← Becker’s Healthcare Podcast
Becker’s Healthcare Podcast
Merage Ghane, Ph.D., Director of Responsible AI in Health at the Coalition for Health AI (CHAI)
00:21:08
Loading summary
Transcript14 lines
- [00:00]
A
Welcome to the Beckers Podcast. I'm Naomi Diaz, health IT reporter here at Beckers. And today we're joined by Dr. Mirage Gain, director of responsible AI in health at the Coalition for Health AI, also known as CHAI. In this episode, we'll discuss what hospital leaders need to know about AI today, how to prepare for the next few years, and where AI is really headed in healthcare. Dr. Gain, thank you so much for joining us today. I appreciate it. I want to start off with our first question here, just a simple one of just introducing yourself and tell us a little bit about your background.
- [00:37]
B
Great to be here. Thanks so much, Naomi. I am Mirage Gain. I have a PhD in clinical psychology actually, and I have a background in computational and clinical neuroscience, behavioral science, and I used a lot of machine learning and AI in my research back in the academic days, but I'm far from those, not super far, but pretty far from the academic days. Since finishing my postdoc, I've been working. I started working as a principal behavioral designer at Ideas42, which is a mission driven nonprofit. And I focus my work in AI and machine learning and health, sort of things around identifying bias and health AI solutions for organizations and helping organizations do that in a behaviorally informed way. I actually started working with CHAI then. So before CHAI was ever a nonprofit, it was a coalition for the Willing, coalition of the willing. And I led their fairness and bias work group in developing sort of responsible AI guidance. And then I transitioned to this role. So here at chai, I work across many sectors. I bring together any sector you can think of in health care. So anybody on the developer side, the clinical side, sometimes policymakers, who else do we have? Patient advocacy groups, community health centers, all of them into one space. And we talk about how to build shared frameworks, tools, processes, best practice guidance to make AI safer, more transparent and responsible in healthcare.
- [02:36]
A
Amazing. Great to hear about your background. And again, just responsible safety of AI implementations is just really what we need right now. Would love if you could just dive in and just tell us about your work with consensus driven frameworks for AI and healthcare. What have you learned over these past years and what do you feel like is most important today?
- [02:55]
B
Great question. Consensus work is hard and traditionally very slow, but we don't actually have that privilege really in the AI space of being slow because things are moving so, so fast. And so we at CHAI use sort of modified Delphi methods to build these consensus driven frameworks. And we convene, like I said, many stakeholders across large academic medical centers like the Mayo's like the Dukes, UT health as well as non academic medical centers, small community health centers and like I said, the developers of these solutions as well, to kind of think about how can we take what currently exists in many different silos with many different incentive structures and bring them together to find the shared motivations. How can we help each other improve the health care system without sacrificing innovation, without sacrificing patient safety and privacy and keep things moving at a pace that this whole field really needs? I think one of the biggest things that I've learned is that everyone is trying to do the right thing with AI. We, most people, we have, I think, different lenses depending on where we're coming at this problem from. And there is no map that says here is how all these different lenses should connect. And that's where we come in to sort of co create practical, practical sort of flexible guardrails, tools, resources to help everyone in this landscape make better, safer decisions. And I think it's more important than ever now because the consequences of AI misuse in healthcare in particular is huge. And we have a healthcare system that is already struggling both as a business and for patients. And so how can we really take this opportunity, and I see this as a big opportunity to do better without hurting the most vulnerable people.
- [05:23]
A
I love what you touched on there too is not stifling that innovation, but also ensuring that you have those safety guardrails for a technology that's continuously evolving. I think that's super important there, I think for our audience obviously and our listeners of healthcare execs. And I would love to ask you as a follow up here as well, what do you feel like hospital executive teams really need to know about healthcare AI today in this such rapidly changing landscape? And what do they need to do to prepare their teams for the next two to three years?
