Validity of a 2-Regression Model for Estimating Physical Activity in Youth Using an Ankle Accelerometer

Abstract

PURPOSE: The purpose of this study was to develop three 2-regression models (2RM) for use with ankle ActiGraph data in youth using the 1) vector magnitude (VM), 2) vertical axis (VA), and 3) all three axes (3ax) as individual predictors. The models were validated using unstructured physical activity (PA). METHODS: For development of the 2RM, 84 girls and 97 boys (mean ± SD; age, 12.0 ± 1.5 yrs) completed 30-min of lying rest and two to seven structured activities (from a list of 25) for 8-min each. For cross-validation, 27 boys and 15 girls (age, 12.6 ± 0.8 yrs) completed approximately 2-hrs of unstructured PA. PA data were collected using an ActiGraph GT3X or GT3X+, positioned on the non-dominant ankle and were converted to 5-sec epochs for each axis and the VM. Measured energy expenditure (Cosmed K4b2) was converted to youth METs (METy; activity VO2 divided by resting VO2). A coefficient of variation (CV) was calculated for each activity to determine if the activity was continuous walking/running or an intermittent activity and separate regression equations were developed for each activity group using the VA only, VM only, and 3ax as individual predictors. Repeated measures ANOVAs were used to compare measured and predicted METy and time spent in sedentary (SED; <1.5 METy), light (LPA; 1.5-2.99 METy), moderate (MPA; 3.0-5.99 METy), and vigorous (VPA; ≥ 6.0 METy) PA during the unstructured PA. RESULTS: For the VA and VM, counts/5-sec ≤ 1 and ≤10, respectively were used to distinguish SED time and a CV of ≤ 15 counts/5-sec was used to distinguish walking/running. For the 3ax 2RM, walking/running was best predicted by using the anterior-posterior and medial-lateral axes and the VA only was the best predictor for intermittent activities. The VA, VM, and 3ax 2RMs were within 0.42 METy of measured METy during the unstructured PA (P>0.05). The VM and 3ax 2RMs were within 5.7 min of measured time spent in SED, LPA, MPA, and VPA (P>0.05). The VA 2RM over- and under-estimated measured SED and LPA time by 9.3 and 8.6 min, respectively (P<0.019), but was not different from measured MPA and VPA time (P>0.05). CONCLUSION: Compared to the Cosmed, the VM and 3ax 2RMs have similar errors on a group basis that are lower than published errors for a 2RM developed for the hip location in youth, using the ActiGraph accelerometer. Study supported by NIH grant R21HL093407