Applying AI to smartphone movement data to determine the onset and severity of COVID-19 infection

The biggest challenge COVID-19 presents to health authorities is knowing who in the population is infected. A real time view of infected individuals and the severity of their symptoms would allow authorities to better manage medical resources. Once quarantine restrictions are lifted, this insight would enable a targeted contact tracing program. Together, these measures would lead to a dramatic lowering of healthcare costs, along with a reduction in the spread of infection and lowering of fatalities.

This proposal is focused on using data collected from the motion sensors on smartphones to model the health behaviour of individuals. All smartphones have motion-tracking sensors capable of detecting when a user is moving or stationary. Our solution will develop AI to profile periods of inactivity. Those exhibiting a change in movement behaviour over an extended period will be categorised as 'ill', and the degree to which their behaviour has changed will be used as an indication of the severity of their symptoms.

Using passively collected data from sensors on smartphones to infer health status is a highly innovative idea, and is different from other health monitoring solutions currently being trialled, which rely on individuals to self-report their symptoms, and sometimes requires the use of specialist equipment.

The benefits of our proposal are numerous. Unlike other attempts at remotely monitoring patient health, our approach leverages technology widely used in the day-to-day lives of the vast majority of the population, from the young to the elderly, making it a solution applicable to all, regardless of socioeconomic status. The insight generated allows health authorities to conduct real time targeted contact tracing to effectively limit the spread of the virus, which will be imperative once quarantine restrictions are relaxed. It provides a granular, real time view of infections across a community, allowing hospitals and GP practices to better target care and optimally manage resources, saving more lives and reducing care cost. And it gives doctors the confidence to discharge patients knowing that the severity of their symptoms can be remotely monitored.

There is an opportunity to make the IP created in this project available to any company developing their own health software. This would mean releasing the solution in the form of a software development kit (SDK). This would be a highly scalable way to monetise the investment made into this work, and would potentially see the IP used far more widely than if it were restricted to a single health tracking app. A project extension will provide additional time and funding to assemble an SDK offering.