Professional Documents
Culture Documents
net/publication/23959454
CITATIONS READS
291 10,909
1 author:
Lynne Turner-Stokes
King's College London
259 PUBLICATIONS 6,156 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 Lynne Turner-Stokes on 29 January 2014.
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/23/4/362.refs.html
What is This?
In the paper by L. Turner-Stokes, there were errors in the formulas presented on pages 364 and 367. The
correct formulas are published below. The publisher apologises for this error and the inconvenience
caused to the author, editors, and readers.
Where wi ¼ the weight assigned to the ith goal (if equal weights, wi ¼ 1), xi ¼ the numerical value
achieved (between 2 and þ2), r (rho) ¼ the expected correlation of the goal scales.
p P P
Starting with: ð0:7 w2i þ 0:3ð wi Þ2 Þ
for scaling. This article gives practical guidance on positive therapeutic value in encouraging the
measuring outcome using this method. patients to reach their goals.9 In particular, the
more formalized process of goal setting before
the start of intervention, and defining and agreeing
What is goal attainment scaling? expected levels of achievement with the patient
and their family, supports the sharing of informa-
tion at an early stage of rehabilitation and the
Goal attainment scaling (sometimes abbreviated negotiation of realistic goals.
to ‘GAS’), is a method of scoring the extent to Second, there is growing evidence that goal
which patient’s individual goals are achieved in attainment scaling is a good measure of outcome,
the course of intervention. In effect, each patient being at least as sensitive to change (and
has their own outcome measure, but this is scored probably more so) when compared with standard
in a standardized way as to allow statistical measures.12–15 It potentially avoids some of the
analysis. problems of standardized measures including:
In contrast to ‘traditional’ standardized
measures (e.g. the Barthel ADL Index5) which
floor and ceiling effects;
comprise a standard set of tasks (items) each
lack of sensitivity – particularly of global
rated in a standard way according to pre-set
measures, where individuals make change in
‘levels’, when using goal attainment scaling, tasks
one or two important items but this change is
are individually identified to suit the patient, and
lost in the overall scores, because a large
the levels are individually set around their current
number of irrelevant items do not change; and
and expected levels of performance.
disjunction between a patient’s main concerns
Measurement through goal attainment scaling
and the domain(s) of the standard measure.
was first introduced in health care in the 1960s
by Kiresuk and Sherman6 for assessing outcomes
The literature encompasses a range of
in mental health settings. Since then it has been
different approaches to scaling and measuring
modified and applied in many other areas
the achievement of goals, ranging from simple
including:
recording of goals achieved, partially achieved or
not achieved16 to rating scales of up to seven
elderly care settings,7,8 points.17 The procedure described below is based
chronic pain,9 on that used in the context of upper limb spasticity
cognitive rehabilitation,10 by Ashford and Turner-Stokes.15 It represents an
amputee rehabilitation.11 attempt to establish a more consistent approach.
This method offers a number of potential
advantages as an outcome measure for rehabilita-
tion. First, because goal setting is already a part How is achievement rated?
of routine clinical practice in many centres, it
builds on this already established process to An important feature of goal attainment scaling
encourage: is the establishment of criteria for a ‘successful’
outcome for that individual, which are agreed
communication and collaboration between the with the patient and family before intervention
multidisciplinary team members as they meet starts so that everyone has a realistic expectation
together for goal setting and scoring, and of what is likely to be achieved, and agrees that
patient involvement. this would be worth striving for.
Each goal is rated on a 5-point scale, with the
There is emerging evidence that goals are more degree of attainment captured for each goal area:
likely to be achieved if patients are involved in
setting them. Moreover, there is Downloaded
also evidence If the patient achieves the expected level, this is
from cre.sagepub.com at DUKE UNIV on January 16, 2013
that using goal attainment scaling may have scored at 0.
364 L Turner-Stokes
If they achieve more than the expected outcome For practical purposes, according to Kirusek
this is scored at: and Sherman, most commonly approximates
þ1 (somewhat more) to 0.3, so the equation simplifies to:
þ2 (much more) 10ðWi Xi Þ
If they achieve less than the expected outcome Overall GAS ¼ 50 þ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
this is scored at: 0:7W2i þ 0:3ðW2i Þ
1 (somewhat less) or
2 (much less) (NB: Mathematically challenged readers take
heart – there are calculation tables in the book
by Kiresuk.19 Alternatively, a simple spreadsheet
Note that this scaling can apply both when calculator is available on the Clinical
improvement and when deterioration is the Rehabilitation server or from the author!)
