The world is not as it appears to be. Our judgments are clouded by biases and mental shortcuts, otherwise known as heuristics. The availability heuristic, for example, causes us to disregard information that isn't top-of-mind. Humans often only consider the available information when confronted with a new decision.
Daniel Kahneman, along with authors Cass Sunstein and Olivier Sibony, recently published their newest exploration into the problems of human decision making, Noise. “Most of us, most of the time, live with the unquestioned belief that the world looks as it does because that’s the way it is,” they write. The culprit for this muddled view of reality, they argue, is noise, a form of systematic deviation and random scatter of outcomes in situations that one would expect to result in precision and accuracy. Kahneman and his co-authors find, “Professional forecasters offer highly variable predictions about likely sales of a new product, likely growth in the unemployment rate...just about everything else.”
The world of medicine is not immune to the problems brought about by noise. Reducing noise in medicine seems, at first glance, like a behavioral challenge. Must facilities implement new practices to ensure systematic noise is accounted for? These changes can be costly and require significant updates in clinical practice, which may be non-starters. The challenge arises in finding a technical solution to reduce noise in healthcare that is sufficiently interesting, yet sufficiently difficult that it hasn’t yet been solved, but the time for solving it has come now.
One such case is predicting surgery durations. Currently, the primary solutions used by hospitals and ambulatory surgery centers utilize moving averages and surgeon estimates. Both of these are inherently noisy. A surgery predicted to go too short will push back the next operation and one that is predicted with a time that is too long leaves a perfectly good OR unoccupied. This is where artificial intelligence based solutions like Opollo can fill the gap.
In Opollo Technologies’ latest rounds of customer interviews, we are finding that customers are hungry for AI solutions to business problems. They want to know how those solutions work, what the return on investment looks like, and who is building these products.
Although the perceived paths to achieving these goals can vary between customers, there’s almost universal agreement that advanced technology is a must for facilities to compete moving forward. One Chief of Anesthesiology at a major national children’s hospital told us a great story highlighting this need. Two weeks ago, he was standing next to a colleague looking at the daily surgery schedule. He said to his colleague, “Wouldn’t it be cool if you had some AI system that could provide accurate duration forecasts and make suggestions to improve on what’s going on with surgeries that we’re scheduling?” We couldn’t help but agree!
A healthcare manager at a system in Central Pennsylvania confirmed a wide-spread interest in AI based solutions, but cautioned integration and buy-in can be roadblocks. The takeaway is that any technical solution must have a clear business case which can be easily understood by decision makers. The use case must be clearly demonstrated.
All of this must be built on a foundation of expertise. Expertise in medicine is required to understand the numerous challenges associated with scheduling surgeries in the operating room. You need to know about the surgeon’s needs, the patient’s background, and the complexities of different procedures. A simple linear regression of these variables does not produce optimal results. Good machine learning in medicine starts with subject matter experts. Did we mention our founder is an MD/MBA with experience in quantitative finance, artificial intelligence, and medical natural language processing?
- - -