How do digital range of motion measurement devices ‘measure-up’ to traditional goniometry in assessing shoulder range of motion? A systematic review and meta-analysis

Abstract

Background

Shoulder range of motion (ROM) is traditionally measured using universal goniometry. However, novel devices to measure shoulder ROM digitally are becoming increasingly available. We aimed to synthesise the current evidence to answer: 1) what technologies are currently in use? 2) Are they reliable? 3) How do they compare to goniometry?

Methods

Systematic review of the literature was conducted according to PRISMA guidelines. MEDLINE, Embase, CINAHL, Emcare and Cochrane databases were searched to identify studies comparing a digital device measuring shoulder ROM to goniometry in participants > = 18years. Quality of studies was assessed using COSMIN risk of bias tool. End points included device validity compared to goniometry and intra-rater reliability.

Results

15 articles were included, representing 372 participants and 608 shoulders, and reporting data for five device categories; infrared/RGB-D, 3D-motion-analysis, combined 3D/infra-red, 2D-video-analysis and virtual-reality. Nine studies reported mean bias and 95% limits of agreement (LOA) compared to goniometry. Pooled mean bias was −0.25 degrees (−1.25, 0.75 95% LOA, random effects model) overall. This did not differ by device type (p = 0.83), sensor or non-sensor-based devices (p = 0.62) or plane of movement (p = 0.91).

Conclusions

These devices compare well to goniometry and represent a possible means of increasing efficiency and facilitating telemedicine.

Introduction

Shoulder range of motion in clinical assessment

Assessment of shoulder range of motion (ROM) is an important part of the overall assessment of patients presenting with shoulder pain or stiffness. Reduced ROM is observed in a variety of shoulder pathologies, including rotator cuff tears and arthropathy, adhesive capsulitis, impingement syndromes and glenohumeral osteoarthritis. ¹ In addition to its diagnostic role, assessment of shoulder ROM is also key in assessing functional impact of pathology on patients, with adequate ROM essential for the ability to perform upper-limb associated activities of daily living. ² Traditionally, shoulder ROM is measured using a universal goniometer, however reported reliability of universal goniometry is highly variable and open to influence of human error. ³

Digital measurement of shoulder range of motion

Novel devices are becoming available which present a means of measuring shoulder ROM digitally. A range of technologies are available on the market, including 3D motion analysis, infrared cameras, digital goniometers and smartphone applications.^4,5 The current NHS Long-Term plan emphasises the target for better use of digital technology within healthcare in order to improve access and efficiency. ⁶ Indeed, the use of telemedicine within the NHS increased dramatically following the onset of the COVID-19 pandemic. ⁷ Its use specifically within trauma and orthopaedic surgery has been shown to be successful for diagnostic, pre-operative optimisation, follow-up and rehabilitation consultations. ⁸ It has also been demonstrated to be an efficient means of conducting post-operative follow-up in a randomised controlled trial of orthopaedic patients, demonstrating equal patient satisfaction. ⁹

Novel digital shoulder ROM measurement technologies present potential advantages to both patients and healthcare systems through increased efficiency, lack of human subjectivity and error and facilitation of telemedicine. Studies to date have assessed the overall reliability of these devices, reporting high overall reliability between testers.^4,5

However, to the best of our knowledge, no review to date has compared these devices with traditional goniometry measurement of shoulder ROM in order to truly establish whether their implementation within health services is a viable option compared to current practice. This review therefore aimed to establish:

• – What technologies are in use for digital measurement of shoulder range of motion?
• – Are digital range of motion measurement technologies reliable?
• – How do digital range of motion measurement technologies compare to standard universal goniometry?

Methods

Systematic review

A systematic review was conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA). ¹⁰ The protocol for the review was registered with the international prospective register of systematic reviews (PROSPERO), registration ID CRD42023384174.

MEDLINE, Embase, CINAHL, Emcare and Cochrane central databases were searched from inception to January 2023 using the search strategy (Appendix 1) comprising the following key words: shoulder, shoulder joint, range of motion, goniometry, artificial intelligence, video recording and 3D motion.

Studies were included if they met the following inclusion criteria: 1) study comparing a digital technology for measurement of shoulder range of motion (sensor or non-sensor based) to standard universal (non-digital) goniometry measurement and 2) participants aged 18 years or older. Studies were excluded if they compared digital measurement methods with a measurement technique other than goniometry, with digital goniometry alone without standard non-digital universal goniometry, utilised human visual estimation of joint range of motion, or compared range of motion of joints other than the shoulder.

