Research Report

Test-Retest Reliability and Minimal

Detectable Change Scores for the
Timed "Up & Go" Test, the Six-Minute
Walk Test, and Gait Speed in People
With Alzheimer Disease
Julie D. Ries, John L Ecbternach, Leah Nof, Micbelle Gagnon Blodgett
).D. Ries, PT, PhD, CCS, is Assistant
Professor, Program in Physical
Background, with the increasing incidence of Alzheimer disease (AD), deter- Therapy, Marymount University,
mining the validity and reliability of outcome measures for people with this disease 2807 N Clebe Rd, Ariington, VA
is necessary. 22207 (USA). Address ail corre-
spondence to Dr Ries at: Julie.
Objective. The goals of this study were to assess test-retest reliability of data for [email protected].
the Timed "Up & Go" Test (TUG), the Six-Minute Walk Test (6MWT), and gait speed ].L. Echternach, PT, DPT, EdD,
and to calculate minimal detectable change (MDC) scores for each outcome measure. ECS, FAPTA, is Professor and Emi-
Performance differences between groups with mild to moderate AD and moderately nent Schoiar Emeritus, Schooi of
Physicai Therapy, Old Dominion
severe to severe AD (as determined by the Functional Assessment Staging [FAST] University, Norfolk, Virginia, and
scale) were studied. Adjunct Professor, Department of
Physical Therapy, Nova Southeast-
Design. This was a prospective, nonexperimental, descriptive methodological ern University, Fort Lauderdaie,
study. Fiorida.

L. Nof, PT, PhD, is Professor, De-

M e t h o d s . Background data collected for 51 people with AD included: use of an partment of Physicai Therapy,
assistive device, Mini-Mental Status Examination scores, and FAST scale scores. Each Nova Southeastern University.
participant engaged in 2 test sessions, separated by a 30- to 60-minute rest period, M. Cagnon Blodgett, PsyD, is
which included 2 TUG trials, 1 6MWT trial, and 2 gait speed trials using a comput- Ciinicai Assistant Professor, De-
erized gait assessment system. A specific cuing protocol was followed to achieve partment of Ceriatrics, Nova
optimal performance during test sessions. Southeastern University.

[Ries |D, Echternach )L, Nof L,

Results. Test-retest reliability values for the TUG, the MAXHT, and gait speed were Cagnon Blodgett M. Test-Retest
high for all participants together and for the mild to moderate AD and moderately reiiabiiity and minimai detectabie
change scores for the Timed "Up
severe to severe AD groups separately (intraclass correlation coefficients >.973); & Co" Test, the Six-Minute Waii<
however, individual variability of performance also was high. Calculated MDC scores Test, and gait speed in peopie
at the 90% confidence interval were: TUG=4.09 seconds, 6MWT=33.5 m (110 ft), with Alzheimer disease. Phys Ther.
and gait speed=9.4 cm/s. The 2 groups were significantly different in performance 2009;89:569-579.]
of clinical tests, with the participants who were more cognitively impaired being 2009 American Physical Therapy
more physically and functionally impaired. Association

Limitations, A single researcher for data collection limited sample numbers and
prohibited blinding to dementia level.

Conclusions. The TUG, the 6MWT, and gait speed are reliable outcome measures
for use with people with AD, recognizing that individual variability of performance
is high. Minimal detectable change scores at the 90% confidence interval can be used
to assess change in performance over time and the impact of treatment.
Post a Rapid Response or
find The Bottom Line:
www.ptjournal. org

|une 2009 Volume 89 Number 6 Physicai Therapy 569

 Reliability and Validity of Outcome Measures for Alzheimer Disease

A
lzheimer disease (AD) is the our review of the literature. Mixed mation to delineate the "expected"
most common form of de- results from studies make it difficult changes from "trtie" changes in per-
mentia in elderly people and to know which outcome measures formance. Statistically, absolute reli-
affects an estimated 5.2 million indi- will best serve physical therapists' ability is determined by the standard
' viduals ill the United States.' It is needs in monitoring change in per- error of measurement (SEM), or the
estimated that 13% of people aged formance in individuals with AD. standard deviation of the measure-
65 years and older are diagnosed Outcome measures that have been ment errors,"" and a clinically use-
with AD, and the incidence and prev- studied for reliability with individu- ftil mechanism for looking at abso-
alence increase considerably with als with AD or dementia include: the lute reliability is the minimal
age.' With the aging of the popula- Timed "Up & Go" Test (TUG),'?^ detectable change (MDC) score.'''
tion, physical therapists in geriatrics the Six-Minute Walk Test (MWT),".
will be treating an increasing num- and gait speed.''''" Recent literature presenting TUG'"
ber of people with AD. Given the and gait speed'" data for individuals
need to measure outcomes to assess Reliability measurements indicate with dementia highlights the impor-
progress or decline in function, spe- the degree to which scores of a clin- tance of understanding relative ver-
cific clinical tools should be tested ical test are free from measurement sus absolute reliability. Even though
for reliability and validity with indi- errors," and although conceptually test-retest reliability coefficients for
viduals with AD. straightforward, the application of clinical tests are high, individual vari-
this notion can be complex."'^ Re- ability and measurement error make
There are recent publications sup- liability can be expressed as relative it very difficult to identify a "true"
porting the physical and functional reliability or as absolute reliability. If change in performance over time.
benefits of exercise in the manage- a measurement has high relative re- Minimal detectable change scores
ment of AD.^5 Identification of ap- liability, this indicates that repeated proyide researchers and clinicians
propriate and useful outcome mea- measurements will reveal consistent with the opportunity to determine
sures for people with AD would positioning or ranking of individuals' whether a change in performance is
enhance the ability to assess the ef- scores within a group. ' ' If a measure- a meaningful change (ie, beyond ex-
fectiveness of interventions in clini- ment has high absolute reliability, pected measurement error and indi-
cal and research environments. Our this indicates that, upon repeated vidual variability).
current understanding of the psycho- measurement, scores show little vari-
metric properties of specific clinical ability." Relative reliability is mea- Clinical observation in people with
tests with this population is limited. sured with correlation coefficients. AD reveals increasing variability of
Methodological studies assessing the The intraclass correlation coefficient performance with increasing levels
reliability of clinical tools for people (ICC) evaluates correlation based of dementia. The existing literature
with AD or dementia are scarce, but upon variance estimates from analy- supports this observation. Although
not nonexistent.'*"^ Given the ex- sis of variance'-"*; the more common Thomas and Hageman^ found the
tremely limited research available ex- the variance between sets of mea- TUG to have reasonable test-retest
clusively vvfith people with a diagno- surements, the higher the ICC. '^ The reliability in subjects in day care set-
sis of AD, information gleaned from ICC is an appropriate statistic for ex- tings who ^vere considered to have
research with individuals with other amining test-retest reliability." As a mild to moderate dementia (Mini-
types of dementia was included in general guideline, an ICC above .75 Mental Status Examination [MMSE]
is considered to demonstrate good [SD] = 16.9 [7.3]), Tappen et al''
reliability; for clinical measures, it is found the TUG to be impracticable
Available With suggested that reliability should ex- for use in subjects with moderate to
This Article at ceed .90 to ensure reasonable severe AD (MMSE=9.3 [6.0]). Miller
www.ptjournal.org validity.'' et al, in a post hoc assessment of
performance on the 6MWT (as a
The Bottom Line clinical Excellent test-retest reliability does component of assessing test-retest
summary
not necessarily ensure that individu- reliability of the Senior Fitness Test),
The Bottom Line Podcast als' repeated performance will be found that subjects who were cogni-
Audio Abstracts Podcast consistent from test to test. Scores tively impaired showed greater vari-
This article was published ahead of
may vary, given expected variability ability than subjects wbo were cog-
print on April 23 2009, at of individual performance and mea- nitively intact; they suggested that
www.ptjournal.org. surement error. A measure of abso- the 6MWT is not reliable for use witb
lute variability provides useful infor- elderly people who are cognitively

