Do wearable fitness devices correlate with performance-based tests of work-related functional capacity?

Abstract

BACKGROUND:

The use of wearable accelerometers in conjunction with Functional Capacity Evaluation (FCE) may provide additional useful information about maximum performance in workers and enhance the validity of functional testing. However, little research has been conducted to compare accelerometer output with performance during FCE.

OBJECTIVE:

The objectives of this study were to: (1) Determine the magnitude and direction of correlation between participant performance on five FCE tasks and scores from Actigraph activity monitors; and (2) Compare the results of two different placements of Actigraph devices.

METHOD:

We used a cross-sectional design and convenience sampling to collect data from 46 healthy participants. Each participant completed 5 functional tasks selected from the WorkWell FCE protocol while wearing 2 Actigraph devices, 1 on the dominant side waist and 1 on the non-dominant wrist. The FCE tasks included 5-repetition maximum lifting (floor-to-waist, waist to crown and front carry), a sustained overhead work endurance task, and the 6-minute walk test. Analysis included calculating Pearson regression coefficients between maximum FCE item performance and Actigraph vector magnitudes (VM) along with Intraclass Correlation Coefficients (ICC) to compare VM activity counts derived from the Actigraphs on the waist and wrist.

RESULTS:

Thirty-Nine (84.8%) participants had complete data and were included in analysis. Findings indicate Actigraph VM data from the device worn on the waist correlated positively with maximum lift performance (r = 0.39–0.64, p < 0.001 to 0.08) and 6-minute walk distance (r = 0.66, p < 0.001). Actigraph data from wrist placement were not significantly correlated with FCE performance on any of the functional tasks, except when comparing average VM data and waist to crown lift (r = 0.44, p < 0.001). There was no significant correlation in either Actigraph placement for VM and overhead work time. ICCs between the two Actigraph placements ranged from poor to acceptable agreement (ICC = 0.24–0.70, p < 0.001 to 0.19).

CONCLUSIONS:

Actigraph device output correlated moderately with maximum performance on FCE lift and ambulation tests. Waist placement appears more suitable than wrist during performance-based tests.

Keywords

Accelerometry disability evaluation motor activity return to work exercise test

1 Introduction

Functional capacity evaluations (FCE) are performance-based tests typically conducted by physical, occupational and exercise therapists. They measure an individual’s ability to perform meaningful tasks on a safe and dependable basis. FCEs evaluate activity and activity limitation in accordance with the International Classification of Functioning, Disability and Health [1]. FCE findings are then extrapolated to the domain of participation in work and other activities of daily living. FCEs have two specific purposes. First, they measure an individual’s potential to carry out job related tasks (i.e. lifting, carrying, climbing, etc.) and second, they identify if an individual has adequately recovered from an injury in order to return to workplace demands [2]. However, FCEs can be lengthy, expensive, and they only modestly predict future return to work [3, 4]. It is also unknown how well FCE performance correlates with actual ‘real-world’ work ability.

Recently, more and more physical activity research has involved the use of accelerometer-based activity monitors [5 –8]. These monitors have the potential to provide more detailed and comprehensive information about functional ability in both clinical and lifestyle research. In terms of FCE, this is important as wearable activity monitors may enhance the validity of functional testing, including allowing measurement of real-time function while the patient conducts actual work activity at the workplace. Currently FCEs are largely dependent on judgments of safe maximum levels from either patients (psychophysical testing) or therapists (kinesiophysical testing). Improved measurement of function will improve workability determinations and communication between stakeholders involved in making return-to-work decisions. These monitors also have the potential to evaluate patients outside of the workplace to understand barriers to function and participation in other activities of daily living. Furthermore these monitors are commercially available and affordably priced [9].

Accelerometers have the ability to measure uniaxial or triaxial movements and are usually worn on the waist or wrist. However, there are few studies that use accelerometer data to classify specific movements or activities of individuals. One such study by Karantonis et al. studied the ability to classify movements based on real time accelerometer data. Movement classes involved postural orientation, walking and falls. The overall accuracy of prediction ranged from 63.3% for walking at a slow speed to 100% accuracy for stand-to-sit, lying-to-sit, sit-to-lying, and falling events both active and inactive [10]. Another such study by Bonomi et al. used a decision tree model to attempt to identify types of physical activity, the duration of activities, and their intensity. This model was found to be 100% accurate in classifying lying and running but only 59% accurate in classifying standing [11]. While these studies were able to classify movement reasonably well, there are no current studies comparing activity counts with time of activities or weight of resistance during activity. Also, despite the potential utility of accelerometers in evaluating physical function, it is unknown how accelerometry data correlates with performance during FCE.

The use of accelerometers and wearable fitness devices in conjunction with FCEs may provide additional useful information about day-to-day function in workers. These devices are readily available and have the potential for evaluating activity within the context of work, allowing direct comparisons between levels of activity demonstrated during FCEs and levels required at work. Accelerometers are an easy, accurate way to measure level of activity and provide information about activity counts. Activity count information is used to derive information about activity intensity (i.e. light, moderate, vigorous) and can be displayed over varying epochs, or units of time. Despite the potential utility of accelerometers in evaluating physical function, it is unknown how accelerometry data correlates with FCE performance. The application of wearable fitness devices into rehabilitation settings has the potential to add another objective measure of activity and function to the process of assessment. As these devices become more advanced and mainstream it is important for health care practitioners to understand their clinical potential. Likewise, there are potential benefits for improving the accuracy of FCE while potentially decreasing assessment and evaluation cost in addition to assessment duration.

