What if you could predict if a given patient were at a higher risk for hospitalization in the coming year? You could potentially save money, and lives, by pulling out all the stops to prevent that hospital visit, if possible. And that's why the Heritage Provider Network (HPN) has put up $3 million for a Netflix Prize-style competition that will pit coders against each other to devise the most effective predictive algorithm for incipient hospitalizations. HPN will be announcing a launch date for the prize this week.
"If you can predict it, you can get ahead of it," Jonathan Gluck, a senior executive at HPN, tells Fast Company. "You can tell a physician, 'This person has a combination of factors that we've seen leads to hospitalization,'" prompting that doctor to pay particular attention, to treat aggressively, and to emphasize the severity of the situation to the patient. "The ultimate goal is to provide better medical care," says Gluck.
HPN has assembled data on 100,000 patients, which it will be sharing with contest entrants. ("It's all HIPAA-compliant," assures Gluck; the patients cannot be reidentified.) Lab data, prescription information, treatment plans—it's all there. "Teams then look at the data and create an algorithm that says, in the year following the data, did they wind up going to hospital?" Since the data is all from a few years back, the answers are available, so the coders can test themselves.
The data will be parceled out into three subsets, actually. The first and largest set will be the practice data set and will include information about where the patients wound up. The latter two sets, the "quiz set" and "test set," will be used to test the algorithms the teams develop. Whoever comes up with the most accurate algorithm scoops up the $3 million prize. HPN teamed up with the Australian data mining group Kaggle to devise the contest.
But is having doctors rely on an algorithm to single out patients in danger such a good idea? Isn't that a skill you want the doctor to have herself? A shortage of primary care physicians in the U.S. means that doctors don't always have time to pick up on the subtle connections that might lead to a Gregory House-style epiphany of what's ailing a patient. More importantly, though, the algorithms may point to connections that a human mind simply would never make in the first place. "By looking at the data, we're hoping to pick out the combination of factors that someone might miss, because they haven't seen a population with that combination, or they've never made a connection between this drug, that visit, missing a visit, and going to the hospital. It's just too many permutations to keep in somebody's head." (All of which reminds us of DARPA's crime-predicting software it has in mind.)
If doctors won't abuse the system, then, could insurance providers? Ultimately, the algorithm will help reduce the roughly $30 billion dollars wasted on preventable hospital visits, improving doctors' radars for who is at risk for hospitalization and improving the care for those patients along the way. "We need to start looking at the situation this way: If someone went to the hospital for an avoidable admission, we all failed," says Gluck.
[Image: Flickr user erix]