Self-recordings of upper arm elevation during cleaning – comparison between analyses using a simplified reference posture and a standard reference posture

This was a field study including two parts. In part one (self-recordings), workers received a protocol with instructions on how to attach a triaxial accelerometer (GC inclinometer) to the upper arm, and how to adopt a reference posture. It was adopted by the cleaners themselves, without the need of extra material, and referred to as the simplified reference posture. A researcher then instructed each of them to adopt the standard reference posture. The workers wore the GC inclinometer continuously, both day and night, for 2 or 3 days. They repeated the simplified reference posture each morning and noted starting and stopping times of work and lunch breaks for each day in a provided form.

In part two (researchers’ recordings), which was conducted on different days than the self-recordings, the researchers attached the GC inclinometer to the worker’s right upper arm and instructed each subject to perform the standard reference posture. The researchers followed each worker during the one-day recording and noted exact starting and stopping times for work and breaks.

Triaxial accelerometers with an integrated data logger (USB Accelerometer Model X16-mini, Gulf Coast Data Concepts, LLC, Waveland, MS, USA, “GC inclinometer”) with a sampling frequency of 25 Hz were used. This frequency is sufficient as it has been shown that 99.5% of the signal power for wrist (and it is not expected to be higher for the upper arms) was contained in the 0–5 Hz band in occupational repetitive work [26]. The size was 5 × 2.4 × 1.3 cm and they contained a 2 GB memory for data logging, a female micro USB-connector and a rechargeable battery [14]. The accelerometer was attached to the upper arm, just below the insertion of the deltoid muscle, with double-sided adhesive tape and fixed with plastic film (Tegaderm™, 3 M Health Care, St Paul, MN, USA) to secure them from falling off.

The data were processed with the EMINGO software suite, developed by the Division of Occupational and Environmental Medicine in Lund, Sweden using MATLAB (version 2016b, Math Works INC., Natick, MA, USA). The data were resampled at 20 Hz, anti-aliased, low-pass filtered (5 Hz), and visually inspected.

No subjects were excluded due to an incorrect placement of the GC inclinometer. Nevertheless, we improved the protocol during the study. These changes appeared to make it easier for the subjects to follow, as the help needed decreased with improved versions of the protocol. An additional change (version 4) was made after the analyses, with instructions not to replace the GC inclinometer if it falls off.

The protocol includes three parts (see Additional file 1):

Concerning self-recordings, the median upper arm velocity was higher in hotel housekeeping than in hotel housekeeping+ (82 vs 63 °/s; Table 5). There were no differences between self-recordings and researchers’ recordings of hotel housekeeping (Table 5).

Five individuals participated in both the researchers’ recordings and the self-recordings. The 90^th percentile of upper arm elevation and the median generalised angular velocity for these individuals are shown in Figure 5. For them, the group mean difference for the 90^th percentile of upper arm elevation between the researchers’ recording on 1 day and the self-recording on several days, using the standard reference, was 1° (range -2 – 8°). The group mean difference for the upper arm velocity was -7 °/s (range -21 °/s – 2 °/s).

The subjects’ perceptions are reported in Table 7. One subject answered “Bad” to one of the questions. All other answers were positive. Additionally, 87% of the subjects stated that the GC inclinometer had not interfered during work or leisure time during the three-day recording, and 96% were willing to wear the GC inclinometer again.

For recordings of arm elevations, it has been suggested that it is sufficient to attach an inclinometer with one of its axes aligned with the upper arm (humerus) without adopting a reference posture [15, 32]. However, since the humerus may not be parallel to the line of gravity, for example in subjects with voluminous upper arms (strong or obese), we believe that it is important to perform a reference posture to define 0° inclination. When using the standard reference posture the arm hangs out from the body (see Fig. 1a). Thus, this should be a minor problem. In the simplified reference posture the arm is closer to the body (see Fig. 1b). We therefore plotted the difference between the two reference postures from day 1 (Fig. 4) versus BMI. We saw no correlation, and do not suspect a major influence of BMI.

