Technologies for the Instrumental Evaluation of Physical Function in Persons Affected by Chronic Obstructive Pulmonary Disease: A Systematic Review

Several systems, sensors, and devices are now available for the instrumental evaluation of physical function in persons with Chronic Obstructive Pulmonary Disease (COPD). We aimed to systematically review the literature about such technologies. The literature search was conducted in all major scientific databases, including articles published between January 2001 and April 2022. Studies reporting measures derived from the instrumental assessment of physical function in individuals with COPD were included and were divided into application and validation studies. The quality of validation studies was assessed with the Consensus-based Standards for the selection of health Measurement Instruments (COSMIN) risk of bias tool. From 8752 articles retrieved, 21 application and 4 validation studies were included in the systematic review. Most application studies employed accelerometers, gait analysis systems, instrumented mattresses, or force plates to evaluate walking. Surface electro-myography or near-infrared spectroscopy were employed in four studies. Validation studies were heterogeneous and presented a risk of bias ranging from inadequate to doubtful. A variety of data regarding physical function can be retrieved from technologies used in COPD studies. However, a general lack of standardization and limitations in study design and sample size hinder the implementation of the instrumental evaluation of function in clinical practice.


Introduction
Chronic obstructive pulmonary disease (COPD) is a respiratory condition characterized by a high prevalence and a strong association with disability and mortality; it has been estimated that almost 10% of the worldwide population [1] is affected by this disease and that, in high-income countries, COPD is the fourth leading cause of death [2].
The evaluation of physical performance and function, by investigating-for examplethe reduction in self-walked distance, is pivotal to the clinical management of persons affected by COPD, as suggested by current guidelines [3]. Several mechanisms may impact the physical performance of persons with COPD, for example, impaired ventilatory mechanics (such as dynamic hyperinflation), modification of the ventilation-perfusion relationship and hypoxemia, pulmonary hypertension, and other cardiovascular factors. The 6 min walking test (6MWT) is currently implemented to measure the impact of the COPD on physical performance. However, physical function tests may also help to gain important information about the risk of exacerbation and the impact of the disease on several organs and systems, returning a global measure of health [4,5]. A strong link between physical functioning and quality of life has also been reported [6].
The technological advancements achieved in the last decades led to the development of several systems, devices, and sensors that might be used to instrumentally evaluate physical function and performance. Exploiting such solutions has two main advantages. In first place, it increases the reliability of the measurements, by limiting the intra-and inter-operator variability [7,8]. Secondly, it can be used to obtain an objective measurement of physical function characteristics that are only seldom qualitatively evaluated in current practice, for example, measures of symmetry and balance can be acquired while performing a 6 min walking test [9]. Tri-axial accelerometers [9], high-speed cameras and markerless motion capture systems [10], and instrumented mattresses and force plates [11], as well as surface electromyography (sEMG) [12], are devices that may be suitable for the evaluation of physical function in persons affected by COPD. However, these devices considerably differ one from the other in terms of performance, price, dimension, and the tests implemented for the instrumental evaluation of physical performance.
Although previous reviews reported on the application of specific tools for the assessment of subjects with COPD, to date, no systematic reviews provided a comprehensive perspective on the application of the technological solutions for the functional evaluation of their physical performance in clinical and research settings [9,13,14].
Therefore, we aimed to systematically review the available scientific literature about the implementation of technologies-including systems, devices, and sensors-for the evaluation of physical function in subjects presenting COPD.

Methods
In accordance with PRISMA guidelines, the protocol for this systematic review was registered with the Open Science Framework (OSF) on 7 December 2020 (DOI:10.17605/ OSF.IO/GXW6S; https://osf.io/gxw6s (accessed on 7 December 2020)) [15]. The article was written in keeping with the Guidelines for Meta-Analyses and Systematic Reviews of Observational Studies (MOOSE).

