The Wessex Head Injury Matrix
The Wessex Head Injury Matrix
The Wessex Head Injury Matrix
net/publication/12376214
CITATIONS READS
574 14,308
6 authors, including:
Michael J Campbell
The University of Sheffield
478 PUBLICATIONS 27,069 CITATIONS
SEE PROFILE
Some of the authors of this publication are also working on these related projects:
All content following this page was uploaded by Agnes Shiel on 29 April 2015.
The Wessex Head Injury Matrix (WHIM) main scale: a preliminary report on a scale to assess
and monitor patient recovery after severe head injury
A Shiel, S A Horn, B A Wilson, M J Watson, M J Campbell and D L Mclellan
Clin Rehabil 2000 14: 408
DOI: 10.1191/0269215500cr326oa
Published by:
http://www.sagepublications.com
Additional services and information for Clinical Rehabilitation can be found at:
Subscriptions: http://cre.sagepub.com/subscriptions
Reprints: http://www.sagepub.com/journalsReprints.nav
Permissions: http://www.sagepub.com/journalsPermissions.nav
Citations: http://cre.sagepub.com/content/14/4/408.refs.html
What is This?
Received 8th October 1998; returned for revisions 23rd March 1999; revised manuscript accepted 4th August 1999.
ery is occurring, even though slow but subtle response, eye opening and verbal response. It has
progress is continuing over weeks or months. been shown to be sensitive to change in the early
Objective evidence of progress could perhaps stages of coma and is still considered to be the
enable more accurate long-term predictions to be best method of monitoring response in the acute
made for individuals. Failure to recognize such stage of coma. It is the only scale to be widely
changes could suggest a falsely poor prognosis used in routine clinical practice in the UK. In
and impair staff motivation and the quality of part, this is probably due to the fact that brain-
patient care. injured people are initially cared for in a wide
In the early stages of recovery, systematic mon- range of different hospital settings, including gen-
itoring of the recovery of functions such as motor eral orthopaedic and surgical wards which may
skills, cognitive abilities, social interaction and not be geared specifically for detailed neurologi-
the control of behaviour cannot be undertaken cal or cognitive assessment and monitoring.
because of the lack of appropriate instruments. Current scales for monitoring recovery and
Wilson2 argued that methods of assessment for classifying outcome after patients after severe
severely head-injured patients were inadequate head injury have been shown to have limited use.
because of the following factors. A comprehensive review of the most commonly
used scales can be found in Horn et al.5 Most are
1) They do not provide a clear relationship not used because of the following limitations.
between performance on tests and perfor-
mance in everyday life. 1) The categories are broad so it may not be
2) They cannot be employed at all stages fol- possible to detect subtle behavioural changes
lowing traumatic brain injury. For example, e.g. confused/nonagitated and confused/
most neuropsychological tests cannot be agitated on the Rancho Los Amigos Scale.3
employed on people as they are emerging 2) Items in the scales cross several behavioural
from coma. dimensions (e.g. cognitive functioning and
3) They cannot be used across a spectrum of social awareness) so it may be difficult to
varying degrees of severity. For example, identify improvement in a single area.
many tests cannot be used with people with 3) Patients recovering from head injury may
severe motor, sensory or cognitive impair- exhibit behaviour appropriate to more than
ments, although precise and objective one category. For example, Rancho Los Ami-
measures of these impairments might be gos level 3 is described as ‘confused agitated’
crucial. and level 4 as ‘confused nonagitated’.
4) They have very little (or only very crude) pre- Patients may exhibit both behaviours on dif-
dictive power regarding final outcome. ferent occasions or in different environments.
