Intra-rater and inter-rater reliability of gait ... - Gait & Posture

Gait and Posture 17 (2003) 59 /67 www.elsevier.com/locate/gaitpost

Intra-rater and inter-rater reliability of gait measurements with CODA mpx30 motion analysis system V. Maynard a, A.M.O. Bakheit b,*, J. Oldham b, J. Freeman a a

b

Plymouth Institute of Health Studies, Plymouth, UK Plymouth Motion Analysis Laboratory, Beauchamp Centre, Mount Gould Hospital, University of Plymouth, PlymouthPL4 7QD, UK Received 29 November 2001; received in revised form 24 March 2002; accepted 8 April 2002

Abstract Variations between measurements of clinical outcomes sometimes result from observer errors and are not due to the therapeutic intervention. For the correct interpretation of clinical data it is, therefore, important to ascertain the reliability of the measurement methods used. In the present study we evaluated the test /retest and between observer reliability of gait measurements with the Cartesian Optoelectronic Dynamic Anthropometer (CODA mpx30) motion analysis system. Our findings suggest a better intrarater and inter-rater reliability of kinetic than kinematic data. Disagreements between the investigators were most marked for the kinematic data, especially the hip angle, and disagreements were least for the spatio-temporal gait parameters. The poor reproducibility of kinematic variables may be due to problems with the accurate placement of markers on the surface anatomical landmarks. # 2002 Elsevier Science B.V. All rights reserved. Keywords: Gait; Motion analysis; Reliability

1. Introduction Computerised three-dimensional (3-D) gait analysis is being increasingly used for the diagnostic work up of patients with locomotor disability and also in the assessment of the effectiveness of therapeutic interventions. However, this investigation is only likely to receive wider acceptance in clinical practice if its reliability can be demonstrated. It is essential that those measurements by the same user at different times yield similar results and there is good agreement between measurements made by different investigators. Several studies have examined the reliability of motion analysis systems in the evaluation of human gait. However, the reported findings have generally lacked consistency. For example, Winter [1] and Kadaba * Corresponding author. Tel.: /44-1752-272-481; fax: /44-1752272-483 E-mail address: [email protected] (A.M.O. Bakheit).

et al. [2] have found significant variability in the measurements of hip and knee kinetics. Similar observations were also reported recently by Cowman et al. [3] who attributed the high variability in their measurements entirely to observer error. However, these reports were contradicted by other investigators who found better repeatability of the kinetic than the kinematic gait parameters.[4] Interestingly, other studies have shown excellent repeatability of both kinetic and kinematic gait analysis measurements [5]. There have been no major studies that examined the reliability of measurements with the Cartesian Optoelectronic Dynamic Anthropometer (CODA mpx30) motion analysis system. Similarly, only a few researchers have examined the reliability of measurements of the spatio-temporal, kinetic and kinematic gait parameters in the same group of individuals. This is an important shortcoming of previous research as information obtained from all of these parameters is usually necessary for the correct interpretation of gait analysis data. The aim of the present study was to assess the intra-rater and inter-rater reliability of measurements of a wide range of clinically relevant gait parameters.

0966-6362/02/$ - see front matter # 2002 Elsevier Science B.V. All rights reserved. PII: S 0 9 6 6 - 6 3 6 2 ( 0 2 ) 0 0 0 5 1 - 6

60

V. Maynard et al. / Gait and Posture 17 (2003) 59 /67

2. Subjects and methods 2.1. Subjects Healthy volunteers who had no past history of neurological or musculoskeletal disease gave an informed written consent to take part in the study. Two groups of subjects were recruited,one for the intra-rater and the other for the inter-rater, reliability studies. The project was approved by the local Ethics Committee. 2.2. Equipment and data acquisition Selected kinematic, kinetic and spatio-temporal gait parameters (see below) were recorded during level walking. Data acquisition was made using a dual CODA mpx30 (Charnwood Dynamics, Barrow on Soar, Leicestershire, England) motion analysis system. The system was fully integrated with two AMTI Biomechanics Force Platforms model BP2416-1000 (Advanced Mechanical Technology Inc, 176 Waltham Street, Watertown, MA 02172-4800, USA). CODA mpx30 is a 3-D pre-calibrated system consisting of three optical sensors mounted on a rigid frame within a scanner unit. The scanner captures infra red light signals pulsed sequentially by markers placed on anatomical landmarks, a pelvic frame and thigh and shin wands. The middle sensor tracks vertical marker movements, whilst the outer sensors track the horizontal movements of the markers. Kinetic parameters were computed from data measured in three planes. The computation was based on the method of link-segmental analysis and inverse dynamics [6]. This mathematical model predicts the joint’s centre from data recorded in two co-ordinates and anthropometric measurements, namely the patient’s body mass, height, width of the ankle and knee joints and the width and depth of the pelvis (a detailed description of this model can be found in Part 3, Section 4 of the CODA mpx30 User Guide, June 2000). The patient’s height was measured using a metal ruler mounted on a wall scale and the body weight was measured with digital scales. The width of the ankle and knee joints was measured with dial calipers. A metal ruler was used to measure pelvic width from one anterior superior iliac spine (ASIS) to the other, and pelvic depth from the ASIS to the posterior superior iliac spine. An alignment frame was used to define the axis of rotation of the ankle joint. An anatomical co-ordinate was defined with a set of infrared light emitting diode markers. Three markers were placed on each side of the pelvic frame in order to track pelvic movement. Further markers were positioned over the knee, the lateral malleolus, and the end of the heel and the fifth metatarsal bone. The knee marker was placed on the line of the knee joint, halfway