The objectives of this study are to (1) quantify peak and average activity during selected FCE tasks, (2) to examine relationships between peak or average activity counts and peak performance during the FCE tasks and (3) to exam the relationship between activity counts from two placements of the Actigraph devices.

1.1 Hypothesis

The overall objective of this research was to gather preliminary information and data on the use of accelerometry during FCE and the association between accelerometry outputs and FCE performance. As such, this research should be considered exploratory in nature. Since no similar research has been conducted in the context of FCE, it was unknown to what magnitude the measures might correlate or whether the correlations would be positive or negative. Therefore, our null hypothesis was that there is no statistically significant correlation between activity monitor outputs and maximum performance on the floor to waist lift, waist to crown lift, carrying FCE items, or the sustained elevated work task. It was hypothesized for the 6-minute walk test that there would be a positive correlation of fair to moderate magnitude (r = 0.25–0.5) between total activity counts and distance of walking.

2 Methods

2.1 Study design

This was a validation study using a cross sectional design. The goal was to evaluate the data from Acti Graph devices worn by participants during FCE activities. A cross sectional design allowed for data to be collected efficiently and at a specific point in time. Since the data from the ActiGraph is being compared with results from the FCE and FCEs are considered the gold standard in return to work, the validation is concurrent in nature. This study was approved by the University of Alberta’s Health Research Ethics Board.

2.2 Sampling

Convenience sampling was used to enroll participants. Specifically, participants were recruited via word of mouth and social media. Participants either male or female, needed to be healthy individuals between the ages of 18–65 years. Participants were excluded if they were injured or had any physical limitations that would hinder their ability to complete certain exercise components from the FCE. This was determined using the Physical Activity Readiness Questionnaire (PAR-Q - revised 2002) [12]. Once a volunteer had expressed interest in taking part in this study, they were asked to complete the PAR-Q to ensure they met inclusion criteria. If the participant answered yes to any questions on the PAR-Q we determined if the individual was suitable to participate on a case-by-case scenario through review of the medical history and discussion with a senior researcher and physical therapist (DPG). In 2 instances an individual answered yes to the question: do you lose your balance because of dizziness or do you ever lose consciousness, in 2 other instances individual answered yes to the question: do you have a bone or joint problem that could be made worse by change in your physical activity and in one instance an individual answered yes to the question: in the past month have you had chest pain when not doing physical activity. However in all situations they had previously consulted a doctor regarding the condition and were not currently limited by these factors. As such they were deemed eligible to participate. If there was no reason to exclude the individual, then a time suitable for both the participant and researcher was arranged to conduct testing.

A sample size of 40 was required based on an alpha level of 0.05 and a power of 80%. This value was calculated using Power and Statistical Software (PASS) based on the following parameters: ρ1 = 0.00 and ρ2 = 0.4 (midpoint of predicted fair to moderate correlation). We added an additional 6 participants to account for attrition (i.e. participants stopping testing) or errors during data collection, making a total sample size required of 46.

2.3 Data collection

After informed consent was obtained, all participants were equipped with 2 ActiGraph wGT3X-BT devices that had been initialized using the ActiGraph 6 software. One device was worn on the non-dominant wrist and a second device was worn on the waist located on the anterior superior iliac spine on the dominant side using a belt style strap. Previous studies have compared Actigraph output based on device placement, with a significantly higher raw output seen from devices worn on the wrist in comparison to those worn on the hip [13]. Other studies have shown that device placement on the waist provided more accurate estimates of energy expenditure over wrist placement [14, 15]. However the wrist placement may be more comfortable and practical, both during the assessment and in future lifestyle-based studies.

The Actigraph device has been found to be reliable and valid for various forms of lifestyle and physical activity related research [5 –8]. Motion data is collected from vertical (Y), horizontal right-left (X) and horizontal front-back (Z) axes and “raw” mode output allows the researcher to select various sampling rates ranging from 30 to 100 Hz [16]. The wG3TX-BT is able to accurately record accelerations ranging in magnitude between 0.05–2.5 Gravitational units on all three axes. Activity counts are then calculated using the absolute change in acceleration over cycle period times or epochs. Vector magnitudes (VM) can be calculated using the vector-summed value. These can be used to show a larger representation of motion, as they give information on movement in all three planes and can be beneficial in more complex motion as compared with simple uniplanar movements. Participants were also asked to wear a Polar heart rate monitor, which was part of the FCE protocol for determining maximum heart rate levels.

After devices were attached, each participant completed 5 tasks taken from the WorkWell FCE including three lifting tasks, sustained, weighted elevated work, and a 6-minute walk test (6MWT). These FCE tasks are described in more detail in the Measures section. After all 5 functional tasks had been completed the ActiGraph data was downloaded using the Actilife 6 software (Pensacola, Florida) and exported to a Microsoft Excel file (Redmond, Washington).

