Search

Walking cadence (steps/min) has emerged as a valid proxy of physical activity intensity, with consensus across numerous laboratory-based treadmill studies that ≥100 steps/min approximates absolutely defined moderate intensity (≥3 metabolic equivalents; METs). We recently reported that this cadence threshold had a classification accuracy of 73.3% for identifying moderate intensity during preferred pace overground walking in young adults. The purpose of this study was to evaluate and compare the performance of a cadence threshold of ≥100 steps/min for correctly classifying moderate intensity during overground walking in middle- and older-aged adults.Participants (N = 174, 48.3% female, 41-85 years of age) completed laboratory-based cross-sectional study involving an indoor 5-min overground walking trial at their preferred pace. Steps were manually counted and converted to cadence (total steps/5 min). Intensity was measured using indirect calorimetry and expressed as METs. Classification accuracy (sensitivity, specificity, accuracy) of a cadence threshold of ≥100 steps/min to identify individuals walking at ≥3 METs was calculated.The ≥100 steps/min threshold demonstrated accuracy of 74.7% for classifying moderate intensity. When comparing middle- vs. older-aged adults, similar accuracy (73.4% vs. 75.8%, respectively) and specificity (33.3% vs. 34.5%) were observed. Sensitivity was high, but was lower for middle- vs. older-aged adults (85.2% vs. 93.9%, respectively).A cadence threshold of ≥100 steps/min accurately identified moderate-intensity overground walking. Furthermore, accuracy was similar when comparing middle- and older-aged adults. These findings extend our previous analysis in younger adults and confirm the appropriateness of applying this cadence threshold across the adult lifespan.

This study aimed to assess the association of a wrist-worn, device-based metric of 24-h movement with cognitive function and subjective cognitive complaints among older adults, 60 yr and older.This is a cross-sectional analysis of the 2011-2012 and 2013-2014 National Health and Nutrition Examination Survey (NHANES) cycles. A wrist-worn ActiGraph GT3X+ accelerometer captured total 24-h movement activity, analyzed as Monitor-Independent Movement Summary units (MIMS-units), and quantified into placement based on an age- and sex-standardized percentile. Cognitive tests in the domains of memory, language/verbal fluency, and executive performance were administered. Test-specific cognitive z -scores were generated. Subjective cognitive complaints included perceived difficulty remembering and confusion/memory loss.The analytical sample included 2708 U.S. older adults (69.5 ± 0.2 yr, 55% female, 20.9% non-White). Multivariable linear regressions revealed those in quartiles 3 (50th-74th percentile) and 4 (≥75th percentile) for their age and sex had higher cognitive function z -scores across all domains compared with those in quartile 1. Logistic regressions demonstrated those in quartiles 3 and 4 also had lower odds of reporting difficulty remembering (adjusted odds ratio [AOR] = 0.52, 95% confidence interval [CI] = 0.31-0.89; AOR = 0.57, 95% CI = 0.37-0.88) and confusion/memory loss (AOR = 0.49, 95% CI = 0.27-0.91; AOR = 0.49, 95% CI = 0.27-0.98), respectively, compared with those in quartile 1.In a representative sample of U.S. older adults, higher cognitive functioning occurs among those that perform total 24-h movement activity at or above the 50th percentile for their age and sex. Future studies should consider movement behaviors across a 24-h period on cognitive health outcomes in older adults. More research exploring prospective associations of MIMS-units and time-use behaviors across midlife and older adulthood that may affect cognitive functioning across diverse populations is needed.

Step-based metrics provide simple measures of ambulatory activity, yet device software either includes undisclosed proprietary step detection algorithms or simply does not compute step-based metrics. We aimed to develop and validate a simple algorithm to accurately detect steps across various ambulatory and nonambulatory activities. Seventy-five adults (21–39 years) completed seven simulated activities of daily living (e.g., sitting, vacuuming, folding laundry) and an incremental treadmill protocol from 0.22 to 2.2 m/s. Directly observed steps were hand-tallied. Participants wore GENEActiv and ActiGraph accelerometers, one of each on their waist and on their nondominant wrist. Raw acceleration (g) signals from the anterior–posterior, medial–lateral, vertical, and vector magnitude directions were assessed separately for each device. Signals were demeaned across all activities and band-pass filtered (0.25, 2.5 Hz). Steps were detected via peak picking, with optimal thresholds (i.e., minimized absolute error from accumulated hand counted) determined by iterating minimum acceleration values to detect steps. Step counts were converted into cadence (steps/minute), and k-fold cross-validation quantified error (root mean squared error [RMSE]). We report optimal thresholds for use of either device on the waist (threshold = 0.0267g) and wrist (threshold = 0.0359g) using the vector magnitude signal. These thresholds yielded low error for the waist (RMSE