570 Physical Therapy Volume 89 Number 6 |une 2009

 Reliability and Validity of Outcome Measures for Alzheimer Disease

impaired. The combined findings of of participants. Inclusion criteria historical information; creating a
these studies''-''' and the previously were: probable diagnosis of AD, low-stress environment; using
noted clinical observation suggest medical stability, and ambulation friendly facial expressions, eye con-
that test-retest reliability of physical with or without an assistive device tact, and a pleasant, but firm, voice;
and functional performance mea- or with handheld guiding assistance one-step commands; and stating
sures with individuals with AD may of one person. Exclusion criteria meaningful goals,2"-'^^ The progres-
be influenced by level of dementia. were: overt neuromuscular or mus- sion of cuing for all tests also was
culoskeletal problems, acute cardiac specific and based on the published
The purposes of this research were: or pulmonary conditions, and sur- literature,^"-'*' Cuing began with ver-
(1) to determine test-retest reliability gery within the previous 6 months. bal instruction with a concurrent vi-
of data for the TUG, the 6MWT, and sual cue or gesture, followed by
gait speed with individuals with AD; Background data were collected pri- modeling or demonstration, fol-
(2) to determine MDC scores for marily from the facility chart and in- lowed by tactile guidance, and fi-
each of the outcome measures; and cluded: age, sex, living environment, nally, if necessary, physical assis-
(3) to identify performance differ- and use of an assistive device (classi- tance. Participants were given 10
ences between participant groups fied as "none," "use of a cane," or seconds to respond to a cue before
stratified by level of dementia. "use of a walker or rolling walker") the tester defaulted to a higher level
or handheld guiding assistance for of cuing, A 7-level scale of assistance
The existing literature guided the ambulation. Personal information developed by Beck et aP' for elderly
choice of outcome measures for the (eg, vocation, avocations, family people with cognitive impairments
present study. We hypothesized that members' names, likes and dislikes) was used to classify the type of cuing
the test-retest reliability of the clini- was collected from the facility or assistance required for each par-
cal tools would decrease with in- record and staff. This information ticipant. If a participant required
creased level of dementia, such that proved useful in establishing rapport handheld guiding assistance, every
the measures would be reliable for with the participants. The primary effort was made to allow the partic-
use with individuals with mild to researcher G,D,R,) administered the ipant to drive the movement; how-
moderate AD, but not for use with MMSE to all participants. The pri- ever, if the participant stopped or
individuals with moderately severe mary researcher scored the Func- veered from the intended path of
to severe AD, We also hypothesized tional Assessment Staging (FAST) movement, the researcher guided
that, when stratified by level of de- scale'5-'8 using a caregiver or staff the participant's motion back on
mentia, the participants who were informant. The FAST scale has been task.
less cognitively impaired would per- established as a reliable and valid as-
form better on the clinical tests com- sessment tool for people with AD,'^ Two testing sessions for each partic-
pared with the participants who The FAST instalment identifies 16 ipant were performed on the same
were more cognitively impaired. levels of ftmctioning, separated into day with a 30- to 60-minute rest pe-
7 stages (Tab, 1), and provided the riod separating testing sessions. Ev-
Method operational definitions for level of ery effort was made to keep all fac-
Participants and Procedure AD in this study. The FAST scale was tors associated with the testing
This methodological study used a used to stratify the participants into sessions consistent (eg, general time
prospective, nonexperimental, de- 2 groups based on level of dementia: of day, staff member assisting with
scriptive research design. Guardian a mild to moderate AD group (FAST testing, room or area in which test-
informed consent was obtained for scale score=4 or 5) and a moderately ing was performed). Participants
all participants, with the exception severe to severe AD group (FAST performed the TUG, the 6MWT, and
of one participant who signed her scale score=6 or 7), the test of gait speed.
own informed consent statement
with her family's approval. When Practical tips on interaction and The TUG52 ^ ^ (ggt of the time re-
possible, assent forms were signed communication with individuals quired for an individual to stand up
by participants in conjunction with with AD have been reported in the from a chair with armrests, walk 3 m,
guardian informed consent. Four literature. Every effort was made to turn, walk back to the chair, and sit
sites providing care to individuals integrate these concepts into the down. In the present study, partici-
with AD (2 inpatient programs and 2 protocol for the present study to pants circled a small orange cone
day care programs) participated in maximize success of interactions, in- placed at the 3-m mark. Participants
the study. The administrative contact cluding: creating a personal connec- were instructed to "go as fast as you
at each site aided in the recniitment tion with the patient using personal safely can," The stopwatch timing

