Excuse us if we interrupt the flow of information for a little bragging. It’s not every day a blog like this scoops The Wall Street Journal, but it’s just caught up with the fact that Melbourne-based Kaggle is running a $3 million contest sponsored by Heritage Healthcare for new data mining algorithms that could “change the world”. Its story, headlined May the Best Algorithm Win [pay wall] is, let’s see now, almost two months behind the news, which we ran here and even anticipated more than a month earlier here.
The WSJ does have some interesting quotes from Kaggle founder, Anthony Goldbloom – who at the moment is leading a peripatetic existence commuting between Melbourne and San Francisco, as the start-up shifts its HQ – which might well be taken on board by those responsible for health datasets:
Preparing for the Heritage prize has required months of work by two Kaggle employees, in part because the real-world data is complicated and messy, Mr. Goldbloom said.
“You might see somebody who was diagnosed with a breast neoplasm—a lump on the breast—and the treatment code was a circumcision. There’s clearly something not right there,” he said.
And Kaggle is working with a specialist to render the information anonymous, so it can’t be traced back to any patients. This includes not only removing names and addresses, but taking out information about dates and even treatment codes. The focus on privacy comes after Netflix had to scuttle a second prize when researchers were able to extract personal information from some of the Netflix data.
What the Journal didn’t pick up, though, is that to get the competitive juices flowing the Heritage Provider Network has announced the $3 million Heritage Health Prize – aimed at developing “a breakthrough algorithm that uses available patient data, including health records and claims data, to predict and prevent unnecessary hospitalisations” – will also include progress prizes totaling $230,000. These prizes will be awarded to teams leading the competition at various milestones:
- $30,000 to the first place team at the six month mark
- $20,000 to the second place team at the six month mark
- $50,000 to the first place team at the 12 month mark
- $30,000 to the second place team at the 12 month mark
- $60,000 to the first place team at the 18 month mark
- $40,000 to the second place team at the 18 month mark