Assessment of Physical Activity and Sedentary Behavior Using Wearable Sensors: Measurement Considerations and Clinical Applications

The aims of study one of this dissertation were two-fold: 1) To understand the variability of sedentary behavior (SB) throughout a seven-day measurement period and 2) To determine the number of days of SB measurement required to reliably estimate seven-days of measurement. To address these aims, we utilized activPAL (AP) data from two existing data sets from the Physical Activity (PA) and Health Laboratory. Data were from 62 adults and 34 adolescents who had seven consecutive days of valid monitor wear. We found that adults were more sedentary on weekdays compared to weekend days, but there were no differences by type of day in adolescents. Results also suggest that two days of wear in adults and three days of wear in adolescents provide estimates comparable to a seven-day day measurement period. Our observation that less than seven-days of data are needed to reliably estimate SB is a novel finding and will reduce measurement burden for researchers and participants in future studies. For study two, we sought to determine the feasibility of assessing SB/PA with wearable sensors in a clinical population. More specifically, the aims of this part of the dissertation were 1) To quantify changes in objectively measured PA and SB using an AP during the natural progression of osteoarthritis (OA) over a nine-month period and 2) To determine the relationship between change in objectively measured PA and SB and patient reported pain and function. Participants were monitored immediately following a baseline clinic visit and after three, six, and nine-month follow-up visits. In addition, we correlated these PA/SB metrics with patient-reported changes in pain and function. We found no changes in any SB metrics (% of time spent sedentary, breaks per day, and break-rate) at three, six, and nine months in comparison to baseline. There were, however, significant declines in activity metrics (% of time spent stepping, guideline minutes, guideline bouts, % of time spent in moderate-to-vigorous PA and steps per day) at each subsequent time-point. The percent change for declines in activity metrics ranged from 0.1% to 42%. The R2 values describing the relationship between the SB/PA variables and the patient reported pain and function using the WOMAC were all low (range: 0 to 0.14). The results from this part of the dissertation demonstrate that it is a feasible approach to longitudinally quantify natural changes in PA/SB as OA progresses and they also provide insight into how activity behavior changes over time and informs the literature that interventions to help individuals maintain their PA levels are warranted. The aims of study three were: 1) To determine if the ActiGraph (AG) and AP provide comparable estimates of PA and SB in a sample of patients with OA and 2) To compare the relationships between patient-reported pain and function and PA/SB measures from the AP and AG accelerometers. Two SB cutpoints from the AG were utilized (100 counts∙min-1 (AG100) and 150 counts∙min-1 (AG150)). There were significant differences between the AP and AG150 for measures of % of time spent sedentary (difference of 6.3%), breaks per day (difference of 30 breaks), and steps per day (difference of 879 steps). There were significant differences between the AP and AG100 for breaks per day (difference of 30 breaks) and steps per day (difference 640 steps). Average guidelines minutes and number of guidelines bouts were comparable among the AP, AG100 and AG150. The association between the SB and PA variables and the patient reported pain and function were very weak (all R2 values \u3c 0.07) for the AP and AG. This last study provides evidence that researchers should be cautious in comparing PA/SB findings across studies that use different wearable sensors to assess SB/PA behavior