Titles and abstracts were screened against this inclusion and exclusion criteria by two independent reviewers (J.S. and S.H.), with ineligible or duplicate studies excluded. Where this was unclear from the title and abstract, full-text was reviewed in order to determine eligibility prior to data extraction. Any disagreements between the two reviewers regarding study inclusion were resolved independently by a 3^rd reviewer (P.R.).

Quality assessment

Quality of included studies was assessed independently by two independent reviewers (J.S. and S.H.) using the COSMIN risk of bias tool for assessing quality of studies on reliability and measurement error of outcome measurement instruments. Where there was disagreement, the two reviewers discussed to reach consensus, as advised in the COSMIN risk of bias tool user manual and the “worst-score-counts” principle applied to determine overall quality of the study as “very good”, “adequate”, “doubtful” or “inadequate” as per the tool user manual. ¹¹

Outcome measures & data extraction

Data extracted included study design, year of publication and study population age, male:female ratio and history of shoulder pathology. Primary outcome included device validity compared to universal goniometry, reported as intra-class correlation co-efficient (ICC) and/or mean bias and 95% limits of agreement (LOA). Secondary outcome included device reliability, reported as ICC +/− 95% confidence intervals (CI).

Meta-analysis

Random-effects meta-analysis and subsequent sub-group analysis was performed using Stata 17.0 BE statistical software to calculate pooled mean bias between digital and goniometric measurements and sub-group analysis performed by plane of movement, device type and use of sensors. For the purposes of calculating study weighting, study size was determined based on number of shoulders included in the study. Influence of publication bias was assessed using funnel plot and Egger's test.

Results

Literature search

A total of 2603 studies were identified from the initial search, with 912 duplicates subsequently removed, leaving 1691 studies to be screened. 1659 were excluded based on title and abstract screening and therefore 31 full-text articles were examined for eligibility, in addition to one full-text not available despite request for access. Sixteen were subsequently excluded for reasons specified in Figure 1 , including one study where authors were unable to obtain article translation sufficient to allow adequate data extraction. Cohen's kappa was 0.71 and 0.6 for title & abstract and full-text screening respectively, representing substantial and moderate agreement. ¹² Fifteen studies were therefore included and underwent data extraction.^13–27

Figure 1.

PRISMA flow diagram.

Quality assessment

Study characteristics are displayed in Table 1 , including level of evidence derived from quality assessment as performed by COSMIN risk of bias tool for assessing quality of studies on reliability and measurement error of outcome measurement instruments. No studies were rated as “inadequate”, with eight “very good” and seven “doubtful”, and therefore all were included in further analysis.

Table 1.

Description of study characteristics and quality assessment outcome (as assessed by COSMIN risk of bias tool).

