The Effect of Height on the Validity of Three Accelerometer Models

Movement/step frequency has been shown (at least in one accelerometer model) to influence the prediction of accelerometry-based energy expenditure (EE). Uniaxial accelerometers are also limited in their ability to distinguish the intensity of faster running speeds and their data outputs may be confounded by a frequency (i.e., step) dependant filtering limitation within the device. These observations have called into question the validity of accelerometers for physical activity monitoring, particularly among individuals of varying height and consequently stride frequency. PUrPose: To determine whether, or the degree to which, the concurrent validity of three popular accelerometer models (Actical, Actigraph, and RT3) are affected by differences in subject height. METHODS: Eighty six participants age 8 to 40 (17.6 ± 8.0) performed three 10-minute bouts of treadmill activity (speed 1 = walk; 2 = walk/jog; 3 = run) ranging in speed from 4 to 12 km/hr. Height quintiles (1 = 125.5–138.7 cm, 2=139.0–155.3 cm, 3=156-166.2 cm, 4=166.8–179.6 cm, 5=181.0–196.1 cm) were created to investigate mean differences between accelerometer predicted EE and measured EE from respiratory gas analysis across speed categories. Validity coefficients between models and within height categories across speed were determined to assess which models provided the most valid EE estimates for individuals of different heights and whether validity changed at different speeds. All analyses were performed using ANOVA and a Bonferroni correction for multiple comparisons (p<0.05). RESULTS: All speeds considered, each model provided the most valid estimate of EE for height categories 3 and 4 (r2 = 0.54 to 0.86). For the tallest individuals, validity coefficients from all models were weak during treadmill running (r2 = 0.14 to 0.39) and mean differences between estimated and measured EE were significantly greater for the tallest individuals in comparison to the shortest individuals across all speeds. All height categories combined, the Actical underestimated EE while the RT3 overestimated EE across all speeds. The Actigraph underestimated EE during treadmill walking, however overestimated EE for taller individuals (i.e., categories 3 to 5) as speed increased. Across all conditions and heights, the RT3 provided the most valid estimate of EE(r2 = 0.58). CONCLUSIONS: When using accelerometry to estimate EE, validity may be affected by changes in treadmill speed and individual differences in height. These factors should be considered when using accelerometry to predict EE or time spent in physical activity, especially in groups with varied heights. Funded in part by statistics Canada