Enroll America is an organization with a big job: to find the uninsured around the country and persuade them to sign up for health insurance. It needs a good way to find them.
The organization realized early that it would need a detailed picture of just who was uninsured int he country and where they lived. [Enroll America] partnered with the data group Civis Analytics […]
To start, it conducted a huge phone survey of 12,000 adults around the country who use cellphones and land lines, asking people 10 basic questions in either English or Spanish. Then it folded in publicly available data about those individuals and their communities, using commercial databases and statistics from the census. It considered about 500 variables; the final model used a smaller number, 30, that together made a strong prediction. The goal was to create a score for each person that reflected the odds that the person would be uninsured.
“We started using the model initially to try to get to people at an individual level, and there was a degree of accuracy there,” said Gregg Ross, Enroll’s organizing director in North Carolina, one of 11 states where the group made a big push. “Once we started using it for this geographic targeting and coupling it with our knowledge of the communities, we were finding that upward of 80 percent of the people we contacted were uninsured.”