Purpose: To compare the accuracy and precision of a hip-worn accelerometer to predict energy cost during structured activities across motor performance and disease conditions. Methods: 118 adults self-identifying as healthy (n = 44) and those with arthritis (n = 23), multiple sclerosis (n = 18), Parkinson’s disease (n = 17), and stroke (n = 18) underwent measures of motor performance and were categorized into groups: Group 1, usual; Group 2, moderate impairment; and Group 3, severe impairment. The participants completed structured activities while wearing an accelerometer and a portable metabolic measurement system. Accelerometer-predicted energy cost (metabolic equivalent of tasks [METs]) were compared with measured METs and evaluated across functional impairment and disease conditions. Statistical significance was assessed using linear mixed effect models and Bayesian information criteria to assess model fit. Results: All activities’ accelerometer counts per minute (CPM) were 29.5–72.6% less for those with disease compared with those who were healthy. The predicted MET bias was similar across disease, −0.49 (−0.71, −0.27) for arthritis, −0.38 (−0.53, −0.22) for healthy, −0.44 (−0.68, −0.20) for MS, −0.34 (−0.58, −0.09) for Parkinson’s, and −0.30 (−0.54, −0.06) for stroke. For functional impairment, there was a graded reduction in CPM for all activities: Group 1, 1,215 CPM (1,129, 1,301); Group 2, 789 CPM (695, 884); and Group 3, 343 CPM (220, 466). The predicted MET bias revealed similar results across the Group 1, −0.37 METs (−0.52, −0.23); Group 2, −0.44 METs (−0.60, −0.28); and Group 3, −0.33 METs (−0.55, −0.13). The Bayesian information criteria showed a better model fit for functional impairment compared with disease condition. Conclusion: Using functionality to improve accelerometer calibration could decrease variability and warrants further exploration to improve accelerometer prediction of physical activity.
Scott J. Strath, Taylor W. Rowley, Chi C. Cho, Allison Hyngstrom, Ann M. Swartz, Kevin G. Keenan, Julian Martinez, and John W. Staudenmayer
Supun Nakandala, Marta M. Jankowska, Fatima Tuz-Zahra, John Bellettiere, Jordan A. Carlson, Andrea Z. LaCroix, Sheri J. Hartman, Dori E. Rosenberg, Jingjing Zou, Arun Kumar, and Loki Natarajan
Background: Machine learning has been used for classification of physical behavior bouts from hip-worn accelerometers; however, this research has been limited due to the challenges of directly observing and coding human behavior “in the wild.” Deep learning algorithms, such as convolutional neural networks (CNNs), may offer better representation of data than other machine learning algorithms without the need for engineered features and may be better suited to dealing with free-living data. The purpose of this study was to develop a modeling pipeline for evaluation of a CNN model on a free-living data set and compare CNN inputs and results with the commonly used machine learning random forest and logistic regression algorithms. Method: Twenty-eight free-living women wore an ActiGraph GT3X+ accelerometer on their right hip for 7 days. A concurrently worn thigh-mounted activPAL device captured ground truth activity labels. The authors evaluated logistic regression, random forest, and CNN models for classifying sitting, standing, and stepping bouts. The authors also assessed the benefit of performing feature engineering for this task. Results: The CNN classifier performed best (average balanced accuracy for bout classification of sitting, standing, and stepping was 84%) compared with the other methods (56% for logistic regression and 76% for random forest), even without performing any feature engineering. Conclusion: Using the recent advancements in deep neural networks, the authors showed that a CNN model can outperform other methods even without feature engineering. This has important implications for both the model’s ability to deal with the complexity of free-living data and its potential transferability to new populations.
Stacey Alvarez-Alvarado and Gershon Tenenbaum
Inquiry of the psychological states during the exercise experience failed to fully account for the role of motivation to adhere and the disposition of exertion tolerance (ET). The current study expands the scope of the integrated cognitive–perceptual–affective framework by measuring the motivation to sustain effort in two physical tasks and accounting for ET. Thirty male participants performed cycling and isometric handgrip tasks to assess the progression of the rating of perceived exertion, attentional focus, affective responses, and motivation to adhere, along with an incremental workload. The ET was determined by a handgrip task time to voluntary exhaustion. The findings indicated significant time effects and linear trends for perceived exertion, attentional focus, affect, and perceived arousal but not motivation to adhere during the handgrip and cycling tasks. The ET played a key role in the integrity of the model, particularly in perceptual, attentional, and affective responses. The intended model serves to stimulate new research into adaptation mechanisms.