expected direction of change. The important In effect, therefore, the composite goal score
point is to ensure that doing better than expected (the sum of the attainment levels the relative
is associated with a positive score, and vice versa. weights for each goal) is transformed into a
Some goals will be more important to the patient standardized measure or T-score with a mean of
than others; and some goals set may be more 50 and standard deviation of 10. If goals are set
difficult for the rehabilitation process to attain in an unbiased fashion so that results exceed
than others. Goals may therefore be weighted to and fall short of expectations in roughly equal
take account of the relative importance of the proportions, over a sufficiently large number of
goal to the individual, and/or the difficulty that patients, one would expect a normal distribution
the rehabilitation team anticipates in achieving of scores.
it.11 Whilst Kiresuk and Sherman allowed for Demonstrating that the mean goal attainment
goal weighting in the formula below, the benefits T-score for the study population is around
of goal weighting remain uncertain (see below),18 50 is a useful quality check of the team’s ability
and it should be regarded as optional. to set and negotiate achievable goals. If a team
attempts to inflate their results by setting goals
over-cautiously, the mean score will be 450.
Similarly, if they are consistently overambitious
How is the overall goal attainment scaling it will be 550. In some contexts, it may be
score calculated? argued that setting more challenging goals
may be associated with greater improvement
and therefore that mean scores of 550
The method allows one to set as many or as few are not necessarily a bad thing.3 In others,
goals as wished, and still gives a single numerical however, it may reflect a less than perfect
outcome. However, goal setting can be understanding of the factors (including external
time-consuming and, in our experience, three to factors) that mitigate against goal attainment, or
five goals usually represent a feasible number to a failure to negotiate realistic expectations for
capture the patient’s key priorities. The goal outcome.20
outcome scores are then incorporated into the
single aggregated T-score by applying the
following formula:
Procedure
10ðWi Xi Þ
Overall GAS ¼ 50 þ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ð1 ÞW2i þ ðW2i Þ The procedure needed is as follows.
priority goal areas (with the help of the team) for function, but realistically using the affected hand
achievement by an agreed date (usually as a prop is the expected outcome, then the active
discharge or the end of the programme). Set function task can be set at level þ2, and use as a
goals should follow the SMART principle – that prop at level 0. This way, the patient’s goal is not
is, they should be specific, measurable, attainable, totally dismissed, but is clearly defined as beyond
realistic and timely. The patient largely determines the level of expectation.
the domains of goals, and the team determines the
levels likely to be achieved. More detail is given by
Bovend’Erdt et al.4 (4) Scoring baseline
Because change is built into the way that
goal attainment scaling is derived, the outcome
(2) Weight the goals (optional) T-score is by definition a measure of change,
Goals may be weighted by asking the patient and avoids the computation of change scores
to rank them in order of importance or by applying which may be unreliable where baseline and
a simple weighting scale, using the 4-point scale in outcome scores are highly correlated.22
Table 1. Similarly the team may rank or weight the Nevertheless, some authors advocate the recording
goals the terms of the anticipated difficulty in of baseline scores.10,11,13 These are usually
achieving them. The weight then attributed rated 1, unless there is no clinically plausible
to a goal is the multiplicand of importance and worse condition with respect to that goal,
difficulty: weight ¼ importance difficulty. in which case the baseline rating is 2. An
Using this system and scoring, goals that carry a aggregated baseline score may then be calculated
zero weighting are effectively cancelled out of the using the same formula.
formula, so that the weighting scale resolves to a
score of 1–3.
If goal weighting is not used, values of ‘1’ are (5) Goal attainment scoring
simply applied to ‘weight’ in the formula. Finally, the outcome score for each goal is rated
at the appointed review date, judging actual
performance against the predefined levels.
(3) Define expected outcome Ideally, this is undertaken by the team in
The ‘expected outcome’ is the most probable conjunction with the patient/family.
result if the patient receives the expected The goal attainment ‘T-score’ is then calculated
treatment Ideally descriptors should also be pre- by applying the formula (or by using the simple
defined for each of the achievement levels (2, 1, spreadsheet calculator).
0, þ1, þ2) and recorded in a ‘follow-up guide’.12,21 If baseline ratings were recorded, goal
Each goal level is defined by the team or attainment change scores may be determined by
investigator, and should be as objective and subtracting the baseline aggregate score from that
observable as possible. This process also at outcome.10,11,13 However, in practice the change
provides an opportunity to negotiate with the score is usually highly correlated with the T-score,23
patient if they have unrealistic expectations. and offers little further advantage.
