A sensitivity analysis on the variability in accelerometer data processing for monitoring physical activity

A standardized workflow for long-term longitudinal actigraphy data processing using one year of continuous actigraphy from the CAN-BIND Wellness Monitoring Study

Monitoring sleep and activity through wearable devices such as wrist-worn actigraphs has the potential for long-term measurement in the individual's own environment. Long periods of data collection require a complex approach, including standardized pre-processing and data trimming, and robust algorithms to address non-wear and missing data. In this study, we used a data-driven approach to quality control, pre-processing and analysis of longitudinal actigraphy data collected over the course of 1 year in a sample of 95 participants. We implemented a data processing pipeline using open-source packages for longitudinal data thereby providing a framework for treating missing data patterns, non-wear scoring, sleep/wake scoring, and conducted a sensitivity analysis to demonstrate the impact of non-wear and missing data on the relationship between sleep variables and depressive symptoms. Compliance with actigraph wear decreased over time, with missing data proportion increasing from a mean of 4.8% in the first week to 23.6% at the end of the 12 months of data collection. Sensitivity analyses demonstrated the importance of defining a pre-processing threshold, as it substantially impacts the predictive value of variables on sleep-related outcomes. We developed a novel non-wear algorithm which outperformed several other algorithms and a capacitive wear sensor in quality control. These findings provide essential insight informing study design in digital health research.

Back to school after lockdown: The effect of COVID-19 restrictions on children's device-based physical activity metrics

BACKGROUND

The coronavirus disease-2019 (COVID-19) pandemic and national lockdowns took away opportunities for children to be physically active. This study aimed to determine the effect of the COVID-19 lockdown on accelerometer-assessed physical activity (PA) in children in Wales.

METHODS

Eight hundred participants (8-18 years old), stratified by sex, age, and socio-economic status, wore Axivity AX3 accelerometers for 7 days in February 2021, during the lockdown, and in May 2021, while in school. Raw accelerometer data were processed in R-package GGIR, and cut-point data, average acceleration (AvAcc), intensity gradient, and the acceleration above which the most active X minutes are accumulated (MX) metrics were extracted. Linear mixed models were used to assess the influence of time-point, sex, age, and socioeconomic status (SES) on PA.

RESULTS

During lockdown, moderate-to-vigorous PA was 38.4 ± 24.3 min/day; sedentary time was 849.4 ± 196.6 min/day; mean ± SD. PA levels increased significantly upon return to school (all variables p < 0.001). While there were no sex differences during lockdown (p = 0.233), girls engaged in significantly less moderate-to-vigorous PA than boys once back in school (p < 0.001). Furthermore, boys had more favorable intensity profiles than girls (intensity gradient: p < 0.001), regardless of time-point. PA levels decreased with age at both time-points; upper secondary school girls were the least active group, with an average M30 of 195.2 mg (while in school).

CONCLUSION

The lockdown affected boys more than girls, as reflected by the disappearance of the typical sex difference in PA levels during lockdown, although these were re-established on return to school. Upper secondary school (especially girls) might need specific COVID-recovery intervention.

Evaluating the performance of raw and epoch non-wear algorithms using multiple accelerometers and electrocardiogram recordings

Accurate detection of accelerometer non-wear time is crucial for calculating physical activity summary statistics. In this study, we evaluated three epoch-based non-wear algorithms (Hecht, Troiano, and Choi) and one raw-based algorithm (Hees). In addition, we performed a sensitivity analysis to provide insight into the relationship between the algorithms’ hyperparameters and classification performance, as well as to generate tuned hyperparameter values to better detect episodes of wear and non-wear time. We used machine learning to construct a gold-standard dataset by combining two accelerometers and electrocardiogram recordings. The Hecht and Troiano algorithms achieved poor classification performance, while Choi exhibited moderate performance. Meanwhile, Hees outperformed all epoch-based algorithms. The sensitivity analysis and hyperparameter tuning revealed that all algorithms were able to achieve increased classification performance by employing larger intervals and windows, while more stringently defining artificial movement. These classification gains were associated with the ability to lower the false positives (type I error) and do not necessarily indicate a more accurate detection of the total non-wear time. Moreover, our results indicate that with tuned hyperparameters, epoch-based non-wear algorithms are able to perform just as well as raw-based non-wear algorithms with respect to their ability to correctly detect true wear and non-wear episodes.

