Carnegie Mellon University
May 12, 2021

Delphi Research Group, Collaborators Honored for COVIDcast

By Stacy Kish

The American Statistical Association (ASA) has presented the 2021 Statistical Partnerships Among Academe, Industry, and Government (SPAIG) Award to Roni Rosenfeld and Ryan Tibshirani with Carnegie Mellon University’s Delphi Research Group, as well as their COVIDcast partners.

The team was recognized for its “commitment to the theory and practice of epidemic tracking and forecasting through building and modeling unique public health data streams.” 

This award also honors Carrie Reed, Matt Biggerstaff, Michael Johansson, Rachel Slayton, Velma Lopez, Jo Walker and the CDC COVID-19 Modeling Team; Hal Varian, Brett Slatkin, the Google Surveys Team and Google.org's CMU-Delphi Fellows; Kang-Xing Jin, Curtiss Cobb and the Demography and Survey Science, Data for Good and Health teams at Facebook; Swami Sivasubramanian, Alex Smola and Amazon AI at Amazon Web Services (AWS); Tim Suther, Craig Midgett, Andrew Harris, Mina Atia, Anil Konda and Jaydeep Kulkarni at Change Healthcare; John Santelli, Paul Nielsen, Danita Kiser and the Optum data team at Optum and John Tamerius, Jhobe Steadman and Torsten Auhorn at Quidel, Inc.

“We grew from a pre-pandemic team of seven or eight to more than 50 members, who are mostly volunteers,” said Tibshirani, associate professor in the Statistics & Data Science and Machine Learning departments at CMU. “They are giving us time from their normal ‘day jobs’ throughout the pandemic, so it is extremely rewarding to receive recognition for everyone’s contribution to these efforts.”

A Quick Pivot from Flu to COVID

The Delphi Research Group was assembled to make epidemiological tools to improve forecasting efforts to help healthcare organizations. While originally focused on seasonal influenza, the team shifted its approach to focus on tracking COVID-19 in the early months of 2020. The group worked with industry partners to evaluate large databases to track illness, monitor mask wearing and explore vaccine hesitancy across the country.

“There are very few good things about a pandemic, but the one thing that was uplifting was the large number of people and organizations that were moved by it to operate in a way that you don’t often see,” said Rosenfeld, professor and head of the Machine Learning Department. “In a sense, it was like we are fighting a war together and in a war you discover your shared humanity. That made me feel good.”

While the Delphi Group has worked with the CDC since 2012, most of the partnerships with industry and healthcare are new. Quidel Corp shared de-identified antigen test data. Change Healthcare and Optum shared de-identified medical insurance claims. 

“It has been useful to look at both public health records as well as medical insurance claims because public health reporting often can have unpredictable delays or exhibit other data errors,” said Tibshirani. “These different data streams provided a check for our work.” 

Facebook offered its platform and access to large membership to run anonymous surveys to gain insight in the spread of symptoms, concerns about personal finances, mask-wearing and vaccine hesitancy, to name only a few. To date, the team has gathered more than 20 million survey results. Google also ran surveys for Delphi towards the beginning of the pandemic, and more recently helped to track search inquiries for COVID symptoms, like loss of the sense of smell or taste, and provided 13 full-time fellows for six months. Google also donated $1 million to the project. Finally, AWS provided COVIDcast cloud computing support.

A Path Forward

The Delphi approach has been unprecedented. While most governmental information is targeted at the state level, few datasets offer the local perspective needed for public health. The endeavors of Delphi and its partners made it possible to dig into the granularity of data at the county level and lower, allowing them to view changes that could be impactful for public health.

“A silver lining of this whole experience is that it provides the blueprint on what to do next,” said Tibshirani. “Everyone is now keenly aware of the importance of the long-game of epidemic and pandemic tracking, moving it to an operational science to understand the importance of auxiliary data streams.”

According to Rosenfeld, last year was an opening salvo in the war against viruses and disease. While tragic and unnerving, it provides the world a window of clarity to understand the risk and danger inherent in these microscopic threats and the opportunity to prepare so the world can manage the next battle more effectively.

“The general consensus is that this is not a once in a hundred years pandemic, but something like it will happen again sooner,” said Rosenfeld. “We are likely to encounter something as challenging again in the next decade, so I feel a sense of urgency. We need to leverage this and other partnerships to improve how we handle public health emergencies moving forward.”

The annual ASA SPAIG Award, which was established in 2002, highlights outstanding partnerships between academe, industry and government organizations and promotes new partnerships. The Delphi Group was joined this year by the Intermediate Clinical Endpoints of Cancer of the Prostate (ICECaP) Working Group at the Dana-Farber Cancer Institute in receiving this honor. The award is sponsored by the SPAIG committee of the ASA and is distinct from other ASA awards in that it recognizes outstanding collaborations between organizations, while recognizing key individual contributors.