Content uploaded by Michel Tousignant
Author content
All content in this area was uploaded by Michel Tousignant on Apr 16, 2015
Content may be subject to copyright.
Clinical evaluation of patient following stroke:
Proposed stroke patient taxonomy based on cluster
analysis method
Michel Tousignant, A. B. Arsenault, H. Corriveau, and P. Philippe
INTRODUCTION
Classification refers to the grouping of objects into
homogeneous groups or sets on the basis of their
similarities and differentiating between the sets on
the basis of their differences (Norman and
Streiner, 1997). In the medical sciences, one of the
main purposes of classification is to develop taxon-
omies which enable the health professional to
build categories of diagnoses, and to direct treat-
ment. This concept of classification has recently
gained interest among researchers and clinicians
in the rehabilitation field (Riddle, 1998).
The act of naming particular groups of
homogeneous patients by classifying them into
Michel Tousignant, Physiotherapy Programme,
School of Rehabilitation Science, Faculty of Health
Sciences, University of Ottawa, 451 Smyth Road,
Ottawa, Ontario, Canada K1H 8M5
A.B. Arsenault, Physiotherapy Program, School of
Rehabilitation, Faculty of Medicine, University of
Montreal and Research Center, Montreal
Rehabilitation Institute, 2375 Côte-Ste-Catherine,
Montréal, Québec, Canada H3T 1A8
H. Corriveau, University of Sherbrooke and
Physiotherapy Program, School of Rehabilitation
Science, Faculty of Health Sciences, University of
Ottawa, 451 Smyth, Ottawa, Ontario, Canada
K1H 8M5
P. Philippe, Department of Social and Preventive
Medicine, Faculty of Medicine, University of
Montreal, 2375 Côte-Ste-Catherine, Montréal,
Québec, Canada H3T 1A8
Accepted for publication October 1999
Physiotherapy Theory and Prac tice (2000) 16, 81–93
© 2000 Taylor & Francis
This study describes the use of a cluster analysis method to develop a stroke
patient taxonomy. Fifty-five volunteer patients, average age 50.4 ± 14.7 years, who
had suffered a cerebrovascular accident (CVA) were included in the study. The
average time from CVA to beginning active rehabilitation was 3.0 ± 1.7 months.
On arrival at the rehabilitation centre, the patients were evaluated by two
physiotherapists with a method based on the Bobath approach. The method uses
four variables: tonus, postural reactions, reflex activity, and active movement.
Agglomerative hierarchical cluster analysis was the statistical method used.
Analysis of the dendrogram led to the identification of five taxonomy categories.
There are two main categories: one that includes a set of items that describe a
stroke patient whose motor performance is relatively good, and another category
that includes a set of items which describe patients presenting weaker motor
performance.
It appears that the proposed taxonomy could guide the development of
treatment plans for stroke patients. A multicentre study needs to be conducted in
order to replicate these results using larger numbers of subjects.
Physiother Theory Pract Downloaded from informahealthcare.com by Universite De Sherbrooke on 04/16/15
For personal use only.
subgroups based on signs and symptoms is the out-
come of the clinician’s clinical reasoning process.
The name or label given to a subgroup of patients
must guide the development of a treatment plan in
order to be useful to the clinician (Sahrmann,
1988). For example, assigning the label of
“cerebrovascular accident” to a patient is not
useful in the physical therapy management of this
patient. The additional label of “flaccid” or “spas-
tic” is relatively meaningless when considering
treatment or prognosis, but a label such as “non-
fragmented volitional movement with severe
muscle tone dysfunction” could be more useful to
the physiotherapist developing the treatment plan
(Sahrmann, 1988).
In this context, there is a need in physical
therapy for patient classification (Rothstein,
1993). In the course of its evolution, rehabilitation
science has had to develop taxonomies. Spinal
disorder taxonomies (Bernard and Kirkaldy-
Willis, 1987; Diletto, Erhard, and Bowling, 1995;
McKenzie, 1981; Spitzer, 1987; Spitzer et al, 1995)
and stroke patient taxonomies (Bobath, 1990;
Brunstrom, 1970; Guarna et al., 1988) are exam-
ples of such efforts.
Most of the taxonomies used in rehabilitation
arose out of the clinical experience and the
advancement of theoretical knowledge. In this
respect, they are built in a subjective way and repre-
sent only one point of view with regard to the
similarities and disparities between groups of
patients. Even when efforts are made to improve
these taxonomies, they still remain an improve-
ment of a subjective classification.
