ORIGINAL ARTICLE

Gait in Children With Cerebral Palsy

Observer Reliability of Physician Rating Scale and Edinburgh
Visual Gait Analysis Interval Testing Scale
Karel G. B. Maathuis, MD, PhD,*† Cees P. van der Schans, PhD,‡ Andries van Iperen, MD,*
Hans S. Rietman, MD,*† and Jan H. B. Geertzen, MD, PhD*†

cluding computerized kinematics and kinetics, electromyog-

Abstract: The aim of this study was to test the inter- and intra- raphy, and videotaping, is increasingly used in the evaluation
observer reliability of the Physician Rating Scale (PRS) and the of gait pattern of CP patients and is considered the gold
Edinburgh Visual Gait Analysis Interval Testing (GAIT) scale for use standard for gait assessment.7–9 However, because this
in children with cerebral palsy (CP). Both assessment scales are assessment is complex, expensive, and time-consuming and
quantitative observational scales, evaluating gait. The study involved is not generally available, it is impractical for routine use. In
24 patients ages 3 to 10 years (mean age 6.7 years) with an abnormal the past decade simplified methods have been developed to
gait caused by CP. They were all able to walk independently with or quantify walking in children with a spastic gait by using
without walking aids. Of the children 15 had spastic diplegia and 9 a standardized observation scoring system with videotaping
had spastic hemiplegia. With a minimum time interval of 6 weeks, only,10–14 but existing measures are either not easily accessed
video recordings of the gait of these 24 patients were scored twice by or untested.
three independent observers using the PRS and the GAIT scale. The One of these instruments is the Physician Rating Scale
study showed that both the GAIT scale and the PRS had excellent (PRS), an observational clinical evaluation of gait originally
intraobserver reliability but poor interobserver reliability for children reported by Koman et al in 199315 and modified by others.16–18
with CP. In the total scores of the GAIT scale and the PRS, the three This simple scale records gait in the sagittal plane only. A more
observers showed systematic differences. Consequently, the authors systematic and extended gait-evaluating instrument is the
recommend that longitudinal assessments of a patient should be done Edinburgh Visual Gait Analysis Interval Testing (GAIT) scale,
by one observer only. developed by Read et al in 1998.19,20 In 2002 the GAIT scale
Key Words: cerebral palsy, video gait assessment, gait analysis, visual was refined and renamed the Edinburgh Visual Gait Score.21 It
gait assessment was developed to give a quantitative assessment of gait where
instrumented gait analysis is not available. The PRS and the
(J Pediatr Orthop 2005;25:268–272) GAIT scale were used for this study because, to our knowledge
in 2002, a good validation study for observer reliability in CP
for these instruments had not been carried out before. The aim
of this study was to test the inter- and intraobserver reliability
A bnormal gait is a common problem in children with cere-
bral palsy (CP). These children are at great risk of dete-
rioration in their walking ability as they grow up. Many
of the PRS and the GAIT scale for use in children with CP.

treatment modalities have been developed in the past decade, MATERIALS AND METHODS
depending on the age of the child and the nature and severity of The study population consisted of 24 children with CP
the restricted walking ability. Because of the importance of with a mean age of 6.7 years (range 3.3–9.9 years); 18 (75%)
planning in the timing of interventions and the difficulty in of them were boys. Of the children, 15 had spastic diplegia and
predicting the outcome of different interventions, monitoring 9 had spastic hemiplegia (right, n = 8; left, n = 1). All children
the patient, including gait analysis, before and after an had an abnormal gait caused by CP but were able to walk
intervention is essential.1–6 Instrumented gait analysis, in- independently with or without walking aids. All patients were
assessed in the University Hospital of Groningen, the
Netherlands, between 1999 and 2001. Frontal and sagittal
From the *Centre for Rehabilitation University Hospital, Groningen, The video recordings were used, taped on a split-screen video. The
Netherlands; †Northern Centre for Health Care Research, University observers were three physicians in rehabilitation medicine
Groningen, The Netherlands; and ‡University for Professional Education,
Hanzehogeschool, Groningen, The Netherlands. (A.van I., C.M., J.R.); two of them were experienced in the
Study conducted at the Department of Rehabilitation, University Hospital field (C.M., J.R.). They all scored the video recordings
Groningen, Groningen, The Netherlands. independently. Guidelines for the PRS and the GAIT scale
None of the authors received financial support for this study. were provided to the observers, and they received a short
Reprints: Karel G. B. Maathuis, MD, PhD, Department of Rehabilitation,
University Hospital Groningen, Hanzeplein 1, P. O. Box 30.001, 9700 RB,
training (1 hour) in scoring using the PRS and GAIT scale.
Groningen, the Netherlands (e-mail: c.g.b.maathuis@rev.umcg.nl). PRS variables are given in Table 1. The last subscale (change)
Copyright Ó 2005 by Lippincott Williams & Wilkins was not used for the purpose of this cross-sectional study.

