“That’ll be $189, please.”
Two researchers at Indiana University, Kris Hauser and Casey Bennett, have developed a predictive system for determining how to treat patients that adapts to new health information over time, and it could reduce costs by up to 50 percent while similarly improving patient outcomes. The system, published in Artificial Intelligence in Medicine journal in January (PDF), builds upon the researchers’ earlier work, which didn’t include the ability to adapt over time.
Courtesy of Indiana University.
Savings were $189 for the system’s suggested treatments versus $497 for the treatment as usual case when doctors of the human variety were on the case. Patient outcome improvements were between 30 to 35 percent, although Hauser and Bennett believe this could be improved to around 50 percent.
The system data mines electronic health records and uses two particular types of machine learning algorithms, Markov decision processes, and dynamic decision networks to learn from clinical data, effectively training the system to “think like a doctor.” Hauser and Bennett are looking at improving the system further by including more information in the training steps, such as genomic data and history of disease progression, as well as trying to train it to take account of events such as patient non-treatment (where medication is not taken or taken incorrectly) and the variability of existing doctors’ effectiveness.
It’s a matter of time before this new and improved system gets a supporting role in a science fiction TV series and begins demanding salary and benefits.
I got it, but maybe next time you might want to explain who and what the character in the photo is, and what sci-fi TV series he’s from (“The Doctor”, a computer-driven holographic physician on Star Trek: Voyager)