John Staudenmayer

Validation of Wearable Monitors for Assessing Sedentary Behavior

PURPOSE A primary barrier to elucidating the association between sedentary behavior (SB) and health outcomes is the lack of valid monitors to assess SB in a free-living environment. The purpose of this study was to examine the validity of commercially available monitors to assess SB. METHODS Twenty overweight (mean ± SD: body mass index = 33.7 ± 5.7 kg·m(-2)) inactive, office workers age 46.5 ± 10.7 yr were directly observed for two 6-h periods while wearing an activPAL (AP) and an ActiGraph GT3X (AG). During the second observation, participants were instructed to reduce sitting time. We assessed the validity of the commonly used cut point of 100 counts per minute (AG100) and several additional AG cut points for defining SB. We used direct observation (DO) using focal sampling with duration coding to record either sedentary (sitting/lying) or nonsedentary behavior. The accuracy and precision of the monitors and the sensitivity of the monitors to detect reductions in sitting time were assessed using mixed-model repeated-measures analyses. RESULTS On average, the AP and the AG100 underestimated sitting time by 2.8% and 4.9%, respectively. The correlation between the AP and DO was R2 = 0.94, and the AG100 and DO sedentary minutes was R2 = 0.39. Only the AP was able to detect reductions in sitting time. The AG 150-counts-per-minute threshold demonstrated the lowest bias (1.8%) of the AG cut points. CONCLUSIONS The AP was more precise and more sensitive to reductions in sitting time than the AG, and thus, studies designed to assess SB should consider using the AP. When the AG monitor is used, 150 counts per minute may be the most appropriate cut point to define SB.

An artificial neural network to estimate physical activity energy expenditure and identify physical activity type from an accelerometer

The aim of this investigation was to develop and test two artificial neural networks (ANN) to apply to physical activity data collected with a commonly used uniaxial accelerometer. The first ANN model estimated physical activity metabolic equivalents (METs), and the second ANN identified activity type. Subjects (n = 24 men and 24 women, mean age = 35 yr) completed a menu of activities that included sedentary, light, moderate, and vigorous intensities, and each activity was performed for 10 min. There were three different activity menus, and 20 participants completed each menu. Oxygen consumption (in ml x kg(-1) x min(-1)) was measured continuously, and the average of minutes 4-9 was used to represent the oxygen cost of each activity. To calculate METs, activity oxygen consumption was divided by 3.5 ml x kg(-1) x min(-1) (1 MET). Accelerometer data were collected second by second using the Actigraph model 7164. For the analysis, we used the distribution of counts (10th, 25th, 50th, 75th, and 90th percentiles of a minutes second-by-second counts) and temporal dynamics of counts (lag, one autocorrelation) as the accelerometer feature inputs to the ANN. To examine model performance, we used the leave-one-out cross-validation technique. The ANN prediction of METs root-mean-squared error was 1.22 METs (confidence interval: 1.14-1.30). For the prediction of activity type, the ANN correctly classified activity type 88.8% of the time (confidence interval: 86.4-91.2%). Activity types were low-level activities, locomotion, vigorous sports, and household activities/other activities. This novel approach of applying ANNs for processing Actigraph accelerometer data is promising and shows that we can successfully estimate activity METs and identify activity type using ANN analytic procedures.

Validity of two wearable monitors to estimate breaks from sedentary time.

UNLABELLED Investigations using wearable monitors have begun to examine how sedentary time behaviors influence health. PURPOSE The objective of this study is to demonstrate the use of a measure of sedentary behavior and to validate the activPAL (PAL Technologies Ltd., Glasgow, Scotland) and ActiGraph GT3X (Actigraph, Pensacola, FL) for estimating measures of sedentary behavior: absolute number of breaks and break rate. METHODS Thirteen participants completed two 10-h conditions. During the baseline condition, participants performed normal daily activity, and during the treatment condition, participants were asked to reduce and break up their sedentary time. In each condition, participants wore two ActiGraph GT3X monitors and one activPAL. The ActiGraph was tested using the low-frequency extension filter (AG-LFE) and the normal filter (AG-Norm). For both ActiGraph monitors, two count cut points to estimate sedentary time were examined: 100 and 150 counts per minute. Direct observation served as the criterion measure of total sedentary time, absolute number of breaks from sedentary time, and break rate (number of breaks per sedentary hour (brk·sed-h)). RESULTS Break rate was the only metric sensitive to changes in behavior between baseline (5.1 [3.3-6.8] brk·sed-h) and treatment conditions (7.3 [4.7-9.8] brk·sed-h) (mean (95% confidence interval)). The activPAL produced valid estimates of all sedentary behavior measures and was sensitive to changes in break rate between conditions (baseline, 5.1 [2.8-7.1] brk·sed-h; treatment, 8.0 [5.8-10.2] brk·sed-h). In general, the AG-LFE and AG-Norm were not accurate in estimating break rate or the absolute number of breaks and were not sensitive to changes between conditions. CONCLUSION This study demonstrates the use of expressing breaks from sedentary time as a rate per sedentary hour, a metric specifically relevant to free-living behavior, and provides further evidence that the activPAL is a valid tool to measure components of sedentary behavior in free-living environments.