For example if the patient wants active hand
Pain 3 2 6 1 þ2
Many have reported that applying goal attainment Ease of 2 2 4 1 0
scaling in the way originally described by Kiresuk dressing
is too time-consuming for routine clinical use. On Driving 1 2 2 2 1
Sum ¼ 12
the Regional Rehabilitation Unit at Northwick
Table 2 Worked example patient AB: baseline, expected and achieved goal levels
(1) Reducing She had severe We expected to reduce Her pain had completely
shoulder pain shoulder pain rating her pain to around 4/10 resolved, both day and night
8/10 at rest, on and to limit her night so she scored þ2, since the
movement and also time waking to once outcome could not have been
disturbing her sleep a night any better
at night (Score 1)
(2) Ease of She needed help to We expected that she She achieved her goal of being
dressing dress her upper would be able to dress able to dress her upper body
body (Score 1) her upper body unaided without help (Score 0)
(3) Able to drive She was unable We expected that she Although she had had a successful
to drive (Score 2) would be able to return driving assessment she was still
to driving using an waiting for her adapted car to
adapted car arrive and so was not driving at
the point of discharge – she
therefore scored 1, even though
Downloaded from cre.sagepub.com at DUKE UNIV on January 16, 2013 this was beyond our control
Goal attainment scaling in rehabilitation 367
Box 1 shows the application of the GAS formula reduction) tend to be critically important and so
Applying the formula:
feature regularly on their wishlists.
10ðWi Xi Þ
The wording of individual goals can be time-
Overall Goal Attainment Score ¼ 50 þ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi consuming, and some of these goals are well-
ð0:7Wi2 þ0:3ðWi Þ2 Þ
defined in existing hierarchical scales. For
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Starting with: ð0:7Wi2 þ 0:3ðWi Þ2 Þ we have:
example, a goal to ‘reduce pain’ may be defined
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
in terms of expected score on a 10-cm visual
ð0:7 ð36 þ 16 þ 4Þ þ 0:3 ð12Þ2 Þ analogue score. Similarly, improved walking
ability may typically be defined in terms of
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
¼ ð39:2 þ 43:2Þ distance covered, the type of support needed, or
¼ 9.07 the type of terrain. Over time we have developed
some menus of pre-worded goal statements in
Then applying the full formula: these more common areas that span a wide
10ð14Þ range of abilities. Suitable goals may be chosen
The baseline score is 50 þ ¼ 50 þ (140/9.07)
¼ 50–15.4 ¼ 34.6 9:07 or adapted from these menus to save starting
10ðþ10Þ from scratch with each new patient. We have
The outcome score is 50 þ ¼ 50 þ (100/9.07) also undertaken mapping of goals onto the
¼ 50 þ 11.0 ¼ 61.0 9:07
International Classification of Functioning
The change in score, should one wish to measure it, Disability and Health (ICF).24
is therefore 26.4
Without the goal weighting, the baseline outcome and
change scores would be 31.7, 54.6
and 22.8, respectively
In this particular case, goal attainment scoring shows (c) Weighting
a better than expected result, but not all cases will be Although weighting for ‘importance’ has
as positive as this. a consistent effect on overall GAS scores in
the expected direction, weighting for
‘difficulty’ can, in some circumstances, lead to a
programme or admission is agreed with the perverse bias. In practice, the weighted and
patient/family at the start of the programme. unweighted scores are very closely correlated.18
These are then broken down into a series Therefore, it is perfectly adequate, and
of ‘staged’ (short-term) goals towards these simpler for many purposes, to use unweighted
objectives, which are reviewed at fortnightly scores in the calculation (i.e. a weighting of
intervals throughout the programme. 1 throughout).
Goal attainment scaling is not applied to every On the other hand, recording the importance
staged goal, but just to three to five key objectives and difficulty of key goals can be helpful for
that form key priorities for the patient. Baseline qualitative interpretation. For example, if
rating is undertaken within the first 7–10 some goals were achieved and others not, it is
working days of admission; the outcome level pertinent to know which of them were most
of achievement is rated just once at discharge important to the patient. Similarly, if a goal
(or at a defined team review date for long admis- was not achieved, it is often helpful for the
sions) – usually about three months after the team to reflect on the extent to which they
original goal setting. had already identified this as a difficult goal
with a correspondingly greater chance of failure.
Over time, this team reflection may be expected
to lead to more accurate prediction of goal
(b) Wording of goals attainment. Therefore in our unit, we record
Whilst we try to encourage the patient to iden- both ‘importance’ and ‘difficulty’ as in Table 1,
tify personal goals that address areas not already for qualitative purposes, but currently we only
covered by our routinely recordedDownloaded
standardized include ‘importance’ in our goal-weighting for
from cre.sagepub.com at DUKE UNIV on January 16, 2013
measures, some goals (such as walking and pain aggregate T-scores.