Taxonomies in the stroke
rehabilitation field
The field of stroke rehabilitation is a good
example of the development and application of
taxonomies. Following Twitchell (1951), the most
common approaches to the evaluation and treat-
ment of stroke patients have led to the creation of
taxonomies that are useful and of great interest in
categorising this type of patient (Bobath, 1990;
Brunstrom, 1970). Table 1 summarises these
taxonomies.
Brunstrom’s theory (1970) is based on the
capacity of a stroke patient to perform muscular
activities that require a progressive refinement of
neuromuscular control. To some extent, such a
refinement is an evaluation of the degree of
recovery of the nervous system. It divides the
process of motor recovery into six stages which
the patient has to go through to attain a complete
functional recovery. It comprises pre-synergic
(stage 1), synergic (stages 2 and 3), and post-
synergic (stages 4 to 6) elements, which range
from deviations in a pattern of basic synergies to
normal co-ordination in the affected stroke
extremity. Brunstrom’s assessment measures the
82 M. TOUSIGNANT ET AL.
Table 1
Summary of Bobath’s and Brunstrom’s taxonomies used with stroke patients
Bobath (1990) Brunstrom (1970)
Stage 1:
Flaccidity
Stage 1:
Flaccidity; no active movement
Stage 2:
Spasticity
Stage 2:
Beginning of spasticity; first appearance of synergies or associated reactions
Stage 3:
Relative recovery
Stage 3:
Increasing spasticity; basic synergies are present and can be voluntarily
executed
Stage 4:
Decreasing spasticity; movements beginning to go off course from synergies
Stage 5:
Stronger spasticity decreasing; relative independence of basic synergies
Stage 6:
No spasticity; movements are isolated in each joint
Physiother Theory Pract Downloaded from informahealthcare.com by Universite De Sherbrooke on 04/16/15
For personal use only.
quality of the voluntary movement and takes into
account the influence of posture, primal reflexes,
and associated reactions.
Bobath used three stages to describe patient
recovery: (1) flaccidity; (2) spasticity; and (3) rela-
tive recovery (Bobath, 1990). She suggested that
the main difficulty for a stroke patient is abnormal
co-ordination. Following stroke patients have diffi-
culty executing active movements, or execute
them in an abnormal way, due to a lack of normal
postural reactions which leads to a co-ordination
problem. Unlike Brunstrom’s approach, synergies
are not considered in Bobath’s description of post-
cerebrovascular accident (CVA) motor recovery.
In this approach, the assessment of the patient’s
motor patterns is qualitative rather than quantita-
tive. It is based on observation of the patient’s
motor functions on the affected side.
One of the drawbacks encountered with this
approach is the existence of grey areas between
the recovery stages, which makes it difficult to
classify certain stroke patients in one of the pro-
posed stages (Bobath, 1990). In such a context,
Guarna et al. (1988) introduced a theoretical
model to improve discrimination of the stroke
patient’s recovery state, according to the stages
proposed by Bobath. They suggested adding
another three stages, thus increasing the number
of recovery stages to six: (1) initial recovery; (2)
hypotonicity; (3) imbalance tonicity; (4) hyper-
tonicity; (5) relative recovery; and (6) return to
normal. Initial recovery and hypotonicity
correspond to stage 1 of the Bobath approach;
imbalance tonicity and hypertonicity to stage 2,
and relative recovery and return to normal corre-
spond to stage 3.
In the process of testing this model, Corriveau
et al. (1988) proposed an evaluation tool for
muscle function based on the Bobath approach to
treatment. The purpose of this tool was to classify
patients among the six stages proposed by Guarna
et al.’s theory (1988). The tool contains the follow-
ing parameters: sensorium, muscle tone, reflex
activity, active movement, postural reactions, and
pain. Despite the development and validation of
this evaluation tool (Arsenault et al., 1988;
Corriveau et al., 1988), the classification model of
Guarna et al. (1988) has never been verified with
respect to its validity in classifying stroke patients.
However, one point links all three models: the
evaluation of stroke patients is based on the
mutual influence of muscle tone, reflex activity,
and postural reactions on active movement recov-
ery. According to Bobath and Brunstrom, these
parameters also form the basis of stroke patient
treatment. Furthermore, each taxonomy conceals
pieces of information that should help the clini-
cian to develop the treatment procedure but this
information particularly concerns muscle tone
quality.
In such a context, classical taxonomies do not
seem to achieve their goal, which is to provide
information that will be useful to the clinician
developing a treatment plan. The present study
attempts to improve on the existing stroke patient
taxonomies. The new taxonomy should provide
more information that will be useful in developing
treatment plans for stroke patients.