General design Bayesian generalized linear mixed models

Linear mixed models are able to handle an extraordinary range of complications in regression-type analyses. Their most common use is to account for within-subject correlation in longitudinal data analysis. They are also the standard vehicle for smoothing spatial count data. However, when treated in full generality, mixed models can also handle spline-type smoothing and closely approximate kriging. This allows for nonparametric regression models (e.g., additive models and varying coefficient models) to be handled within the mixed model framework. The key is to allow the ran- dom effects design matrix to have general structure; hence our label general design. For continuous response data, particularly when Gaussianity of the response is reasonably assumed, computation is now quite mature and sup- ported by the R, SAS and S-PLUS packages. Such is not the case for bi- nary and count responses, where generalized linear mixed models (GLMMs) are required, but are hindered by the presence of intractable multivariate in- tegrals. Software known to us supports special cases of the GLMM (e.g., PROC NLMIXED in SAS or glmmML in R) or relies on the sometimes crude Laplace-type approximation of integrals (e.g., the SAS macro glimmix or glmmPQL in R). This paper describes the fitting of general design general- ized linear mixed models. A Bayesian approach is taken and Markov chain Monte Carlo (MCMC) is used for estimation and inference. In this gener- alized setting, MCMC requires sampling from nonstandard distributions. In this article, we demonstrate that the MCMC package WinBUGS facilitates sound fitting of general design Bayesian generalized linear mixed models in practice.

Accelerometer Output and MET Values of Common Physical Activities

PURPOSE This article 1) provides the calibration procedures and methods for metabolic and activity monitor data collection, 2) compares measured MET values to the MET values from the compendium of physical activities, and 3) examines the relationship between accelerometer output and METs for a range of physical activities. METHODS Participants (N = 277) completed 11 activities for 7 min each from a menu of 23 physical activities. Oxygen consumption (V O2) was measured using a portable metabolic system, and an accelerometer was worn. MET values were defined as measured METs (V O2/measured resting metabolic rate) and standard METs (V O2/3.5 mL.kg.min). For the total sample and by subgroup (age [young < 40 yr], sex, and body mass index [normal weight < 25 kg.m]), measured METs and standard METs were compared with the compendium, using 95% confidence intervals to determine statistical significance (alpha = 0.05). Average counts per minute for each activity and the linear association between counts per minute and METs are presented. RESULTS Compendium METs were different than measured METs for 17/21 activities (81%). The number of activities different than the compendium was similar between subgroups or when standard METs were used. The average counts for the activities ranged from 11 counts per minute (dishes) to 7490 counts per minute (treadmill: 2.23 m.s, 3%). The r between counts and METs was 0.65. CONCLUSIONS This study provides valuable information about data collection, metabolic responses, and accelerometer output for common physical activities in a diverse participant sample. The compendium should be updated with additional empirical data, and linear regression models are inappropriate for accurately predicting METs from accelerometer output.

Comparison of the ActiGraph 7164 and the ActiGraph GT1M during self-paced locomotion.

PURPOSE This study compared the ActiGraph accelerometer model 7164 (AM1) with the ActiGraph GT1M (AM2) during self-paced locomotion. METHODS Participants (n = 116, aged 18-73 yr, mean body mass index = 26.1 kg x m(-2)) walked at self-selected slow, medium, and fast speeds around an indoor circular hallway (0.47 km). Both activity monitors were attached to a belt secured to the hip and simultaneously collected data in 60-s epochs. To compare differences between monitors, the average difference (bias) in count output and steps output was computed at each speed. Time spent in different activity intensities (light, moderate, and vigorous) based on the cut points of Freedson et al. was compared for each minute. RESULTS The mean +/- SD walking speed was 0.7 +/- 0.22 m x s(-1) for the slow speed, 1.3 +/- 0.17 m x s(-1) for medium, and 2.1 +/- 0.61 m x s(-1) for fast speeds. Ninety-five percent confidence intervals (95% CI) were used to determine significance. Across all speeds, step output was significantly higher for the AM1 (bias = 19.8%, 95% CI = -23.2% to -16.4%) because of the large differences in step output at slow speed. The count output from AM2 was a significantly higher (2.7%, 95% CI = 0.8%-4.7%) than that from AM1. Overall, 96.1% of the minutes were classified into the same MET intensity category by both monitors. CONCLUSIONS The step output between models was not comparable at slow speeds, and comparisons of step data collected with both models should be interpreted with caution. The count output from AM2 was slightly but significantly higher than that from AM1 during the self-paced locomotion, but this difference did not result in meaningful differences in activity intensity classifications. Thus, data collected with AM1 should be comparable to AM2 across studies for estimating habitual activity levels.

