Accuracy of a smartphone pedometer application according to different speeds and mobile phone locations in a laboratory context

Abstract

It is well documented that a prolonged decrease of physical activity results in the rise of chronic diseases such as type 2 diabetes,1 obesity,2 hypertension,3 coronary diseases4 and therefore increases the healthcare costs.5 Considering that 44% of the European population does not exercise6 and that the WHO recommendations of 150 min of moderate physical activity per week are only fulfilled by 30% of the Swiss population,7,8 it is a priority for public health policies to encourage individuals to be more active.9 Being physically active by walking is free of charge, presents limited risks of injury and can be practiced in many places by those who can walk. That's the reason why Public health institutions are developing walking-based programs to encourage people to increase their level of physical activity.10,11 These programs are supported by large amounts of studies showing that walking 30 min/day, 5 days/week diminishes risks of cardiovascular accident by 19%11 and also has a positive impact on psychological well-being and diminishes the risks of depression.12,13 Ideally, it is recommended for adults between 26 and 65 years to reach at least 7000 steps per day.14 In an educational perspective, pedometers are often used in health promotion as they are easy to use, low-cost, motivational and self-monitoring tools for sedentary persons.15 In addition, a recent study showed that using a pedometer is likely to be a more precise way to assess the level of physical activity as compared with subjective measure, especially in sedentary sub-population.16 Pedometer-based programs are considered to be efficient to increase the volume of physical activity. In their systematic review, Bravata and collaborators10 showed that when the use of a pedometer is associated with daily step goal, walking performance can be increased by an average of 2187 steps/day. Interestingly, an extra 2000 steps/day in men with very low physical activity (i.e. 2000 steps/day) has been associated with reduced waist circumference17 supporting the assumption that increase of steps in sedentary population is likely to have major impact on health outcomes. The Yamax Digiwalker products range of pedometers (Yamasa Tokei Keiki Co. Ltd., Tokyo, Japan) are usually used as reference devices because they showed the best accuracy scores and reliability for step counting.18,19 However, it is also generally reported that at lower speed the error rate increases.20 The work of Basset and colleagues (1996) already revealed that between walking speeds of 50 m/min and 70 m/min, the accuracy was less than 80% steps counted. Considering that preferred walking speed can be very low in general population,21 the accuracy of these devices at very low speeds (i.e. at 2 km/h) in real world setting may be lower. The aim of this study is twofold: to evaluate the effect of different speeds on the accuracy of one pedometer application but also to test this accuracy when the smartphone is attached at different locations. First, the increased number of pedometer applications and the rapid evolution of technology allow smartphones to be used as step counters 22 Second, few studies have been conducted to validate the accuracy of Smartphone pedometer applications so far, while the location of the device seems to influence the accuracy of the step counts.23,24 Two studies have found applications to be inaccurate.25,26 In two other studies, the results are less straightforward. Åkerberg and colleagues found one application (Pedometer 24/7©) out of ten to be accurate27 and Leong and Wong found one (Pedometer Tayutau©) application out of three to be accurate.20 In this study, we tested the Runtastic Pedometer© application, which was one of the most popular pedometer application.28 Because of its popularity among the Swiss population, the Iphone6© was selected among the multiple mobile phone models (Iphone6 is the best-selling mobile phone in Switzerland with 56% of the Swiss customers that possess this model at the time of the study versus 39% for its direct concurrent Android).29 In addition, Åkerberg and collaborators concluded that the Iphone4© model was accurate with reasonable low standard deviation.27 Probably, because it is equipped with a BMA280 accelerometer discriminating accelerations between 1/512 g and 1/4906 g, whereas walking from 2 to 8 km/h induces 0.1–0.61 g accelerations at hip level.30 Our hypotheses were that 1) the smartphone application would be as accurate as the pedometer; 2) the sensibility of the smartphone accelerometers would be more accurate than the mechanical lever of the pedometer at slow speeds; and 3) the accuracy of the smartphone accelerometers would disrupt the measures in the loosest position (“jacket”).