ActiGraph has moved!
Please make a note of our new corporate address:70 North Baylen Street, Suite 400
Pensacola, FL 32502
Measuring Transportation Physical Activity: Self-Report vs. GPS
- Presented on February 26 2013
Background and Purpose Research investigating relationships between physical activity (PA) for transportation and the built and social environments relies upon accurate measures of PA. Developing accurate and reliable measures has been an objective of the ALR program, as well as of behavioral researchers more generally. Generally, objectives measures are preferred over self-report measures. One review of studies comparing self-report and objective measures of PA (Prince et al, 2008) found low-to-moderate correlations between the two (mean 0.37), with slightly higher correlations for male-only studies compared with female-only studies. For studies with comparable measures, a majority found that self-reported PA was higher than the objectively measured PA, including PA measured with accelerometers. At the same time, transportation researchers have noted that travel survey respondents often underreport trips, particularly short, walking trips; personal GPS units have been proposed as a solution to this problem (Stecher & Stopher, 2006). While objective measures may be more accurate, they are not perfect. In addition to their own underreporting issues (e.g. forgetting to wear the equipment), issues include higher costs and sampling bias. In addition, to capture relatively rare activity, e.g. bicycling, objective measures may require longer data collection periods, posing a greater burden on participants.
Objectives The objective of this research is to compare an objective measure of transportation PA collected through GPS with self-reported data on habit and typical walking/bicycling travel collected through a survey. The research questions include: 1). How does walking and bicycling travel as measured with a GPS unit over a 5-day sample period compare to self-reported typical behavior? 2). If there are differences, what factors help explain those differences (e.g. demographics, attitudes, travel mode, etc.)? 3). Do any differences between the measures lead to significantly different conclusions regarding the correlates (e.g. built environment, psycho-social factors, demographics, etc.) of transportation PA?
Methods The research utilizes data from the Family Activity Study, a multi-year longitudinal intervention study in Portland, Oregon. The Family Activity Study is seeking to determine the effect of traffic calming interventions on the active travel behaviors, perceptions, and attitudes of families in the Portland area. Data collection includes surveys, GPS and accelerometer data. The sample includes 480 adult parents of children under 17 from 320 households. Each participant carried a GPS unit with them when they left their homes for five days, including at least on weekend day. The unit recorded a location every five seconds. These data points, which include speed, were used to identify travel mode for each trip, including walking and bicycling. Surveys completed just prior to GPS data collection included questions about bouts of walking and bicycling PA within the past 7 days, as well as frequency of walking & bicycling over the past 30 days and during typical good- and poor-weather months. Therefore, the comparisons are between objective data for five days and self-reported data for the immediate past or typical behavior.
Results On average, participants walked and biked fewer minutes while collecting GPS data than they stated they did in the past seven days (scaled to 5 days):Bicycling: 59 (102 s.d.) vs. 30 (62)Walking: 127 (190 s.d.) vs. 75 (82), respectively The correlation between the self-reported and GPS data was much higher for bicycling (0.623) than for walking (0.137). Over 60% of the participants did not bicycle during the GPS data collection. Removing those participants, the correlation coefficient in reduced, but only to 0.544. Correlation coefficients were higher for women than men for walking (0.185 vs. 0.037), but about the same for bicycling. The presentation will explore these data further, including an assessment of agreement using the Bland Altman method. Finally, we will test several multiple regression models with walking and bicycling PA as the dependent variable, measured both objectively and self-reported. Independent variables will include demographics, the built environment, and psycho-social factors. The models will be compared with respect to model fit and coefficients to detect whether any differences in measurement methods lead to different conclusions regarding correlates of transport PA.
Conclusions The analysis to date indicates that adults may report higher levels of transport PA in the past or in a typical month than recorded using GPS. The final paper will draw conclusions to explain these differences and assess whether they make a difference in understanding correlates of transport PA.
References 1. Prince, SA, et al. A comparison of direct versus self-report measures for assessing physical activity in adults: a systematic review. International Journal of Behavioral Nutrition and Physical Activity 2008,5:56 2. Stopher, P.R., Stecher, C.C. (eds.) Travel Survey Methods—Standards and Future Directions, Elsevier, Oxford, 2006
Support/Funding Source The Family Activity Study is funded by the Active Living Research program of the Robert Wood Johnson Foundation and the Oregon Transportation Research & Education Consortium