Introduction: Accelerometers are commonly used to assess time-use behaviors related to physical activity, sedentary behavior, and sleep; however, as new accelerometer technologies emerge, it is important to ensure consistency with previous devices. This study aimed to evaluate the concurrent validity of the commonly used accelerometer, ActiGraph GT3X+, and the relatively new Axivity AX3 (fastened to the lower back) for detecting physical activity intensity and body postures when using direct observation as the criterion measure. Methods: A total of 41 children (aged 6–16 years) and 33 adults (aged 28–59 years) wore both monitors concurrently while performing 10 prescribed activities under laboratory conditions. The GT3X+ data were categorized into different physical activity intensity and posture categories using intensity-based cut points and ActiGraph proprietary inclinometer algorithms, respectively. The AX3 data were first converted to ActiGraph counts before being categorized into different physical activity intensity categories, while activity recognition models were used to detect the target postures. Sensitivity, specificity, and the balanced accuracy for intensity and posture category classification were calculated for each accelerometer. Differences in balanced accuracy between the devices and between children and adults were also calculated. Results: Both accelerometers obtained 74–96% balanced accuracy, with the AX3 performing slightly better (∼4% higher, p < .01) for detecting postures and physical activity intensity. Error in both devices was greatest when contrasting sitting/standing, sedentary/light intensity, and moderate/light intensity. Conclusion: In comparison with the GT3X+ accelerometer, AX3 was able to detect various postures and activity intensities with slightly higher balanced accuracy in children and adults.
Browse
Concurrent Validity of ActiGraph GT3X+ and Axivity AX3 Accelerometers for Estimating Physical Activity and Sedentary Behavior
Leila Hedayatrad, Tom Stewart, and Scott Duncan
Concurrent Measurement of Global Positioning System and Event-Based Physical Activity Data: A Methodological Framework for Integration
Anna M.J. Iveson, Malcolm H. Granat, Brian M. Ellis, and Philippa M. Dall
Objective: Global positioning system (GPS) data can add context to physical activity data and have previously been integrated with epoch-based physical activity data. The current study aimed to develop a framework for integrating GPS data and event-based physical activity data (suitable for assessing patterns of behavior). Methods: A convenience data set of concurrent GPS (AMOD) and physical activity (activPAL) data were collected from 69 adults. The GPS data were (semi)regularly sampled every 5 s. The physical activity data output was presented as walking events, which are continuous periods of walking with a time-stamped start time and duration (to nearest 0.1 s). The GPS outcome measures and the potential correspondence of their timing with walking events were identified and a framework was developed describing data integration for each combination of GPS outcome and walking event correspondence. Results: The GPS outcome measures were categorized as those deriving from a single GPS point (e.g., location) or from the difference between successive GPS points (e.g., distance), and could be categorical, scale, or rate outcomes. Walking events were categorized as having zero (13% of walking events, 3% of walking duration), or one or more (52% of walking events, 75% of walking duration) GPS points occurring during the event. Additionally, some walking events did not have GPS points suitably close to allow calculation of outcome measures (31% of walking events, 22% of walking duration). The framework required different integration approaches for each GPS outcome type, and walking events containing zero or more than one GPS points.