between the posterior margin of the head of the fibula and the anterior margin of the tibial tuberosity. A second co-ordinate was used to establish the orientation of the femur and tibia. Thigh and shank wands were fastened with Velcro straps above and below the knee as the subject sat whilst holding their feet apart. The thigh wand was adjusted into a position parallel to a T-bar placed between the knees. Alignment of the shank wand was made as the subject sat. The wand was adjusted perpendicular to the axis of the ankle joint, i.e. perpendicular to the line joining the medial and lateral malleolus. Markers were placed at each of the anterior and posterior ends of the thigh and shank wands. The following spatiotemporal gait parameters were studied: walking velocity, the duration of the stance and swing phases of the gait cycle. The kinematic variables studied were the hip, knee and ankle angles on initial contact, mid stance and mid swing. These kinematic parameters were chosen because of their importance for safe and energy efficient ambulation.[7] Hip, knee and ankle moments and powers were measured on initial contact and in mid stance. The phases of the gait cycle were determined from inspection of the animated ‘stick diagram’ and confirmed by the position of the ground reaction force vector relative to the foot. Although mid stance and mid swing are not discrete points in time, they were defined in this study (for simplicity) as 50% of the respective phase of the gait cycle. Only the data in the sagittal plane were analysed. The kinetic and kinematic data were acquired and digitised with a sampling frequency of 200 Hz. Data synchronisation and normalisation of moments and powers to body mass were made automatically by the CODA MPX30 software. The gait laboratory was approximately 12 m long and 6 m wide and the force plates were embedded in the middle of the walkway. Two CODA mpx30 motion analysis systems were used to record the gait cycle from both lower limbs simultaneously. They were placed 1.5 m lateral to the long axis of the force plates. Subjects were asked to walk the length of the laboratory barefoot and at their usual speed. One ‘clean’ gait cycle for each patient/test was identified and saved for future analysis. 2.3. Assessments Gait analysis was carried out twice on the same day (in the morning and the afternoon) and was repeated (only once) a week later. The investigators were blind to previous measurements. 2.4. Statistical analysis Intra-rater and inter-rater reliability was assessed with the interclass correlation (ICC) method and the Bland


61

Table 1 Bland and Altman limits of agreement and intraclass correlation coefficient (ICC) for intra-rater reliability of the kinematic data (joint moments) Initial contact