5) They provide inadequate information for 4) The scales are rating scales rather than obser-
designing and guiding treatment pro- vations or assessments and, as such, are open
grammes. to subjective interpretation which may not
reflect the true level of the behaviour accu-
Several scales have been developed to monitor rately.
recovery after emergence from coma. Examples 5) The scales do not specify the sequence of
include the Rancho Los Amigos Scale3 and the recovery in sufficient detail. For example,
Glasgow Coma Scale4 (GCS). The Rancho Los item 4 on the Glasgow Outcome Scale
Amigos Scale is an eight-item scale with broad (GOS)6 is defined as ‘can travel by public
categories. When assessed on this scale, patients transport and work in a sheltered environ-
may show behaviours which are appropriate to ment and can, therefore, be independent in
more than one category at the same time and can so far as daily life is concerned.’ The ques-
also change categories several times in the same tion remains as to whether ‘ability to travel
day, e.g. ‘confused and agitated’ and ‘confused on public transport’ precedes, follows on, or
and non-agitated’. The GCS is one of the most always goes along with ‘ability to work in a
widely used measures of level of consciousness. sheltered environment’.
It assesses three aspects of behaviour: motor 6) Most scales have ordinal scoring; that is, a
Downloaded from cre.sagepub.com at NUI, Galway on June 3, 2014
410 A Shiel et al.
cognitive social or communicative behaviour and which the behaviour was first observed to occur.
was included in all three scales initially. Where subjects were seen daily, this allowed cal-
Examination of the behaviours in each scale culation of the number of days between the
showed that overlaps occurred across four sub- injury and the first observed occurrence of the
scales : cognition, communication, social behav- behaviour. These data were used to calculate the
iour and visual awareness. Furthermore rank order of recovery of items of behaviour.
recording the assessment became more compli- Where the subjects were seen less frequently, it
cated when behaviours were divided into sub- could not be assumed that the date of an obser-
scales as patients tended to recover early items vation of a new behaviour was the date on which
in several domains simultaneously so that all rel- it would first have been observed. In such cases,
evant subscales had to be administered concur- ‘first observations’ of behaviour tended to clus-
rently. In order to avoid these difficulties, the ter. This created disparities in accuracy of timing
subscales were amalgamated to form a single cog- between subjects and the large amount of ‘miss-
nitive scale. Duplicated behaviours were ing data’ presented a challenge for analysis.
removed and the result was a 58-item scale of Traditional analysis requires complete or
cognitive behaviours. mainly complete data sets. However, Allen and
The gross motor and upper limb scales have Yen15 described a technique of developing ordi-
undergone no further development to date and nal scales whereby rather than examining scaling
the feeding, washing, dressing and continence properties, a categorical analysis between pairs of
scales were abandoned. There were two reasons variables is carried out. For each pair of vari-
for this: very few data on these behaviours had ables, it is determined whether one is likely to
been collected and it was not possible to record recover sooner than the other in more than 50%
a sequence of recovery of these items as they of the observed cases. With the data collected in
were so strongly affected by environmental fac- this study, each ‘first observed on’ date was used
tors. When patients were looked after on general to identify a rank order of emergence of behav-
wards, nursing staff tended to wash, dress and iours and a matrix of paired comparisons was
feed patients rather than allow them to attempt generated. The procedure is described fully else-
to carry out these tasks themselves. where.16 The advantage of using this technique
was that all data, including that which was incom-
Development of operational definitions plete could be used. This allowed analysis of
Each behaviour included in the scale was oper- incomplete data and the data could be used
ationally defined.14 Operational definitions need to identify a potential sequence or trend of
to be concise and concrete and, for some com- recovery.