- [05:55]
B
Great question. And we're doing a lot of work in this space. But I think the first thing to realize is AI is already being used and you know, sometimes it's chatgpt on somebody's phone and it's also being used for real use cases and documentation in triage and diagno diagnostics. Radiology is a big area and administrative tasks. So people think about AI and health care like oh, it must be a clinical use. But there are so many operational and administrative processes in health care that are struggling and leading to workforce issues and costs of having to retrain people because people are so burned out. So just recognizing that it is already being used and there are lots of spaces that you can use it within the healthcare landscape. I think the second thing is that folks need a plan, and that means establishing internal policies, governance structures, processes, working together with your teams and within your network. So some organizations are much larger and have very robust internal teams. Some organizations like safety net hospitals, FQHCs, community health centers, they don't have that luxury, so to speak, but they do have access to things like their health center control network, their primary care association network. And so how can we get really creative about what it means to establish internal policies, governance structures, and not, you know, burden those that don't have resources and be able to develop that? So we're doing that in partnership with nac, the national association for Community Health Centers, but also with just our member organizations that have come up and said we really care about this. Even the larger organizations will come up and say, if we're proposing this as a process, how will the smaller entities, the less resource entities, actually make this happen? So we're really thinking about this carefully. So as part of that plan, folks need to understand the risks. They need to determine what are some of the responsible use cases, where can they start in this AI journey for their particular organization? And again, context matters a lot. And then thinking about the processes that you're developing, the governance structures, the internal policies, should address things like bias or risk of bias, security and privacy issues, the usability and usefulness of that solution. You're paying so much money not just to procure this, but to use it. And if it's not helpful, if it's just going to make your staff angry, like shouldn't, shouldn't be the, you know, even if it's bright and shiny and new, there's a lot of change management that needs to go in to bringing new processes in and making sure that they're helpful, and then safety and transparency. So those are, I think, areas that folks really need to start developing plans around. And then finally, I think investing in capacity and in education for staff. So upskilling existing staff, training, new staff to sort of have that come in with the mindset of, like, where AI is used, when it's helpful, how it's helpful, and what is it for many folks, that's the first question. So investing in that capacity now and investing in a roadmap for the future, data infrastructures, assets, all of that, making sure that you have a sense of, like, where your data's at, how it could be used, how it could be leveraged to actually improve the business of your organization in this sort of, in these Times where data, data is king and queen. All of the above.
- [10:08]
A
I, I really loved your takeaways. Dr. Gain. Really hit the nail on the head when it comes to just things. I'm even hearing from our audience here on just the, the obstacles they're going through in their own organizations there. Really appreciate you diving deep. I want to stay on that track of what you mentioned of, you know, those governance, those guards rails and ask you, you know, how should hospitals and really health systems just address regular regulatory concerns with AI and health care? And I know you touched on this a bit in your answer as well, but how do you really create ethical standards within the organization to deploy this safety and ethically as well?
- [10:45]
B
Yeah, I think, I think we're in this sort of regulatory purgatory, I would call it. And, and I don't think, I think at the end of the day we shouldn't have to wait, nobody should have to wait for full regulation to act responsibly. If we're going to only do the right thing when somebody tells us to do the right thing, I think, you know, maybe it loses its value a little. So for me, ethical standards has to be, they have to be built into the AI life cycle. So all the way from procurement to development, if that's something you're doing internally in your organization or co development with, with a vendor or purchasing, all the way to monitoring and oversight. So when we think about ethics in healthcare, how can we then extend that to ethics and AI? And I think there is this sense that sometimes people try to reinvent the wheel. They're like, oh, AI is so different. Everything about it needs to be different. But safety still matters, security, privacy still matters. And these things existed, you know, transparency, consent, like these things have existed in health care even pre AI. Now the question is how do we identify where AI uniquely needs ethics in these spaces and how can we just add that to the processes that already exist without fully reinventing the wheel? And I think at chai, we try to align our work with some of the regulatory momentum that's out there. You know, FDA has action plans out there, ONC has their HTI1 rule. But I think what we hear from our community, and that's we are a community led nonprofit, is they want practical playbooks, they want practical tools, resources to make these things possible. So one of the things we're doing in partnership with the joint Commission is we're developing a set of governance playbooks for organizations of all, all sizes. And hopefully down the line, I mean in the near future, Accompanying tools, templates, resources that folks can leverage. And as part of that playbook, I think one of the biggest things that we want to prioritize, not only getting feedback from our community, which we are going to do, we have a series of workshops coming up, but also illustrating how differently sized organizations are aligning with some of these ethical guidelines that we're putting out there or some of these governance controls that we're putting out there in the playbook. So if we say everybody should have AI policy, what does that look like? How are different organizations doing this? So that folks have kind of a roadmap or template to look to say, oh, that worked for this size organization, which is similar to me, so maybe I could start there. And also being very clear about what the challenges are that people are facing in aligning with some of these ethical guidance guidelines and you know, recommended governance controls and being able to say like, okay, are there tools, are there templates, resources that we can provide to help reduce some of that barrier and making sure that the ones who are resourced, who have it, and I can develop those things, great. For the ones who can't, how can we help them get a step up in that process and not widen that digital divide? And I think ethical AI isn't, shouldn't just be within the healthcare system. I think a big part of what is now happening if we really want to create person centered or human centered solutions, it is a lot about the developers also understanding what ethics means in healthcare and being able to align their products with that so that when they go to their customers, the healthcare systems and they say I have this really cool thing that they're already speaking the same language. And I think that's a lot of what we're trying to do is get folks to hear each other's different languages and at least understand like do a little bit of translation there. And, and it's about the relationships. So it's not, it's also about how are we supporting each other, how are developers supporting their customers in aligning and how are the health systems supporting the developers and improving the tools to better fit their needs. And so I think that's another big piece of ethical AI is that relationship between the developer and the implementer.