Eligibility Criteria and Information Sources
We reviewed studies providing information about the instrumental evaluation of physical function in persons with a diagnosis of COPD. We included studies that (1) addressed either a function (such as walking) or a test of physical performance (such as the 6-Meter Walking Test-6MWT), (2) studies that reported quantitative measures derived from the instrumental assessment of physical function, and (3) studies conducted either in a clinical or a laboratory setting.
Case-series studies, letters to the editor, reviews, and metanalyses, as well as commentaries, were excluded from this systematic review. Studies evaluating an isolated movement or characteristic (e.g., hand-grip strength), a specific body part, non-functional activities (e.g., cycling), or that reported only the assessment of respiratory function were also excluded. Finally, we excluded studies that performed remote/home-based measurements (e.g., studies assessing physical activity using an accelerometer worn at home for several days were excluded). This review was limited to articles in English or Italian, published between January 2001 and April 2022. The literature search was conducted in PubMed, Web of Science, Embase, and Scopus. An ethical committee's approval was not needed for the conduction of this study.

Search Strategy
The search strategy for PubMed (available in the Supplementary Material Table S1) was adapted from the search filter proposed by a previous study [16]. In brief, the search string included terms for (1) construct search, (2) population search, and (3) instrument search. The search included terms related to physical function (e.g., walking, gait, balance, etc.), the population of interest (e.g., chronic obstructive pulmonary disease), and the instrument used for the assessment (e.g., actigraphy, kinematic analysis, and wearable sensors). Then, these searches were combined with the search filter for measurement properties and the exclusion filter to remove irrelevant records. The search terms were subsequently adapted for use with the other databases. References and additional files from selected articles were checked to identify further studies eligible for inclusion. The search was re-run just before the final analysis to retrieve new studies suitable for inclusion.

Selection Process
Two researchers (A.Z. and S.P.) independently screened the articles' titles and abstracts after duplicate exclusion. Conflicts were resolved via consensus. In case a consensus was not reached, a third assessor (NFL) was included in the discussion. The full texts from all selected abstracts were retrieved, and they were screened and selected using the same procedure. An online application [17] was used to simplify the process of abstract and full-text screening. Study characteristics and information were independently extracted from selected papers by two assessors. Extracted data were then compared and possible inconsistencies were resolved.

Data Items and Presentation of Results
We collected data on (1) the study (i.e., author, year, country, and design), (2) participants (i.e., sample size, demographic, and clinical characteristics), and (3) instrumental evaluation (i.e., device features, application procedures, and provided parameters). Then, the selected studies were divided into two groups: the first one included those employing technology-derived metrics to describe the functional characteristics of COPD participants, to compare healthy and COPD participants, or to investigate the association between physical function and other health-related outcomes. This group of studies was named "application studies". The second group, including articles aiming to evaluate the performance (e.g., reliability, measurement error, precision, and validity) of the instrumental evaluation of physical function, was named "validation studies".

Risk of Bias Assessment
We assessed the risk of bias for the validation studies employing the Consensus-based Standards for the selection of health Measurement Instruments COSMIN Risk of Bias tool [18]. The tool offers two different sets of criteria for studies' evaluation according to their aim (reliability studies and measurement error studies). The worst-score-count method was applied to determine the risk of bias. The risk of bias was independently evaluated by the two reviewers: conflicts were resolved by the third assessor.

Study Characteristics
Most of the studies (n = 22) presented an observational design, and 18 aimed at comparing the functional characteristics of participants with COPD with those of individuals without this condition (mostly healthy controls). The number of persons with COPD included in the selected studies ranged between 6 and 80, with a proportion of female patients comprised between 0% and 83% and a mean age comprised between 60 and 72 years old. The mean % of predicted Forced Expiratory Volume in the 1 s (FEV1) ranged from 35.1% to 58.4%. A recent COPD exacerbation was reported as an exclusion criterion in nine studies, whereas hypoxemia and/or chronic oxygen supplementation were considered exclusion criteria by four studies. The presence of comorbidities able to interfere with physical function tests was reported as an exclusion criterion in most studies (n = 18). Most studies (n = 17) reported that the diagnosis of COPD was made in accordance with the Global Initiative for Chronic Obstructive Lung Disease (GOLD) guidelines, whereas the others mainly reported that the diagnosis was made by a physician or reported in the medical charts.

