Comparison Of Hip And Wrist Accelerometers In Pre-adolescents During Free-living And Semi-structured Physical Activity

Abstract

The dose-response relationship between physical activity (PA) and health is not well understood. The development of accelerometer-based motion sensors make it possible to objectively measure several dimensions of PA in free-living contexts. These devices have traditionally been worn on the hip, however there has been a recent trend to place these monitors on the wrist. PURPOSE: 1) To examine the accuracy of a hip (Evenson algorithm) and wrist-worn (Crouter algorithm) accelerometer for categorizing PA intensity in pre-adolescent girls during dance classes using direct observation (D.O.) as the criterion measure. 2) To compare the validity of the hip and wrist-worn accelerometer algorithms for classifying girls as meeting or not meeting PA guidelines. METHODS: For aim 1, participants (N = 6; Age = 10.22 ± 2.38) were video recorded while wearing hip and wrist accelerometers during a dance class. Data was analyzed using the Kruskal-Wallis Test. For aim 2, participants (N = 20; Age = 8.6 ± 1.6) wore a hip and wrist accelerometer concurrently for seven consecutive days. Fisher Exact Test was used to compare similarity between wrist and hip accelerometry data. RESULTS: For aim 1, compared to D.O., the wrist-worn accelerometer was inaccurate in measuring time spent in light PA (D.O. = 44.77 ± 6.82; wrist = 5.27 ± 4.98), vigorous PA (D.O. = 0.50 ± 1.01; wrist = 27.65 ± 22.87) and MVPA (D.O. = 6.59 ± 5.34; wrist = 44.14 ± 7.57). The hip-worn accelerometer was inaccurate in measuring time spent in sedentary time (D.O. = 1.39 ± 2.18; hip = 12.38 ± 8.25), light PA (D.O. = 44.77 ± 6.82; hip = 30.23 ± 5.47), vigorous PA (D.O. = 0.50 ± 1.01; hip = 4.05 ± 3.56) and total PA (D.O. = 51.36 ± 2.19; hip = 40.46 ± 8.25). For aim 2, there was no location differences for meeting PA guidelines for 1-2 days. However, there was a significant difference for 3-4 days (OR = 7.01) and ≥5 days (OR = 7.01). CONCLUSION: Both the hip and wrist-worn accelerometer algorithms provided poor classification accuracy for PA during dance class. Relative to the hip-worn accelerometer, the wrist-worn accelerometer was more likely to classify girls as meeting guidelines for PA. Future research should move away from cut-point methods and use pattern recognition algorithms that leverage the rich data available in the acceleration signal.