When and where will the next COVID-19 hotspot appear in the U.S? What populations will be impacted most? What interventions might best reduce the spread and save lives?
While no one can predict such outcomes with 100% accuracy, there are experts who can come pretty close. One of those experts is Thomas McAndrew, a professor in Lehigh University’s new College of Health. McAndrew, whose research focuses on forecasting infectious diseases, has developed a promising approach that uses computational models based on data, in concert with human judgement, to produce more accurate forecasts of COVID-19 outbreaks and determine the optimal interventions to lessen the impact.
“While computational models are great for making predictions from structured data, information from human forecasters who understand the social factors that contribute to health outcomes is also important,” says McAndrew. “The urgency and evolving nature of the COVID-19 pandemic requires that we marshal all resources at our disposal to stay ahead of the disease. The hybrid ensemble we plan to develop has the potential to forecast outbreaks more accurately than current ensemble predictions because it incorporates computational models trained on formatted datasets and predictions from humans who have access to unstructured data.”
McAndrew’s innovative approach recently received support from the MIDAS Coordination Center based at the University of Pittsburgh. MIDAS, or the Models of Infectious Disease Agent Study, aims “to advance science to improve global preparedness and response against infectious disease threats.” It is funded by the National Institutes of Health, National Institute of General Medical Sciences.
“This award will allow Dr. McAndrew to expand his innovative biostatistical modeling approach to predict COVID-19 outbreaks,” says Whitney P. Witt, Inaugural Dean of Lehigh’s College of Health. “This exciting and impactful new research stands to provide more accuracy in the tracking of the pandemic, inform policy and practice, and reduce morbidity and mortality.”
McAndrew is among a group of forecasting researchers who are urging a substantial shift from the current status quo.
“While computational models alone are usually sufficient to provide accurate forecasts when data is plentiful, a fast-moving pandemic requires a shift to a new paradigm. This work aims to mobilize our expert knowledge-base and produce actionable predictions that can have an impact on improving health outcomes,” says McAndrew.
His research is designed to inform public health decision-making and make a tangible impact on the course of the pandemic in the U.S. McAndrew and his team will provide probabilistic forecasts of the U.S. COVID-19 outbreak via monthly summary forecasts and reports for members of the MIDAS community, members of the Centers for Disease Control and Prevention (CDC) and the Council of State and Territorial Epidemiologists (CSTE), as well as to the general public. Summaries will include combined (or ensemble) predictions from statistical models and subject matter experts. A unique aspect of these reports will be commentary solicited from subject matter experts and trained forecasters about the pandemic’s trajectory—information only humans can provide, says McAndrew.
The work will tightly integrate with current computational forecasts of COVID-19 that are being led by Dr. Nicholas G. Reich, one of two directors of the CDC’s Influenza Forecasting Centers of Excellence established in 2019. McAndrew did his post-doctoral research with Reich and was part of a team that developed forecasting models of COVID-19 that were regularly shared with public health officials, including the CDC, to guide decision-making.
The work McAndrew and colleagues are doing will also model possible interventions by collecting probabilistic predictions from experts on optimal interventions to reduce the spread of COVID-19 in the U.S., offering insight into what strategies might be most effective in preventing community spread and adverse health outcomes.
“An expert consensus of optimal interventions can help translate statistical predictions and the current political and social atmosphere into meaningful actions to reduce the impact of COVID-19,” says McAndrew.
Ultimately, McAndrew plans to produce combined statistical and human judgment meta-forecasts to support decisions made by public health officials, making the code and data open source and available to support the scientific community and future work combining human judgment and computational models.
“My goal is to use statistical and computational methods to do good in our world: to support public health decision making, and to support my scientific community,” says McAndrew.