Bronwyn Clark, Elisabeth Winker, Matthew Ahmadi, and Stewart Trost
Accurate measurement of time spent sitting, standing, and stepping is important in studies seeking to evaluate interventions to reduce sedentary behavior. In this study, the authors evaluated the agreement in classification of these activities from three algorithms applied to thigh-worn ActiGraph accelerometers using predictions from the widely used activPAL device as a criterion. Participants (n = 29, 72% female, age 23–68 years) wore the activPAL3™ micro (processed by PAL software, version 7.2.32) and the ActiGraph™ GT9X accelerometer on the right front thigh concurrently for working hours on one full workday (7.2 ± 1.2 hr). ActiGraph output was classified via the three test algorithms: ActiGraph’s ActiLife software (inclinometer); an open source method; and, a machine-learning algorithm reported in the literature (Acti4). Performance at an instance level was evaluated by computing classification accuracy (F scores) for 15-s windows. The F scores showed high accuracy relative to the criterion for identifying sitting (96.7–97.1) and were 84.7–85.1 for identifying standing and 78.1–80.6 for identifying stepping. The four methods agreed strongly in total time spent sitting, standing, and stepping, with intraclass correlation coefficients of .96 (95% confidence interval [.92, .96]), .92 (95% confidence interval [.81, .96]), and .87 (95% confidence interval [.53, .95]) but sometimes overestimated sitting time and underestimated standing time relative to activPAL. These algorithms for identifying sitting, standing, and stepping from thigh-worn accelerometers provide estimates that are very similar to those obtained using the activPAL.
Katie Potter, Robert T. Marcotte, Greg J. Petrucci, Caitlin Rajala, Deborah E. Linder, and Laura B. Balzer
Given high rates of obesity and chronic disease in both people and dogs, it is important to understand how dogs and dog owners influence each other’s health, including physical activity (PA) levels. Research suggests that dog owners who walk their dogs are more likely to meet PA guidelines than those who do not, but few studies have investigated dog walking intensity or its contribution to dog owners’ total moderate-to-vigorous PA using accelerometry. Furthermore, no studies have examined the contribution of dog walking to dogs’ total PA or the relationship between dog and dog owner PA using accelerometers on dogs. The authors used accelerometers on 33 dog owner–dog pairs to investigate (a) the intensity of dog walking behavior, (b) the contribution of dog walking to dog owners’ overall moderate-to-vigorous PA and dogs’ overall PA, and (c) the correlation between dog and dog owner PA. Dog owners wore an ActiGraph accelerometer and logged all dog walking for 7 days; dogs wore a Fitbark activity monitor. On average, 64.1% (95% confidence interval [55.2, 73.1]) of daily dog walking was moderate to vigorous intensity, and dog walking accounted for 51.2% (95% confidence interval [44.1, 58.3]) of dog owners’ daily moderate-to-vigorous PA. Dog walking accounted for 41.2% (95% confidence interval [36.0, 46.4]) of dogs’ daily PA. Dog owners’ daily steps were moderately correlated (r = .54) with dogs’ daily activity points. These findings demonstrate the interdependence of dog and dog owner PA and can inform interventions that leverage the dog–owner bond to promote PA and health in both species.
Julian Martinez, Autumn E. Decker, Chi C. Cho, Aiden Doherty, Ann M. Swartz, John Staudenmayer, and Scott J. Strath
Purpose: To assess the convergent validity of body-worn wearable camera still images (IMGs) for determining posture compared with activPAL (AP) classifications. Methods: The participants (n = 16, mean age 46.7 ± 23.8 years, 9 F) wore an Autographer wearable camera and an AP during three 2-hr free-living visits. IMGs were on average 8.47 s apart and were annotated with output consisting of events, transitory states, unknown, and gaps. The events were annotations that matched AP classifications (sit, stand, and move), consisting of at least three IMGs; the transitory states were posture annotations fewer than three IMGs; the unknowns were IMGs that could not be accurately classified; and the gaps were the time between annotations. For the analyses, the annotation and AP output were converted to 1-s epochs. The total and average length of visits and events were reported in minutes. Bias and 95% confidence intervals for event posture times from IMGs to AP were calculated to determine accuracy and precision. Confusion matrices using total AP posture times were computed to determine misclassification. Results: Forty-three visits were analyzed, with a total visit and event time of 5,027.73 and 4,237.23 min, respectively, and the average visit and event lengths being 116.92 and 98.54 min, respectively. Bias was not statistically significant for sitting, but was significant for standing and movement (0.84, −6.87, and 6.04 min, respectively). From confusion matrices, IMGs correctly classified sitting, standing, and movement (85.69%, 54.87%, and 69.41%, respectively) of total AP time. Conclusion: Wearable camera IMGs provide a good estimation of overall sitting time. Future work is warranted to improve posture classifications and examine the validity of IMGs in assessing activity-type behaviors.
