Can a physical activity monitor provide a valid measure of arm elevation angle? A study to assess agreement between the SenseWear Mini Armband and the universal goniometer

The primary aims of the current study were: i) to demonstrate proof of concept, that an accelerometer-based physical activity monitor (the SMA) could be used to provide retrospective data on static arm elevation angle, hence an indication of upper limb ROM; and ii) to assess agreement between static arm elevation measures obtained using the SMA and those obtained with a universal goniometer. We found that we could derive static arm elevation measures from SMA data, and indeed there were significant and very strong correlations between measures derived using SMA data and those set with the universal goniometer. The use of a manufacturer-provided algorithm to calculate arm elevation from SMA data, however, resulted in a progressive discrepancy between the two methods of measurement, i.e. the greater the goniometer-set elevation angle, the greater the mean difference between goniometer-set and SMA-derived angles. This progressive discrepancy could be corrected for by adjusting the algorithm used to derive angles from SMA data, resulting in limits of agreement ± 12 degrees (flexion) and ± 13 degrees (abduction) across the full range of arm elevation.

A likely explanation for the progressive discrepancy observed with increasing arm elevation in the current study relates to progressive ‘tilting’ of the SMA relative to gravity and the virtual line joining bony landmarks. It has been observed that both arm flexion, and to a greater degree, arm abduction are associated with adjunct humeral rotation [14]. While not measured in the current study, said humeral rotation might be expected to progressively alter SMA attitude, notwithstanding that surface tissue may not rotate precisely in concert with underlying bony structures.

Similarly, uneven ‘seating’ of the SMA on the contours of each subject’s upper arm may have contributed to SMA tilt, hence to the observed progressive discrepancy. The upper arm only approximates a cylindrical shape; subcutaneous muscle distorts the skin surface. Our experimental set-up was designed to minimise muscle contraction in each static position, but passive lengthening and shortening of opposing upper arm muscle groups as the arm was repositioned in progressively more elevated positions might also have affected SMA attitude. That the relationship between goniometer-set and SMA-derived angles was subject-dependent (for abduction) supports this ‘uneven seating’ hypothesis, as individual subjects differed in their upper arm morphology. Alternative but unlikely explanations for the progressive discrepancy include that the researcher setting arm elevation with the universal goniometer systematically worsened in accuracy as the elevation angle increased, or that the SMA’s accelerometer itself introduced a systematic error as it rotated around its longitudinal axis.

Several previous studies have reported on the agreement between other proprietary accelerometer-based devices and criterion standard devices for the measurement of arm and/or shoulder elevation. Bernmark & Wiktorin measured agreement between a tri-axial accelerometer device (Logger Technology HB) and an optoelectric measuring system across a range of shoulder flexion and abduction positions (0 to 180 degrees), [7] finding ‘almost perfect correspondence’ between measures taken with the two devices. Notably, the accelerometer device was calibrated for each subject prior to testing. Amasay et al. validated a similar tri-axial accelerometer device (the Virtual Corset) against a digital protractor, [15] seating both devices on a vice rather than on a human arm; for measurement of static angles the Virtual Corset differed by up to 14 degrees from the digital protractor, the largest differences occurring at 0 and 180 degrees of elevation.

Kolber et al. and El-Zayat et al. reported on agreement between other, non-accelerometer based digital devices (a digital inclinometer and a three-dimensional gyroscope respectively) and the universal goniometer, for the measurement of end shoulder flexion and abduction ROM in healthy subjects [5,6]. Both studies reported ‘acceptable’ agreement with the universal goniometer (95% limits of agreement: flexion: −20 to 5 degrees, −8 to 9 degrees; abduction: −17 to 14 degrees). The ROM over which agreement was measured in both studies was limited (Kolber et al. [5]: goniometer-measured flexion: mean 156 ± 9 degrees; abduction mean 161 ± 11 degrees; El-Zayat et al. [6]: goniometer-measured flexion: range: 129 to 186 degrees; abduction: 116 to 186 degrees).

A comparative strength of the current study is that we assessed agreement between the SMA and the universal goniometer across the full range of arm elevation angles (arm-by-side to maximal active elevation). We note that in practice, physiotherapists typically are interested in maximal ROM, as an objective descriptor of functional limitation. In the presence of pathology however, arm elevation is often limited, certainly to less than that tested in previous studies. Given the progressive discrepancies observed in the current study, agreement between devices for maximal ROM measurement in healthy subjects might not assure agreement for maximal ROM measurement in patients with pathology. By way of example, if we had only examined data at goniometer-set elevation angles of 45 and 60 degrees in the current study (see Table 1), we would have reported an unjustifiably high level of agreement between the universal goniometer and the SMA.

Are the limits of agreement observed in the current study ‘tight’ enough to commend the use of the SMA as a valid alternative to universal goniometry in the clinical setting? Arguably not, if one considers the universal goniometer itself to be a valid criterion standard measure of arm elevation. A potential variation of 24 to 26 degrees (± 12 to 13 degrees) around the ‘true’ value of arm elevation is similar to that observed for the digital inclinometer and gyroscope, but would likely be perceived by physiotherapists as being unsatisfactory if one considered a 5 degree change in ROM to be clinically significant [1]. As noted by Kolber et al. however, there are limited data on the validity of the universal goniometer against e.g. radiographic measures of movement, [5] – the argument could be drawn that digital devices should rather be considered the criterion standard against which goniometry should be compared.

One clinical benefit of using a physical activity monitor with data recorder, such as the SMA, is that a physiotherapist could measure e.g. daily progression of ROM limitation, without the need for a patient to physically attend the clinic. Patients could flex/abduct the arm to its limit at pre-defined intervals, ‘mark’ SMA data electronically using the timestamp button, and return the SMA to the clinic ex post facto for data download and analysis. An important consideration in such use would be the patient’s capacity to ensure consistent placement of the SMA on the arm. If one could account for dynamic acceleration of the arm, as discussed by Amasay et al. [15] theoretically one could also retrospectively track arm elevation over a period of time, e.g. to assess work-related arm movement demands or patient adherence to physiotherapist-imposed arm movement restrictions. Such clinical benefits might offset the relatively high cost of the SMA (AU$1722 including software at the time of study conduct) compared to the universal goniometer.

A weakness of the current study is that we calculated adjusted SMA-derived arm elevation angles using the same algorithms for all subjects. As described, however, there was a significant effect of subject on differences between the goniometer-set and SMA-derived elevation angles for abduction. Calibration of the SMA for individual patients (i.e. use of patient-specific algorithms) might provide for better agreement with the universal goniometer in the clinical/field setting - albeit if a patient did have limited range, said calibration would be problematic.

A second weakness of the current study is that we did not assess the reliability of arm elevation measures derived from the SMA. As an electronic device, we would argue that any unreliability we might have observed would only be related to the reliability of the researcher/universal goniometer in setting elevation angles (neither did we assess inter-tester reliability of this), and/or alterations in how the SMA was seated on the arm. Nonetheless, further research into the clinical feasibility of the SMA to measure/record arm elevation should include assessment of both inter- and intra-tester reliability.