Harnessing Opportunities of Smartphones and Smartwatches for Epidemiological Research

Student thesis: Phd


Smartphones and smartwatches are ubiquitous and often used daily. Their touchscreens and sensors provide unprecedented opportunities for collecting individual-level data on health, behaviour and the surrounding environment. This granular data can be used to examine the relationship between exposures and health outcomes − the scientific discipline of epidemiology. This thesis explores methods to collect epidemiological data through smartphones and smartwatches and how to analyse it in a scientifically rigorous way. Epidemiological research is only as good as the data that it is based on. The limitations of traditional epidemiological methods hamper our ability to answer some research questions, such as the age age-old question of a possible association between weather conditions and chronic pain severity. Most studies investigating this association enrolled fewer than 100 participants, collected data for less than a month and assumed that participants were always exposed to the weather at their home location (Chapter 2). The national smartphone study Cloudy with a Chance of Pain sought to examine the same question in this digital era. It recruited over 10,000 people with chronic pain from across the UK who tracked their daily symptoms for up to 15 months in a smartphone app. The app collected hourly location data to link to weather reports of the closest weather station. The study showed that participants were more likely to report increased pain severity on days with high relative humidity and wind speed, and low atmospheric pressure (Chapter 4). Missing data was a challenge. For many participants, pain data was missing because they did not report their pain every day. It hampered application of traditional cohort analysis methods and was a reason to use the case-crossover analysis. In addition, location data was often missing, especially for participants using iPhones, during the night and for participants that had not used the app for a while (Chapter 3). To determine the exposure to the weather when location data was missing, the most likely location based on participants’ usual behaviour was imputed. Smartwatches provide similar opportunities for collecting self-reported and passively-measured data. As smartwatches are body-worn, sensor data is easier to interpret and can provide more accurate measurements of some types of behaviour, like physical activity. They can therefore potentially collect granular data to identify how physical activity influences knee pain and vice versa in people with osteoarthritis. To assess the feasibility of such a smartwatch study, I was involved in development of a smartwatch app that launched pain questionnaires multiple times a day, while continuously collecting raw physical activity data (Chapter 5). In a 90-day feasibility study of 26 people, good engagement was observed: participants wore the watch on 73% of days and answered the majority of the questionnaires. Battery life hampered some participants answering the evening pain questionnaire. Combining quantitative and qualitative methods, I identified facilitators and barriers to engagement (Chapter 6). Ease of use and low burden of data entry were among these facilitators. For some participants, expectations of the watch’s performance and limited battery life hampered engagement. This thesis demonstrates that smartphones and smartwatches can be used to collect granular epidemiological data from many participants over long durations. Attrition and missing data are inherent to data collection using consumer devices, and requires careful study design and appropriate analysis methods. With the appropriate methods, this rich data from participants’ daily lives can provide new insights in epidemiological questions. Smartphones and smartwatches are valuable tools for determining the associations between behaviour, environment, health and disease.
Date of Award1 Aug 2020
Original languageEnglish
Awarding Institution
  • The University of Manchester
SupervisorJohn Mcbeth (Supervisor), William Dixon (Supervisor), David Schultz (Supervisor) & Jamie Sergeant (Supervisor)


  • rheumatology
  • healthcare informatics
  • meteorology
  • weather
  • smartwatches
  • smartphones
  • pain
  • digital epidemiology
  • missing data
  • case-crossover
  • biostatistics
  • data science
  • mobile health

Cite this