The purpose of this study is methodological. It
describes the use of a cluster analysis method in an
attempt to develop a stroke patient taxonomy.
Since most of the taxonomies used in rehabilita-
tion are based on theoretical knowledge and/or
clinical experience, the original contribution of
this study is that it is the first study whose purpose
was to develop a taxonomy based on an objective
methodology for this group of patients.
After describing the development process, this
new taxonomy is discussed with regard to its useful-
ness in planning specific treatment for stroke
patients. It is compared with the taxonomies pro-
posed by Bobath (1990), Brunstrom (1970), and
Guarna et al. (1988).
METHODS
Subjects
The sample comprised 55 stroke patients (30 men
and 25 women) who were hospitalised at the
Montreal Rehabilitation Institute. Their average
age was 50.4 (± 14.7) years. They had all suffered
from a cerebrovascular accident (CVA) and the
average time from CVA to beginning active
rehabilitation was 3.0 ± (1.7) months. Thirty-one
of the subjects presented with left hemiplegia and
EVALUATION OF PATIENT FOLLOWING STROKE 83
Physiother Theory Pract Downloaded from informahealthcare.com by Universite De Sherbrooke on 04/16/15
For personal use only.
24 with right hemiplegia. All subjects attended
intensive therapy based on the Bobath approach
for a period of at least 2 months. Based on these
characteristics, the sample was judged to be
relatively homogeneous.
All subjects were evaluated by two physio-
therapists who had received training on the use of
the evaluation protocol developed by Corriveau et
al. (1988) based on the Bobath approach. They
evaluated the six variables included in this proto-
col: muscle tone, active movement, postural reac-
tions, reflex activity, sensorium, and pain. A full
description of the parameters has been published
previously (Corriveau et al., 1988) and is summa-
rised below. Sensorium tests the ability of the
patient to understand and co-operate in the test-
ing procedures. His or her responses to certain
stimuli such as verbal, nociceptive, olfactory, and
tactile stimuli are evaluated. Muscle tone includes
passive movement of the neck, trunk, and extremi-
ties along with the elicitation of the biceps reflex.
Reflex activity includes the disclosure of any associ-
ated reactions present in the upper extremity and
lower extremity, Babinski’s sign and the eliciting
of the positive supporting reaction. Voluntary
movement requires the patient to execute the
same movements done passively by the therapist in
the evaluation of the muscle tone. Automatic reac-
tions include the righting reactions, the equilib-
rium and protective reactions. Pain consists of
questions on the localisation, type, and the
frequency of pain in the upper extremity.
Cluster analysis method
Cluster analysis is a generic term which covers a
large variety of methods used to create classi-
fications. These methods are based on multi-
variate statistical procedures, starting with a
sample of entities. The ultimate purpose is to
organise these entities into homogeneous sub-
groups (Aldenderfer and Blashfield, 1984).
Cluster analysis attempts to classify subjects into
subgroups and to distinguish them from other sub-
groups of subjects. Applications of this method are
increasingly common in the literature as well as in
other fields; for example, psychiatry (Everitt,
Gourlay, and Kendell, 1971; Matussek, Soldner,
and Nagel, 1981; Playkel, 1978), dentistry (Wastell
and Gray, 1987), and low back pain (Coste, 1991).
The cluster analysis method used in
this research is the hierarchical approach
(Aldenderfer and Blashfield, 1984). The under-
lying assumption of this approach is that each sub-
group is unique, but all the subgroups within a
given group have something in common.
Although the groups are different from each
other, one can move up in the hierarchy to a
higher level of organisation in which all of the
groups are related somehow. This reflects the
belief that there is a hierarchy among the groups.
With this approach, there are numerous criteria
for forming the groups. One of these methods
is the Ward technique which seems to be the
most and most frequently used method
(Aldenderfer and Blashfield, 1984). The first step
in the process is to establish an index of distance
between the subjects. In the Ward technique, the
index used is the Euclidean distance between
the variables of two subjects:
( )
d X X
ij ik jk
k
p
= -
=
å
2
1
where
d
ij
is the distance between cases
i
and
j
and
X
ik
is the value of variable
k
for case
i
With this approach, a subject is joined to the
group to which it adds the least within-group vari-
ability. That is, it tries to outreach each group in
turn, and sticks with the one where its membership
increases the variability by the smallest amount. At
first, there are as many groups as there are individ-
uals; as the classification progresses, each subject
ends up in one of the groups. The final result can
be illustrated by a tree diagram called a dendro-
gram (Figure 1). The X axis represents the first
level in the classification of entities and represents
each subject. The Y axis represents the classifica-
tion criterion value. Reading the dendrogram
starting at the bottom line of the Y axis and moving
upwards, shows an increase in similarity across the
subjects. This increasing similarity leads to group-
ing some subjects together.
