Underlying Structure of the National Institutes of

Health Stroke Scale

Results of a Factor Analysis
Patrick Lyden, MD; Mei Lu, PhD; Christy Jackson, MD; John Marler, MD; Rashmi Kothari, MD;
Thomas Brott, MD; Justin Zivin, MD, PhD; and the NINDS tPA Stroke Trial Investigators

Background and Purpose—No stroke scale has been validated as an outcome measure using data from a clinical trial
demonstrating a positive therapeutic effect. Therefore, we proposed to use data from the National Institute of
Neurological Disorders and Stroke (NINDS) tPA Stroke Trial to determine whether the National Institutes of Health
Stroke Scale (NIHSS) was valid in patients treated with tissue plasminogen activator (tPA) and to explore the underlying
clinimetric structure of the NIHSS.
Methods—We performed an exploratory factor analysis of NIHSS data from Part 1 (n5291) of the NINDS tPA Stroke
Trial to derive a hypothesized underlying factor structure. We then performed a confirmatory factor analysis of this
structure using NIHSS data from Part 2 of the same trial (n5333). We then tested whether this final factor structure
could be found in tPA- and placebo-treated patients serially over time after stroke treatment. Using 3-month outcome
data, we tested for an association between the NIHSS and other measures of stroke outcome.
Results—The exploratory analysis suggested that there were 2 factors underlying the NIHSS, representing left and right
brain function, confirming the content validity of the scale. An alternative structure composed of 4 factors could be
derived, with a better goodness of fit: the first 2 factors could represent left brain cortical and motor function,
respectively, and the second 2 factors could represent right brain cortical and motor function, respectively. The same
factor structures were then found in tPA and placebo patient groups studied serially over time, confirming the
exploratory analysis. All 3-month clinical outcomes were associated with each other at subsequent time points,
confirming predictive validity.
Downloaded from http://ahajournals.org by on June 7, 2021

Conclusions—This is the first study of the validity of a stroke scale in patients treated with effective stroke therapy. The
NIHSS appeared to be valid in patients with acute stroke and for finding treatment-related differences. The scale was
valid when used serially over time after stroke, up to 3 months, and showed good agreement with other measures of
outcome. (Stroke. 1999;30:2347-2354.)
Key Words: cerebrovascular disorders n clinimetrics n factor analysis, statistical n neuropsychological tests

D emonstration of effective stroke therapy requires a

standardized measurement of outcome.1–7 For use in
multicenter stroke treatment trials, a rating scale should be
reliable, valid, and time efficient.8,9 Currently, no stroke scale
fulfills all the requirements of rigorous psychometric scale
design; however, a few scales, including the National Insti-
tutes of Health Stroke Scale (NIHSS), are generally accepted
because of their simplicity and relatively rigorous design.5–
7,10,11 The NIHSS contains 15 items, including level of
consciousness, eye movement, visual field deficit, and motor
and sensory involvement.12–14 Scale items are scored by
degree of severity using weighted scores; reliability has been
demonstrated for neurologists, other physicians, and nonphy-
sicians.13–18 The items contained in the NIHSS were selected
on the basis of expert opinion and literature review, thus
satisfying the requirements for content validity.1,4,5,7 The
NIHSS correlates with lesion volume, suggesting that the
NIHSS can be used to estimate current clinical status (con-
current criterion validity) on the basis of 1 complementary
outcome measure.19,20 The scale correlates with alternative
measures of neurological outcome, such as Activities of Daily
Living (ADL) scales, and other deficit scales, further suggest-
ing concurrent criterion validity.21 Another type of criterion
validity, known as predictive validity, is demonstrated by
using a scale to predict future health status; this type of
validity has not previously been demonstrated for the NIHSS.

Received September 28, 1998; final revision received July 27, 1999; accepted July 27, 1999.
From the Department of Neurology, Veterans Administration Medical Center, San Diego, Calif, and Department of Neurosciences, University of
California at San Diego School of Medicine (P.L., C.J., J.Z.); Department of Biostatistics and Research Epidemiology, Henry Ford Health Science Center,
Detroit, Mich (M.L.); National Institute of Neurological Disorders and Stroke, Bethesda, Md (J.M.); and Departments of Emergency Medicine and
Neurology, University of Cincinnati Medical Center (Ohio) (R.K., T.B.).
A list of all NINDS tPA Trial Investigators is found in the Appendix.
Correspondence to Patrick D. Lyden, MD, Stroke Center (8466), 3rd Floor, OPC, Suite 3, 200 W Arbor Dr, San Diego, CA 92103-8466.
© 1999 American Heart Association, Inc.
Stroke is available at http://www.strokeaha.org