Study ID	Year of Publication	Journal	Country	Study design	Population	History of Shoulder Pathology	Total no. of participants	Total no. of shoulders	Age (mean +/- SD unless otherwise stated)	Sex - M:F	Name of digital range of motion measurement device/technology	Type of device	Type of movement/position measured	Nature of dynamic movement measured:	Planes of movement measured:	Quality Assessment Outcome:
Huber 2015	2015	Physiotherapy	United States	Measurement device reliability/validity study	Healthy volunteers	No known pathology	10	10	22.1 (0.9)	06:04	Microsoft Kinect (infrared-based camera system)	Non-sensor-based	Both	Maximal range of motion	Flexion; Abduction; External rotation	Very Good
Cui 2019	2019	Journal of Healthcare Engineering	China & Taiwan	Measurement device reliability/validity study	Adhesive capsulitis patients	Adhesive capsulitis	25	25	56.25	10:15	3D motion analysis + IMU sensors + Virtual Reality software	Sensor-based	Dynamic movement	Maximal range of motion	Flexion; Abduction; External rotation; Internal rotation	Very Good
Henmi 2006	2006	Modern Rheumatology	Japan	Measurement device reliability/validity study	Healthy volunteers	No known pathology	5	5	23 (range 20–28)	02:03	3D motion capture system (Vicon 512) + optical markers	Sensor-based	Dynamic movement	Maximal range of motion	Flexion	Doubtful
Hawi 2014	2014	Technology and Health Care	Germany	Measurement device reliability/validity study	Healthy volunteers	No known pathology	7	14	Not stated	Not stated	Microsoft Kinect (infrared-based camera system)	Non-sensor-based	Dynamic movement	Maximal range of motion	Abduction; Adduction	Doubtful
Lee 2015	2015	PLOS One	Korea	Measurement device reliability/validity study	Healthy volunteers Adhesive capsulitis patients	Healthy volunteers and adhesive capsulitis patients	27 15 healthy volunteers 12 adhesive capsulitis patients	42 30 healthy volunteers 12 adhesive capsulitis patients	Healthy volunteers: 45.47 (9.05) Adhesive capsulitis patients: 51.5 (9.04)	Healthy volunteers - 8:7 Adhesive capsulitis patients 6:6	Microsoft Kinect (infrared-based camera system)	Non-sensor-based	Dynamic movement	Maximal range of motion	Flexion; Abduction; External rotation	Doubtful
Zulkarnain 2017	2017	Journal of Shoulder and Elbow Surgery	Korea	Measurement device reliability/validity study	Healthy volunteers	No known pathology	10	10	25.8 (4.6)	10:00	3D motion analysis (OptiTrack) + optical markers	Sensor-based	Static position	N/A e.g., static position measured	Flexion; Abduction; External rotation; Internal rotation	Very Good
Rigoni 2019	2019	Sensors	Australia	Measurement device reliability/validity study	Healthy volunteers	No known pathology	30	60	32.8 (range 24–62)	12:18	3D motion analysis + IMU sensors (Biokin)	Sensor-based	Dynamic movement	Maximal range of motion	Flexion; Abduction; External rotation; Internal rotation	Very Good
Cubukcu 2020	2019	Medical Engineering and Physics	Turkey	Measurement device reliability/validity study	Healthy volunteers	No known pathology	40	80	22.08 (3.11)	22:18	Microsoft Kinect (infrared-based camera system)	Non-sensor-based	Dynamic movement	Maximal range of motion	Flexion; Extension; Abduction; External rotation; Internal rotation	Doubtful
Beshara 2020	2020	Sensors	Australia	Measurement device reliability/validity study	Healthy participants	No known pathology	50	100	32.2	23:27	Microsoft Kinect (infrared-based camera system)+ IMU sensors (combined system)	Sensor-based	Both	Maximal range of motion	Flexion; Abduction	Very Good
Bravi 2021	2021	Sensors	Italy	Measurement device reliability/validity study	8 x healthy controls 8 x cervical spinal cord injury patients	No known pathology	16	16 (dominant shoulder)	Healthy controls: 44 (18) CSCI: 50 (12)	***	3D motion analysis (Raspberry Pi gateway) + IMU sensors	Sensor-based	Dynamic movement	Maximal range of motion	Flexion; Abduction; External rotation; Internal rotation	Very Good
Tanioka 2022	2022	The Journal of Medical Investigation	Japan	Measurement device reliability/validity study	Patients with psychiatric disorders	No known pathology	54	108	75.72 (11.3)	20:34	ImageJ 2D video analysis	Non-sensor-based	Dynamic movement	Maximal range of motion	Flexion; Abduction	Doubtful
Chapman 2022	2023	Seminars in Arthroplasty	Canada	Case control study	Reverse total shoulder arthroplasty patients and healthy controls	Cases: (9x rotator cuff arthropathy 1x OA) Healthy controls	20 (10x health controls, 10rTSA patients)	20	rTSA: 82 (5) Controls: 69 (20)	rTSA: 1:9 Controls: 4:6	3D motion analysis + IMU sensors - weekly continuous measurement	Sensor-based	Dynamic movement	Movement during specific activity e.g., ADL/exercise	Flexion; External rotation	Doubtful
Gauci 2022	2022	International Orthopaedics	France	Measurement device reliability/validity study	Healthy volunteers	No known pathology	30	30	33 (7)	20:10	Unspecified RGB-D/infrared camera + Artificial Intelligence algorithm	Non-sensor-based	Static position	N/A e.g., static position measured	Flexion; Extension; Abduction; Adduction; External rotation	Very Good
Rettig 2015	2015	BMC Musculoskeletal Disorders	Germany	Measurement device reliability/validity study	Healthy volunteers	No known pathology	8	8	29.1 (12)	04:04	3D motion analysis (Vicon 612) + optical markers	Sensor-based	Static position	N/A e.g., static position measured	Abduction; External rotation	Doubtful
Carley 2021	2021	Journal of Allied Health	United States	Measurement device reliability/validity study	Healthy volunteers	No known pathology	40	80	28.8	12:28	Virtual Reality device - XRHealth/OculusRift	Other: Virtual reality	Dynamic movement	Maximal range of motion	Flexion; Abduction; Adduction	Very Good

