GPS and Accelerometer Enhanced Prompted Recall as an Ontological Tool in Travel Behavior and Physical Activity Research Laura Wilson Westat, Inc. May 2016
Outline Physical Activity Research Prompted Recall Methods Physical Activity Case Study Methodology Discussion Prompted Recall in Household Travel Surveys Next Steps 2
Introduction Physical Activity (PA) Research Goals: Quantify physical activity of free living subjects to gauge if people are meeting CDC recommendations for exercise and PA. Methods Self-report - least reliable, individuals do not estimate personal PA well GPS and / or accelerometer passive data collection much higher accuracy of PA durations by intensity level, lacks subjective context GPS, accelerometer with PR survey most accurate PA recall along with attitudinal context about activities 3
Prompted Recall Method Purpose Allows for collection of contextual data to supplement the objective sensor data thereby creating a more robust data set for analysis Method Instrument participants with sensors (GPS and / or accelerometers) to collect passive objective data for several days Post process sensor data into useable / recognizable pieces (GPS trips, bouts of high PA) Engage participant in survey using GPS trips and / or PA bouts to probe for attitudinal and contextual attributes 4
PA Case Study Moving Across Places (MAPS) study Salt Lake City, UT Evaluate a Complete Streets infrastructure implementation for changes in PA of residents living near to the project 918 complete subjects, 856 with bouts 7 day data collection with passive GPS loggers and triaxial accelerometers 5
PA Case Study Moving Across Places (MAPS) study Bout detection Bouts detected on the fly via cloud based processor Time periods of Moderate / Vigorous PA at least 3 or more minutes in duration Prompted Recall survey of MVPA bouts in order of recency to: confirm recall of the activity collect activity attributes Trip start and end geocodes and places of interest (POIs) attitudinal and behavioral context regarding the detected MVPA bout 6
MAPS: PA Bout Reviewer 7
MAPS: PA Bout Reviewer 8
MAPS: PA Bout Reviewer 9
MAPS GPS, PA, and PR results 120.0% Prompted Recall Results 100.0% 80.0% 82.8% 91.8% 76.6% 94.7% 81.9% 97.1% 88.3% 66.8% 64.9% 60.0% 50.0% 40.0% 20.0% 0.0% 100 m 200 m 400 m 800 m 1600 m % of MVPA Bouts Recalled with Maps % of MVPA Bouts Recalled without Maps 10
MAPS GPS, PA, and PR results Getting someplace was rated stronger than either exercise or leisure/recreation motives. Exercise and leisure/recreation motives were strongly correlated and the other correlations were modest but significant 11
Discussion Prompting individuals to recall details of their sensor detected MVPA bouts with spatio-temporal reference allows more boutspecific information than many self-reports. This method illustrates how researchers can clarify contextual and experiential qualities of healthy bouts of activity. 12
Prompted Recall in Household Travel Surveys Travel Behavior Research / Household Travel Surveys (HTS) Goals: collect travel information from a representative sample for use in updating travel demand models at the regional and state levels. Methods Self-report diary - Least reliable, individuals tend to under-report trips/places GPS and diary matching - Better than self-report alone. Identifies missing trips but no opportunity to correct/reconcile discrepancies GPS with Prompted Recall Provides most complete combination of spatial resolution and accuracy (from GPS data) and attribute data (from self-report data) 13
Trip Builder Web GPS + PR Interface A web-based system that supports both travel log reporting and GPSbased prompted recall. Used for telephone interviews, web surveys, or mail-back travel log data entry. Integrated online map (using Google Maps API) for real-time geocoding and point-of-interest lookup. 14
HTS GPS with PR Project Summary Study Attributes Michigan Cleveland Jerusalem New York Length 1 year 1 year 1 year 1 year Sensors Wearable GPS Wearable GPS Wearable GPS Wearable GPS Deployment Period 3 days 2-4 days 1 day 2-4 days Age 16-75 13-75 15 and up 16-75 # Households Completed 2,200 4,545 3,000 1,900 15
Next Steps, New Technologies Smartphones, mobile devices Real time survey prompts based on GPS, possibly accelerometer data Reduce time between actual travel and capture of details Allow participants to BYOD or get pre-loaded phone shipped ahead of travel Long term usability in the estimation of travel demand models yet unknown 16
Acknowledgements University of Utah - MAPS study team Dr. Barbara Brown Calvin Tribby Dr. Carol Werner Dr. Harvey Miller Dr. Ken Smith 17
Questions? Contact Information: Laura Wilson Westat, Inc. laurawilson@westat.com 18