Application Studies
Eleven application studies employed a single device for the evaluation of physical function, whereas in the remaining papers, multiple instruments were used, as shown in Tables 2 and 3. Accelerometers (either alone or in combination with other devices) were employed in five studies, whereas force plates or instrumented mattresses were used in eight papers. The employment of surface electromyography (sEMG) was reported in five studies. Gait analysis systems, three-dimensional (3D) motion captures systems, or highspeed cameras were employed in seven studies, and two studies reported the utilization of near-infrared spectroscopy (NIRS). Walking was the most commonly assessed function (15 studies). The other application studies evaluated a variety of tests and functions: 6 min step test, pegboard and ring test, balance and perturbation tests, and domestic activities of daily living (simulated in the laboratory setting). Table 4 shows the parameters derived from the employment of each technology: accelerometers were used mainly to obtain information about activity intensity (i.e., n. counts/unit of time) or volume (i.e., total n. counts or steps), although some spatiotemporal parameters were inferred from accelerometric data (step length, step time, gait speed). Spatiotemporal parameters of gait were mainly obtained using instrumented mattress and gait analysis systems. Force plates were employed in a variety of tests (among which, sitto-stand tests, perturbed balance tests, and jump tests). The parameters obtained from such technologies were strongly related to the test performed: they ranged from the duration of meaningful function segments (such as the time taken for standing from the sitting position) to inferred measures of muscle power. Surface EMGs and NIRSs were employed to investigate specific muscles' activation time, duration, signal intensity, or infer oxygen consumption during the execution of functional tests.

Validation Studies
Four studies were classified as validation studies ( Table 1). As shown in Table 5, validation studies were heterogeneous in terms of technology employed and aim. Walking parameters were evaluated using 6MWT in two studies [37,38], using a 5 min walking protocol in one study [39] and a 10 m walkway in the remaining one [23]. Three [23,37,38] out of four studies measured gait speed, whereas the remaining one [39] assessed walking distance. One study [37] estimated gait speed using a mobile phone application based on an accelerometric measurement and compared such measure with manual walkway testing. Another study [38] investigated the validity of an instrumented treadmill with virtual reality and 3D motion analysis system evaluation by comparing the results with overground 6MWT. A third paper [23] evaluated the test-retest reliability of gait speed assessment exploiting a tri-axial accelerometer. Lastly, one study [39] investigated the correlation between the measurement obtained from a motion sensor and criterion methods during the assessment of walking distance at two different speeds.

Risk of Bias (Validation Studies)
According to the COSMIN tool, study quality was doubtful in three studies [23,38,39] and inadequate in one study [37]. Unclear concealed administration of tests and score assignment, or different conditions between repeated measurements, were the main reasons contributing to the doubtful quality of the studies. One study [37] was considered of inadequate quality due to important flaws in the sample composition.   A tri-axial accelerometer system (Mimamori-gait system, LSI Medience Corporation, Japan) The accelerometer was fixed to a belt around the level of the subject's third lumbar vertebra.

Gait speed
Step length Cadence Walk ratio Acceleration magnitude Step time

Discussion
In this systematic review, we evaluated the implementation of different technologiesincluding systems, devices, and sensors-to assess the physical function of persons affected by COPD. We found that a wide variety of technologies, including small wearable devices (such as accelerometers), as well as cumbersome gait analysis systems, were employed in the selected studies. In most cases, these technological solutions were implemented for evaluation purposes, whereas a limited number of trials aimed to validate or assess the reliability of different tools in the specific context given by patients with COPD; unfortunately, this subset of studies showed doubtful or inadequate quality.

