Winning the predictive prize—at least in the interim

Hospitals and health plans aren't the only ones interested in developing an algorithm for predicting hospitalizations, as I wrote about in this week's magazine.

In California, one medical group network has taken another approach to the question of predicting hospital stays. The tactic has drawn physicists, mathematicians, computer scientists and stochastic modelers (a method to estimate probability, sometimes used in financial markets) to wrestle with the question.

Dr. Richard Merkin, president and CEO of the Heritage Provider Network, a group of nine affiliated medical groups, pledged a $3 million prize for a predictive model that would identify 85% of patients who end up hospitalized using the network's historical data.

The contest, launched in early 2011 and open for three years, awarded its first interim prize in September to the team leading the field of 4,500 entrants. David Vogel, a hedge fund manager with Voloridge Investment Management, Palm Beach, Fla.; healthcare consultant Dr. Randy Axelrod; and Australian data-miner Phil Brierley claimed the $30,000 interim award.

Vogel, who has a decade of experience with predictive modeling in healthcare, said he applied the same technology to predicting hospitalization as he uses to forecast changes in securities prices.

The $3 million competition, Vogel said, will demonstrate the potential to improve the accuracy of predictive modeling in healthcare.

Merkin said rules allow rivals to join together to strengthen entries in what has become a fiercely competitive contest.

Merkin said he would like to see funding for research grow as healthcare delivery grows more efficient.

He said he launched the contest because a prize-winning predictive model could help reduce wasteful healthcare spending.

“The best way to predict the future is to try to create it,” Merkin said.

You can follow Melanie Evans on Twitter @MHMEvans.

You can follow Melanie Evans on Twitter: @MHmevans.