For each individual, the difference in upper arm elevation during work between the two analyses was lower than the difference between the two references, and may be explained by the triangle inequality (see Fig. 7). The distance between the two reference points can be seen as the length of one side of a triangle (a). The distance between one of the reference points and a specific elevation point during work can then be seen as the length of a second side of the triangle (b), while the distance between the other reference point and the same specific elevation point can be seen as the length of the third side of the triangle (c). Thus, as the length of one side in a triangle is less than the difference (∆) of the lengths of the two other sides, the difference between the two reference analyses will be less than the difference between the two references (∆ = b – c < a). For an elevation point that is equally far from the two reference points, the triangle becomes isosceles and the difference between the two reference analyses will be zero (∆ = b – c = 0). If the elevation point is in line with the two references, the triangle becomes a line and the difference between the two reference analyses will be the same as the difference between the two references (∆ = b – c = a). In this study, the difference during work was never more than 10°, while the difference between the two references was up to 21°. In addition, the group mean difference during work was as low as 0.2° (range -7 – 10°). We therefore consider, on group level, the simplified reference posture sufficient for recording of elevations of the upper arm, given that it is the same work tasks and a low degree of freedom in work performance for all individuals [33]. In the current group of cleaners, the simplified reference posture deviated from the standard reference in a uniform pattern (i.e. in all directions, see Fig. 4). Consequently, deviations during work were balanced on group level. However, this may not be the case in other populations. A non-uniform deviation pattern will introduce a systematic error. Concerning upper arm velocity, the self-recording method can be used on individual level, as this measure is not dependent on the reference.

According to the questionnaire which the subjects answered after the study, all but one of the subjects were positive to self-recordings of upper arm elevations and velocities. Only one person answered “Bad” to the question “How did you experience to put on more plastic film?” Since eight subjects reported that it had not been necessary, we think this negative answer was due to language barriers, and this subject also meant that it had not been necessary.

Our research group has performed technical measurements of upper arm elevations and velocities for about thirty years in about sixty different occupations. Most of these occupational groups have also been clinically examined using the standardised Health Surveillance in Adverse Ergonomics Conditions (HECO) method [35, 36] which quantifies the prevalence of WMSDs and diagnoses of the neck and upper extremities. Exposure-response relationships were obtained by compiling the data from the technical measurements and the clinical examinations, and we found strong associations between upper arm velocity and several diagnoses [2]. Based on this knowledge, we have recently proposed action levels for the prevention of WMSDs. The proposed action level for the median generalised angular velocity is 60 °/s [37]. This is well in line with the findings in a recent study by Dalbøge et al., where it was indicated that a median generalised angular velocity of the upper arm below 45 °/s was safe [38]. Based on previous studies [2, 39‐42], we have proposed an action level of 60° for the 90^th percentile of upper arm elevation. The action level for elevation was exceeded in office cleaning, while the action levels for both elevation and angular velocity were exceeded in hotel housekeeping (both self-recordings and researchers’ recordings) and hotel housekeeping+, indicating the need for preventive actions. Hence, it was highly relevant to test the self-recording method among cleaners.

To the best of our knowledge, this is the first time self-recordings have been made of upper arm elevation and velocity. This required a protocol explaining how to perform the self-recording. A strength of the study was that the protocol was tested and improved in an occupation with a high proportion of immigrants. Even if the subjects spoke poor Swedish and English, they were able to perform self-recordings. This indicates that the protocol is easy to follow and may be used by most employees. A weakness is that we did not improve the protocol systematically and did not evaluate the different steps of improvements in a systematic manner. Instead, we made changes in the protocol based on how comfortable and secure the subjects appeared to be when they attached the GC inclinometer and performed the simplified reference posture. On a visual inspection of Fig. 2 we did not see any improvement concerning the individual differences between the two analyses. Thus, we do not think that different versions of the protocol impacted on our data.

Considering recordings of upper arm elevation, we judge a difference of 5° to be clinically relevant. Prior to the study we did not know the distribution of the differences between the analyses with the two different reference postures. As this was about 5° for both the 50th and the 90th percentiles we would have needed 11 subjects to be able to detect a 5° difference between the two analyses with an 80% power. As 28 cleaners were included, we could detect a difference of 3°.