2347
 2348 Stroke November 1999
Another approach to validity is to explore the internal
structures or dimensions underlying a scale.2–5,22,23 This
sort of data demonstrates the construct validity of a scale:
a valid scale should measure 1 or a small number of
underlying constructs. We sought to study the underlying
structure of the NIHSS using factor analysis, a widely
accepted method for deriving the internal structure of a
scale.5,22,24 –30 Such factors should reflect biological phe-
nomena that make sense to the investigator, such as right
or left hemispheric function. Scale items that do not
contribute to such factors can be eliminated, thus simpli-
fying the scale and improving its internal reliability.22 We
also desired to determine whether the scale structure
identified at baseline was independent of therapy and time,
that is, we questioned whether the same dimensions could
be identified at later time points and in patients who
received tissue plasminogen activator (tPA) compared with
placebo. This property of a scale is critical; if the structure
underlying the scale differs between 2 treatment groups,
then the scale is invalid for treatment studies because scale
scores would not be comparable between treatment groups.
Essentially, 1 group would be tested with 1 version of the
scale, and the other group would be tested with a different
version. The National Institute of Neurological Disorders
and Stroke (NINDS) tPA Stroke Trial afforded an appro-
priate opportunity for studying this clinimetric property,
since the 2 treatment groups differed significantly.31
Subjects and Methods
Downloaded from http://ahajournals.org by on June 7, 2021
The NINDS tPA Stroke Trial results and the methods used to ensure
reliable use of the scale have been published.13,31 To participate in
the NINDS tPA Stroke Trial, the investigators underwent NIHSS
certification by viewing training and testing videotapes.13,14 Recer-
tification, using another tape, was conducted every 6 months. Each
patient in the trial was scored by a trained, certified investigator at
the time of initial stroke evaluation (baseline) and 2 hours, 24 hours,
7 to 10 days, and 3 months later.
To explore the structure underlying the NIHSS, a factor analysis
was conducted.24 The NINDS tPA Stroke Trial was conducted in 2
parts using nearly identical methods: We elected to use baseline
(pretreatment) NIHSS scores from part 1 (291 patients randomly
divided between treatment and placebo) for an exploratory factor
analysis. On the basis of the findings of the exploratory analysis, we
planned a confirmatory factor analysis using the baseline part 2 data
(333 patients randomly divided between treatment and placebo).
The purpose of factor analysis is to describe and explain a large set
of independent variables in terms of a few underlying new variables,
called factors. If a variable correlates well with the factor, it is said
to “load” on that factor. By studying the factor loadings, which is
interpreted as a correlation coefficient, one can determine how well
the factors explain the data. A group of items in an outcome scale
may represent any number of underlying factors, from a single factor
to the total number of items. In the latter case, each item represents
a unique factor, which is an undesirable property for outcome scales.
In general, an ideal scale represents a small number of underlying
factors.
To gain an initial estimate of how many factors may underlie the
NIHSS, we examined the Scree plot (Figure 1). Once the initial
number of factors was selected, a factor analysis was conducted, and
we examined the factor loadings on each respective factor. We
assessed the goodness of fit of the factor structure using Bentler’s
Comparative Fit Index (CFI).32,33 CFI ranges from 0 to 1 and is
viewed as the percentage of variation of the observed measure (the
scale items) explained by a given structure (such as R2 in a regression
model); values .0.90 indicate excellent goodness of fit.32 Several
factor structures were examined in this way until we obtained the
best solution, defined as the structure that had reasonable goodness
of fit (CFI .0.90) and made clinical sense.
A confirmatory factor analysis on a new data set (baseline scores
in part 2) was conducted to validate the factor structure identified in
the previous, exploratory analyses. CFI was calculated on the basis
of the new data; the structure was considered valid if it had
reasonable goodness of fit and was consistent with that seen in the
previous analyses. Data collected after placebo or tPA treatment
were used to determine whether the factor structure identified from
baseline data were independent of time and tPA therapy. The data
(placebo or tPA) could be used to identify a new factor structure if,
in fact, the factor structure depends on time and/or treatment.
Patients from parts 1 and 2 were analyzed together for this aspect of
the study to allow for greater power. In this part of the analysis only,
patients who died, missed 1 of the follow-up NIHSS examinations,
or had a scale item recorded as unknown were excluded from the
specific analysis involving the missing datum. The factor structure
would be considered independent of time or therapy if the CFI
goodness of fit at each time point was .0.90 using the same factor
structure derived from baseline data.32,33
In addition, the exploratory and confirmatory analyses were
conducted, including 15 NIHSS items and 2 extra items regarding
distal motor function in the left arm or the right arm. These distal
motor items were attached to the scale at the time the trial was begun
but were never validated, in response to the criticism that the NIHSS
did not measure distal limb strength.7,31
To assess predictive validity, the associations between the NIHSS
at each time point and late outcome at 3 months, as measured by
Barthel Index, Glasgow Outcome Scale, and Rankin Scale, were
calculated with Spearman rank correlation coefficients. We expected
significant correlation between the NIHSS at several time intervals
and the 3-month clinical outcomes.
Results
There are several versions of the NIHSS; the stroke scale and
item labeling used in this study are shown in Table 1. We
obtained valid baseline scores on 284 of 291 part 1 patients
for the exploratory analyses and 331 of 333 part 2 patients for
the confirmatory analyses; invalid scores were either incom-
plete or missing. The demographic data for these patients are
contained in our prior report.31 There were similar numbers of
lacunar, large-vessel, and cardioembolic strokes in the 2
parts, and there was a similar distribution of subtypes be-
tween tPA treatment and placebo.
The initial exploratory analysis of the part 1 baseline
scores was unstructured and yielded the Scree plot in
Figure 1. From the plot, it is clear that 2 factors account for
the majority of the variance in the data. The first 2 factors
explained 88% and the first 4 factors explained 100% of
the variance in the data; the 4-factor solution was consid-
ered further. The ataxia item loaded weakly (loadings
#0.40) on all factors in the exploratory phase, and
therefore this item was excluded; it did not correlate with
any factors underlying the scale. The consciousness item
(item 1A) loaded on all 4 factors with equal loads #0.4.
Similarly, item 4, facial palsy, exhibited loading values of
,0.40, suggesting that it contributed little to the scale and
could be excluded from the analysis. The final exploratory
4-factor solution using the remaining 12 items produced
the best goodness of fit (CFI50.96). Table 2 lists each
factor and the loading of each variable on each factor. The
final column in Table 2, R2 (also called communality),
represents the percentage of variance in the variable that is
explained by all the factors. Additional variance could be
 Lyden et al NIHSS Factor Analysis 2349