Each FCE task was completed after an explanation and demonstration had been given and any questions had been answered. The evaluator (JK) was trained in how to conduct these FCE items by a trained and experienced occupational therapist that is a practicing WorkWell FCE evaluator. A practice session was also held with the research team before formal testing began. All testing was conducted in the Functional Performance lab in Corbett Hall at the University of Alberta. The time to complete each task was logged using the computer’s clock to which the devices were previously synced (to the nearest 5 millisecond interval) at the beginning and end of each set of 5 repetitions along with the corresponding weight lifted. The Bluetooth feature on the Actigraph devices allowed for data to be seen in real time on the Actilife software. Progressive isoinertial testing was completed. After each set of lifts, if the participant believed they could safely increase the weight, and there were no signs of observable stopping criteria [17, 18], then more weight was added and the participant once again completed a 5-repetition set. A rest period of maximum two minutes was provided between sets, or once heart rate had dropped to around 100 beats per minute. This time was used to change the weights and isolate activity counts between sets.

2.4 Measures

Actigraph - The ActiGraph is a piezoelectric sensor that generates signal based on movement [5]. The wG3TX-BT is a 3-axis accelerometer. Previous studies have shown the Actigraph to have high inter-monitor reliability [6] and some validity for estimating energy expenditure, tracking movements and exercise repetitions [5 –8]. The Actigraph displays data for each individual axis or as the square root of the squared sum of each axis (known as the vector magnitude). The Actigraph collects raw data, which was displayed at varying epoch times. These epoch times are selected at the time of importing the raw data into the Actilife software. Due to the short nature of the selected FCE tasks, 5-second epochs were chosen. Each set of weighted lifting exercise took a maximum of 90 seconds. With 5-second epochs a maximum of 18 VM were recorded. The first 5-second epoch used for analysis was once the exercise began with the final 5-second epoch being upon completion of the exercise. Ultimately, the measure used was vector magnitudes over a 5-second epoch at 60 hz for this study.

Functional Capacity Evaluation - The 5 FCE items were selected from the broader WorkWell FCE protocol. Three different manual handling tasks (floor-to-waist and waist-to-crown lift along with 25 m front carry) were completed. Two systematic reviews have been conducted on the WorkWell FCE protocol. Kuijer et al. found that 13/14 studies suggested the lifting items are predictive of RTW [19]. Bieniek and Bethge found the manual handling exercises to have high inter-rater reliability, intra-rater reliability and test-retest reliability [20]. Weighted elevated work was also conducted, which within the protocol assesses posture, mobility and upper extremity endurance. Brouwer et al. found the overhead work test to have a high kappa agreement percentage, but a low ICC in test-retest reliability [21]. The 6-Minute Walk Test (6MWT) was also conducted to assess walking capacity. The 6MWT is a common test across a variety of FCE protocols. This test has shown to be a valid instrument in testing exercise capacity across all ages [22, 23]. In more description, the FCE items are as follows:

5-repetition max floor to waist level lift - For this lift, the maximum weight lifted (in kilograms) was used for analyses and compared to average vector magnitude and peak vector magnitude from the ActiGraph device during the repetition in which maximum performance was reached.

5-repetition max waist to crown level lift - For this lift, the maximum weight lifted (in kilograms) was used for analyses and compared to average vector magnitude and peak vector magnitude from the ActiGraph device during the repetition in which maximum performance was reached.

Front (horizontal) carry - For this lift, the maximum weight lifted (in kilograms) was used for analyses and compared to average vector magnitude and peak vector magnitude from the ActiGraph device during the repetition in which maximum performance was reached.

Weighted overhead work until fatigue - A perforated board at about crown level had bolts and nuts facing outward. A 3-pound Velcro weight was worn on each wrist. The participant works at crown level, being kept busy by screwing and unscrewing the nuts and moving them from one bolt to another. There was no set pattern in which the participant followed, as long as both hands remained elevated at a height in line with the bolts. Time started as soon as the participant raised their arms to begin work. The tester watched for thoracic extension, attempts to rest arms, shoulder hiking or elbow dropping. If any of those actions were noticed or if the participant dropped one or both arms, the test was stopped. For this task total time of exercise was used for analyses and compared to the average vector magnitude from the ActiGraph device.

6-Minute Walk Test - In this study, each turn around (lap) represented 25 meters. The tester had a timer and lap counter to keep count of the distance travelled. The tester ensured that a steady pace was maintained, there were no gait deviations, or changes in weight bearing. If any of those actions were noticed, or if the participant reported discomfort, then testing was stopped. For this exercise, distance (in meters) was used for analysis and compared to total activity count from the Actigraph.

Potential confounders –One of the potential confounding variables of concern was the effect of feedback or comments from the FCE examiner. To limit this, there was no feedback given to participants during the exercises and all exercises were explained using a standard script. As many participants were recruited via word of mouth, it was also asked by the researcher that the exercises not be explained to other potential participants in detail to prevent any form of preparation that could have impacted the results.

