- Email: [email protected]
Inter- and intra-tester reliability when measuring seated spinal postures with inertial sensors
International Journal of Industrial Ergonomics 44 (2014) 732e738
Contents lists available at ScienceDirect
International Journal of Industrial Ergonomics journal homepage: www.elsevier.com/locate/ergon
Inter- and intra-tester reliability when measuring seated spinal postures with inertial sensors b € € m b, c, O. Lindroos d, C. Ha €ger a, B. Rehn a T.C. Stenlund a, *, F. Ohberg , R. Lundstro a
Dept. of Community Medicine and Rehabilitation, Physiotherapy, Umeå University, 90187 Umeå, Sweden Dept. of Radiation Sciences, Biomedical Engineering, Umeå University, Umeå, Sweden c Dept. of Public Health and Clinical Medicine, Occupational Medicine, Umeå University, Sweden d Dept. of Forest Biomaterials & Technology, Swedish University of Agricultural Sciences, Sweden b
a r t i c l e i n f o
a b s t r a c t
Article history: Received 10 January 2014 Received in revised form 5 June 2014 Accepted 11 June 2014 Available online
Prolonged awkward sitting postures may be associated with neck or back pain, but it is often unclear which specific postures cause most problems and which mechanisms that may underlie the pain. In order to increase the knowledge in this field, it seems crucial first of all to be able to analyse, in depth, different seated spinal postures. A problem is however the lack of reliable and direct measurement methods of the posture, especially for sitting. Recently developed systems with inertial sensor attached along the spine have potential for this purpose. The aim of the present study was therefore to test the reliability of using such a system to assess various seated postures. Inter- and intra-tester as well as intra-subject relative reliability was estimated with intra-class correlation coefficient (ICC). Absolute reliability was estimated with standard error of measurement (SEM) and smallest detectable change (SDC). Ten þ ten healthy subjects and four testers participated. Three standardised unsupported seated postures (lumbar lordosis, lumbar kyphosis and neutral posture) and two standing postures (neutral and lumbar kyphosis) were evaluated using five sensors attached to the head, the thorax (high and low), the lumbar spine and the pelvis. The ICC for intra-tester reliability ranged from 0.37 to 0.90, SEM 2.5e12.0 , and SDC 7.1e33.3 where the largest measurement error was from the head. Intra-tester reliability was higher than inter-tester reliability but not as good as intrasubject reliability. The intra-tester absolute reliability was nevertheless not considered sufficient to distinguish smaller differences. The low reliability may depend on inertial sensor size and attachment but also on the tester's accuracy. This study shows that assessing unsupported seated spinal postures with inertial sensors could be performed with higher reliability if done by the same, rather than different, testers. Relevance to industry: Prolonged awkward seated postures at work may be associated with back and neck pain and should therefore be analysed. Inertial sensor units is a promising tool to measure spinal posture. Smaller sensors attached by one skilled tester directly onto the body will most likely improve assessment in the future. © 2014 Elsevier B.V. All rights reserved.
Keywords: Biomechanics Human engineering Observer variations
1. Introduction Spinal pain and/or discomfort are common ailments in the general population throughout the world and strike almost everyone at some point in life (Cote et al., 2009; Hoy et al., 2010a, 2010b). Type of workplace may influence the prevalence of neck and back pain; for example office workers are more likely to suffer from neck pain. They have a higher one-year prevalence of neck
* Corresponding author. Tel.: þ46 907868040. E-mail address: [email protected] (T.C. Stenlund). http://dx.doi.org/10.1016/j.ergon.2014.06.002 0169-8141/© 2014 Elsevier B.V. All rights reserved.
pain than the general population (Kamwendo et al., 1991; Chiu et al., 2002; Ariens et al., 2001), possibly because of prolonged seated postures (Yue et al., 2012; Ariens et al., 2000). Likewise, excessive sitting periods have commonly been reported as an aggravating factor for low back pain (Williams et al., 1991; BieringSorensen, 1983). People are spending more and more time sedentary because of the demands of modern working life. The use of computers and normally in seated postures, among office employees in Sweden, reaches 75%, and close to 40% use the computer for the majority of their working day (Rackner et al., 2012). Research suggests that sitting duration alone is not a causal factor for developing pain (Roffey et al., 2010; Lis et al., 2007; Kwon et al.,
T.C. Stenlund et al. / International Journal of Industrial Ergonomics 44 (2014) 732e738
2011; Chen et al., 2009; Bakker et al., 2009); there may be other explanatory factors, such as the adopted posture or work task. There is no consensus on an ideal posture for sitting (Pynt et al., 2001; O'Sullivan, 2005; O'Sullivan et al., 2012; Claus et al., 2009). A proper posture in general maintains spinal curvatures, keeping the joints in neutral positions (O'Sullivan et al., 2006; Barrero and Hedge, 2002) to avoid excessive tissue strain which can lead to musculoskeletal disorders (Scannell and McGill, 2003). In previous epidemiological studies, spinal curvatures have seldom been reported, rather the duration and frequency of different postures have been investigated (e.g., sitting versus standing) (Roffey et al., 2010; Kwon et al., 2011). One explanation for the lack of evidence could be non-sensitive measurement techniques. Digital photography and video analysis are commonly used for field studies, but their ability to detect regional changes of spinal posture is limited. Regional changes have only been reported previously in laboratory studies, and differences between symptomatic and asymptomatic groups were revealed only when regional changes were detected (Mitchell et al., 2008; Dankaerts et al., 2006). In order to identify postures divergent from the neutral posture during real working conditions, more sensitive instruments appear necessary. Technology has brought forth inertial sensors that may be used for taking direct longer-duration posture measurements at the workplace. Compared to ordinary laboratory technology, these sensors have the advantages of being small, low-cost, and wearable. Inertial sensors have been used for various purposes such as evaluating movements occurring during gait (Findlow et al., 2008; Mayagoitia et al., 2002), head rotations (Jasiewicz et al., 2007), flexion/extension, lateral flexion, or rotation in the upper body (Wong and Wong, 2008; Plamondon et al., 2007; Ha et al., 2013). None of these studies has focused on spinal postures in seated postures. Inertial sensors have been found to be a feasible method for measuring range of motion (ROM) in the cervical spine and trunk motions (Intolo et al., 2010; Theobald et al., 2012; Ohberg et al., 2013). Intolo and colleagues (Intolo et al., 2010) used one sensor on the lateral iliac crest and found that the reliability of five repeated movements in standing gave an excellent intra-class correlation coefficient (ICC) and a repeatability coefficient in percentages that were higher for small movements than for large. Strain gauge instruments can give high ICC values for inter- and intratester reliability in laboratory settings for measuring spine ROM in standing and ordinary sitting, but there is a lack of angular output i.e. no absolute positions (O'Sullivan et al., 2011). Previous laboratory studies are well defined, but exposure measurements in workplace settings may be less ideal partly because there can be various testers with different skills. Reliability issues could lead to risk assessments being imprecise. The purpose of the present study was to evaluate the inter- and intra-tester reliability among testers using a novel inertial sensors system to measure spinal postures in healthy persons when they are seated in different postures. Our hypothesis was that there would be no difference between testers or within a tester. 2. Methods 2.1. Participants Four testers were involved in the inter-tester reliability test: one physiotherapist (two years of clinical experience), two physiotherapy students, and one biomedical engineer (experience with clinical measurements), each with different knowledge about clinical anatomy. They were chosen to mimic the real-life conditions where various skills among practitioners could affect the outcome when they use inertial sensor units (ISU:s) measuring
733
postures. All testers had knowledge about the technical features of ISU:s but were inexperienced with postural measurements. Young subjects were chosen to eliminate age-related problems such as degeneration and rigidity of the spine. The subjects were excluded according to a health-screening protocol if they reported any neurological conditions or reduced ability to work during the last €m et al., 12 months because of back or neck problems (Lundstro 2004). The subjects of height below 160 cm were excluded because of the risk that the sensors would collide during movements. Ten healthy subjects, two women and eight men, were included in the inter-tester reliability tests, with a mean age (SD) of 28 (6) years, height of 179 (7) cm, and BMI of 23.3 (2.0) kg/m2. For the intra-tester reliability, performed by one and the same physiotherapist as from the inter-reliability test, another ten healthy subjects were recruited, six men and four women, with a mean age of 31 (5) years, height of 176 (9) cm, and BMI of 24.3(3.4) kg/m2. Written informed consent was obtained from each participant, and the Regional Ethical Review Board (No 2012-24-31M) approved all procedures. 2.2. Equipment The Department of Biomedical Engineering and Informatics, University Hospital of Umeå, Sweden, developed the movement capture and analysis system, in this setting consisting of one data collecting unit and five tri-axial inertial sensor units (ISU:s). Each ISU (ADIS 16364 Analog Devices, USA) included three axis gyros and three axis accelerometers, which together detect the threedimensional posture angles and motions of the object. The size of an ISU is L76 W52 H46 mm and the weight is 40 g. The sensors use a local Cartesian coordinate system, which is determined at the time of initial setting. The sensors are connected by cables to a collection unit, which in turn communicates with a laptop via Bluetooth. Customised software (AnyMo, The Department of Biomedical Engineering and Informatics, Umeå University Hospital, Sweden) calculates the real-time orientation of the sensors, using data from the gyros and accelerometers (Ohberg et al., 2013). A stool was adjusted to the level of the posterior knee crease (popliteal height). The stool had no backrest and had a flat wooden surface covered with two layers of 2 mm foam and an anti-slip surface. 2.3. Protocol The study used a repeated-measurement design, test retest with four testers for the inter-tester reliability and one tester for the intra-tester reliability. Following our written study protocol, the tester mounted the ISU:s on the subject, calibrated them, and gave instructions as to the postures that the subject should then take. There was no prior training for this specific procedure. The subjects performed for each tester, as similarly as possible, three tests, two while seated unsupported and one while standing. The reliability within a subject was also analysed to investigate their precision in repeating postures. The standing test, similar to the test by Intolo and colleagues (Intolo et al., 2010), was added because we suspected it might be difficult for the subjects to adopt the seated postures repeatedly with good accuracy. Five ISU:s were used. The position of the lower three units was chosen based on information from a study by Dankaerts and colleagues (Dankaerts et al., 2006), where sensors were taped on the skin over the spinal processes of S2, L3, and T12, because the lower part of the lumbar spine is the most common area for low back pain (Biering-Sorensen, 1983). The two upper units were meant to be positioned on the forehead and over the spinal process of C7, according to Jasiewicz and colleagues (Jasiewicz et al., 2007). The
734
T.C. Stenlund et al. / International Journal of Industrial Ergonomics 44 (2014) 732e738
forehead position remained, but because the sensor size restricted movement, the C7 sensor was placed instead on the spinal process of T2. The placements of the units were palpated while the subjects stood. The units were attached on elastic Velcro straps and on a customised vest, which was chosen because this would probably be the most feasible situation for the upcoming field measurements. The physiotherapist gave each subject short oral information and allowed them to familiarise themselves with the upcoming movements. However, during the actual tests for the inter-tester reliability, the testers were randomised in order and hidden from one another. Thus, each tester individually instructed the subjects and mounted the ISU:s. When the subjects felt confident with the movements, the test began. Between testers, the systems were switched off, demounted by the previous tester, and then remounted and recalibrated by the next tester. Between trials and between testers, the subjects were able to stand or sit freely for up to 5 min to minimise any effects of fatigue. For the intra-tester reliability test, one tester, the physiotherapist, repeated four measurements (with ISU:s remounted and recalibrated between measurement), so that the same number of measurements were carried out as for the inter-tester reliability settings. The measurements took approximately 1.5e2 h per subject and all measurements for a given subject were taken the same day. 2.4. Test procedure During the tests, subjects were seated towards the front of the stool, with approximately 15 cm support from ischial tuberosities to upper thighs. They were facing forward fixing their gaze at a point on the wall at eye height, hips flexed less than 90 , knees flexed and heels under the stool, hands resting lightly on their thighs so that the subjects could achieve the lordosis posture by sufficiently tilting their pelvis (Claus et al., 2009). Analyses of postures were made during the neutral and end range postures (See Fig. 1AeE). The subjects started in a self-selected and unsupported “neutral” posture. Every 5 s a new command with a new posture was adopted going to the maximum lumbar lordosis and, at a new command, back to a neutral posture. This procedure was repeated until both postures were achieved/obtained a total of 5 times. The same procedure was repeated for the maximum lumbar kyphosis posture. The subjects were also asked to stand 10 cm behind the stool, facing forward, feet and shoulders vertically aligned with
straight knees, neutral self-selected spine, and straight arms. Upon command, the subjects flexed their spine to touch a marked line 1 cm from the stool edge with their fingertip, and told to remain that way until a new command was given. Every postural change was initiated upon a verbal command from the tester. 2.5. Data analysis Orientation data (i.e., segment angles) from the five units attached to the forehead and along the spinal processes were processed in Matlab, (version 7.10.0 (R2010a), The MathWorks, Inc., USA). The processed data described the sensors' normal vector relative to the gravitation vector in degrees. The study initialises movement in the sagittal plane. Data were collected at a sample rate of 240 Hz for the sensors. All orientation data was low-pass filtered with a second order Butterworth filter at a cut-off frequency of 10 Hz. The Euler sequence used for all segment angles was XYZ, where X was rotations in the sagittal plane, Y was rotations in the frontal plane and Z was rotations in the transverse plane. A more detailed description of the algorithms used can be € found in the article by Ohberg et al. (Ohberg et al., 2013). A mean value of the segments angles during 0.5 s of a stable phase for every repetition in each posture was collected. All angles were reported for each posture as means and standard deviations. Data are presented as absolute angles, that is, each unit's normal vector in relation to the gravitational vector and not relative to the other. 2.6. Statistical analysis Inter- and intra-tester as well as intra-subject relative reliability was estimated by the intra-class correlation coefficients (ICC). The ICC was calculated with a 95% confidence interval (CI) using model ICC2,1 absolute agreement (Shrout and Fleiss, 1979). A two-way random model was used because tester and subjects were considered random effects. Inter- and intra-tester reliability was calculated from the mean value of 5 repetitions from each specific posture and segment. Included in the inter-tester reliability were those subjects where all measurements for four testers were valid. Included in the intra-tester reliability were those subjects where four measurements for one tester were valid. Intra-subject reliability, calculated from each repetition of a posture, was calculated to act as reference to tester reliability, indicating the individual variability. The strength of the correlation was assessed according
Fig. 1. A-E. Lumbar kyphosis seated: Thoraco-lumbar and lumbar angles kyphosed approaching end of range (Fig. 1A). Neutral posture seated: self-selected spinal posture (Fig. 1B). Lumbar lordosis seated: Thoraco-lumbar and lumbar angles lordosed approaching end of range (Fig. 1C). Neutral posture standing: self-selected spinal posture (Fig. 1D). Lumbar kyphosis standing: Thoraco-lumbar and lumbar angles kyphosed (Fig. 1E).
