Defining Adherence

Increasingly, people are collecting detailed personal activity data from commercial trackers. Such data should be able to give important insights about their activity levels. However, people do not wear or carry tracking devices all day, every day and this means that tracker data is typically incomplete. This paper aims to provide a systematic way to take account of this incompleteness, by defining adherence, a measure of data completeness, based on how much people wore their tracker. We show the impact of different adherence definitions on 12 diverse datasets, for 753 users, with over 77,000 days with data, interspersed with over 73,000 days without data. For example, in one data set, one adherence measure gives an average step count of 6,952 where another gives 9,423. Our results show the importance of adherence when analysing and reporting activity tracker data. We provide guidelines for defining adherence, analysing its impact and reporting it along with the results of the tracker data analysis. Our key contribution is the foundation for analysis of physical activity data, to take account of data incompleteness.