2.5 Data analysis

Descriptive statistics were calculated including mean and minimum/maximum values for both demographic data and measures. Exercises that were weight based were analyzed using descriptive statistics in the form of mean, standard deviation, and minimum/maximum values. The mean, standard deviation, and minimum/maximum values were found for maximum weight for each manual-handling task. Exercises that were time based were also analyzed using descriptive statistics in the form of mean, standard deviation, and minimum/maximum values.

Next, two sets of inferential statistics were calculated. First, the association between FCE variables (time or weight lifted) and Actigraph variables (peak or average vector magnitudes) from the Actigraph placed on the dominant side waist were analyzed using Pearson Correlation with 95% confidence intervals. Second, the association between FCE variables (time or weight lifted) and Actigraph variables (peak or average vector magnitudes) placed on the non-dominant wrist was analyzed using Pearson Correlation with 95% confidence intervals. Magnitude of the Pearson correlation values was evaluated using criteria from Portney & Walkins, with 0–0.25 indicating none to poor, 0.25–0.5 indicating fair to moderate, and 0.5–0.75 indicating moderate to good, and above 0.75 indicating excellent correlations [24].

Intraclass Correlation Coefficients (ICC) were used to determine if the two Actigraph placements provided consistent measures using two-way random effect models, consistency agreement and average measures. A two-way random effect model was used as both the sample and the Actigraph devices are from random samples. As it was expected that data from the wrist would be higher than the waist, consistency agreement was used to analyze linear trend. Average measure considers the average from both devices being used through out all the data. ICC values were evaluated using criteria from McDowell [25], in which ICC above 0.75 indicate excellent inter-rater agreement, 0.6 to 0.74 shows good agreement; 0.4 to 0.59 indicates fair to moderate, and below 0.4 is poor agreement.

Statistical analysis was performed using SPSS (version 23.0, Armonk, New York). All data used an alpha value of α= 0.05.

3 Results

3.1 Participant characteristics

In total 46 participants were recruited for this study. 54.3% of the participants were male and 89.1% were right handed. The mean age of the sample was 23.7 years, and the mean height and weight were 170 cm and 73.2 kg respectively. Additional details are available in Table 1. Thirty-nine of the 46 participants (84.78%) had no missing data on the FCE and Actigraph measures (see Fig. 1). Missing data was a result of Actigraph recording error, where the device stopped recording data before the testing was completed.

Table 1
Participant Characteristics (n = 46)

Total Male Female Total Male Female

Mean (SD) Min, Max values

Age 23.7 (3.7) 24.5 (4.1) 22.8 (3.0) 19, 40 19, 40 19, 34

Height (cm) 170 (10.8) 177 (8.5) 161.6 (6.1) 152, 194 157, 194 152, 172

Weight (kg) 73.2 (18.0) 82.6 (17.0) 62.0 (11.7) 43, 135 46, 135 43, 90

Male 54.3%

Right-Handed 89.1%

Total	Male	Female	Total	Male	Female
Age	23.7 (3.7)	24.5 (4.1)	22.8 (3.0)	19, 40	19, 40	19, 34
Height (cm)	170 (10.8)	177 (8.5)	161.6 (6.1)	152, 194	157, 194	152, 172
Weight (kg)	73.2 (18.0)	82.6 (17.0)	62.0 (11.7)	43, 135	46, 135	43, 90
Male	54.3%
Right-Handed	89.1%

Fig. 1

Flow chart showing usable participant data sets from the various FCE exercises.

3.2 Functional Capacity Evaluation performance characteristics

The front carry was found to have the highest mean maximum weight lifted (36.9 kg) followed by the floor to waist level lift (30.7 kg) and the waist to crown level lift (20.9 kg). Both sexes followed the same trend as above for sex-specific average maximum weight lifted. The front carry was also found to have the largest difference between maximum and minimum weight at 52.2 kg, followed by floor to waist with 44.2 kg and lastly waist to crown level with 35.2 kg. Again, both sexes followed the same trend as above for sex-specific ranges. These results are presented in Table 2.

Table 2
Descriptive statistics for Functional Capacity Evaluation item performance

Total Male Female Total Male Female

Mean (SD) Min, Max Values

Floor to Waist weight (kg) 30.7 (11.8) 38.8 (9.0) 21.0 (6.1) 14.0, 58.2 23.0, 58.2 14.0, 37.8

Waist to Crown weight (kg) 20.9 (8.0) 26.5 (6.3) 14.3 (3.3) 9.4, 44.6 16.2, 44.6 9.4, 21.9

Front Carry weight (kg) 36.9 (13.4) 45.7 (10.1) 26.4 (8.1) 15.1, 67.3 24.2, 67.3 15.1, 43.5

Overhead Work time (minutes) 2 : 31 (1 : 27) 3 : 00 (1 : 43) 1 : 56 (0 : 44) 0 : 58, 8 : 18 1 : 05, 8 : 18 0 : 58, 3 : 38