T.C. Stenlund et al. / International Journal of Industrial Ergonomics 44 (2014) 732e738
to the following criteria: <0.75, Poor to moderate; 0.75e0.90, Good; 0.91-1, Adequate reliability for clinical measurement (Portney and Watkins, 2000). The calculated ICC depends on several variance components which are presented separately to clarify the magnitude of the different sources of variance. s2p represents the variance due to systematic differences in the subjects, s2o represents the variance due to systematic differences in testers (inter-tester reliability) or test-retest (intra-tester and intra-subject), s2residual represent the random error of variance. ICC ¼ s2p =ðs2p þ s2o þ s2residual Þ (de Vet et al., 2011). The Standard Error of Measurement (SEM) and the Smallest Detectable Change (SDC) is calculated for the absolute reliability (agreement) between testers or repeated measurements. SEM represents the measurement error and includes the error variance, systematic and random SEM ¼ √ðs2o þ s2residual Þ (de Vet et al., 2011). pSDC is based on the SEM, and is defined as ffiffiffi SDC ¼ 1*96 2*SEM (de Vet et al., 2011). The SDC provides information about the smallest degree of change between two measurements within a 95% CI. When interpreting the results of accuracy the upper limit of 5 SEM, set by the American Medical Association as an acceptable finding for evaluation of movement in a clinical context, was used (Nitschke et al., 1999). The guidance is set for Range of Motion and here used for absolute angles but still considered as a useful reference. The proportion of differences fewer than 5 between the first and the following measurements or repetition was retrieved to present the agreement in percentage between measurements or repetition (de Winter et al., 2004). Five degrees was chosen because of an assumed clinical significance. Mean values and standard deviations for each posture and segment in respective reliability tests were calculated from every valid value. Analyses were performed using IBM SPSS version 20 (SPSS Inc., Chicago III USA). 3. Results 3.1. Reliability The inter-test reliability was poor to moderate. ICC:s between testers varied between 0.18 and 0.76, SEM varied between 4.1 and 11.5 , and SDC between 9.9 and 31.9 . The number of included tests varied where the head segment was of lowest value. Agreement within 5 varied between 26 and 67%. There was a significant difference between testers at three levels on three occasions; head in neutral posture seated, segment L3 in lumbar lordosis seated and S2 in neutral posture standing. (Table 1). The intra-test reliability proved to be poor to moderate in nine and good in 16 combinations of postures and segments. ICC:s for the same tester varied between 0.37 and 0.90, SEM between 2.5 and 12.0 , and SDC between 7.1 and 33.3 . Number of included tests varied where the head segment was of the lowest value. Agreement within 5 varies between 38 and 81% (Table 2). The intra-subject reliability proved to be adequate in all cases except at level head in neutral posture standing. ICC:s for within a subject varied between 0.81 and 0.98, SEM between 1.3 and 5.1, and SDC between 3.6 and 14.2 . Agreement within 5 varied between 57 and 98% (Table 3). 4. Discussion The purpose of this study was to evaluate if an inertial sensor system may be used reliably by different testers to measure seated spinal postures. The initiated movements and measurements were performed in the sagittal plane. The inter-tester reliability using ISU:s was considered poor to moderate, whereas the intra-tester reliability was considered good in the majority of segments and
735
tests, but not reaching adequate reliability (Portney and Watkins, 2000). The hypothesis was rejected in three cases; at segment head in neutral posture seated, L3 in lumbar lordosis seated and S2 in neutral posture standing, due to a significant difference between testers. Variability within the subjects' test retest situation was rather low. The SEM and SDC for intra-tester reliability are better than for the inter-tester reliability but still considered too high to detect smaller changes. For the intra-tester reliability, the correlation was good in the majority of segments with small variation for ICC:s between the seated lumbar kyphosis and standing kyphosis, but with more differences for seated lumbar lordosis in the three lowest segments. This is in accordance with the study by Claus and colleagues (Claus et al., 2009) showing that a slump posture, named “seated lumbar kyphosis” in our study, is easier to repeat. Further, the ICC:s from intra-tester reliability indicate small variation between the seated lumbar kyphosis and a neutral posture, except for Th12. The possibility to repeat a neutral posture with high reliability in the lowest segment is in accordance with the report of O’Sullivan and colleagues (O'Sullivan et al., 2010) showing that pain-free subjects' own ideal seated posture and a tester perceived neutral posture are not significantly different and may be tested with very high intertester reliability. 4.1. Reliability considerations The inter- and intra-tester reliability for measurement of seated and standing posture shows variation that cannot be explained by the intra-subject reliability. There is a variance in the tested subjects, shown by the standard deviations in the angle outcome, affecting the subject variance component. The difference from the intra-subject reliability to inter-tester reliability is not mainly caused by the subject variance component but by the fact that the error variance components have increased. There are several causes for the variation, such as finding the spinal processes, attachment of the units, and the movements of the Velcro straps and the vest on skin and clothes. Different skills and accuracy among the testers could influence these results i.e. how they identify the specific spinal processes or how tight the Velcro straps were set. The testers were all novices in respect of this specific set-up and training would probably improve accuracy. However, all had the same written information about the task. The ICC:s for the inter-tester test show poor to moderate correlation with more moderate values in the maximal lumbar kyphosis and lordosis, whereas correlations in the neutral posture are poorer. The agreement between testers within 5 is more equally distributed, approximately 25e50%. The absolute reliability is influenced by high residual variances and occasionally systematic error variances why the majority of SEM:s exceeds the 5 limit for acceptability findings (Nitschke et al., 1999). In addition, high SDC:s indicate that small differences can not be detected. The ICC:s for intra-tester reliability show, on the other hand, good correlation in most cases with SEMs below the 5 limit for the majority of segments (Nitschke et al., 1999) and improved SDC:s. Still, in the majority of segments there has to be >10 differences between two measures before a clinical difference is found. Adding more testers and subjects may influence the results as this study involves only one tester measuring ten subjects repeatedly. However, as expected our results indicate, that one tester has better reliability than several testers. Still, the reliability can most likely be improved. According to O’Sullivan and colleagues (O'Sullivan et al., 2011), a better outcome could be expected with attachment on the skin, but it should be noted that their results are based on measurement in ROM and not angular output. One concern in the test procedure was regarding the subject's ability to occupy the same posture repeatedly in the test-retest
736
T.C. Stenlund et al. / International Journal of Industrial Ergonomics 44 (2014) 732e738
Table 1 Inter-tester reliability. Absolute angles presented as means and standard deviations, intra-class correlation coefficients (ICC2,1 absolute agreement), s2p variance component of subjects, s2o variance component of testers, s2residual variance component of random error variance standard error of measurement (SEM), smallest detectable change (SDC), and measurements within 5 (1 ¼ 100%) between 4 measurements made by 4 different testers in a test retest for different postures and placement of inertial sensor units (ISU). Posture
n
Segment
Mean (SD)
ICC2,1 (95% CI)
s2p
s2o
s2residual
SEM
Lumbar kyphosis seated
6 10 10 10 10 4 9 9 9 9 4 9 9 9 9 5 8 9 9 9 5 8 8 9 9
Head Th2 Th12 L3 S2 Head Th2 Th12 L3 S2 Head Th2 Th12 L3 S2 Head Th2 Th12 L3 S2 Head Th2 Th12 L3 S2
106.3 41.5 88.3 98.7 103.6 117.9 64.7 96.9 95.9 93.1 122.6 79.7 104.9 92.9 85.5 114.8 67.0 103.6 91.0 71.9 48.9 10.4 50.9 51.6 46.2
0.68 0.60 0.63 0.67 0.74 0.47 0.49 0.65 0.50 0.73 0.67 0.59 0.54 0.24 0.61 0.65 0.49 0.43 0.49 0.18 0.62 0.63 0.62 0.66 0.76
140.5 87.0 54.1 41.0 86.6 117.6 38.0 31.2 20.9 87.8 132.3 63.8 53.7 12.5 59.2 96.8 28.7 17.5 30.2 9.7 103.7 39.8 53.0 24.3 84.1
0 4.8 9.6 2.9 3.8 0 2.4 1.7 0.6 3.4 0 1.2 6.4 2.9 5.4 0 0 0 9.1 2.5 0 1.3 0.2 0 4.3
64.5 52.4 21.9 17.4 26.4 132.8 37.5 15.3 20.2 29.3 64.5 43.5 40.2 35.8 31.8 51.3 29.8 22.8 21.8 41.1 64.5 23.5 33.0 12.7 22.1
8.0 7.6 5.6 4.5 5.5 11.5 6.3 4.1 4.6 5.7 8.0 6.7 6.8 6.2 6.1 7.2 5.5 4.8 5.6 6.6 8.0 5.0 5.8 3.6 5.1
Neutral posture seated
Lumbar lordosis seated
Neutral posture standing
Lumbar kyphosis standing
(15.1) (11.7) (8.9) (7.6) (10.5) (12.0) (9.5) (7.3) (5.9) (10.3) (13.4) (10.0) (9.4) (6.9) (9.1) (11.1) (7.3) (5.8) (7.3) (7.0) (11.1) (7.9) (8.7) (6.8) (10.8)
(0.20e0.92) (0.31e0.86) (0.32e0.87) (0.39e0.89) (0.49e0.92) (0.02e0.94)a (0.17e0.82) (0.34e0.89) (0.18e0.82) (0.46e0.92) (0.18e0.97) (0.27e0.86) (0.22e0.84) (0.03e0.66)a (0.30e0.87) (0.21e0.95) (0.14e0.84) (0.07e0.81) (0.18e0.82) (0.07e0.61)a (0.18e0.94) (0.29e0.89) (0.28e0.89) (0.35e0.89) (0.50e0.93)
SDC
22.3 21.0 15.6 12.5 15.2 31.9 17.5 11.4 12.6 15.9 22.3 18.5 18.9 17.2 16.9 19.9 15.1 13.2 15.4 18.3 22.3 13.8 16.0 9.9 14.2
<5 0.32 0.30 0.45 0.40 0.28 0.35 0.28 0.45 0.40 0.28 0.41 0.24 0.39 0.26 0.34 0.38 0.31 0.47 0.45 0.32 0.31 0.47 0.67 0.53 0.26
SD, standard deviation; CI, confidence interval; n number of subjects included in the ICC analysis. a
Significant difference between testers.