plex behaviours, salient and more easily identi-
fied components of the behaviour were Reliability
described. Although the fine definitions do not The reliability of the scale was evaluated with
therefore encompass all aspects of the behaviour a different set of subjects and raters. Two
of which these elements were a part, it was clear researchers, unfamiliar with the scale, underwent
that without definitions there was little chance of training in administration of the scale using video
developing a useable tool. There were three recordings of subjects and behaviours were
stages in the development of these operational recorded on the WHIM. As reliability between
definitions. First, draft definitions were devel- raters at this stage was found to be unacceptably
oped by the research observers. Second, they poor, further training was provided. Subse-
were discussed and revised by the whole project quently, a further group of subjects were
team. Third, they were tested prospectively using recruited using the same recruitment criteria as
subjects and finally, further revised as necessary. for the scale development. These subjects were
observed daily for times ranging from 15 minutes
Analysis of data to 21/2 hours. Two types of reliability were exam-
All behavioural observations were recorded ined: inter-rater reliability and test–retest relia-
and particular attention was paid to the date on bility. Inter-rater reliability referred to the degree
Downloaded from cre.sagepub.com at NUI, Galway on June 3, 2014
Wessex Head Injury Matrix 413
to which trained raters agree and test–retest reli- apparently ‘random eye movement’. Others
ability referred to the stability of measures over appear to be purposeful in nature, for example
at least two observational sessions. To test inter- looking at a person briefly, looking at the person
rater reliability, researchers observed the subject who is talking or following with the eyes a per-
and recorded the behaviours observed during the son moving in the line of vision. All of these
same session; test–retest reliability was evaluated behaviours have been noted to occur prior to
by assessing the subject twice on the same day at obeying a command.
least two hours apart. Reliability was tested using The sequence of behaviours between obeying
kappa statistics which gave an agreement mea- a command and the first spoken utterance is
sure for both the occurrence and nonoccurrence dominated by visual behaviours (Table 2). We
of behaviour corrected for chance agreement can then see the emergence of behaviours
between raters.17 required for social interaction and communica-
tion, for example ‘vocalizing’, showing selective
responses to preferred people and frowning to
Results show dislike. These behaviours were observed
both in patients who had no mechanical barrier
Eighty-eight subjects were recruited to the study. to speech and those with a tracheostomy and they
They ranged in age from 14 to 67 years. Median appeared to emerge prior to return of spoken
age was 26 years (mean = 30 years, SD =14). communication.
Median coma duration was six days (mean = 14 The next segment of the scale (Table 3)
days, SD = 21). Median duration of PTA was 30 includes behaviours which imply an increasing
days (mean = 56 days, SD = 65). recovery of attention and cognitive organization
including such items as looking at TV or books,
Main scale being distracted by any external stimulus and
Paired comparisons between each pair of items then continuing to being distracted but becoming
were calculated to identify an order of recovery. able to return to the task. Smiling spontaneously
The procedure has been described in detail else- typically returns at this stage.
where.16 The main scale of the matrix contained The final segment of the sequence demon-
58 items of behaviour. These are described in strates the recovery of orientation and continu-
four sections. ous memory and signals emergence from PTA
The first part of the scale includes basic behav- which was assessed prospectively using a modi-
iours (Table 1), some of which may be partly fied version of the Westmead PTA test (Table
reflexive in origin for example, ‘eye opening’ and 4).12 As can be seen from the behaviours, the ini-
tial recovery of orientation and memory func-
Table 1 Items of behaviour 1–15
Table 2 Items of behaviour 16–29
1) Eyes open briefly
2) Eyes open for extended period 16) Turns head/eyes to look when someone is talking
3) Eyes open and move but do not focus on 17) Watches person moving in line of vision
object/person 18) Tracks for 3–5 seconds
4) Attention held momentarily by dominant stimulus 19) Vocalizes to express mood or needs
5) Looks at person briefly 20) Tracks a source of sound
6) Volitional vocalization, to express feelings 21) Shows selective response to preferred people
7) Distressed when cloth put on face 22) Maintains eye contact over 5 seconds
8) Makes eye contact 23) Removes cloth from face by headshake, hand grasp,
9) Looks at person talking etc. × 3
10) Expletive utterance (‘Get off!’, etc.) 24) Silent mouthing
11) Eyes follow person moving in line of vision 25) Frowns, grimaces, etc. to show dislike
12) Looks at person giving attention 26) Is able to ignore distraction
13) Closes eyes and becomes quiescent when cloth 27) Looks at object when requested
put on face 28) Imitates gesture (blink x 2, thumb up, etc.)
