Women's Health

How weather apps could predict your risk of COVID

August 17, 2022 – Tapio Schneider is a climatologist and his wife a mechanical engineer. In many ways, they were like many other families affected by COVID: two young, out-of-school children and endless Zoom meetings from home. But the two weren’t just baking sourdough bread and walking around during the lockdown: they were thinking about how they could use their expertise to help.

“We were locked up at home like everyone else, talking about how isolation or lockdowns could be avoided,” recalls Schneider, professor of environmental science and engineering at the California Institute of Technology and principal investigator at the NASA’s Jet Propulsion Laboratory.

At the time, lockdowns were the only known way to control the virus, but Schneider felt they didn’t work well.

“Even at the height of the pandemic, 1 or 2% of the population was actually contagious,” he says. “Ninety-eight percent wouldn’t need to self-isolate.” But the problem was who these contagious people were.

Then it hit him: what if he could create a COVID “forecast” using the same technology used by weather apps?

Schneider’s wife, who is also a professor at Caltech, was studying body temperature sensors. Perhaps, they reasoned, data from similar devices could be combined with data from COVID tests to predict a person’s chances of contracting the virus. Send that data to an app, and each user could get their own personalized risk right on their smartphone.

This seed of idea became a study in Computational Biology PLOS. Schneider partnered with a global team — including a German computer scientist and a disease modeler from Columbia University in New York — to find out if an app like this could help control a pandemic like COVID. And the results are promising.

How a COVID forecast app works

If you’ve ever used a weather app, you’ve probably noticed that weekend forecasts can be very different on Mondays compared to Fridays. And it’s not because meteorologists don’t know what they’re doing: it’s a reflection of the vast glut of data that is constantly being imported, increasing the accuracy of forecasts as the actual date approaches.

Every 12 hours, weather apps run a scan. The first stage captures the atmospheric state just now – elements such as temperature, humidity and wind speed, as measured by sources such as weather stations and satellites. This information is mixed with the forecast from 12 hours earlier, then connected to an atmospheric model. An algorithm predicts conditions in 12 hours, the weather app updates, and half a day later the cycle repeats.

Imagine an app that uses a similar method, except it connects COVID data to a disease tracking model, tracing the path from risk, to exposure, to infection, and finally to recovery, to hospitalization. or upon death. The data would include the obvious — rapid test and antigen test results, self-reported symptoms — as well as the more unexpected, like smartphone data and the amount of virus in local sewage, which is quickly becoming an invaluable tool for predicting illnesses. COVID outbreaks.

“The key is that it’s specific to individuals,” says Schneider. The app will not only predict the percentage of infected people in your city; he would rather assess your unique risk of getting the virus, based on the data your Bluetooth-enabled device picks up.

Existing exposure notification apps, which are more widely used in Europe and Asia than in the US, ping you after you’ve been exposed to the virus, but they don’t update you between alerts. Schneider envisions using the data these apps use more efficiently, drawing on other data sources, providing regularly updated infectivity predictions, and advising you to self-isolate after likely exposure.

How effective would the application be?

In the study, Schneider and his team created a simulation city, designed to mimic New York City during the early stages of the pandemic. This data network included thousands of intersection points, each representing a person – some with many daily interactions, some with few. Each was assigned an age because age impacts the route COVID takes.

What their simulations revealed: If 75% of people used a COVID prediction app and self-isolated as recommended, the pandemic could be effectively controlled – as long as diagnostic testing rates are high.

“It’s just as effective as a lockdown, except at any given time only a small fraction of the population self-isolates,” says Schneider, noting that in this case a “small fraction” is about 10% Population. “Most people could live their lives normally.”

But as low COVID vaccination rates have revealed, near-universal compliance may be a goal that cannot be achieved.

Another potential challenge is overcoming privacy issues, even if the data would be anonymized. According to Schneider, starting with smaller communities, like college campuses or workplaces, could foster broader acceptance as people see the benefit of sharing their data. Younger people, he observes, seem more comfortable with disclosing health information, which means they may be more willing to use such an app, especially if it can avoid another lockdown.

The Future of Infectious Disease Tracking: Empowering Every Person

Mathematical modeling of infectious diseases is not new. In 2009, during the H1N1 (swine flu) pandemic, the CDC used data from several sources to help slow the spread of the flu. During the Zika outbreak of 2016 to 2017, modeling helped researchers identify the link between the virus and microcephaly, or a condition in which a baby’s head is much smaller than normal, from the start. beginning. In fact, mathematical predictions have been useful for everything from the flu to HIV, according to a 2022 journal article inClinical infectious diseases.

Then came COVID-19 – the worst pandemic in US history, requiring a new level of reckoning.

In partnership with the University of Massachusetts at Amherst, the CDC created The Hub, a data repository that merged multiple independent forecasts to predict COVID cases, hospitalizations, and deaths. This massive undertaking not only helped inform public policy, but also revealed the importance of prompt contact tracing: if identification of close contacts took longer than 6.5 days after exposure, it was practically useless.

Schneider echoes this concern with what was once hailed as the COVID control method. In his team’s app-based prediction simulations, “you reduce death rates by a factor of between 2 and 4, simply because you identify more people who are likely to be infectious than you would by testing , tracing and isolating,” he says. Contact tracing is limited in its ability to control the spread of COVID, due to the high rate of transmission without symptoms and the short latency period of the virus. By combining multiple data sources with a disease transmission model, you gain efficiency.

“You know how it propagates through the network,” says Schneider. “And once you get that in, you get more effective control of the outbreak.”

Applying this mathematical approach to individuals – rather than whole populations – is the real innovation in Schneider’s vision. In the past, we could predict, for example, the possibility of finding an infectious person throughout New York City. But the app Schneider hopes to develop would determine the unique risk of infection for each user. It empowers you to make informed decisions – Am I going out tonight? Am I isolating myself? – more squarely in everyone’s hands.

“We have technology here that can lead to outbreak management, even containment altogether, if adopted widely enough and combined with testing,” Schneider says, “and it’s just as effective as our lockdowns, without having to isolate a large part of the population.”

This innovation could help track infectious diseases like the flu or even curb the next COVID, Schneider says.

“You want to control epidemics, you want to minimize disease and suffering,” he says. “At the same time, you want to minimize economic disruption and disruption to life, to schooling. The hope is that with digital means like the ones we have described, you can achieve both of these goals.


Back to top button