- [15:39]
A
Wow, you just mentioned so many important topics here. I think one of just an incredible quote you just said, you know, said in the beginning as well, just nobody should wait for regulations to do the right thing. I think in essence that was just amazingly said a Dr. Gaines. So appreciate that. I want to, I want to look as well to the future a little bit here for, for our last question and ask you just where do you really see AI headed in healthcare over the next two years and what do you expect will be a little bit different because of AI?
- [16:13]
B
Yeah, so there's a lot happening in the space. I think we'll see definitely some expanded AI use in areas like ambient documentation, prior authorization, both on the provider and the payer side and patient risk stratification. I think that's going to continue sort of predictive prescriptive analytics types pieces aren't, it's not going to go away. Even though Gen has, has come up. I think Janaya will help improve some of those processes. I think we have started focusing on a lot of back office tools as a first pass for AI in healthcare. Like that's been the cautious approach. I think this will expand to frontline clinical decision support and I think it will also move towards patients taking agency over their health care differently and interacting differently with their providers. Right. So with tools like ChatGPT, people are inevitably going to ask their health questions, bring a different kind of knowledge to their appointments. And so I think that will really shift the landscape quite a bit. I think there will be greater demand for transparency and evaluation over time. And I think we at CHAI are developing sort of certification programs for assurance resource providers which might be folks that are validating like third party validation of AI solutions, individuals that enable sort of local validation or training in a secure, privacy enhanced way. Really making sure that people can actually access these tools in a way that is, is safe. And then I think we'll see a lot more involvement from nursing staff, from community health centers, from under resourced providers. But we need to equip them with the tools and the training to really participate in this space. So one of the things that we are doing, I think there is quite a bit of curricula out there for physicians and we are working on expanding that into specialty society type work. But we also are partnering with Florida State University to bring a responsible AI education to nurses. And I think that will be pretty important. We have cross functional teams and if only some parts of the team understand what's happening and other parts don't, it can continue to increase the tensions, the silos in a space where we want actually better care continuity and better collaboration. So I think that's, that's some of it and then agentic things hopefully maybe will come out in healthcare. Maybe not in the next two years, but definitely down the line especially for some of that back office stuff. I think for the administrative and operational work it makes a little bit more sense, but there are no solutions that I know of currently that are being deployed in the healthcare system there. But it will come up down the line for sure.
- [19:37]
A
Lots to look forward to, lots evolving so quickly. I think Dr. Gain, you just shed such great light for our audience on what this means for them and just the great work Chai is doing as well so it doesn't become a place of AI have and have nots as I like to describe it, where you know, smaller systems versus larger systems can't get that access to really deploying this technology. Like you said, ethically and safety. So I want to just give you a huge thank you for shedding light on there, sharing your insights on this and as well as to our listeners for taking the time out listening to the Becker's podcast today. That's it for our episode today. Dr. Gain, any last words or anything you would like to share with our audience before we end today's episode?
- [20:20]
B
No, I think join our newsletter list. Keep up to date even if you're not a member. If you want to be a member, feel free to reach out to us. We would love chat and see how we can be helpful and yeah, I'm just great. I'm grateful for the community. Without the community like the the things we put out, the understanding we have of the field just like really would not be possible and so I'm grateful for that. We, we can only do this together.
- [20:50]
A
Actualizing that learning health system and learning community. Well, thank you again Dr. Gain and I appreciate your time today.
- [20:56]
B
Appreciate you. Thanks so much Naomi.