Comparing the activPAL CREA and GHLA algorithms for the classification of postures and activity in free-living children

Duncan S. Buchan*, Ukadike C. Ugbolue

*Corresponding author for this work

Research output: Contribution to journalArticlepeer-review

1 Citation (Scopus)
21 Downloads (Pure)

Abstract

The activPAL accelerometer has been used extensively in research to assess sedentary behaviour (SB) and physical activity (PA) outcomes. The aim of this study was to assess the comparability of PA and SB outcomes from two automated algorithms (CREA and GHLA) applied to the activPAL accelerometer. One hundred and twenty participants aged 8–12 years wore an activPAL accelerometer on their right thigh continuously for seven days on two occasions, providing valid data from 1058 days. The PALbatch software downloaded the data after applying the CREA and GHLA (latest) algorithms. The comparability of the algorithms were assessed using the mean absolute percent error (MAPE), intra-class correlation coefficients (ICC), and equivalence testing. Comparisons for daily wear time, primary lying, sitting and standing time, sedentary and stepping time, upright time, total number of steps, sit–stand transitions and stepping time ≤ 1 min revealed mainly small MAPE (≤2%), excellent ICCs (lower bound 95% CI ≥ 0.97), and equivalent outcomes. Time spent in sitting bouts > 60 min and stepping bouts > 5 min were not equivalent with the absolute zone needed to reach equivalence (≥7%). Comparable outcomes were provided for wear time and postural outcomes using the CREA or GHLA algorithms, but not for time spent in sitting bouts > 60 min and stepping bouts > 5 min.

Original languageEnglish
Article number15962
Number of pages10
JournalInternational Journal of Environmental Research and Public Health
Volume19
Issue number23
DOIs
Publication statusPublished - 30 Nov 2022

Keywords

  • agreement
  • equivalence
  • free-living
  • sedentary behaviour
  • accelerometry

Fingerprint

Dive into the research topics of 'Comparing the activPAL CREA and GHLA algorithms for the classification of postures and activity in free-living children'. Together they form a unique fingerprint.

Cite this