Skip to content Skip to navigation

Trying to Solve Preventable Readmissions with Predictive Analytics

June 3, 2013
by Gabriel Perna
| Reprints

Heritage Provider Network, an integrated care delivery network, based out of Northridge, Calif., recently announced the results of an analytics competition it held, handing out a $500,000 award to a team that had come closest to reaching the benchmark it set. The announcement was made at Health Datapalooza IV, in Washington D.C.

The competition asked data miners to take four years of sequential data from anonymous Southern California health plan enrollees and create an algorithm that predicts how many days a patient will spend in the hospital in a given year. The winning team, named POWERDOT, comprised of former rivals who joined forces to work on getting to the benchmark collectively, came the closest.

In addition, Heritage announced that because no one had reached or surpassed the benchmark, it is extending the competition for the “best of the best.” It will dish out $3 million to the first teams that creates an algorithm that reaches or exceeds the benchmark of .40.

In an interview with Healthcare Informatics, Heritage’s senior executive, Jonathan Gluck says the purpose of the competition was to reduce preventable readmissions. “If you can predict who is going to get hospitalized, you can figure out which ones are preventable through providing preventable primary care, and prevent them,” he says.

The second time around Gluck says it is extending the dataset, no longer restricted by HIPAA [the Health Insurance Portability and Accountability Act], Gluck says. “It’s all HIPAA compliant [this time] therefore they have more data in which to create even better algorithms,” he adds.  

The idea of a challenge, Gluck says, has allowed Heritage to think outside the box, compared to using a vendor solution. Ultimately, Heritage hopes these algorithms can be plugged in and used for all 700,000 of its members. The information can then be sent to physicians, where they’ll be informed of which patients are in danger of being admitted to the hospital unnecessarily.

“At the end of the day it’s about providing better healthcare to the population, and that’s what the algorithim allows us to do and that’s why we’re running this challenge,” Gluck says.