Disability and Rehabilitation, March 2006; 28(6): 389 395

Assessment of shoulder pain in hemiplegia: Sensitivity of the ShoulderQ

LYNNE TURNER-STOKES & DIANA JACKSON

Regional Rehabilitation Unit, Northwick Park Hospital, Kings College London

Abstract
Background. The ShoulderQ is a structured questionnaire designed to assess timing and severity of hemiplegic shoulder
pain (HSP), in order to target pain relief effectively. It includes both verbal and visual graphic rating scale questions, simply
presented for patients with language/visuo-spatial deficits following stroke.
Objective. To assess the sensitivity of the ShoulderQ to clinical improvement in shoulder pain following multi-disciplinary
intervention.
Design and setting. Retrospective analysis of serial questionnaires collected in the course of clinical treatment in an in-patient
neurological rehabilitation unit.
Subjects and interventions. Thirty consecutive adults with cognitive and communicative deficits, presenting with hemiplegic
shoulder pain following acquired brain injury. Multi-disciplinary treatment was delivered through an integrated care
pathway, and ShoulderQs recorded fortnightly, including at baseline and end of treatment.
Results. Changes on visual graphic rating scale (VGRS) were associated with verbal reports of improvement (rho 0.665,
p 5 0.001). Patients were divided retrospectively on the basis of their overall clinical response into responders (n 18) and
non-responders (n 12). Responders showed significant change in both VGRS and verbal scores, whereas the nonresponder group did not. A change in summed VGRS score of
3 showed 77% sensitivity and 91.3% specificity for
identifying the responders, with a positive predictive value of 93.3%. Summed VGRS scores of 2 had a negative predictive
value of 73.3%.
Conclusion. In this preliminary evaluation of clinical data, the ShoulderQ appears to provide a sensitive measure of shoulder
pain which is responsive to change in pain experience for those able to complete the questionnaire, despite the difficulties
that many of this group of patients may have in reporting their symptoms. Set alongside previously reported test-retest
reliability, the results support the utility of the ShoulderQ as a simple and practical tool for evaluation of shoulder pain in
patients with severe complex disabilities.

Keywords: Rehabilitation, integrated care pathway, hemiplegic shoulder pain, pain assessment

Introduction
Hemiplegic shoulder pain is a common complication
of stroke which may interfere with rehabilitation and
delay discharge from hospital. Following a systematic
review [1], an evidence-based integrated care pathway (ICP) has been developed [2]. This ICP offers a
holistic multi-disciplinary approach to management
of hemiplegic shoulder pain, including suitably
targeted analgesia, support for the paralysed arm,
and physical treatment as appropriate to the underlying presentation [1]. Regular accurate assessment
of shoulder pain is integral to the effective application
of the care pathway and decisions regarding pain
interventions.

Standardized questionnaires provide a useful

method for the systematic gathering of information
and are applied in many areas of medicine and social
life. However, patients with stroke often have
cognitive and communicative impairments which
may compromise their ability to complete a questionnaire. Patients with aphasia, for example, may
have difficulty understanding and responding to
verbally based questions, whilst those with visuoperceptual problems may find visual analogue scales
hard to interpret. It has been theorized that a vertical
visual analogue scale may be preferable to a horizontal scale [3], especially in patients with unilateral
visuo-spatial neglect. Formal studies, however, have
still found these to be disappointing [4,5].

Correspondence: Professor Lynne Turner-Stokes, Regional Rehabilitation Unit, Northwick Park Hospital, Watford Road, Harrow, Middlesex., HA1 3UJ, UK.
Tel: 020 8869 2800. Fax: 020 8869 2803. E-mail: lynne.turner-stokes@dial.pipex.com
ISSN 0963-8288 print/ISSN 1464-5165 online 2006 Taylor & Francis
DOI: 10.1080/09638280500287692

390

L. Turner-Stokes & D. Jackson

As part of the ICP we have developed the

ShoulderQ, a simple questionnaire to assess the
timing and severity of shoulder pain which includes
both verbal and visual graphic rating scale (VGRS)1
questions, to suit the patients strengths. A prescreening questionnaire (the AbilityQ) is used to
identify the individuals ability to respond to verbal
and VGRS questions, and to determine the type of
assistance that they may need to complete a
questionnaire successfully. Preliminary evaluation
has demonstrated test-re-test reliability of this system
and has confirmed the predictive value of the
AbilityQ in determining reliability of response to
the ShoulderQ [6].
The purpose of this further study is to determine
sensitivity of the ShoulderQ in terms of response to
changes in pain experience, as pain is resolved (or
not) through the interventions of the ICP.
Specific questions are:
1.

