Reliability and Concurrent Validity of GT1M and MTI ActiGraph Accelerometers using a Mechanical Setup

Introduction The MTI and GT1M (Actigraph,LLC, Florida) uniaxial accelerometers have been used interchangeably in field research, however their internal technologies are different. The GT1M detects acceleration via a polysilicon surface micromachined sensor while the MTI utilises a cantilevered beam sensor. Determining the reliability across accelerometers and the validity between models is important prior to their use in the field. The aim of this study was to determine the intra and inter-unit reliability of the MTI and the GT1M accelerometers and to investigate their concurrent validity using a mechanical set-up.

Methods A mechanical unit, based on a model by Brage et al [1], consisting of 3 wheels connected by horizontally placed linkage was developed. Six accelerometers could be positioned along their x-axis on the unit. Rotation of the wheels caused the accelerometers to move through a sinusoidal acceleration in two axes (anteroposterior and mediolateral).

The accelerometers were set to record acceleration at 1-s epochs. Six accelerometers were tested 3 times at 3 different frequencies: 3 Hz, 2 Hz and 1 Hz. Each test lasted 5 minutes. To test for validity 3 accelerometers of each model were tested simultaneously at each frequency.

Data were imported to SPSS (version 17.0) for analysis. To determine the intra-unit variability the standard error of measurement (SEM) and the coefficient of variation (CV) were calculated. To determine the inter-unit variability a intraclass correlation coefficient (ICC) was calculated. Pearson’s correlation and the mean difference with 95% limits of agreement (LOA) were calculated to determine the concurrent validity between models.

Results The mean intra-instrument CV for the GT1M was 0.66% while the MTI was 8.28%. The mean SEM for the GT1M was 0.62 and 2.42 for the MTI. The ICC for the GT1M was 0.91 while the MTI was 0.98. Pearson’s correlation was 0.94 (p<0.01) and the mean difference between models was 3.19 counts per second (cps) (95% LOA : 0.16-6.22 cps).

Discussion and Conclusion The results suggest that both models had high inter-unit reliability with an ICC value >0.8 [2]. The CV and SEM of the MTI accelerometer were larger than the GT1M suggesting it had lower intra-unit reliability. While there was a significant correlation between accelerometer models, there was a mean difference of 3.19 cps. This difference of 191.4 counts per minute may not be biologically meaningful when considering applying cut-points to data i.e. 3,200 counts per minute for moderate to vigorous activity in children [3].

References [1] Brage, S., Brage, N., Wedderkopp, N., Froberg, K. Reliability and validity of the Computer Science and Applications accelerometer in a mechanical setting. Meas Phys Educ Exerc Sci, 2003; 7(2):101-119. [2] Matthews, C. E. Calibration of accelerometer Output for Adults.Med Sci Sports Exerc, 2005; 37(Suppl 11):S512-S522. [3] Puyau, M. R., Adolph, A. L., Vohra, F. A., Butte, N. F. Validation and calibration of physical activity monitors in children.Obes Res, 2002; 10(3):150-157.