Brady, R (2020) THE DEVELOPMENT AND CROSS-VALIDATION OF RAW ACCELEROMETER SEDENTARY AND PHYSICAL ACTIVITY THRESHOLDS IN UNIVERSITY STUDENTS. Masters thesis, Liverpool John Moores University.

Preview

Text 2020 Brady, Ruth MPhil.pdf - Published Version Download (1MB) | Preview

Abstract

Background: Regular participation in physical activity (PA) has been associated with the primary prevention of 25 chronic medical conditions. University is considered as a major period of transition as it is associated with full independence from parents for the first time. The prevalence of inactivity in university students is reported to be as high as 60%, though few studies have focussed on university students or used device-based methods to assess physical activity. Previously, researchers in the discipline have used device specific, proprietary, dimensionless units called counts to process and report accelerometer data. Researchers can now process raw acceleration signals rather than rely on proprietary counts. This is advantageous as these transparent methods improve comparability across different accelerometers and studies. Currently there is a lack of raw accelerometer thresholds for use in university students. Aim: The primary aim of this study was to calculate and cross-validate accelerometer thresholds to classify PA and SB in university students. Methods: Thirty-five university students enrolled during the 2018/2019 academic year at Liverpool John Moores University with a mean age of 21.4 years (n=21 females) completed a circuit of 12 activities in laboratory conditions. Each participant wore 3 ActiGraph GT9X accelerometers (both wrists and hip). Thresholds were generated using Receiver Operating Characteristic (ROC) curve analysis in a calibration group (n= 21) using indirect calorimetry as the criterion reference. Sensitivity, specificity, percentage agreement, mean absolute percent error (MAPE), and Cohens’ Kappa coefficients were investigated. These thresholds were then fine-turned using equivalency analysis. These resultant thresholds were then cross-validated in an independent sample (12 participants, n=7 females). Once again, sensitivity, specificity, MAPE, percentage agreement and Cohens’ Kappa coefficients were investigated. Results: The final SB thresholds range from <8 (hip) to <40 mg (dominant) (sensitivity: 0.76 (non-dominant) to 0.84 (hip), specificity: 0.81 (hip) to 0.92 (non-dominant), MAPE: 9.5 (dominant) to 20.7 (hip), percentage agreement: 82.2 (hip) to 87.2 (dominant), Cohens’ Kappa coefficients: 0.61 (hip) to 0.70 (dominant)). LPA ≥8 to ≥40 mg (sensitivity: 0.25 (hip) to 0.39 (dominant), specificity: 0.82 (dominant) to 0.88 (hip), MAPE: 9.8 (dominant) to 22.7 (hip), percentage agreement: 71.6 (non-dominant) to 72.8 (hip), Cohens’ Kappa coefficients: 0.15 (hip) to 0.21 (non-dominant and dominant)). MPA ≥50 to ≥110 mg (sensitivity: 0.45 (dominant) to 0.68 (hip), specificity: 0.76 (non-dominant) to 0.83 (hip), MAPE: 6.8 (hip) to 12.7 (non-dominant), percentage agreement: 69.6 (non-dominant) to 78.8 (hip), Cohens’ Kappa coefficients: 0.24 (dominant) to 0.49 (hip)). VPA ≥225 to ≥315 mg (sensitivity: 0.64 (hip) to 0.72 (dominant), specificity: 0.92 (dominant) to 0.95 (hip), MAPE: 11.3 (hip) to 12.1 (non-dominant), percentage agreement: 88.5 (dominant) to 90.1 (hip), Cohens’ Kappa coefficients: 0.60 non-dominant and dominant) to 0.62 (hip)). MVPA ≥50 to ≥110 mg (sensitivity: 0.74 (dominant) to 0.84 (hip), specificity: 0.74 (non-dominant) to 0.85 (hip), MAPE: 8.8 (hip) to 13.6 (non-dominant), percentage agreement: 75.0 (dominant) to 84.6 (hip), Cohens’ Kappa coefficients: 0.50 (non-dominant) to 0.69 (hip)). In comparison to the criterion reference, for the non-dominant wrist, the SB threshold provided equivalent estimates at ±20%. For the hip placement, MVPA estimates were equivalent at ±15% and MPA was equivalent at ±20%. For the dominant wrist placement, SB and MPA were equivalent at ±15%. LPA and MVPA estimates were equivalent at ±20%. VPA estimates were not statistically equivalent in comparison to the criterion for any placement. Conclusion: The thresholds for SB, LPA, MPA and MVPA showed acceptable levels of agreement between the accelerometer and criterion reference in regards to specificity, MAPE and percentage agreement. VPA was the only threshold to show no equivalency between accelerometer and criterion reference (Metamax) on any placement site. The hip placement generally provided better agreement between the criterion reference and accelerometer and therefore could be considered the optimum placement to provide accurate estimates of SB and PA levels of university students. The thresholds generated in this study can be used to help researchers estimate the habitual activity levels of university students. Future research should aim to validate these thresholds in a free-living or simulated free-living situations to examine their performance in an ecologically valid setting.

Item Type:	Thesis (Masters)
Uncontrolled Keywords:	Area Under the Curve; Euclidean Norm Minus One (ENMO); Mean Absolute Percent Error; Physical Activity (PA); Receiver Operating Characteristic Curve
Subjects:	R Medicine > RC Internal medicine > RC1200 Sports Medicine
Divisions:	Sport & Exercise Sciences
Date Deposited:	04 Feb 2020 12:47
Last Modified:	21 Nov 2022 15:01
DOI or ID number:	10.24377/LJMU.t.00012181
Supervisors:	Boddy, L, Graves, L and Foweather, L
URI:	https://researchonline.ljmu.ac.uk/id/eprint/12181

View Item