)une 2009 Volume 89 Number 6 Physical Therapy 571

 Reliability and Validity of Outcome Measures for Alzheimer Disease

Table 1
Functional Assessment Staging (FAST) Scale for People With Alzheimer Disease

Stage Level of Functioning Ciinical Diagnosis

No decrement "Normal" adult
Subjective deficit in word finding Normal-aged adult
Deficit In demanding employment settings Compatible with incipient AD
Assistance required in complex tasks (eg, handling finances, marketing, Mild AD
planning dinner for guests)

Assistance required in choosing proper clothing Moderate AD

6a Assistance required in putting on clothing Moderately severe AD
Assistance required in bathing properly

Assistance required with the mechanics of toileting (eg, flushing, wiping)

6d Urinary incontinence

6'e Fecal incontinence

7a Speech ability limited to approximately a half-dozen intelligible words Severe AD

7|b Intelligible vocabulary limited to a single word

7c Ambulatory ability lost

Ability to sit up lost

Ability to smile lost

Ability to hold up head lost

started when the participant's bot- profile of temporal and spatial pa- combinations for each participant.
torn left the chair and ended when rameters of gait and is considered a The test order administration re-
"th bottom made contact with the valid and reliable quantitative gait as- mained constant from test session 1
chair after the walk. sessment tool."5'^ Participants were to test session 2 for each participant.
instructed to walk at a "comfortable"
The 6; is the distance walked in pace for the length of the mat (4.57 Data Management and Analysis
a period of 6 minutes. This test was m [15 ft]), and the walking path was We used SPSS 15.0 for Windows^ for
initially considered an endurance established such that acceleration data management and analysis. Level
measure but more recently has and deceleration did not occur on of significance was predetermined to
been coisidered a broader measure the mat. be P<.05 for all statistical analyses.
of mobility and function.*''" The Descriptive statistics for compari-
6MWT was performed in long hall- Testing took place at the participat- sons of groups included independent-
ways of the participating facilities. ing facilities. Patierits performed one samples t tests for parametric data
Participants walked at a "comfort- practice run of the TUG and one and chi square and Mann-Whitney
able paae," were discouraged from practice pass on the GAITRite walk- U tests for nonparametric data. All
talking during the test, and were no- way. They did not perform a practice descriptive comparisons between
tified of each passing minute. If par- run of the 6MWT, but were oriented groups were 2-tailed, as no assump-
ticipants were distracted or stopped to the walking course. Each testing tions of directionality -were made.
walking, they were prompted to session included 2 trials of the TUG,
keep walking" and were advised of 2 passes at a comfortable pace on the Test-retest reliability of data for all
the time remammg. GAITRite mat, and 1 trial of the tests was assessed using the ICC
6MWT. Tests were performed in (model 2), which is appropriate for
Self-selected gait speed was assessed variable order to control for variabil- methodological research. " ' 3 Reli-
using the GAITRite walkway.* This ity of performance from first to last ability of data obtained for the TUG
portable mat with embedded sensors test as a confounding factor. The or- and gait speed was assessed using
and companion software creates a der of test administration was ran- the ICC (2,2), as mean scores from 2
domized, determined by blind draw-
CIR Systems Inc, 60 Garlor Dr, Havertown, ing of test order from all possible * SPSS Inc, 233 S Wacker Dr, Chicago, IL
PA 19083. 60606-6412.

572 'hysical Therapy Volume 89 Number 6 June 2009

 Reliability and Validity of Outcome Measures for Alzheimer Disease

trials from each test session were (2) MDCyo = SEM X 1.65 X ^ groups as well. All participants were
used in the calculations. Mean scores consistent in their level of cuing
are considered better estimates of In this equation, SEM was calculated needs from test session 1 to test ses-
true scores and can increase reliabil- as described previously. The 1.65 in sion 2. The 2 groups were sig-
ity estimates.'* For calculation of the MDCtio equation represents the nificantly different in their level of
test-retest reliability for the 6MWT, z-score at the 90% confidence level. cuing needs, as evidenced by Mann-
the ICC (2,1) was used, as there was The product of SEM multiplied by Whitney 7test analysis, with the par-
only one test score from each ses- 1.65 is multiplied by the square root ticipants who were more cognitiveiy
sion. For each clinical test, the reli- of 2 to account for errors associated impaired requiring higher levels of
ability coefficient was calculated for with repeated measurements. cuing than those who were less cog-
the entire sample and then sepa- nitiveiy impaired (Tab. 2). The use of
rately for the mild to moderate AD Results assistive devices (classified as
group and the moderately severe to Data were collected on a total of 53 "none," "use of a cane," or "use of a
severe AD group. participants. Two participants' data walker or rolling walker") was simi-
were eliminated from analysis be- lar between groups. More than 50%
Independent-samples t tests were cause their dementia was later deter- of the participants of both groups
used to determine differences in per- mined to be caused by factors other were ambulatory without assistive
formance on the clinical tests be- than AD. One setting was not condu- devices. Six participants, all in the
tween the 2 groups. Comparisons cive to the performance of the moderately severe to severe AD
were made using the mean score of 6MWT, so that test was not per- group, required handheld guiding as-
all trials for each participant on the formed with participants in that set- sistance for ambulation.
given test (ie, mean of 4 TUG scores, ting. On 2 occasions, individuals at
mean of 2 6MWT scores, and mean other settings declined to perform Intraclass correlation coefficients for
of 4 gait speed measurements). One- the 6MWT. The Figure diagrams a test-retest reliability were very high
tailed tests were used to assess these flowchart of participants, explaining for all outcome measures and for the
data, as there is evidence to suggest any differences between numbers of entire sample and each group (for
that a decrease in speed occurs in participants tested and data used in the TUG, ICC=.985-.988, P<.00\;
patients with dementia'*'*-''^; there- analysis of the results. The remaining for the 6MWT, ICC = .982-.987,
fore, an assumption of directionality 51 participants were stratified into P<.001; and for gait speed,
was thought to be reasonable. the mild to moderate AD group ICC=.973-.977, P<.001).
(n=20) and the moderately severe to
Standard errors of measurement and severe AD group (n=31). There were statistically significant
MDC scores were calculated for the differences between the mild to
TUG, the 6MWT, and gait speed. Descriptive statistics for the 51 sub- moderate AD group and the moder-
Standard errors of measurement" jects are shown in Table 2. The 2 ately severe to severe AD group on
were calculated using the following groups were similar in age, as deter- TUG, 6MWT, and gait speed perfor-
equation: mined by the independent-samples t mance. The participants who were
test, and similar in sex and living more cognitiveiy impaired were
(1) SEM = sdX environment (ie, home versus inpa- slower on the TUG and the test of
tient), as determined by analysis of gait speed and walked shorter dis-
In this equation, sd is the standard frequencies using the chi-square test. tances in the 6MWT compared with
deviation of the measure, and r is Mini-Mental Status Examination the participants who w^ere less cog-
the reliability coefficient (test-retest scores also are presented in Table 2. nitiveiy impaired (Tab. 3).
reliability in the form of ICC for the The MMSE scores of the 2 groups
subject group). For repeated mea- were compared using the Mann- Repeated-measures SEMs were calcu-
sures, the SEM was multiplied by the Whitney U test, a nonparametric sta- lated for the TUG, the 6MWT, and
square root of the number of tistical analysis, given the ordinal na- gait speed to provide a comparison
measurements. ' ' ture of the MMSE data. Given the of individual variability of perfor-
desire to compare the participants in mance across groups (Tab. 4). Al-
Minimal detectable change scores the present study with those in many though there was little difference in
were calculated for the TUG, 6MWT, published studies that reported SEMs for the mild to moderate AD
and gait speed data at the 90% con- MMSE findings using parametric sta- group compared with the moder-
fidence interval. The formula used tistics, mean scores and standard de- ately severe to severe AD group for
for calculating MDCi,,,''''''^ was: viations are presented for both the 6MWT and gait speed (-10% dif-