situation. A way to reduce the risk of too excessive variability was to perform all tests in a single session. Previous studies have used similar seated postures (Claus et al., 2009; O'Sullivan et al., 2010; Caneiro et al., 2010) but also facilitated their posture with verbal and manual guidance. Subjects need facilitation to find an upright sitting posture, and to mimic a posture just by looking at a picture may not be enough (Claus et al., 2009). This suggests that our standardised instructions may be insufficient and that some of the
error is due to the protocol itself. To assess that the standardised seated posture was not problematic to repeat, a standing posture similar to that set by Intolo and colleagues which tested that posture as ROM with an ICC value of 0.90 was added to the protocol (Intolo et al., 2010). The intra-subject reliability shows adequate correlation, except for the head segment in the neutral standing posture, with only a small variation for ICC:s between the seated and standing postures. All tests for a subject were made
Table 2 Intra-tester reliability. Absolute angles presented as means and standard deviations, intra-class correlation coefficients (ICC2,1 absolute agreement), s2p variance component of subjects, s2o variance component of test-retest, s2residual variance component of random error, standard error of measurement (SEM), smallest detectable change (SDC), and measurements within 5 (1 ¼ 100%) between 4 measurements made by one tester in a test retest for different postures and placement of inertial sensor units (ISU). Posture
n
Segment
Mean (SD)
ICC2,1 (95% CI)
s2p
s2o
s2residual
SEM
Lumbar kyphosis seated
8 7 7 8 8 8 8 8 8 8 8 8 8 8 8 4 6 6 6 6 4 6 6 6 6
Head Th2 Th12 L3 S2 Head Th2 Th12 L3 S2 Head Th2 Th12 L3 S2 Head Th2 Th12 L3 S2 Head Th2 Th12 L3 S2
97.5 46.7 91.8 107.3 108.8 110.7 64.1 94.4 104.7 103.5 123.2 80.3 101.1 101.5 97.4 111.6 66.4 101.6 90.8 78.2 37.3 5.7 46.5 55.3 53.1
0.81 0.74 0.82 0.80 0.77 0.84 0.74 0.37 0.73 0.88 0.90 0.90 0.56 0.63 0.89 0.84 0.57 0.49 0.81 0.61 0.84 0.87 0.87 0.86 0.76
108.7 50.8 39.0 56.9 52.4 58.9 26.4 12.7 38.2 104.5 224.1 123.4 64.0 32.1 141.3 38.9 13.8 22.4 120.9 37.9 752.8 78.3 43.0 179.8 81.3
0 0.2 0.9 0 0 0.8 1.4 0 0 0 0 1.0 0 0 0 0 0 0 4.6 2.8 0 1.3 0.2 0 0
25.7 17.9 7.4 14.1 15.4 10.8 7.8 21.2 13.8 15.0 25.7 12.3 50.1 19.2 18.2 7.7 10.3 22.0 23.3 21.6 144.5 10.3 6.3 28.2 25.5
5.1 4.3 2.9 3.8 3.9 3.4 3.0 4.6 3.7 3.9 5.1 3.6 7.1 4.4 4.3 2.8 3.2 4.7 5.3 4.9 12.0 3.4 2.5 5.3 5.0
Neutral posture seated
Lumbar lordosis seated
Neutral posture standing
Lumbar kyphosis standing
(12.5) (8.8) (6.5) (7.9) (7.6) (8.2) (5.6) (5.8) (6.9) (10.0) (14.7) (10.5) (9.6) (6.8) (11.6) (7.7) (4.7) (6.9) (10.5) (8.5) (21.3) (7.6) (9.1) (11.8) (9.5)
(0.55e0.95) (0.43e0.94) (0.57e0.96) (0.54e0.95) (0.49e0.94) (0.62e0.96) (0.45e0.93) (0.02e0.78) (0.44e0.93) (0.69e0.97) (0.74e0.98) (0.75e0.98) (0.20e0.87) (0.29e0.89) (0.71e0.97) (0.48e0.99) (0.15e0.91) (0.06e0.89) (0.52e0.97) (0.23e0.92) (0.44e0.99) (0.65e0.98) (0.64e0.98) (0.61e0.98) (0.41e0.96)
SD, standard deviation; CI, confidence interval; n, number of subjects included in the ICC analysis.
SDC 14.1 11.8 8.0 10.4 10.9 9.4 8.4 12.8 10.3 10.7 14.1 10.1 19.6 12.1 11.8 7.7 8.9 13.0 14.6 13.7 33.3 9.4 7.1 14.7 14.0
<5 0.39 0.56 0.78 0.79 0.64 0.71 0.68 0.68 0.71 0.57 0.57 0.64 0.36 0.61 0.61 0.81 0.60 0.64 0.48 0.60 0.38 0.76 0.80 0.52 0.48
T.C. Stenlund et al. / International Journal of Industrial Ergonomics 44 (2014) 732e738
737
Table 3 Intra-subject reliability. Absolute angles presented as means and standard deviations, intra-class correlation coefficients (ICC2,1 absolute agreement), s2p variance component of subjects, s2o variance component of repetitions, s2residual variance component of random error, standard error of measurement (SEM), smallest detectable change (SDC), and measurements within 5 (1 ¼ 100%) between repetitions within each measurement in a test retest situation in different postures and placement of inertial sensor units (ISU). Posture
n
Segment
Mean (SD)
ICC2,1 (95% CI)
s2p
s2o
s2residual
SEM
Lumbar kyphosis seated
72 76 76 76 73 71 76 76 76 74 71 75 75 75 76 65 73 73 74 74 64 72 76 73 73
Head Th2 Th12 L3 S2 Head Th2 Th12 L3 S2 Head Th2 Th12 L3 S2 Head Th2 Th12 L3 S2 Head Th2 Th12 L3 S2
101.7 44.0 90.0 102.8 106.1 114.1 65.1 95.6 100.5 98.4 120.8 79.9 103.0 97.1 91.3 113.2 66.7 102.6 90.9 74.8 43.4 8.1 48.7 53.4 49.5
0.93 0.95 0.96 0.97 0.96 0.91 0.92 0.95 0.97 0.98 0.98 0.97 0.95 0.97 0.98 0.81 0.91 0.96 0.96 0.96 0.92 0.96 0.96 0.95 0.94
207.6 114.0 58.3 74.4 90.8 104.5 52.5 39.9 59.0 126.3 392.5 101.0 93.9 65.7 142.8 91.9 37.4 40.2 78.4 69.4 310.3 64.0 72.5 92.3 115.3
0.2 0.4 0.1 0.1 0.2 0 0.5 0.1 0.4 0.5 0 0 0.2 0.1 0.4 1.9 0.2 0.1 0.3 0.3 6.1 0.3 0.3 0.8 1.6
15.1 6.3 2.4 2.4 3.2 9.9 3.9 2.2 1.7 2.7 8.3 3.0 4.9 2.1 2.0 20.4 3.6 1.6 3.0 2.5 20.1 2.4 2.9 4.1 5.5
3.9 2.9 1.6 1.6 1.8 3.1 2.1 1.5 1.4 1.8 2.9 1.7 2.3 1.5 1.5 4.7 1.9 1.3 1.8 1.7 5.1 1.6 1.8 2.2 2.7
Neutral posture seated
Lumbar lordosis seated
Neutral posture standing
Lumbar kyphosis standing
(14.8) (8.0) (8.0) (8.8) (9.6) (10.6) (7.5) (6.5) (7.8) (11.3) (15.5) (10.3) (9.8) (8.1) (12.0) (10.6) (6.4) (6.4) (8.9) (8.5) (18.1) (8.1) (9.1) (9.8) (10.9)
(0.91e0.95) (0.92e0.96) (0.94e0.97) (0.95e0.98) (0.95e0.98) (0.88e0.94) (0.89e0.95) (0.93e0.96) (0.95e0.98) (0.96e0.98) (0.97e0.99) (0.96e0.98) (0.93e0.97) (0.95e0.98) (0.97e0.99) (0.73e0.87) (0.87e0.94) (0.94e0.97) (0.94e0.97) (0.94e0.97) (0.88e0.95) (0.94e0.97) (0.94e0.97) (0.92e0.97) (0.91e0.96)
SDC 10.8 7.2 4.4 4.4 5.1 8.7 5.8 4.2 4.0 5.0 8.0 4.8 6.3 4.1 4.3 13.1 5.4 3.6 5.0 4.6 14.2 4.6 5.0 6.1 7.4
<5 0.73 0.90 0.93 0.92 0.93 0.76 0.80 0.95 0.93 0.91 0.71 0.91 0.86 0.95 0.92 0.67 0.92 0.98 0.90 0.93 0.57 0.93 0.93 0.89 0.77
SD, standard deviation; CI, confidence interval; n number of measurements included in the ICC analysis.