Description of studies & study populations

Included studies present a total of 372 participants and 608 shoulders from which range of motion (ROM) measurements were obtained. Fourteen of the included studies were studies of reliability and measurement error of outcome measurement instruments, with the remaining one study a case-control study design. Mean age of study population ranged from 22 to 82. Study populations consisted of health volunteers (10 studies), patients with conditions other than shoulder pathology (two studies), adhesive capsulitis (one study) and combination of both healthy volunteers and adhesive capsulitis and reverse total shoulder arthroplasty (rTSA) patients (two studies).

ROM measurements were reported for variable planes of movement including; flexion (13 studies), abduction (13 studies), external rotation (10 studies), internal rotation (five studies), adduction (three studies) and extension (two studies). Dynamic movement was recorded in 12 of the included studies, with static position measurements reported in five studies, three of which reported only measurement of static positions. For the purposes of this review, when both dynamic and static data were available, dynamic data was included in analysis.

Digital range of motion measurement technologies in use

Within the included literature, five broad categories of digital ROM measurement technologies are described: infrared/red-green-blue-digital based camera systems (five studies); 3D motion analysis (seven studies); 2D video analysis software (one study); virtual reality (one study) and a combined infrared camera and inertial measurement unit (IMU) sensor system (one study). Overall, devices reported in eight of the included studies were sensor or marker-based, six were non-sensor-based and one used a virtual reality gaming device.

Reliability

Eight articles reported device reliability with overall intra-class correlation co-efficient (ICC) ranging from 0.62 to 0.99. This ranged from 0.62 to 0.99 and 0.71 to 0.99 for infra-red/RGB-D and 3D motion analysis device types, respectively. Five studies reported the concurrent reliability of goniometry measurements obtained, with reported ICC ranging from 0.448 to 0.99.

Validity

Two of the 15 studies did not report validity, comparing devices with goniometry with regards to inter-rater reliability only. Of the 13 studies reporting validity ( Table 2 ), nine studies reported mean bias and 95% LOA. While 10 studies reported ICC, only five of these reported concurrent 95% confidence intervals, standard deviation (SD) or standard error of mean (SEM). One study reported median ROM in degrees and compared between device and goniometry using Wilcoxon signed ranks test which showed no difference (p = 0.35), but did not report mean bias, LOA or ICC.

Table 2.

Validity of digital shoulder range of motion measurement devices compared to universal goniometry, presented as mean difference (degrees), 95% limits of agreement (LoA) and intra-class correlation co-efficient (ICC) reported by included studies, arranged by device type.