Same Metrics and Different Technologies
In our systematic review, we found that similar metrics were retrieved using a variety of technologies. For example, the spatiotemporal parameters of gait were obtained from accelerometers, gait analysis systems, and instrumented mattresses. In addition to the importance of maintaining the ecology of movement, containing operating costs and ensuring its applicability in clinical contexts, the validity of measurements obtained by different technologies is of pivotal importance for the implementation of an instrumental evaluation of mobility. Instrumented mattresses and walkway of force plates have been reported as valid technologies for the assessment of spatiotemporal parameters in healthy individuals when compared with video-based systems as gold standards [40,41]. The spatiotemporal parameters of gait obtained from a tri-axial accelerometer located near the center of grav-ity or from instrumented mattresses have also been shown to exhibit good-to-excellent collinearity in healthy individuals [42,43]. In addition, a recent meta-analysis [8] showed that the validity of Inertial Measurement Unit (IMU)-derived spatiotemporal parameters was generally excellent (using either instrumented walkway, instrumented mattresses, or motion capture systems as reference). However, the authors of the latter study [8] highlighted how this finding was strongly limited by the quality of the investigated studies, generally characterized by low statistical power. Our systematic review extends the issue of the absence of high-quality studies investigating the validity of the instrumental evaluation of physical function to the setting of COPD. For example, only one study [39] included in this systematic review compared measurements obtained from an accelerometer with a video-based assessment of gait: the collinearity between the two methodologies for the measurement of simple characteristics of gait (total distance walked and intensity) was moderate-to-low (Pearson's rho ranging between 0.47 and 0.63). It is also worth mentioning that it is likely that the validity of an instrumented evaluation of function in persons affected by chronic conditions is lower than the one reported in healthy individuals. A recent study [44] comparing gait events recognition obtained from magneto-IMUs and instrumented mattresses, for example, showed that the errors in the estimation of the initial contact, stride time, and step time were significantly lower for healthy older adults in comparison with participants affected by Parkinson's disease. This result is likely to be explained by the higher heterogeneity that can be found in pathological patterns of gait and the consequent difficulty in finding rules and algorithm for the identification of gait events. However, a previous review of this topic was not able to define the characteristic traits of gait patterns in individuals with COPD due to contrasting results in the existing literature [14]. However, the impact of possible measurement errors in the instrumental evaluation of physical function in COPD-related clinical or research practice is unknown. Almost the totality of the application studies included in our systematic review were cross-sectional, mostly with a case-control design; the possible association of specific technology-derived metrics with meaningful clinical outcome was not investigated, as well as possible confounders that may explain the differences found in terms of instrumental metrics between persons with COPD and healthy controls. It is crucial, for the implementation of technologies in clinical practice, to comprehensively investigate the possible associations between technology-derived metrics and both poor-health-related outcomes and specific indicators of disease progression in COPD.

The Instrumental Evaluation of the 6 min Walking Test
Out of eighteen studies (including both application and validation ones) evaluating gait, nine were conducted by performing the 6MWT. This result is not surprising given the current importance of this test protocol in COPD clinical practice [3]. The 6MWT, performed either on a treadmill or on the ground, was used to investigate a variety of parameters, exploiting different technologies; accelerometers, instrumented mattresses, gait analysis system, and sEMG were used to retrieve gait volume and intensity, spatiotemporal parameters, and muscle activation and fatigue.
Such novel technologies allow retrieving information about the variability of specific parameters or metrics while the test is performed: for example, a previous work has shown that fatigue may impact the temporal variability of spatiotemporal parameters [45] or that the distance walked may significantly impact gait rhythmicity [46].
Rutkowski and colleagues [32] reported that, both in participants with COPD and healthy controls, stride lengths were higher at the beginning of the test in comparison with those measured after 3 min. They also reported that stride length seemed to revert to baseline values at the end of the 6MWT, probably because of the provided instructions and motivation. Liu and colleagues [26] reported the coefficients of variation (i.e., standard deviation divided by mean) of several spatiotemporal parameters of gait obtained during the whole 6MWT (performed on a treadmill and investigated using a gait analysis laboratory); these ranged between 2% for stride time and 14% for step width, although the temporal patterns of such variations were not investigated. Interestingly, these authors showed, in a sub-analysis of participants with comparable walking speeds, that the mean values of spatiotemporal parameters of gait were similar between participants with COPD and healthy controls, but the variability of stride length, double support time, and step lengths seemed to be higher among cases. These results also extend our knowledge about the potential role of the variability of instrumental parameters as an early sign of deterioration of the gait pattern in persons with COPD, as already described for the general population of older adults [47][48][49].

