A novel attachment-method for the use of accelerometers in physical activity measurement

Abstract

Accelerometry is the method-of-choice for the assessment of habitual physical activity (PA). However, a number of limitations have been identified for existing validated accelerometer attachment methods. One of the major limitations is that existing attachment sites have been highlighted as the cause to low participant compliance. Therefore, this thesis aimed to develop a novel attachment site using the popular research accelerometer (Actigraph GT3X+). Consequently, the design of the new site - attachment by a lanyard (resembling a necklace), worn around the neck; hanging down to the chest, underneath clothing, has been thoroughly tested in this thesis. Five studies are presented in this thesis to examine the feasibility (measurement validity and user acceptability) of this novel attachment-method. The first study (chapter 3) evaluated the validity of the raw accelerometer output (i.e. vector magnitude counts; VMC) measured by a neck-worn GT3X+. The results demonstrated that VMC measured by a neck-worn GT3X+ more closely resembled VMC measured by a waist-worn GT3X+. Errors were much lower (higher agreement) for the neck-worn GT3X+ (9.23 – 15.5%) compared to the wrist-worn GT3X+ (19.7 – 54.9%). User acceptability of a neck-, wrist- and waist-worn GT3X+ was also compared in study one. The results suggest that the neck-worn GT3X+ appeared more acceptable than the waist-worn GT3X+ based on observed ranking frequencies (first, second, and third choice respectively), with 29.5%, 36.4% and 34.1% selections for the neck; 47.7, 36.4% and 15.8% selections for the wrist; and 22.7%, 27.3% and 50.0% selections for the waist. The next two studies (chapters 4 & 5) evaluated the validity of a neck-worn GT3X+ to estimate two frequently used PA metrics (steps-counts and physical activity energy expenditure, PAEE), delimited to using existing GT3X+ data processing methods provided by the manufacturer. The results of study two (chapter 4) demonstrated step outputs obtained from waist-, wrist- and neck-worn accelerometers against the criterion of manual counts were generally not comparable. The results of study three (chapter 5) suggest that the neck-worn GT3X+ (overall mean percentage error = 34.9%) can be more accurate in estimating PAEE than the wrist-worn GT3X+ (overall mean percentage error = 92.8%). To improve the accuracy of a neck-worn GT3X+ for the estimation of PA, the last two studies (chapters 6 & 7) developed and evaluated neck-specific vector magnitude cut-points for measuring time spent in different physical activity intensities. Cut-points with good-to-excellent classification accuracy were developed for the neck-site (chapter 6). The leave-one-out cross validation analysis demonstrated that the cut-points developed for the new neck-site correctly identified 93.4%, 82.2%, 78.8% and 87.9% of Sedentary-, Light-, Moderate-, and Vigorous PA-minutes, respectively. The free-living validation revealed that a neck-worn GT3+ was more accurate at classifying minutes spent in different PA intensities when compared with a wrist-worn GT3X+ (chapter 7). In conclusion, the feasibility of a novel neck-worn accelerometer has been thoroughly demonstrated in this thesis and with specifically designed cut-points it was able to measure PA with acceptable accuracy (comparable with waist-measurements) during free-living studies, thus demonstrating is future research potential.