TABLE 1. Current Form of the NIHSS TABLE 1. Continued

Item Name Response
1a Level of consciousness 05Alert
15Not alert, arousable
25Not alert, obtunded
35Unresponsive
1b Questions 05Answers both correctly
15Answers one correctly
25Answers neither correctly
1c Commands 05Performs both tasks correctly
15Performs one task correctly
25Performs neither task
2 Gaze 05Normal
15Partial gaze palsy
25Total gaze palsy
3 Visual fields 05No visual loss
15Partial hemianopsia
25Complete hemianopsia
35Bilateral hemianopsia
4 Facial palsy 05Normal
15Minor paralysis
25Partial paralysis
35Complete paralysis
5a Left motor arm 05No drift
15Drift before 10 seconds
25Falls before 10 seconds
Downloaded from http://ahajournals.org by on June 7, 2021
Item Name Response
9 Language 05Normal
15Mild aphasia
25Severe aphasia
35Mute or global aphasia
10 Dysarthria 05Normal
15Mild
25Severe
11 Extinction/inattention 05Normal
15Mild
25Severe
attributed to other sources, such as interindividual varia-
tion or examiner-to-examiner error. From Table 2, it is
apparent that the first factor relates to language function,
since the aphasia and level-of-consciousness items load
most heavily. The second factor is difficult to interpret
because it includes the gaze, neglect, visual field, and
sensory items. We suspect that a large right hemisphere
cortical lesion would impair these items and thus underlie
this factor. Factors 3 and 4 clearly represent right and left
brain motor functions, respectively.
We used the baseline data from part 2 for the confirmatory
analysis, excluding the level-of-consciousness, face palsy,
and ataxia items. The results of the confirmatory analysis on

35No effort against gravity the remaining 12 items (Table 3) again showed excellent
goodness of fit (CFI50.93). The factor structure is identical
45No movement
to that of the exploratory analysis: factor 1 appears to
5b Right motor arm 05No drift
represent left cortical function, and factor 2 represents right
15Drift before 10 seconds
cortical function. Factors 3 and 4 again seem to represent left
25Falls before 10 seconds and right brain motor function, respectively.
35No effort against gravity To determine whether the scale is valid within treatment
45No movement groups, we repeated the analyses in tPA- and placebo-
6a Left motor leg 05No drift treated patients using data obtained 2 and 24 hours, 7 to 10
15Drift before 5 seconds days, and 3 months after stroke. The resulting goodness-
25Falls before 5 seconds of-fit statistics are shown in Table 4. The factor structures
35No effort against gravity (ie, loadings) were identical to that at baseline (data not
shown). The consistency of the structure and the goodness-
45No movement
of-fit statistics over time and treatment (Table 4) suggest
6b Right motor leg 05No drift
that the scale remains valid, regardless of time from stroke
15Drift before 5 seconds
onset or treatment given.
25Falls before 5 seconds To assess the predictive validity of the NIHSS using
35No effort against gravity alternative scales, we compared the NIHSS over time with
45No movement the 3-month outcome using the Barthel Index, Rankin
7 Ataxia 05Absent Scale, and Glasgow Outcome Scale. The correlations
15One limb between the scale and the other clinical outcomes were
25Two limbs significant (P,0.001; Table 5) but modest in magnitude at
8 Sensory 05Normal baseline and 2 hours after stroke. The correlation between
the NIHSS at baseline and the measures at 90 days
15Mild loss
demonstrates predictive validity. The absolute values of
25Severe loss
the correlation coefficients were greater for the later
measurements, suggesting that after 2 hours from stroke,
the NIHSS values may have greater predictive validity
with respect to the 3-month outcome.
 2350 Stroke November 1999

Figure 1. From the initial set of data, an exploratory factor analysis yields the eigenvalues associated with the data. A Scree plot
shows the eigenvalues plotted over the number of factors that could be extracted from the data set. The plot shows that 2 fac-
tors, with eigenvalues .2.5, account for the majority of the variance in the data set. However, the third and fourth factors,
Downloaded from http://ahajournals.org by on June 7, 2021

although ,1.0, contribute to more of the variance than the remaining factors. Thus, we selected 2-, 3-, and 4-factor solutions for
further study.