2.

3.

Does serial administration accurately record

changes in pain levels in keeping with the
clinical experience of individuals undergoing
treatment?
Does the ShoulderQ demonstrate pain reduction in individuals who make a clinical
response to treatment?
Does it accurately identify individuals whose
pain resolves (responders) from those whose
pain persists (non-responders), i.e., is it
sensitive and specific?

Methods
The study was conducted in a regional rehabilitation service for young adults (16 65 years) with
severe complex neurological disability (i.e., physical
cognitive and/or communicative deficits), mainly
following acquired brain injury of any cause
(including stroke, trauma, hypoxia etc). Of 39
consecutive adults who presented with hemiplegic
shoulder pain and were enrolled in the ICP over a
two-year period, 30 were demonstrated, through
screening with the AbilityQ, to be able to respond
to questionnaires with acceptable accuracy [6] and
were included in this analysis. Twenty-three
patients were able to complete both verbal and
VGRS scores correctly, and the remaining seven
were able to complete either VGRS (n 4) or
verbal (n 3) scores.
The ShoulderQ takes about 5 10 minutes to
complete for most patients. It includes verbal assessments of overall severity of pain (graded 0 4) and
pain in comparison to the previous week (the same,
better or worse on a 5-point scale). Three visual
graphic rating scales, with numbered tick-marks
(1 10) at 1 cm intervals, provide separate assessment

of the timing of shoulder pain (a) at rest during the

day, (b) on movement (e.g., during physiotherapy
sessions), and (c) at night.
Following baseline assessment, the ICP includes
routine fortnightly re-assessment and adjustment of
intervention as necessary. Administration of the
ShoulderQ at each assessment provides a serial
record of pain experience, in order that analgesia
may be accurately targeted to provide effective pain
relief, according to the treatment protocol. At the
final assessment, the overall resolution of pain is
recorded by the treating team as complete, partial
or non-responsive [7]. Assignment to these groups
is made on the basis of overall clinical evaluation in
consultation with the patient, rather than by any
formal measurement. For the purpose of analysis,
patients were divided into 18 responders (eight
complete and 10 partial) and 12 non-responders.
Data and statistical handling
Ethical permission has been obtained for this analysis.
Serial data from the ShoulderQ entered on the
customized ICP database (FileMaker pro) were
extracted and transferred to the Statistical Package
for Social Science v 10.5 (SPSS) for statistical
analysis. The majority of VGRS data were shown to
be within acceptable limits for normal distribution
(Kolmogorov-Smirnov p 4 0.05) with closely coinciding mean and median, so parametric tests were
applied throughout for consistency. Paired t-tests
were used to evaluate change scores, and group
differences were tested using independent sample ttests. The 5-level ordinal verbal severity scale was not
quite normally distributed and so the relationship
between this and visual analogue scales was assessed
using a non-parametric (Spearman rank) correlation.
Results
The mean age of the group was 47.2 years (+SE 2.2,
range 19 65), male: female ratio 2:1, and the mean
time on the ICP was 8 weeks (+SE 1.3). Fifteen
(50%) patients had left sided hemiparesis, 12 (40%)
right sided hemiparesis, and three (10%) had
bilateral deficits. All subjects were able to respond
to either verbal or VGRS questions with adequate
accuracy [6]: on the AbilityQ, 23 (77%) achieved at
least 3/4 accuracy on verbal questions and 27 (90%)
achieved at least 2/3 accuracy on VGRS questions.
There was no difference in demographic characteristics between the responder and non-responder
groups. Reasons for non-response were as follows:
. Three (25%) had very short admissions for
assessment only, and were discharged before
treatment had properly commenced;

Assessment of shoulder pain in hemiplegia

391

Figure 1. Serial VGRS scores in one of the six patients who made an initial response, but whose pain relapsed when they left the support of
their wheelchair and started to walk.