Volume 3 (2020): Issue 4 (Dec 2020)
Accuracy of Wearable Trackers for Measuring Moderate- to Vigorous-Intensity Physical Activity: A Systematic Review and Meta-Analysis
Jessica Gorzelitz, Chloe Farber, Ronald Gangnon, and Lisa Cadmus-Bertram
Background: The evidence base regarding validity of wearable fitness trackers for assessment and/or modification of physical activity behavior is evolving. Accurate assessment of moderate- to vigorous-intensity physical activity (MVPA) is important for measuring adherence to physical activity guidelines in the United States and abroad. Therefore, this systematic review synthesizes the state of the validation literature regarding wearable trackers and MVPA. Methods: A systematic search of the PubMed, Scopus, SPORTDiscus, and Cochrane Library databases was conducted through October 2019 (PROSPERO registration number: CRD42018103808). Studies were eligible if they reported on the validity of MVPA and used devices from Fitbit, Apple, or Garmin released in 2012 or later or available on the market at the time of review. A meta-analysis was conducted on the correlation measures comparing wearables with the ActiGraph. Results: Twenty-two studies met the inclusion criteria; all used a Fitbit device; one included a Garmin model and no Apple-device studies were found. Moderate to high correlations (.7–.9) were found between MVPA from the wearable tracker versus criterion measure (ActiGraph n = 14). Considerable heterogeneity was seen with respect to the specific definition of MVPA for the criterion device, the statistical techniques used to assess validity, and the correlations between wearable trackers and ActiGraph across studies. Conclusions: There is a need for standardization of validation methods and reporting outcomes in individual studies to allow for comparability across the evidence base. Despite the different methods utilized within studies, nearly all concluded that wearable trackers are valid for measuring MVPA.
Comparison of Energy Expenditure and Step Count Measured by ActiGraph Accelerometers Among Dominant and Nondominant Wrist and Hip Sites
Kayla J. Nuss, Nicholas A. Hulett, Alden Erickson, Eric Burton, Kyle Carr, Lauren Mooney, Jacob Anderson, Ashley Comstock, Ethan J. Schlemer, Lucas J. Archambault, and Kaigang Li
Objective: To validate and compare the accuracy of energy expenditure (EE) and step counts measured by ActiGraph accelerometers (ACT) at dominant and nondominant wrist and hip sites. Methods: Thirty young adults (15 females, age 22.93 ± 3.30 years) wore four ActiGraph wGT3X accelerometers while walking and running on a treadmill for 7 min at seven different speeds (1.7, 2.5, 3.4, 4.2, 5.0, 5.5, and 6.0 mph). The EE from each ACT was calculated using the Freedson Adult equation, and the “worn on the wrist” option was selected for the wrist data. Indirect calorimetry and manually counted steps were used as criterion measures. Mean absolute percentage error and two one-sided test procedures for equivalence were used for the analyses. Results: All ACTs underestimated the EE with mean absolute percentage errors over 30% for wrist placement and over 20% for hip placement. The wrist-worn ACTs underestimated the step count with mean absolute percentage errors above 30% for both dominant and nondominant placements. The hip-worn ACTs accurately assessed steps for the whole sample and for women and men (p < .001 to .05 for two one-sided tests procedures), but not at speeds slower than 2.0 mph. Conclusion: Neither hip nor wrist placements assess EE accurately. More algorithms and methods to derive EE estimates from wrist-worn ACTs must be developed and validated. For step counts, both dominant and nondominant hip placements, but not wrist placements, lead to accurate results for both men and women.
Remote Monitoring of Cancer Patient Participation in a 12-Week Online Yoga Study: Challenges and Directions for Future Research
Ryan Eckert, Jennifer Huberty, Heidi Kosiorek, Shannon Clark-Sienkiewicz, Linda Larkey, and Ruben Mesa
Introduction: The delivery of online interventions in cancer patients/survivors has increased. The measurement of participation in online interventions is important to consider, namely, the challenges of the remote assessment of activity. The purpose of this study was to report the measures used to assess intervention compliance and other physical activity participation in two online yoga studies, the relationship between the multimethod measures used, and the ability of cancer patients to complete these measures. Methods: The methods described are of two online yoga studies (feasibility and pilot). Cancer patients were asked to participate in 60 min/week of online yoga for 12 weeks, complete a weekly yoga log, wear a Fitbit daily for 12 weeks, and complete a weekly physical activity log. Finally, Clicky®, a web analytics software, was used to track online yoga participation. Results: Eighty-four people participated across both studies, with 63/84 participating in online yoga, averaging 57.5 ± 33.2 min/week of self-reported yoga participation compared to 41.4 ± 26.1 min/week of Clicky® yoga participation (Lin concordance = 0.28). All 84 participants averaged 95.5 ± 111.8 min/week of self-reported moderate/vigorous physical activity compared with 98.1 ± 115.9 min/week of Fitbit-determined moderate/vigorous physical activity (Lin concordance = 0.33). Across both studies, 82.9% of the yoga logs were completed, the Fitbit was worn on 75.2% of the days, and 78.7% of the physical activity logs were completed. Conclusions: Weak relationships between self-report and objective measures were demonstrated, but the compliance rates were above 75% for the study measures. Future research is needed, investigating the intricacies of self-report physical activity participation in remote interventions and the validation of a gold standard measurement for online interventions.
