gms | German Medical Science

Background: The walking distance of a patient is an important medical feature. For example, it allows physicians to evaluate the severity of peripheral artery disease (PAD) and can help to monitor the healing process after surgical revascularization [1]. Widely used measures for walking capacity are pain-free walking distance (PFWD) and maximal walking distance (MWD) [2].

Measuring the walking distance of a patient in a real-world setting is non-trivial. The common approach is to use a treadmill [3] or a periodically marked hallway [4]. Using mobile technology like smartphones could help physicians to assess healing processes afar from laboratory conditions [5]. It could further enable patients to track rehabilitation or secondary prevention successes at home by means of technologically supervised exercise therapy [6].

Methods: An empirical evaluation was conducted to evaluate the accuracy of eight publicly available applications for walking distance estimation on smartphones. The applications were tested on four different systems based on Android (LG G2 and Samsung Galaxy J5) and iOS (Apple iPhone 6 and Apple iPhone 8). The eight selected applications either use geolocations for walking distance estimation (Runtastic, Strava, Nike Run Club and Walkmeter) or step counting with step length estimation (Google Fit, Apple Health, Footsteps and Step Counter).

A 300 meter and a 20 meter distance were measured and selected as test tracks. 672 test were conducted. A test is defined as measuring the accuracy of one application running on one smartphone. All four smartphones were carried simultaneously each running two applications at the same time. All tests have been conducted by one healthy person walking with a uniform, normal speed. In accordance with ISO 5725-1, accuracy is considered a combination of trueness and precision [7]. The accuracy of applications is therefore modeled as mean (trueness) ± standard deviation (precision) as well as the mean absolute percentage error (overall accuracy).

Results: On the 300 meter track, Strava and Nike Run Club were the most accurate candidates overall (i.e., on both platforms) with 300.6±26.2m;6.0% and 286.9±20.0m;6.0%, respectively. The most accurate candidate on Android was Strava with 292.5±18.9m;3.9% and Nike Run Club with 286.3±24.6m;7.2% on iOS. Google Fit and Apple Health achieved 275.7±68.8m;9.4% and 269.4±33.9m;12.5%, respectively. On the 20 meter track, Apple Health was most accurate with 19.0±3.3m;14.8%. In general, applications based on step counting with step length estimation were more accurate on the short distance than solutions using geolocations, and vice versa.

Conclusion: Results showed that accuracies differ greatly between applications and platforms. When determining the most accurate solution, it should be noted that while Strava is more true, Nike Run Club is more precise. Therefore, if repeatability and comparability between the walking distances are required, Nike Run Club is the application to choose. However, if the walking distance should be estimated in roofed areas or for short distances, the pedometer-based solutions Google Fit and Apple Health are the most fitting candidates. A hybrid GPS- and pedometer-based solution could prove beneficial. It must be considered that all tests have been conducted by a healthy person. Future work could include patients instead.

The authors declare that they have no competing interests.

The authors declare that an ethics committee vote is not required.