Statistical approaches to account for missing values in accelerometer data: Applications to modeling physical activity

Xu, Selene; Nelson, Sandahl H; Kerr, Jacqueline; Godbole, Suneeta; Patterson, Ruth E.; Merchant, Gina; Abramson, Ian; Staudenmayer, John

doi:10.1177/0962280216657119

Cited by 21 publications

(7 citation statements)

References 31 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…We simulated missingness evenly throughout the entire 24-hour period in relation to the observed patterns of missingness in the NoHoW trial. This is contrary to a previous study observing that missing data patterns more frequently occur at the beginning and end of the day [42]. It is of note that we utilised wrist-worn devices compared to the aforementioned study, which utilised hip worn accelerometers.…”

Section: Discussioncontrasting

confidence: 59%

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

et al. 2020

View full text Add to dashboard Cite

show abstract

Section: Discussioncontrasting

confidence: 59%

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

et al. 2020

View full text Add to dashboard Cite

show abstract

“…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

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

“…This resulted in the recoding of <7% of cases. As in previous publications [ 34 , 44 ], missing PA data from the FIT trial were accounted for using a weighted mixed model approach with variance weighting by the inverse of daily wear-time proportions to evaluate the effect of log-in duration on PA [ 45 ]. The weighted regression approach is an efficient approach for dealing with missing accelerometry data because cases with a higher proportion of missing wear time are down weighted compared with cases with less proportion of missing wear time, which has been shown to improve the precision in estimating accelerometry-estimated PA [ 45 ].…”

Section: Methodsmentioning

confidence: 99%

Engagement With Tailored Physical Activity Content: Secondary Findings From the Families Improving Together for Weight Loss Randomized Controlled Trial

Sweeney¹,

Wilson²,

Resnicow³

et al. 2023

J Med Internet Res

View full text Add to dashboard Cite

Background Web-based tailored interventions offer rich opportunities for improved access to and personalization of behavioral interventions. However, despite the promise of this approach, the engagement and underrepresentation of minority groups remain major issues. Objective This study evaluated whether engagement (log-in status and log-in duration) with different types of tailored behavioral content from the Families Improving Together for weight loss web-based intervention was associated with changes in moderate to vigorous physical activity (MVPA) among African American families with overweight or obesity. Methods Parent-adolescent dyads were randomized to a web-based tailored intervention or web-based health education comparison program. The web-based intervention (N=119) was completed by parents and targeted 6 weight-related behaviors to support their adolescent children’s weight loss goals (session contents included energy balance, fast food, fruits and vegetables, physical activity [PA], sedentary behavior, and sweetened beverages). MVPA was measured using accelerometers at baseline and after the intervention. Results Using a hierarchical approach, the log-in status and duration for each web-based session were used to evaluate the additive effects of engagement with different types of tailored behavioral content on MVPA after the web-based intervention. Among parents, logging in to the PA session was not associated with greater MVPA (B=−12.561, 95% CI −18.759 to −6.367), but MVPA increased with greater log-in duration for the PA (B=0.008, 95% CI 0.004-0.012) and sedentary behavior (B= 0.008, 95% CI 0.004-0.012) sessions. These results suggest that parents who logged in to the PA session had lower MVPA, but MVPA increased with greater log-in duration for the PA and sedentary behavior sessions. These associations remained even after accounting for engagement with other content sessions. However, these engagement effects did not translate to the adolescents. Conclusions The results of this study highlight the need to disentangle the impact of engagement with different tailored content to improve the efficacy of tailored web-based interventions, especially for promoting PA in African American families. Trial Registration ClinicalTrials.gov NCT01796067; https://clinicaltrials.gov/ct2/show/NCT01796067

show abstract

Statistical approaches to account for missing values in accelerometer data: Applications to modeling physical activity

Cited by 21 publications

References 31 publications

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

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

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

Engagement With Tailored Physical Activity Content: Secondary Findings From the Families Improving Together for Weight Loss Randomized Controlled Trial

Contact Info

Product

Resources

About