268 J Pediatr Orthop  Volume 25, Number 3, May/June 2005

J Pediatr Orthop  Volume 25, Number 3, May/June 2005 Observer Reliability of Two Gait Scales

GAIT scale variables are given in Table 2. It contains 17

TABLE 1. Physician Rating Scale15
variables of observation during gait at six anatomic levels
Definition Right Left (foot, ankle, knee, hip, pelvis, and trunk), including sagittal22
Crouch and frontal23 observations. Recordings are made using a three-
Severe (.20° hip, knee, ankle) 0 0 point ordinal scale: 0 (normal), 1 (moderate deviation), and 2
Moderate (5–20° hip, knee, ankle) 1 1 (marked deviation). Both sides of the patients were scored
Mild (,5° hip, knee, ankle) 2 2 separately.
None 3 3 Observers were recommended to use slow-motion facil-
Knee ities, to stop or repeat the video if necessary, and to take their
Recurvatum .5° 0 0 time. They were instructed not to measure degrees directly
Recurvatum 0–5° 1 1 from the video screen but to give their best visual estimate. For
Neutral (no recurvatum) 2 2 all patients, either with hemiplegia or diplegia, both sides were
Foot contact scored. All video recordings were scored twice using both the
Toe 0 0 PRS and GAIT scale with a minimal time interval of 6 weeks
Toe-heel 1 1 to avoid any effects of memory; this also corresponds with
Flat 2 2 clinical practice.
Occasional heel-toe 3 3
Heel-toe 4 4
Statistical Analysis
Change
All statistical analyses were performed using SPSS 11.0.
Worse 21 21
Reliability analysis was done using analysis of variance
None 0 0
(ANOVA). As we were interested in the inter- and intra-
Better 1 1
observer reliability of the total scores and in the sources of

TABLE 2. Edinburgh Visual Gait Analysis Interval Testing Scale25

Movement Movement
Sagittal 2 1 0 1 2 Frontal 2 1 0 1 2
FOOT FOOT
1 foot clearance none reduced full n.a n.a 5 stance position .15 6–15 5–0–5 6–15 .15
hind foot in load valgus valgus neutral varus varus
2 initial contact toe flat foot heel n.a n.a 6 foot progression .15 ir 6–15 ir 5–0–5 6–15 er .15 er
angle neutral
3 heel lift none early normal delayed n.a
4 max dorsiflexion .10 10–0–9 10–20 21–30 dor .30
hind foot in stance plan plan/dor dor dor
KNEE KNEE
7 terminal swing .30 15–30 0–15 .0 n.a 10 knee part all cap ir neutral all part
flex flex flex hyperext progression cap ir cap er cap er
angle mid-stance
8 peak stance knee .30 16–30 0–15 1–10 .10
extension flex flex flex hyperext hyperext
9 peak knee flexion .80 65–80 60–64 30–59 .30
in swing flex flex flex flex flex
HIP HIP
11 peak hip .30 16–30 15–0–15 n.a n.a 13 position .15 5–15 4–0–9 10–20 .20
extension in stance flex flex flex/ext in swing add add add/abd abd abd
12 peak hip flexion .75 51–75 30–50 15–29 ,15
in swing flex flex flex flex flex
PELVIS PELVIS
14 pelvic rotation .15 6–15 5–0–5 6–15 .15 15 contra
midstance fwd fwd neutral bwd bwd lateral drop
in stance marked mod normal n.a n.a
TRUNK TRUNK
16 peak sagittal .15 6–15 5–0–5 6–15 .15 17 max lateral
position in stance fwd fwd neutral bwd bwd shift in stance marked mod neutral n.a n.a
TOTAL TOTAL
Score 2 means marked deviation, score 1 is moderate deviation, score 0 is normal range.
n.a, not available; plan, plantarflexion; dor, dorsiflexion; flex, flexion; hyperext, hyperextension; fwd, forward rotation; bwd, backward rotation; ir, internal rotation; er, external
rotation; part cap, only a part of the knee cap is visible; all cap, whole knee cap is visible; add, adduction; abd, abduction; lat, lateral; mod, moderate.