d

S.E. of d

95% CI for d

S.D. diff

95% limits of agreement

ICC

95% CI

Ankle angle Am /pm pm /wk am /wk

0.88 0.22 1.1

0.1 1.02 0.96

/1.370/3.13 /2.080/2.51 /1.090/3.28

3.15 3.21 3.05

/5.420/7.18 /6.180/6.62 /5.00/7.2

0.28 /0.07 /0.10

/0.360/0.75 /0.620/0.55 /0.640/0.53

Knee angle am /pm pm /wk am /wk

0.64 0.78 1.42

1.53 1.11 1.80

/2.820/4.09 /1.730/3.29 /2.660/5.50

4.83 3.51 5.70

/9.020/10.30 /6.240/7.8 /9.980/12.82

/0.49 /0.07 /0.41

/0.830/0.15 /0.620/0.55 /.800/0.25

Hip angle am /pm pm /wk am /wk

2.85 1.39 4.23

2.55 1.61 3.30

/2.930/8.62 /2.260/5.03 /3.240/11.71

8.07 5.10 10.45

/13.590/18.99 /10.200/11.59 /16.670/25.13

0.46 0.71 0.09

/0.160/0.83 0.220/0.92 /0.520/0.65

Ankle angle am /pm pm /wk am /wk

0.164 0.602 0.766

1.097 1.115 0.821

/2.3190/2.646 /1.920/3.123 /1.0910/2.622

3.47 3.53 2.59

/6.780/7.1 /6.440/7.66 /4.410/5.95

0.45 0.35 0.58

/0.170/0.83 /0.290/0.79 /0.000/0.87


0.007 1.208 1.215

0.725 1.286 1.342

/1.6330/1.647 /1.7020/4.117 /1.8200/4.249

2.29 4.07 4.24

/4.600/4.61 /6.940/9.34 /7.260/9.70

0.87 0.64 0.59

0.590/0.97 0.090/0.90 0.010/0.88

/0.784 4.147 3.364

1.923 1.67 2.36

/5.1350/3.568 0.3690/7.925 /1.9860/8.713

6.083 5.28 7.48

/12.940/11.38 /6.410/14.71 /11.60/18.32

0.62 0.50 0.21

0.060/0.89 /0.120/0.84 /0.420/0.72

1.58 0.024 1.60

0.649 0.836 0.670

0.1120/3.048 /1.8670/1.91 0.0890/3.120

2.052 2.642 2.118

/2.50/5.70 /5.260/5.30 /2.640/5.84

0.56 0.51 0.38

/0.030/0.87 /0.100/0.85 /0.260/0.80


/2.228 1.530 /0.698

2.074 1.136 1.786

/6.9190/2.463 /1.0390/4.10 /4.7380/3.342

6.557 3.59 5.648

/15.350/10.89 /5.670/8.73 /12.00/10.6

0.32 0.51 0.51

/0.320/0.77 /0.100/0.85 /0.100/0.85

Hip angle am /pm pm /wk am /wk

/2.68 4.40 1.72

1.49 1.56 1.61

/6.0560/0.695 0.870/7.94 /1.920/5.37

4.72 4.94 5.10

/12.080/6.72 /5.480/14.28 /8.480/11.92

0.71 0.58 0.68

0.220/0.92 /0.010/0.87 0.160/0.91

Mid stance

Hip angle am /pm pm /wk am /wk Mid swing Ankle angle am /pm pm /wk am /wk

d , the mean difference; S.E. of d , standard error of the mean difference; 95% CI for d , the 95% confidence interval for the mean difference; SDdiff, the standard deviation for the differences. The 95% limits of agreement are calculated as d9/(S.D. diff multiplied by 2).

and Altman test [8]. This approach is considered the most appropriate statistical method for studying agreement between sets of interval data [9]. An ICC coefficient of /0.75 was accepted as evidence of good agreement [10]. The data were analysed on a personal computer using version 9 of SPSS for WINDOWS software.

3. Results 3.1. Intra-rater reliability Ten subjects (five males and five females) were recruited and all completed the intra-rater reliability study. The mean age of the group was 39.2 years.


62

Table 2 Bland and Altman limits of agreement and intraclass ICC for intra-rater reliability of the kinetic data (joint moments) Initial contact

