A novel scaling methodology to reduce the biases associated with missing data from commercial activity monitors

O’Driscoll, Ruairi; Turicchi, Jake; Duarte, Cristiana; Michałowska, Joanna; Larsen, Sofus C.; Palmeira, António Labisa; Heitmann, Berit Lilienthal; Horgan, Graham; Stubbs, R. James

doi:10.1371/journal.pone.0235144

Cited by 7 publications

(4 citation statements)

References 53 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Despite the small bias, this is unlikely to be a feasible means of assessing free-living energy balance behaviors. Participant discomfort and sensor removal present additional biases (ie, missing data), which may require additional modeling approaches to address [48][49][50]. The threshold of practicality varies depending on the size, duration, computational resources, and specific aims of the research study.…”

Section: Xsl • Fomentioning

confidence: 99%

Comparison of the Validity and Generalizability of Machine Learning Algorithms for the Prediction of Energy Expenditure: Validation Study

O’Driscoll¹,

Turicchi²,

Hopkins³

et al. 2021

JMIR Mhealth Uhealth

Self Cite

View full text Add to dashboard Cite

Background Accurate solutions for the estimation of physical activity and energy expenditure at scale are needed for a range of medical and health research fields. Machine learning techniques show promise in research-grade accelerometers, and some evidence indicates that these techniques can be applied to more scalable commercial devices. Objective This study aims to test the validity and out-of-sample generalizability of algorithms for the prediction of energy expenditure in several wearables (ie, Fitbit Charge 2, ActiGraph GT3-x, SenseWear Armband Mini, and Polar H7) using two laboratory data sets comprising different activities. Methods Two laboratory studies (study 1: n=59, age 44.4 years, weight 75.7 kg; study 2: n=30, age=31.9 years, weight=70.6 kg), in which adult participants performed a sequential lab-based activity protocol consisting of resting, household, ambulatory, and nonambulatory tasks, were combined in this study. In both studies, accelerometer and physiological data were collected from the wearables alongside energy expenditure using indirect calorimetry. Three regression algorithms were used to predict metabolic equivalents (METs; ie, random forest, gradient boosting, and neural networks), and five classification algorithms (ie, k-nearest neighbor, support vector machine, random forest, gradient boosting, and neural networks) were used for physical activity intensity classification as sedentary, light, or moderate to vigorous. Algorithms were evaluated using leave-one-subject-out cross-validations and out-of-sample validations. Results The root mean square error (RMSE) was lowest for gradient boosting applied to SenseWear and Polar H7 data (0.91 METs), and in the classification task, gradient boost applied to SenseWear and Polar H7 was the most accurate (85.5%). Fitbit models achieved an RMSE of 1.36 METs and 78.2% accuracy for classification. Errors tended to increase in out-of-sample validations with the SenseWear neural network achieving RMSE values of 1.22 METs in the regression tasks and the SenseWear gradient boost and random forest achieving an accuracy of 80% in classification tasks. Conclusions Algorithms trained on combined data sets demonstrated high predictive accuracy, with a tendency for superior performance of random forests and gradient boosting for most but not all wearable devices. Predictions were poorer in the between-study validations, which creates uncertainty regarding the generalizability of the tested algorithms.

show abstract

Section: Xsl • Fomentioning

confidence: 99%

Comparison of the Validity and Generalizability of Machine Learning Algorithms for the Prediction of Energy Expenditure: Validation Study

O’Driscoll¹,

Turicchi²,

Hopkins³

et al. 2021

JMIR Mhealth Uhealth

Self Cite

View full text Add to dashboard Cite

show abstract

“…Similarly, we used 14 days of data close to each visit at month 6 and month 12. To reduce biases associated with missing data and maximize data retention, we used a scaling methodology previously developed using NoHoW data and described in detail elsewhere [ 24 ]. This means that all participants had exactly 24 hours of data available for each of the included days.…”

Section: Methodsmentioning

confidence: 99%

Substituting sedentary time with sleep or physical activity and subsequent weight‐loss maintenance

