In studies that compare physical activity between groups of individuals, it is common for physical activity to be quantified by step count, which is measured by accelerometers or other wearable devices. Missing step count data often arise in these settings and can lead to bias or imprecision in the estimated effect if handled inappropriately. Replacing each missing value in accelerometer data with a single value using the Expectation–Maximization (EM) algorithm has been advocated in the literature, but it can lead to underestimation of variances and could seriously compromise study conclusions. We compare the performance in terms of bias and variance of two missing data methods, the EM algorithm and Multiple Imputation (MI), through a simulation study where data are generated from a parametric model to reflect characteristics of a trial on physical activity. We also conduct a reanalysis of the 2019 MOVE-IT trial. The EM algorithm leads to an underestimate of the variance of effects of interest, in both the simulation study and the reanalysis of the MOVE-IT trial. MI should be the preferred approach to handling missing data in accelerometer, which provides valid point and variance estimates.
Browse
Missing Step Count Data? Step Away From the Expectation–Maximization Algorithm
Mia S. Tackney, Daniel Stahl, Elizabeth Williamson, and James Carpenter
Erratum. The 7th International Conference on Ambulatory Monitoring of Physical Activity and Movement
Agreement of Step-Based Metrics From ActiGraph and ActivPAL Accelerometers Worn Concurrently Among Older Adults
Eric T. Hyde, Steve Nguyen, Fatima Tuz-Zahra, Christopher C. Moore, Mikael Anne Greenwood-Hickman, Rod L. Walker, Loki Natarajan, Dori Rosenberg, and John Bellettiere
Purpose : Our study evaluated the agreement of mean daily step counts, peak 1-min cadence, and peak 30-min cadence between the hip-worn ActiGraph GT3X+ accelerometer, using the normal filter (AGN) and the low frequency extension (AGLFE), and the thigh-worn activPAL3 micro (AP) accelerometer among older adults. Methods: Nine-hundred and fifty-three older adults (≥65 years) were recruited to wear the ActiGraph device concurrently with the AP for 4–7 days beginning in 2016. Using the AP as the reference measure, device agreement for each step-based metric was assessed using mean differences (AGN − AP and AGLFE − AP), mean absolute percentage error (MAPE), and Pearson and concordance correlation coefficients. Results: For AGN − AP, the mean differences and MAPE were: daily steps −1,851 steps/day and 27.2%, peak 1-min cadence −16.2 steps/min and 16.3%, and peak 30-min cadence −17.7 steps/min and 24.0%. Pearson coefficients were .94, .85, and .91 and concordance coefficients were .81, .65, and .73, respectively. For AGLFE − AP, the mean differences and MAPE were: daily steps 4,968 steps/day and 72.7%, peak 1-min cadence −1.4 steps/min and 4.7%, and peak 30-min cadence 1.4 steps/min and 7.0%. Pearson coefficients were .91, .91, and .95 and concordance coefficients were .49, .91, and .94, respectively. Conclusions: Compared with estimates from the AP, the AGN underestimated daily step counts by approximately 1,800 steps/day, while the AGLFE overestimated by approximately 5,000 steps/day. However, peak step cadence estimates generated from the AGLFE and AP had high agreement (MAPE ≤ 7.0%). Additional convergent validation studies of step-based metrics from concurrently worn accelerometers are needed for improved understanding of between-device agreement.