Discussion 2-factor solution (Tables 2 and 3). Each of the hemisphere

We explored the NIHSS to identify the constructs, or factors,
underlying the scale.22 The Scree plot (Figure 1) suggested 2
factors, and the underlying structure analysis suggested that
these 2 factors related to the functions of the 2 cerebral
hemispheres. The 4-factor solution, which actually resulted in
better goodness of fit, may be viewed as a refinement of the
factors seemed to resolve into 2 subfactors, one representing
cortical and the other representing motor function (Figure 2).
Our data may also be interpreted as confirmation of the
construct validity of the scale, since it was designed to detect
and measure deficit in either cerebral hemisphere.12 Brain
stem deficits fail to appear as a separate factor here, perhaps

TABLE 2. Exploratory 4-Factor Solution of the 12-Item Scale TABLE 3. Confirmatory 4-Factor Solution of the 12-Item Scale
2
Item Factor 1 Factor 2 Factor 3 Factor 4 R Item Factor 1 Factor 2 Factor 3 Factor 4 R2
Language 0.97 0.93 Language 0.95 0.91
LOC questions 0.84 0.71 LOC questions 0.84 0.71
LOC commands 0.77 0.59 LOC commands 0.70 0.49
Gaze 0.71 0.51 Gaze 0.67 0.45
Visual 0.69 0.48 Extinction/neglect 0.71 0.50
Extinction/neglect 0.69 0.48 Visual 0.75 0.57
Sensory 0.60 0.36 Sensory 0.67 0.45
Right arm 0.98 0.97 Right arm 0.97 0.94
Right leg 0.89 0.79 Right leg 0.90 0.80
Dysarthria 0.49 0.24 Dysarthria 0.49 0.24
Left arm 1.00 1.00 Left arm 1.00 1.00
Left leg 0.87 0.75 Left leg 0.77 0.60
Goodness of fit CFI50.96 Goodness of fit CFI50.93
LOC indicates level of consciousness. LOC indicates level of consciousness.
 Lyden et al NIHSS Factor Analysis 2351

TABLE 4. Stability of the 12-Item NIHSS Over Time

and Treatment
4-Factor Structure

Time Placebo tPA

2 hours 0.92 0.88
24 hours 0.94 0.93
7–10 days 0.93 0.93
3 months 0.92 0.93
Figure 2. The 4-factor solution can be viewed as a subset or
refinement of the 2-factor solution. GOF indicates goodness
of fit.
because of the low numbers of such patients included in our
data set (,15% of the subjects). Thus, the factor structure we administrations, when different examiners were used for
determined here may not apply in studies that contain large patients seen up to 3 months after stroke, the same factors
numbers of patients with brain stem events. were found. These repeated confirmations suggest that the
Factor analysis depends heavily on the specific data set 4-factor solution is likely to be found in future study
used; the factor structure we obtained may not necessarily be
populations.
found in another set of patients.22 To obtain some assurance
Our data provide a unique opportunity to explore the
that the exploratory results are generalizable, a second,
response of the scale to treatment effects, since the treated
confirmatory analysis was required. As shown in Table 3, that
group differed significantly from the placebo group. This
analysis confirmed the exploratory study: the NIHSS con-
is an important property of the scale, for if the factor
tained 4 factors corresponding to the 2 cerebral hemispheres,
analysis showed that the scale behaved differently in
with loadings essentially identical to the exploratory analysis
(Table 3). In interpreting these results, it is important to recall placebo- versus tPA-treated groups, the scale could not be
that the 2 parts of the NINDS trial were conducted in used to measure outcome in further treatment trials. In
sequence by investigators who were blinded to the results of such an event, the NIHSS would become essentially 2
part 1 during part 2; the protocol was essentially unchanged different scales, 1 for placebo- and 1 for tPA-treated
during the 2 parts.31 The confirmatory analysis is somewhat patients, an untenable state. From our study, it is clear that
Downloaded from http://ahajournals.org by on June 7, 2021

limited, however, by the fact that the 2 study populations
were quite homogeneous. In a future study, using a different
study population, it is very likely, but not certain, that the
internal structure of the scale would be the same as we
determined here.
Further confirmation of the robustness of these findings
comes from the analyses conducted on the data collected
serially after stroke treatment (Table 4). On repeated
the NIHSS clearly reports deficits in placebo- and tPA-
treated patients in a like manner (Table 4). The 4-factor
solutions in the 2 treatment groups were similar in the
exploratory and confirmatory analyses. The consistency of
these repeated analyses suggests that the internal scale
structure is the same in patients who receive tPA or
placebo. Furthermore, our prior report clearly showed that
the scale reported true differences between the groups.31

