How a COVID-19 Symptom-Tracking App Could Be Used to Ease Lockdowns

Scientists have used data from a COVID-19 symptom-tracking app to predict whether a person has the disease with some degree of accuracy. Now, they are looking to see if it could be used to alert people they may have the disease and forecast future coronavirus outbreaks.

The COVID-19 Symptom Study app tracks potential cases by asking users to anonymously state whether they feel well. If the answer is "no," their phone prompts them to detail symptoms—including hallmarks of COVID-19 such as a fever, persistent cough and shortness of breath, as well as those which have come to attention more recently, like a loss of smell and taste, stomach pains and diarrhea.

Those who are sick can document how their disease has progressed, whether they have been hospitalized, and if they have any underlying health conditions. Users are also asked to document their demographic information, and if they have had what is known as a reverse transcription-polymerase chain reaction (RT-PCR) test, the scientific name for the nasal swab that pick up the virus' genetic material in a person's body.

Created by researchers at King College London and health science firm ZOE, the free app was launched in the U.K. on 24 March 2020 and 29 March in the U.S. Three weeks later, more than 2.6 million people had downloaded it.

covid19, coronavirus, getty
A cleaning company worker unrelated to the app writes on his cell phone while waiting to be tested of COVID-19 during the coronavirus pandemic on May 11, 2020 in Quito, Ecuador. Agencia Press South/Getty Images

In a study published by Nature Medicine, the team detailed how their app could be used to predict whether users had COVID-19. The researchers looked at data provided by the 2,450,569 people in the U.K. and 168,293 in the U.S. who had used the app to report whether they felt ill between March 24 and April 21.

Using this information, the team created a mathematical model to see if a person's symptoms could accurately show whether they had COVID-19. The researchers checked the reliability of their model by looking at what are known as the positive predictive values and the negative predictive values. While the positive predictive value shows the proportion of the people thought to have the coronavirus who indeed tested positive, the negative predictive value is the proportion of the people who tested negative after the app predicted this outcome.

Out of 805,753 users who reported symptoms but had not been tested for COVID-19, the model showed that 140,312 likely had the disease. In the U.S., it correctly found 58 percent of those thought to be positive were, and 69 percent in the U.K. The negative predictive value, meanwhile, was 87 percent in the U.S. data and 75 percent in the U.K. data.

In the U.K. group, all ten symptoms listed on the app were linked with testing positive for the coronavirus. Those were a fever, persistent cough, fatigue, shortness of breath, diarrhea, delirium, skipped meals, abdominal pain, chest pain and a hoarse voice. But only loss of smell and taste, fatigue and skipped meals were linked with testing positive for the virus in the U.S.

The study also revealed that a loss of smell and taste was found to be a predictor of a person having the coronavirus, alongside other more commonly known symptoms. Of the total 6,452 people who tested positive for the coronavirus in the U.K., 4,178 said they lost their sense of smell and taste, versus 2,083 of those who tested negative for the coronavirus.

The team is now close to launching two clinical trials in the U.S. and the U.K. to test whether more accurate algorithms can be created for the app, to alert a person has COVID-19 based on two day's worth of reported symptoms. That includes a trial at Massachusetts General Hospital involving around 5,000 people. At the start of the trial, they will be tested for antibodies to the coronavirus and chart daily whether they have symptoms in the app. If participants show symptoms, they will be tested for the coronavirus. A similar trial will soon start at King's College London.

Highlighting the benefits of the app, Tim Spector, professor of genetic epidemiology at King's College London and co-author of the Nature Medicine study, told Newsweek it's free for individuals and countries to use, and doesn't require health insurance. Unlike other apps, it doesn't trace or track users "so it alleviates any privacy concerns." However, it under-represents those in very deprived areas, some ethnic minorities, and the over 70s, he said, adding that a new feature enabling users to track the symptoms of others has been introduced to tackle this.

The app complements RT-PCR testing because the infection may be missed if swabs are taken too late. "The COVID Symptom Study app would be able to pick up these cases where early testing was not possible," he said.

"There are issues with all the tests both viral and antibody and that's why we need other methods like this symptom algorithm to help clinics and researchers," Spector added.

Spector explained the tech could be used as a base level indicator by officials who are on the look-out for emerging outbreaks. "That's why it's important that high numbers of people in a population are logging daily so the data is as accurate as possible," he said, adding: "This approach will work well in the absence of conventional testing."

The app is also accurate enough to help leaders decide on when to re-open an economy, he said. "When it comes to using the app to decide whether it is safe to lift lockdown [measures], it's about following the data curve downwards and checking it against hospital data and other data to get the full picture of the virus in the population."

Experts in the field not involved in the project praised the team's Nature Medicine paper, but said the app can't replace traditional testing methods.

Simon Gubbins, head of the Transmission Biology Group at the U.K.'s Pirbright Institute, said in a statement that although the level of accuracy was too low to replace testing, the model would potentially be useful for rapidly alerting someone who reports that they have symptoms indicative of COVID-19 to self-isolate and follow local guidelines.

Kevin McConway, emeritus professor of applied statistics at The Open University, said: "Sometimes such a model will track too closely the characteristics of the people who provided the data for it, and therefore not work well in different groups of people—but the researchers checked this by seeing how good the predictions were for app users in the U.S. The model passed this validation test pretty well."

Peter Bannister, a biomedical engineer and Executive Chair of the Institution of Engineering and Technology, said the authors acknowledged that the algorithm could overestimate the number of users who are infected, as RT-PCR testing is only carried out when a person has clear symptoms or is at high risk of exposure to the coronavirus.

"A more comprehensive program of testing would need to be in place to validate such an algorithm for widespread use and to handle the inevitable false alerts raised by the algorithm given its reported accuracy," he said.

Dennis Wang, senior lecturer in genomic medicine and bioinformatics at the U.K.'s University of Sheffield, said: "The performance of their diagnostic model is still quite poor compared to genetic tests that achieve >95 percent accuracy, however, the symptoms highlighted in their model can be used by the government to prioritize individuals for COVID-19 genetic testing, especially when testing supplies are limited."

Linda Bauld, professor of public health at the U.K.'s University of Edinburgh, said: "It is imperative that we collect data on COVID-19 in 'real-time' and in community settings, as most studies to date have focused on hospital patients. We now know that people who come into contact with the SARS-CoV-2 virus won't require treatment in hospital, so in the absence of mass community testing we need to harness digital technology to encourage people to record symptoms and report outcomes. This has limitations—people who download the app won't be representative of the whole population, recall and reporting might not always be correct, and estimates of accuracy stated by the authors may not be sufficient.

She said: "These findings could help health authorities and governments to improve guidance on COVID-19 symptomatology and alert the public to be more aware of this symptom. It also highlights the value of smartphone apps to help us learn more about COVID-19, and how these apps can complement other essential parts of the public health response including testing, tracking and contact tracing."

Since the COVID-19 pandemic started late last year, more than 4.2 million people have tested positive for the coronavirus, 292,619 people have died, and over 1.5 million are known to have survived, according to Johns Hopkins University. The U.S. is the country with the most known cases, as the graph below by Statista shows.

Coronavirus, Statista, Update
Number of cases of COVID-19 reported worldwide as of May 13. Statista