Transcript
Speaker: Greetings.
Speaker: Welcome to the Health Data Ethics Podcast.
Speaker: I'm your host, Jenny Owens.
Speaker: And in today's episode, we're going to talk a little bit about the American Medical Association's recently released report on the future of health, the emerging landscape of augmented intelligence and health care.
Speaker: I thought this was a really good report.
Speaker: Lots of really nice summaries of the current state of artificial intelligence or what AMA would like us to call augmented intelligence.
Speaker: And we'll dive into this a little bit.
Speaker: And some really good takeaways for how to properly engage your clinical stakeholders when you're doing pilots and implementations.
Speaker: So let's get started.
Speaker: So first of all, I thought this report was interesting because it opens with a survey of 1,000 clinicians, 1,000 physicians actually,
Speaker: And what AMA asked them was, are you more concerned than excited?
Speaker: Are you more excited than concerned?
Speaker: Or are you equal parts concerned and excited about the use of artificial intelligence across healthcare?
Speaker: And the pie chart, as you might imagine, looks almost like equally divided into thirds.
Speaker: A slight majority, 41% of our physicians said that they were
Speaker: equally concerned and excited about the use of artificial or augmented intelligence in healthcare.
Speaker: So this really sets the stage for a ecosystem that is curious and engaged and wants to know more about what this can do for them while still having very valid concerns about ethics, about privacy, and about what does implementation actually mean?
Speaker: You know, we don't want to dilute the care that we're able to provide for our patients.
Speaker: So this report, and I will link to it in the notes, has a nice summary of what some of the artificial intelligence platforms and the large models of artificial intelligence do, right?
Speaker: What is the difference between unsupervised learning versus supervised learning versus reinforcement learning?
Speaker: Talks about large language models versus deep learning and really makes a compelling case for the use of augmented intelligence
Speaker: as opposed to artificial intelligence, in particular in the healthcare space, because most of the AI tools that are on the market are not really looking to replace clinical judgment.
Speaker: These are really assistive tools.
Speaker: They're doing chart summaries.
Speaker: They're collecting risk and presenting it in a concise manner.
Speaker: So this is really not intended to replace any sort of human thinking.
Speaker: It is intended to be supportive and to enhance human capabilities and decision-making.
Speaker: They call out a few really interesting use cases.
Speaker: There's some neat stuff towards the end with specific use cases.
Speaker: And then they talk about non-clinical AI use cases, talk about ways to use artificial intelligence or augmented intelligence to improve access to care, ways to cut down on our administrative burden in administration and revenue cycle.
Speaker: We talk about ways to enhance hospital operations with predicting staffing volumes and staffing needs, looking at tracking inventory, monitoring equipment availability.
Speaker: I still think that one of the great places where we could really use some additional analytics and some augmented intelligence is in sterile processing.
Speaker: I think there's a lot of opportunity there, and I would be really, really excited to dig into this more.
Speaker: AMA also calls out specifically
Speaker: opportunities in research, right?
Speaker: Can we optimize research target outreach?
Speaker: Can we, target is the wrong word there.
Speaker: Can we optimize research participant outreach?
Speaker: Can we use this to simplify enrollment evaluation?
Speaker: Can we take a look at electronic medical records and simplify follow-up data provision?
Speaker: All in all, this is a really nice summary.
Speaker: They have some really helpful considerations about using augmented intelligence in your practice.
Speaker: And I just want to go through some of these almost verbatim because they are really good.
Speaker: They talk about four distinct phases in the use of any sort of augmented intelligence tool.
Speaker: So they talk about phase one, which is to identify the challenge and the use cases.
Speaker: Phase two is to evaluate the tools.
Speaker: So our whole series that we've been doing on how to evaluate an AI tool in your hospital setting, this would all fall into phase two.
Speaker: Phase three, which is where we're getting with some projects that I'm doing during my day job and I'm really excited, is to implement AI tools.
Speaker: And they call out specifically needs for training and making sure that training is instilling confidence.
Speaker: So in our previous episode, when we talked about that financial report that said AI moves at the speed of trust, training is a really important component of that trust.
Speaker: Do people feel confident using the tool?
Speaker: Do they feel confident in its output?
Speaker: Do they understand how to exercise their clinical judgment when they disagree with the tool?
Speaker: In this phase three implementation, we also talk about workflow, talk about technology support.
Speaker: So who do you call when it's not working?
Speaker: And what sort of infrastructure do you need for this tool to function at peak capacity?
Speaker: So like a sepsis prediction model is stellar.
Speaker: But if it takes 45 to 90 seconds to return a prediction on a patient and I'm sitting there.
Speaker: at my workstation on wheels, drumming my fingers, waiting for my score to appear, that's probably not a winner.
Speaker: In this phase, they also talk about bias.
Speaker: So how can we make sure that we're noticing bias when it comes up?
Speaker: How can we make sure that we're rectifying the bias if it arises?
Speaker: And then in phase four, we talk about managing AI tools.
Speaker: So we've moved past implementation.
Speaker: These tools are now a part of our everyday life.
Speaker: How do we make sure that we're maintaining it, that it's keeping up with medical practice, that we're monitoring and that we're administering access appropriately, that we're measuring our financial implications, right?
Speaker: If we budgeted for this tool with the understanding that it was going to save us so much time or so much money, are we actually doing that?
Speaker: And then underpinning all of this is great.
Speaker: It's the health equity.
Speaker: We want to make sure that in every single stage of our evaluation, so even when we're still looking at what are the challenges, what are the use cases, where might augmented intelligence or artificial intelligence be helpful, all the way through to this is now part of our ecosystem.
Speaker: How are we making sure that we're ensuring equity across all of these stages and all of the patients and all of the caregivers that this touches?
Speaker: I think this is great.
Speaker: They end with AMA augmented intelligence commitments to develop AI principles for the use of AI in healthcare.
Speaker: Very helpful.
Speaker: The AMA actually released new principles just in November, 2023, but even since then, the field has moved on.
Speaker: So I look forward to continuing to read the AMA's guidance on this point.
Speaker: They mentioned that they want to support the development of state and federal policies.
Speaker: that will ensure appropriate oversight and continued innovation in AI.
Speaker: I think it's nice that we balance both the oversight and the innovation.
Speaker: I think that's very thoughtfully stated.
Speaker: They end with a, I think what is the really highest priority and that's to prepare and inform physicians by providing high value insights and actionable resources.
Speaker: So the information that your tech folks
Speaker: that your IT crew, that your innovation-minded leaders are providing you.
Speaker: Is it helpful?
Speaker: Is the information clear?
Speaker: As an intended end user of this technology, is this relevant to me?
Speaker: Is it going to solve the problems that I see?
Speaker: And then just as quickly, you can flip this on its head and say, here is a real opportunity to involve your stakeholders early on.
Speaker: So a lot of the physicians that the AMA consulted for this said, we would like to be at the table.
Speaker: We would like to be consulted.
Speaker: In the previous episode, we had talked about the need for early stakeholder involvement.
Speaker: And this is a really clear sign that the appetite is there clinically.
Speaker: Our physicians want to be involved in conversations about using technology to make clinical practice better, to provide better, safer health care.
Speaker: And I think that we should consider this from the very earliest stages, even as we're starting to define our use cases and look at what technology might be able to help us.
