Accelerometer data reduction: A comparison of reduction algorithms with three ActiGraph analysis ] software packages

Purpose: The aim of this study was to compare common reduction algorithms for ActiGraph accelerometer data using three software packages in order to identify their impact on the number of subjects with valid data.

Methods: 200 adolescents from the UP&DOWN Study wore the GT1M and GT3X ActiGraph accelerometers during 7 days. Data were processed applying eight data reduction algorithms with three software packages (i.e. Actilife, Meterplus, and Propero). Sixty minutes of continuous zeros were used to define non-wearing time. Algorithms 1 to 4 required four valid days, whereas algorithms 5 to 8 required three valid days. Algorithms 2, 4, 6 and 8 also required at least one weekend day. Algorithms 1, 2, 5, and 6 defined a valid day as having accelerometer data for ten hours while algorithms 3, 4, 7, and 8 for eight hours. Percentages of valid cases were calculated and the McNemar test was used to determinate differences between software outputs.

Results: Significant differences were found between Actilife and Propero for all algorithms (P<0.001) with the exception of the algorithm 7. We also found significant differences in the results derived from Meterplus and Actilife in all algorithms (P<0.05). Significant differences between Propero and Meterplus for algorithms 1, 2, 6 and 8 were found (P<0.05) whereas no differences were found for the rest of algorithms.

Conclusions: Choice of the ActiGraph analysis software may have an important impact on the number of subjects with valid data, consequently in subsequent analyses and potentially in study results.