Revising Free Text Inputs In Physical Activity Self-report Methods

Abstract

PURPOSE: To examine the change in total Metabolic Equivalent of Task (MET) minutes reported by the Activities Completed over Time in 24 Hours (ACT24) physical activity self-report recall for different activity types after recoding free text entries and assigning updated MET values. METHODS: Men and women aged 50-70 years participating in a measurement error study on diet and physical activity were administered 6 ACT24 recalls spaced evenly over 12 months. Participants that completed at least 1 recall (n=1,023) were included in the analyses. Free text activity entries were recoded and either assigned to an existing ACT24 activity category or placed into a new activity category. Recoded free text activities, pre-defined activities, and gaps (no information recorded) were then assigned MET values from the 2011 Compendium of Physical Activities. The subsequent change in MET minutes was calculated after each of these adjustments. RESULTS: Participants completed 5,311 ACT24 recalls, of which 2,712 (51.1%) contained at least 1 free text entry; free text entries constituted less than 5% of all activities. Recoding free text entries increased the total reported MET minutes by 0.004% (+434.8 MET minutes), and applying the updated compendium values to all entries further increased the total by 6.3% (+743,920.9 MET minutes). Recoding free text had the largest positive effect for the “caring for or playing with others” activity category (+36.3%, or +6,455.1 MET minutes), and the largest negative effect for the “shopping, errands, and appointments” activity category (-15.4%, or -4,925.0 MET minutes). Updating MET values had the largest positive effect on “bicycle repairs” (+81.9%, or +1,226.5 MET minutes), and the largest negative effect on “covering plants” (-49.0%, or -202.0 MET minutes), although these activities were rarely reported (12 and 3 times, respectively). CONCLUSIONS: Recoding free text activities negligibly changes the total MET minutes reported in the ACT24, although more dramatic adjustments do occur within certain specific activity categories. Presence of free text is a data management burden that may provide little additional information, and efforts should be made for best activity classification in instrument design to eliminate the need for free text.