Towards Automatic Modeling of Volleyball Players’ Behavior for Analysis, Feedback, and Hybrid Training
Fahim A. Salim, Fasih Haider, Dees Postma, Robby van Delden, Dennis Reidsma, Saturnino Luz, and Bert-Jan van Beijnum
Automatic tagging of video recordings of sports matches and training sessions can be helpful to coaches and players and provide access to structured data at a scale that would be unfeasible if one were to rely on manual tagging. Recognition of different actions forms an essential part of sports video tagging. In this paper, the authors employ machine learning techniques to automatically recognize specific types of volleyball actions (i.e., underhand serve, overhead pass, serve, forearm pass, one hand pass, smash, and block which are manually annotated) during matches and training sessions (uncontrolled, in the wild data) based on motion data captured by inertial measurement unit sensors strapped on the wrists of eight female volleyball players. Analysis of the results suggests that all sensors in the inertial measurement unit (i.e., magnetometer, accelerometer, barometer, and gyroscope) contribute unique information in the classification of volleyball actions types. The authors demonstrate that while the accelerometer feature set provides better results than other sensors, overall (i.e., gyroscope, magnetometer, and barometer) feature fusion of the accelerometer, magnetometer, and gyroscope provides the bests results (unweighted average recall = 67.87%, unweighted average precision = 68.68%, and κ = .727), well above the chance level of 14.28%. Interestingly, it is also demonstrated that the dominant hand (unweighted average recall = 61.45%, unweighted average precision = 65.41%, and κ = .652) provides better results than the nondominant (unweighted average recall = 45.56%, unweighted average precision = 55.45, and κ = .553) hand. Apart from machine learning models, this paper also discusses a modular architecture for a system to automatically supplement video recording by detecting events of interests in volleyball matches and training sessions and to provide tailored and interactive multimodal feedback by utilizing an HTML5/JavaScript application. A proof of concept prototype developed based on this architecture is also described.
Validation of Garmin Fenix 3 HR Fitness Tracker Biomechanics and Metabolics (VO2max)
Bryson Carrier, Andrew Creer, Lauren R. Williams, Timothy M. Holmes, Brayden D. Jolley, Siri Dahl, Elizabeth Weber, and Tyler Standifird
The purpose of this study was to determine the validity of the Garmin fēnix® 3 HR fitness tracker. Methods: A total of 34 healthy recreational runners participated in biomechanical or metabolic testing. Biomechanics participants completed three running conditions (flat, incline, and decline) at a self-selected running pace, on an instrumented treadmill while running biomechanics were tracked using a motion capture system. Variables extracted were compared with data collected by the Garmin fēnix 3 HR (worn on the wrist) that was paired with a chest heart rate monitor and a Garmin Foot Pod (worn on the shoe). Metabolic testing involved two separate tests; a graded exercise test to exhaustion utilizing a metabolic cart and treadmill, and a 15-min submaximal outdoor track session while wearing the Garmin. 2 × 3 analysis of variances with post hoc t tests, mean absolute percentage errors, Pearson’s correlation (R), and a t test were used to determine validity. Results: The fēnix kinematics had a mean absolute percentage errors of 9.44%, 0.21%, 26.38%, and 5.77% for stride length, run cadence, vertical oscillation, and ground contact time, respectively. The fēnix overestimated (p < .05) VO2max with a mean absolute percentage error of 8.05% and an R value of .917. Conclusion: The Garmin fēnix 3 HR appears to produce a valid measure of run cadence and ground contact time during running, while it overestimated vertical oscillation in every condition (p < .05) and should be used with caution when determining stride length. The fēnix appears to produce a valid VO2max estimate and may be used when more accurate methods are not available.
