M = 10.59, SD = 6.42; CDR-SOB:
M = 1.53, SD = 1.66). Ranges appear in Table 1.

Statistical analyses
IRT assumes unidimensionality of the data, so we conducted exploratory factor
analyses to test the unidimensionality. Specifically, we tested a three-variable model,
where each measure (i.e., the ADAS-Cog, MMSE, and CDR-SOB) served as the vari-
ables. Then, we determined whether the first factor was sufficiently large relative to the
second and third factors to assume unidimensionality. We then conducted IRT analyses
in Multilog (Thissen, 1991) using Samejima’s graded response model (Samejima, 1969)
to determine how the ADAS-Cog, MMSE, and CDR-SOB each function across the
range of latent cognitive dysfunction. Although it can be considered atypical to use full
measures as “items” in IRT, it is a statistically coherent practice within an IRT frame-
work. We also used Multilog to map raw scores on each of these measures to corre-
sponding ranges of latent AD-related cognitive dysfunction, or theta. Finally, we
developed a “crosswalk” table that details how raw scores on each of these three mea-
sures correspond to one another based on their relationship to theta.

RESULTS
Findings indicated that data from the ADAS-Cog, MMSE, and CDR-SOB were
adequately unidimensional for IRT analyses. Indeed, the results of an exploratory factor
analysis with maximum likelihood estimation were nearly perfectly unidimensional.
The first-to-second eigenvalue ratio was 2.45:.28 or 8.73:1. The third eigenvalue also
was .28, and revealed no significant drop-off from the second eigenvalue. Our ratio of
first-to-second eigenvalue, 8.73:1, far exceeded the 3:1 ratio standard which is com-
monly cited for sufficient unidimensionality (see Embretson & Reise, 2000). This is not
 ADAS-COG, MMSE AND CDR-SOB 5

Table 1. Corresponding scores across three mea-

sures of cognitive dysfunction

ADAS-Cog MMSE CDR-SOB

0 30 0
1 30 0
2 30 0
3 30 0
4 30 0
5 30 0
6 29 0
7 29 .5
8 29 .5
Downloaded by [Orta Dogu Teknik Universitesi] at 04:07 15 January 2016

9 28 1
10 28 1
11 27 1.5
12 27 1.5
13 26 2
14 26 2.5
15 26 2.5
16 25 3
17 25 3
18 24 3.5
19 24 4
20 24 4
21 23 4.5
22 23 4.5
23 23 4.5
24 22 5
25 22 5
26 22 5.5
27 22 5.5
28 21 5.5
29 21 6
30 21 6
31 21 6
32 21 6.5
33 21 6.5
34 20 6.5
35 20 6.5
36 20 6.5
37 20 6.5

Notes: ADAS-Cog, Alzheimer’s Disease

Assessment Scale—Cognitive Subscale. MMSE,
Mini-Mental State Examination. CDR-SOB, Clini-
cal Dementia Rating scale—Sum of the Boxes.

to say that these measures assess only one aspect of cognitive dysfunction. Rather, this
result suggests that all of the cognitive processes that these measures assess covary
quite strongly in this sample, and strongly enough to justify the application of
unidimensional IRT procedures.
The IRT analyses revealed that the measures indicated cognitive dysfunction in
the expected direction. Scores on the CDR-SOB and ADAS-Cog were higher at greater
 6 STEVE BALSIS ET AL.
Downloaded by [Orta Dogu Teknik Universitesi] at 04:07 15 January 2016

Figure 1. The correspondence among three measure of cognitive dysfunction in Alzheimer’s disease.
Notes: ADAS-Cog, Alzheimer’s Disease Assessment Scale—Cognitive Subscale. MMSE, Mini-Mental State
Examination. CDR-SOB, Clinical Dementia scale—Sum of the Boxes.