on the same day. Probably, a less ideal ICC value would be the outcome if the intra-subject reliability were to be repeated one day later. 4.2. Methodological considerations Firstly, there were different subjects within inter- and intratester reliability tests because those tests were made on separate occasions. We attempted to match subjects as much as possible but there exists a small difference in age, height, BMI and gender. When comparing the outcome for each segment from the two groups, a small difference is noticed for the two lowest segments, L3 and S2. Gender e women have more lumbar lordosis (Norton et al., 2004) and BMI, found to give regional differences in different sitting postures (Mitchell et al., 2008) e could influence the regional differences. Still, we cannot conclude whether this is because of the subjects, the tester, or a combination of both. Secondly, the number of subjects is similar or fewer than in other studies. On the other hand, the number of testers and repetitions is higher compared to the same studies (Theobald et al., 2012; O'Sullivan et al., 2011; O'Sullivan et al., 2010). Including more testers for the test-retest would probably result in improved intra-tester reliability. Nevertheless, the results show that one and the same tester is superior compared to combining several testers. Some technical and operating errors led to some loss of data that could have influenced the results. However, the achieved results were quite consistent and we believe that the valid measures are representative. Thirdly, the results are described as absolute angular outputs instead of a relative angular outputs (Wong and Wong, 2008) or percentage of ROM (Jasiewicz et al., 2007; O'Sullivan et al., 2011). A relative angle could contribute more knowledge of two segments' relative movement to each other but could lack information if both segments move in the same direction. In contrast, such movement could be detected with absolute angle output. Information about ROM gives good clinical information about movement but less information about actual posture. Technically, the ISU:s would be
able to describe the movement and the position in several dimensions but in this study the data was reduced to describe the gyros absolute angular output in predefined postures. To not add more dimensions was a choice due to the fact that seated postures earlier have been described in the sagittal plane but also that adding more dimensions may increase the risk of errors. However, the results strongly depend on the placements of the ISU:s and therefore are a reflection of the testers' skill and experimental accuracy, which was evaluated here. Fourthly, the attachment of the units seems to be of high importance. In this study, we attached the units to an elastic Velcro strap which was placed on the participant's thin clothing. This created a shield between the spine and units. Other studies have used only adhesive tape directly onto the skin (Theobald et al., 2012) or a combination of tape and straps (Wong and Wong, 2008; Ha et al., 2013). The major reason to use a Velcro strap and a vest is that future field measurements at a workplace would probably not be possible without subjects wearing their working clothes. Fifthly, the sensor size in this setting is quite large compared to some other inertial sensors (Wong and Wong, 2008) or other techniques like strain gauge (O'Sullivan et al., 2011).The size is important because a bigger sensor tends to cover several segments in the spinal column. There is also a greater risk that large units could collide when placed near each other and could tend to float on the belly muscle instead of being in contact with the spinal processes. Smaller units have already been developed, so future measurements should be improved. 4.3. Practical implications As the inter-tester reliability was low it suggests that performing repeated measurements of seated postures with ISU:s in the field should preferably be carried out by one and the same assessor. The reliability would probably be even poorer during real life working conditions, due to high variability within subjects when seated without instructions. Another difference from real conditions is the possibility of using backrests. We chose not to use a backrest as
738
T.C. Stenlund et al. / International Journal of Industrial Ergonomics 44 (2014) 732e738
mechanical contact would lead to artefacts from the ISU:s. Further, the size and placement of ISU:s would prevent the use of a backrest. In future field measurements, it seems necessary to find alternative ISU placements - perhaps on the frontal side of the trunk or to use fewer but representative ISU:s. 4.4. Conclusion This study shows that measuring unsupported seated spinal postures with inertial sensors could be performed with higher reliability if done by the same rather than different testers. The reason for this difference is mainly due to the individual tester's level of skill and experience to assess postures. The reliability for a single tester is mostly good, but still insufficient, to distinguish small differences in posture. The reliability figures will most likely been higher with access to smaller inertial sensors that also would have facilitated unit attachment, particularly over the spine. Acknowledgements This study was made possible through the foundation created by from Centre for Environmental Research in Umeå, AFA Insurance, and the Swedish Agency for Economic and Regional Growth e Goal 2 (Project: 158715 CMTF 2 e AnyMo). The authors also would like to acknowledge funding support from the Inga Britt & Arne Lundberg Research Foundation. The authors are grateful to physiotherapy students Frans Unnemyr and Alex Nassiri for their contribution to the data collection. References Ariens, G.A., van Mechelen, W., Bongers, P.M., Bouter, L.M., van der Wal, G., 2000. Physical risk factors for neck pain. Scand. J. Work Environ. Health 26, 7e19. Ariens, G.A., Bongers, P.M., Douwes, M., Miedema, M.C., Hoogendoorn, W.E., van der Wal, G., Bouter, L.M., van Mechelen, W., 2001. Are neck flexion, neck rotation, and sitting at work risk factors for neck pain? Results of a prospective cohort study. Occup. Environ. Med. 58, 200e207. Bakker, E.W., Verhagen, A.P., van Trijffel, E., Lucas, C., Koes, B.W., 2009. Spinal mechanical load as a risk factor for low back pain: a systematic review of prospective cohort studies. Spine (Phila Pa 1976) 34, E281eE293. Barrero, M., Hedge, A., 2002. Computer environments for children: a review of design issues. Work 18, 227e237. Biering-Sorensen, F., 1983. A prospective study of low back pain in a general population. II. Location, character, aggravating and relieving factors. Scand. J. Rehabil. Med. 15, 81e88. Caneiro, J.P., O'Sullivan, P., Burnett, A., Barach, A., O'Neil, D., Tveit, O., Olafsdottir, K., 2010. The influence of different sitting postures on head/neck posture and muscle activity. Man. Ther. 15, 54e60. Chen, S.M., Liu, M.F., Cook, J., Bass, S., Lo, S.K., 2009. Sedentary lifestyle as a risk factor for low back pain: a systematic review. Int. Arch. Occup. Environ. Health 82, 797e806. Chiu, T.T., Ku, W.Y., Lee, M.H., Sum, W.K., Wan, M.P., Wong, C.Y., Yuen, C.K., 2002. A study on the prevalence of and risk factors for neck pain among university academic staff in Hong Kong. J. Occup. Rehabil. 12, 77e91. Claus, A.P., Hides, J.A., Moseley, G.L., Hodges, P.W., 2009. Is 'ideal' sitting posture real? Measurement of spinal curves in four sitting postures. Man Ther. 14, 404e408. Cote, P., van der Velde, G., Cassidy, J.D., Carroll, L.J., Hogg-Johnson, S., Holm, L.W., Carragee, E.J., Haldeman, S., Nordin, M., Hurwitz, E.L., Guzman, J., Peloso, P.M., 2009. The burden and determinants of neck pain in workers: results of the bone and joint decade 2000e2010 task force on neck pain and its associated disorders. J. Manipulative Physiol. Ther. 32, S70eS86. Dankaerts, W., O'Sullivan, P., Burnett, A., Straker, L., 2006. Differences in sitting postures are associated with nonspecific chronic low back pain disorders when patients are subclassified. Spine (Phila Pa 1976) 31, 698e704. , W., van de Winter, A., Heemskerk, M., Terwee, C., Jans, M., Deville Schaardenburg, D.-J., Scholten, R., Bouter, L., 2004. Inter-observer reproducibility of measurements of range of motion in patients with shoulder pain using a digital inclinometer. BMC Musculoskelet. Disord. 5, 1e8. Findlow, A., Goulermas, J.Y., Nester, C., Howard, D., Kenney, L.P., 2008. Predicting lower limb joint kinematics using wearable motion sensors. Gait Posture 28, 120e126. Ha, T.H., Saber-Sheikh, K., Moore, A.P., Jones, M.P., 2013. Measurement of lumbar spine range of movement and coupled motion using inertial sensors e A protocol validity study. Man. Ther. 18, 87e91.
Hoy, D., Brooks, P., Blyth, F., Buchbinder, R., 2010. The epidemiology of low back pain. Best Pract. Res. Clin. Rheumatol. 24, 769e781. Hoy, D.G., Protani, M., De, R., Buchbinder, R., 2010. The epidemiology of neck pain. Best Pract. Res. Clin. Rheumatol. 24, 783e792. Intolo, P., Carman, A.B., Milosavljevic, S., Abbott, J.H., Baxter, G.D., 2010. The spineangel: examining the validity and reliability of a novel clinical device for monitoring trunk motion. Man Ther. 15, 160e166. Jasiewicz, J.M., Treleaven, J., Condie, P., Jull, G., 2007. Wireless orientation sensors: their suitability to measure head movement for neck pain assessment. Man Ther. 12, 380e385. Kamwendo, K., Linton, S.J., Moritz, U., 1991. Neck and shoulder disorders in medical secretaries. Part I. Pain prevalence and risk factors. Scand. J. Rehabil. Med. 23, 127e133. Kwon, B.K., Roffey, D.M., Bishop, P.B., Dagenais, S., Wai, E.K., 2011. Systematic review: occupational physical activity and low back pain. Occup. Med. (Lond) 61, 541e548. Lis, A.M., Black, K.M., Korn, H., Nordin, M., 2007. Association between sitting and occupational LBP. Eur. Spine J. 16, 283e298. €m, R., Burstro €m, L., Hagberg, M., Nilsson, T., 2004. WBV initial questionLundstro naire: Swedish translation. Risks of Occupational Vibration Exposures (VIBRISKS). Final report. Supplement 3 to Annex 12, Appendix 4a. European Commission. Quality of Life and Management of Living Resources Programme. Key Action 4 e Environment and Health. FP5 Project No. QLK4-2002-02650, January 2003 to December 2006. www.vibrisks.soton.ac.uk/. Mayagoitia, R.E., Nene, A.V., Veltink, P.H., 2002. Accelerometer and rate gyroscope measurement of kinematics: an inexpensive alternative to optical motion analysis systems. J. Biomech. 35, 537e542. Mitchell, T., O'Sullivan, P.B., Burnett, A.F., Straker, L., Smith, A., 2008. Regional differences in lumbar spinal posture and the influence of low back pain. BMC Musculoskelet. Disord. 9, 152. Nitschke, J.E., Nattrass, C.L., Disler, P.B., Chou, M.J., Ooi, K.T., 1999. Reliability of the American Medical Association Guides' model for measuring spinal range of motion e Its implication for whole-person impairment rating. Spine 24, 262e268. Norton, B.J., Sahrmann, S.A., Van Dillen, F.L., 2004. Differences in measurements of lumbar curvature related to gender and low back pain. J. Orthop. Sports Phys. Ther. 34, 524e534. O'Sullivan, P.B., Dankaerts, W., Burnett, A.F., Farrell, G.T., Jefford, E., Naylor, C.S., O'Sullivan, K.J., 2006. Effect of different upright sitting postures on spinal-pelvic curvature and trunk muscle activation in a pain-free population. Spine (Phila Pa 1976) 31, E707eE712. O'Sullivan, K., O'Dea, P., Dankaerts, W., O'Sullivan, P., Clifford, A., O'Sullivan, L., 2010. Neutral lumbar spine sitting posture in pain-free subjects. Man. Ther. 15, 557e561. O'Sullivan, K., Galeotti, L., Dankaerts, W., O'Sullivan, L., O'Sullivan, P., 2011. The between-day and inter-rater reliability of a novel wireless system to analyse lumbar spine posture. Ergonomics 54, 82e90. O'Sullivan, K., O'Sullivan, P., O'Sullivan, L., Dankaerts, W., 2012. What do physiotherapists consider to be the best sitting spinal posture? Man Ther. 17, 432e437. O'Sullivan, P., 2005. Diagnosis and classification of chronic low back pain disorders: maladaptive movement and motor control impairments as underlying mechanism. Man. Ther. 10, 242e255. Ohberg, F., Lundstrom, R., Grip, H., 2013. Comparative analysis of different adaptive filters for tracking lower segments of a human body using inertial motion sensors. Meas. Sci. Technol. 24. Plamondon, A., Delisle, A., Larue, C., Brouillette, D., McFadden, D., Desjardins, P., Lariviere, C., 2007. Evaluation of a hybrid system for three-dimensional measurement of trunk posture in motion. Appl. Ergon. 38, 697e712. Portney, L.G., Watkins, M.P., 2000. Foundations of Clinical Research: Application to Practice, second ed. Prentice Hall, Upper Saddle River, N.J. Pynt, J., Higgs, J., Mackey, M., 2001. Seeking the optimal posture of the seated lumbar spine. Physiother. Theory Pract. 17, 5e21. Rackner, A.D., Samuelsson, C., 2012. Statistics Sweden. Living conditions report 126, working hours and work environment 2010e2011. In: Sweden, S. (Ed.), Living Condition Report, Statistics Sweden. Social Welfare Statistics Unit, Stockholm, p. 114. Roffey, D.M., Wai, E.K., Bishop, P., Kwon, B.K., Dagenais, S., 2010. Causal assessment of occupational sitting and low back pain: results of a systematic review. Spine J. 10, 252e261. Scannell, J.P., McGill, S.M., 2003. Lumbar postureeshould it, and can it, be modified? A study of passive tissue stiffness and lumbar position during activities of daily living. Phys. Ther. 83, 907e917. Shrout, P.E., Fleiss, J.L., 1979. Intraclass correlations: uses in assessing rater reliability. Psychol. Bull. 86, 420e428. Theobald, P.S., Jones, M.D., Williams, J.M., 2012. Do inertial sensors represent a viable method to reliably measure cervical spine range of motion? Man. Ther. 17, 92e96. de Vet, H.C.W., Terwee, C.B., Mokkink, L.B., Knol, D.L., 2011. Measurement in Medicine : a Practical Guide. Cambridge University Press, New York, United States of America. Williams, M.M., Hawley, J.A., McKenzie, R.A., van Wijmen, P.M., 1991. A comparison of the effects of two sitting postures on back and referred pain. Spine (Phila Pa 1976) 16, 1185e1191. Wong, W.Y., Wong, M.S., 2008. Trunk posture monitoring with inertial sensors. Eur. Spine J. 17, 743e753. Yue, P., Liu, F., Li, L., 2012. Neck/shoulder pain and low back pain among school teachers in China, prevalence and risk factors. BMC Public Health 12, 789.