TABLE 5. Correlation Coefficients of Outcome Measures Serially Over Time

After Randomization
NIHSS at Barthel at Rankin at GOS at
90 Days 90 Days 90 Days 90 Days
Placebo-treated patients
NIHSS at baseline 0.55 20.48 0.51 0.49
NIHSS at 2 hours 0.62 20.58 0.61 0.60
NIHSS at 24 hours 0.78 20.72 0.73 0.71
NIHSS at 7–10 days 0.85 20.78 0.81 0.79
NIHSS at 90 days 20.83 0.86 0.83
tPA-treated patients
NIHSS at baseline 0.45 20.51 0.56 0.56
NIHSS at 2 hours 0.67 20.65 0.70 0.68
NIHSS at 24 hours 0.76 20.77 0.82 0.80
NIHSS at 7–10 days 0.85 20.79 0.86 0.82
NIHSS at 90 days 20.79 0.87 0.86
GOS indicates Glasgow Outcome Scale. Correlation coefficients were significant from zero
(P,0.001).
 2352 Stroke November 1999
Taken together, these findings confirm the utility of the
scale as an outcome measure, its most important function
in large clinical trials of putative therapies. However, this
assertion will be stronger when a factor analysis yields
similar results in another trial of a different, also effica-
cious, compound.
The ataxia item did not correlate with any factor in the
structures we examined. The variance in this question
exceeded the variance attributable to the factors, suggest-
ing that this item is either a unique factor or unreliable in
its administration. We favor the latter explanation because
this item has been shown to have very low reproducibility
in some studies13,18,21,34 but was reliable in the Trial of
ORG 10172 in Acute Stroke Treatment (TOAST) certifi-
cation study.14 In a series of videotaped patients, multiple
examiners could not clearly grade degrees of ataxia.13
Using rating scale analysis of the predecessor of the
NIHSS, the University of Cincinnati Stroke Rating Scale,
a study of rehabilitation inpatients showed results similar
to ours34: ataxia was found to be an unreliable item that
contributed little to the scale; reliability measures im-
proved when this item was deleted. The ataxia item was
included in earlier versions of the scale to increase the
sensitivity for detecting brain stem deficits, but only a
small number of brain stem strokes occurred in our study
group. If our sample is highly representative of the
distribution of lesions typically entered into multicenter
acute therapeutic trials, then it is likely that future stroke
trials will likewise contain few brain stem strokes. Delet-
Downloaded from http://ahajournals.org by on June 7, 2021
ing ataxia from the scale will not affect validity and may
increase reliability of the NIHSS when used acutely (data
not shown). Similarly, the items relating to level of
consciousness and facial palsy also exhibited smaller
loadings in the acute phase, as well as poor reliability in
our prior study, and were deleted in the confirmatory
analyses. The facial item exhibited poor reliability in a
study of the Unified Stroke Scale.2 The sensory and
dysarthria items showed moderate loadings, but low com-
munalities, throughout our analyses, consistent with their
known poor reliability.21 It would be appropriate to col-
lapse these items into responses with fewer choices or to
eliminate them in a future version of the scale. Our data
suggest further, however, that the elimination of the
dysarthria item would not change the predictive validity of
the NIHSS, given its loadings of 0.49 (Tables 2 and 3).
Finally, our data showed that the unvalidated item 12
(distal motor function) should be deleted from the scale
because it contributes little to the measurement of the
structures underlying the scale. This is true, despite the
common assertion that distal limb function should be
measured in addition to proximal limb function.
It may seem inconsistent that the questions concerning
level of consciousness (items 1b and 1c) load on the left
hemisphere factor, but this result likely reflects the role
that language plays in clinical assessment of these items.
Although these questions are intended to measure level of
consciousness, in fact they depend heavily on the presence
of intact comprehension. Similarly, the visual field ques-
tion (item 3) loads heavily on the right hemisphere factor,
although there were an equal number of patients with left-
and right-sided visual field deficits (data not shown).
Patients with left cerebral stroke, because of aphasia, may
be more difficult to evaluate, and visual field testing may
not be reliable. Right brain stroke patients may exhibit
neglect and could therefore appear to exhibit a visual field
deficit. Both phenomena might work to enhance the
correlation of the visual field item with the right hemi-
sphere factor, although other explanations of this phenom-
enon might be explored.
Factor analysis has been used previously to study the
inherent properties of other scales. Wade and Hewer35 re-
ported finding 2 factors underlying the Barthel Index admin-
istered 6 months after stroke, but .67% of the variance in the
data was explained by the first factor. This study suggested
that the Barthel Index measured only 1 underlying construct,
functional independence. A similar analysis of the Fugl-
Meyer recovery scale and 3 measures of functional indepen-
dence was performed on data obtained in the first week after
stroke.36 Again, although 3 factors could reasonably be
extracted from the data, the first factor explained .80% of
the variance. In this study the recovery scale and all 3
functional independence measures correlated very highly
with each other.36 Factor analysis has been used to derive the
structure underlying global outcome and ADL scales.27,29,30
In a study of 1328 patients with presumed transient ischemic
attack, factor analysis determined the correlation of various
symptoms with vascular territories and neurologists’ pre-
sumptive localization.26 A caregiver burden scale was factor
analyzed to derive key dimensions.28
We identified 2 underlying constructs of the NIHSS when
used in the first 24 hours after stroke. These 2 constructs seem
to reflect the function of the 2 cerebral hemispheres, confirm-
ing the construct validity of the scale. Most importantly, the
internal scale structure appears to remain consistent in treated
and placebo groups and when administered serially over time.
These findings support the validity of the scale for use in
future treatment trials as an outcome measure. Some NIHSS
items were found to exhibit poor concordance with the 2
constructs and with other items in the scale; when combined
with our prior investigations of the scale’s clinimetric prop-
erties, these results suggest that it may be possible to simplify
the NIHSS. A proposal for a simplified scale will be the
subject of a future publication.
Appendix
The following persons and institutions participated in the NINDS
tPA Stroke Trial: Clinical Centers: University of Cincinnati (150
patients): Principal Investigator: T. Brott; Co-investigators: J.
Broderick, R. Kothari; M. O’Donoghue, W. Barsan, T. Tomsick;
Study Coordinators: J. Spilker, R. Miller, L. Sauerbeck; Affiliated
Sites: St Elizabeth Hospital (South), J. Farrell, J. Kelly, T. Perkins,
R. Miller; University Hospital, T. McDonald; Bethesda North
Hospital, M. Rorick, C. Hickey; St Luke Hospital (East), J.
Armitage, C. Perry; Providence Hospital, K. Thalinger, R. Rhude;
The Christ Hospital, J. Armitage, J. Schill; St Luke Hospital (West),
P.S. Becker, R.S. Heath, D. Adams; Good Samaritan Hospital, R.
Reed, M. Klei; St Francis/St George Hospital, A. Hughes, R. Rhude;
Bethesda Oak Hospital, J. Anthony, D. Baudendistel; St Elizabeth
Hospital (North), C. Zadicoff, R. Miller; St Luke’s Hospital–Kansas
City, M. Rymer, I. Bettinger, P. Laubinger; University of California,