The Assessment of 24-Hr Physical Behavior in Children and Adolescents via Wearables: A Systematic Review of Laboratory Validation Studies
Marco Giurgiu, Carina Nigg, Janis Fiedler, Irina Timm, Ellen Rulf, Johannes B.J. Bussmann, Claudio R. Nigg, Alexander Woll, and Ulrich W. Ebner-Priemer
Purpose: To raise attention to the quality of published validation protocols while comparing (in)consistencies and providing an overview on wearables, and whether they show promise or not. Methods: Searches from five electronic databases were included concerning the following eligibility criteria: (a) laboratory conditions with humans (<18 years), (b) device outcome must belong to one dimension of the 24-hr physical behavior construct (i.e., intensity, posture/activity type outcomes, biological state), (c) must include a criterion measure, and (d) published in a peer-reviewed English language journal between 1980 and 2021. Results: Out of 13,285 unique search results, 123 articles were included. In 86 studies, children <13 years were recruited, whereas in 26 studies adolescents (13–18 years) were recruited. Most studies (73.2%) validated an intensity outcome such as energy expenditure; only 20.3% and 13.8% of studies validated biological state or posture/activity type outcomes, respectively. We identified 14 wearables that had been used to validate outcomes from two or three different dimensions. Most (n = 72) of the identified 88 wearables were only validated once. Risk of bias assessment resulted in 7.3% of studies being classified as “low risk,” 28.5% as “some concerns,” and 71.5% as “high risk.” Conclusion: Overall, laboratory validation studies of wearables are characterized by low methodological quality, large variability in design, and a focus on intensity. No identified wearable provides valid results across all three dimensions of the 24-hr physical behavior construct. Future research should more strongly aim at biological state and posture/activity type outcomes, and strive for standardized protocols embedded in a validation framework.
Evaluation of Two Thigh-Worn Accelerometer Brands in Laboratory and Free-Living Settings
Alexander H.K. Montoye, Olivia Coolman, Amberly Keyes, Megan Ready, Jaedyn Shelton, Ethan Willett, and Brian C. Rider
Background: Given the popularity of thigh-worn accelerometers, it is important to understand their reliability and validity. Purpose: Our study evaluated laboratory validity and free-living intermonitor reliability of the Fibion monitor and free-living intermonitor reliability of the activPAL monitor. Free-living comparability of the Fibion and activPAL monitors was also assessed. Methods: Nineteen adult participants wore Fibion monitors on both thighs while performing 11 activities in a laboratory setting. Then, participants wore Fibion and activPAL monitors on both thighs for 3 days during waking hours. Accuracy of the Fibion monitor was determined for recognizing lying/sitting, standing, slow walking, fast walking, jogging, and cycling. For the 3-day free-living wear, outputs from the Fibion monitors were compared, with similar analyses conducted for the activPAL monitors. Finally, free-living comparability of the Fibion and activPAL monitors was determined for nonwear, sitting, standing, stepping, and cycling. Results: The Fibion monitor had an overall accuracy of 85%–89%, with high accuracy (94%–100%) for detecting prone and supine lying, sitting, and standing but some misclassification among ambulatory activities and for left-/right-side lying with standing. Intermonitor reliability was similar for the Fibion and activPAL monitors, with best reliability for sitting but poorer reliability for activities performed least often (e.g., cycling). The Fibion and activPAL monitors were not equivalent for most tested metrics. Conclusion: The Fibion monitor appears suitable for assessment of sedentary and nonsedentary waking postures, and the Fibion and activPAL monitors have comparable intermonitor reliability. However, studies using thigh-worn monitors should use the same monitor brand worn on the same leg to optimize reliability.
Impact of COVID-19 Pandemic on Physical Activity, Pain, Mood, and Sleep in Adults With Knee Osteoarthritis
Michael J. Rose, Michael P. LaValley, S. Reza Jafarzadeh, Kerry E. Costello, Nirali Shah, Soyoung Lee, Belinda Borrelli, Stephen P. Messier, Tuhina Neogi, and Deepak Kumar
Objective: To examine changes in physical activity, sleep, pain, and mood in people with knee osteoarthritis during the ongoing COVID-19 pandemic by leveraging an ongoing randomized clinical trial. Methods: Participants enrolled in a 12-month parallel two-arm randomized clinical trial (NCT03064139) interrupted by the COVID-19 pandemic wore an activity monitor (Fitbit Charge 3) and filled out custom weekly surveys rating knee pain, mood, and sleep as part of the study. Data from 30 weeks of the parent study were used for this analysis. Daily step count and sleep duration were extracted from activity monitor data, and participants self-reported knee pain, positive mood, and negative mood via surveys. Metrics were averaged within each participant and then across all participants for prepandemic, stay-at-home, and reopening periods, reflecting the phased reopening in the state of Massachusetts. Results: Data from 28 participants showed small changes with inconclusive clinical significance during the stay-at-home and reopening periods compared with prepandemic for all outcomes. Summary statistics suggested substantial variability across participants with some participants showing persistent declines in physical activity during the observation period. Conclusion: Effects of the COVID-19 pandemic on physical activity, sleep, pain, and mood were variable across individuals with osteoarthritis. Specific reasons for this variability could not be determined. Identifying factors that could affect individuals with knee osteoarthritis who may exhibit reduced physical activity and/or worse symptoms during major lifestyle changes (such as the ongoing pandemic) is important for providing targeted health-care services and management advice toward those that could benefit from it the most.