q 2005 Lippincott Williams & Wilkins 269

Maathuis et al J Pediatr Orthop  Volume 25, Number 3, May/June 2005

variance (child, observer, and repetition), for each total score

TABLE 4. Total Scores of GAIT Scale and PRS of All Three
we chose this method, not kappa statistics of the separate
Observers
items. Total GAIT scale and PRS and subscores of both scales
Observer 1 Observer 2 Observer 3
for the right and left side were taken as independent factors. (AvI) (CM) (JR)
For each independent factor the estimated variances were Mean (sd) Mean (sd) Mean (sd) P Values*
calculated. Post hoc comparison of differences between the
GAIT scale right 9.1 (5) 10.3 (5.2) 13.5 (6.7) ,0.001
three observers was done using the Friedman test. P , 0.05
GAIT scale left 7.8 (6.8) 8.5 (6.9) 11.1 (9.4) 0.003
was considered statistically significant.
PRS right 5.5 (1.9) 4.4 (2.1) 4.4 (2.2) ,0.001
PRS left 6.1 (2.4) 5.4 (2.7) 5.5 (3.0) 0.004
RESULTS *Friedman test.
Two subjects were excluded from analysis: in one patient
the frontal video imaging failed; in the other the sagittal one
failed. The 22 patients who remained for the reliability relevant because the differences in the outcome of 1.1 in the
analysis were assessed at random by the observers. On the PRS and 4.4 in the GAIT scale (Table 4) for the same person
right as well as the left side, both observer and child proved to should be clearly visible when observing gait. Besides, if the
be significant sources of variance in the GAIT scale and in the same difference in the PRS and GAIT scale could be measured
PRS. Repetition was not a significant source of variance (Table as a result of an intervention in CP patients, it would be
3); in most cases it even approached P = 1. In fact, both the considered a clinically relevant difference.
GAIT scale and the PRS showed excellent intraobserver The total scores of the GAIT scale and the PRS, between
reliability. The interobserver reliability of both assessment the three observers, also showed systematic differences. The
scales, the GAIT scale and the PRS, was considered poor. reason is not clear. Probably, angle estimation of the different
Post hoc analysis showed considerable differences anatomic levels from a video screen was done systematically
between the three observers. The mean (SD) scores for each different between the observers.
observer are given in Table 4; box plots are shown in Figure 1. Although the PRS is used often in research, we only
In Table 5, the GAIT scale is subdivided into seven found one study in the literature23 that reported interobserver
subscales: the first five subscales correspond to the different reliability; we found no study that reported intraobserver
anatomic levels, and the last two correspond to the different reliability data. Corry et al23 studied the interobserver reliability
directions of observing gait (frontal and sagittal views). Only of the PRS in a group of 20 CP children with a dynamic
with the ankle subscale on the left side did the observer appear component in spastic equinus, treated by serial casting or
not to be a significant factor in the source of variation. In all
other GAIT subscales the observer appeared to be a significant
source of variance.

DISCUSSION
The differences in mean total scores of the three
observers were considerable and were considered clinically

TABLE 3. Sources of Variation in the GAIT Scale and PRS

SS MS P Value Variance Estimates
GAIT scale right
Observer 240 120 ,0.001 2.6
Repetition 0.2 0.2 0.846 0
Child 3904 186 ,0.001 30.1
GAIT scale left
Observer 162 81 ,0.001 1.7
Repetition 0.1 0.1 0.941 0
Child 7571 361 ,0.001 59.2
PRS right
Observer 21 10.5 ,0.001 0.2
Repetition 0.4 0.4 0.395 0
Child 509 24 ,0.001 4.0
PRS left
Observer 14.6 7.3 ,0.001 0.1
Repetition 0 0 1 0 FIGURE 1. Box plots of GAIT scale and PRS. The ends of the
Child 872 42 ,0.001 6.8 rectangle reflect the interquartile range; the horizontal line in
SS, sum of squares; MS, mean square. the rectangle reflects the median value; the whiskers indicate
the minimum and maximum values.

