AMHERST — As leaders around the world use COVID-19 forecast models to guide responses to the pandemic, a research team at the University of Massachusetts Amherst has created an interactive, publicly available model compiling data from leading international reports.
The “COVID-19 forecast hub,” developed by the UMass Influenza Forecasting Center of Excellence team, compiles information from 11 models created by seven expert teams from around the world, which is posted as an interactive graphic displaying state-by-state and nationwide figures at reichlab.io/covid19-forecast-hub/.
The university’s Influenza Forecasting Center, headed by UMass biostatistician Nicholas Reich, is one of just two U.S. Centers for Disease Control and Prevention flu forecasting facilities in the country. The center was established in October and granted up to $3 million to improve flu forecasting accessibility.
Now, the team is expanding this goal to cover COVID-19 forecasting — an important but complicated task, according to Reich.
“When you’re forecasting seasonal flu, you have 15 to 20 years of data you can look to and have your models learn patterns from these previous years of data,” Reich said. “With COVID-19, we have nothing like that. We have barely two months and are learning as we go.”
But the method of collecting and aggregating the COVID-19 data is “in many ways the same,” Reich said. The team’s “ensemble model” draws from a range of forecasts with sometimes vastly contrasting messages.
“The different models are saying pretty different things,” Reich said. “There are some models that feel like we’re fairly quickly coming down off a peak of deaths in the U.S., and that we’ll be sort of finished or nearly finished in a month, and there are other models saying that we should expect to be seeing significant elevated deaths for several months.”
The answer is typically somewhere in the middle, Reich said. But, he added, “I also feel like this is such an unprecedented situation, and there’s so much that could change, that it’s still really hard to know.”
Although the idea of a forecast may inspire comparisons to daily weather reports, Reich said it’s important to note that forecasts for infectious diseases such as COVID-19 can change greatly depending on what measures are taken to combat the spread of the virus. In contrast, by the time rain is in the forecast, choices such as taking an umbrella to work can protect you from getting wet, but cannot stop the weather.
“Your personal choices aren’t going to have an impact on whether it rains today,” Reich said. “But for infectious disease, especially for some of these decision makers … the decisions they make can change the course of the outbreak, and that can cut both ways.”
A pessimistic forecast may lead to stricter social distancing measures that mitigate the predicted impact of the disease, Reich said. But the opposite can also occur: If a model shows a more positive outlook, leaders may loosen restrictions, allowing the virus to further spread. Some models such factors into account, providing best-case and worst-case scenarios under different policy decisions.
In addition to guiding social distancing decisions, Reich said, predictive models may also be used by hospital management teams to determine surge capacities and how many ventilators might be needed.
“With so much pressure to reopen and to get society and the economy working again,” Reich said, “I think that people are looking to these models, or decision-makers may be looking to these models, to provide some guidance about how bad the outbreak can be in coming weeks.”
Jacquelyn Voghel can be reached at jvoghel@gazettenet.com.