june 2009 Volume 89 Number 6 Physical Therapy 573

Reliability and Validity of Outcome Measures for Alzheimer Disease

53 participants
with
diagnosis of
AD tested in
4 settings

2 participants' data
omitted when
diagnosis of AD
rescinded, leaving
51 participants

One setting was not

1 participant conducive to 6MWT
became agitated (16 participants did not
with MMSE, and perform 6MWT at that
test was aborted setting) and 2
participants from other
settings declined to
perform 6MWT

MMSE TUG and 6MWT

data on 50 GAITRite data on 33
participants data on 51 participants
participants

Figure.
Flowchart of study participants. AD=Alzheimer disease, MMSE=Mini-Mental Status Examination, 6MWT= Six-Minute Walk Test,
TUG=Timed "Up & Co" Test, CAITRite=computerized walkway test of gait parameters.

ference), there was a substantial dif- TUG, the 6MWT, and gait speed for The TUG appears to be the most
ference in SEMs for TUG scores be- individuals with AD. All 3 outcome widely studied of the tools. We cal-
tween the 2 groups (100% measures were found to have excel- culated an ICC of .987 for test-retest
difference), with the participants lent test-retest reliability, well ex- reliability of TUG scores for all par-
who were more cognitively impaired ceeding the r=.9O threshold" for ticipants. Tappen et aH had such dif-
showing greater variability of perfor- minimal acceptable reliability for a ficulty getting their subjects with
mance compared with the partici- clinical test and indicating that these moderately severe to severe AD
pants who were less cognitively im- tests can be used clinically with good (MMSE=9.3 [6.0]) to perform the
paired. Table 4 also presents the confidence in their test-retest reli- TUG, that they had to modify the test
MDC90 values for the TUG, the ability (ie, relative reliability). Test- beyond recognition. Our partici-
6MWT, and gait speed for all retest reliability was not influenced pants in the moderately severe to
participants. by level of dementia, as was hypoth- severe AD group had comparable
esized. Existing literature shows MMSE scores (10.2 [8.8]) and were
Discussion mixed results for relative reliability able to perform the test with excel-
Our initial purpose was to determine of these tools in people with AD and lent relative reliability results. Rock-
test-retest reliability of data for the dementia. wood et aF reported an ICC of .56

574 Physical Therapy Volume 89 Number 6 lune 2009

 Reliability and Validity of Outcome Measures for Alzheimer Disease

Tabie 2.
Descriptive Statistics for Participants (N=51)
Moderateiy Severe to
All Participants Miid to Moderate Severe AD Croup Statisticai Comparison
Variable (N=51) AD Croup ( n = 2 0 ) (n=31) Between Croups

Age (y), XSD 80.71 8.77 81.059.48 80.488.43 (=.223

df=49
P=.824

Sex (no. female [%]) 34 (66.7) 12(60.0) 22(70.7) ;^=.658

df=^
P=.417

Living environment (no. living at home [%]) 39 (76.5) 16 (80.0) 23 (74.2) ;^=.228
df=^
P=.633

MMSE M-W U= 134.5

XSD 13.1 8.2 17.404.50 10.208.83 P=.OO1*
Range 0-30 10-26 0-30
Median 14 16.5 7.5
Mean rani<: 33.78 Mean rank: 19.98
Sum of ranks: 675.50 Sum of ranks: 599.50

Levis of assistance for eiderly peopie with Mean rank: 19.05 Mean rank: 30.48 M-W U=171.00
cognitive impairment (ievel of cuing) Sum of ranks: 381.00 Sum of ranks: 945.00 P=.OO4*

o AD=A2heimer disease, MMSE=Mini-Mentai Status Examination, (=independent-sampies test, df=degrees of freedom, ;j^=chi-square test of
independence, M-W (7= Mann-Whitney U test of independent samples. Asterisk indicates statistically significant difference between dementia groups.