Two Days of Measurement Provides Reliable Estimates of Physical Activity Poststroke: An Observational Study

OBJECTIVE

The aim of this study was to determine the duration of physical activity (PA) monitoring required for reliable measurements following stroke.

DESIGN

Single-center, prospective, observational study.

SETTING

PA was measured in a community setting.

PARTICIPANTS

Adults (N=70) poststroke.

MAIN OUTCOME MEASURES

The SenseWear armband was used to monitor PA for 5 days (≥10 hours wear per day).

DATA ANALYSIS

Variance among 2, 3, 4, and 5 days of consecutive measurements for PA variables was examined using intraclass correlation coefficients (ICCs). The minimum number of days to achieve acceptable reliability (ICC ≥0.8) was calculated. Differences between weekdays and weekend days were investigated using paired t tests and Wilcoxon signed rank tests.

RESULTS

Two days of measurement was sufficient to achieve an ICC ≥0.8 for daily averages of total energy expenditure, step count, and time spent sedentary (≤1.5 metabolic equivalent tasks [METs]) and in light (1.5-3 METs) and moderate- to vigorous-intensity (>3 METs) PA. At least 3 days were required to achieve an ICC ≥0.8 when investigating the number of and time spent in bouts (≥10 minutes) of moderate to vigorous PA and sedentary behavior. Participants took significantly more steps (P=.03) and spent more time in light PA (P=.03) on weekdays than weekends.

CONCLUSION

Following stroke, 2 days of measurement appears sufficient to represent habitual PA for many simple variables. Three or more days may be necessary for reliable estimates of bouts of PA and sedentary behavior. Consistent inclusion or exclusion of a weekend day is recommended for measuring step count and light PA. Short periods of monitoring provide reliable PA information and may make PA measurement more feasible in the clinical setting.

The Impact of Low Accelerometer Wear Time on the Estimates and Application of Sedentary Behavior and Physical Activity Data in Adults

BACKGROUND

This investigation sought to determine how accelerometer wear (1) biased estimates of sedentary behavior (SB) and physical activity (PA), (2) affected misclassifications for meeting the Physical Activity Guidelines for Americans, and (3) impacted the results of regression models examining the association between moderate to vigorous physical activity (MVPA) and a clinically relevant health outcome.

METHODS

A total of 100 participants [age: 20.6 (7.9) y] wore an ActiGraph GT3X+ accelerometer for 15.9 (1.6) hours per day (reference dataset) on the hip. The BOD POD was used to determine body fat percentage. A data removal technique was applied to the reference dataset to create individual datasets with wear time ranging from 15 to 10 hours per day for SB and each intensity of PA.

RESULTS

Underestimations of SB and each intensity of PA increased as accelerometer wear time decreased by up to 167.2 minutes per day. These underestimations resulted in Physical Activity Guidelines for Americans misclassification rates of up to 42.9%. The regression models for the association between MVPA and body fat percentage demonstrated changes in the estimates for each wear-time adherence level when compared to the model using the reference MVPA data.

CONCLUSIONS

Increasing accelerometer wear improves daily estimates of SB and PA, thereby also improving the precision of statistical inferences that are made from accelerometer data.

Comparison of tri-axial accelerometers step-count accuracy in slow walking conditions

Accelerometers have shown great promise and popularity for monitoring gait. However, the accuracy of accelerometers for gait analysis in slow walking conditions is largely unknown. In this study, we compared the accuracy of three accelerometers recommended for gait analysis - Axivity AX3, APDM Opal, and the Actigraph wGT3X-BT, by holding the step-count algorithm constant. We evaluated device accuracy in four minutes of treadmill walking at the speeds of 0.9m/s, 1.1m/s, and 1.3m/s. We constructed a symbolization of the gait data to count the steps using Piecewise Aggregate Approximation and compared the estimated step counts with observer counted steps from video recordings. Our results highlight the variation between the performance of devices - the Axivity AX3 provides more accurate step counts than the other two devices. In this, we provide evidence for future scientific teams to make decisions on selecting accelerometers which can more accurately measure steps taken at slower walking speeds, and suggest ways to improve the design of algorithms and accelerometers.