San Diego (146 patients): Principal Investigator: P. Lyden; Co-
investigators: J. Dunford, J. Zivin; Study Coordinators: K. Rapp, T.
Babcock, P. Daum, D. Persona; Affiliated Sites: University of
California, San Diego, M. Brody, C. Jackson, S. Lewis, J. Liss, Z.
Mahdavi, J. Rothrock, T. Tom, R. Zweifler; Sharp Memorial
Hospital, R. Kobayashi, J. Kunin, J. Licht, R. Rowen, D. Stein;
Mercy Hospital, J. Grisolia, F. Martin; Scripps Memorial Hospital,
Chaplin, N. Kaplitz, J. Nelson, A. Neuren, D. Silver; Tri-City
Medical Center, T. Chippendale, E. Diamond, M. Lobatz, D.
Murphy, D. Rosenberg, T. Ruel, M. Sadoff, J. Schim, J. Schleimer;
Mercy General Hospital, Sacramento, R. Atkinson, D. Wentworth,
R. Cummings, R. Frink, P. Heublein; San Diego Veterans Admin-
istration Medical Center. University of Texas Medical School,
Houston (104 patients): Principal Investigator: J.C. Grotta; Co-
investigators: T. DeGraba, M. Fisher, A. Ramirez, S. Hanson, L.
Morgenstern, C. Sills, W. Pasteur, F. Yatsu, K. Andrews,
C. Villar-Cordova, P. Pepe; Study Coordinators: P. Bratina, L.
Greenberg, S. Rozek, K. Simmons; Affiliated Sites: Hermann
Hospital; St Luke’s Episcopal Hospital; Lyndon Baines Johnson
General Hospital; Memorial Northwest Hospital; Memorial South-
west Hospital; Heights Hospital; Park Plaza Hospital; Twelve Oaks
Hospital; Houston Fire Department Emergency Medical Services.
Long Island Jewish Medical Center (72 patients): Principal Investi-
gators: T.G. Kwiatkowski, S.H. Horowitz; Co-investigators: R.
Libman, R. Kanner, R. Silverman, J. LaMantia, C. Mealie, R.
Duarte; Study Coordinators: R. Donnarumma, M. Okola, V. Cullin,
E. Mitchell. Henry Ford Hospital (62 patients): Principal Investiga-
tor: S.R. Levine; Co-investigators: C.A. Lewandowski, G. Tokarski,
N.M. Ramadan, P. Mitsias, M. Gorman, B. Zarowitz, J. Kokkinos, J.
Dayno, P. Verro, C. Gymnopoulos, R. Dafer, L. D’Olhaberriague;
Study Coordinators: K. Sawaya, S. Daley, M. Mitchell. Emory
University School of Medicine (39 patients): Principal Investigators:
M. Frankel, B. Mackay; Co-investigators: J. Weissman, J. Washington,
B. Nguyen, A. Cook, H. Karp, M. Williams, T. Williamson; Study
Downloaded from http://ahajournals.org by on June 7, 2021
Coordinators: C. Barch, J. Braimah, B. Faherty, J. MacDonald, S.
Sailor; Affiliated Sites: Grady Memorial Hospital; Crawford Long
Hospital; Emory University Hospital; South Fulton Hospital, M.
Kozinn, L. Hellwick. University of Virginia Health System (37
patients): Principal Investigator: E.C. Haley, Jr; Co-investigators:
T.P. Bleck, W.S. Cail, G.H. Lindbeck, M.A. Granner, S.S. Wolf,
M.W. Gwynn, R.W. Mettetal, Jr, C.W.J. Chang, N.J. Solenski, D.G.
Brock, G.F.Ford; Study Coordinators: G.L. Kongable, K.N. Parks,
S.S. Wilkinson, M.K. Davis; Affiliated Sites: University of Virginia
Health System, E.C. Haley, Jr; Winchester Medical Center, G.L.
Sheppard, D.W. Zontine, K.H. Gustin, N.M. Crowe, S.L. Massey.
University of Tennessee (14 patients): Principal Investigator: M.
Meyer, K. Gaines; Study Coordinators: A. Payne, C. Bales, J.
Malcolm, R. Barlow, M.Wilson; Affiliated Sites: Baptist Memorial
Hospital, C. Cape; Methodist Hospital Central, T. Bertorini; Jackson
Madison County General Hospital, K. Misulis; University of Ten-
nessee Medical Center, W. Paulsen, D. Shepard.
Other participants are as follows: Coordinating Center: Henry
Ford Health Sciences Center: Principal Investigator: B.C. Tilley;
Co-investigators: K.M.A. Welch, S.C. Fagan, M. Lu, S. Patel, E.