for test-retest reliability of TUG w^ith the higher relative reliability authors suggested that the MWT
scores in elderly individuals with found in our study. may be the preferred test of physical
cognitive impairment, but their performance for people with AD.
methodology was fraught with diffi- The 6MWT has not been widely used We calculated an ICC of .987 for test-
culties of working within the limita- in people with AD or dementia; how- retest reliability of 6MWT scores in
tions of retrospective data. A study ever. Tappen et al"* reported that our study. Thomas and Hageman'
by Thomas and Hageman' with a their participants with AD who were and van Iersel et al'" reported ICCs
small sample of individuals with unable to perform the TUG were of .92 and .77, respectively, for test-
mild to moderate dementia (MMSE= able to perform the 6MWT. They did retest reliability of measurements of
16.9 [7.31) revealed an ICC of .87 for not report test-retest reliability of the self-selected gait speed in people
test-retest reliability of TUG scores, 6MWT scores, although the research with dementia. We calculated an ICC
and van Iersel et al'" examined test- design was such that their ICCs of of .977 for test-retest reliability of
retest reliability in people with de- .76 to .90 for intrarater reliability gait speed measurements. Our find-
mentia (MMSE=19.1 [5.2]) and (one rater observing 2 different ses- ings consistently showed higher test-
found an ICC of .97 for the TUG. sions) could potentially be inter- retest reliability on all 3 outcome
Thesefindingswere more consistent preted as test-retest reliability. The

Table 3.
Performance Differences on Timed "Up & Co" Test, Six-Minute Walk Test, and Gait Speed Between Dementia Croups
Outcome Measure Dementia Croup (n) XSD independent-Sampies t Test, t (</f) P

Timed "Up & Co" Test (s) Mild to moderate AD (20) 19.959.81 -1.876(49) .0335*

Moderately severe to severe AD (31) 28.01 17.49

Six-Minute Walk Test (ft) Mild to moderate AD (16) 938.78428.62 2.411 (31) .011*

Moderately severe to severe AD (17) 615.71 338.34

Gait speed (cm/s) Mild to moderate AD (20) 66.0729.63 1.823 (49) .037*'

Moderately severe to severe AD (31) 52.4323.58

' AD=Al2heimer disease. Asterisk indicates statistically significant at P<.05.

June 2009 Volume 89 Number 6 Physical Therapy 575

Reliability and Validity of Outcome Measures for Alzheimer Disease

Table 4
Standard Error of Measurement (SEM) for Repeated Measures and Minimal Detectable Change Scores at the 90% Confidence
Interval (MDC90) for the Timed "Up & Go" Test, the Six-Minute Walk Test, and Gait Speed

SEM
All
Participants Mild to Moderate Moderateiy Severe to
Outcome Measure (N=S1) AD Croup (n=20) Severe AD Croup (n=31) MDC90, Aii Participants
Timed "Up & Go" Test (s) 2.48 1.52 3.03 4.09
Six-Minute Walk test (m) 20.28 (66.53 ft) 21.86 (71.72 ft) 19.57 (64.20 ft) 33.47 (109.8 ft)
Gait speed (cm/s) 5.72 6.07 5.48 9.44

measures compared with previous documented (verbal cue/gesture, clinician or researcher to repeatedly
research. modeling/demonstration, physical/ elicit the most favorable perfor-
tactile prompt, progressive amounts mance from an individual with AD.
One factor that may have enhanced of physical guidance) and were con-
performance on all clinical tests in sistent from one testing session to Our findings demonstrate that al-
the present study was the careful use the next. Not surprisingly, partici- though test-retest reliability (relative
and progression of cuing to facilitate pants with moderately severe to se- reliability) for the clinical tests was
optimal I performance. Although we vere AD required more substantive excellent, there was still a substantial
anticipated that performance consis- prompting and guiding for perfor- degree of variability of performance
tency from one trial to the next per- mance of the outcome measures for individual participants from one
haps would suffer with increasing than those with mild to moderate test session to the next (absolute re-
dementia, the steady and scripted AD. Six of our 51 participants, all liability). The SEM and MDCy,, were
use of verbal and tactile cuing to from the moderately severe to severe calculated to objectify these find-
optimize performance was carefully AD group, required handheld guid- ings. Because the SEM is based on an
implemented; this may have consis- ing assistance of one person to com- assumption of normal distribution,
tently facilitated the best perfor- plete the outcome measures. With- probabilities of the normal curve can
mance. Both Nordin et al^ and van out the physical guidance of the be applied to SEM values." Values
Iersel et al'" commented that the use researcher, these participants would from Table 4 can be translated to
of cuing was the key to the success- not have been able to complete the clinical performance using these
ful administration of the TUG in sub- tests. principles. Eor instance, there is a
jects with cognitive impairment in 68% probability that a repeated mea-
their recent reliability studies. In all A recent publication by Hauer and sure of the TUG will be within 1.52
of the studies review^ed that ad- Oster*'^ reiterates that measuring seconds (1 SEM) of the original score
dressed cuing, the authors either ex- functional performance in people for an individual with mild to mod-
pressed simply a general statement with dementia is very complex and erate AD and a 96% probability that a
that cuing was aUow^ed'*'"' or re- cautions researchers that when we repeated measure, will be within
ported a dichotomy of cuing versus provide external cues to patients, 3.04 seconds (2 SEMs) of the original
no cuing.^ perhaps w^e are measuring the reli- score. For an individual with moder-
ability and quality of the external cu- ately severe to severe AD, there is a
We believe that careful use of cuing ing (ie, the researcher's perfor- 96% probability that a repeated mea-
was an asset to consistency of per- mance) as opposed to, or as well as, sure of the TUG will be within 6.06
formance, contributing to the high the patients' performance. In con- seconds (2 SEMs) of the original
test-retest reliability findings for the trast, we contend that a consistent score. This could be useful informa-
clinical tests in our study. Perhaps progression of cuing to facilitate best tion when examining repeat perfor-
our careful progression of cuing was possible performance may be the op- mances of individuals with AD on
pivotal in the successful administra- timal way to administer clinical tests the TUG. The dichotomy of demen-
tion of the TUG in the moderately to people with AD or dementia. The tia levels is important in interpreting
severe to severe AD group, as Tap- use of a consistent cuing paradigm, clinical findings here, as a difference
pen et al"* were unable to administer in conjunction with following other in performance of approximately 4
the TUG; to their subjects with com- suggestions related to establishing to 5 seconds likely represents a
parable MMSE scores. Our partici- rapport and maintaining a nonthreat- change beyond the expected vari-
pants' cuing needs were rated and ening environment, may allow the ability in performance in a patient