Contents lists available at ScienceDirect
International Journal of Industrial Ergonomics journal homepage: www.elsevier.com/locate/ergon
Inter- and intra-tester reliability when measuring seated spinal postures with inertial sensors b € € m b, c, O. Lindroos d, C. Ha €ger a, B. Rehn a T.C. Stenlund a, *, F. Ohberg , R. Lundstro a
Dept. of Community Medicine and Rehabilitation, Physiotherapy, Umeå University, 90187 Umeå, Sweden Dept. of Radiation Sciences, Biomedical Engineering, Umeå University, Umeå, Sweden c Dept. of Public Health and Clinical Medicine, Occupational Medicine, Umeå University, Sweden d Dept. of Forest Biomaterials & Technology, Swedish University of Agricultural Sciences, Sweden b
a r t i c l e i n f o
a b s t r a c t
Article history: Received 10 January 2014 Received in revised form 5 June 2014 Accepted 11 June 2014 Available online
Prolonged awkward sitting postures may be associated with neck or back pain, but it is often unclear which specific postures cause most problems and which mechanisms that may underlie the pain. In order to increase the knowledge in this field, it seems crucial first of all to be able to analyse, in depth, different seated spinal postures. A problem is however the lack of reliable and direct measurement methods of the posture, especially for sitting. Recently developed systems with inertial sensor attached along the spine have potential for this purpose. The aim of the present study was therefore to test the reliability of using such a system to assess various seated postures. Inter- and intra-tester as well as intra-subject relative reliability was estimated with intra-class correlation coefficient (ICC). Absolute reliability was estimated with standard error of measurement (SEM) and smallest detectable change (SDC). Ten þ ten healthy subjects and four testers participated. Three standardised unsupported seated postures (lumbar lordosis, lumbar kyphosis and neutral posture) and two standing postures (neutral and lumbar kyphosis) were evaluated using five sensors attached to the head, the thorax (high and low), the lumbar spine and the pelvis. The ICC for intra-tester reliability ranged from 0.37 to 0.90, SEM 2.5e12.0 , and SDC 7.1e33.3 where the largest measurement error was from the head. Intra-tester reliability was higher than inter-tester reliability but not as good as intrasubject reliability. The intra-tester absolute reliability was nevertheless not considered sufficient to distinguish smaller differences. The low reliability may depend on inertial sensor size and attachment but also on the tester's accuracy. This study shows that assessing unsupported seated spinal postures with inertial sensors could be performed with higher reliability if done by the same, rather than different, testers. Relevance to industry: Prolonged awkward seated postures at work may be associated with back and neck pain and should therefore be analysed. Inertial sensor units is a promising tool to measure spinal posture. Smaller sensors attached by one skilled tester directly onto the body will most likely improve assessment in the future. © 2014 Elsevier B.V. All rights reserved.
Keywords: Biomechanics Human engineering Observer variations
1. Introduction Spinal pain and/or discomfort are common ailments in the general population throughout the world and strike almost everyone at some point in life (Cote et al., 2009; Hoy et al., 2010a, 2010b). Type of workplace may influence the prevalence of neck and back pain; for example office workers are more likely to suffer from neck pain. They have a higher one-year prevalence of neck
* Corresponding author. Tel.: þ46 907868040. E-mail address: [email protected] (T.C. Stenlund). http://dx.doi.org/10.1016/j.ergon.2014.06.002 0169-8141/© 2014 Elsevier B.V. All rights reserved.
pain than the general population (Kamwendo et al., 1991; Chiu et al., 2002; Ariens et al., 2001), possibly because of prolonged seated postures (Yue et al., 2012; Ariens et al., 2000). Likewise, excessive sitting periods have commonly been reported as an aggravating factor for low back pain (Williams et al., 1991; BieringSorensen, 1983). People are spending more and more time sedentary because of the demands of modern working life. The use of computers and normally in seated postures, among office employees in Sweden, reaches 75%, and close to 40% use the computer for the majority of their working day (Rackner et al., 2012). Research suggests that sitting duration alone is not a causal factor for developing pain (Roffey et al., 2010; Lis et al., 2007; Kwon et al.,
T.C. Stenlund et al. / International Journal of Industrial Ergonomics 44 (2014) 732e738
2011; Chen et al., 2009; Bakker et al., 2009); there may be other explanatory factors, such as the adopted posture or work task. There is no consensus on an ideal posture for sitting (Pynt et al., 2001; O'Sullivan, 2005; O'Sullivan et al., 2012; Claus et al., 2009). A proper posture in general maintains spinal curvatures, keeping the joints in neutral positions (O'Sullivan et al., 2006; Barrero and Hedge, 2002) to avoid excessive tissue strain which can lead to musculoskeletal disorders (Scannell and McGill, 2003). In previous epidemiological studies, spinal curvatures have seldom been reported, rather the duration and frequency of different postures have been investigated (e.g., sitting versus standing) (Roffey et al., 2010; Kwon et al., 2011). One explanation for the lack of evidence could be non-sensitive measurement techniques. Digital photography and video analysis are commonly used for field studies, but their ability to detect regional changes of spinal posture is limited. Regional changes have only been reported previously in laboratory studies, and differences between symptomatic and asymptomatic groups were revealed only when regional changes were detected (Mitchell et al., 2008; Dankaerts et al., 2006). In order to identify postures divergent from the neutral posture during real working conditions, more sensitive instruments appear necessary. Technology has brought forth inertial sensors that may be used for taking direct longer-duration posture measurements at the workplace. Compared to ordinary laboratory technology, these sensors have the advantages of being small, low-cost, and wearable. Inertial sensors have been used for various purposes such as evaluating movements occurring during gait (Findlow et al., 2008; Mayagoitia et al., 2002), head rotations (Jasiewicz et al., 2007), flexion/extension, lateral flexion, or rotation in the upper body (Wong and Wong, 2008; Plamondon et al., 2007; Ha et al., 2013). None of these studies has focused on spinal postures in seated postures. Inertial sensors have been found to be a feasible method for measuring range of motion (ROM) in the cervical spine and trunk motions (Intolo et al., 2010; Theobald et al., 2012; Ohberg et al., 2013). Intolo and colleagues (Intolo et al., 2010) used one sensor on the lateral iliac crest and found that the reliability of five repeated movements in standing gave an excellent intra-class correlation coefficient (ICC) and a repeatability coefficient in percentages that were higher for small movements than for large. Strain gauge instruments can give high ICC values for inter- and intratester reliability in laboratory settings for measuring spine ROM in standing and ordinary sitting, but there is a lack of angular output i.e. no absolute positions (O'Sullivan et al., 2011). Previous laboratory studies are well defined, but exposure measurements in workplace settings may be less ideal partly because there can be various testers with different skills. Reliability issues could lead to risk assessments being imprecise. The purpose of the present study was to evaluate the inter- and intra-tester reliability among testers using a novel inertial sensors system to measure spinal postures in healthy persons when they are seated in different postures. Our hypothesis was that there would be no difference between testers or within a tester. 2. Methods 2.1. Participants Four testers were involved in the inter-tester reliability test: one physiotherapist (two years of clinical experience), two physiotherapy students, and one biomedical engineer (experience with clinical measurements), each with different knowledge about clinical anatomy. They were chosen to mimic the real-life conditions where various skills among practitioners could affect the outcome when they use inertial sensor units (ISU:s) measuring
733
postures. All testers had knowledge about the technical features of ISU:s but were inexperienced with postural measurements. Young subjects were chosen to eliminate age-related problems such as degeneration and rigidity of the spine. The subjects were excluded according to a health-screening protocol if they reported any neurological conditions or reduced ability to work during the last €m et al., 12 months because of back or neck problems (Lundstro 2004). The subjects of height below 160 cm were excluded because of the risk that the sensors would collide during movements. Ten healthy subjects, two women and eight men, were included in the inter-tester reliability tests, with a mean age (SD) of 28 (6) years, height of 179 (7) cm, and BMI of 23.3 (2.0) kg/m2. For the intra-tester reliability, performed by one and the same physiotherapist as from the inter-reliability test, another ten healthy subjects were recruited, six men and four women, with a mean age of 31 (5) years, height of 176 (9) cm, and BMI of 24.3(3.4) kg/m2. Written informed consent was obtained from each participant, and the Regional Ethical Review Board (No 2012-24-31M) approved all procedures. 2.2. Equipment The Department of Biomedical Engineering and Informatics, University Hospital of Umeå, Sweden, developed the movement capture and analysis system, in this setting consisting of one data collecting unit and five tri-axial inertial sensor units (ISU:s). Each ISU (ADIS 16364 Analog Devices, USA) included three axis gyros and three axis accelerometers, which together detect the threedimensional posture angles and motions of the object. The size of an ISU is L76 W52 H46 mm and the weight is 40 g. The sensors use a local Cartesian coordinate system, which is determined at the time of initial setting. The sensors are connected by cables to a collection unit, which in turn communicates with a laptop via Bluetooth. Customised software (AnyMo, The Department of Biomedical Engineering and Informatics, Umeå University Hospital, Sweden) calculates the real-time orientation of the sensors, using data from the gyros and accelerometers (Ohberg et al., 2013). A stool was adjusted to the level of the posterior knee crease (popliteal height). The stool had no backrest and had a flat wooden surface covered with two layers of 2 mm foam and an anti-slip surface. 2.3. Protocol The study used a repeated-measurement design, test retest with four testers for the inter-tester reliability and one tester for the intra-tester reliability. Following our written study protocol, the tester mounted the ISU:s on the subject, calibrated them, and gave instructions as to the postures that the subject should then take. There was no prior training for this specific procedure. The subjects performed for each tester, as similarly as possible, three tests, two while seated unsupported and one while standing. The reliability within a subject was also analysed to investigate their precision in repeating postures. The standing test, similar to the test by Intolo and colleagues (Intolo et al., 2010), was added because we suspected it might be difficult for the subjects to adopt the seated postures repeatedly with good accuracy. Five ISU:s were used. The position of the lower three units was chosen based on information from a study by Dankaerts and colleagues (Dankaerts et al., 2006), where sensors were taped on the skin over the spinal processes of S2, L3, and T12, because the lower part of the lumbar spine is the most common area for low back pain (Biering-Sorensen, 1983). The two upper units were meant to be positioned on the forehead and over the spinal process of C7, according to Jasiewicz and colleagues (Jasiewicz et al., 2007). The
734
T.C. Stenlund et al. / International Journal of Industrial Ergonomics 44 (2014) 732e738
forehead position remained, but because the sensor size restricted movement, the C7 sensor was placed instead on the spinal process of T2. The placements of the units were palpated while the subjects stood. The units were attached on elastic Velcro straps and on a customised vest, which was chosen because this would probably be the most feasible situation for the upcoming field measurements. The physiotherapist gave each subject short oral information and allowed them to familiarise themselves with the upcoming movements. However, during the actual tests for the inter-tester reliability, the testers were randomised in order and hidden from one another. Thus, each tester individually instructed the subjects and mounted the ISU:s. When the subjects felt confident with the movements, the test began. Between testers, the systems were switched off, demounted by the previous tester, and then remounted and recalibrated by the next tester. Between trials and between testers, the subjects were able to stand or sit freely for up to 5 min to minimise any effects of fatigue. For the intra-tester reliability test, one tester, the physiotherapist, repeated four measurements (with ISU:s remounted and recalibrated between measurement), so that the same number of measurements were carried out as for the inter-tester reliability settings. The measurements took approximately 1.5e2 h per subject and all measurements for a given subject were taken the same day. 2.4. Test procedure During the tests, subjects were seated towards the front of the stool, with approximately 15 cm support from ischial tuberosities to upper thighs. They were facing forward fixing their gaze at a point on the wall at eye height, hips flexed less than 90 , knees flexed and heels under the stool, hands resting lightly on their thighs so that the subjects could achieve the lordosis posture by sufficiently tilting their pelvis (Claus et al., 2009). Analyses of postures were made during the neutral and end range postures (See Fig. 1AeE). The subjects started in a self-selected and unsupported “neutral” posture. Every 5 s a new command with a new posture was adopted going to the maximum lumbar lordosis and, at a new command, back to a neutral posture. This procedure was repeated until both postures were achieved/obtained a total of 5 times. The same procedure was repeated for the maximum lumbar kyphosis posture. The subjects were also asked to stand 10 cm behind the stool, facing forward, feet and shoulders vertically aligned with
straight knees, neutral self-selected spine, and straight arms. Upon command, the subjects flexed their spine to touch a marked line 1 cm from the stool edge with their fingertip, and told to remain that way until a new command was given. Every postural change was initiated upon a verbal command from the tester. 2.5. Data analysis Orientation data (i.e., segment angles) from the five units attached to the forehead and along the spinal processes were processed in Matlab, (version 7.10.0 (R2010a), The MathWorks, Inc., USA). The processed data described the sensors' normal vector relative to the gravitation vector in degrees. The study initialises movement in the sagittal plane. Data were collected at a sample rate of 240 Hz for the sensors. All orientation data was low-pass filtered with a second order Butterworth filter at a cut-off frequency of 10 Hz. The Euler sequence used for all segment angles was XYZ, where X was rotations in the sagittal plane, Y was rotations in the frontal plane and Z was rotations in the transverse plane. A more detailed description of the algorithms used can be € found in the article by Ohberg et al. (Ohberg et al., 2013). A mean value of the segments angles during 0.5 s of a stable phase for every repetition in each posture was collected. All angles were reported for each posture as means and standard deviations. Data are presented as absolute angles, that is, each unit's normal vector in relation to the gravitational vector and not relative to the other. 2.6. Statistical analysis Inter- and intra-tester as well as intra-subject relative reliability was estimated by the intra-class correlation coefficients (ICC). The ICC was calculated with a 95% confidence interval (CI) using model ICC2,1 absolute agreement (Shrout and Fleiss, 1979). A two-way random model was used because tester and subjects were considered random effects. Inter- and intra-tester reliability was calculated from the mean value of 5 repetitions from each specific posture and segment. Included in the inter-tester reliability were those subjects where all measurements for four testers were valid. Included in the intra-tester reliability were those subjects where four measurements for one tester were valid. Intra-subject reliability, calculated from each repetition of a posture, was calculated to act as reference to tester reliability, indicating the individual variability. The strength of the correlation was assessed according
Fig. 1. A-E. Lumbar kyphosis seated: Thoraco-lumbar and lumbar angles kyphosed approaching end of range (Fig. 1A). Neutral posture seated: self-selected spinal posture (Fig. 1B). Lumbar lordosis seated: Thoraco-lumbar and lumbar angles lordosed approaching end of range (Fig. 1C). Neutral posture standing: self-selected spinal posture (Fig. 1D). Lumbar kyphosis standing: Thoraco-lumbar and lumbar angles kyphosed (Fig. 1E).