14) Mechanical vocalization (with yawn, sigh, etc.) 29) Indicates understanding by headshake, nod,
15) Performs physical movement on verbal request gesture, etc.
tions (for example, the ability to say what part of communication aid. There is increasing evidence
the day it is) may predate first ‘day-to-day’ mem- of social interaction including initiation of con-
ory item – knowing the name of a member of versation and the ability voluntarily to attract
staff. This segment also introduces evidence of attention of others.
learning new information such as the name of a
staff member, or a new skill such as the use of a Reliability of the scale
Reliability was evaluated using 25 subjects
recruited to a second study over a nine-month
Table 3 Items of behaviour 30–46 period. The inclusion criteria were the same as in
30) Seeks eye contact
the first study. The median age of the reliability
31) Monosyllabic or single words in response to sample was 36 years (mean = 42, SD 21 years)
questions and median duration of coma was seven days
32) Looks at, and apparently explores, pictures, (mean = 13, SD = 20 days). There were no sig-
magazine, TV etc. nificant differences between the study sample and
33) Switches gaze from one person to another,
spontaneously this group on either of these parameters.
34) Speech is fluent but rambling. Lots of words but Initial data showed that neither inter-rater nor
meaning hard to discern test–retest reliability reached acceptable levels
35) Looks for object that has been shown and then when the scale was employed by previously
removed from line of vision
36) Can attend to task, TV etc., but concentration is
untrained observers immediately after a brief
vulnerable training session. (κ = 025 to 0.84 for inter-rater
37) Monosyllabic or single words to express mood or reliability and – 0.66 to 0.12 for test–retest relia-
need bility). Following this, more extensive training
38) Is momentarily distracted by external stimulus but was provided and an amended data collection
can return to task
39) Can find a specific playing card from a selection of procedure introduced. The training procedure
four comprised practice using the scales with video
40) Smiles demonstrations of the behaviours. Training took
41) Uses writing, typing or other communication aid, approximately two hours. The data-collection
but is hard to understand
42) Can say what part of day it is
procedure was amended so that instead of
43) Brief phrases recording new behaviours only, raters recorded
44) Points with eyes all the behaviours observed in a session. Follow-
45) Initiates conversation ing such training and using the amended record-
46) Vocalizes to attract attention ing system the level of reliability reached was
acceptable for both inter-rater and test–retest
reliability. Mean kappa scores were 0.86 for inter-
Table 4 Items of behaviour 47–58 rater reliability and 0.74 for test–retest reliability.
The range of scores for test retest reliability
47) Conventional speech usage but with very few words
48) Uses one or two gestures (scores ranged from 0.62 to 1 for inter-rater reli-
49) 1 or 2 orientation items correct (day of week, month, ability and 0.22 to 1 for test–retest reliability) was
year, age, place) considerably greater than for inter-rater reliabil-
50) Knows the price of three common objects ity. The lower scores were due to the fact that
51) Recognizes coins (eye-point or touch named coin)
52) Knows the name of one member of staff
two subjects changed significantly between the
53) Names or indicates left and right on self two assessments (carried out 4 hours and 8 hours
54) Uses writing, typing or other communication aid apart respectively) but all other kappa scores
fluently were high as confirmed by the mean kappa score
55) 3–5 orientation items correct (0.74).
56) Remembers something from the day before
(e.g. show a coin, key, watch, etc. from your
pocket and say you will ask them to remember it
tomorrow)
57) Remembers something from earlier in the day
(e.g. ‘Have you been to physio yet?’)
58) Completes PTA test
equally likely in those who recover very rapidly cognitive functioning. In: Professional Staff
and those who recover very slowly. Association of Rancho Los Amigos Hospital eds.
