The Alliance is a network of physicians in New England covering roughly 475,000 patients. Cantor was struck during a biannual conference held by the network by a panel of patients that had successfully used care management to improve their lives. The patients, he realized, were mostly retirees.
“They were at a point in their lives when they were ready to make changes,” he said. How could the network know what types of patients are receptive to such changes? Or for whom such changes would be most effective?
“Because care management programs are difficult to manage, and they're difficult to staff, and they're very expensive, we want to make sure we target those people who are most likely to agree to participate in those programs,” he said.
Supplementing medical data with consumer data might lead to better predictions, he, and the alliance, reasoned.
In the pilot program, the network will send its health data to a modeler, which will pair that information with consumer data, such as credit card and Google usage. The modeler doesn't necessarily have a hypothesis going in, Cantor said.
“They're identifying correlations between the consumer data and healthcare outcomes,” he said.
The modelers won't necessarily give Cantor's group the key predictive attributes; they'll give the network a list. “Here's a list of 100,000 people: who's likely to end up in the ER, who's likely to take their medicines. They're still early on; it's not like there's a definitive model,” he explained.
That uncertainty makes Cantor a bit uncomfortable. He knows that these models have been used in other sectors, from finance to retail and beyond. But using such models in healthcare creates some issues.
For example, he said, “there are five people named Mike Cantor who live in the Boston area. If I get someone's LL Bean catalogue instead of the other Mike Cantor, OK, no harm no foul.”
It's entirely different if, instead of a catalogue, Cantor gets a call from a care manager warning him about his risk of diabetes. “As we move into healthcare, the need for precision is that much greater. How are we going to make sure we accomplish that?”
Cantor has seen a lot of problems in the data from consumer entities—name confusions, or just plain wrong data. As such, he sees the need for more precise data, to ensure it's more responsibly used.
“If I look up information on someone's chart, hopefully that information is extremely precise. It is the right data, with the right patient, with the right results, in the right format,” he said. “Because there's so much at stake, because decisions are made based on those data, we've spent a huge amount of time and energy making sure those data are accurate.” That hasn't necessarily been true for consumer databases.
The other issue from using such models is validation; some of the models haven't necessarily been rigorously compared to their peers, let alone versus human instinct.
The patients' awareness of such tools, he thinks, also has been lagging, and the healthcare sector needs to educate them on that subject. “People should understand these tools are evolving, and we should be open to hear how people think about that.”
Follow Darius Tahir on Twitter: @dariustahir
