Refined two-regression model for the ActiGraph accelerometer.

Area of Science:

• Physical activity monitoring
• Biomedical engineering
• Human movement analysis

Background:

• The ActiGraph 2006 two-regression model has limitations in accurately classifying activity transitions within a minute.
• Misclassification of walking or running during activity starts/stops on the ActiGraph clock affects energy expenditure estimates.
• Accurate assessment of free-living physical activity is crucial for public health research.

Purpose of the Study:

• To refine the 2006 Crouter two-regression model for improved accuracy.
• To eliminate misclassification errors in walking or running detection during activity transitions.
• To enhance the estimation of free-living energy expenditure using accelerometry.

Main Methods:

• Forty-eight participants performed 18 different activities ranging from sedentary to vigorous.
• ActiGraph accelerometers were worn on the hip, and energy expenditure was measured using portable indirect calorimetry.
• A refined two-regression model was developed using 45 routines and validated with 15 routines, employing coefficient of variation (CV) for activity classification.

Main Results:

• New exponential and cubic regression equations were developed to predict METs (Metabolic Equivalent of Task) every 10 seconds.
• The refined method analyzes 10-second epochs and surrounding epochs to minimize CV.
• Cross-validation showed the refined method was not significantly different from measured METs, except for cycling, and demonstrated similar accuracy to the 2006 model for structured activities.

Conclusions:

• The refined two-regression model effectively addresses misclassification of transitional minutes on the ActiGraph clock.
• This improved model enhances the accuracy of free-living energy expenditure estimation.
• The study provides a more reliable method for analyzing physical activity patterns from accelerometry data.