270 q 2005 Lippincott Williams & Wilkins

J Pediatr Orthop  Volume 25, Number 3, May/June 2005 Observer Reliability of Two Gait Scales

gait. This means that only a small part of the gait is available
TABLE 5. Observer Variance of Subscores of the GAIT Scale
for evaluation of gait in the sagittal direction. We tried to
Variance approach daily practice as much as possible. With the tech-
SS MS P Value Estimates
nique used, it is possible to reproduce the study in any con-
GAIT subscore sulting room in a hospital. Second, the children did not wear
Ankle (6) standardized clothing at the time of the measurement. In some
Right 13.2 6.6 0.002 0.1 video recordings, children wore a T-shirt or sweater, which
Left 3.0 1.5 0.254 0 might have influenced the ability to estimate the amount of
Knee (4) flexion and extension in the hip, pelvis, and trunk in the GAIT
Right 17.5 8.7 ,0.001 0.2 scale. To exclude such possible failures we calculated the
Left 16.1 8.0 ,0.001 0.2 interobserver reliability of each anatomic level itself and the
Hip (3) subscores of the frontal and sagittal plane. Our hypothesis was
Right 13.5 6.7 ,0.001 0.1 that the reliability of the assessments of the foot, ankle, and
Left 9.5 4.7 ,0.001 0 knee level would have been better compared with the hip,
Pelvis (2) pelvis, and trunk level (see Table 5). It appeared that only
Right 13.9 6.9 ,0.001 0.1 the subscore of the left ankle was not a significant factor for
Left 14.7 7.3 ,0.001 0.2 the source of variation. Of the 22 left sides we tested, 8 sides
Trunk (2) were not affected sides and were scored almost equally by all
Right 11.2 5.6 ,0.001 0.1 observers, compared with only 1 unaffected right side. This
Left 8.4 4.2 ,0.001 0.1 might explain the difference in interobserver reliability be-
Frontal (6) tween the subscores at ankle level, respectively, on the right
Right 68.0 34.0 ,0.001 0.7 and left side. All other subscores of the GAIT scale, including
Left 53.8 26.9 ,0.001 0.6 the frontal and sagittal plane, showed poor interobserver reli-
Sagittal (11) ability data. Post hoc analysis of the more affected side only of
Right 109.1 54.6 ,0.001 1.2 all patients showed similar results (data not presented).
Left 43.7 21.8 ,0.001 0.4 Noonan et al6 reported a interobserver reliability study
SS, sum of squares; MS, mean square. of patients with CP evaluated with instrumented gait analysis
Values in parentheses represent the number of items on the GAIT scale. at four different centers. The results were poor.
In 2003 editorials in the Journal of Pediatric Ortho-
paedics, Gage22 and Wright25 made contrasting comments about
Botulinum toxin A. They found a moderate agreement in interobserver variability in gait analysis. They agreed that we
crouch assessment and in the overall impression in gait change need a careful assessment of the analysis of the pathology of
in the affected side using the PRS (weighted kappa test 0.55– the CP child. The importance of instrumented gait analysis is
0.67). The agreement in the knee assessment was poor. The clear for this assessment, but it is only one of the methods of
‘‘change’’ section was added to provide a more discriminating examination and should be seen as complementary to the
difference. Unfortunately, no information was given about the generation of the problem list, the physical examination, and
interobserver variation of the total PRS. Moreover, only two radiographs. Further research into the different sources that
observers were used in Corry et al’s study. contribute to the variability in gait analysis will be necessary.
Post hoc analysis of the results of our two most To evaluate the gait of a CP patient in the consulting room by
experienced observers (observers 2 and 3) showed no dif- means of video recording is very useful. To score gait with
ferences between these observers for the PRS. In other words, PRS took about 5 minutes; the GAIT scale took 25 minutes per
the information concerning the interobserver reliability for this patient on average. We also recommend that longitudinal
scale may have been influenced by the number of observers assessments of a patient should be done by the same observer.
used in the study and the degree of experience in observing
gait. This may explain the different results between the Corry REFERENCES
et al study, which included only two observers, and ours, in 1. Calderon-Gonzalez R, Calderon-Sepulveda R, Rincon-Reyes M, et al.
which three observers took part. Botulinum toxin A in management of cerebral palsy. Pediatr Neurol.
1994;4:284–288.
In 1999 Read et al20 presented data about the inter- and 2. Cook RE, Schneider I, Hazlewood ME, et al. Gait analysis alters decision-
intraobserver reliability of the GAIT scale in a study making in cerebral palsy. J Pediatr Orthop. 2003;23:292–295.
population of four patients with CP and one ‘‘normal’’ person. 3. Gage JR. The role of gait analysis in the treatment of cerebral palsy
Because of the very small group, these data are of limited [editorial]. J Pediatr Orthop. 1994;14:701–702.
value, however. In 2003 Read et al24 reported a good intra- and 4. Hailey D, Tomie J. An assessment of gait analysis in the rehabilitation of
children with walking difficulties. Disabil Rehabil. 2000;6:275–280.
interobserver reliability for the Edinburgh Visual Gait Score 5. Molenaers G, Desloovere K, Eyssen M, et al. Botulinum toxin A treatment
system. They used kappa statistics for their calculations; we, of cerebral palsy: an integrated approach. Eur J Neurol. 1999;6:S1–S7.
however, were more interested in the source of variance. In our 6. Noonan KJ, Halliday S, Brown R, et al. Interobserver variability of gait
population the child and the observer proved to be a significant analysis on patients with cerebral palsy. J Pediatr Orthop. 2003;23:279–
287.
factor in this respect for the source of variation. 7. Drouin LM, Malouin F, Richards CL, et al. Correlation between the gross
We are aware of some limitations in our study. First, we motor function measure and gait spatiotemporal measures in children with
used a fixed video camera framing for recording the sagittal neurological impairments. Dev Med Child Neurol. 1996;38:1007–1019.