PURPOSE This study aimed to present age- and sex-specific percentiles for daily wrist-worn movement metrics in US youth and adults. This metric represents a summary of all recorded movement, regardless of the purpose, context, or intensity. METHODS Wrist-worn accelerometer data from the combined 2011-2014 National Health and Nutrition Examination Survey cycles and the 2012 National Health and Nutrition Examination Survey National Youth Fitness Survey were used for this analysis. Monitor-Independent Movement Summary units (MIMS-units) from raw triaxial accelerometer data were used. We removed the partial first and last assessment days and days with ≥5% nonwear time. Participants with ≥1 valid day were included. Mean MIMS-units were calculated across all valid days. Percentile tables and smoothed curves of daily MIMS-units were calculated for each age and sex using the Generalized Additive Models for Location Shape and Scale. RESULTS The analytical sample included 14,705 participants age ≥3 yr. The MIMS-unit activity among youth was similar for both sexes, whereas adult females generally had higher MIMS-unit activity than did males. Median daily MIMS-units peaked at age 6 yr for both sexes (males, 20,613; females, 20,706). Lowest activity was observed for males and females 80+ yr of age: 8799 and 9503, respectively. CONCLUSIONS Population referenced MIMS-unit percentiles for US youth and adults are a novel means of characterizing total activity volume. By using MIMS-units, we provide a standardized reference that can be applied across various wrist-worn accelerometer devices. Further work is needed to link these metrics to activity intensity categories and health outcomes.

Purpose: Develop a direct observation (DO) system to serve as a criterion measure for the calibration of models applied to free-living (FL) accelerometer data. Methods: Ten participants (19.4 ± 0.8 years) were video-recorded during four, one-hour FL sessions in different settings: 1) school, 2) home, 3) community, and 4) physical activity. For each setting, 10-minute clips from three randomly selected sessions were extracted and coded by one expert coder and up to 20 trained coders using the Observer XT software (Noldus, Wageningen, the Netherlands). The coder defines each whole-body movement which was further described with three modifiers: 1) locomotion, 2) activity type, and 3) MET value (used to categorize intensity level). Percent agreement was calculated for intra- and inter-rater reliability. For intra-rater reliability, the criterion coder coded all 12 clips twice, separated by at least one week between coding sessions. For inter-rater reliability, coded clips by trained coders were compared to the expert coder. Intraclass correlations (ICCs) were calculated to assess the agreement of intensity category for intra- and inter-rater comparisons described above. Results: For intra-rater reliability, mean percent agreement ranged from 91.9 ± 3.9% to 100.0 ± 0.0% across all variables in all settings. For inter-rater reliability, mean percent agreement ranged from 88.2 ± 3.5% to 100.0 ± 0.0% across all variables in all settings. ICCs for intensity category ranged from 0.74–1.00 and 0.81–1.00 for intra- and inter-rater comparisons, respectively. Conclusion: The DO system is reliable and feasible to serve as a criterion measure of FL physical activity in young adults to calibrate accelerometers, subsequently improving interpretation of surveillance and intervention research.

