RESUMEN
PURPOSE: The purpose of this study was to determine the criterion validity of Activities Completed over Time (ACT24), an automated previous-day recall tool designed for mobile devices for 1) estimating sedentary versus active time compared with an activPAL, and 2) estimating time spent in activity domains (e.g., work, household, leisure) compared with direct observation (DO). METHODS: Over a 7-d study period, 53 participants were sent invitations to complete three automated ACT24 recalls and to wear an activPAL device. A subset ( N = 24) consented to two, 3-h video-recorded DO sessions. activPAL and ACT24 data were matched by date, and agreement for sedentary versus active time was compared between methods using paired t -tests for mean differences and spearman correlations. We compared DO and ACT24 results by domain for overall time use and separately for sedentary and active time using κ statistics and tested mean differences with linear mixed models. RESULTS: Compared with the activPAL, the mean difference in ACT24 sedentary time was 1.9% (mean (95% confidence interval), -0.17 (-0.75 to 0.40) h·d -1 ), and the mean difference for ACT24 active time was 2.2% (0.14 (-0.32 to 0.60) h·d -1 ). Correlations were R = 0.61 (95% confidence interval, 0.39 to 0.76) and R = 0.65 (0.44 to0.78) for sedentary and active time, respectively. Domain-specific agreement was substantial for leisure-time, work, and shopping/errands ( κ range, 0.63-0.79), moderate for transportation ( κ = 0.49), and fair for personal care and household activities ( κ = 0.24 and 0.33). ACT24 estimates of average time within each domain were not significantly different from DO. CONCLUSIONS: The present study confirms that ACT24 is accurate for group-level estimation of active and sedentary time. Domain-specific agreement tended to be higher for more commonly reported activities and those that were of longer duration.