6 Minute Walk Test distance (meters) 457.1 (55.2) 459.8 (41.9) 454.0 (68.9) 350, 625 381, 560 350, 625

	Total	Male	Female	Total	Male	Female
Floor to Waist weight (kg)	30.7 (11.8)	38.8 (9.0)	21.0 (6.1)	14.0, 58.2	23.0, 58.2	14.0, 37.8
Waist to Crown weight (kg)	20.9 (8.0)	26.5 (6.3)	14.3 (3.3)	9.4, 44.6	16.2, 44.6	9.4, 21.9
Front Carry weight (kg)	36.9 (13.4)	45.7 (10.1)	26.4 (8.1)	15.1, 67.3	24.2, 67.3	15.1, 43.5
Overhead Work time (minutes)	2 : 31 (1 : 27)	3 : 00 (1 : 43)	1 : 56 (0 : 44)	0 : 58, 8 : 18	1 : 05, 8 : 18	0 : 58, 3 : 38
6 Minute Walk Test distance (meters)	457.1 (55.2)	459.8 (41.9)	454.0 (68.9)	350, 625	381, 560	350, 625

The sustained, weighted overhead work exercise took on average 2 minutes and 31 seconds. Males were found to have a notably larger range than females. The 6-minute walk test was the last exercise completed. The mean distance walked was 457.2 m, with only a slight difference in means between males (459.8 m) and females (454.0 m). Full descriptive Statistics for FCE item performance are shown in Table 2.

3.3 Pearson correlations between functional capacity evaluation performance and Actigraph data

The correlational analysis was completed amongst the vector magnitudes from the devices placed on the waist and wrist. The maximum weight lifted during the floor to waist lift was found to have a statistically significant positive correlation at a magnitude judged to be fair to moderate for both peak (r = 0.40) and average (0.45) vector magnitudes. Both peak and average vector magnitude correlations from the waist placement were significant and a non-significant correlation with both peak and average vector magnitudes was observed from the wrist placement (r = 0.18, r = 0.18) for the floor to waist lift. The maximum weight lifted during the waist to crown level correlated significantly and positively at a magnitude judged to be fair to moderate with both peak (r = 0.39) and average (r = 0.39) vector magnitudes from the waist. The average vector magnitude from with wrist was also significant (r = 0.44). A non-significant correlation was found from the peak vector magnitude at the wrist (r = 0.15). The maximum weight lifted during front carry was found to correlate significantly and positively at a magnitude judged to be moderate to good, with both peak (r = 0.57) and average (r = 0.64) vector magnitudes from the waist placement being significant. Non-significant correlations were observed for peak (r = –0.13) and average (r = 0.24) vector magnitudes from the wrist. Correlation results for the FCE lift items are presented in Table 3.

Table 3
Correlations between Functional Capacity Evaluation results and vector magnitudes from waist and wrist Actigraph placements

Peak Waist Average Waist Peak Wrist Average Wrist

Vector Magnitude Vector Magnitude Vector Magnitude Vector Magnitude

Floor to Waist Max Weight Pearson Correlation 0.40 0.45 0.18 0.18

Sig. (2-tailed) 0.005 0.002 0.230 0.245

n 46 46 46 46

Waist to Crown Max Weight Pearson Correlation 0.39 0.39 0.15 0.44

Sig. (2-tailed) 0.008 0.007 0.314 0.002

n 46 46 46 46

Front Carry Pearson Correlation 0.57 0.64 –0.13 0.24

Max Weight Sig. (2-tailed) <0.001 <0.001 0.402 0.117

n 44 44 44 44

Total Time for Sustained Overhead Work Pearson Correlation –0.07 –0.21

Sig. (2-tailed) 0.665 0.188

n 42 42

		Peak Waist	Average Waist	Peak Wrist	Average Wrist
Floor to Waist Max Weight	Pearson Correlation	0.40	0.45	0.18	0.18
	Sig. (2-tailed)	0.005	0.002	0.230	0.245
	n	46	46	46	46
Waist to Crown Max Weight	Pearson Correlation	0.39	0.39	0.15	0.44
	Sig. (2-tailed)	0.008	0.007	0.314	0.002
	n	46	46	46	46
Front Carry	Pearson Correlation	0.57	0.64	–0.13	0.24
Max Weight	Sig. (2-tailed)	<0.001	<0.001	0.402	0.117
	n	44	44	44	44
Total Time for Sustained Overhead Work	Pearson Correlation		–0.07		–0.21
	Sig. (2-tailed)		0.665		0.188
	n		42		42

Correlations were also calculated between weighted overhead work time with average vector magnitude from both the dominant side waist and non-dominant side wrist placement. Peak was not calculated as this exercise was endurance and not maximal ability based. A non-significant correlation was found between total time in minutes and both average vector magnitude from the waist and wrist placement. These results are presented in Table 3.

Distance covered during the six-minute walk test was significantly correlated with total activity count (sum of vector magnitudes over the full 6-minute period) from the waist placement (r = 0.66). A non-significant correlation was observed between total distance walked and total counts from the wrist placement. Results are presented in Table 4.