Variable selection
Selecting the variables used in an automated classi-
fication is the most important step (Aldenderfer
84 M. TOUSIGNANT ET AL.
Physiother Theory Pract Downloaded from informahealthcare.com by Universite De Sherbrooke on 04/16/15
For personal use only.
and Blashfield, 1984). It is preferable to choose a
set of variables which are not correlated to each
other and where the informational content is not
redundant. The variables must provide a maxi-
mum of different information describing the stud-
ied entity. They must also be discriminating, i.e.,
present a potential similarity or disparity between
each of them. A resemblance or difference
between the variables affects the variance and will
permit the association of similar subjects, leaving
aside the aliens. The formation of homogeneous
categories then becomes possible.
In this study, the protocol introduced by
Corriveau et al. (1988) was used to describe the
stroke patient. This protocol proposes six vari-
ables: sensorium, postural reactions, active move-
ment, reflex activity, muscle tone, and pain. Each
of these variables was rated using an operational
definition. Table 2 gives an example of the opera-
tional definition for the rating of the active move-
ment of the upper extremity. All the other
operational definitions are given in Corriveau et
al. (1988). Its validity (Arsenault et al., 1988) and
reliability (Corriveau et al., 1992) have been
analysed through empirical studies. Intrarater
reliability was very high, with intraclass correlation
coefficients (ICC) of .95 and .97 for the upper and
lower limbs, respectively. For the interrater
reliability study, the ICCs were .79 and .77 for the
upper and lower limbs, respectively (Corriveau et
al., 1992). Furthermore, this protocol is sensitive
in depicting progress in motor recovery through
three measurements taken over 2 months of treat-
ment and was significantly correlated with the
results of other testing procedures evaluating
motor performance, such as Brunstrom’s (1970)
(Spearman’s
r
= .79;
p
< .001), and Fugl-Meyer’s
(1979) evaluation (Spearman’s
r
= .85;
p
< .001)
(Arsenault et al., 1988).
To examine the importance and redundancy of
the variables included in this protocol, Corriveau
et al. (1992) did a principal components analysis
which generated three main components explain-
ing 69% of the total variance of the data. These
components were: (1) postural reactions and
active movement (32.3%); (2) muscle tone
(22.2%); and (3) sensorium (14.4%).
Based on these results, the present study on
stroke patient taxonomy used the elements of two
first main components as the most informative and
non-redundant variables: postural reactions,
active movement, and muscle tone. Based on
EVALUATION OF PATIENT FOLLOWING STROKE 85
Fig. 1 Dendrogram or tree diagram
Physiother Theory Pract Downloaded from informahealthcare.com by Universite De Sherbrooke on 04/16/15
For personal use only.
clinical experience, another variable which
seemed to be important, i.e., reflex activity, was
added since it was felt that it could be related to the
motor recovery profile of the stroke patient. Fur-
thermore, in the protocol used for this study,
reflex activity is composed of three important ele-
ments related to the patient following stroke: asso-
ciated reactions, Babinski’s sign, and positive
supporting reaction (Bobath, 1990; Twitchell,
1951).
Statistical analysis
Statistical analysis used the hierarchical clustering
method based on Ward’s technique. The outcome
of the clustering method was presented as a
dendrogram and visual observation was used to
select the number of clusters. Descriptive statistics
of the clusters completed the data description. An
analysis was performed on Systat Software version
5.1 for Macintosh.
RESULTS
Ward’s hierarchical method was used, taking into
account four variables: postural reactions, active
movement, muscle tone, and reflex activity. The
results are presented in a dendrogram (Figure 2).
The X axis represents the first level in the individ-
ual classification of stroke patients. The Y axis
represents the classification criterion value. Read-
ing the dendrogram starting at the bottom line of
the Y axis and moving upward, shows a progressive
grouping of some of the stroke patients, thus
reducing the number of clusters.