Which Technology to Implement?
Several factors should be taken into consideration before implementing a technology for the instrumental evaluation of physical function in clinical or research practice. In the first place, size and cost may be considerably different from one instrument to another: tri-axial accelerometers are easily worn on different parts of the body and are generally affordable, whereas 3D gait analysis systems typically need dedicated spaces or infrastructures, and their cost is significantly higher. The physical test or function that needs to be objectively measured is another important factor for the choice of the system to implement: some gait labs are combined with treadmills that allow a detailed evaluation of gait or run, whereas force plates are often used both for balance tests' evaluation (e.g., posturography) and assessment of strength or power during functional tests involving the lower limbs (e.g., jumps and chair stands). The capability to assess different functions with the same device may also be worth considering; among the application studies included in this systematic review, instrumented mattresses have been implemented only for the evaluation of gait, whereas force plates, accelerometers, NIRS, and sEMGs were employed to retrieve parameters from a variety of tests and function (i.e., gait, balance, jump, and replication of domestic activities). In addition, the aim of an instrumental evaluation of function should be clearly defined before the implementation of any technology. For example, in our study, we found that accelerometers have been used to obtain data about the intensity of gait (cadence), spatiotemporal parameters of gait (stride length, step length, and gait speed) and measures of gait symmetry. Such data cannot be retrieved using NIRS sensors or sEMGs which, in turn, have been implemented to obtain detailed information about the activation and usage of particular muscles (or muscular groups) during the performance of specific actions or functions.

Current Limitations to Technology Implementation
Our study highlights a significant implementation of technologies for the instrumental evaluation of physical function in persons with COPD over the last two decades. We also found that there was a significant heterogeneity in terms of the type of device used, applications, tested function(s), and retrieved metrics. This finding is of particular interest given that the lack of standardization and device-dependent assessment results were deemed as among the most important factors that hinder technology-aided assessments of physical function [50]. Furthermore, the possibility of deriving a general conclusion and giving precise indications regarding the application of these technologies to patients with COPD is limited by the fact that the studies included in our review recruited a limited number of participants. Indeed, 9 out of 24 papers included fewer than 20 participants with COPD. Furthermore, most authors selected their study population by excluding individuals affected by conditions potentially impacting the performance in physical function tests: cardiological, neurological, and musculoskeletal comorbidities were the most cited exclusion criteria. However, COPD is known to be frequently associated with multiple chronic diseases; in particular, due to shared risk factors, cardiological and neurovascular diseases are frequent comorbidities of COPD. These considerations suggest that the results from the studies included in this systematic review may not be directly generalizable to a significant share of older persons affected by this respiratory condition.

Conclusions
Novel technological solutions are more and more used to investigate physical function in persons affected by COPD, and a variety of potentially interesting metrics can be retrieved from such instrumental evaluation. However, a general lack of standardization and limitations in study design and sample size hinder the implementation of the instrumental evaluation of function in clinical practice.

Future Directions
Shared protocols for the validation of instrumental evaluation of function in persons affected by COPD are warranted. Future longitudinal studies should focus on the association between metrics derived from the instrumental evaluation of function and clinically meaningful outcomes for persons with COPD.