Identifying Determinants and Patterns of Physical Activity Behaviour Using Smartphone Data

Physical inactivity is a global health concern. Personal, spatial, and temporal characteristics
all influence how likely an individual is to be physically active. Secondary smartphone and
wearable data are well placed to objectively capture these characteristics alongside physical
activity behaviour, whilst simultaneously addressing some of the limitations of other research
methods and providing new objective methodological opportunities. Use of such secondary
physical activity data however is in its infancy, the feasibility and suitability of these data to
answer key physical activity research questions needs to be investigated.
This alternative format thesis, supported by Fuell Ltd (UKRI data partnership) investigates
the socio-demographic, temporal and spatial determinants of physical activity using secondary
physical activity data from over 30,000 Bounts app users over the course of a year. Analysis of
the Bounts app data highlighted the socio-demographic profile of the users and key variations in
activity type, frequency, and meeting of physical activity guidelines by age gender and socioe-
conomic status. We evaluate the findings in the context of other methodical developments in
the objective measurement of physical activity and the environment. Focusing on reproducibil-
ity, we also apply and develop new machine learning and spatial methods to these app data,
identifying key clustered weekly and seasonal behavioural patterns. Moreover, through develop-
ment of a reporting framework for objective studies and an evaluation of standardised activity
space measures on environmental exposures this thesis strives to ensure the development of
transparent and reproducible research practices in this developing field.
This research demonstrates the scope of these data balanced alongside the considerations needed
to successfully utilise them and the challenges they present. The discussion makes the case that
these data, if used correctly, have many methodological advantages which go a long way to
address current research limitations and provide valuable insights for physical activity policy
development.