Study ID	Population description	Total number of participants	Total number of shoulders measured	Device	Type of measurement device:	Type of movement measured	Planes of movement measured:	ICC (95% CI)	Mean diff. (degrees)	95% LOA (degrees)
Infrared/RGB-D Based Camera Systems
Huber 2015	Healthy volunteers	10	10	Microsoft Kinect	Non-sensor-based	Both	Flexion		Dynamic: -15.9 Static: -16.6	Dynamic: -36.5, 4.7 Static: -52.1, 18.9
							Abduction		Static: -1.5 (N.B. dynamic not measured.)	Static: -7, 4.1 (N.B. dynamic not measured.)
							External rotation		Dynamic: 1.3 N.B. static not measured.	Dynamic: -14.3, 16.9 N.B. static not measured.
Hawi 2014	Healthy volunteers	7	14	Microsoft Kinect	Non-sensor-based	Dynamic movement	Abduction	0.184 (-0.41–0.647)	2.38	-55.46, 60.21
							Adduction	0.471 (-0.053–0.795)	-13.5	-41.07, 14.07
Lee 2015	-Healthy volunteers -Adhesive capsulitis patients	27 15 healthy volunteers 12 adhesive capsulitis patients	42 30 healthy volunteers 12 adhesive capsulitis patients	Microsoft Kinect	Non-sensor-based	Dynamic movement	Flexion	Active: 0.864 Passive: 0.906	Active: -0.12 Passive: -6.86	Active: -37.6, 37.3 Passive: -32.2, 18.5
							Abduction	Active: 0.932 Passive: 0.942	Active: 4.17 Passive: -0.71	Active: -29.6, 37.9 Passive: -30.8, 29.3
							External rotation	Active: 0.925 Passive: 0.965	Active: 1.61 Passive: -8.39	Active: -25.2, 28.4 Passive: -38, 21.2
Cubukcu 2020	Healthy volunteers	40	80	Microsoft Kinect V2	Non-sensor-based	Dynamic movement	Flexion		-2.83	-9.88, 4.23
							Extension		-0.1	-1.69, 1.49
							Abduction		0.33	-4.86, 5.51
							External rotation		-0.5	-7.55, 6.55
							Internal rotation		-6.67	-21.42, 8.07
Gauci 2022	Healthy volunteers	30	30	Unspecified RGB-D camera + Artificial Intelligence algorithm	Non-sensor-based	Static position	Flexion	0.97 (0.95–0.98)	-1.14	-17.6, 15.3
							Extension	0.9 (0.84–0.94)	-2.58	-12.2, 7.08
							Abduction	0.98 (0.96–0.99)	-0.33	-14.6, 13.9
							Adduction	0.91 (0.85–0.95)	-1.13	-11.4, 9.13
							External rotation	0.82 (0.72–0.89)	7.5	-17.9, 32.9
3D Motion Analysis
Optical Markers:
Henmi 2006	Healthy volunteers	5	5	Vicon 512	Sensor-based	Dynamic movement	Flexion	0.94
Rettig 2015	Healthy volunteers	8	8	Vicon 612	Sensor-based	Static position	Abduction	Only reported reliability
							External rotation
Zulkarnain 2017	Healthy volunteers	10	10	OptiTrak	Sensor-based	Static position	Flexion	0.965	0.74	-8.7, 10.2
							Abduction	0.993	6.1	-0.1, 12.3
							External rotation	0.898	4.8	-4.6, 14.2
							Internal rotation	0.712	-5.3	-14.7, 4.1
IMU Sensors:
Cui 2019	Adhesive capsulitis patients	25	25	IMU sensors + Virtual Reality software	Sensor-based	Dynamic movement	Flexion	0.997	*Wilcoxon signed-ranks test
							Abduction	0.978
							External rotation	0.897
							Internal rotation	0.984
Rigoni 2019	Healthy volunteers	30	60	IMU (Biokin)	Sensor-based	Dynamic movement	Flexion	0.99 (0.99–0.99)	0	-3.2, 3.2
							Abduction	0.99 (0.99–0.99)	-0.8	-4.5, 2.9
							External rotation	0.99 (0.99–0.99)	-0.3	-3.3, 2.7
							Internal rotation	0.99 (0.99–0.99)	-0.9	-4.2, 2.4
Bravi 2021	-Healthy controls -Cervical spinal cord injury (CSCI) patients	16	16	Raspberry Pi gateway + IMU sensors	Sensor-based	Dynamic movement	Flexion	Overall: 0.86 (0.75–0.92) CSCI: 0.86 (0.65–0.93) Healthy: 0.61 (0.31–0.78)	Overall: 6 CSCI: 7 Healthy: 4	Overall: -19, 30 CSCI: -17, 32 Healthy: -20, 28
							Abduction	Overall: 0.95 (0.91–0.97) CSCI: 0.95 (0.91–0.97) Healthy: 0.87 (0.53–0.95)	Overall: -3 CSCI: -1 Healthy: -6	Overall: -19, 13 CSCI: -18, 15 Healthy: -20, 9
							External rotation	Overall: 0.94 (0.91–0.96) CSCI: 0.88 (0.78–0.93) Healthy: 0.97 (0.94–0.98)	Overall: 1 CSCI: 1 Healthy: 1	Overall: -15, 16 CSCI: -19, 20 Healthy: -9, 11
							Internal rotation	Overall: 0.9 (0.81–0.94) CSCI: 0.9 (0.55–0.96) Healthy: 0.89 (0.81–0.94)	Overall: -3 CSCI: -7 Healthy: 0	Overall: -19, 12 CSCI: -22, 8 Healthy: -14, 13
Chapman 2022	-Healthy controls -Reverse total shoulder arthroplasty patients	20 -10 healthy controls -10 rTSA patients	20	IMU sensors - weekly continuous measurement	Sensor-based	Dynamic movement	Flexion	Average ROM: 0.02 Max ROM: 0.14
							External rotation	Average ROM: 0.08 Max ROM: 0.66
Combined Infrared + IMU Sensors
Beshara 2020	Healthy volunteers	50	100	Microsoft Kinect (HumanTrak) + IMU sensors (coupled system)	Sensor-based	Both	Flexion	Dynamic: 0.84 (0.72–0.87) Static: 0.98 (0.97–0.99)	Dynamic: 2.05 Static: 0	Dynamic: -11.3, 15.4 Static: -4.8, 4.8
							Abduction	Dynamic: 0.59 (0.6–0.82) Static: 0.91 (0.84–0.94)	Dynamic: 3.05 Static: 2.18	Dynamic: -15.6, 21.7 Static: -7.8, 12.1
2D Video Analysis Software
Tanioka 2022	Patients with psychiatric disorders	54	108	ImageJ 2D video analysis	Non-sensor-based	Dynamic movement	Flexion	*medians and Wilcoxon signed ranks reported. Goniometer	Goniometer: median = 120, 120 (L, R) degrees	ImageJ median = 127, 120 (L, R) p = 0.35, p = 0.75 (L, R)
							Abduction		Goniometer: median = 135, 140 (L, R)	ImageJ median = 121, 130 (L, R) p < 0.001, p < 0.01 (L, R)
Virtual Reality
Carley 2021	Healthy volunteers	40	80	Virtual Reality - XRHealth/OculusRift VR device	Other: Virtual reality	Dynamic movement	Flexion	Only reported reliability
							Abduction
							Adduction

