Related Experiment Video
Updated: Jun 16, 2026
Design and Analysis for Fall Detection System Simplification
Published on: April 6, 2020
Beyond the (Log)book: Comparing Accelerometer Nonwear Detection Techniques in Toddlers
Elyse Letts1, Sarah M da Silva1, Natascja Di Cristofaro1
1Child Health & Exercise Medicine Program, Department of Pediatrics, McMaster University, Hamilton, Canada.
Insights
Several automated methods accurately validate accelerometer wear time in toddlers, aiding physical activity research. These methods offer reliable alternatives to manual logbooks for distinguishing wear and nonwear time in young children.
Area of Science:
- Pediatrics
- Biomedical Engineering
- Physical Activity Research
Background:
- Accelerometers are widely used to measure physical activity and sedentary behavior in toddlers.
- Accurate wear time validation is crucial for reliable data, especially in young children.
- Previous studies have not systematically compared automated wear time validation strategies in toddlers.
Purpose of the Study:
- To compare fully automated accelerometer wear time validation methods (counts and raw data algorithms) against a semi-automated logbook criterion in toddlers.
- To evaluate the accuracy and equivalence of different automated methods for distinguishing wear and nonwear time.
Main Methods:
- 109 toddlers (12-35 months) wore an accelerometer for ~7 days.
- Parents maintained logbooks to record monitor removal and nap times.
- 15 automated nonwear detection methods were tested and compared to the semi-automated logbook criterion using accuracy and F1 scores.
Main Results:
- Accuracy and F1 scores for automated methods ranged from 86% to 95%.
- Five methods demonstrated equivalence to the 'AllWear' criterion (including sleep time).
- Only one method was equivalent to the 'AwakeWear' criterion, with mean absolute differences varying significantly.
Conclusions:
- Automated methods like 5min0count, 10min_0count, 30min_0count, Troiano60s, and Ahmadi offer high accuracy and equivalency compared to semi-automated logbook cleaning.
- These findings provide valuable insights for researchers selecting appropriate wear time validation strategies based on their study population and protocol.
- The study supports the use of specific automated methods for reliable accelerometer data processing in toddler physical activity research.
Background:
Accelerometers are increasingly used to measure physical activity and sedentary time in toddlers. Data cleaning or wear time validation can impact outcomes of interest, particularly in young children who spend less time awake. However, no study has systematically compared wear time validation strategies in toddlers. As such, the objective of this study is to compare different fully automated methods of distinguishing wear and nonwear time (counts and raw data algorithms) methods to the semi automated (counts with logbooks) criterion method in toddlers.
Methods:
We recruited 109 toddlers (age 12-35 months) as part of the iPLAY study to wear an ActiGraph w-GT3X-BT accelerometer on the right hip for ~7 consecutive days (removed for sleep and water activities). Parents completed a logbook to indicate monitor removal and nap times. We tested 15 nonwear detection methods grouped into four main categories: semi-automated logbook, consecutive 0 counts, modified consecutive 0 counts (Troiano and Choi) and raw data methods (van Hees and Ahmadi). Using semi-automated logbooks as the criterion standard (all wear and wake-time only wear), we calculated the accuracy and F1 scores (a metric which balances precision and recall) and compared overall wear time with a two one-sided test of equivalence.
Results:
Participant daily wear time ranged from 556 to 684 min/day depending on method. Accuracy and F1 score ranged from 86% to 95%. Five methods were considered equivalent to the AllWear nonwear criterion (true wear time including sleep-time wear), with only one equivalent to the AwakeWear criterion. Mean absolute differences were lower for the AllWear criterion but ranged from 49 to 192 min/day.
Conclusions:
The 5min0count, 10min_0count, 30min_0count, Troiano60s and Ahmadi methods provide high accuracy and equivalency when compared with semi-automated cleaning using logbooks. This paper provides insights and quantitative results that can help researchers decide which method may be the most appropriate given their population of interest, sample size and study protocol.
Related Concept Videos
01:12Measuring Acceleration Due to Gravity
A simple pendulum can be described as a point mass and a string. Meanwhile, a physical pendulum is any object whose oscillations are similar to a simple pendulum, but cannot be modeled as a point mass on a string because its mass is distributed over a larger area. The behavior of a physical pendulum can be modeled using the principles of...
01:10Relative Motion Analysis - Acceleration