The Centers for Disease Control and Prevention (CDC) plays a huge role in making sure healthcare providers and hospitals around the country are prepared for seasonal flu outbreaks by predicting and forecasting what to expect every year.
Epidemiological forecasting is actually much like weather forecasting because a lot of data is required to make an intelligent forecast for the coming flu season. Correct forecasting will affect the type of vaccine being used and the preparation that may be required.
But what if we could use data to predict the spread of the flu virus? Data from the retail sales of flu medications, Google searches about flu symptoms or tweets about symptoms all create a picture in near real-time that tells us a great deal about flu outbreaks. If we could use this information wisely, it would make predicting the spread of a disease as common as predicting a rainstorm.
CDC forecasting research initiative
For the past four years, the CDC has run an initiative called “Predict the Influenza Season Challenge.” The goal is to build models and methods to predict what a particular flu season might bring. Participants were asked to forecast the timing, peak and intensity of the coming flu season using digital data.
A variety of sources such as data from social media sites like Facebook and Twitter, and Internet search engines like Google, as well as innovative modeling approaches, were used to forecast national and regional flu activity every year. When the initiative was first started in November 2013, there were 11 participating organizations or groups.
This year, participation in the CDC’s initiative grew to 28 participants, and their forecasting was judged on their ability to accurately predict the coming flu season based on four criteria, including:
1. When the season will start
2. When it will peak
3. How severe will the peak be
4. How severe will it be in one, two, three and four weeks’ time
“Wisdom of the crowds’ approach”
As it turns out, two forecasts developed by Carnegie Mellon University’s Delphi research group claimed the number 1 and 2 spots among 28 submissions from other universities, government agencies, and private organizations this year. They used a “wisdom of the crowd” approach in achieving the “highest skill” award.
CMU used artificial intelligence, Delphi-Stat — machine learning technology — “to make predictions based on past patterns and on input from the CDC’s domestic influenza surveillance system,” the university said in a release. CMU’s second system used a “wisdom of the crowds” concept that turned out to be quite interesting.
The “Wisdom of the crowd” or Delphi-Epicast method involved about 100 volunteers who competed week after week with predictions based in part on historic flu data provided to them. Delphi-Stat did slightly better on short-term forecasts while Delphi-Epicast did slightly better on long-term forecasts, said Ryan Tibshirani, a Delphi group member, an associate professor of statistics and machine learning, in a news statement.
The Delphi group is not stopping with only forecasting seasonal flu outbreaks. They are also working on developing a forecasting method for dengue fever, with future plans to use its forecasting tools for various diseases including the Human Immunodeficiency Virus, drug resistance, and such epidemic viral infections as Ebola, Zika, and Chikungunya.
“We’re gratified that our forecasting methods continue to perform as well as they do, but it’s important to remember that epidemiological forecasting remains in its infancy,” said Roni Rosenfeld, Delphi leader and professor in the CMU School of Computer Science’s Machine Learning Department and Language Technologies Institute.
Rosenfeld likens epidemiological forecasting to early weather forecasting. “At the time that it started, people didn’t realize how useful it would be economically and socially, and how much it could progress,” he said of weather forecasting’s early years. “It took many, many years — many, many decades — of development across multiple dimensions.”
