Flu forecasts within large metro areas like New York City might be improved by adding in data about the flow of commuters. Christopher Intagliata reports.
Every flu season—that’s now through the spring—epidemiologists track flu infections as they break out across the country. And they forecast how bad it's going to get: at the national level, regionally, state by state. They even forecast for metro areas, like New York City, and L.A. Which sounds pretty fine-grained, until you consider that New York City is made up of five boroughs. And that there are actually more than 80 cities… in L.A. County.
So there might be an advantage to forecasting at even smaller scales. "Public health decision-making and interventions are done at small scales, they're done at the municipal and county scale." Jeffrey Shaman, an infectious disease modeler at Columbia University.
He and his team built a model to forecast flu within New York City neighborhoods and boroughs, using data on flu cases from 2008 through 2013. They added in something they called "network connectivity"—commuter data, basically. The commuter data didn't improve the accuracy of hyper-local, neighborhood-level forecasts. But it did improve predictions at the borough level, compared to models without that sort of commuter flow built in. The results are in the journal PLoS Computational Biology. [Wan Yang, Donald R. Olson, Jeffrey Shaman: Forecasting Influenza Outbreaks in Boroughs and Neighborhoods of New York City]
Shaman says fine-tuned forecasts could warn local hospitals before a big outbreak. "Knowing when that's going to be will allow them to plan the resources out. Have the staff available. They need gloves, beds, they need ventilators, they need to have those appropriately available in time so they can meet that patient surge." And—so they can stop the virus' spread in that most local of networks: within the hospital itself.
[The above text is a transcript of this podcast.]