I had a professor in undergrad who worked on order entry research. He talked about one of the problems facing Vanderbilt as far as scanning costs.
CT scans and MRI are crazy expensive. Trauma patient comes in, doc orders a CT scan, it comes back and he says, "shit, this doesn't tell me what I need to. I knew I should have ordered an MRI instead."
They used a decision tree learning algorithm and trained it using attributes of incoming patients and which type of scan would have been the most useful. Then they did a trail where ER docs would enter in patient information and what they would have picked, but then defer to the algorithm's judgement instead.
I can't remember the number, but they significantly reduced the number of unnecessary scans.
CT scans and MRI are crazy expensive. Trauma patient comes in, doc orders a CT scan, it comes back and he says, "shit, this doesn't tell me what I need to. I knew I should have ordered an MRI instead."
They used a decision tree learning algorithm and trained it using attributes of incoming patients and which type of scan would have been the most useful. Then they did a trail where ER docs would enter in patient information and what they would have picked, but then defer to the algorithm's judgement instead.
I can't remember the number, but they significantly reduced the number of unnecessary scans.