Masha, J. Verter; Study Coordinators: J. Boura, J. Main, L. Gordon;
Programmers: N. Maddy, T. Chociemski; CT Reading Centers: Part
A—Henry Ford Health Sciences Center, J. Windham, H. Soltanian
Zadeh; Part B—University of Virginia Medical Center, W. Alves,
M.F. Keller, J.R. Wenzel; Central Laboratory: Henry Ford Hospital,
N. Raman, L. Cantwell; Drug Distribution Center: A. Warren, K.
Smith, E. Bailey. NINDS, Project Officer: J.R. Marler. Data and
Safety Monitoring Committee: J.D. Easton, J.F. Hallenbeck, G. Lan,
J.D. Marsh, M.D. Walker. Genentech, Inc, Participants: J. Froelich,
MD, J. Breed, F. Wang-Chow.
Acknowledgments
This study was supported by grants from the National Stroke
Association, the NINDS (N01-NS02382, N01-NS02374, N01-
Lyden et al NIHSS Factor Analysis 2353
NS02377, N01-NS02381, N01-NS02379, N01-NS02373, N01-
NS02376, N01-NS02378, N01-NS02380), and the Veterans Affairs
Research Service.
References
1. Boysen G. Stroke scores and scales. Cerebrovasc Dis. 1992;2:239 –247.
2. Edwards DF, Chen YW, Diringer MN. Unified neurological stroke scale
is valid in ischemic and hemorrhagic stroke. Stroke. 1995;26:1852–1858.
3. Fullerton KJ, Steiner TJ, Orgogozo JM, Adams RJ, Nichols FT,
Thompson WO, Yatsu FM, Yeakley JW, Fenstermacher MJ, Donner A,
Eliasziw M, Poungvarin N, Edwards JMR, Haley EC Jr, Kassell NF,
Torner JC, van Gijn J, Adams HP Jr, Wade DT, Kornhuber HH, Backhaus
B, Kornhuber AW, Kornhuber J, Warlow C, Koudstaal PJ, Smets P,
Candelise L, Harrison MJG. Clinical Trial Methodology in Stroke.
London, England: Bailliere Tindall; 1989:3–286.
4. Feinstein AR. Clinimetrics. New Haven, Conn: Yale University Press;
1987.
5. Lyden PD, Lau GT. A critical appraisal of stroke evaluation and rating
scales. Stroke. 1991;22:1345–1352.
6. Lyden PD, Hantson L. Assessment scales for the evaluation of stroke
patients. J Stroke Cerebrovasc Dis. 1998;7:113–127.
7. Hantson L, De Keyser J. Neurological scales in the assessment of cerebral
infarction. Cerebrovasc Dis. 1994;4(suppl 2):4 –14.
8. Cote R, Battista RN, Wolfson CM. Stroke assessment scales: guidelines
for development, validation, and reliability assessment. Can J Neurol Sci.
1988;15:261–265.
9. Asplund K. Clinimetrics in stroke research. Stroke. 1987;18:528 –530.
10. de Haan R, Horn J, Limburg M, Van Der Meulen J, Bossuyt P. A
comparison of five stroke scales with measures of disability, handicap,
and quality of life. Stroke. 1993;24:1178 –1181.
11. Muir KW, Weir CJ, Murray GD, Povey C, Lees KR. Comparison of
neurological scales and scoring systems for acute stroke prognosis.
Stroke. 1996;27:1817–1820.
12. Brott T, Adams HP, Olinger CP, Marler JR, Barsan WG, Biller J, Spilker
J, Holleran R, Eberle R, Hertzberg V, Rorick M, Moomaw CJ, Walker M.
Measurements of acute cerebral infarction: a clinical examination scale.
Stroke. 1989;20:864 – 870.
13. Lyden P, Brott T, Tilley B, Welch KMA, Mascha EJ, Levine S, Haley
EC, Grotta J, Marler J, NINDS TPA Stroke Study Group. Improved
reliability of the NIH Stroke Scale using video training. Stroke. 1994;25:
2220 –2226.
14. Albanese MA, Clarke WR, Adams HP Jr, Woolson RF. Ensuring reli-
ability of outcome measures on multicenter clinical trials of treatments for
acute ischemic stroke: the program developed for the trial of ORG 10172
in Acute Stroke Treatment (TOAST). Stroke. 1994;25:1746 –1751.
15. D’Olhaberriague L, Litvan I, Mitsias P, Mansbach H. A reappraisal of
reliability and validity studies in stroke. Stroke. 1996;27:2331–2336.