Pooled mean bias was evaluated in order to maximise inclusion, given that only five studies reported ICC with either 95% CI or data sufficient to calculate 95% CI or SEM required for inclusion in meta-analysis. The nine studies reporting mean bias and 95% LOA were therefore included in the meta-analysis to determine overall validity of the reported digital ROM measurement devices compared to goniometry. Pooled mean bias was −0.25 degrees overall (95% LOA −1.25, 0.75, random effects model, I²= 0).

Mean bias did not differ by device type (Infrared/RGB-D: −0.45, 95% LOA −1.8, 0.89; 3D Motion Analysis: −0.05, 95% LOA −1.55, 1.44; Combined Infrared + IMU: 2.39, 95% LOA −8.47, 13.24; p = 0.83), sensor or non-sensor-based devices (Sensor: −0.01, 95% LOA −1.49, 1.47; Non-sensor: −0.45, 95% LOA −1.8, 0.89; p = 0.66) or plane of movement (Abduction: 0.35, 95% LOA −2, 2.7; Adduction: −2.64, 95% LOA −12.26, 6.96; External Rotation: 0.22, 95% LOA −2.33, 2.77; Extension: −0.17, 95% LOA −1.73, 1.4; Flexion: −0.48, 95% LOA −3.13, 2.15; Internal Rotation: −1.66, 95% LOA −4.65, 1.33; p = 0.91) - (Figures 2, 3, 4 respectively).

Figure 2.

Meta-analysis forest plot displaying sub-group analysis evaluating mean bias by device type.

Figure 3.

Meta-analysis forest plot displaying sub-group analysis evaluating mean bias by sensor vs non-sensor-based devices.

Figure 4.

Meta-analysis forest plot displaying sub-group analysis evaluating mean bias by plane of movement.

Publication bias

The funnel plot produced ( Figure 5 ), depicting effect size plotted against standard error for each study, showed a symmetrical distribution, suggesting no evidence of publication bias. This was further confirmed by Egger's test which showed no evidence of publication bias (p = 0.7497).

Figure 5.

Publication bias funnel plot presented as effect size (x-axis) plotted against standard error (y-axis) for each individual study (blue circle).