Context Insulin resistance is a risk factor for breast cancer recurrence. How exercise training changes fasting and postglucose insulin resistance in breast cancer survivors is unknown. Objective To evaluate exercise-induced changes in postglucose ingestion insulin concentrations, insulin resistance, and their associations with cancer-relevant biomarkers in breast cancer survivors. Setting The University of Massachusetts Kinesiology Department. Participants 15 postmenopausal breast cancer survivors not meeting the physical activity guidelines (150 min/week of exercise). Intervention A supervised 12-week aerobic exercise program (60 min/day, 3–4 days/week). Main outcome measures Postglucose ingestion insulin was determined by peak insulin and area under the insulin curve (iAUC) during a 5-sample oral glucose tolerance test. Insulin sensitivity was estimated from the Matsuda composite insulin sensitivity index (C-ISI). Changes in fitness and body composition were determined from submaximal VO2peak and dual energy X-ray absorptiometry. Results Participants averaged 156.8 ± 16.6 min/week of supervised exercise. Estimated VO2peak significantly increased (+2.8 ± 1.4 mL/kg/min, P < .05) and body weight significantly decreased (–1.1 ± 0.8 kg, P < .05) following the intervention. There were no differences in fasting insulin, iAUC, C-ISI, or peak insulin following the intervention. Insulin was only significantly lower 120 min following glucose consumption (68.8 ± 34.5 vs 56.2 ± 31.9 uU/mL, P < .05), and there was a significant interaction with past/present aromatase inhibitor (AI) use for peak insulin (–11.99 non-AI vs +13.91 AI uU/mL) and iAUC (-24.03 non-AI vs +32.73 AI uU/mL). Conclusions Exercise training had limited overall benefits on insulin concentrations following glucose ingestion in breast cancer survivors but was strongly influenced by AI use.

Higher insulin following sedentary behavior may be due to increased insulin secretion (IS), decreased hepatic insulin extraction (HIE), or a combination of both. Ten healthy adults completed glucose tolerance tests following 7 days of normal activity and 7 days of increased sitting. There were no differences in IS; however, HIE at 120 min after ingestion (85.4% ± 7.2% vs. 74.6% ± 6.6%, p < 0.05) and the area under the curve (73.6% ± 9.4% vs. 67.5% ± 11.3%, p < 0.05) were reduced following 7 days of increased sedentary time.

This study aimed to determine the validity of existing methods to estimate sedentary behavior (SB) under free-living conditions using ActiGraph GT3X+ accelerometers (AG).Forty-eight young (18-25 yr) adults wore an AG on the right hip and nondominant wrist and were video recorded during four 1-h sessions in free-living settings (home, community, school, and exercise). Direct observation videos were coded for postural orientation, activity type (e.g., walking), and METs derived from the Compendium of Physical Activities, which served as the criterion measure of SB (sitting or lying posture,1.5 METs). Thirteen methods using cut points from vertical counts per minute (CPM), counts per 15-s (CP15s), and vector magnitude (VM) counts (e.g., CPM1853VM), raw acceleration and arm angle (sedentary sphere), Euclidean norm minus one (ENMO) corrected for gravity (mg) thresholds, uni- or triaxial sojourn hybrid machine learning models (Soj1x and Soj3x), random forest (RF), and decision tree (TR) models were used to estimate SB minutes from AG data. Method bias, mean absolute percent error, and their 95% confidence intervals were estimated using repeated-measures linear mixed models.On average, participants spent 34.1 min per session in SB. CPM100, CPM150, Soj1x, and Soj3x were the only methods to accurately estimate SB from the hip. Sedentary sphere and ENMO44.8 overestimated SB by 3.9 and 6.1 min, respectively, whereas the remaining wrist methods underestimated SB (range, 9.5-2.5 min). In general, mean absolute percent error was lower using hip methods compared with wrist methods.Accurate group-level estimates of SB from a hip-worn AG can be achieved using either simpler count-based approaches (CPM100 and CPM150) or machine learning models (Soj1x and Soj3x). Wrist methods did not provide accurate or precise estimates of SB. The development of large open-source free-living calibration data sets may lead to improvements in SB estimates.

We evaluated the accuracy and precision of wearable activity trackers and a pedometer (ATPs) in estimating steps and sedentary time (ST) in free-living settings. Thirty-two healthy men and women (M...

