Digital data streams obtained from smart phones and wearable devices have the potential to advance our ability to improve population health. The ‘How Engineering, Mathematics, Health Tech and AI has impacted on the COVID-19 pandemic’ event at Reuben College on January 27- the second in Reuben’s Hilary Term Pandemic themed seminar series - brought together Lionel Tarassenko and Christophe Fraser to discuss how digital technologies combined with modern data processing tools have quickly advanced our understanding of COVID-19 and how they may be used in the future.
Lionel described how sensor data, primarily vital signs like blood oxygen levels, respiratory and heart rates, from wearable devices were used in the Oxford hospital COVID-19 isolation wards to reduce the time nurses spent taking measurements, and to learn about the health of each patient in their disease progression. These data were then transferred through the local and secure WIFI network and displayed on tablets in a ‘patient control room’. These technologies were deployed rapidly at the start of the pandemic in the UK on March 23 2020. Since then, they have not only improved the temporal resolution of data, but also protected many health care workers, as less contact time with patients reduced their risk of becoming infected with this novel infectious disease.
Machine learning was used to evaluate the test results of thousands of Lateral Flow Tests (LFTs), which became the primary tool of testing during the more recent phase of the COVID-19 pandemic in the UK. LFTs can be self-administered and provide results of infection status in ~15 minutes. However, there are challenges associated with evaluating the results of these tests. A team in Oxford partnered with two colleges to evaluate the testing results sent from their smartphones using a combination of image classification algorithms and deep learning algorithms. These algorithms outperformed human readers by up to ~60% in some settings.
After hearing about wearable technologies and machine learning and how they have improved our ability to interpret data from the COVID-19 pandemic, we moved into a concrete application of how mobile technology aided classical contact tracing during an emergency.
Contact tracing has two principal uses during a pandemic:
a) Breaking chains of transmission by asking contacts of infectious individuals to isolate or test;
b) Identifying the pathways of transmission to inform the public about routes of transmission.
Traditionally, contact tracing was performed manually but was quickly abandoned in the UK during the early stages of the pandemic due to limited capacity. Prof. Fraser from the Big Data Institute described the journey from early in the pandemic, when researchers learned that the bulk of transmission was occurring before symptom onset, making SARS-CoV-2 difficult to control. Speed was of the essence, so why not use digital tools, such as our smartphones that we carry around all day?
The race to develop a secure and accurate, yet actionable and integrated platform proved to be a technical and political challenge. However, as nicely shown in this Nature paper, the rollout of the contact tracing app had measurable impact in reducing cases of COVID-19 in the UK, specifically at local scales.
We can think of digital contact tracing technology as a tool to enable more precise interventions, but their impact during future epidemics will depend on the epidemiological parameters of the disease (e.g., when does transmission occur, before or after symptom onset) and how quickly such technologies can be deployed.
Moritz Kraemer is a Research Fellow at Reuben College, within the Artificial Intelligence and Machine Learning theme. His research interests lie at the intersection of data-science, network science, climate science, epidemiology, public health and genomics.