Table 4

Correlations between Six-Minute Walk Test distance and total activity counts from waist and wrist Actigraph placements

		Total activity counts - waist	Total activity counts - wrist
Total Distance	Pearson Correlation	0.66	0.23
	Sig. (2-tailed)	<0.0001	0.163
	N	39	39

3.4 Intraclass correlation between Actigraph device placements

Output from the Actigraph Devices placed on the waist and wrist were compared using ICC (for all 5 exercises. ICCs are calculated for two-way random effect models, consistency agreement and average measures. For floor to waist lift and front carry exercises the peak vector magnitude between the two placements were found to have poor agreement (ICC = 0.36, ICC = 0.24 respectively). For waist to crown lift peak vector magnitudes between the two placements were found to have fair agreement (ICC = 0.45). Average vector magnitudes for the Actigraph placements of the floor to waist lift and waist to crown lift were found to have fair agreement (ICC = 0.53, ICC = 0.59) respectively, while good agreement was found for the front carry lifts (ICC = 0.70), although the lower boundary of the 95% confidence intervals at times dropped into the poor range (<0.3). Results are presented in Table 5.

Table 5
Intraclass Correlation Coefficients between two placements of Actigraph during three lifting exercises, sustained overhead work, and 6MWT

Intraclass Correlation 95% Confidence Interval

Lower Bound Upper Bound

Floor to waist using peak VM 0.36 0 0.64

Waist to crown using peak VM 0.45 0.01 0.70

Front carry using peak VM 0.24 0 0.59

Floor to waist using average VM 0.53 0.15 0.74

Waist to crown using average VM 0.59 0.26 0.77

Front carry using average VM 0.70 0.46 0.84

Weighted overhead work using average VM 0.30 0 0.62

6-minute walk test using total activity counts 0.27 0 0.62

	Intraclass Correlation	95% Confidence Interval
Floor to waist using peak VM	0.36	0	0.64
Waist to crown using peak VM	0.45	0.01	0.70
Front carry using peak VM	0.24	0	0.59
Floor to waist using average VM	0.53	0.15	0.74
Waist to crown using average VM	0.59	0.26	0.77
Front carry using average VM	0.70	0.46	0.84
Weighted overhead work using average VM	0.30	0	0.62
6-minute walk test using total activity counts	0.27	0	0.62

ICCs were also calculated for device placement for the weighted overhead work exercise and six-minute walk test. For the overheard work exercise only average vector magnitudes were used. Placement of the two Actigraph devices was found to have poor agreement (ICC≤0.4) for overhead work. For the six-minute walk test total activity counts were used in the analysis. Placement of the two devices was found to also have poor agreement (ICC≤0.4) for the 6-minute walk test. Results are presented in Table 5.

4 Discussion

4.1 Key findings

Four key findings emerged from this study. First, there is a positive, moderate correlation between Actigraph device vector magnitudes and maximum weight lifted when the Actigraph device was placed on the participant’s waist. Second, there was a non-significant correlation between Actigraph device vector magnitudes and performance on the sustained, weighted overhead work test. Thirdly, there was a positive, large correlation between total activity counts from the Actigraph device and total distance walked during the six-minute walk test when the Actigraph was worn on the participant’s waist. Finally, there was a poor agreement between Actigraph device vector magnitude scores from the two placements of the Actigraph device on the dominant side waist and non-dominant side wrist. Although our study was exploratory in nature, our study may have implications for use of these Actigraph devices in occupational rehabilitation and assessment of work ability.

Our hypothesis was that a positive correlation of moderate magnitude would be found between total activity counts and walk distance during the six-minute walk test. This was again only supported when the Actigraph device was worn on the participant’s waist. We thought that we would find a good agreement when comparing Actigraph outputs from the waist and wrist placements. Surprisingly, this was not supported. This indicates that the waist is likely the optimal placement of the Actigraph device for measuring level of activity during FCE testing. Placement on the wrist likely generated too much ‘noise’ for accurate measurement of activity levels, and the Actigraph does not appear to be a good measure of sustained postural activities.

Out of all the tasks conducted during the study, sustained overhead work was the one in which there was the most uncertainty. There was no previous literature found on endurance tasks and activity monitor output. As well, out of all the exercises this one had the most variability in terms of how participants could complete it. The same instructions were given to the participants in that they could chose the pattern in which the task was completed (i.e. nuts were moved on the bolts) and that the number of nuts moved was not being recorded. The screwing/unscrewing was designed more to be a distraction during the test of upper extremity muscle endurance. Most participants found a comfortable stance and stayed in that position. In regards to efficiency, it is of interest that although no significant correlations were found, both correlations were found to be negative. This would suggest that those with lower average vector magnitudes were able to complete the exercises for a longer period of time indicating that perhaps there was less overall movement of the upper extremity.

Findings also did not support our hypotheses that a good agreement would be found on the 6-minute walk test. The poor agreement found in the 6MWT came as a surprise but can likely be explained by variations in walking mechanics, specifically arm swinging. Generally during walking the arms swing contralaterally with the legs on a 1 : 1 ratio [26]. However, this is not always the case as mechanics can vary significantly between individuals. One specific participant was noted to be holding his hands behind his back while walking. Other motions such as adjusting clothing, stretching, or even bringing the hands up to cover a cough or sneeze would also have increased VM readings on the wrist device and in turn resulted in a weaker agreement.

