Wednesday, September 28, 2022

UMass Amherst outfits most accurate model for predicting COVID-19 deaths: The Forecast Hub is the largest infectious disease prediction project ever

The University of Massachusetts Amherst-based US COVID-19 Forecast Hub, a collaborative research consortium, has the most consistently accurate predictions of pandemic deaths at the state and national levels, according to a paper published on April 8. Proceedings of National Academies of Science, Every week since the beginning of April 2020, this international effort has produced a multi-modal ensemble forecast of short-term COVID-19 trends in the US

The COVID-19 pandemic has highlighted the critical role that collaboration and coordination among public health agencies, academic teams and industry partners play in developing modern modeling capabilities to support local, state and federal responses to infectious disease outbreaks. can perform.

“Predicting outbreak change is critical for optimal resource allocation and response,” says lead author Estee Kramer, a UMass Amherst Ph.D. Candidate of Epidemiology in the School of Public Health and Health Sciences. “These forecasting models provide specific, quantitative and evaluable predictions that inform short-term decisions, such as health care staffing needs, school closures and the allocation of medical supplies.”

An unprecedented global cooperative effort, the Forecast Hub represents the largest infectious disease prediction project to date. The collective research involved only 300 authors affiliated with 85 groups, including US government agencies such as the Centers for Disease Control and Prevention (CDC); Universities in the US, Canada, China, England, France and Germany; and scientific industry partners in the US and India. The authors also include independent data analysts without affiliation, such as Yuyang Gu, who took the internet by storm with his early successful modeling efforts of the pandemic.

The Forecast Hub is directed by Nicholas Reich and Evan Ray, faculty of the UMass School of Public Health and Health Sciences. “Collaborating directly with so many talented and inspired groups to build this ensemble forecast has been an incredible experience,” says Reich, a biostatistician and senior author of the paper. “In addition to the operational aspect of the hub, where forecasts have been used by CDC every week for the past two years, this paper shows how we can use these data, collected in real time throughout the pandemic, to better It’s going to take many years to unpack all the lessons of the past few years to really understand which modeling approaches worked and which didn’t, and why. In some ways, this is just the beginning.”

In April 2020, CDC partnered with Reach Labs to create and fund a COVID-19 Forecast Hub. At this time, Hub began collecting, disseminating and synthesizing specific predictions from various academic, industry and independent research groups. The effort grew rapidly, and in its first two years the US Forecast Hub collected more than half a billion rows of forecast data from nearly 100 research groups. The CDC uses the Hub’s weekly forecast in official public communications about the pandemic.

The paper compared the accuracy of short-term forecasts of US-based COVID-19 deaths during the first year and a half of the pandemic. The 27 different models that produced consistent forecasts during that period showed high variation in accuracy across time, locations and forecast horizon. The ensemble model combining individual predictions was more accurate than those individual predictions.

“This project demonstrates the importance of diversity in modeling approaches and modeling assumptions,” Kramer says. “The inclusion of different models in the ensemble contributes to its robustness and ability to address individual model biases. It is really important for public health agencies to use forecasts to inform policies during an outbreak of any size. important consideration.”

The Forecast Hub ensemble was the only model that ranked in the top half of all models for more than 85% of forecasts, had better overall accuracy than baseline forecasts in each location, and was overall better four-week-ahead. Was. Accuracy compared to baseline forecast in each week.

All forecasts, including the ensemble model, produced less consistent and less accurate forecasts during the four waves of the pandemic that occurred during the study period: the summer 2020 wave in the South and Southwest, an increase in deaths in late fall 2020, spring 2021 in the Upper Midwest, Michigan The alpha version wave and the nationwide delta version wave in the summer of 2021. “In general the model systematically lowered the mortality curve as trends were rising and trends were falling,” the paper states.

Forecasts became less accurate as models made longer-term predictions. The probabilistic error on the 20-week horizon was three to five times greater than that predicted on the one-week horizon. This resulted in an underestimation of the possibility of a future increase in cases, concludes the paper. “Since many of us interact with weather forecasters on our phones almost every day, we know not to rely on daily precipitation forecasts past the two-week horizon,” Reich says. “But we don’t yet have a society-like intuition about infectious disease forecasts. This work shows that the accuracy of forecasting deaths is very good for the next four weeks, but on the horizon of six weeks or more, accuracy.” is generally quite bad.”

The open-source infrastructure built by the US COVID-19 Forecast Hub team has also been used around the world, including the hub run by the European Centers for Disease Control and Prevention, long overseen by German academic researchers and other US researchers. Huh. Modeling of various “what if” scenarios.

Nation World News Desk
Nation World News Deskhttps://nationworldnews.com
Nation World News is the fastest emerging news website covering all the latest news, world’s top stories, science news entertainment sports cricket’s latest discoveries, new technology gadgets, politics news, and more.
Latest news
Related news
- Advertisement -