Sometimes it takes “a beautiful mind” to answer a simple question. Where, for example, should a community build a hospital?
GE data scientists set out to answer that question using game theory, the fiendishly difficult branch of mathematics pioneered by John von Neumann and John Forbes Nash, Jr. (portrayed by Russell Crowe in the 2001 biopic A Beautiful Mind). Mathematicians, economists and generals have used game theory to model complex behavior and make strategic decisions for winning in everything from shopping, socializing and market competition to politics and nuclear war. The results can be profound. Eight game theorists, including Nash, have received the Nobel Prize.
The GE team has applied insights from the discipline to healthcare. They gathered demographic, economic, and epidemiological data and built a system, called Corvix, to simulate, improve and expand the treatment of heart disease. “We are converting big data into intelligent data,” says Mitch Higashi, chief economist at GE Healthcare. “We are creating digital clones of human beings and they live out their lives in the cloud. It’s almost like running different versions of the future. How does the population health trajectory change over time? We can accelerate getting to the right answer.”
The team is testing the system in the Indian state of Andhra Pradesh.
Higashi says that the system seeks to find an “optimal solution” to a problem, a point where the most “players” receive the most benefit. Game theorists call it the Nash equilibrium.
Healthcare, for example, has a number of stakeholders, from patients and government officials, to doctors, equipment makers, and hospitals. The GE system could help planners build hospitals in the right place for the right price and fill them with the right equipment so that the most people can live longer and better lives. “Quality-adjusted life-years is one of the outputs of the model,” Higashi says. “That’s the utility measure we use.”
Data scientists and health economists at GE Healthcare spent two years developing the system. First they partnered with the U.S. Department of Energy’s Argonne National Laboratory to help them build a new operating system for Corvix that converts data sets into virtual beings.
The Atom Group, a cloud architecture firm, then helped the GE team load and store the massive amounts of information, and Mad*Pow, a game design company, developed the intuitive interface. “The idea was that anyone in the planning space could interact with the data in the cloud in real time,” Higashi says. “You can drag and drop a virtual hospital right on top of a map and then activate the big data in the cloud to start running a forecast.”
The GE team currently feeds the system with two sets of census data and information from an Indian national health survey. But in the future Corvix could learn directly from anonymous patient and imaging equipment data. “The ‘cloud’ cities of Andhra Pradesh could be a mirror snapshot of what’s happening there on day-to-day basis,” Higashi says.
Higashi says that healthcare is only the beginning for systems like Corvix. “In the future, all policy planning will be developed in the digital world before being implemented in the real world,” Higashi says. “We see this going beyond healthcare to help develop environmental, energy, unemployment, and education policies, those tough, intractable problems that society has had a hard time with.”
March 25, 2013