To our knowledge this is the first study to compare data from a wearable fitness device and maximal performance on specific FCE items. Our study appears to be the first to examine using a wearable fitness device for short burst, maximal performance exercises as opposed to lifestyle activities. In the past, wearable devices have been used to determine cut points or activity levels on a day-to-day basis [5 , 15] or to compare energy expenditure results, [7 , 27] but we have not found studies that specifically compared device output and maximal performance. The study conducted by Stec and Rawson predicting energy expenditure was the only study in which Actigraph devices were placed on various positions on the body while conducting weight based exercises [7]. Our research was similar to that of Stec and Rawson in that both found the wrist placement of the Actigraph provided greater VM counts than the waist placement [7]. Stec and Rawson also found that net energy expenditure was significantly correlated with the VM counts at the waist but not the wrist, while we found that weight lifted was significantly correlated with VM counts at waist but not always at the wrist.

4.2 Limitations

There are a few important limitations inherent to this study. First, our sample consisted of all healthy individuals. FCEs are tests generally conducted on injured workers and this limits the applicability of our research to the intended target of these evaluations. With a healthy population, we did not have to be concerned with pain or psychosocial factors influencing the result. Due to the exploratory nature of our research we believe this was a necessary first step. However, given the activity limitations that come with age and physical conditions, there is a need to conduct future research with various population groups to further develop the association between Actigraph device outputs and functional performance. Second, our sample was collected using convenience sampling on a University campus and as such the mean age of the sample is not reflective of the mean age in the workforce. Naturally, those around a university campus are of a younger age and many have yet to join the workforce. We did not limit our study to these age ranges, they were, however, the most readily available given our location. With respect to the Actigraph measure of vector magnitudes, we used this measure since we were trying to assess full 3-dimensional motion. An alternate approach would have been to consider only X or Y-axis data, which may have led to different results as not every activity contains movements in all 3 planes. For example, the floor to waist lift would incorporate movement in the sagittal (Y) and frontal planes (Z), when placing the box on the shelf, but little movement in the transverse plane (X). However, because our raw data is available from the devices this could be analyzed in future studies.

4.3 Future research

To our knowledge this was the first study to examine the relationship between data collected from a wearable device and performance on work-related functional tasks. Further research should be conducted with a refined design to improve the scope of this research. First, a wider age range should be studied to examine similar results across a sample more representative of the workforce. Second, future studies should look to evaluate workers with injuries, beginning with minor injuries and progressing toward more severe or chronic disabling injuries or conditions. Third, to fully understand the applicability of these devices, research should be conducted comparing the Actigraph data from clinical functional evaluations and activity performed in the workplace. Actigraph data can be gathered while a worker completes an FCE and then continued to be gathered either in the community (work and home) or during a graduated return to work program. This could allow an objective measure of how much activity the worker is performing in these contexts to determine whether progressions could or should be made. However, future research is needed to determine the validity and usefulness of activity monitoring of injured workers within these contexts. These changes would increase the quality of future research while broadening the scope and applicability of our conclusions.

4.4 Conclusions

A significant correlation was observed between data from the G3TX tri-axial Actigraph device and performance on FCEs. In general, the waist placement of the Actigraph device appears more optimal than the wrist placement due to higher correlations observed with waist placement. Furthermore, average vector magnitudes were found to have a stronger correlation than peak vector magnitudes. Additionally, agreement between device placement (waist and wrist) ranged from poor to good agreement, again with average vector magnitude having a stronger agreement than peak vector magnitudes in the lifting exercises. These findings have important implications for trying to introduce new technology, specifically wearable devices, into clinical settings. With the rapid development of these products and the widespread acceptance, it is important to determine if, and how they should be introduced into clinical practice. Relatively little research has been conducting on this topic in the past and further research is require before definitive statements regarding the utility of these devices in conjunction with FCE can be made.

Conflict of interest

None to report.

References

Gibson

, Strong

. A conceptual framework of functional capacity evaluation for occupational therapy in work rehabilitation. Australian Occupational Therapy Journal. 2003;50(2):64–71.

Morgan

, Allison

, Duhon

. Heart rate changes in functional capacity evaluations in a workers’ compensation population. Work. 2012;42(2):253–7.

Gross

, Battié

, Cassidy

. The prognostic value of functional capacity evaluation in patients with chronic low back pain: part 1: timely return to work. Spine. 2004;29(8):914–9.

Matheson

, Isernhagen

, Hart

. Relationships among lifting ability, grip force, and return to work. Physical Therapy. 2002;82(3):249–56.

Santos-Lozano

, Santin-Medeiros

, Cardon

, Torres-Luque

, Bailon

, Bergmeir

, et al. ActiGraph GT3X: validation and determination of physical activity intensity cut points. International Journal of Sports Medicine. 2013;34(11):975–82.