OBJECTIVE Slow walking speed paired with increased energy cost is a strong predictor for mortality and disability in older adults but has yet to be examined in a heterogeneous sample (ie, age, sex, disease status). The aim of this study was to examine energy cost of slow and normal walking speeds among low- and normal-functioning adults. DESIGN Adults aged 20-90 yrs were recruited for this study. Participants completed a 10-m functional walk test at a self-selected normal walking speed and were categorized as low functioning or normal functioning based on expected age- and sex-adjusted average gait speed. Participants completed two successive 3-min walking stages, at slower than normal and normal walking speeds, respectively. Gas exchange was measured and energy cost per meter (milliliter per kilogram per meter) was calculated for both walking speeds. RESULTS Energy cost per meter was higher (P < 0.0001) in the low-functioning group (n = 76; female = 59.21%; mean ± SD age = 61.13 ± 14.68 yrs) during the slower than normal and normal (P < 0.0001) walking speed bouts compared with the normal-functioning group (n = 42; female = 54.76%; mean ± SD age = 51.55 ± 19.51 yrs). CONCLUSIONS Low-functioning adults rely on greater energy cost per meter of walking at slower and normal speeds. This has implications for total daily energy expenditure in low-functioning, adult populations.

Breaking up sitting with light physical activity (PA) is effective in reducing hyperglycemia in the laboratory. Whether the same effects are observed in the free-living environment remains unknown. We evaluated how daily and postprandial glycemia is impacted by 20, 40, or 60 min of activity performed as either breaks from sitting after each meal (BR) or as one continuous walk after breakfast (WALK). Thirty individuals with type 2 diabetes completed three experimental conditions [BR, WALK, and control (CON)] in a randomized crossover design. Conditions were performed in a free-living environment with strict dietary control over 7 days. Participants increased PA in BR and WALK by 20, 40, or 60 min ( n = 10 in each group) and maintained habitual levels of PA during CON. A continuous glucose monitor (iPro2) and activPAL activity monitor were worn to quantify glycemic control and PA. Using linear mixed models with repeated measures, we 1) compared postprandial glucose (PPG) across conditions and 2) assessed the relationship between activity volume and glucose responses. Whereas WALK tended to shorten the daily duration of hyperglycemia compared with CON ( P = 0.0875), BR was not different from CON. BR and WALK significantly attenuated the breakfast PPG versus CON ( P ≤ 0.05), but lunch and dinner PPG were unaffected by BR and WALK. In conclusion, continuous walking was more effective than breaks from sitting in lowering daily hyperglycemia for the group, but both conditions lowered breakfast PPG. In contrast to tightly controlled laboratory studies, breaks from sitting did not lower hyperglycemia in the free-living environment. NEW & NOTEWORTHY Our “ecolabical” approach is new and noteworthy. This approach combines the external validity of the free-living environment (ecological) with the control of key confounding variables in the laboratory and allows for highly translatable findings by minimizing confounding variables. We found that both postmeal continuous walking and short breaks from sitting similarly attenuated the postprandial glucose (PPG) response to breakfast. Unlike previous laboratory studies, neither condition (walk after breakfast or postmeal breaks) significantly impacted PPG at lunch or dinner.

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.

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.

Objective To determine classes of motor performance based on community deployable motor impairment and functional tests in a heterogeneous adult population. Design Sixteen tests of limb-specific and whole-body measures of motor impairment and function were obtained. Linear regression analysis was used to dichotomize performance on each test as falling within or outside the age- and sex-predicted values. Latent class analysis was used to determine 3 classes of motor performance. The chi-square test of association and the Fisher exact test were used for categorical variables, and analysis of variance and the Kruskal-Wallis test were used for continuous variables to evaluate the relationship between demographic characteristics and latent classes. Setting General community. Participants Individuals (N=118; 50 men) participated in the study. Quota sampling was used to recruit individuals who self-identified as healthy (n=44) or currently living with a preexisting chronic health condition, including arthritis (n=19), multiple sclerosis (n=18), Parkinson disease (n=17), stroke (n=18), or low functioning (n=2). Intervention Not applicable. Main Outcome Measure Latent classes of motor performance. Results Across the entire sample, 3 latent classes of motor performance were determined that clustered individuals with motor performance falling: (1) within predicted values on most of the tests (expected class), (2) outside predicted values on some of the tests (moderate class), and (3) outside predicted values on most of the tests (severe class).The ability to distinguish between the respective classes based on the percent chance of falling outside predicted values was achieved using the following community deployable motor performance tests: 10-meter walk test (22%, 80%, and 100%), 6-minute walk test (14.5%, 37.5%, and 100%), grooved pegboard test (23%, 38%, and 100%), and modified physical performance test (3%, 54%, and 96%). Conclusions In this heterogeneous group of adults, we found 3 distinct classes of motor performance, with the sample clustering into an expected test score group, a moderate test score deficiency group, and a severed test score deficiency group. Based on the motor performance tests, we established that community deployable, easily administered testing could accurately predict the established clusters of motor performance.