Elif Inan-Eroglu, Bo-Huei Huang, Leah Shepherd, Natalie Pearson, Annemarie Koster, Peter Palm, Peter A. Cistulli, Mark Hamer, and Emmanuel Stamatakis
Background: Thigh-worn accelerometers have established reliability and validity for measurement of free-living physical activity-related behaviors. However, comparisons of methods for measuring sleep and time in bed using the thigh-worn accelerometer are rare. The authors compared the thigh-worn accelerometer algorithm that estimates time in bed with the output of a sleep diary (time in bed and time asleep). Methods: Participants (N = 5,498), from the 1970 British Cohort Study, wore an activPAL device on their thigh continuously for 7 days and completed a sleep diary. Bland–Altman plots and Pearson correlation coefficients were used to examine associations between the algorithm derived and diary time in bed and asleep. Results: The algorithm estimated acceptable levels of agreement with time in bed when compared with diary time in bed (mean bias of −11.4 min; limits of agreement −264.6 to 241.8). The algorithm-derived time in bed overestimated diary sleep time (mean bias of 55.2 min; limits of agreement −204.5 to 314.8 min). Algorithm and sleep diary are reasonably correlated (ρ = .48, 95% confidence interval [.45, .52] for women and ρ = .51, 95% confidence interval [.47, .55] for men) and provide broadly comparable estimates of time in bed but not for sleep time. Conclusions: The algorithm showed acceptable estimates of time in bed compared with diary at the group level. However, about half of the participants were outside of the ±30 min difference of a clinically relevant limit at an individual level.
Oliver W.A. Wilson, Kelsey E. Holland, Lucas D. Elliott, Michele Duffey, and Melissa Bopp
Background: Investigating the impact of the COVID-19 pandemic on both physical activity (PA) and mental health is important to demonstrate the need for interventions. This study examined the apparent impact of the pandemic on college students’ PA, perceived stress, and depressive symptoms. Methods: From 2015 through 2020, data were collected at the beginning and end of the spring semester at a large Northeastern US university via an online survey assessing student demographics, PA, perceived stress, and depressive symptoms. Mixed ANOVA examined differences in PA and mental health changes over the spring semester between “normal” and COVID-19 circumstances. Two-way ANOVA examined the interaction between circumstance and changes in PA in relation to changes in mental health. Results: Participants (n = 1019) were predominately women and non-Hispanic white. There was a significant decline in PA and an increase in perceived stress under COVID-19, but not normal, circumstances and a significant increase in depressive symptoms under COVID-19, but not normal, circumstances among women. Conclusions: A significant decline in PA and mental health among college students occurred under COVID-19 circumstances, and PA did not appear to protect against deterioration in mental health. Proactive and innovative policies, programs, and practices to promote student health and well-being must be explored immediately.
Bryndan W. Lindsey, Ali Boolani, Justin J. Merrigan, Nelson Cortes, Shane V. Caswell, and Joel R. Martin
Background: The COVID-19 pandemic has changed our working environment and divided workers into essential or nonessential statuses. Employment status is a major factor determining the amount of physical activity performed. Our purpose was to understand how employment status affects physical activity and sitting time. Methods: Between April 13 and May 4, 2020, 735 full-time employed individuals responded to a survey investigating daily life and overall health during the COVID-19 pandemic. Participants reported how much physical activity they had performed in the last 7 days. Multiple linear regressions were performed for physical activity and sitting time. Results: Physical activity was not associated with employment status. An interaction effect between hours worked and employment status was found for sitting time. Conclusions: Employment status was not related to physical activity; however, it did affect the amount of time spent sitting, with nonessential employees sitting more and working more hours than essential employees. Because greater amounts of daily total sitting time have been associated with increased risk of all-cause mortality, it is important that increased sitting time be attenuated by greater physical activity.
Stephanie G. Kerrigan, Evan M. Forman, Dave Williams, Mitesh Patel, Caitlin Loyka, Fengqing Zhang, Ross D. Crosby, and Meghan L. Butryn
Background: Financial incentives and feedback on behavior offer promise for promoting physical activity. However, evidence for the effect of each of these techniques is inadequate. The present study evaluated the effects of daily versus weekly feedback and incentives contingent on reaching a daily walking goal versus noncontingent incentives in a 2 × 2 trial. Methods: Participants (N = 57) had a body mass index >25 kg/m2 and were insufficiently active. Participants received a daily walking goal that adapted weekly. Results: Participants receiving daily feedback increased daily steps (P = .03) more than those receiving weekly feedback. Participants receiving contingent incentives did not significantly increase steps (P = .12) more than those receiving noncontingent incentives. A trend-level effect (P = .09) suggested that there may be an interaction such that the combination of daily feedback and contingent incentives is most effective. Conclusions: Results indicate that feedback is an important component of remotely delivered PA interventions and that evaluating each component of low-intensity interventions may help to improve efficacy. Moreover, results indicate that possible synergistic effects of feedback and rewards should be investigated further to help optimize interventions.