Santos-Lozano

, Torres-Luque

, Marín

, Ruiz

, Lucia

, Garatachea

. Intermonitor Variability of GT3X Accelerometer. International Journal of Sports Medicine. 2012;33(12):994–9.

Stec

, Rawson

. Estimation of resistance exercise energy expenditure using triaxial accelerometry. Journal of strength and conditioning research / National Strength & Conditioning Association. 2012;26(5):1413–22.

Rothney

, Schaefer

, Neumann

, Choi

, Chen

. Validity of physical activity intensity predictions by ActiGraph, Actical, and RT3 accelerometers. Obesity. 2008;16(8):1946–52.

Kimura

, Ozasa

, Nomura

, Yoshioka

, Endo

. Estimation of muscle strength from ActiGraph data in Duchenne muscular dystrophy. Pediatr Int. 2014;56(5):748?752. doi:10.1111/ped.12348.

10.

Karantonis

, Narayanan

, Mathie

, Lovell

, Celler

. Implementation of a real-time human movement classifier using a triaxial accelerometer for ambulatory monitoring. IEEE Transactions On Information Technology In Biomedicine: A Publication Of The IEEE Engineering In Medicine And Biology Society. 2006;10(1):156–67.

11.

Bonomi

, Goris

AHC

, Bin

, Westerterp

. Detection of Type, Duration, and Intensity of Physical Activity Using an Accelerometer. Medicine & Science in Sports & Exercise. 2009;41(9):1770–7.

12.

PAR-Q & You. Canadian Society for Exercise Physiology; 2012. http://www.csep.ca/cmfiles/publications/parq/par-q.pdf.

13.

Hildebrand

, VT

VANH

, Hansen

, Ekelund

. Age group comparability of raw accelerometer output from wrist- and hip-worn monitors. Medicine and Science in Sports and Exercise. 2014;46(9):1816–24.

14.

Chen

, Acra

, Majchrzak

, Donahue

, Baker

, Clemens

, et al. Military metabolic monitoring. Predicting energy expenditure of physical activity using hip- and wrist-worn accelerometers. Diabetes Technology & Therapeutics. 2003;5(6):1023–33.

15.

Rosenberger

, Haskell

, Albinali

, Mota

, Nawyn

, Intille

. Estimating Activity and Sedentary Behavior From an Accelerometer on the Hip or Wrist. Medicine and Science in Sports and Exercise. 2013;45(5):964–75.

16.

John

, Freedson

. ActiGraph and Actical physical activity monitors: a peek under the hood. Medicine and Science in Sports and Exercise.S. 2012;44(1 Suppl 1):86–9.

17.

Reneman

, Fokkens

, Dijkstra

, Geertzen

JHB

, Groothoff

. Testing lifting capacity: Validity of determining effort level by means of observation. Spine. 2005;30(2):E40–E6.

18.

Matheson

. Test Instructions and Administration. Epic Lift Capacity Examiner’s Manual. 2003;54-62.

19.

Kuijer

, Gouttebarge

, Brouwer

, Reneman

, Frings-Dresen

. Are performance-based measures predictive of work participation in patients with musculoskeletal disorders? A systematic review. International Archives of Occupational and Environmental Health.. 2012;85(2):109–23.

20.

Bieniek

, Bethge

. The reliability of WorkWell Systems Functional Capacity Evaluation: a systematic review. BMC Musculoskeletal Disorders. 2014;15(1):106.

21.

Brouwer

, Reneman

, Dijkstra

, Groothoff

, Schellekens

JMH

, Göeken

LNH

. Test–Retest Reliability of the Isernhagen Work Systems Functional Capacity Evaluation in Patients with Chronic Low Back Pain. Journal of Occupational Rehabilitation. 2003;13(4):207–18.

22.

Bautmans

, Lambert

, Mets

. The six-minute walk test in community dwelling elderly: influence of health status. BMC Geriatrics. 2004;4(1):6.

23.

, Yin

, Yu

, Tsang

, So

, Wong

, et al. The six-minute walk test in healthy children: reliability and validity. European Respiratory Journal. 2005;25(6):1057–60.

24.

Portney

, Watkins

. Validity of measurements. Foundations of Clinical Research: Applications to Practice. 2000;69-86.

25.

McDowell

. Measuring health: a guide to rating scales and questionnaires. Oxford University Press. 2006:10–54.

26.

Meyns

, Bruijn

, Duysens

. The how and why of arm swing during human walking. Gait & Posture. 2013;38(4):555–62.

27.

Lee

, Kim

, Welk

. Validity of consumer-based physical activity monitors. Medicine and Science in Sports and Exercise. 2014;46(9):1840–8.

	Intraclass Correlation	95% Confidence Interval
		Lower Bound	Upper Bound
Floor to waist using peak VM	0.36	0	0.64
Waist to crown using peak VM	0.45	0.01	0.70
Front carry using peak VM	0.24	0	0.59
Floor to waist using average VM	0.53	0.15	0.74
Waist to crown using average VM	0.59	0.26	0.77
Front carry using average VM	0.70	0.46	0.84
Weighted overhead work using average VM	0.30	0	0.62
6-minute walk test using total activity counts	0.27	0	0.62