levels of cognitive dysfunction. In contrast, scores on the MMSE were lower at greater
levels of cognitive dysfunction (see Figure 1). Next, expected scores on the three mea-
sures at each level of cognitive dysfunction were compared. Results indicated multiple
inflections in CDR-SOB and ADAS-Cog scores within a given MMSE score (see
Table 1), suggesting that the CDR-SOB and ADAS-Cog are more precise in measuring
the severity of cognitive dysfunction than the MMSE. Figure 1 illustrates how an
MMSE score of 22 is equivalent to an ADAS-cog score of 25 and a CDR-SOB score
of 5.0. Table 1 summarizes the correspondence among each raw score in the relatively
mild range of cognitive dysfunction (ADAS-cog = 0 through 37, CDR-SOB = 0
through 6.5, MMSE = 20 through 30) across the three measures. Note that a threshold
procedure ensured that the relationship among the scores was computed consistently.
For example, a raw score on the ADAS-Cog of 28 would need to be fully reached
before the raw score on the chart switched from 28 to 29.

DISCUSSION
This study examined the relationship among raw scores on three commonly used
measures of AD severity related to cognitive dysfunction. Using IRT, we were able to
determine the correspondence among scores from each of the three measures (Table 1
provides a useful catalog for raw score equivalence across the measures). This table is
designed to allow clinicians and investigators to compare patients or research partici-
pants who were assessed using one measure to those who were assessed using another
measure. Results are consistent with previous research in that they replicate the well-
documented general correspondence among these measures. Higher scores on the
ADAS-Cog and the CDR-SOB are associated with lower scores on the MMSE, and
vice versa. The associations detailed in the table we present are important because they
 ADAS-COG, MMSE AND CDR-SOB 7

show how individual raw scores correspond among the measures. Likewise, the table
shows the meaningful model-based floor and ceiling effects of these measures across
the continuum of cognitive dysfunction.
Of particular note, we replicated the structure among these three measures via a
confirmatory factor analysis using MPlus software (Muthen & Muthen, 2007) in the
database maintained at Baylor College of Medicine’s Alzheimer’s Disease and Memory
Disorders Center (n = 859), (see Doody et al., 2005). Results confirmed equivalent fac-
tor loadings between the samples. Notably, the differences between the samples did not
exceed the required magnitude to impose additional constraints as recommended by
simulations (Cheung & Rensvold, 2002). Confirmatory fit indices between the con-
strained (CFI = .99) and unconstrained (CFI = 1.00) models did not differ more than
.01. This simple analysis helps to further the replicability of the findings.
Downloaded by [Orta Dogu Teknik Universitesi] at 04:07 15 January 2016

A clinician, for example, who is treating a client and has assessed the client using
the MMSE would be able to use our table to understand the patient’s score in relation
to published literature using the ADAS-Cog, for example. Or, if a clinician had an
MMSE score for the patient, but did not have the required collateral source to arrive at
a CDR-SOB score, the clinician could use our table to estimate the client’s CDR-SOB
scores and to arrive at a dementia stage (see O’Bryant et al., 2008, for how to stage
dementia using a CDR-SOB score). The clinical applications extend beyond these two
examples, of course, into the clinical trial literature where the ADAS-Cog is ubiquitous,
yet these other measures are not always used. Our findings provide a way for clinicians
or researchers to directly apply clinical trial findings using the ADAS-Cog to their own
patients, who might have been assessed using the MMSE or CDR-SOB scores.
It is important to note that these main findings are based on just one sample,
although the structural correspondence among the measures was replicated in a second
sample, and the ADNI sample is one of the best characterized and largest samples in
Alzheimer’s research. That said, the ADNI sample is skewed in terms of education and
race, and one should exercise caution when applying this table to underrepresented
minority groups. Also, participants had relatively mild cognitive dysfunction, and this
study does not address scores in the range of severe cognitive dysfunction. While this
is a limitation of the study, it is also a potential strength because most of the relevant
comparisons among these measures will be made for individuals and studies that
address mild cognitive dysfunction.