d

S.E. of d

95% CI for d

S.D. diff


ICC

95% CI

Ankle moments am /pm pm /wk am /wk

/0.015 0.0026 /0.013

0.0082 0.0057 0.0089

/0.0340/3.39 /0.0100/0.016 /0.0330/0.0075

0.026 0.018 0.028

/0.0670/0.037 /0.0330/0.039 /0.0430/0.043

/0.03 0.37 0.22

/0.600/0.58 /0.270/0.79 /0.420/0.72

Knee moments am /pm pm /wk am /wk

/0.028 0.013 /0.015

0.020 0.018 0.033

/0.0730/0.016 /0.0270/0.053 /0.90 0/0.060

0.062 0.056 0.10

/0.1520/0.096 /0.0990/0.125 /0.2150/0.185

0.50 0.71 0.25

/0.110/0.85 0.210/0.92 /0.390/0.74

Hip moments am /pm pm /wk am /wk

0.166 /0.078 0.088

0.096 0.087 0.131

/0.05 0/0.382 /0.27 0/0.12 /0.21 0/0.38

0.30 0.27 0.41

/0.430/0.77 /0.620/0.46 /0.730/0.91

0.31 0.31 0.30

/0.330/0.77 /0.340/0.76 /0.340/0.76


/0.025 0.040 0.015

0.056 0.049 0.042

/0.1500/0.101 /0.0710/0.150 /0.0800/0.110

0.176 0.154 0.133

/0.390/0.34 /0.260/0.34 /0.250/0.28

0.41 0.57 0.64

/0.230/0.81 /0.010/0.87 0.090/0.89


0.016 /0.007 0.010

0.030 0.027 0.018

/0.0520/0.084 /0.0670/0.054 /0.0310/0.050

0.095 0.084 0.056

/0.170/0.21 /0.150/0.17 /0.10/0.12

0.84 0.86 0.94

0.500/0.96 0.570/0.96 0.770/0.98


/0.051 /0.002 /0.053

0.051 0.031 0.036

/0.1660/0.065 /0.0720/0.068 /0.1340/0.029

0.161 0.098 0.114

/0.370/0.27 /0.200/0.19 /0.080/0.17

0.19 0.75 0.36

/0.440/0.70 0.290/0.93 /0.290/0.79


/0.0003 /0.001 /0.0015

0.0008 0.0009 0.0007

/0.0020/0.0014 /0.0030/0.0008 /0.0030/0.00009

0.0024 0.0028 0.0023

/0.00430/0.0037 /0.00680/0.0044 /0.0060/0.002

0.17 0.11 0.29

/0.460/0.69 /0.540/0.66 /0.350/0.76


0.0014 /0.0058 /0.0044

0.009 0.0085 0.0037

/0.0180/0.021 /0.0250/0.013 /0.0130/0.0039

0.028 0.027 0.012

/0.060/0.06 /0.060/0.048 /0.0280/0.02

0.49 0.55 0.87

/0.120/0.84 /0.040/0.87 0.590/0.97


0.013 0.017 0.030

0.0098 0.011 0.014

/0.0090/0.035 /0.0080/0.041 /0.00190/0.061

0.03 0.034 0.04

/0.0470/0.073 /0.050/0.09 /0.050/0.11

0.58 0.47 0.24

/0.010/0.87 /0.150/0.83 /0.390/0.73

Mid stance

Mid swing

Results of the kinematic, kinetic and spatio-temporal gait parameters are given in tables Tables 1/4. Generally there was poor correlation (ICC coefficientB/ 0.75) between the measurements made in the morning and afternoon of the same day and between them and the measurements taken a week later. This was also true for all gait parameters except the knee angles and knee moments in mid stance. There was also a trend for agreement between the measurements of the duration of the swing phase of the gait cycle. The ICC coefficients were corroborated by results from the corresponding Bland and Altman test that showed good limits of agreement.

3.2. Inter-rater reliability For the inter-rater reliability part of the study, the gait of another 19 subjects was measured once by each of three examiners. There were four males and 15 females. The mean age (range) of the group was 34.4 (20 /49) years. As shown in Tables 5/8, there was good agreement (demonstrated by both statistical tests) between the three investigators for some of the gait parameters. These were knee angle in mid stance, ankle and knee angles in mid swing, knee and ankle moments on initial contact and mid stance, respectively, and the spatio-temporal parameters. Interestingly, the disagree-


63

Table 3 Bland and Altman limits of agreement and intraclass ICC for intra-rater reliability of the kinetic data (joint powers) Initial contact

