Physical activity in the era of mHealth big data: considerations on accuracy and bias

Abstract

Recently smartphones have been considered a powerful tool with which to study large-scale population health on a global scale. Landmark Nature publications highlight the global uptake of mobile technology and the staggering potential for big data to promote large observational studies with big datasets from electronic health records. However, these data have to be generated through unbiased, accurate and validated measurement techniques. This paper discusses in detail some of the inconsistencies in smartphone health apps’ data collection and the issues arising from the increasing availability of low-quality health data. These threats to valid inference from observational studies using big data remain a great challenge to overcome and future research should focus on the development and validation of more accurate algorithms for health-related smartphone apps.