In the introduction, several criticisms of the Rehabilitation of the head injured adult:
comprehensive physical management. Downey, CA:
scales already available were made. These Rancho Los Amigos Hospital Inc, 1987.
included their poor relationship with everyday 4 Teasdale G, Jennett B. Assessment of coma and
life, poor predictive power, their lack of useful- impaired consciousness: a practical scale. Lancet
ness in designing and guiding treatment pro- 1974; i: 81–84
grammes, and their lack of ability to detect subtle 5 Horn S, Shiel A, McLellan DL, Campbell M,
changes in functioning. Other criticisms included Watson M, Wilson BA. A review of behavioural
assessment scales for monitoring recovery in and
the fact that scales crossed several behavioural after coma with pilot data on a new scale of visual
dimensions and could not be employed at all awareness. Neuropsychol Rehabil 1993; 3: 121–137.
stages after head injury. To what extent could 6 Jennett B, Bond M. Assessment of outcome after
similar criticism be levelled at the WHIM? The severe brain damage – a practical scale. Lancet 1975;
above account demonstrates that an explicit 1: 480–84.
attempt has been made to address these issues. 7 Wilson BA, Vizor A, Bryant T. Predicting severity
of cognitive impairment after severe brain injury.
Although other studies have examined reliability Brain Inj 1991; 5: 189–97.
and validity of the scale further,19 more prospec- 8 Sparrow S, Balla DA, Cicchetti DV. The Vineland
tive studies are required to clarify the order of Adaptive Behaviour Scales. Circle Pines, MN:
recovery and the individual variability of recov- American Guidance Service Ltd, 1984.
ery patterns. 9 Eson ME, Yen JK, Bourke RS. Assessment of
recovery from serious head injury. J Neurol
The Wessex Head Injury Matrix will be available Neurosurg Psychiatry 1978; 41: 1036–42.
10 Goldfried FC, Linahan MM. Basic issues in
from Thames Valley Test Company, 7–9 The behavioural assessment. In: Ciminero AR, Calhovn
Green, Flempton, Bury St Edmonds, Suffolk, KS, Adams HE eds. Handbook of behavioural
UK. assessment. New York: John Wiley & Sons, 1977.
11 Horn S, Watson M, Campbell MJ, McLellan DL,
Acknowledgements Wilson BA. Assessing post traumatic amnesia. J Clin
This research was supported by grants from the Exp Neuropsychol, 1991; 13: 19 (abstract).
12 Medical Disability Society. The management of
Medical Research Council (grant number traumatic brain injury: a Working Party Report of
SPG8715555) and Remedii through the Wessex the Medical Disability Society. London: Medical
Medical School Trust (grant number 309/22 Disability Society, February 1988.
WMST). 13 Brinkmann R, von Cramon D, Schulz H. The
We would like to thank the subjects and their Munich Coma Scale. J Neurol Neurosurg Psychiatry
families and the staff at Southampton Rehabili- 1976; 39: 788–93.
14 Kazdin A. Research design in clinical psychology.
tation Unit and the TBI Team at Poole Hospital New York: Harper and Row, 1980.
for participating in this project. We also thank 15 Allen MJ, Yen WM. Introduction to measurement
the anonymous reviewer for Clinical Rehabilita- theory. Monterey, CA: Brooks Cole, 1979.
tion for helpful comments and recommendations. 16 Watson M, Horn S, Wilson BA, Shiel A, McLellan
Our thanks also to Karen Hinton for typing DL. The application of a paired comparisons
and layout of the paper. technique to identify sequence of recovery after
severe head injury. Neurospsychol Rehabil 1997; 7:
441–58.
17 Nourish MJ/SPSS Inc. SPSS for Windows. Base
References system users guide. Chicago: SPSS Inc, 1993, 214–15.
18 Adamovich BB. Information processing, cognition,
1 Chamberlain MA, Neumann V, Tennant A eds. attention, and communication following closed head
Traumatic brain injury rehabilitation. London: injury. Folia Phoniatr 1990; 42: 11.
Chapman and Hall, 1995. 19 Majerus S, Van der Linden M, Shiel A. The Wessex
2 Wilson BA. Adapting ‘portage’ for neurological Head Injury Matrix and the Glasgow/Glasgow-Liège
patients. J Rehabil Res Dev 1988; 7: 6–8. Coma Scale: a validation and comparison study.
3 Hagen C, Malkmus D, Durham P. Levels of Neuropsychol Rehabil 2000; 10: 167–84.