Discussion

Reliability & validity of digital shoulder range of motion measurement technologies

Our findings show a difference of less than one degree between measurements obtained by traditional goniometry methods or digitally using novel devices. Even within the overall 95% Limits of Agreement of −1.25 to 0.75 degrees, this would not represent a clinically relevant mean difference between measurement of ROM digitally compared to goniometry. Furthermore, this was the case independent of device type, plane of movement and utilisation of sensors in measurement. Thus, it could be inferred that measurements obtained using these devices can be utilised in the same way as that goniometric measurements of shoulder ROM and that this is not limited by the above factors.

Furthermore, the reported reliability of devices within included studies of 0.62–0.99 ICC compares favourably to that of goniometry, and may even be superior with regards to reliability. ³ However, direct comparison of reliability between goniometry and digital devices is only reported in 5 of the included studies and this should therefore be interpreted with caution.

Strengths and limitations

Previous reviews have examined the reliability of infrared and sensor-based tracking systems through inter-rater reliability, reflecting the ability of these devices to measure consistently over time and between users. However, to the best of our knowledge, this review is the first to assess validity of these devices, comparing these systems to clinical goniometry measurements. We present the collated evidence on this through a systematic search strategy and data extraction, adhering to PRISMA guidelines, ¹⁰ and use meta-analysis to quantitatively synthesise the data to produce pooled mean bias between digital and goniometric measurement of shoulder ROM.

However, the technology market is fast-paced and ever-changing and this review provides evidence of only those devices reported to date. We have sought to draw more generalisable conclusions based on broad categories of device type rather than specific device by manufacturer to draw more meaningful generalisable results. There is, however, inevitable heterogeneity introduced by the differing devices available on the market and we therefore present a broad overview of validity of device types as a whole, accepting this potential limitation.

A random effects model was chosen for meta-analysis due to the known clinical heterogeneity in terms of population age and shoulder pathology history, device type, and planes of movement measured. Despite this, I² value for this model was 0. Given the low heterogeneity reflected by this, sensitivity analysis was conducted by employing a fixed effects model, in which the same results were observed. Thus, random effects model was deemed appropriate given the known stated clinical heterogeneity and the knowledge that I² value can be biased in analyses which include a small number of studies. ²⁸

Several studies within the review were rated as “doubtful” in the quality assessment. On further inspection, this was predominantly due to lack of blinding of assessors and the use of a single “measurer”. However, no studies were assessed as “inadequate” and therefore all studies were included in the analysis. However, as any meta-analysis is limited by the quality of its included studies, this should still be taken into account when drawing conclusions. Furthermore, this does emphasise the importance of robust study design when conducting studies evaluating measurement devices, particularly with regards to blinding of assessors to other measurements and as is also suggested in previous studies, ⁴ this should be considered in future studies evaluating these, or similar, devices.

The heterogenous study population of studies included in this review includes predominantly healthy volunteers as study participants. While some data on a small number of shoulder pathologies, including adhesive capsulitis and reverse total shoulder arthroplasty patients, is included in this review, these devices do remain predominantly unvalidated in the specific target population of patients with shoulder pathologies.

Potential applications

The findings of this review suggest that these novel devices represent a viable option for digital measurement of shoulder ROM, regardless of plane of movement required.

This is important as, due to the costly nature of many of these digital systems, particularly gold standard optical motion capture technology, ²⁹ clinical goniometry is still the standard practice. However, as novel technologies at lower costs are becoming available on the market, this represents a potential application in telemedicine, the use of which has increased dramatically since 2020, ⁸ by facilitating accurate assessment of shoulder ROM remotely. In order to determine the viability of their use in this context it is important to establish their validity compared to standard goniometric measurements. This is supported by the findings of this review.

Furthermore, our findings suggest infrared RGB-D cameras are not inferior to 3D motion analysis when compared to goniometry. This presents a key potential benefit in the application of digital measurement of shoulder ROM as, unlike 3D motion analysis systems, infrared RGB-D camera systems use infrared to track joint position in space, ³⁰ negating the need for the application of physical sensors onto the patient prior to recording the movement, as patient movement is detected by the camera independently. Clearly, this presents a more practically feasible option, particularly in remote consultations. Thus, the use of these devices to facilitate telemedicine and increased efficiency is feasible. This is in keeping with the Key Priorities set out by the National Institute for Health and Care Research (NIHR) which emphasise the importance of engagement with digital health and integration of novel technologies in the UK ³¹ and the NHS Long-Term plan which dictates use of digital technology within healthcare in order to improve services, access and efficiency.