On the basis of the best separation of the clus-
ters by visual inspection, analysis of the dendro-
gram led to the identification of five clusters of
patients: A, B, C, D, and E (Figure 2). Once these
clusters were identified, a description of the
patients belonging to each of these clusters was
developed. Table 3 gives descriptive statistics for
each cluster of patients. Considering that the main
characteristic of patients following stroke is diffi-
culty executing an active movement (Bobath,
1990), the “active movement” variable was chosen
as the reference variable. This allowed us to study
the other three variables in comparison with this
reference variable in order to identify similarities
or disparities across the five clusters of patients. In
Table 4, two main clusters are highlighted based
on the similarity of the descriptive statistics of the
“active movement” variable across the five clusters
of patients: Two clusters of patients present a
similar magnitude of their means (clusters A and
86 M. TOUSIGNANT ET AL.
Table 2
Bobath evaluation: Operational definition of muscle tone of the upper extremity
Procedures
Support is given at the elbow and wrist by examiner and the following passive movements are performed:
1. With the shoulder in external rotation and the elbow in extension, flex shoulder to a range from zero to full
flexion.
2. With shoulder flexed at 90°, flex and extend elbow bringing palm of hand to opposite shoulder and to the top
of the head.
3. With elbow flexed and palm of hand on opposite shoulder, flex and extend the shoulder.
4. With elbow extended, and shoulder flexed at 90°, protract the shoulder.
5. With arm next to trunk, forearm and hand supported and with the elbow at 90°, pronate and supinate forearm.
6. With elbow extended and shoulder at 90°, (with forearm in neutral position and arm supported at elbow and
wrist) flex and extend the wrist.
7. With elbow extended, shoulder flexed at 90°, with wrist and forearm in neutral position (arm supported at
elbow and wrist), flex and extend fingers.
Ratings
0 = Flaccidity: flaccidity noted in all the muscle groups including distal.
1 = Mainly flaccid with some spasticity: This category includes patients who have one of the following—flaccidity
in shoulder girdle and shoulder with spasticity in fingers, wrist, and pronators; flaccidity distally and spasticity
proximally; weakness and weak tendon reflexes.
2 = Mainly spastic with some flaccidity: Spasticity dominates in upper extremity with possibility of some specific
muscle flaccidity.
3 = Normal: Normal tone with possible residual weakness. The tendon reflexes are normal.
Physiother Theory Pract Downloaded from informahealthcare.com by Universite De Sherbrooke on 04/16/15
For personal use only.
87
Fig. 2 Dendogram for the cluster analysis of the 55 stroke subjects using Ward Method
Physiother Theory Pract Downloaded from informahealthcare.com by Universite De Sherbrooke on 04/16/15
For personal use only.
B) compared with the other three (clusters C, D,
and E). From these groupings, we can identify two
main categories of patients: one includes clusters
A and B which describe stroke patients presenting
weaker motor performance, and the second
category includes clusters C, D, and E which
describe stroke patients whose motor perfor-
mance is relatively good.
The next step consisted of identifying the
differences between the three variables (reflex
activity, postural reactions, and muscle tone) for
each cluster now featured in two main categories:
weak motor performance and good motor perfor-
mance. As in the previous step, we examined the
similarities or discrepancies of the means of the
three variables. Within clusters A and B of the weak
motor performance main category, a variable can
be distinguished from the others that allows
separation of the two clusters based on the descrip-
tive statistics as shown in Table 5: (1) Patients in
cluster A present a weak motor performance and a
stronger tonus; (2) patients in cluster B present
88 M. TOUSIGNANT ET AL.
Table 3
Descriptive statistics of the variables for stroke subjects in each category of the cluster analysis
Category Tonus
Mean
(standard deviation)
Reflex
Mean
(standard deviation)
Postural reaction
Mean
(standard deviation)
Active movement
Mean
(standard deviation)
Cluster A
n = 15
1.63 (0.44) 1.0 (0.38) 0.87 (0.35) 0.4 (0.34)
Cluster B
n = 14
0.93 (0.47) 2.0 (0) 0.93 (0.27) 0.75 (0.58)
Cluster C
n = 8
1.56 (0.32) 0.88 (0.36) 1.25 (0.46) 1.88 (0.35)
Cluster D
n = 10
1.34 (0.67) 2.2 (0.42) 2.5 (0.53) 2.0 (0.53)
Cluster E
n = 8
2.06 (0.32) 2.25 (0.46) 1.25 (0.71) 2.0 (0.39)
Table 4
Category groupings: Patient following stroke presenting good motor performance (subcategories 3, 4, and 5)
and weak motor performance (subcategories 1 and 2)
Category Tonus
Mean
(standard deviation)
Reflex
Mean
(standard deviation)
Postural reaction
Mean
(standard deviation)
Active movement
1
Mean
(standard deviation)
Weak motor
performance
Cluster A
n = 15
1.63 (0.44) 1.0 (0.38) 0.87 (0.35) 0.4 (0.34)
Cluster B
n = 14 0.93 (0.47) 2.0 (0) 0.93 (0.27) 0.75 (0.58)
Good motor
performance
Cluster C
n = 8
1.56 (0.32) 0.88 (0.36) 1.25 (0.46) 1.88 (0.35)
Cluster D
n = 10
1.34 (0.67) 2.2 (0.42) 2.5 (0.53) 2.0 (0.53)
Cluster E
n = 8
2.06 (0.32) 2.25 (0.46) 1.25 (0.71) 2.0 (0.39)
1
reference variable.