576 Physical Therapy Volume 89 Number 6 )une 2009

 Reliability and Validity of Outcome Measures for Alzheimer Disease

who is less cognitively impaired, lated not to cognitive level, but to normal aging. Admittedly, this is
whereas this same change of approx- time to complete the TUG. Also like piecing together data cross-section-
imately 4 to 5 seconds would be our study, although their calculated ally; a longitudinal study would be
within the expected variability of ICCs were high (.91 and .92 for in- helpftil to confirm this observation
performance in a patient who is trarater and intertester reliability, re- and would be a useful contribution
more profoundly impaired. spectively), individual variability also to the literature.
was high. Using logarithmically
The SEM findings for the TUG were transformed data, the authors cre- The present study indicates that the
consistent with what was antici- ated a calculation for expected vari- TUG, the 6MWT, and gait speed
pated, with the group with a higher ability of TUG performance. This (using the GAITRite system) are reli-
level of dementia showing more vari- method revealed a large degree of able measures for use with individu-
ability of performance compared variability or measurement error, als with AD. Recently, interpreting
with the group with a lower level of such that if an individual performed change scores and identifying clini-
dementia. However, this was not the the TUG in 20 seconds, the expected cally significant changes in perfor-
case with the 6MWT or gait speed range of performance on a repeated mance have become an explicit
data. Differences in SEM between measure could be between 13.2 and focus of the physical therapy profes-
groups for the 6MWT and gait speed 30.3 seconds. If an individual's per- sion.'' Clinicians are encouraged to
were small (~10%), with the mild to formance was 30 seconds, the ex- understand how changes in scores
moderate AD group showing greater pected range of a repeated measure translate to clinical relevance. To
variability of performance than the could be between 26.4 and 60.6 sec- that end, this study presents MDCy,,
moderately severe to severe AD onds. Despite similarities in our gen- scores that provide meaningful cri-
group. Given the small difference be- eral study findings, we used substan- teria for assessing performance
tween groups, clinically, it seems tially different statistical mechanisms changes for people with AD on the
appropriate to use the SEM for all to assess absolute reliability. Our TUG, the 6MWT, and the gait speed
individuals if calculating expected findings suggest that a smaller test (Tab. 4). Minimal detectable
performance on repeated measures change in performance on the TUG change is the magnitude of change
of the 6MWT and gait speed, irre- (ie, 4.09 seconds) than proposed by that a measurement must demon-
spective of dementia level. Based on Nordin et al may represent a clini- strate to exceed the anticipated mea-
these findings, there is a 96% proba- cally significant change. Again, it is surement error and variability.''*'"' If
bility that a repeated measure of the possible that our structured and con- a change in score occurs, in either
6MWT will be within 40.5 m (133 sistent use of cuing and our efforts to direction, that is greater than MDC^,,,
ft) (2 SEMs) of the initial score. There maximize participant comfort and one can be 90% confident that the
is a 96% probability that a repeated minimize environmental stress were difference was not due to measure-
measure of gait speed will be within effective in minimizing variability of ment error or patient variability. In
11.44 cm/s (2 SEMs) of the initial performance, resulting in more con- comparison with the SEM, this pro-
measurement. These numbers give sistency across trials. vides an even more conservative es-
wide ranges of performance that timate of a change in score that is
would fall into the "expected" level Our final research goal was to iden- clinically meaningftil.
of variability on these tests, but still tify performance differences be-
could be clinically useftil in the iden- tween groups stratified by level of Rabheru52 recently published a call
tification of "tme" changes in indi- dementia. There were significant dif- for the expansion of the mechanism
viduals with AD. ferences in performance between for disease staging and milestones in
the mild to moderate AD group and people with AD, stating that al-
The TUG is the only one of the out- the moderately severe to severe AD though cognitive milestones are im-
come measures we studied that has group for the TUG, the 6MWT, and portant, functional and behavioral
previously been assessed for abso- gait speed. The findings of the milestones may help to enhance the
lute reliability. Nordin et al" studied present study, within the context of general picture of the progression of
the reliability of TUG scores with published data for the TUG,5'"'*'<y AD. The functional measures in the
participants stratified by cognitive the 6MWT,'<f.5() and gait speed'"'' present study could potentially be a
level. As in our study, they hypothe- in individuals with dementia, clearly component of the staging process.
sized that increased cognitive impair- represent a degradation of perfor- Van lersel et aH'' suggested that a
ment would increase the variability mance with the progression of de- reasonable goal of research should
of TUG scores, but they found that mentia, and this performance de- be to identify the minimal clinically
variability of performance was re- cline is beyond that seen with important changes in gait variables