T.C. Stenlund et al. / International Journal of Industrial Ergonomics 44 (2014) 732e738
to the following criteria: <0.75, Poor to moderate; 0.75e0.90, Good; 0.91-1, Adequate reliability for clinical measurement (Portney and Watkins, 2000). The calculated ICC depends on several variance components which are presented separately to clarify the magnitude of the different sources of variance. s2p represents the variance due to systematic differences in the subjects, s2o represents the variance due to systematic differences in testers (inter-tester reliability) or test-retest (intra-tester and intra-subject), s2residual represent the random error of variance. ICC ¼ s2p =ðs2p þ s2o þ s2residual Þ (de Vet et al., 2011). The Standard Error of Measurement (SEM) and the Smallest Detectable Change (SDC) is calculated for the absolute reliability (agreement) between testers or repeated measurements. SEM represents the measurement error and includes the error variance, systematic and random SEM ¼ √ðs2o þ s2residual Þ (de Vet et al., 2011). pSDC is based on the SEM, and is defined as ffiffiffi SDC ¼ 1*96 2*SEM (de Vet et al., 2011). The SDC provides information about the smallest degree of change between two measurements within a 95% CI. When interpreting the results of accuracy the upper limit of 5 SEM, set by the American Medical Association as an acceptable finding for evaluation of movement in a clinical context, was used (Nitschke et al., 1999). The guidance is set for Range of Motion and here used for absolute angles but still considered as a useful reference. The proportion of differences fewer than 5 between the first and the following measurements or repetition was retrieved to present the agreement in percentage between measurements or repetition (de Winter et al., 2004). Five degrees was chosen because of an assumed clinical significance. Mean values and standard deviations for each posture and segment in respective reliability tests were calculated from every valid value. Analyses were performed using IBM SPSS version 20 (SPSS Inc., Chicago III USA). 3. Results 3.1. Reliability The inter-test reliability was poor to moderate. ICC:s between testers varied between 0.18 and 0.76, SEM varied between 4.1 and 11.5 , and SDC between 9.9 and 31.9 . The number of included tests varied where the head segment was of lowest value. Agreement within 5 varied between 26 and 67%. There was a significant difference between testers at three levels on three occasions; head in neutral posture seated, segment L3 in lumbar lordosis seated and S2 in neutral posture standing. (Table 1). The intra-test reliability proved to be poor to moderate in nine and good in 16 combinations of postures and segments. ICC:s for the same tester varied between 0.37 and 0.90, SEM between 2.5 and 12.0 , and SDC between 7.1 and 33.3 . Number of included tests varied where the head segment was of the lowest value. Agreement within 5 varies between 38 and 81% (Table 2). The intra-subject reliability proved to be adequate in all cases except at level head in neutral posture standing. ICC:s for within a subject varied between 0.81 and 0.98, SEM between 1.3 and 5.1, and SDC between 3.6 and 14.2 . Agreement within 5 varied between 57 and 98% (Table 3). 4. Discussion The purpose of this study was to evaluate if an inertial sensor system may be used reliably by different testers to measure seated spinal postures. The initiated movements and measurements were performed in the sagittal plane. The inter-tester reliability using ISU:s was considered poor to moderate, whereas the intra-tester reliability was considered good in the majority of segments and
735
tests, but not reaching adequate reliability (Portney and Watkins, 2000). The hypothesis was rejected in three cases; at segment head in neutral posture seated, L3 in lumbar lordosis seated and S2 in neutral posture standing, due to a significant difference between testers. Variability within the subjects' test retest situation was rather low. The SEM and SDC for intra-tester reliability are better than for the inter-tester reliability but still considered too high to detect smaller changes. For the intra-tester reliability, the correlation was good in the majority of segments with small variation for ICC:s between the seated lumbar kyphosis and standing kyphosis, but with more differences for seated lumbar lordosis in the three lowest segments. This is in accordance with the study by Claus and colleagues (Claus et al., 2009) showing that a slump posture, named “seated lumbar kyphosis” in our study, is easier to repeat. Further, the ICC:s from intra-tester reliability indicate small variation between the seated lumbar kyphosis and a neutral posture, except for Th12. The possibility to repeat a neutral posture with high reliability in the lowest segment is in accordance with the report of O’Sullivan and colleagues (O'Sullivan et al., 2010) showing that pain-free subjects' own ideal seated posture and a tester perceived neutral posture are not significantly different and may be tested with very high intertester reliability. 4.1. Reliability considerations The inter- and intra-tester reliability for measurement of seated and standing posture shows variation that cannot be explained by the intra-subject reliability. There is a variance in the tested subjects, shown by the standard deviations in the angle outcome, affecting the subject variance component. The difference from the intra-subject reliability to inter-tester reliability is not mainly caused by the subject variance component but by the fact that the error variance components have increased. There are several causes for the variation, such as finding the spinal processes, attachment of the units, and the movements of the Velcro straps and the vest on skin and clothes. Different skills and accuracy among the testers could influence these results i.e. how they identify the specific spinal processes or how tight the Velcro straps were set. The testers were all novices in respect of this specific set-up and training would probably improve accuracy. However, all had the same written information about the task. The ICC:s for the inter-tester test show poor to moderate correlation with more moderate values in the maximal lumbar kyphosis and lordosis, whereas correlations in the neutral posture are poorer. The agreement between testers within 5 is more equally distributed, approximately 25e50%. The absolute reliability is influenced by high residual variances and occasionally systematic error variances why the majority of SEM:s exceeds the 5 limit for acceptability findings (Nitschke et al., 1999). In addition, high SDC:s indicate that small differences can not be detected. The ICC:s for intra-tester reliability show, on the other hand, good correlation in most cases with SEMs below the 5 limit for the majority of segments (Nitschke et al., 1999) and improved SDC:s. Still, in the majority of segments there has to be >10 differences between two measures before a clinical difference is found. Adding more testers and subjects may influence the results as this study involves only one tester measuring ten subjects repeatedly. However, as expected our results indicate, that one tester has better reliability than several testers. Still, the reliability can most likely be improved. According to O’Sullivan and colleagues (O'Sullivan et al., 2011), a better outcome could be expected with attachment on the skin, but it should be noted that their results are based on measurement in ROM and not angular output. One concern in the test procedure was regarding the subject's ability to occupy the same posture repeatedly in the test-retest
736
T.C. Stenlund et al. / International Journal of Industrial Ergonomics 44 (2014) 732e738
Table 1 Inter-tester reliability. Absolute angles presented as means and standard deviations, intra-class correlation coefficients (ICC2,1 absolute agreement), s2p variance component of subjects, s2o variance component of testers, s2residual variance component of random error variance standard error of measurement (SEM), smallest detectable change (SDC), and measurements within 5 (1 ¼ 100%) between 4 measurements made by 4 different testers in a test retest for different postures and placement of inertial sensor units (ISU). Posture
n
Segment
Mean (SD)
ICC2,1 (95% CI)
s2p
s2o
s2residual
SEM
Lumbar kyphosis seated
6 10 10 10 10 4 9 9 9 9 4 9 9 9 9 5 8 9 9 9 5 8 8 9 9
Head Th2 Th12 L3 S2 Head Th2 Th12 L3 S2 Head Th2 Th12 L3 S2 Head Th2 Th12 L3 S2 Head Th2 Th12 L3 S2
106.3 41.5 88.3 98.7 103.6 117.9 64.7 96.9 95.9 93.1 122.6 79.7 104.9 92.9 85.5 114.8 67.0 103.6 91.0 71.9 48.9 10.4 50.9 51.6 46.2
0.68 0.60 0.63 0.67 0.74 0.47 0.49 0.65 0.50 0.73 0.67 0.59 0.54 0.24 0.61 0.65 0.49 0.43 0.49 0.18 0.62 0.63 0.62 0.66 0.76
140.5 87.0 54.1 41.0 86.6 117.6 38.0 31.2 20.9 87.8 132.3 63.8 53.7 12.5 59.2 96.8 28.7 17.5 30.2 9.7 103.7 39.8 53.0 24.3 84.1
0 4.8 9.6 2.9 3.8 0 2.4 1.7 0.6 3.4 0 1.2 6.4 2.9 5.4 0 0 0 9.1 2.5 0 1.3 0.2 0 4.3
64.5 52.4 21.9 17.4 26.4 132.8 37.5 15.3 20.2 29.3 64.5 43.5 40.2 35.8 31.8 51.3 29.8 22.8 21.8 41.1 64.5 23.5 33.0 12.7 22.1
8.0 7.6 5.6 4.5 5.5 11.5 6.3 4.1 4.6 5.7 8.0 6.7 6.8 6.2 6.1 7.2 5.5 4.8 5.6 6.6 8.0 5.0 5.8 3.6 5.1
Neutral posture seated
Lumbar lordosis seated
Neutral posture standing
Lumbar kyphosis standing
(15.1) (11.7) (8.9) (7.6) (10.5) (12.0) (9.5) (7.3) (5.9) (10.3) (13.4) (10.0) (9.4) (6.9) (9.1) (11.1) (7.3) (5.8) (7.3) (7.0) (11.1) (7.9) (8.7) (6.8) (10.8)
(0.20e0.92) (0.31e0.86) (0.32e0.87) (0.39e0.89) (0.49e0.92) (0.02e0.94)a (0.17e0.82) (0.34e0.89) (0.18e0.82) (0.46e0.92) (0.18e0.97) (0.27e0.86) (0.22e0.84) (0.03e0.66)a (0.30e0.87) (0.21e0.95) (0.14e0.84) (0.07e0.81) (0.18e0.82) (0.07e0.61)a (0.18e0.94) (0.29e0.89) (0.28e0.89) (0.35e0.89) (0.50e0.93)
SDC
22.3 21.0 15.6 12.5 15.2 31.9 17.5 11.4 12.6 15.9 22.3 18.5 18.9 17.2 16.9 19.9 15.1 13.2 15.4 18.3 22.3 13.8 16.0 9.9 14.2
<5 0.32 0.30 0.45 0.40 0.28 0.35 0.28 0.45 0.40 0.28 0.41 0.24 0.39 0.26 0.34 0.38 0.31 0.47 0.45 0.32 0.31 0.47 0.67 0.53 0.26
SD, standard deviation; CI, confidence interval; n number of subjects included in the ICC analysis. a
Significant difference between testers.