.
.

Three (25%) patients had very mild pain

(VGRS 1 2/10) to start with;
Six patients (50%) responded initially, but
developed increased pain when they left the
support of their wheelchair and started to walk.
These changes were recorded through the
serial ShoulderQ assessments as illustrated in
Figure 1.

The correlation between change in summed

VGRS score (pain at night during movement
at rest) and verbal response was only moderate
(Spearman rank correlation: rho .665, p 5 0.001),
reflecting the fact, as predicted on the AbilityQ, that
some patients were able to respond more reliably on
visual graphic rating scales than on verbal scales, and
others vice versa. The distribution of summed VGRS
scores between the non-responder, partial, and
complete resolution groups is shown in Figure 2.
Table I shows the change in VGRS scores from
baseline to final assessment. Taking partial and
complete pain resolution together in one responder group (n 18), significant changes were
recorded in VGRS scores for pain at night and
during physiotherapy; and also in the summed
VGRS score and on the verbal scale. The absence
of change in pain at rest is thought to reflect the low
starting score. By contrast, the non-responder group
showed no significant change on any parameter
(indeed pain during physiotherapy showed a significant increase in this group, which was again
consistent with reported symptoms).
A change in summed VGRS score of
3 showed
77% sensitivity and 91.3% specificity for identifying
the responders, with a positive predictive value of
93.3%. Similarly, summed VGRS scores of 2 had a
negative predictive value of 73.3%. A change in

Figure 2. Box Plot of change in summed VGRS score between

different responder groups. Note: T-tests show the partial
responder group to make significantly greater change than the
non-responders (mean difference 4.2, SE 2.1; CI 0.0, 8.4;
p 0.05), but significantly less change than the complete
responder group (mean difference 5.9, SE 2.4; CI 0.7, 11.2
p 0.03).

verbal score of
1 showed 66% sensitivity and 75%

specificity for identifying responders, with a positive
predictive value of 80%. Zero or minus changes (i.e.,
increase in pain level) carried a negative predictive
value of 60%.
Discussion
This preliminary evaluation includes a relatively
small number of patients, but nevertheless provides
encouraging evidence for the responsiveness of the
ShoulderQ to changes in hemiplegic shoulder pain,

392

L. Turner-Stokes & D. Jackson

Table I. Change in VGRS scores from baseline to final assessment.

Timing of pain

Start score Mean (SE)

Responders (n 18)
At night (VGRS)
On movement (VGRS)
At rest (VGRS)
Summed VGRS score
Verbal score
Non-responders (n 12)
At night (VGRS)
On movement (VGRS)
At rest (VGRS)
Summed VGRS score
Verbal score

Change Mean (SE)

Confidence Intervals*

Significance*

4.0
5.8
2.5
11.7
2.2

(0.8)
(0.6)
(0.6)
(1.5)
(0.2)

2.2
3.2
0.4
5.0
1.0

(0.7)
(0.8)
(0.6)
(1.7)
(0.3)

0.6,
1.5,
70.9,
1.4,
0.4,

3.7
5.0
1.7
8.6
1.6

0.009
0.001
0.533
0.009
0.002

2.3
3.9
2.7
9.7
1.8

(0.7)
(0.9)
(0.4)
(1.4)
(0.2)

70.5
71.7
0.4
0.7
70.2

(0.9)
(0.7)
(0.6)
(1.6)
(0.3)

72.7,
73.4,
70.9,
72.8,
70.9,

1.7
0.1
1.8
4.3
0.4

0.623
0.040
0.526
0.654
0.389

*Tested by paired sample t-tests.