)une 2009 Volume 89 Number 6 Physical Therapy 577

Reliability and Validity of Outcome Measures for Alzheimer Disease

..> population. The present

in the D are confident that the consistent and 5 Thomas VS, Hagenian PA. A preliminary
study on the reliability of physical perfor-
study suggests that if a gait speed scripted use of cuing with partici- mance measures in older day-care center
change of greater than 9.44 cm/s is pants was instrumental in the suc- clients with dementia. Int Psychogeriatr.
2002;l4:17-23.
detected in an individual with AD, cessful administration of the out-
6 Miller E, Whalen R, Biehl J, et al. The reli-
one can be 90% confident that this come measures and contributed to ability of the senior fitness test for assisted
represents a "true" change. This in- consistency of participant perfor- living dwellers with different cognitive
abilities. Paper presented at: Annual Con-
formation could be helpful in inter- mance across trials. These findings ference & Exposition of the American
preting clinical and research findings are relevant for both clinicians and Physical Therapy Association; June 3 0 -
July 3, 2004; Chicago, Illinois.
related to performance changes in researchers. It is our hope that
MDC90 values for the TUG, the 7 Rockwood K, Await E, Carver D, Mac-
gait speed. This study provides the Knight C. Eeasibility and measurement
information to make similar judg- 6MWT, and gait speed presented properties of the Functional Reach and the
here will be helpful in monitoring Timed Up and Go tests in the Canadian
ments with repeated TUG and Study of Health and Aging. J Oerontol A
6MWT scores in individuals with AD. performance changes over time and Biot Sei Med Sei. 2OOO;55:M7O-M73.
assessing the effectiveness of physi- 8 Nordin E, Rosendahl E, Lundin-Olsson L.
cal therapy and exercise interven- Timed 'Up & Go" Test: reliability in older
Limitati ons people dependent in activities of daily liv-
There were some limitations of this tions in individuals with AD. ingfocus on cognitive state. Phys Ther.
2006;86:646-655.
study. The varied clinical presenta-
9 Rockwood K, Strang D, MacKnight C,
tion of participants bodes well for Dr Ries and Dr Echternach provided con- et al. Interrater reliability of the Clinical
the generalizability of the study find- cept/idea/research design. Dr Ries provided Dementia Rating in a niulticenter trial.
J Atn Geriatr Soc. 2OOO;48:558-559.
ings, but the limited geographical writing, data collection and analysis, project
10 van Iersel MB, Benraad CE, Rikkert MG.
(northern Virginia) and socioeco- management, participants, facilities/equip-
Validity and reliability of quantitative gait
nomic (supper middle class) variabil- ment, institutional liaisons, and clerical sup- analysis in geriatric patients with and with-
port. All authors provided consultation (in- out dementia. J Am Geriatr Soc. 2007;
ity of tne group may threaten the 55:632-634.
cluding review of the manuscript before
external validity of the study. The submission). 11 Domholdt E. Rehabititation Research:
logistics of using a single researcher Principies and Applications. 3rd ed. St
for data collection influenced sample This research project was approved by the Louis, MO: Elsevier Saunders; 2005.
Nova Southeastern University Institutional 12 Rothstein JM, Echternach JL. Primer on
size and made it impossible to blind Review Board (Research Protocol No. Measurement: An introductory Guide to
the SCO!'er to the dementia level of HPD-ALL08280604Exp). Measurement Issues. Alexandria, VA:
the patient, which would be ideal in American Physical Therapy Association;
This study was conducted in partial fulfill- 1993.
this type of study. ment of the requirements for Dr Ries' PhD 13 Ponney LG, Watkins MP. Eoundations of
degree in physical therapy from Nova South- Clinical Research: Applications to Prac-
tice. 2nd ed. Upper Saddle River, NJ: Pren-
Conclusions eastern University. tice Hall Health; 2000.
This study demonstrated excellent A poster presentation of this research was 14 Haley SM, Fragala-Pinkham MA. Interpret-
test-retest reliability for the TUG, the given at the Combined Section Meeting of ing change scores of tests and measures
used in physical therapy. Phys Ther.
6MWT, and gait speed in individuals the American Physical Therapy Association; 2006;86:735-743.
w^ith AD. Significant differences February 9-12, 2009; Las Vegas, Nevada. 15 Reisberg B. Functional assessment staging
were found in performance of these This article was received August 23, 2008, and (FAST). Psychopharmacol Bull 1988;24:
outcome measures between the mild was accepted March 9, 2009. 653-659.
16 Reisberg B, Ferris SH, Franssen E. An ordi-
to moderate AD group and the mod- DOI: 10.2522/ptj.20080258 nal functional assessment tool for Alzhei-
erately severe to severe AD group, mer's-type dementia. Hosp Community
Psychiatry. 1985;36:593-595.
with the participants who were less
cognitiveiy impaired outperforming References 17 Reisberg B, Ferris SH, Anand R, et al. Func-
tional staging of dementia of the Alzhei-
1 2008 Alzheimer's disease facts and figures.
those who were more cognitiveiy Alzheimers Dement. 2OO8;4:11O-133. mer type. Ann NY Acad Sei. 1984;435:
481-483.
impaired. Despite very high ICCs for 2 Rolland Y, Pillard F, Klapouszczak A, et al.
18 Reisberg B, Ferris SH, Franssen E. Practical
test-retest reliability, there was nota- Exercise program for nursing home resi-
geriatrics: an ordinal functional assess-
dents with Alzheinier's disease: a 1-year
ble individual variability in the per- randomized, controlled trsX.JAm Geriatr ment tool for Alzheimer's-type dementia.
Hosp Community Psychiatry. 1985;36:
formance of these measures. Presen- Soc. 2007;55:158-l65.
593-595.
tation of SEM and MDCgo, for each of 3 Santana-Sosa E, Barriopedro MI, Lopez- 19 Sclan SG, Reisberg B. Functional assess-
Mojares LM, et al. Exercise training is ben-
the measurement tools, provides cli- eficial for Alzheimer's patients. IntJ Sports ment staging (FAS'O in Alzheimer s dis-
ease: reliability, validity, and ordinality. Int
nicians with meaningful thresholds Med 2008;29:845-850.
Psychogeriatr. 1992;4(suppl l):55-69.
for identifying changes beyond those 4 Tappen RM, Roach KE, Bchner D, et al.
Reliability of physical performance mea-
expected from measurement error sures in nursing hoine residents with Alz-
and individual variability (ie, "true" heimer's disease. / Gerontot A Biol Sei
Med Sei. 1997;52:M52-M55.
change) in individuals with AD. We