16. Goldstein L, Samsa G. Reliability of the National Institutes of Health
Stroke Scale. Stroke. 1997;28:307–310.
17. Goldstein LB, Bartels C, Davis JN. Interrater reliability of the NIH Stroke
Scale. Arch Neurol. 1989;46:660 – 662.
18. Schmulling S, Grond M, Rudolf J, Kiencke P. Training as a prerequisite
for reliable use of NIH Stroke Scale. Stroke. 1998;29:1258 –1259.
19. Brott T, Marler JR, Olinger CP, Adams HP, Tomsick T, Barsan WG,
Biller J, Eberle R, Hertzberg V, Walker M. Measurements of acute
cerebral infarction: lesion size by computerized tomography. Stroke.
1989;20:871– 875.
20. Tong DC, Yenari MA, Albers GW, O’Brien MD, Marks MP, Moseley
ME. Correlation of perfusion- and diffusion-weighted MRI with NIHSS
score in acute (,6.5 hour) ischemic stroke. Neurology. 1998;50:
864 – 870.
21. Duncan PW, Goldstein LB, Matchar D, Divine GW, Feussner J. Mea-
surement of motor recovery after stroke: outcome assessment and sample
size requirements. Stroke. 1992;23:1084 –1089.
22. Nunnally JC. Psychometric Theory. 2nd ed. New York, NY:
McGraw-Hill; 1978.
23. Kaplan RM. Basic Statistics for the Behavioral Sciences. Boston, Mass:
Allyn and Bacon; 1987.
24. Child D. The Essentials of Factor Analysis. 5th ed. London, England:
Holt, Rinehart & Winston, Inc; 1978:1–107.
25. LaRocca NC. Statistical and methodologic considerations in scale con-
struction. In: Munsat TL, ed. Quantification of Neurologic Deficit.
Boston, Mass: Butterworth; 1989:49 – 61.
 2354 Stroke November 1999
26. Futty D, Conbneally M, Dyken M, Price T, Haerer A, Poskanzer D,
Swanson P, Calanchini P, Gotshall R. Cooperative study of hospital
frequency and character of transient ischemic attacks, V: symptom anal-
ysis. JAMA. 1977;238:2386 –2390.
27. Holbrook M, Skilbeck CE. An activities index for use with stroke
patients. Age Ageing. 1983;12:166 –170.
28. Elmstahl S, Malmberg B, Annerstedt L. Caregiver’s burden of patients 3
years after stroke assessed by a novel caregiver burden scale. Arch Phys
Med Rehabil. 1996;77:177–182.
29. Hajek VE, Gagnon S, Ruderman J. Cognitive and functional assessments
of stroke patients: an analysis of their relation. Arch Phys Med Rehabil.
1997;78:1331–1337.
30. Essink-Bot M, Krabbe P, Bonsel B, Aaronson N. An empirical com-
parison of four generic health status measures: the Nottingham Health
Profile, the Medical Outcomes Study 36-Item Short-Form Health Survey,
the COOP/WONCA charts, and the EuroQol instrument. Med Care.
1997;35:522–537.
Downloaded from http://ahajournals.org by on June 7, 2021
31. The National Institute of Neurological Disorders and Stroke rt-PA Stroke
Study Group. Tissue plasminogen activator for acute ischemic stroke.
N Engl J Med. 1995;333:1581–1587.
32. Anderson J, Gerbing D. Structural equation modeling in practice: a
review and recommended two-step approach. Psychol Bull. 1988;103:
411– 423.
33. Bentler P, Bonett D. Significance tests and goodness-of-fit in the analysis
of covariance structures. Psychol Bull. 1980;88:588 – 606.
34. Heinemann AW, Harvey RL, McGuire JR, Ingberman D, Lovell L, Semik
P, Roth EJ. Measurement properties of the NIH Stroke Scale during acute
rehabilitation. Stroke. 1997;28:1174 –1180.
35. Wade DT, Hewer RL. Functional abilities after stroke: measurement,
natural history and prognosis. J Neurol Neurosurg Psychiatry. 1987;50:
177–182.
36. Lindmark B, Hamrin E. Evaluation of functional capacity after stroke as
a basis for active intervention: validation of a modified chart for motor
capacity assessment. Scand J Rehabil Med. 1988;20:111–115.