ASSESSMENT OF PHYSICAL ACTIVITY WITH THE TRITRAC ACCELEROMETER: LABORATORY VS. FIELD MEASURES

585 The purpose of this study was to validate the laboratory generated equation that predicted speed, and indirectly METS, from vector magnitude (VM) counts generated by the Tritrac, in a field setting. Thirty non-obese young adults were instructed, to "stroll", "walk briskly" and "jog at a comfortable pace" for two laps each around a flat, outdoor, cork 400m track. Exercise bouts were timed and actual speeds calculated. Projected speeds calculated using field generated VM data and the laboratory developed regression equation that predicted speed from VM counts were found to be significantly different from actual speeds (p < .05). Actual speeds were underestimated by 0.69, 0.53, 0.94, and 0.73 mph for the light, moderate, vigorous, and total activity. Correlation demonstrated a strong linear relationship between pooled actual speeds and VM data (R2e = .90, p<.01). The results of a two-way repeated measures ANOVA indicated a significant speed effect (p=.001), a significant gender effect (p=.019), and a significant interaction between speed and gender (p=.001). Women ran significantly slower than men during the vigorous activity but VM did not underestimate the speed as it did in men. Using the laboratory developed VM cut-points for light, moderate, and vigorous activity, 33% of the light and 20% of the moderate activity bouts were incorrectly classified. These results may indicate that laboratory developed VM cut-points established to classify activity may provide misleading results in a field setting. Possible differences in the biomechanics of treadmill versus overland walking and running may explain the results. A new regression equation that estimated overground walking and running speeds from VM data was developed (speed = .001175*VM + .474). Supported by NIH 49505