Validity of Individual versus Group Actigraph Prediction Equations During Structured Walking Sessions
- Presented on May 30, 2014
Background: Accelerometers are increasingly used to objectively estimate free-living energy expenditure (EE) in population-based studies. Published group-level calibration equations are typically used to convert accelerometer count output into applicable EE units. The use of subject-speciﬁc calibration equations has been proposed as a way to increase the accuracy of EE estimates but few studies to date have used this approach.
Purpose: To compare the validity of ActiGraph estimates of EE derived from individual calibration equations versus a published group-level calibration equation.
Methods: A convenience sample of 37 premenopausal women (aged 35 ± 0.8 years) completed two study visits over a 4-day period. During Visit 1 women completed calibration protocols for both the Actiheart monitor (AH) [standardized step-test] and ActiGraph GT3X+ (AG) [graded treadmill test at 2, 3, and 4 mph with VO2 measured by metabolic cart]. The AG calibration data was used to generate three different EE (kcal) prediction equations 1) regression of individual data (IND), 2) mixed model of individual and group level data (MIX), 3) regression of group level data (GRP). Participants then simultaneously wore AG and AH monitors while completing bouts of self-selected moderate and brisk paced walking on both an indoor track and outdoor community walking course. Estimated EE values for each bout and overall were then calculated using each of the derived AG equations and the published group-level AG equation of Sasaki et al. (2011).
Results: Based on our criterion measure (AH), mean EE across all bouts was 4.1 kcal/min. Compared to this estimate, the individually derived prediction equations resulted in less bias (+1.4-1.7 kcal/min) and were more accurate (Root Mean Square Error (RMSE): 1.8-2.3 kcal/min) than the published group-level equation (bias: +2.8 kcal/min; RMSE: 3.2 kcal/min).
Conclusion: Individual calibration of AG monitors may improve the precision of free-living estimates of EE. However, potential increases in precision must be weighed against the additional costs and subject burden associated with performing individual calibration testing.