Background Heuristic cadence (steps/min) thresholds of ≥100 and ≥ 130 steps/min correspond with absolutely-defined moderate (3 metabolic equivalents [METs]; 1 MET = 3.5 mL O2·kg− 1·min− 1) and vigorous (6 METs) intensity, respectively. Scarce evidence informs cadence thresholds for relatively-defined moderate (≥ 64% heart rate maximum [HRmax = 220-age], ≥ 40%HR reserve [HRR = HRmax -HRresting, and ≥ 12 Rating of Perceived Exertion [RPE]); or vigorous intensity (≥ 77%HRmax, ≥ 60%HRR, and ≥ 14 RPE). Purpose To identify heuristic cadence thresholds corresponding with relatively-defined moderate and vigorous intensity in 21–60-year-olds. Methods In this cross-sectional study, 157 adults (40.4 ± 11.5 years; 50.6% men) completed up to twelve 5-min treadmill bouts, beginning at 0.5 mph and increasing by 0.5 mph. Steps were directly observed, HR was measured with chest-worn monitors, and RPE was queried in the final minute of each bout. Segmented mixed model regression and Receiver Operating Characteristic (ROC) curve analyses identified optimal cadence thresholds, stratified by age (21–30, 31–40, 41–50, and 51–60 years). Reconciliation of the two analytical models, including trade-offs between sensitivity, specificity, positive and negative predictive values, and overall accuracy, yielded final heuristic cadences. Results Across all moderate intensity indicators, the segmented regression models estimated optimal cadence thresholds ranging from 123.8–127.5 (ages 21–30), 120.2–126.0 (ages 31–40), 117.7–122.7 (ages 41–50), and 113.3–116.1 steps/min (ages 51–60). Corresponding values for vigorous intensity were 140.3–144.1, 139.6–142.6, 139.3–143.6, and 131.6–132.8 steps/min, respectively. ROC analysis estimated chronologically-arranged age groups’ cadence thresholds ranging from 114.5–118, 113.5–114.5, 104.6–112.9, and 103.6–106.0 across all moderate intensity indicators, and 124.5, 121.5, 117.2–122.2, and 113.0 steps/min, respectively, for vigorous intensity. Conclusions Heuristic cadence thresholds corresponding to relatively-defined moderate intensity for the chronologically-arranged age groups were ≥ 120, 120, 115, and 110 steps/min, regardless of the intensity indicator (i.e., % HRmax, %HRR, or RPE). Corresponding heuristic values for vigorous intensity indicators were ≥ 135, 130, 125, and 120 steps/min. These cadences are useful for predicting/programming intensity aligned with age-associated differences in physiological response to, and perceived experiences of, moderate and/or vigorous intensity. Trial registration Clinicaltrials.gov NCT02650258. Registered 24 December 2015.

PurposeTo compare estimates of moderate-vigorous physical activity (MVPA) duration derived from accelerometers calibrated only to walking and running activities to estimates from calibrations based on a broader range of lifestyle and ambulatory activities.MethodsIn a study of 932 older (50–7

A person's physical activity has important health implications, so it is important to be able to measure aspects of physical activity objectively. One approach to doing that is to use data from an accelerometer to classify physical activity according to activity type (e.g., lying down, sitting, standing, or walking) or intensity (e.g., sedentary, light, moderate, or vigorous). This can be formulated as a labeled classification problem, where the model relates a feature vector summarizing the accelerometer signal in a window of time to the activity type or intensity in that window. These data exhibit two key characteristics: (1) the activity classes in different time windows are not independent, and (2) the accelerometer features have moderately high dimension and follow complex distributions. Through a simulation study and applications to three datasets, we demonstrate that a model's classification performance is related to how it addresses these aspects of the data. Dynamic methods that account for temporal dependence achieve better performance than static methods that do not. Generative methods that explicitly model the distribution of the accelerometer signal features do not perform as well as methods that take a discriminative approach to establishing the relationship between the accelerometer signal and the activity class. Specifically, Conditional Random Fields consistently have better performance than commonly employed methods that ignore temporal dependence or attempt to model the accelerometer features.