Physiother Theory Pract Downloaded from informahealthcare.com by Universite De Sherbrooke on 04/16/15
For personal use only.
a weak motor performance and more accurate
reflexes.
The same process can be followed to distin-
guish between the three clusters in the second
main category, good motor performance: (1)
Patients in cluster C present a good motor perfor-
mance and a more serious reflex disorder; (2)
patients in cluster D present a good motor perfor-
mance and a more serious muscle tonus disorder;
and (3) patients in cluster E present a good motor
performance and a more serious postural reaction
disorder.
The final step in the procedure consisted of
developing a classification giving the best possible
description of a stroke patient belonging to any of
the established subcategories. Table 6 describes
this taxonomy.
EVALUATION OF PATIENT FOLLOWING STROKE 89
Table 5
Identification of similarities and disparities for tonus, reflex, and postural reaction variables for each
sub-category
Category Tonus
Mean
(standard deviation)
Reflex
Mean
(standard deviation)
Postural reaction
Mean
(standard deviation)
Active movement
1
Mean
(standard deviation)
Weak motor
performance
Cluster A
n = 15
1.63 (0.44) 1.0 (0.38) 0.87 (0.35) 0.4 (0.34)
Cluster B
n = 14 0.93 (0.47) 2.0 (0) 0.93 (0.27) 0.75 (0.58)
Good motor
performance
Cluster C
n = 8
1.56 (0.32) 0.88 (0.36) 1.25 (0.46) 1.88 (0.35)
Cluster D
n = 10
1.34 (0.67) 2.2 (0.42) 2.5 (0.53) 2.0 (0.53)
Cluster E
n = 8
2.06 (0.32) 2.25 (0.46) 1.25 (0.71) 2.0 (0.39)
1
reference variable.
Table 6
Taxonomy of stroke patients after cluster analysis
Main Clusters Characteristics of
the stroke patients
Suggested taxonomy
Main category 1:
Weak motor performance
Stroke patients presenting reflex activity
and postural reaction problems
Disturbed motor performance—group of
stroke patients with reflex activity and
postural reactions problems
Stroke patients presenting reflex activity
and muscle tone problems
Disturbed motor performance—group of
stroke patients with reflex activity and
muscle tone problems
Main category 2:
Good motor performance
Stroke patients presenting reflex activity
problems
Relative recovery—group of stroke
patients with reflex activity problems
Stroke patients presenting muscle tone
problems
Relative recovery—group of stroke
patients with muscle tone problems
Stroke patients presenting postural
reaction problems
Relative recovery—group of stroke
patients with postural reaction problems
Physiother Theory Pract Downloaded from informahealthcare.com by Universite De Sherbrooke on 04/16/15
For personal use only.
DISCUSSION
The main objective of this study was to present a
classification of clinical entities that could be
useful in rehabilitation. The automated classifi-
cation introduced by Ward (Aldenderfer and
Blashfield, 1984) was used in an effort to produce
a taxonomy of stroke patients going through
recovery stages. More specifically, this study
sought to answer the following question: Can
applying an automated classification to stroke
subject data be helpful in describing the character-
istics of these subjects in order to identify clusters
that could be clinically useful in planning
treatment?
The taxonomy generated by Ward’s automated
classification appears to be more specific than
existing taxonomies. In an effort to identify the
clinical usefulness of the suggested taxonomy, it
was compared with Bobath’s (1990), Brunstrom’s
(1970) and Guarna et al.’s (1988) taxonomies. We
hypothesised that the proposed taxonomy would
provide information that is not provided by classi-
cal taxonomies and that this additional informa-
tion would be useful to clinicians. Table 7
compares both classical taxonomies with the new
empirical taxonomy with respect to the informa-
tion each provides. Regardless of the taxonomy
used, clinicians base their evaluation of the
patient’s state of recovery upon the interaction
between the four parameters evaluated (muscle
tone, reflex activity, postural reaction, and active
movement) in an attempt to plan the most person-
alised treatment. Clinicians have to consider how
reflexes and postural reactions interfere with
tonus, and how tonus interferes with each active
movement. The treatment plan must consider the
specific characteristics of each stroke patient.