Larsen

O’Driscoll

Horgan

et al. 2022

Obesity

View full text Add to dashboard Cite

Objective: In this study, the associations between the substitution of sedentary time with sleep or physical activity at different intensities and subsequent weight-loss maintenance were examined.Methods: This prospective study included 1152 adults from the NoHoW trial who had achieved a successful weight loss of ≥5% during the 12 months prior to baseline and had BMI ≥25 kg/m 2 before losing weight. Physical activity and sleep were objectively measured during a 14-day period at baseline. Change in body weight was included as the primary outcome. Secondary outcomes were changes in body fat percentage and waist circumference. Cardiometabolic variables were included as exploratory outcomes.Results: Using isotemporal substitution models, no associations were found between activity substitutions and changes in body weight or waist circumference. However, the substitution of sedentary behavior with moderate-to-vigorous physical activity was associated with a decrease in body fat percentage during the first 6 months of the trial (À0.33% per 30 minutes higher moderate-to-vigorous physical activity [95% CI: À0.60% to À0.07%], p = 0.013).Conclusions: Sedentary behavior had little or no influence on subsequent weight-loss maintenance, but during the early stages of a weight-loss maintenance program,

show abstract

“…Age (continuous variable) and sex was also included. Objectively measured physical activity was measured using the Fitbit Charge 2 ( 33 , 34 ). The device was updated to the latest firmware and worn on the nondominant wrist of the participant.…”

Section: Methodsmentioning

confidence: 99%

Hair Cortisol Concentration, Weight Loss Maintenance and Body Weight Variability: A Prospective Study Based on Data From the European NoHoW Trial

et al. 2021

View full text Add to dashboard Cite

Several cross-sectional studies have shown hair cortisol concentration to be associated with adiposity, but the relationship between hair cortisol concentration and longitudinal changes in measures of adiposity are largely unknown. We included 786 adults from the NoHoW trial, who had achieved a successful weight loss of ≥5% and had a body mass index of ≥25 kg/m2 prior to losing weight. Hair cortisol concentration (pg/mg hair) was measured at baseline and after 12 months. Body weight and body fat percentage were measured at baseline, 6-month, 12-month and 18-month visits. Participants weighed themselves at home ≥2 weekly using a Wi-Fi scale for the 18-month study duration, from which body weight variability was estimated using linear and non-linear approaches. Regression models were conducted to examine log hair cortisol concentration and change in log hair cortisol concentration as predictors of changes in body weight, change in body fat percentage and body weight variability. After adjustment for lifestyle and demographic factors, no associations between baseline log hair cortisol concentration and outcome measures were observed. Similar results were seen when analysing the association between 12-month concurrent development in log hair cortisol concentration and outcomes. However, an initial 12-month increase in log hair cortisol concentration was associated with a higher subsequent body weight variability between month 12 and 18, based on deviations from a nonlinear trend (β: 0.02% per unit increase in log hair cortisol concentration [95% CI: 0.00, 0.04]; P=0.016). Our data suggest that an association between hair cortisol concentration and subsequent change in body weight or body fat percentage is absent or marginal, but that an increase in hair cortisol concentration during a 12-month weight loss maintenance effort may predict a slightly higher subsequent 6-months body weight variability.Clinical Trial RegistrationISRCTN registry, identifier ISRCTN88405328.

show abstract

A novel scaling methodology to reduce the biases associated with missing data from commercial activity monitors

Cited by 7 publications

References 53 publications

Comparison of the Validity and Generalizability of Machine Learning Algorithms for the Prediction of Energy Expenditure: Validation Study

Comparison of the Validity and Generalizability of Machine Learning Algorithms for the Prediction of Energy Expenditure: Validation Study

Substituting sedentary time with sleep or physical activity and subsequent weight‐loss maintenance

Hair Cortisol Concentration, Weight Loss Maintenance and Body Weight Variability: A Prospective Study Based on Data From the European NoHoW Trial

Contact Info

Product

Resources

About