Purpose: Determine the sensitivity of the Misfit Shine™ (MS) to detect changes in physical activity (PA) measures (steps, “points,” kCals) in laboratory (LAB) and free-living (FL) conditions. Methods: Twenty-one participants wore the MS and ActiGraph GT3X+™ accelerometer (AG) at the hip and dominant-wrist during three, one-hour LAB sessions: sedentary (SS), sedentary plus walking (SW), and sedentary plus jogging (SJ). Direct observation (DO) of steps served as the criterion measure. Devices were also worn during two FL conditions: 1) active week (ACT) and 2) inactive week (INACT). For LAB and FL, significant differences were examined using paired t-tests and linear mixed effects models, respectively. Linear mixed effects models were used to estimate differences between MS estimated steps and DO (α ≤ 0.05). Results: For all hip-worn MS measures and wrist-worn MS estimates of steps and “points,” there was a significant increase (p

We take a functional data approach to longitudinal studies with complex bivariate outcomes. This work is motivated by data from a physical activity study that measured two responses over time in five-minute intervals. One response is the proportion of time active in each interval, a continuous proportions with excess zeros and ones. The other response, energy expenditure rate in the interval, is a continuous variable with excess zeros and skewness. This outcome is complex because there are three possible activity patterns in each interval (inactive, partially active and completely active), and those patterns, which are observed, induce both non-random and random associations between the responses. More specifically, the inactive pattern requires a zero value in both the proportion for active behavior and the energy expenditure rate; a partially active pattern means that the proportion of activity is strictly between zero and one and that the energy expenditure rate is greater than zero and likely to be moderate, and the completely active pattern means that the proportion active is exactly one, and the energy expenditure rate is greater than zero and likely to be higher. To address these challenges, we propose a three-part functional data joint modeling approach. The first part is a continuation-ratio model to reorder the ordinal valued three activity patterns. The second part models the proportions when they are in interval (0, 1). The last component specifies the skewed continuous energy expenditure rate with Box-Cox transformations when they are greater than zero. In this three-part model, the regression structures are specified as smooth curves measured at various time-points with random effects that have a correlation structure. The smoothed random curves for each variable are summarized using a few important principal components, and the association of the three longitudinal components is modeled through the association of the principal component scores. The difficulties in handling the ordinal and proportional variables are addressed using a quasilikelihood type approximation. We develop an efficient algorithm to fit the model which also involves the selection of the number of principal components. The method is applied to physical activity data and is evaluated empirically by a simulation study.

The activPAL (AP) monitor is well established for distinguishing sitting, standing, and stepping time. However, its validity in predicting time in physical activity intensity categories in a free-living environment has not been determined.This study aimed to determine the validity of the AP in estimating time spent in sedentary, light, and moderate-to-vigorous physical activity (MVPA) in a free-living environment.Thirteen participants (mean ± SD age = 24.8 ± 5.2 yr, BMI = 23.8 ± 1.9 kg·m) were directly observed for three 10-h periods wearing an AP. A custom R program was developed and used to summarize detailed active and sedentary behavior variables from the AP. AP estimates were compared with direct observation.The AP accurately and precisely estimated time in activity intensity categories (bias [95% confidence interval]; sedentary = 0.8 min [-2.9 to 4.5], light = 1.7 min [2.2-5.7], and -2.6 min [-5.8 to 0.7]). The overall accuracy rate for time in intensity categories was 96.2%. The AP also accurately estimated guideline minutes, guideline bouts, prolonged sitting minutes, and prolonged sitting bouts.The AP can be used to accurately capture individualized estimates of active and sedentary behavior variables in free-living settings.