578 Physical Therapy Volume 89 Number 6 june 2009

 Reliability and Validity of Outcome Measures for Alzheimer Disease

20 Staples S. Alzheimer's disease: modifying 31 Beck CK, Heacock P, Rapp CG, Mercer 42 Reed R, Yochum K, Alcott S, Meredith K.
instructions and approach to enhance per- SO. Assisting cognitively impaired elders Gait characteristics of subjects with senile
formance. Paper presented at: Combined with activities of daily living. American dementia of the Alzheimer's type |al>
Sections Meeting of the American Physical Journat of Atzheimers Care and Retated stract|.y/lH7 Geriatr Soe. 1992;4O:SA79.
Therapy Association; February 23-27, Disorders & Research. 1993;8(6): 11-20.
2005; New Orleans, Louisiana. 43 Tanaka A, Oktiztimi H, Kobayashi I, et al.
32 Podsiadlo D, Richardson S. The timed 'Up Gait disturbance of patients with vascular
21 Haak NJ. Maintaining connections: under- & Go": a test of basic functional mobility and Alzheimer-type dementias. Percept
standing communication from the per- for frail elderly persons. I Am Geriatr Soe. Mot Skitts. I995;8O(3 Pt l):735-738.
spective of the person with dementia. I991;39:I42-148.
Atzheimers Cave Quarterty. 2002;3: 44 van lersel MB, Hoefsloot W, Munneke M,
33 Guyatt GH, Sullivan MJ, Thompson PJ, et al. Systematic review of qtiantitative
116-131.
et al. The 6-minute walk: a new meastire of clinical gait analysis in patients with de-
22 Kovach CR, Henschel H. Planning activi- exercise capacity in patients with chronic mentia. Z Gerontot Geriatr. 2OO4;37:
ties for patients with dementia: a descrip- heart failure. Can Med AssoeJ. 1985;132: 27-32.
tive study of therapeutic activities on spe- 919-923.
45 Visser H. Gait and balance in senile demen-
cial care units, y Gerontot Nurs. 1996;22: 34 Harada ND, Chiu V, Stewart AL. Mobility- tia of Alzheimer's type. Age Ageing.
33-38. related function in older adults: assess- 1983;12:296-301.
23 Patterson JT, Wessel J. Strategies for re- ment with a 6-niinute walk test. Arch Phys
46 Stratford PW. Getting more from the liter-
training functional movement in persons MedRetjabit. 1999;80:837-84l. ature: estimating the standard error of
with Alzheimer disease: a review. Phys- measurement from reliability studies.
35 Lord SR, Menz HB. Physiologic, psycho-
iother Can. 2002;54:274 -280. logic, and health predictors of 6-niinute Physiother Can. 2OO4;56:27-3O.
24 Auer SR, Sclan SG, Yaffee RA, Reisberg B. walk performance in older people. Arch
The neglected half of Alzheimer disease: Phys Med Retjabit. 2002;83:9O7-911. 47 Hauer K, Oster P. Measuring functional
cognitive and functional concomitants of performance in persons with dementia.
severe dementia, y/4;n Geriatr Soe. 1994; 36 Bilney B, Morris M, Webster K. Concur- J Am Geriatr Soe. 2008;56:949-950.
42:1266-1272. rent related validity of the GAITRite walk-
way system for qttantication of the spatial 48 Steffen TM, Hacker TA, Mollinger L. Age-
and temporal parameters of gait. Gait Pos- and gender-related test performance in
25 Davis CM. The role of the physical and
ture. 2003; 17:68-74. community-dwelling elderly people: Six-
occupational therapist in caring for the
Minute Walk Test, Berg Balance Scale,
victim of Alzheimer's disease. Phys Oeeup
37 Menz HB, Latt MD, Tiedemann A, et al Timed Up & Go Test, and gait speeds.
Ther Geriatr. 1986;4(3):15-28. Reliability of the GAITRite walkway sys- Phys Wer. 2002;82:128-137.
26 Hoppes S, Davis LA, Thompson D. Envi- tem for the quantication of teniporo-
ronmental effects on the assessment of spatial parameters of gait in yoting and 49 Toulotte C, Fabre C, Dangremont B, et al.
people with dementia: a pilot study. AinJ older people. Gait Posture. 2004;20: Effects of physical training on the physical
Occup Ther 2003;57:396-402. 20-25. capacity of fniil, demented patients with a
history of falling: a randomised controlled
27 Davis LA, Hoppes S, Chesbro SB. 38 Alexander NB, Mollo JM, Giordani B, et al. trial. Age Ageing. 2003;32:67-73.
Cognitive-communicative and indepen- Maintenance of balance, gait patterns, and
dent living skills assessment in individuals obstacle clearance in Alzheimer's disease. 50 Roach KE, Tappen RM. Factors related to
with dementia: a pilot study of environ- Neurotogy. 1995;45:908-9l4. decline in ambulation ability in institution-
mental impact. Top Geriatr Rehabit. alized individuals with Alzheimer's disease
2005;21:136-l43. 39 MeroryJR, Wittwer JE, Rowe CC, Webster [abstractl.y Geriatr Phys TIjer. 2004;27:
KE. Quantitative gait analysis in patients 119-120.
28 Souren LE, Franssen EH, Reisberg B. Neu- with dementia with Lewy bodies and Alz-
roniotor changes in Alzheimer's disease: 51 Verghese J, Wang C, Upton RB, et al.
heimer's disease. Gait Posture. 2007;26: Quantitative gait dysfunction and risk of
implications for patient care. / Geriatr 414-419.
Psychiatry Neurot. 1997; 10:93-98. cognitive decline and dementia. J Neurot
40 Nakamura T, Meguro K, Yamazaki H, et al. Neurosurg Psychiatry. 2007;78:929-935.
29 Oddy R. Promoting mobility in patients Postural and gait disturbance correlated
with dementia: some suggested strategies with decreased frontal cerebral blood flow 52 Rabhen.1 K. Disease staging and mile-
stones. Can J Neurot Sei. 2OO7;34(Suppl
for physiotherapists. Physiother Pract. in Alzheimer disease. Atzheimer Dis Assoc 1):S62-S66.
1987;3:18-27. Disord. 1997;11:132-139.
30 Vogelpohl TS, Beck CK, Heacock P, Mer- 41 O'Keeffe ST, Kazeeni H, Philpott RM, et al.
cer SO. "I can do it!" Dressing: promoting Gait disturbance in Alzheimer's disease: a
independence through individualized clinical study. Age Ageing. 1996;25:
strategies. J Gerontot Nurs. 1996;22(3): 313-316.
39-42; quiz 48.

)une 2009 Volume 89 Number 6 Physical Therapy 579