Bidirectional Day-to-Day Associations of Reported Sleep Duration With Accelerometer Measured Physical Activity and Sedentary Time Among Dutch Adolescents: An Observational Study
Nathalie Berninger, Gregory Knell, Kelley Pettee Gabriel, Guy Plasqui, Rik Crutzen, and Gill Ten Hoor
Objectives: To examine the bidirectional association of sleep duration with proportions of time spent in physical behaviors among Dutch adolescents. Methods: Adolescents (n = 294, 11–15 years) completed sleep diaries and wore an accelerometer (ActiGraph) over 1 week. With linear mixed-effects models, the authors estimated the association of sleep categories (short, optimal, and long) with the following day’s proportion in physical behaviors. With generalized linear mixed models with binomial distribution, the authors estimated the association of physical behavior proportions on sleep categories. Physical behavior proportions were operationalized using percentages of wearing time and by applying a compositional approach. All analyses were stratified by gender accounting for differing developmental stages. Results: For males (number of observed days: 345, n = 83), short as compared with optimal sleep was associated with the following day’s proportion spent in sedentary (−2.57%, p = .03, 95% confidence interval [CI] [−4.95, −0.19]) and light-intensity activities (1.96%, p = .02, 95% CI [0.27, 3.65]), which was not significant in the compositional approach models. Among females (number of observed days: 427, n = 104), long sleep was associated with the proportions spent in moderate- to vigorous-intensity physical activity (1.69%, p < .001, 95% CI [0.75, 2.64]) and in sedentary behavior (−3.02%, p < .01, 95% CI [−5.09, −0.96]), which was replicated by the compositional approach models. None of the associations between daytime activity and sleep were significant (number of obs.: 844, n = 204). Conclusions: Results indicate partial associations between sleep and the following day’s physical behaviors, and no associations between physical behaviors and the following night’s sleep.
Where to Place Which Sensor to Measure Sedentary Behavior? A Method Development and Comparison Among Various Sensor Placements and Signal Types
Roman P. Kuster, Daniel Baumgartner, Maria Hagströmer, and Wilhelmus J.A. Grooten
Background: Sedentary behavior (SB) is associated with several chronic diseases and office workers especially are at increased risk. SB is defined by a sitting or reclined body posture with an energy expenditure of ≤1.5 metabolic equivalents. However, current objective methods to measure SB are not consistent with its definition. There is no consensus on which sensor placement and type should be used. Aim: To compare the accuracy of newly developed artificial intelligence models for 15 sensor placements in combination with four signal types (accelerometer only/plus gyroscope and/or magnetometer) to detect posture and physical in-/activity during desk-based activities. Method: Signal features for the model development were extracted from sensor raw data of 30 office workers performing 10 desk-based tasks, each lasting 5 min. Direct observation (posture) and indirect calorimetry (in-/activity) served as the reference criteria. The best classification model for each sensor was identified and compared among the sensor placements, both using Friedman and post hoc Wilcoxon tests (p ≤ .05). Results: Posture was most accurately measured with a lower body sensor, while in-/activity was most accurately measured with an upper body or waist sensor. The inclusion of additional signal types improved the posture classification for some placements, while the acceleration signal already contained the relevant signal information for the in-/activity classification. Overall, the thigh accelerometer most accurately classified desk-based SB. Conclusion: This study favors, in line with previous work, the measurement of SB with a thigh-worn accelerometer and adds the information that this sensor is also accurate in measuring physical in-/activity while sitting and standing.