Validity of the Omron HJ-112 Pedometer during Treadmill Walking

PURPOSE The purpose of this investigation was to examine the validity of step counts measured with the Omron HJ-112 pedometer and to assess the effect of pedometer placement. METHODS Ninety-two subjects (44 males and 48 females; 71 with body mass index [BMI] <30 kg.m and 21 with BMI >or=30 kg.m) completed three, 12-min bouts of treadmill walking at speeds of 1.12, 1.34, and 1.56 mxs. A subset (21 males and 23 females; 38 BMI <30 kg.m and 6 BMI >or=30 kg.m) completed a variable walking condition. For all conditions, participants wore an Omron HJ-112 pedometer on the hip, in the pants pocket, in the chest shirt pocket, and around the neck. Hip pedometer placement was alternated between right and left sides with the Yamax Digiwalker SW-701. During each walk, an investigator recorded actual steps with a manual hand counter. RESULTS There was no substantial bias with the Omron in any speed condition (-0.1% to 0.5%). Bias was larger with the Yamax (-3.6% to 2.0%). The largest random error for the Omron was 3.7% in the variable-speed condition for the BMI <30 kg.m group, whereas random errors for the Yamax were larger and up to 20%. None of the Omron placement positions produced statistically significant bias. Hip mounting produced the smallest random error (1.2%), followed by shirt pocket (1.7%), neck (2.2%), and pants pocket (5.8%). CONCLUSION The Omron HJ-112 pedometer validly assesses steps in different BMI groups during constant- and variable-speed walking; other than that in the pants pocket, placement of the pedometer has little effect on validity.

The Feasibility of Reducing and Measuring Sedentary Time among Overweight, Non-Exercising Office Workers

This study examined the feasibility of reducing free-living sedentary time (ST) and the convergent validity of various tools to measure ST. Twenty overweight/obese participants wore the activPAL (AP) (criterion measure) and ActiGraph (AG; 100 and 150 count/minute cut-points) for a 7-day baseline period. Next, they received a simple intervention targeting free-living ST reductions (7-day intervention period). ST was measured using two questionnaires following each period. ST significantly decreased from 67% of wear time (baseline period) to 62.7% of wear time (intervention period) according to AP (n = 14, P < 0.01). No other measurement tool detected a reduction in ST. The AG measures were more accurate (lower bias) and more precise (smaller confidence intervals) than the questionnaires. Participants reduced ST by ~5%, which is equivalent to a 48_min reduction over a 16-hour waking day. These data describe ST measurement properties from wearable monitors and self-report tools to inform sample-size estimates for future ST interventions.

Multisensor Data Fusion for Physical Activity Assessment

This paper presents a sensor fusion method for assessing physical activity (PA) of human subjects, based on support vector machines (SVMs). Specifically, acceleration and ventilation measured by a wearable multisensor device on 50 test subjects performing 13 types of activities of varying intensities are analyzed, from which activity type and energy expenditure are derived. The results show that the method correctly recognized the 13 activity types 88.1% of the time, which is 12.3% higher than using a hip accelerometer alone. Also, the method predicted energy expenditure with a root mean square error of 0.42 METs, 22.2% lower than using a hip accelerometer alone. Furthermore, the fusion method was effective in reducing the subject-to-subject variability (standard deviation of recognition accuracies across subjects) in activity recognition, especially when data from the ventilation sensor were added to the fusion model. These results demonstrate that the multisensor fusion technique presented is more effective in identifying activity type and energy expenditure than the traditional accelerometer-alone-based methods.

A Method to Estimate Free-Living Active and Sedentary Behavior from an Accelerometer

INTRODUCTION Methods to estimate physical activity (PA) and sedentary behavior (SB) from wearable monitors need to be validated in free-living settings. PURPOSE The purpose of this study was to develop and validate two novel machine-learning methods (Sojourn-1 Axis [soj-1x] and Sojourn-3 Axis [soj-3x]) in a free-living setting. METHODS Participants were directly observed in their natural environment for 10 consecutive hours on three separate occasions. Physical activity and SB estimated from soj-1x, soj-3x, and a neural network previously calibrated in the laboratory (lab-nnet) were compared with direct observation. RESULTS Compared with lab-nnet, soj-1x and soj-3x improved estimates of MET-hours (lab-nnet: % bias [95% confidence interval] = 33.1 [25.9 to 40.4], root-mean-square error [RMSE] = 5.4 [4.6-6.2]; soj-1x: % bias = 1.9 [-2.0 to 5.9], RMSE = 1.0 [0.6 to 1.3]; soj-3x: % bias = 3.4 [0.0 to 6.7], RMSE = 1.0 [0.6 to 1.5]) and minutes in different intensity categories {lab-nnet: % bias = -8.2 (sedentary), -8.2 (light), and 72.8 (moderate-to-vigorous PA [MVPA]); soj-1x: % bias = 8.8 (sedentary), -18.5 (light), and -1.0 (MVPA); soj-3x: % bias = 0.5 (sedentary), -0.8 (light), and -1.0 (MVPA)}. Soj-1x and soj-3x also produced accurate estimates of guideline minutes and breaks from sedentary time. CONCLUSIONS Compared with the lab-nnet algorithm, soj-1x and soj-3x improved the accuracy and precision in estimating free-living MET-hours, sedentary time, and time spent in light-intensity activity and MVPA. In addition, soj-3x is superior to soj-1x in differentiating SB from light-intensity activity.