Body-worn devices that estimate physical behavior have tremendous potential to address key research gaps. However, there is no consensus on how devices and processing methods should be developed and evaluated, resulting in large differences in summary estimates and confusion for end users. We propose a phase-based framework for developing and evaluating devices that emphasizes robust validation studies in naturalistic conditions.

Physical inactivity is a recognized risk factor for many chronic diseases. Accelerometers are increasingly used as an objective means to measure daily physical activity. One challenge in using these devices is missing data due to device nonwear. We used a well-characterized cohort of 333 overweight postmenopausal breast cancer survivors to examine missing data patterns of accelerometer outputs over the day. Based on these observed missingness patterns, we created psuedo-simulated datasets with realistic missing data patterns. We developed statistical methods to design imputation and variance weighting algorithms to account for missing data effects when fitting regression models. Bias and precision of each method were evaluated and compared. Our results indicated that not accounting for missing data in the analysis yielded unstable estimates in the regression analysis. Incorporating variance weights and/or subject-level imputation improved precision by >50%, compared to ignoring missing data. We recommend that these simple easy-to-implement statistical tools be used to improve analysis of accelerometer data.

AB Purpose: Walking for health is recommended by health agencies, partly based on epidemiological studies of self-reported behaviors. Accelerometers are now replacing survey data, but it is not clear that intensity-based cut points reflect the behaviors previously reported. New computational techniques can help classify raw accelerometer data into behaviors meaningful for public health. Methods: Five hundred twenty days of triaxial 30-Hz accelerometer data from three studies (n = 78) were employed as training data. Study 1 included prescribed activities completed in natural settings. The other two studies included multiple days of free-living data with SenseCam-annotated ground truth. The two populations in the free-living data sets were demographically and physical different. Random forest classifiers were trained on each data set, and the classification accuracy on the training data set and that applied to the other available data sets were assessed. Accelerometer cut points were also compared with the ground truth from the three data sets. Results: The random forest classified all behaviors with over 80% accuracy. Classifiers developed on the prescribed data performed with higher accuracy than the free-living data classifier, but these did not perform as well on the free-living data sets. Many of the observed behaviors occurred at different intensities compared with those identified by existing cut points. Conclusions: New machine learning classifiers developed from prescribed activities (study 1) were considerably less accurate when applied to free-living populations or to a functionally different population (studies 2 and 3). These classifiers, developed on free-living data, may have value when applied to large cohort studies with existing hip accelerometer data.

AB Purpose: Accelerometers are a valuable tool for objective measurement of physical activity (PA). Wrist-worn devices may improve compliance over standard hip placement, but more research is needed to evaluate their validity for measuring PA in free-living settings. Traditional cut-point methods for accelerometers can be inaccurate and need testing in free living with wrist-worn devices. In this study, we developed and tested the performance of machine learning (ML) algorithms for classifying PA types from both hip and wrist accelerometer data. Methods: Forty overweight or obese women (mean age = 55.2 +/- 15.3 yr; BMI = 32.0 +/- 3.7) wore two ActiGraph GT3X+ accelerometers (right hip, nondominant wrist; ActiGraph, Pensacola, FL) for seven free-living days. Wearable cameras captured ground truth activity labels. A classifier consisting of a random forest and hidden Markov model classified the accelerometer data into four activities (sitting, standing, walking/running, and riding in a vehicle). Free-living wrist and hip ML classifiers were compared with each other, with traditional accelerometer cut points, and with an algorithm developed in a laboratory setting. Results: The ML classifier obtained average values of 89.4% and 84.6% balanced accuracy over the four activities using the hip and wrist accelerometer, respectively. In our data set with average values of 28.4 min of walking or running per day, the ML classifier predicted average values of 28.5 and 24.5 min of walking or running using the hip and wrist accelerometer, respectively. Intensity-based cut points and the laboratory algorithm significantly underestimated walking minutes. Conclusions: Our results demonstrate the superior performance of our PA-type classification algorithm, particularly in comparison with traditional cut points. Although the hip algorithm performed better, additional compliance achieved with wrist devices might justify using a slightly lower performing algorithm.