Investigating the Effects of Applying Different Actigraphy Processing Approaches to Examine the Sleep Data of Patients With Neuropathic Pain
Hannah J. Coyle-Asbil, Anuj Bhatia, Andrew Lim, and Mandeep Singh
Individuals suffering from neuropathic pain commonly report issues associated with sleep. To measure sleep in this population, researchers have used actigraphy. Historically, actigraphy data have been analyzed in the form of counts; however, due to the proprietary nature, many opt to quantify data in its raw form. Various processing techniques exist to accomplish this; however, it remains unclear how they compare to one another. This study sought to compare sleep measures derived using the GGIR R package versus the GENEActiv (GA) R Markdown tool in a neuropathic pain population. It was hypothesized that the processing techniques would yield significantly different sleep outcomes. One hundred and twelve individuals (mean age = 52.72 ± 13.01 years; 60 M) with neuropathic pain in their back and/or lower limbs were included. While simultaneously undergoing spinal cord stimulation, actigraphy devices were worn on the wrist for a minimum of 7 days (GA; 50 Hz). Upon completing the protocol, sleep outcome measures were calculated using (a) the GGIR R package and (b) the GA R Markdown tool. To compare these algorithms, paired-samples t tests and Bland–Altman plots were used to compare the total sleep time, sleep efficiency, wake after sleep onset, sleep onset time, and rise times. According to the paired-samples t test, the GA R Markdown yielded lower total sleep time and sleep efficiency and a greater wake after sleep onset, compared with the GGIR package. Furthermore, later sleep onset times and earlier rise times were reported by the GGIR package compared with the GA R Markdown.
CRIB: A Novel Method for Device-Based Physical Behavior Analysis
Paul R. Hibbing, Seth A. Creasy, and Jordan A. Carlson
Physical behaviors (e.g., sleep, sedentary behavior, and physical activity) often occur in sustained bouts that are punctuated with brief interruptions. To detect and classify these interrupted bouts, researchers commonly use wearable devices and specialized algorithms. Most algorithms examine the data in chronological order, initiating and terminating bouts whenever specific criteria are met. Consequently, the bouts may encapsulate or overlap with later periods that also meet the activation and termination criteria (i.e., alternative bout solutions). In some cases, it is desirable to compare these alternative bout solutions before making a final classification. Thus, comparison-focused algorithms are needed, which can be used in isolation or in concert with their chronology-focused counterparts. In this technical note, we present a comparison-focused algorithm called CRIB (Clustered Recognition of Interrupted Bouts). It uses agglomerative hierarchical clustering to facilitate the comparison of different bout solutions, with the final classification being made in favor of the smallest number of bouts that comply with user-specified criteria (i.e., limits on the number, individual duration, and cumulative duration of interruptions). For demonstration, we use CRIB to assess bouts of moderate to vigorous physical activity in accelerometer data from the National Health and Nutrition Examination Survey, and we include a comparison against results from two established chronology-focused algorithms. Our discussion explores strengths and limitations of CRIB, as well as potential considerations and applications for using it in future studies. An online vignette (https://github.com/paulhibbing/PBpatterns/blob/main/vignettes/CRIB.pdf) is available to assist users with implementing CRIB in R.