Examining Non-Linear Associations between Accelerometer-Measured Physical Activity, Sedentary Behavior, and All-Cause Mortality Using Segmented Cox Regression

Healthy adults are advised to perform at least 150 min of moderate-intensity physical activity weekly, but this advice is based on studies using self-reports of questionable validity. This study examined the dose-response relationship of accelerometer-measured physical activity and sedentary behaviors on all-cause mortality using segmented Cox regression to empirically determine the break-points of the dose-response relationship. Data from 7006 adult participants aged 18 or above in the National Health and Nutrition Examination Survey waves 2003-2004 and 2005-2006 were included in the analysis and linked with death certificate data using a probabilistic matching approach in the National Death Index through December 31, 2011. Physical activity and sedentary behavior were measured using ActiGraph model 7164 accelerometer over the right hip for 7 consecutive days. Each minute with accelerometer count <100; 1952-5724; and ≥5725 were classified as sedentary, moderate-intensity physical activity, and vigorous-intensity physical activity, respectively. Segmented Cox regression was used to estimate the hazard ratio (HR) of time spent in sedentary behaviors, moderate-intensity physical activity, and vigorous-intensity physical activity and all-cause mortality, adjusted for demographic characteristics, health behaviors, and health conditions. Data were analyzed in 2016. During 47,119 person-year of follow-up, 608 deaths occurred. Each additional hour per day of sedentary behaviors was associated with a HR of 1.15 (95% CI 1.01, 1.31) among participants who spend at least 10.9 h per day on sedentary behaviors, and each additional minute per day spent on moderate-intensity physical activity was associated with a HR of 0.94 (95% CI 0.91, 0.96) among participants with daily moderate-intensity physical activity ≤14.1 min. Associations of moderate physical activity and sedentary behaviors on all-cause mortality were independent of each other. To conclude, evidence from this study supported at least 15 min per day of moderate-intensity physical activity and no more than 10.9 h per day of sedentary behaviors as recommendations to reduce all-cause mortality.

Correlates of children's compliance with moderate-to-vigorous physical activity recommendations: a multilevel analysis

This study aimed to investigate the association between individual and school characteristics associated with the number of school days children comply with moderate-to-vigorous physical activity (MVPA) recommendations. Sample comprises 612 Portuguese children, aged 9-11 years, from 23 schools. Time spent in MVPA was measured by accelerometry, while individual-level correlates were obtained by anthropometry and questionnaires. School-level variables were collected by questionnaire, and accelerometer wear time and season were also considered. Maximum likelihood estimates of model parameters were obtained via a multilevel analysis with children as level-1, and school as level-2. Children who spent more time in sedentary activities and girls were less likely to comply with MVPA/daily. More mature children and those who use active transportation to school were more likely to attain the PA recommendation. Furthermore, greater accelerometer wear time and spring season increased the chance to achieve the recommended MVPA. In terms of school-level correlates, a greater number of available facilities was negatively associated with children MVPA compliance. Given the set of variables, our results showed that individual characteristics seem to be more relevant for children's compliance rates with PA/day than school context variables, which should be taken into account in the implementation of school policies and practices.

Practical considerations in using accelerometers to assess physical activity, sedentary behavior, and sleep

Increasingly, behavioral and epidemiological research uses activity-based measurements (accelerometry) to provide objective estimates of physical activity, sedentary behavior, and sleep in a variety of study designs. As interest in concurrently assessing these domains grows, there are key methodological considerations that influence the choice of monitoring instrument, analysis algorithm, and protocol for measuring these behaviors. The purpose of this review is to summarize evidence-guided information for 7 areas that are of importance in the design and interpretation of studies using actigraphy: (1) choice of cut-points; (2) impact of epoch length; (3) accelerometer placement; (4) duration of monitoring; (5) approaches for distinguishing sleep, nonwear times, and sedentary behavior; (6) role for a sleep and activity diary; and (7) epidemiological applications. Recommendations for future research are outlined and are intended to enhance the appropriate use of accelerometry for assessing physical activity, sedentary behavior, and sleep behaviors in research studies.