Detecting COVID-19 using AI

I am very excited to share the latest news on our not-for-profit COVID work, published in the leading medical journal Nature Medicine today. We set up Zoe to use machine learning to help solve some of the biggest health issues we face. These initial results suggest that artificial intelligence diagnosis could help significantly without controversial privacy trade-offs.

ZOE in collaboration with Massachusetts General Hospital and King’s College London have developed an artificial intelligence diagnostic that can predict whether someone is likely to have COVID-19 based on their symptoms. The initial phase of the research was published today. 

The AI model uses data from the COVID Symptom Study app, which in just six weeks has grown to over 3 million people reporting their health to help fight the epidemic. The model is able to predict COVID-19 infection without patients having to be tested, by comparing them with people who have used the app to share their symptoms and the results of traditional COVID tests. Two clinical trials are due to start shortly in both the US and the UK. The diagnostic is a not for profit initiative which cannot completely replace clinical tests, but may have great promise for populations where access to testing is limited. The app also does not present the same privacy concerns that track and trace apps do.

More than 3.3 million people globally have used the COVID Symptom Study app to report daily on their health status, whether they feel well or have any new symptoms such as persistent cough, fever, fatigue and loss of taste or smell (anosmia). In the UK, the data is being used to help inform scientific decision making by the National Health Services of England, Wales and Scotland. Following clinical trials, the app may be able to share a diagnosis with study participants, playing a crucial role in helping us around the world to safely lift lockdown and detect a second wave sooner.

In the study published today in Nature Medicine, we analysed data gathered from 2.5 million people in the US and UK who had been regularly logging their health status in the app, around a third of whom had logged symptoms associated with COVID-19. Of participants in the US and the UK, 18,374 reported having had a test for coronavirus, with 7,178 people testing positive.

We investigated which symptoms known to be associated with COVID-19 were most likely to be associated with a positive test. The study demonstrates the wide range of symptoms caused by COVID compared to cold and flu, and the danger of focusing only on fever and cough as the world looks to reopen without reigniting the epidemic. Loss of taste and smell (anosmia) was particularly striking, with two thirds of users testing positive for coronavirus infection reporting this symptom compared with just over a fifth of the participants who tested negative. The findings suggest that anosmia is a stronger predictor of COVID-19 than fever, supporting anecdotal reports of loss of smell and taste as a common symptom of the disease.

We then created a mathematical model that predicted with nearly 80% accuracy whether an individual is likely to have COVID-19 based on their age, sex and a combination of four key symptoms: loss of smell or taste, severe or persistent cough, fatigue and skipping meals. Applying this model to the entire group of over 800,000 app users experiencing symptoms predicted that just under a fifth of those who were unwell (17.42%) were likely to have COVID-19 at that time. 

We are now focused on launching two clinical trials to test more sophisticated AI models to predict COVID based on just two days of symptoms. One trial planned at Massachusetts General Hospital is focused on high risk populations in Boston. Around 5,000 people will be tested for antibodies at the start of the trial, they will use the app daily, and then be retested when showing symptoms. The app will be launched in Spanish as part of this effort. A second trial with King’s College London is due to start shortly in the UK. 

Combining this AI prediction with widespread adoption of the COVID Symptom Study app could help to identify those who are likely to be infectious as soon as the earliest symptoms start to appear, focusing tracking and testing efforts where they are most needed and so helping us to navigate reopening the economy without triggering a resurgence of the virus.?

?Additional notes:

• Find the full paper here: C. Menni et al. (2020) Real time tracking of self-reported symptoms to predict potential COVID-19 infection
• See https://covid.joinzoe.com/us/data for US data and covid.joinzoe.com/data for UK data.
• The COVID Symptom Study app is available to download from the Apple App Store and Google Play Store.
• Daily research updates and data can be found at covid.joinzoe.com/

About ZOE

ZOE is a healthcare science company using data-driven research to tackle the world’s health issues. By using machine learning combined with digital technologies like mobile phones, ZOE enables large-scale scientific studies to tackle issues like COVID-19, inflammation and the impact of nutrition on health. Located in London and Boston, ZOE was founded by Professor Tim Spector of King’s College London, machine learning leader Jonathan Wolf and entrepreneur George Hadjigeorgiou. ZOE has carried out the largest nutritional studies of their kind in the world, and was named one of the Deloitte Fast 50 Rising Stars in 2019 for the company’s contribution to science enabled by technology and machine learning. 

For more information on ZOE’s mission and science visit www.joinzoe.com.?

About King’s College London?

King's College London is one of the top 10 UK universities in the world (QS World University Rankings, 2018/19) and among the oldest in England. King’s has more than 31,000 students (including more than 12,800 postgraduates) from some 150 countries worldwide, and some 8,500 staff.

King's has an outstanding reputation for world-class teaching and cutting-edge research. In the 2014 Research Excellence Framework (REF), eighty-four per cent of research at King’s was deemed ‘world-leading’ or ‘internationally excellent’ (3* and 4*).

Since our foundation, King’s students and staff have dedicated themselves in the service of society. King’s will continue to focus on world-leading education, research and service, and will have an increasingly proactive role to play in a more interconnected, complex world. Visit our website to find out more about Vision 2029, King’s strategic vision for the next 12 years to 2029, which will be the 200th anniversary of the founding of the university. 

World-changing ideas. Life-changing impact. kcl.ac.uk.

?About Massachusetts General Hospital

Massachusetts General Hospital, founded in 1811, is the original and largest teaching hospital of Harvard Medical School. The MGH Research Institute conducts the largest hospital-based research program in the nation, with an annual research budget of more than $925 million and comprises more than 8,500 researchers working across more than 30 institutes, centers and departments. 

In August 2018 the MGH was once again named to the Honor Roll in the U.S. News & World Report list of "America’s Best Hospitals."