q 2005 Lippincott Williams & Wilkins 271

Maathuis et al J Pediatr Orthop  Volume 25, Number 3, May/June 2005

8. Kirkpatrick M, Wytch R, Cole G, et al. Is the objective assessment of 16. Chutorion AM, Root L. Management of spasticity in children with
cerebral palsy gait reproducible? J Pediatr Orthop. 1994;14:705–708. botulinum-A toxin. Int Pediatr. 1994;9:35–42.
9. Sutherland DH, Kaufman KR, Wyatt MP, et al. Injections of botulinum 17. Denislic M, Meh D. Botulinum toxin in the treatment of cerebral palsy.
A toxin into the gastrocnemius muscle of patients with cerebral palsy: a Neuropediatrics. 1995;26:249–252.
3-dimensional motion analysis study. Gait Posture. 1996;4:269–279. 18. Graham HK, Aoki KR, Autti-Rämö, et al. Recommendations for the use
10. Boyd NR, Graham HK. Objective measurement of clinical findings in the of botulinum toxin type A in the management of cerebral palsy. Gait
use of botulinum toxin type A for the management of children with Posture. 2000;11:67–79.
cerebral palsy. Eur J Neurol. 1999;6:523–535. 19. Kerr AM, Hazlewood ME, Linden van der Ml, et al. The Edinburgh Visual
11. Boyd NR, Graham JEA, Nattrass GR, et al. Medium term outcome- Gait Score as an outcome measure after surgical intervention cerebral
response characterisation and risk factor analysis of Botulinum toxin type palsy. Gait Posture. 2002;16:S116.
A in the management of spasticity in children with cerebral palsy. Eur J 20. Read HS, Hillmann SJ, Hazlewood ME, et al. The Edinburgh Visual Gait
Neurol. 1999;6:S37–S45. Analysis Interval Testing (GAIT) scale. Gait Posture. 1999;10:63.
12. Eastlack ME, Arvidson J, Snyder-Mackler L, et al. Interrater reliability 21. Read HS, Hazlewood ME, Hillmann SJ, et al. A visual gait analysis score
of videotaped observational gait-analysis assessments. Phys Ther. 1991;71: for use in cerebral palsy: the Edinburgh Visual Gait Score. Gait Posture.
465–472. 2002;16:S115–S116.
13. Flett PJ, Stern LM, Waddy H, et al. Botulinum toxin A versus fixed cast 22. Gage JR. Con: interobserver variability of gait analysis [editorial].
stretching for dynamic calf tightness in cerebral palsy. J Paediatr Child J Pediatr Orthop. 2003;23:290–291.
Health. 1999;35:71–77. 23. Corry IS, Graham HK. Botulinum toxin A compared with stretching casts
14. Ubhi T, Bhakta BB, Ives HL, et al. Randomised double-blind placebo- in the treatment of spastic equines: a randomised prospective trial.
controlled trial of the effect of botulinum toxin on walking in cerebral J Pediatr Orthop. 1998;18:304–311.
palsy. Arch Dis Child. 2000;83:481–487. 24. Read HS, Hazlewood ME, Hillman SJ, et al. Edinburgh Visual Gait Score.
15. Koman LA, Mooney JF, Smith B, et al. Management of cerebral palsy with J Pediatr Orthop. 2003;23:296–301.
botulinum-A toxin: preliminary investigation. J Pediatr Orthop. 1993;4: 25. Wright JG. Pro: interobserver variability of gait analysis [editorial].
489–495. J Pediatr Orthop. 2003;23:288–289.

272 q 2005 Lippincott Williams & Wilkins