despite the difficulties that this group of patients may

have in reporting their symptoms. These individuals,
who were identified by the AbilityQ as being able to
answer either verbal or VGRS questions correctly,
completed serial records consistently which were in
keeping with their clinical presentation, and the
ShoulderQ appeared to detect fluctuations in pain
experience appropriately.
A recognized limitation of this study is that the
data were collected in the course of clinical practice,
rather than under formal experimental conditions.
Because the serial ShoulderQ data are used to inform
clinical decision-making in the ICP, it may be
expected that they may have influenced the teams
overall clinical assessment of response to treatment
to some degree. On the other hand, in our ICP,
ShoulderQs are normally administered by the
medical staff in advance of the multi-disciplinary
clinic reviews, and are primarily used to influence
analgesic prescribing. During the course of assessment and review in the multi-disciplinary clinic, the
team tends to place more reliance on the patients
verbal and behavioural response to symptom enquiry
than on formally recorded measurement, so the
processes are independent to a certain extent.
Nevertheless, this may represent a source of bias.
These limitations accepted, in this study group,
both verbal and visual graphic rating scales were
sensitive to change and appeared to differentiate
between the responder and non-responder groups.
The change in summed visual graphic rating score
and in verbal severity score each appeared to give a
reasonably accurate overall assessment of pain
response; while changes in individual scores for pain
at night, at rest and on movement provided more
specific information on which clinical management
decisions, such as timing and type of analgesic
prescription, could be based.

Set alongside previously reported test-retest reliability [3], the results provide support for the utility of
the ShoulderQ as a simple and practical tool for
evaluation of shoulder pain in patients with complex
disabilities, who are demonstrated by pre-screening
with the AbilityQ to be able to complete the
questionnaire accurately. So far as the authors are
aware, there is currently no other validated tool for
assessing hemiplegic shoulder pain in this group of
patients. The instrument is now worthy of more
detailed evaluation.
Clinical points
. The ShoulderQ is designed to assess hemiplegic shoulder pain in patients with cognitive
and communicative deficits;
. Serially recorded ShoulderQs appear to detect
fluctuations in pain experience appropriately
in those able to complete the questionnaire;
. Both verbal and visual analogue scales were
sensitive to change and differentiated between
the responder and non-responder groups;
. Changes in individual VGRS scores for pain
at night, at rest and on movement provided
more specific information on which clinical
management decisions, such as analgesic
prescribing, could be based.
Acknowledgements
The authors gratefully acknowledges the hard work
of the RRU staff in collecting the data presented in
this study, and the cooperation of the patients to
whom it belongs. Special thanks are due to Lisa
Knight, Lisa Beatty, Anne OConnell, Hilary Stern,
Rana Das Gupta, Ajoy Nair and Jan Gavronski for
their roles in coordinating the ICP data collection.

Assessment of shoulder pain in hemiplegia

Financial support for preparation of the manuscript
was kindly provided by the Luff Foundation.
Note
1

Simple visual analogue scales (VAS) were used in

the original version of the ShoulderQ, but subsequent evaluation has demonstrated that numbered
visual graphic rating scales (VGRS) improve the
test-retest reliability, so these have now replaced
VAS scales in the current version of the ShoulderQ
and AbilityQ.

References
1. Turner-Stokes L, Jackson D. Shoulder pain after stroke: A
review of the evidence base to inform the development of an
integrated care pathway. Clin Rehab 2002;16(3):276 298.

393

2. Jackson D, Turner-Stokes L, Khatoon A, Stern H, Knight L,

OConnell A. Development of an integrated care pathway for
the management of hemiplegic shoulder pain. Disabil Rehab
2002;24(7):390 398.
3. Scott, Huskinson EC. Vertical or horizontal visual analogue
scales? Ann Rheum Dis 1979;38:560.
4. Price CIM, Curless RH, Rodgers H. Can stroke patients use
visual analogue scales? Stroke 1999;30:1357 1361.
5. Pomeroy VM, Frames C, Faragher EB, et al. Reliability of
a measure of post stroke shoulder-pain with and without
aphasia or unilateral spatial neglect. Clin Rehab 2000;14(6):
584 591.
6. Turner-Stokes L, Rusconi S. Screening for ability to complete a
questionnaire: A preliminary evaluation of the AbilityQ and
ShoulderQ for assessing shoulder pain in stroke patients. Clin
Rehab 2003;17(2):150 157.
7. Jackson D, Turner-Stokes L, Williams H, Das-Gupta R. Use of
an integrated care pathway: A third round audit of the
management of shoulder pain in neurological conditions. J
Rehab Med 2003;35(6):265 270.

394

L. Turner-Stokes & D. Jackson

Assessment of shoulder pain in hemiplegia

395