situation. A way to reduce the risk of too excessive variability was to perform all tests in a single session. Previous studies have used similar seated postures (Claus et al., 2009; O'Sullivan et al., 2010; Caneiro et al., 2010) but also facilitated their posture with verbal and manual guidance. Subjects need facilitation to find an upright sitting posture, and to mimic a posture just by looking at a picture may not be enough (Claus et al., 2009). This suggests that our standardised instructions may be insufficient and that some of the
error is due to the protocol itself. To assess that the standardised seated posture was not problematic to repeat, a standing posture similar to that set by Intolo and colleagues which tested that posture as ROM with an ICC value of 0.90 was added to the protocol (Intolo et al., 2010). The intra-subject reliability shows adequate correlation, except for the head segment in the neutral standing posture, with only a small variation for ICC:s between the seated and standing postures. All tests for a subject were made
Table 2 Intra-tester reliability. Absolute angles presented as means and standard deviations, intra-class correlation coefficients (ICC2,1 absolute agreement), s2p variance component of subjects, s2o variance component of test-retest, s2residual variance component of random error, standard error of measurement (SEM), smallest detectable change (SDC), and measurements within 5 (1 ¼ 100%) between 4 measurements made by one tester in a test retest for different postures and placement of inertial sensor units (ISU). Posture
n
Segment
Mean (SD)
ICC2,1 (95% CI)
s2p
s2o
s2residual
SEM
Lumbar kyphosis seated
8 7 7 8 8 8 8 8 8 8 8 8 8 8 8 4 6 6 6 6 4 6 6 6 6
Head Th2 Th12 L3 S2 Head Th2 Th12 L3 S2 Head Th2 Th12 L3 S2 Head Th2 Th12 L3 S2 Head Th2 Th12 L3 S2
97.5 46.7 91.8 107.3 108.8 110.7 64.1 94.4 104.7 103.5 123.2 80.3 101.1 101.5 97.4 111.6 66.4 101.6 90.8 78.2 37.3 5.7 46.5 55.3 53.1
0.81 0.74 0.82 0.80 0.77 0.84 0.74 0.37 0.73 0.88 0.90 0.90 0.56 0.63 0.89 0.84 0.57 0.49 0.81 0.61 0.84 0.87 0.87 0.86 0.76
108.7 50.8 39.0 56.9 52.4 58.9 26.4 12.7 38.2 104.5 224.1 123.4 64.0 32.1 141.3 38.9 13.8 22.4 120.9 37.9 752.8 78.3 43.0 179.8 81.3
0 0.2 0.9 0 0 0.8 1.4 0 0 0 0 1.0 0 0 0 0 0 0 4.6 2.8 0 1.3 0.2 0 0
25.7 17.9 7.4 14.1 15.4 10.8 7.8 21.2 13.8 15.0 25.7 12.3 50.1 19.2 18.2 7.7 10.3 22.0 23.3 21.6 144.5 10.3 6.3 28.2 25.5
5.1 4.3 2.9 3.8 3.9 3.4 3.0 4.6 3.7 3.9 5.1 3.6 7.1 4.4 4.3 2.8 3.2 4.7 5.3 4.9 12.0 3.4 2.5 5.3 5.0
Neutral posture seated
Lumbar lordosis seated
Neutral posture standing
Lumbar kyphosis standing
(12.5) (8.8) (6.5) (7.9) (7.6) (8.2) (5.6) (5.8) (6.9) (10.0) (14.7) (10.5) (9.6) (6.8) (11.6) (7.7) (4.7) (6.9) (10.5) (8.5) (21.3) (7.6) (9.1) (11.8) (9.5)
(0.55e0.95) (0.43e0.94) (0.57e0.96) (0.54e0.95) (0.49e0.94) (0.62e0.96) (0.45e0.93) (0.02e0.78) (0.44e0.93) (0.69e0.97) (0.74e0.98) (0.75e0.98) (0.20e0.87) (0.29e0.89) (0.71e0.97) (0.48e0.99) (0.15e0.91) (0.06e0.89) (0.52e0.97) (0.23e0.92) (0.44e0.99) (0.65e0.98) (0.64e0.98) (0.61e0.98) (0.41e0.96)
SD, standard deviation; CI, confidence interval; n, number of subjects included in the ICC analysis.
SDC 14.1 11.8 8.0 10.4 10.9 9.4 8.4 12.8 10.3 10.7 14.1 10.1 19.6 12.1 11.8 7.7 8.9 13.0 14.6 13.7 33.3 9.4 7.1 14.7 14.0
<5 0.39 0.56 0.78 0.79 0.64 0.71 0.68 0.68 0.71 0.57 0.57 0.64 0.36 0.61 0.61 0.81 0.60 0.64 0.48 0.60 0.38 0.76 0.80 0.52 0.48
T.C. Stenlund et al. / International Journal of Industrial Ergonomics 44 (2014) 732e738
737
Table 3 Intra-subject reliability. Absolute angles presented as means and standard deviations, intra-class correlation coefficients (ICC2,1 absolute agreement), s2p variance component of subjects, s2o variance component of repetitions, s2residual variance component of random error, standard error of measurement (SEM), smallest detectable change (SDC), and measurements within 5 (1 ¼ 100%) between repetitions within each measurement in a test retest situation in different postures and placement of inertial sensor units (ISU). Posture
n
Segment
Mean (SD)
ICC2,1 (95% CI)
s2p
s2o
s2residual
SEM
Lumbar kyphosis seated
72 76 76 76 73 71 76 76 76 74 71 75 75 75 76 65 73 73 74 74 64 72 76 73 73
Head Th2 Th12 L3 S2 Head Th2 Th12 L3 S2 Head Th2 Th12 L3 S2 Head Th2 Th12 L3 S2 Head Th2 Th12 L3 S2
101.7 44.0 90.0 102.8 106.1 114.1 65.1 95.6 100.5 98.4 120.8 79.9 103.0 97.1 91.3 113.2 66.7 102.6 90.9 74.8 43.4 8.1 48.7 53.4 49.5
0.93 0.95 0.96 0.97 0.96 0.91 0.92 0.95 0.97 0.98 0.98 0.97 0.95 0.97 0.98 0.81 0.91 0.96 0.96 0.96 0.92 0.96 0.96 0.95 0.94
207.6 114.0 58.3 74.4 90.8 104.5 52.5 39.9 59.0 126.3 392.5 101.0 93.9 65.7 142.8 91.9 37.4 40.2 78.4 69.4 310.3 64.0 72.5 92.3 115.3
0.2 0.4 0.1 0.1 0.2 0 0.5 0.1 0.4 0.5 0 0 0.2 0.1 0.4 1.9 0.2 0.1 0.3 0.3 6.1 0.3 0.3 0.8 1.6
15.1 6.3 2.4 2.4 3.2 9.9 3.9 2.2 1.7 2.7 8.3 3.0 4.9 2.1 2.0 20.4 3.6 1.6 3.0 2.5 20.1 2.4 2.9 4.1 5.5
3.9 2.9 1.6 1.6 1.8 3.1 2.1 1.5 1.4 1.8 2.9 1.7 2.3 1.5 1.5 4.7 1.9 1.3 1.8 1.7 5.1 1.6 1.8 2.2 2.7
Neutral posture seated
Lumbar lordosis seated
Neutral posture standing
Lumbar kyphosis standing
(14.8) (8.0) (8.0) (8.8) (9.6) (10.6) (7.5) (6.5) (7.8) (11.3) (15.5) (10.3) (9.8) (8.1) (12.0) (10.6) (6.4) (6.4) (8.9) (8.5) (18.1) (8.1) (9.1) (9.8) (10.9)
(0.91e0.95) (0.92e0.96) (0.94e0.97) (0.95e0.98) (0.95e0.98) (0.88e0.94) (0.89e0.95) (0.93e0.96) (0.95e0.98) (0.96e0.98) (0.97e0.99) (0.96e0.98) (0.93e0.97) (0.95e0.98) (0.97e0.99) (0.73e0.87) (0.87e0.94) (0.94e0.97) (0.94e0.97) (0.94e0.97) (0.88e0.95) (0.94e0.97) (0.94e0.97) (0.92e0.97) (0.91e0.96)
SDC 10.8 7.2 4.4 4.4 5.1 8.7 5.8 4.2 4.0 5.0 8.0 4.8 6.3 4.1 4.3 13.1 5.4 3.6 5.0 4.6 14.2 4.6 5.0 6.1 7.4
<5 0.73 0.90 0.93 0.92 0.93 0.76 0.80 0.95 0.93 0.91 0.71 0.91 0.86 0.95 0.92 0.67 0.92 0.98 0.90 0.93 0.57 0.93 0.93 0.89 0.77
SD, standard deviation; CI, confidence interval; n number of measurements included in the ICC analysis.
