“Let’s build something new and interesting. Really, the focus, especially in health care, needs to be with understanding deeply what the actual problems are that physicians, other clinicians, and operational leaders at health systems face today and build around those leveraging these new technologies, not trying to fit the technology to a problem that isn’t well-defined. We think it’s important to leverage human clinical intelligence throughout any work that anybody does with AI with large language models.”
Excerpt from transcript:
It’s important to see what works and make changes where things aren’t working as well as they can. All of that is sort of consistent with what we’ve been doing as we work with health systems for a few years. So this is again, a new technology that we can deploy to solve existing problems. And the way we think about doing some of this work is really in these two categories.
First is improving risk stratification and triage, making that much more intelligent than what we typically see, making it proactive so you can identify patients who are likely to need particular interventions, likely to have a particular adverse outcome in the absence of an intervention, enabling workflows based on predictive analytics, rules from clinicians, assessments that we see in clinical notes to sort of match patients to the right next clinical step and therefore improve their clinical outcomes and economic outcomes.
And secondly, and I think LLMs do have a lot to offer here, we think a lot about how we can redefine the health insights that we make available to clinicians. And the goal here is to do that in a way that lowers the cognitive burden that we’re placing onclinicians. And I think that burden has increased over years. And I think it’s rarely gone in the opposite direction. And we ask physicians to review more data. We give them less time to do it. The complexity of the patients they’re seeing is going up. And that really makes for a burnt-out population of physicians who have less time to do the kinds of work that they want to do.
And so, as we’re thinking about deploying LLMs and everything I’ll talk about today, we’re really focused on that question. And Scott might weigh in here as well of how we can sort of reduce that burden that physicians are under. Yeah, I totally agree with you there, David. And the reality is that a lot of the burden of information management data collection, even in the EMR, has been pushed down to the front lines increasingly.
That’s where the interface is supposed to happen with information, but there’s very little curation. And there’s kind of a fundamental mismatch between expectations of the EMR being the record of choice
When you think about the demand curve that we’re looking at for the next 10 years in healthcare, we’re talking more than 50% increase in demand. We’re not adding a lot more people to the workforce. We’re losing people.
So the only way to confront that kind of demand curve is to use tools like this to make us, to augment our smarts and to reduce the time that it takes us to be effective decision makers. This is not an academic question. And I think in a lot of industries, as people look at AI, I think people have understandable concerns about whether it puts people out of work. We don’t have enough people to do the work today. So the hope is we can maybe meet that demand that Scott is talking about with the staff that we have here or maybe a shrinking one. So I think, again, the stakes are incredibly high here.
SHARE