d

S.E. of d

95% CI for d

S.D. diff


ICC

95% CI

Ankle power am /pm pm /wk am /wk

/0.036 0.0072 /0.029

0.019 0.012 0.020

/0.0780/5.58 /0.0210/0.035 /0.0740/0.015

0.059 0.039 0.062

/0.1540/0.082 /0.0710/0.085 /0.1530/0.095

/0.03 0.33 0.28

/0.600/0.58 /0.310/0.77 /0.360/0.75

Knee power am /pm pm /wk am /wk

0.139 /0.02 /0.118

0.094 0.068 0.145

/0.0730/0.350 /0.1740/0.134 /0.2100/0.447

0.296 0.215 0.460

/0.460/0.74 /0.440/0.44 /0.860/0.74

0.23 0.64 /0.13

/0.410/0.73 0.070/0.89 /0.660/0.50

Hip power am /pm pm /wk am /wk

/0.029 /0.027 /0.0025

0.035 0.028 0.025

/0.1090/0.050 /0.0380/0.091 /0.0590/0.054

0.111 0.090 0.079

/0.190/0.25 /0.150/0.21 /0.160/0.16

0.11 0.37 0.68

/0.510/0.66 /0.270/0.79 0.160/0.91

0.101 0.084 0.185

0.094 0.091 0.096

/0.1120/0.314 /0.1220/0.291 /0.0310/0.402

0.298 0.289 0.302

/0.500/0.70 /0.500/0.66 /0.420/0.79

0.40 0.46 0.53

/0.240/0.80 /0.170/0.83 /0.080/0.86


0.057 /0.013 0.044

0.049 0.030 0.035

/0.0540/0.167 /0.0820/0.056 /0.0340/0.122

0.155 0.096 0.109

/0.240/0.36 /0.210/0.18 /0.180/0.26

0.27 0.44 0.62

/0.370/0.75 /0.190/0.82 0.560/0.89


/0.098 /0.022 /0.120

0.100 0.065 0.060

/0.3230/0.127 /0.1690/0.125 /0.2570/0.016

0.316 0.205 0.191

/0.740/0.54 /0.400/0.40 /0.50/0.26

0.02 0.65 0.37

/0.570/0.61 0.110/0.90 /0.270/0.79

Ankle power am /pm pm /wk am /wk

/0.003 0.001 /0.002

0.002 0.002 0.003

/0.0070/0.0014 /0.0030/0.006 /0.0070/0.004

0.006 0.006 0.008

/0.0150/0.009 /0.0110/0.013 /0.020/0.01

0.01 /0.06 /0.31

/0.570/0.61 /0.620/0.55 /0.760/0.35


0.018 /0.033 /0.015

0.042 0.041 0.017

/0.0760/0.112 /0.13 0/0.06 /0.0530/0.024

0.132 0.13 0.054

/0.240/0.28 /0.290/0.23 /0.120/0.09

0.48 0.54 0.88

/0.140/0.84 /0.060/0.86 0.600/0.97


/0.014 /0.022 /0.036

0.014 0.016 0.020

/0.0460/0.018 /0.0590/0.015 /0.0810/0.008

0.045 0.052 0.062

/0.10/0.08 /0.120/0.08 /0.0840/0.084

0.58 0.47 0.35

0.000/0.88 /0.160/0.83 /0.290/0.78

Mid stance Ankle power am /pm Pm /wk am /wk

Mid swing

Table 4 Bland and Altman limits of agreement and intraclass ICC for intra-rater reliability of the spatio-temporal gait parameters Velocity

d

S.E. of d

95% CI for d

S.D. diff


ICC

95% CI

am /pm pm /wk am /wk

/0.023 0.007 /0.016

0.078 0.040 0.077

/0.200/0.154 /0.0820/0.096 /0.190/0.16

0.25 0.12 0.24

/0.52 0/0.48 /0.23 0/0.25 /0.50 0/0.46

0.15 0.81 /0.14

/0.470/0.68 0.430/0.95 /0.660/0.50

Duration of stance am /pm /0.041 pm /wk 0.004 am /wk /0.036

0.021 0.011 0.015

/0.0870/0.006 /0.0210/0.030 /0.070/0.0021

0.065 0.035 0.048

/0.17 0/0.09 /0.07 0/0.07 /0.14 0/0.06

0.30 0.84 0.34

/0.340/0.76 0.500/0.96 /0.300/0.78

Duration of swing am /pm 0.0067 /0.0022 pm /wk am /wk 0.0046

0.010 0.005 0.010

/0.0170/0.027 /0.0130/0.009 /0.0170/0.027

0.033 0.015 0.031

/0.06 0/0.07 /0.03 0/0.03 /0.06 0/0.07

0.62 0.86 0.59

0.060/0.89 0.570/0.96 0.020/0.88


64

Table 5 Bland and Altman limits of agreement and intraclass ICC for inter-rater reliability of the kinematic data Initial contact