CHAP-Adult: A Reliable and Valid Algorithm to Classify Sitting and Measure Sitting Patterns Using Data From Hip-Worn Accelerometers in Adults Aged 35+
John Bellettiere, Supun Nakandala, Fatima Tuz-Zahra, Elisabeth A.H. Winkler, Paul R. Hibbing, Genevieve N. Healy, David W. Dunstan, Neville Owen, Mikael Anne Greenwood-Hickman, Dori E. Rosenberg, Jingjing Zou, Jordan A. Carlson, Chongzhi Di, Lindsay W. Dillon, Marta M. Jankowska, Andrea Z. LaCroix, Nicola D. Ridgers, Rong Zablocki, Arun Kumar, and Loki Natarajan
Background: Hip-worn accelerometers are commonly used, but data processed using the 100 counts per minute cut point do not accurately measure sitting patterns. We developed and validated a model to accurately classify sitting and sitting patterns using hip-worn accelerometer data from a wide age range of older adults. Methods: Deep learning models were trained with 30-Hz triaxial hip-worn accelerometer data as inputs and activPAL sitting/nonsitting events as ground truth. Data from 981 adults aged 35–99 years from cohorts in two continents were used to train the model, which we call CHAP-Adult (Convolutional Neural Network Hip Accelerometer Posture-Adult). Validation was conducted among 419 randomly selected adults not included in model training. Results: Mean errors (activPAL − CHAP-Adult) and 95% limits of agreement were: sedentary time −10.5 (−63.0, 42.0) min/day, breaks in sedentary time 1.9 (−9.2, 12.9) breaks/day, mean bout duration −0.6 (−4.0, 2.7) min, usual bout duration −1.4 (−8.3, 5.4) min, alpha .00 (−.04, .04), and time in ≥30-min bouts −15.1 (−84.3, 54.1) min/day. Respective mean (and absolute) percent errors were: −2.0% (4.0%), −4.7% (12.2%), 4.1% (11.6%), −4.4% (9.6%), 0.0% (1.4%), and 5.4% (9.6%). Pearson’s correlations were: .96, .92, .86, .92, .78, and .96. Error was generally consistent across age, gender, and body mass index groups with the largest deviations observed for those with body mass index ≥30 kg/m2. Conclusions: Overall, these strong validation results indicate CHAP-Adult represents a significant advancement in the ambulatory measurement of sitting and sitting patterns using hip-worn accelerometers. Pending external validation, it could be widely applied to data from around the world to extend understanding of the epidemiology and health consequences of sitting.
A Comparison of Wrist- Versus Hip-Worn ActiGraph Sensors for Assessing Physical Activity in Adults: A Systematic Review
Nolan Gall, Ruopeng Sun, and Matthew Smuck
Introduction: Wrist-worn accelerometer has gained popularity recently in commercial and research use for physical activity tracking. Yet, no consensus exists for standardized wrist-worn data processing, and physical activity data derived from wrist-worn accelerometer cannot be directly compared with data derived from the historically used hip-worn accelerometer. In this work, through a systematic review, we aim to identify and analyze discrepancies between wrist-worn versus hip-worn ActiGraph accelerometers in measuring adult physical activity. Methods: A systematic review was conducted on studies involving free-living data comparison between hip- and wrist-worn ActiGraph accelerometers among adult users. We assessed the population, study protocols, data processing criteria (axis, epoch, wear-time correction, etc.), and outcome measures (step count, sedentary activity time, moderate-to-vigorous physical activity, etc.). Step count and activity count discrepancy were analyzed using meta-analysis, while meta-analysis was not attempted for others due to heterogeneous data processing criteria among the studies. Results: We screened 235 studies with 19 studies qualifying for inclusion in the systematic review. Through meta-analysis, the wrist-worn sensor recorded, on average, 3,537 steps/day more than the hip-worn sensor. Regarding sedentary activity time and moderate-to-vigorous physical activity estimation, the wrist sensor consistently overestimates moderate-to-vigorous physical activity time while underestimating sedentary activity time, with discrepancies ranging from a dozen minutes to several hours. Discussions: Our findings quantified the substantial discrepancies between wrist and hip sensors. It calls attention to the need for a cautious approach to interpreting data from different wear locations. These results may also serve as a reference for data comparisons among studies using different sensor locations.
