Is A Smartphone App A Valid And Reliable Method To Measure Bicycling Behavior?

Abstract

PURPOSE: To determine the criterion validity and test-retest reliability of a smartphone app in measuring bicycling behavior. METHODS: A GPS device was used as the gold standard for comparison to the smartphone app. Courses ~2 miles in length at two locations were tested; 1) urban college campus and 2) rural bicycle path. Adult bicyclists (N=50; 21 urban, 29 rural) carried a Global Positioning System (GPS) device and their smartphone while riding a bicycle on a course loop twice (5 minute break in between); both the GPS device and smartphone app recorded location data. Movement time was recorded for the GPS device and the smartphone app, and MET*min were calculated for bicycling assuming 6 METs/min. Using a geographic information system (GIS), a 36-foot buffer was created around the road network for each course. Location data from the GPS and smartphone devices were mapped over the course buffers in GIS, and the percent of location data within the buffers was calculated for each. Validity and reliability of the percent within buffer and validity of MET*min for each device were examined using paired sample t-tests. RESULTS: GPS device data fell within the urban course buffer 69% of the time and within the rural course 37% of the time, while data from the smartphone app was within the urban course buffer 78% of the time and within the rural course 51% of the time. For the GPS device, maximum distance of a single GPS point away from the road buffer averaged 14 meters (range 0-290), while maximum distance of a single GPS point from the smartphone app averaged 15 meters (range 0-92). Mean difference between the GPS and smartphone devices was 9% in favor of the smartphone app on the urban course (n=40, p<0.05), and 13% in favor of the smartphone app on the rural course (n=54, p<0.01). For reliability of the GPS device, mean difference between the two repeated rides was 4% on the urban course (n=20, p>0.05), and 9% on the rural course (n=28, p<0.01). For reliability of the smartphone app, mean difference between the two repeated rides was 3% on the urban course (n=21, p>0.05), and 0.6% on the rural course (n=28, p>0.05). MET*min from the GPS device was 1.22 versus 1.21 from the smartphone app (n=94, p>0.05). CONCLUSIONS: The smartphone app tested is a reasonable alternative to GPS devices in assessing bicycling behavior in urban and rural environments.