Skip to 0 minutes and 14 seconds So what have we learned from our experiences with electronic decision support so far? We know that human factors design is absolutely critical, and yet we still don’t pay enough attention to it. We know that context is important, the hierarchy of the medical profession, how decisions are made? Where decisions are made? We need to understand that. But also we can draw another discipline area such as behavioral economics, and that really studies how do individuals make decisions,
Skip to 0 minutes and 45 seconds and I think this is a lovely quote from Daniel Kahneman: “We think much less than we think we think,” and he’s written some very interesting work on how do individuals make decisions. And some of this research on decision making is very interesting and very applicable to the type of work that we’re doing, so we know for example if you provide people with the option of should they opt in, or opt out, people will tend to maintain the status quo. If you put an item on the top of the list, people will tend to choose that. That’s why every politician wants to be on the top of the ballot paper.
Skip to 1 minute and 27 seconds We know from health care that presenting at antibiotic choice grouped according to narrow or broad spectrum rather than if you list individual drugs can result in more significant reduction in antibiotic reduction and in appropriate antibiotic use. And when we look at tests or medications in an order set, we can increase their use just by including them in that order set, regardless of whether they are clinically appropriate.
Skip to 2 minutes and 3 seconds So choice architecture is what this is all about; it’s how do we format and design the choices for people, and how can we nudge them towards making a better decision. And some of you will be familiar with Thaler’s work in this area on nudging, but also there’s a call for how we can do digital nudging, and how we can use this evidence to improve decision support.
Skip to 2 minutes and 34 seconds So in conclusion, if we are really going to obtain the benefits of AI in supporting clinicians to deliver better care for patients, we need to understand the decision making context. Who’s making the decisions, when are they being made? When do they need support? We need to focus on what are the big safety and quality issues; we shouldn’t just because we can design some decision support, doesn’t mean we should. We need to say where are our biggest problems, and where can we provide decision support for those problems. We need to look at choice architecture, and how can we nudge people to make the right decisions and better decisions.
Skip to 3 minutes and 16 seconds And finally what we have to do is always evaluate and monitor whether we are actually gaining the outcomes that we hope we, um, we would like, and we need to use very robust measurement techniques to do that, we can’t just trust it’s going to produce the outcome that we want. Thank you very much.
Conclusion on Electronic Decision Support
Dr. Westbrook is making a conclusion on electronic decision support. If we are hoping to obtain the benefits of AI in supporting clinicians to deliver better care for patients, we need to understand the decision making context.
Who’s making the decisions? When are they being made? When do they need support? The researchers need to focus on what are the big safety and quality issues. It does not depend on the capability to design some decision support. Instead, clear analysis of the problem is quite more important. The researchers need to define the biggest problems, and where can they provide decision support for those problems.
We also need to look at choice architecture, and how can we nudge people to make the right decisions and better decisions? These are the challenges that we currently have. What researcher can do now is to evaluate and monitor the right outcomes that intended to gain. More measurement and techniques are required to implement so that electronic decision support can be improved in the future.