Theoretically, the best taxonomy should guide
clinicians in their therapeutic decisions. However,
this is not the case if the clinician applies Bobath’s
or Brunstrom’s taxonomies. In fact, the only
indicator emerging from these two taxonomies is
the influence of muscle tone on an active move-
ment. There is no indication regarding reflex
influence or postural reaction influence in these
taxonomies. However, clinicians should take these
parameters into account when developing their
treatment plans. For example, identifying a stroke
patient as a spastic subject does not define the
90 M. TOUSIGNANT ET AL.
Table 7
Comparison of the information concealed in the traditional and empirical stroke patient taxonomies
Bobath (1990) Brunstrom (1978) Guarna et al. (1988) Empirical
Stage 1:
Flaccidity
Stage 2:
Spasticity
Stage 3:
Relative recovery
Stage 1:
Flaccidity, no active
movement
Stage 2:
Beginning of spasticity;
appearance of synergies
or associated reactions
Stage 3:
Increasing spasticity; basic
synergies present and can
be executed voluntarily
Stage 4:
Decreasing spasticity;
movements begin to go
off course from synergies
Stage 5:
Spasticity decreases again;
relative independence of
basic synergies
Stage 6:
No spasticity; movements
isolated at each joint
Stage 1:
Initial recovery; flaccidity
Stage 2:
Hypotonicity
Stage 3:
Imbalanced tonicity
Stage 4:
Hypertonicity
Stage 5:
Relative recovery
Stage 6:
Return to normal
Stage 1:
Disturbed motor
performance: group of stroke
patients with reflex and
postural reactions problems;
group of stroke patients with
muscle tone and reflex
problems
Stage 2:
Relative recovery: group of
stroke patients with reflex
problems; group of stroke
patients with muscle tone
problems; group of stroke
patients with postural
reactions problems
Physiother Theory Pract Downloaded from informahealthcare.com by Universite De Sherbrooke on 04/16/15
For personal use only.
interaction that exists between muscle tone,
postural reactions, and reflexes on the active
movement model, although these parameters are
considered by clinicians when developing treat-
ment plans.
The proposed taxonomy is unique because of
the information it provides on reflexes, tonus and
postural reactions. For example, Figure 3 shows
the results of an assessment of a CVA patient in
which we can identify a major reflex problem but
with a good active movement. This patient should
be identified as a member of the “Relative recov-
ery—group reflex” category. Thus, a treatment
plan for this patient should specifically take into
EVALUATION OF PATIENT FOLLOWING STROKE 91
3
l imb
333
3
3
3
3
3 3
3
3
3
3
3
3
3
Fig. 3 Evaluation of a patient following stroke using the protocol
Physiother Theory Pract Downloaded from informahealthcare.com by Universite De Sherbrooke on 04/16/15
For personal use only.
account the influence of reflexes on the active
movements of that particular patient. Therefore, it
appears that the proposed taxonomy could pro-
vide more specific guidance in the development of
treatment plans for stroke patients. This taxonomy
only systematises what clinicians already know
about their patients: It could be viewed as an
improvement over the classical taxonomies.
Limitations of the study
This analysis must be interpreted with care. The
small number of subjects involved (
N
= 55) makes
the results only interpretable as a first indication of
disparities or similarities between stroke patients
on the basis of the four variables selected for the
study. The small number of subjects in each cate-
gory makes it impossible to determine statistically
if there is a real difference between the categories
as described in this analysis. To test the validity of
the taxonomy proposed in our results, cross-valida-
tion analysis should be conducted. Such a cross-
validation analysis implies randomly splitting the
sample into two subsamples, and using one
subsample to produce a taxonomy, and then
reproducing this taxonomy on the second
subsample. However, our sample size was too small
to split into two subsamples large enough for an
adequate cross-validation.
However, from a methodological viewpoint,
these results are of great interest. From our
sample, it appears that the use of a cluster analysis
method produces a taxonomy which provides a
better description of the parameters used for the
treatment of stroke patients. On the other hand,
interpreting these results while developing a
theoretical model would be hazardous and
inappropriate. This study merely suggests a
research hypothesis which warrants validation
through further studies.
In addition, having a larger number of subjects
involved in a similar study would make it possible
to include other variables in an effort to improve
and to permit statistical validation of the empirical
taxonomy. Among other parameters, differentia-
tion between the degrees of impairment of the
upper and lower extremities would be of interest.
It would also be interesting to do the same analysis
at different stages in the patient’s recovery, for
example, in the acute stage and one year post-
CVA, and to examine the distribution of patients
with the proposed classification.