d

S.E. of d

95% CI for d

S.D. diff

Ankle angle 1 /2 1 /3 2 /3

/0.36 /1.68 /1.32

0.71 0.63 0.48

/1.87 0/1.14 /3.00 0//0.36 /2.33 0/0.31

3.12 2.74 2.09

Knee angle 1 /2 1 /3 2 /3

/1.50 /2.15 /0.66

0.80 0.83 0.77

/3.18 0/0.19 /3.89 0//0.42 /2.28 0/0.96

ICC

95% CI

/6.600/5.87 /7.170/3.80 /5.500/2.86

0.26 0.44 0.73

/0.21 0/0.63 0.00 0/0.74 0.43 0/0.89

3.50 3.60 3.37

/8.490/5.50 /9.350/5.04 /7.390/6.08

0.59 0.58 0.54

0.20 0/0.82 0.19 0/0.82 0.12 0/0.79

2.79 3.30 0.50

1.50 1.59 1.07

/0.36 0/5.95 /0.05 0/6.65 /1.75 0/2.75

6.56 6.95 4.67

/10.320/15.91 /10.610/17.20 /8.840/9.85

0.34 0.15 0.47

/0.13 0/0.68 /0.31 0/0.56 0.03 0/0.75

Ankle angle 1 /2 1 /3 2 /3

/0.32 /2.03 /1.71

0.52 0.50 0.48

/1.41 0/0.77 /3.08/0.97 /2.71 0//0.71

2.26 2.19 2.08

/4.840/4.20 /6.410/2.35 /5.870/2.45

0.53 0.47 0.57

0.11 0/0.79 0.03 0/0.75 0.17 0/0.81

Knee angle 1 /2 1 /3 2 /3

/0.94 /1.62 /0.68

0.73 0.74 0.78

/2.48 0/0.60 /3.16 0//0.07 /2.31 0/0.96

3.20 3.21 3.39

/7.340/5.46 /8.040/4.81 /7.460/6.11

0.70 0.68 0.66

0.37 0/0.87 0.34 0/0.86 0.31 0/0.86

2.69 3.20 0.51

1.20 1.19 1.05

/0.17 0/5.22 0.71 0/0.69 /1.70 0/2.71

5.24 5.17 4.58

/7.790/13.18 /7.130/13.53 /8.650/9.66

0.42 0.38 0.50

/0.03 0/0.73 /0.07 0/0.71 0.08 0/0.77

Ankle angle 1 /2 1 /3 2 /3

/0.06 /0.95 /0.89

0.71 0.61 0.61

/1.56 0/1.44 /2.22 0/0.32 /2.18 0/0.39

3.12 2.64 2.66

/6.290/6.18 /6.230/4.33 /6.220/4.44

0.74 0.73 0.76

0.44 0/0.89 0.43 0/0.89 0.48 0/0.90

Knee angle 1 /2 1 /3 2 /3

/1.06 /1.36 /0.30

1.02 0.94 1.06

/3.21 0/1.08 /3.33 0/0.60 /2.53 0/1.93

4.45 4.08 4.63

/9.960/7.84 /9.530/6.80 /9.560/8.97

0.67 0.72 0.67

0.32 0/0.86 0.40 0/0.88 0.33 0/0.86

Hip angle 1 /2 1 /3 2 /3

2.82 3.14 /0.32

1.28 1.30 1.22

0.13 0/5.51 0.42 0/5.86 /2.87 0/2.24

5.59 5.65 5.30

/8.350/13.99 /8.160/14.44 /10.920/10.28

0.27 0.34 0.23

/0.20 0/0.64 /0.12 0/0.68 /0.24 0/0.61

Hip angle 1 /2 1 /3 2 /3


Mid stance

Hip angle 1 /2 1 /3 2 /3 Mid swing

ments between the investigators were most marked for the kinematic data, especially the hip angle.

4. Discussion Variations between measurements of clinical outcomes are not always due to real changes in the patient’s condition or as a result of a therapeutic intervention. Observer variability may introduce a significant error in measurements and this possibility should always be considered, especially when the assessment tool or the equipment used is new to the clinician. In the present study we sought to establish the extent of the test-retest variability when the same clinician made repeated

measurements with CODA mpx30 motion analysis system. We also examined the level of agreement between different clinicians when they measured the same variables in the same group of individuals using a standard protocol. We assumed that the use of a standard protocol for marker placement and data collection was likely to minimise error. We also examined gait in the sagittal plane because of the low variability in gait pattern in this plane [2]. Our findings suggest a better inter-rater than intra-rater reliability for most of the gait parameters measured with CODA mpx30. Test-retest repeatability of measurements of the joint kinematics was best for the knee angles and poorest for the hip angles. This is consistent with previous observa-


65

Table 6 Bland and Altman limits of agreement and intraclass ICC for inter-rater reliability of the kinetic data (joint moments) Initial contact

