Coronavirus: research reveals a way to predict infection

The UK and some US states have set out plans for easing coronavirus lockdowns. Ultimately, any government that attempts to ease restrictions must keep a close eye on daily infection numbers and the spread of the virus. Ideally, every new case should be traced and managed.

The problem is that most countries lack the resources to test and contact-trace enough people. But our app, which is called the COVID Symptom Study and is based on some 3.4 million users in the UK, US and Sweden logging symptoms daily, could help. In a new study, published in Nature Medicine, we show that this app can estimate whether someone has COVID-19 purely based on their symptoms – with a high degree of accuracy.

The app (formerly known as the COVID Symptom Tracker) was launched by our team at King’s College London in collaboration with the health technology company ZOE (which one of us helped co-found) in March. Users are asked to say whether they are feeling well or experiencing any symptoms related to COVID-19 every day. Within 14 days, with the help of social media, we gathered 2 million users, collecting vital information on the symptoms of coronavirus infection and the spread of the disease across the UK.

For our new study, which has been peer-reviewed, we analyzed data gathered from just under 2.5m people in the UK who had been regularly logging their health status in the app. Around a third had logged multiple symptoms associated with COVID-19. More than 15,000 people reported having had a test for coronavirus, with nearly 6,500 testings positive. We confirmed the findings with data from around 168,000 US-based users of the app – 2,736 of whom had been tested for COVID-19, with 726 testings positive. US users started participating in about one week after UK ones.

Telling symptoms

We then investigated which symptoms known to be associated with COVID-19 were most likely to predict a positive test. Loss of taste and smell were particularly striking, with two-thirds of users testing positive for coronavirus infection reporting them compared with just over a fifth of the participants who tested negative.

Next, we created a mathematical model that can predict 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.

The implications of this are huge: in the absence of widespread, reliable testing for coronavirus, symptom logging through the app is a simple, fast, and cost-effective way to help people know whether or not they are likely to be infected and should take steps to self-isolate and get tested.

We’re now further validating our prediction model by working together in the UK with the Department of Health and Social Care’s coronavirus testing program, offering swab testing to thousands of app users reporting new symptoms every week. In the US, we are planning studies to deploy antibody tests to see if people who reported symptoms in the past were indeed infected with the virus and if antibodies are enough to protect against another infection.

Importantly, our results suggest that loss of taste or smell is a key early warning sign of COVID-19 infection. A loss of appetite and severe fatigue also outperformed the classical symptoms like cough and fever. Focusing on just cough and fever will miss many cases. Although the World Health Organization and the US Centers for Disease Control and Prevention have recently expanded the list of symptoms, many governments like the UK have been slow to change. NHS England still lists cough and fever as main symptoms on its website.

We strongly urge governments and health authorities everywhere to broaden the range of symptoms and advise anyone experiencing a sudden loss of smell or taste to assume that they are infected and follow local self-isolation guidelines.

The detailed symptom data being collected is showing us the enormous diversity of clinical presentations of the virus, such that we are beginning to define distinct clusters over time that have different outcomes and duration. For example, multiple symptoms occurring rapidly have a better prognosis than those coming on more slowly involving fatigue and chest symptoms.

We are also finding many people with symptoms waxing and waning for over a month. Working alongside testing and contact tracing, which most governments are doing to some extent, the COVID Symptom Study app is a potential tool for getting countries out of lockdown more safely. This is especially important as testing resources will remain scarce. Gathering detailed health data from as many people as possible is an essential part of this, while also ensuring that consent and privacy are fully respected.

This data-driven approach relies on millions of people using the app to log their health on a daily basis. Even as we return to our normal lives, we need to stay vigilant - and people need to understand the full range of symptoms. We are asking people to download the app and get in the habit of spending just a minute every day checking in. The app has been endorsed and promoted by charities as well as the governments of Wales and Scotland – but not yet by NHS England.

The rapid roll-out of the COVID Symptom Study app and others like that used in Israel proves the worth of apps like this for real-time epidemiology in the immediate response to a pandemic. There’s an even larger role for the app in research.

Working together with a large team at Massachusetts General Hospital in the US and the charity Stand Up to Cancer, we are producing early data on risk factors across countries like obesity, blood pressure medication, and social deprivation. We are also looking at the risk to healthcare workers. Some of this work hasn’t yet been subject to peer review, the process by which experts scrutinize each other’s work.

The COVID Symptom Study app is available to download from the Apple App Store and Google Play Store in the UK and USA as well as Sweden. Daily research updates and data which are shared with the NHS can be found here.

This article was originally published on The Conversation by Tim Spector at King's College London and Andrew Chan at Harvard Medical School. Read the original article here.