CONCLUSION
In our sample, the stroke patient taxonomy based
on Ward’s approach appears to provide the clini-
cian planning treatment with more specific infor-
mation than the classical taxonomies. In fact, the
clinician has more information about reflexes,
tonus, and postural reactions and could develop a
more specific treatment plan for each stroke
patient. A multicentre study should be planned in
order to reproduce these results, using a sufficient
number of subjects and stroke patients in different
recovery stages.
References
Aldenderfer MS, Blashfield RK 1984 Cluster Analysis, 3rd edn.
Sage Publications, Beverly Hills
Arsenault AB, Dutil E, Lambert J, et al 1988 An evaluation of
the hemiplegic patient based on the Bobath approach. Part
III. A validation study. Scandinavian Journal of
Rehabilitation Medicine 20: 13–16
Bernard TN, Kirkaldy-Willis WH 1987 Recognising specific
characteristics of non-specific low-back pain. Clinical
Orthopaedics 217: 266–280
Bobath B 1990 Adult hemiplegia: Evaluation and treatment,
3rd edn. Heinemann Medical Books, London
Brunstrom S 1970 Movement therapy in hemiplegia: A
neurophysiological approach, 1st edn. Harper and Row,
Philadelphia
Corriveau H, Guarna F, Dutil E et al 1988 An evaluation of the
hemiplegic patient based on the Bobath approach. Part II.
The evaluation protocol. Scandinavian Journal of
Rehabilitation Medicine 20: 5–11
Corriveau H, Arsenault AB, Dutil E et al 1992a An evaluation
of the hemiplegic patient based on the Bobath approach: A
reliability study. Disability and Rehabilitation 14: 81–84
Corriveau H, Arsenault AB, Dutil E et al 1992b Principal
components analysis of an evaluation of the hemiplegic
subject based on the Bobath approach. Disability and
Rehabilitation 14: 85–88
Coste J 1991 Clinical taxonomy of psycho-social diversity of
non-specific low back pain. Journal of Clinical
Epidemiology 44: 1233–1245
Diletto A, Erhard RE, Bowling RW 1995 A treatment-based
classification approach of low-back pain syndrome:
Identifying and staging patients for conservative treatment.
Physical Therapy 75: 470–489
Everitt BS, Gourlay AJ, Kendell RE 1971 An attempt at
validation of traditional psychiatric syndromes by cluster
analysis. British Journal of Psychiatry 119: 399–412
92 M. TOUSIGNANT ET AL.
Physiother Theory Pract Downloaded from informahealthcare.com by Universite De Sherbrooke on 04/16/15
For personal use only.
Guarna F, Corriveau H, Chamberland J et al 1988 An
evaluation of the stroke patient subject based on the
Bobath approach. Part I. The model. Scandinavian Journal
of Rehabilitation Medicine 20: 1–4
Matussek P, Soldner M, Nagel D 1981 Identification of the
endogenous depressive syndrome based on the symptoms
and the characteristics of the course. British Journal of
Psychiatry 138: 361–372
McKenzie RA 1981 The lumbar spine: Mechanical diagnosis
and therapy, Spinal Publications, Waikanae, New Zealand
Norman GR, Streiner DL 1997 PDQ Statistics, 2nd edn.
Mosby, St. Louis
Paykel ES 1978 Classification of suicide attempters by cluster
analysis. British Journal of Psychiatry 133: 42–52
Riddle DL 1998 Classification and low-back pain: A review of
the literature and critical analysis of selected systems.
Physical Therapy 78: 708–737
Rothstein JM 1993 Patient classification. Physical Therapy 73:
214–215
Sahrmann S 1988 Diagnosis by the physical therapist: A
prerequisite for treatment. Physical Therapy 11: 1703–1706
Spitzer WO 1987 Scientific approach to the assessment and
management of activity-related spinal disorders. A
monograph for clinicians. Spine (supplement) S4–S59
Spitzer WO, Salmi R, Skovron ML, Duranceau J, Suissa S, Zeiss
E 1995 Redéfinir le “whiplash” et sa prise en charge. Le
médecin du Québec Mai(S) 1–76
Twitchell TE 1951 The restoration of motor function
following hemiplegia in man. Brain 74: 443–480
Wastell DG, Gray RH 1987 The numerical approach to
classification: A medical application to develop a typology
for facial pain. Statistics in Medicine 6: 137–146
EVALUATION OF PATIENT FOLLOWING STROKE 93
Physiother Theory Pract Downloaded from informahealthcare.com by Universite De Sherbrooke on 04/16/15
For personal use only.