on the same day. Probably, a less ideal ICC value would be the outcome if the intra-subject reliability were to be repeated one day later. 4.2. Methodological considerations Firstly, there were different subjects within inter- and intratester reliability tests because those tests were made on separate occasions. We attempted to match subjects as much as possible but there exists a small difference in age, height, BMI and gender. When comparing the outcome for each segment from the two groups, a small difference is noticed for the two lowest segments, L3 and S2. Gender e women have more lumbar lordosis (Norton et al., 2004) and BMI, found to give regional differences in different sitting postures (Mitchell et al., 2008) e could influence the regional differences. Still, we cannot conclude whether this is because of the subjects, the tester, or a combination of both. Secondly, the number of subjects is similar or fewer than in other studies. On the other hand, the number of testers and repetitions is higher compared to the same studies (Theobald et al., 2012; O'Sullivan et al., 2011; O'Sullivan et al., 2010). Including more testers for the test-retest would probably result in improved intra-tester reliability. Nevertheless, the results show that one and the same tester is superior compared to combining several testers. Some technical and operating errors led to some loss of data that could have influenced the results. However, the achieved results were quite consistent and we believe that the valid measures are representative. Thirdly, the results are described as absolute angular outputs instead of a relative angular outputs (Wong and Wong, 2008) or percentage of ROM (Jasiewicz et al., 2007; O'Sullivan et al., 2011). A relative angle could contribute more knowledge of two segments' relative movement to each other but could lack information if both segments move in the same direction. In contrast, such movement could be detected with absolute angle output. Information about ROM gives good clinical information about movement but less information about actual posture. Technically, the ISU:s would be
able to describe the movement and the position in several dimensions but in this study the data was reduced to describe the gyros absolute angular output in predefined postures. To not add more dimensions was a choice due to the fact that seated postures earlier have been described in the sagittal plane but also that adding more dimensions may increase the risk of errors. However, the results strongly depend on the placements of the ISU:s and therefore are a reflection of the testers' skill and experimental accuracy, which was evaluated here. Fourthly, the attachment of the units seems to be of high importance. In this study, we attached the units to an elastic Velcro strap which was placed on the participant's thin clothing. This created a shield between the spine and units. Other studies have used only adhesive tape directly onto the skin (Theobald et al., 2012) or a combination of tape and straps (Wong and Wong, 2008; Ha et al., 2013). The major reason to use a Velcro strap and a vest is that future field measurements at a workplace would probably not be possible without subjects wearing their working clothes. Fifthly, the sensor size in this setting is quite large compared to some other inertial sensors (Wong and Wong, 2008) or other techniques like strain gauge (O'Sullivan et al., 2011).The size is important because a bigger sensor tends to cover several segments in the spinal column. There is also a greater risk that large units could collide when placed near each other and could tend to float on the belly muscle instead of being in contact with the spinal processes. Smaller units have already been developed, so future measurements should be improved. 4.3. Practical implications As the inter-tester reliability was low it suggests that performing repeated measurements of seated postures with ISU:s in the field should preferably be carried out by one and the same assessor. The reliability would probably be even poorer during real life working conditions, due to high variability within subjects when seated without instructions. Another difference from real conditions is the possibility of using backrests. We chose not to use a backrest as
738
T.C. Stenlund et al. / International Journal of Industrial Ergonomics 44 (2014) 732e738
mechanical contact would lead to artefacts from the ISU:s. Further, the size and placement of ISU:s would prevent the use of a backrest. In future field measurements, it seems necessary to find alternative ISU placements - perhaps on the frontal side of the trunk or to use fewer but representative ISU:s. 4.4. Conclusion This study shows that measuring unsupported seated spinal postures with inertial sensors could be performed with higher reliability if done by the same rather than different testers. The reason for this difference is mainly due to the individual tester's level of skill and experience to assess postures. The reliability for a single tester is mostly good, but still insufficient, to distinguish small differences in posture. The reliability figures will most likely been higher with access to smaller inertial sensors that also would have facilitated unit attachment, particularly over the spine. Acknowledgements This study was made possible through the foundation created by from Centre for Environmental Research in Umeå, AFA Insurance, and the Swedish Agency for Economic and Regional Growth e Goal 2 (Project: 158715 CMTF 2 e AnyMo). The authors also would like to acknowledge funding support from the Inga Britt & Arne Lundberg Research Foundation. The authors are grateful to physiotherapy students Frans Unnemyr and Alex Nassiri for their contribution to the data collection. References Ariens, G.A., van Mechelen, W., Bongers, P.M., Bouter, L.M., van der Wal, G., 2000. Physical risk factors for neck pain. Scand. J. Work Environ. Health 26, 7e19. Ariens, G.A., Bongers, P.M., Douwes, M., Miedema, M.C., Hoogendoorn, W.E., van der Wal, G., Bouter, L.M., van Mechelen, W., 2001. Are neck flexion, neck rotation, and sitting at work risk factors for neck pain? Results of a prospective cohort study. Occup. Environ. Med. 58, 200e207. Bakker, E.W., Verhagen, A.P., van Trijffel, E., Lucas, C., Koes, B.W., 2009. Spinal mechanical load as a risk factor for low back pain: a systematic review of prospective cohort studies. Spine (Phila Pa 1976) 34, E281eE293. Barrero, M., Hedge, A., 2002. Computer environments for children: a review of design issues. Work 18, 227e237. Biering-Sorensen, F., 1983. A prospective study of low back pain in a general population. II. Location, character, aggravating and relieving factors. Scand. J. Rehabil. Med. 15, 81e88. Caneiro, J.P., O'Sullivan, P., Burnett, A., Barach, A., O'Neil, D., Tveit, O., Olafsdottir, K., 2010. The influence of different sitting postures on head/neck posture and muscle activity. Man. Ther. 15, 54e60. Chen, S.M., Liu, M.F., Cook, J., Bass, S., Lo, S.K., 2009. Sedentary lifestyle as a risk factor for low back pain: a systematic review. Int. Arch. Occup. Environ. Health 82, 797e806. Chiu, T.T., Ku, W.Y., Lee, M.H., Sum, W.K., Wan, M.P., Wong, C.Y., Yuen, C.K., 2002. A study on the prevalence of and risk factors for neck pain among university academic staff in Hong Kong. J. Occup. Rehabil. 12, 77e91. Claus, A.P., Hides, J.A., Moseley, G.L., Hodges, P.W., 2009. Is 'ideal' sitting posture real? Measurement of spinal curves in four sitting postures. Man Ther. 14, 404e408. Cote, P., van der Velde, G., Cassidy, J.D., Carroll, L.J., Hogg-Johnson, S., Holm, L.W., Carragee, E.J., Haldeman, S., Nordin, M., Hurwitz, E.L., Guzman, J., Peloso, P.M., 2009. The burden and determinants of neck pain in workers: results of the bone and joint decade 2000e2010 task force on neck pain and its associated disorders. J. Manipulative Physiol. Ther. 32, S70eS86. Dankaerts, W., O'Sullivan, P., Burnett, A., Straker, L., 2006. Differences in sitting postures are associated with nonspecific chronic low back pain disorders when patients are subclassified. Spine (Phila Pa 1976) 31, 698e704. , W., van de Winter, A., Heemskerk, M., Terwee, C., Jans, M., Deville Schaardenburg, D.-J., Scholten, R., Bouter, L., 2004. Inter-observer reproducibility of measurements of range of motion in patients with shoulder pain using a digital inclinometer. BMC Musculoskelet. Disord. 5, 1e8. Findlow, A., Goulermas, J.Y., Nester, C., Howard, D., Kenney, L.P., 2008. Predicting lower limb joint kinematics using wearable motion sensors. Gait Posture 28, 120e126. Ha, T.H., Saber-Sheikh, K., Moore, A.P., Jones, M.P., 2013. Measurement of lumbar spine range of movement and coupled motion using inertial sensors e A protocol validity study. Man. Ther. 18, 87e91.
Hoy, D., Brooks, P., Blyth, F., Buchbinder, R., 2010. The epidemiology of low back pain. Best Pract. Res. Clin. Rheumatol. 24, 769e781. Hoy, D.G., Protani, M., De, R., Buchbinder, R., 2010. The epidemiology of neck pain. Best Pract. Res. Clin. Rheumatol. 24, 783e792. Intolo, P., Carman, A.B., Milosavljevic, S., Abbott, J.H., Baxter, G.D., 2010. The spineangel: examining the validity and reliability of a novel clinical device for monitoring trunk motion. Man Ther. 15, 160e166. Jasiewicz, J.M., Treleaven, J., Condie, P., Jull, G., 2007. Wireless orientation sensors: their suitability to measure head movement for neck pain assessment. Man Ther. 12, 380e385. Kamwendo, K., Linton, S.J., Moritz, U., 1991. Neck and shoulder disorders in medical secretaries. Part I. Pain prevalence and risk factors. Scand. J. Rehabil. Med. 23, 127e133. Kwon, B.K., Roffey, D.M., Bishop, P.B., Dagenais, S., Wai, E.K., 2011. Systematic review: occupational physical activity and low back pain. Occup. Med. (Lond) 61, 541e548. Lis, A.M., Black, K.M., Korn, H., Nordin, M., 2007. Association between sitting and occupational LBP. Eur. Spine J. 16, 283e298. €m, R., Burstro €m, L., Hagberg, M., Nilsson, T., 2004. WBV initial questionLundstro naire: Swedish translation. Risks of Occupational Vibration Exposures (VIBRISKS). Final report. Supplement 3 to Annex 12, Appendix 4a. European Commission. Quality of Life and Management of Living Resources Programme. Key Action 4 e Environment and Health. FP5 Project No. QLK4-2002-02650, January 2003 to December 2006. www.vibrisks.soton.ac.uk/. Mayagoitia, R.E., Nene, A.V., Veltink, P.H., 2002. Accelerometer and rate gyroscope measurement of kinematics: an inexpensive alternative to optical motion analysis systems. J. Biomech. 35, 537e542. Mitchell, T., O'Sullivan, P.B., Burnett, A.F., Straker, L., Smith, A., 2008. Regional differences in lumbar spinal posture and the influence of low back pain. BMC Musculoskelet. Disord. 9, 152. Nitschke, J.E., Nattrass, C.L., Disler, P.B., Chou, M.J., Ooi, K.T., 1999. Reliability of the American Medical Association Guides' model for measuring spinal range of motion e Its implication for whole-person impairment rating. Spine 24, 262e268. Norton, B.J., Sahrmann, S.A., Van Dillen, F.L., 2004. Differences in measurements of lumbar curvature related to gender and low back pain. J. Orthop. Sports Phys. Ther. 34, 524e534. O'Sullivan, P.B., Dankaerts, W., Burnett, A.F., Farrell, G.T., Jefford, E., Naylor, C.S., O'Sullivan, K.J., 2006. Effect of different upright sitting postures on spinal-pelvic curvature and trunk muscle activation in a pain-free population. Spine (Phila Pa 1976) 31, E707eE712. O'Sullivan, K., O'Dea, P., Dankaerts, W., O'Sullivan, P., Clifford, A., O'Sullivan, L., 2010. Neutral lumbar spine sitting posture in pain-free subjects. Man. Ther. 15, 557e561. O'Sullivan, K., Galeotti, L., Dankaerts, W., O'Sullivan, L., O'Sullivan, P., 2011. The between-day and inter-rater reliability of a novel wireless system to analyse lumbar spine posture. Ergonomics 54, 82e90. O'Sullivan, K., O'Sullivan, P., O'Sullivan, L., Dankaerts, W., 2012. What do physiotherapists consider to be the best sitting spinal posture? Man Ther. 17, 432e437. O'Sullivan, P., 2005. Diagnosis and classification of chronic low back pain disorders: maladaptive movement and motor control impairments as underlying mechanism. Man. Ther. 10, 242e255. Ohberg, F., Lundstrom, R., Grip, H., 2013. Comparative analysis of different adaptive filters for tracking lower segments of a human body using inertial motion sensors. Meas. Sci. Technol. 24. Plamondon, A., Delisle, A., Larue, C., Brouillette, D., McFadden, D., Desjardins, P., Lariviere, C., 2007. Evaluation of a hybrid system for three-dimensional measurement of trunk posture in motion. Appl. Ergon. 38, 697e712. Portney, L.G., Watkins, M.P., 2000. Foundations of Clinical Research: Application to Practice, second ed. Prentice Hall, Upper Saddle River, N.J. Pynt, J., Higgs, J., Mackey, M., 2001. Seeking the optimal posture of the seated lumbar spine. Physiother. Theory Pract. 17, 5e21. Rackner, A.D., Samuelsson, C., 2012. Statistics Sweden. Living conditions report 126, working hours and work environment 2010e2011. In: Sweden, S. (Ed.), Living Condition Report, Statistics Sweden. Social Welfare Statistics Unit, Stockholm, p. 114. Roffey, D.M., Wai, E.K., Bishop, P., Kwon, B.K., Dagenais, S., 2010. Causal assessment of occupational sitting and low back pain: results of a systematic review. Spine J. 10, 252e261. Scannell, J.P., McGill, S.M., 2003. Lumbar postureeshould it, and can it, be modified? A study of passive tissue stiffness and lumbar position during activities of daily living. Phys. Ther. 83, 907e917. Shrout, P.E., Fleiss, J.L., 1979. Intraclass correlations: uses in assessing rater reliability. Psychol. Bull. 86, 420e428. Theobald, P.S., Jones, M.D., Williams, J.M., 2012. Do inertial sensors represent a viable method to reliably measure cervical spine range of motion? Man. Ther. 17, 92e96. de Vet, H.C.W., Terwee, C.B., Mokkink, L.B., Knol, D.L., 2011. Measurement in Medicine : a Practical Guide. Cambridge University Press, New York, United States of America. Williams, M.M., Hawley, J.A., McKenzie, R.A., van Wijmen, P.M., 1991. A comparison of the effects of two sitting postures on back and referred pain. Spine (Phila Pa 1976) 16, 1185e1191. Wong, W.Y., Wong, M.S., 2008. Trunk posture monitoring with inertial sensors. Eur. Spine J. 17, 743e753. Yue, P., Liu, F., Li, L., 2012. Neck/shoulder pain and low back pain among school teachers in China, prevalence and risk factors. BMC Public Health 12, 789.