Comparison of a computerized physical activity recall with a triaxial motion sensor in middle-school youth

To develop and test a computerized activity recall (CAR) for capturing activity and energy expenditure (EE) in youth and to further our understanding of the use of the three-dimensional accelerometer (Tritrac) for studying activity and EE.

Forty-five students (25 girls and 20 boys) in grade 6-8, completed 5 consecutive days of the CAR and were randomly assigned to wear the Tritrac during 1 day in which they also completed the CAR. Twenty-two subjects also repeated 5 d of the CAR and 1 d of Tritrac monitoring 1-2 wk later.

The correlation between CAR and Tritrac for total EE was significant (r = 0.510, P = 0.0003). However, the total EE computed from the Tritrac was significantly greater than the CAR (1941 +/- 273 kcal vs 1576 +/- 343 kcal (8.14 +/- 1.14 vs 6.60 +/- 1.44 MJ); P < 0.001). The EE of activities was similar for the Tritrac and CAR, approximately 670 kcal (2.80 MJ), suggesting that the difference between the two methods was their estimates of resting EE. Comparison of the CAR and Tritrac for classifying active and inactive subjects indicated that both methods similarly classified 35 of the 45 subjects or 78% (P = 0.0038). Although significant, the Kappa statistic (kappa = 0.398) indicated a low-level of agreement between methods. The Tritrac indicated 50.4 +/- 29.2 min.d-1 of activities of > or = 3 METs, whereas the CAR indicated 76.7 +/- 71.9 min.d-1 (P = 0.02). The week 1 vs 2 test-retest correlation was 0.0485 (P = 0.022) for the Tritrac and 0.820 (P = 0.0001) for the CAR.

It appears that both methods have acceptable reliability. However, the validity of each method to measure total and activity EE remains unclear, as the Tritrac appears to overestimate resting energy expenditure, whereas the CAR overestimates total minutes of activity.