The $3 Million Heritage Health Prize

Jonathan Gluck (Heritage Provider Network)

Good morning. My name is Jonathan Gluck, and I’m from Heritage Provider Network, which is one of the largest physicians’ groups in the United States. Heritage Provider Network is not an insurance company, but instead is a group of doctors that provides the actual medical care to the members.

So why is a doctors’ group at a data conference? The answer is that better predictive modeling and data mining is precisely what health care needs. The health care system in the United States is currently not a health care system, but instead we can think of it as a sick care system. We provide care to people after they have already become ill. What we need to do is move towards a system that seeks to keep people healthy and prevents them from becoming sick.

So, the challenge we have for you is conceptually simple. Create an algorithm that will predict who will go to the hospital from a given patient population. If an algorithm of that sort can be created, doctors can intercede with the patients and seek to keep them healthy and prevent them from going to the hospital. That will help us move towards a health care system. If you can create the winning algorithm, you’ll win $3 million. There will also be milestone prizes along the way for teams leading at given points in time.

Today, we’re excited to announce that the competition phase will begin on April 4th, 2011. Now let me turn it over to Anthony, who is going to give you a few more details about the prize.

Anthony Goldbloom (Kaggle)

Thanks a lot, Jonathan. I’m Anthony Goldbloom, Kaggle’s founder, and we’ll be running the prize. So, if you choose to participate, you’ll see a lot more of us. Well, first off, welcome to the biggest ever data prize. This is really exciting.

Three things I wanted to say. First of all, why a prize? For those of you who don’t know about Kaggle, we run data predictive modeling competitions. So, we’ve run a whole range of things, and what we find is that by throwing the problem open to a wide audience, everybody tries their own different techniques and you very quickly make rapid improvements on the benchmark. We’ve never hosted a competition where there’s been a benchmark, a model that’s been built in-house, that hasn’t been outperformed.

On the slide at the moment, you can see some of our champions. They come from all over the place. These are the people who have finished in the top three in our competitions. So, we have a physicist from Portugal, a mathematician from Israel, an engineer from New Mexico. Very often, the best answers aren’t coming from where you might expect them to come from.

Our first competition was won by a 25-year-old computer scientist from the University of Ljubljana in Slovenia. He beat a team from MIT, a team from SAS, and I suspect, if I was contracting for that job, I would not have picked him, and I would have missed out on the best model.

What sort of data can you expect to see? HPN are releasing data on things like patients’ hospital visits, doctor’s visits, what lab tests they’ve had, have they had a blood test, what meds they’re on. Your algorithm might say . . . here’s an example of a sick patient that your algorithm might flag as being at risk. They might have diabetes, high cholesterol, hypertension, and they’re not filling their prescriptions. It’s pretty obvious that this guy’s crook, and there’s a high chance he or she is going to hospital. There are many others like him in the dataset, less obvious, and I guess we want you to find them.

I don’t think it’s a stretch to say this is probably the most exciting thing to happen in 2011 in this space. How many here have heard of the Netflix Prize? I suspect a majority. Yeah. How many competed? Yes, so there’s a smattering.

The Netflix Prize solved a grave problem. It helped prevent you from seeing bad movies. The $3 million dollar Heritage Health Prize, aside from the fact the prize is three times bigger, solves a somewhat graver problem. It prevents people from going to the hospital and getting really sick where it’s preventable. I’m very excited about this, and I hope to see many more of you entering the competition.

I have just a couple of things to say. There’s a Heritage Health Prize booth in the Exhibitor Hall. For anybody who wants more information, just come by and have a chat, and I’ll be giving a talk at 1:40 where I’ll set more of this stuff out.