d

S.E. of d

95% CI for d

S.D. diff


ICC

95% CI

Ankle moments 1 /2 1 /3 2 /3

/0.002 0.003 /0.005

0.002 0.001 0.002

/0.0010/0.006 /0.0050/0.000 /0.0090/0.001

0.008 0.005 0.008

/0.0140/0.019 /0.0130/0.008 /0.0220/0.011

0.359 0.577 0.310

/0.1010/0.692 0.1780/0.812 /0.1550/0.663

Knee moments 1 /2 1 /3 2 /3

/0.002 0.004 0.006

0.012 0.010 0.008

/0.0270/0.023 /0.0180/0.025 /0.0110/0.022

0.052 0.044 0.034

/0.1050/0.101 /0.0850/0.092 /0.0620/0.073

0.604 0.714 0.774

0.2190/0.826 0.3950/0.879 0.5030/0.906

0.008 0.031 0.023

0.027 0.025 0.027

/0.0490/0.066 /0.0210/0.083 /0.0330/0.079

0.119 0.109 0.117

/0.2290/0.246 /0.1860/0.248 /0.2110/0.257

0.605 0.721 0.631

0.2200/0.827 0.4080/0.883 0.2600/0.840

Ankle moments 1 /2 1 /3 2 /3

/0.016 /0.045 /0.029

0.026 0.023 0.027

/0.0710/0.038 /0.0940/0.004 /0.0850/0.027

0.114 0.102 0.117

/0.2430/0.211 /0.2490/0.159 /0.2620/0.204

0.707 0.757 0.748

0.3830/0.876 0.4720/0.899 0.4560/0.895

Knee moments 1 /2 1 /3 2 /3

/0.008 0.001 0.008

0.015 0.018 0.017

/0.0400/0.025 /0.0380/0.039 /0.0280/0.045

0.067 0.080 0.076

/0.1420/0.127 /0.1590/0.160 /0.1430/0.160

0.719 0.515 0.594

0.4040/0.882 0.0920/0.780 0.2040/0.821

Hip moments 1 /2 1 /3 2 /3

0.003 /0.046 /0.050

0.024 0.022 0.019

/0.0460/0.053 /0.0930/0.000 /0.0900//0.009

0.104 0.096 0.0.84

/0.2040/0.210 /0.2390/0.147 /0.2170/0.118

0.597 0.548 0.655

0.2090/0.823 0.1380/0.798 0.2980/0.851

Hip moments 1 /2 1 /3 2 /3 Mid stance

tions [3] and could be due to the easier identification of the anatomical landmarks for placement of the markers on the knee. We found the repeatability of the kinetic measurements to be better than that of the kinematic data. However, the levels of agreement were generally low. Calculating the total limb support moment [11] may reduce the variability of repeated measurements of kinetic parameters further. This approach may be more useful clinically than the study of kinetic variables at individual joints. It has been argued that the information derived from ICC coefficients alone has limited utility in clinical practice, as it does not define the magnitude of disagreement between measurements. Previous researchers [9,12] have shown that the interpretation of reliability data is more meaningful when ICC analysis is complemented with other tests, e.g. the Bland and Altman method, as was performed in the present study. The calculations of ICC may be based on any one of several equations depending on the design of the study [13]. The choice of the appropriate equation for the calculation of ICC is important because analysis of the same data using different ICC equations may give opposite results [13]. Equation (2,1) takes into account the variance due to the bias introduced by the raters and a high ICC coefficient using this formula would, therefore, suggest generalisability. We chose equation (2,1)

for the calculation of the ICC because the results are potentially generalisable to many users of the CODA mpx30 motion analysis system. This procedure requires the use of several raters that are chosen randomly from a large sample of clinicians. A limitation of our study is that only three selected investigators collected the data. The present study has not demonstrated complete reproducibility of the gait analysis data when measurements were made with the CODA mpx30. This may be due to causes other than observer error. Large differences between repeated measurements of gait in the same subject may occur in the absence of error by the examiner and when measurement tools of statistically proven reliability are used [14]. This may be merely due to variations of the individual gait cycles, for example depending on walking speed. It has been suggested that a minimum of three gait cycles should be averaged to overcome the effects of cycle-to-cycle variability [1]. A recent study [15] has shown that reliability of kinetic gait parameters was excellent when a mean of 10 trials was used. By contrast, a similar degree of reliability of kinematic data was obtained when five acceptable gait cycles were averaged. It is, therefore, imperative that gait analysis data collected from one representative cycle are interpreted with caution and in conjunction with other relevant clinical information.


66

Table 7 Bland and Altman limits of agreement and intraclass ICC for inter-rater reliability of the kinetic data (joint powers) Initial contact