DISCLOSURE STATEMENT
No potential conflict of interest was reported by the authors.

REFERENCES

Aisen, P. S., Gauthier, S., Ferris, S. H., Saumier, D., Haine, D., Garceau, D., … Sampalis, J.
(2011). Tramiprosate in mild-to-moderate Alzheimer’s disease–a randomized, double-blind,
placebo-controlled, multi-centre study (the Alphase Study). Archives of Medical Science, 7,
102–111.
 8 STEVE BALSIS ET AL.

Cheung, G. W., & Rensvold, R. B. (2002). Evaluating goodness-of-fit indexes for testing
measurement invariance. Structural Equation Modeling: A Multidisciplinary Journal, 9,
233–255.
Doody, R. S., Pavlik, V., Massman, P., Kenan, M., Yeh, S., Powell, S., … Chan, W. (2005).
Changing patient characteristics and survival experience in an Alzheimer’s disease center
patient cohort. Dementia and Geriatric Cognitive Disorders, 20, 198–208.
Embretson, S., & Reise, S. P. (2000). Item response theory for psychologists. Mahwah, NJ:
Lawrence Erlbaum Associates.
Folstein, M. F., Folstein, S. E., & McHugh, P. R. (1975). “Mini-Mental State.” A practical
method for grading the cognitive state of patients for the clinician. Journal of Psychiatric
Research, 12, 189–198
Hambleton, R. K., Swaminathan, H., & Rogers, H. J. (1991). Fundamentals of item response the-
Downloaded by [Orta Dogu Teknik Universitesi] at 04:07 15 January 2016

ory. Newbury Park, CA: Sage.

Hughes, C. P., Berg, L., Danziger, W. L., Coben, L. A., & Martin, R. L. (1982). A new clinical
scale for the staging of dementia. The British Journal of Psychiatry, 140, 566–572.
McKhann, G., Drachman, D., Folstein, M., Katzman, R., Price, D., & Stadlan, E. M. (1984).
Clinical diagnosis of Alzheimer’s disease: Report of the NINCDS-ADRDA Work Group*
under the auspices of Department of Health and Human Services Task Force on Alzheimer’s
Disease. Neurology, 34, 939–939.
Mitchell, S. L., Black, B. S., Ersek, M., Hanson, L. C., Miller, S. C., Sachs, G. A., … Morrison,
R. S. (2012). Advanced dementia: State of the art and priorities for the next decade. Annals
of Internal Medicine, 156, 45–51.
Morris, J. C. (1992). The Clinical Dementia Rating (CDR): Current vision and scoring rules. Neu-
rology, 43, 2412–2414.
Muthen, L. K., & Muthen, B. O. (2007). MPlus user’s guide (6th ed.). Los Angeles, CA: Muthen
& Muthen.
O’Bryant, S. E., Waring, S. C., Cullum, C. M., Hall, J., Lacritz, L., Massman, P. J., … Doody, R.
(2008). Staging dementia using Clinical Dementia Rating Scale Sum of Boxes scores: A
Texas Alzheimer’s research consortium study. Archives of Neurology, 65, 1091–1095.
Rosen, W. G., Mohs, R. C., & Davis, K. L. (1984). A new rating scale for Alzheimer’s disease.
American Journal of Psychiatry, 141, 1356–1364.
Samejima, F. (1969). Estimation of latent ability using a response pattern of graded scores.
Psychometric Monograph (No. 17). Richmond, VA: Psychometric Society.
Thissen, D. (1991). Multilog user’s guide, version 6.3. Mooresville, IN: Scientific Software.
Wang, H., Isenor, A., & Graybeal, J. (2011). Moving between standards (crosswalking). In The
MMI guides: Navigating the world of marine metadata. Retrieved October 8, 2015, from
http://marinemetadata.org/guides/mdatastandards/crosswalks