368 L Turner-Stokes
Baseline Baseline
score -1 score -2
At baseline At outcome
A lot +2 +2
With respect to
this goal, does the Better than
patient have? expected A little +1
+1
Yes
Worse −2
Figure 1 Algorithm for converting verbal scoring by clinicians to the 5-point goal attainment scores. This algorithm
allows clinicians to record goal attainment without reference to the numeric scores, and so avoids the perceived negative
connotations of zero and minus scores. Providing the level at baseline is known, ‘partial achievement’, ‘no change’
and worse can be translated by computerization outside the clinical arena. This incidentally offers the opportunity to
compare the effect of using different scoring systems such as (1, 2 and 3) or (0.5, 1 and 2) and this work is
currently underway.
7 Stolee P, Rockwood K, Fox RA, Streiner DL. 17 Cusick A, McIntyre S, Novak I, Lannin N,
The use of goal attainment scaling in a geriatric Lowe K. A comparison of goal attainment scaling
care setting. J Am Geriatr Soc 1992; 40: 574–78. and the Canadian Occupational Performance
8 Stolee P, Stadnyk K, Myers AM, Rockwood K. Measure for paediatric rehabilitation research.
An individualized approach to outcome Pediatr Rehabil 2006; 9: 149–57.
measurement in geriatric rehabilitation. 18 Cardillo JE, Smith A. Psychometric issues.
J Gerontol Series A – Biol Sci Med Sci 1999; 54: In Kiresuk TJ, Smith A, Cardillo JE. eds.
M641–M647. Goal attainment scaling: applications, theory
9 Williams RC, Steig RL. Validity and therapeutic and measurement. Lawrence Erlbaum, 1994,
efficiency of individual goal attainment 190–92.
procedures in a chronic pain treatment centre. 19 Kiresuk TJ, Chaote RO. Applications of goal
Clin J Pain 1987; 2: 219–28. attainment scaling. In Kiresuk TJ, Smith A,
10 Rockwood K, Joyce B, Stolee P. Use of goal Cardillo JE. eds. Goal attainment scaling:
attainment scaling in measuring clinically applications, theory and measurement.
important change in cognitive rehabilitation Lawrence Erlbaum, 1994, 61–104.
patients. J Clin Epidemiol 1997; 50: 581–88. 20 McCrory P, Turner-Stokes L, Baguley IJ et al.
11 Rushton PW, Miller WC. Goal attainment Botulinum toxin A for treatment of upper limb
scaling in the rehabilitation of patients with lower- spasticity following stroke: a multi-centre
extremity amputations: a pilot study. Arch Phys randomized placebo-controlled study of the
Med Rehabil 2002; 83: 771–75. effects on quality of life and other person-
12 Rockwood K, Stolee P, Fox RA. Use of goal centred outcomes. Submitted for publication 2008.
attainment scaling in measuring clinically 21 Kiresuk T, Smith A, Cardillo JE. Goal attainment
important change in the frail elderly. J Clin scaling: application, theory and measurement.
Epidemiol 1993; 46: 1113–18. Lawrence Erlbaum, 1994.
13 Gordon JE, Powell C, Rockwood K. Goal 22 Becker H, Stuifbergen A, Rogers S, Timmerman G.
attainment scaling as a measure of clinically Goal attainment scaling to measure individual
important change in nursing-home patients. change in intervention studies. Nurs Res 2000; 49:
Age Ageing 1999; 28: 275–81. 176–80.
14 Fisher K, Hardie RJ. Goal attainment scaling 23 Khan F, Pallant JF, Turner-Stokes L. Use of
in evaluating a multidisciplinary pain management goal attainment scaling in inpatient rehabilitation
programme. Clin Rehabil 2002; 16: 871–77. for persons with multiple sclerosis. Arch Phys
15 Ashford S, Turner-Stokes L. Goal attainment Med Rehabil 2008; 89: 652–59.
for spasticity management using botulinum toxin. 24 World Health Organization. International classifi-
Physiother Res Int 2006; 11: 24–34. cation of functioning, disability and health.
16 Macpherson R, Cornelius F, Kilpatrick D, Accessed from: www.who.int/classifications/icf/en/
Blazey K. Outcome of clinical risk management in 25 Tennant A. Goal attainment scaling: current
the Gloucester rehabilitation service. Psychiatr Bull methodological challenges. Disabil Rehabil 2007;
2002; 26: 449–52. 29: 1583–88.
In the paper by L. Turner-Stokes, there were errors in the formulas presented on pages 364 and 367. The
correct formulas are published below. The publisher apologises for this error and the inconvenience
caused to the author, editors, and readers.
Where wi ¼ the weight assigned to the ith goal (if equal weights, wi ¼ 1), xi ¼ the numerical value
achieved (between 2 and þ2), r (rho) ¼ the expected correlation of the goal scales.
p P P
Starting with: ð0:7 w2i þ 0:3ð wi Þ2 Þ