d

S.E. of d

95% CI for d

S.D. diff


ICC

95% CI

Ankle power 1 /2 1 /3 2 /3

0.000 /0.006 /0.006

0.004 0.003 0.004

/0.0080/0.008 /0.0120/0.001 /0.0140/0.002

0.016 0.013 0.017

/0.0320/0.032 /0.0320/0.021 /0.0400/0.028

0.503 0.586 0.407

0.0760/0.774 0.1920/0.817 /0.0450/0.721

Knee power 1 /2 1 /3 2 /3

0.000 /0.001 /0.001

0.031 0.032 0.018

/0.0660/0.065 /0.0680/0.066 /0.0380/0.037

0.136 0.139 0.078

/0.2720/0.272 /0.2790/0.277 /0.1560/0.155

0.722 0.673 0.876

0.4100/0.883 0.3270/0.860 0.7080/0.951

Hip power 2 /3 1 /3 1 /2

0.038 0.030 /0.008

0.033 0.022 0.030

/0.0320/0.108 /0.0170/0.077 /0.0710/0.055

0.145 0.097 0.131

/0.2520/0.327 /0.1640/0.224 /0.2690/0.254

0.453 0.343 0.520

0.0120/0.747 /0.1190/0.683 0.0990/0.783

Ankle power 1 /2 1 /3 2 /3

0.049 0.020 /0.029

0.038 0.034 0.056

/0.0310/0.130 /0.0510/0.091 /0.1470/0.089

0.167 0.147 0.245

/0.2840/0.383 /0.2740/0.315 /0.5190/0.461

0.666 0.700 0.349

0.3150/0.856 0.3710/0.873 /0.1120/0.686

Knee power 1 /2 1 /3 2 /3

/0.008 /0.009 /0.001

0.022 0.020 0.018

/0.0530/0.037 /0.0.510/0.034 /0.0380/0.037

0.094 0.089 0.078

/0.1960/0.180 /0.1860/0.169 /0.1560/0.155

0.549 0.509 0.651

0.1390/0.798 0.0840/0.777 0.2910/0.849

Hip power 2 /3 1 /3 1 /2

0.015 /0.070 /0.084

0.042 0.043 0.035

/0.0740/0.104 /0.1600/0.021 /0.1570/0.012

0.184 0.187 0.151

/0.3540/0.383 /0.4450/0.305 /0.3870/0.218

0.604 0.517 0.695

0.2190/0.826 0.0950/0.782 0.3640/0.871

Mid stance

Table 8 Bland and Altman limits of agreement and intraclass ICC for inter-rater reliability of the spatio-temporal gait parameters Velocity

S.E. of d

95% CI for d

S.D. diff


ICC

95% CI

0.00 0.00 0.01

0.01 0.00 0.00

/0.020/0.01 /0.010/0.01 0.000/0.02

0.02 0.02 0.02

/0.050/0.05 /0.030/0.04 /0.040/0.05

0.69 0.81 0.71

0.350/0.87 0.580/0.92 0.400/0.88

Duration of stance 1 /2 0.00 0.00 1 /3 2 /3 0.00

0.00 0.00 0.00

/0.010/0.01 /0.010/0.01 0.000/0.01

0.02 0.02 0.01

/0.040/0.03 /0.030/0.04 /0.030/0.03

0.74 0.71 0.78

0.450/0.89 0.400/0.88 0.510/0.91

Duration of swing 1 /2 0.02 1 /3 0.01 /0.01 2 /3

0.02 0.01 0.02

/0.010/0.06 /0.010/0.04 /0.040/0.02

0.07 0.05 0.07

/0.120/0.17 /0.080/0.11 /0.150/0.13

0.86 0.93 0.85

0.670/0.94 0.830/0.97 0.660/0.94

1 /2 1 /3 2 /3

d

References [1] Winter DA. Kinematic and kinetic patterns in human gait: variability and compensating effects. Hum Move Sci 1984;3:51 / 76. [2] Kadaba MP, Ramakrishnan HK, Wootten ME, Gainey J, Gorton G, Cochran GVB. Repeatability of kinematic, kinetic, and electromyographic data in normal adult gait. J Orthop Res 1989;7:849 /60.

[3] Cowman JP, Jenkinson A, O’Connell P, O’Brien T. A model for establishing reliability and quantifying error associated with routine gait analysis (abstract). Gait Posture 1998;8:79. [4] Steinwender G, Saraph V, Scheiber S, Zwick EB, Uitz C, Hackl K. Intrasubject repeatability of gait analysis data in normal and spastic children. Clin Biomech 2000;15:134 /9. [5] Post PA, Piotter JM, VandenBerg KJ, Alderink GJ, Bothner KE. Reliability of kinematic and kinetic data in the analysis of normal human gait (abstract). Gait Posture 1998;8:127.

V. Maynard et al. / Gait and Posture 17 (2003) 59 /67 [6] Winter DA. The biomechanics and motor control of human movement, 2. New York: Wiley, 1990. [7] Perry J. Gait analysis. Normal and pathological gait. Thorofare, NJ: Slack Inc, 1992. [8] Bland JM, Altman DG. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet 1986;1:307 /10. [9] Rankin G, Stokes M. Reliability of assessment tools in rehabilitation: an illustration of appropriate statistical analyses. Clin Rehab 1998;12:187 /99. [10] Fleiss JK. The design and analysis of clinical experiments. Toronto: Wiley, 1986.

67

[11] Winter DA. Overall principle of lower limb support during stance phase of gait. J Biomech 1980;13:923 /7. [12] Eliasziw M, Young SL, Woodbury MG, Fryday-Field K. Statistical methodology for the concurrent assessment of interrater and intrarater reliability. Phys Ther 1994;74:89 /100. [13] Shrout PE, Fleiss JL. Interclass correlations: uses in assessing rater reliability. Psychol Bull 1979;86:420 /8. [14] Wade DT. Measurement in neurological rehabilitation. Oxford: Oxford University Press, 1992. [15] Diss CE. The reliability of kinetic and kinematic variables used to analyse normal running gait. Gait Posture 2001;14:98 /103.