← Becker’s Healthcare Podcast
Becker’s Healthcare Podcast
Scaling AI in Healthcare: Insights from Johns Hopkins’ Dr. Alvin Liu
00:09:35
Loading summary
Transcript13 lines
- [00:01]
A
Philips is a health tech leader focused on innovation that improves the health and well being of people. Our healthcare technology and informatics solutions help care teams diagnose, treat and manage more patients with greater precision, speed and confidence. Across the care journey with Philips, clinicians are empowered with streamlined insights in the moments that matter for every patient. Better care for more people. Philips.
- [00:29]
B
This is Grace Lynn Keller with the Becker's Healthcare Podcast and we are recording live at the 2025 Digital Health, Health IT and RCM Conference. I'm currently joined by Dr. Alvin Liu, who is an endowed professor and on the AI Oversight team at Johns Hopkins Medicine. So Alvin, thanks so much for joining me today. We'd love to have you start off by introducing yourself a little bit further and telling us more about your work in healthcare.
- [00:53]
C
Sure. Thank you so much for having me. My name is Alvin Liu. Thank you. I'm a practicing retinal surgeon at Johns Hopkins Medicine. In addition to my clinical care, I'm involved in artificial intelligence in three different ways. First, I'm the director of the Gils AI Innovation center, which is a research center made available and possible by a very generous donation of $10 million from Dr. Gils. Second, I am very involved in the implementation of healthcare AI tools for both clinical and operational purposes across the health system. And third, I'm part of the AI Oversight team which is a leadership team with purview over all things AI related across the health system.
- [01:37]
B
Wonderful. Well, thank you for being here today and let's start our conversation with a topic you know very well, which is AI. So nearly half of medical practices reported using AI in some capacity last year and it remains a key topic for health IT leaders. So from your perspective, what are the use cases that are making a difference right now and how are you leveraging those in your organization?
- [01:59]
C
I will give you two examples, one in the clinical domain and one in the operational domain. The clinical example I'd like to give you pertains to autonomous AI screening for diabetic retinopathy. Diabetes, as you know, is a very common condition. It can affect different parts of the body, including the retina which is the back of the eye. In fact, diabetic retinopathy is the leading cause of blindness in working age population around the world. To maximize the chance of patients diabetes of seeing well in the long run, we really should be screening these patients once a year to look for early signs of diabetic retinopathy. However, despite us knowing the benefits of this regular Screening, only about 50% of Americans undergo the screening every year. This is where autonomous AI comes in, meaning now we can do the screening in the primary care setting and it will be quite convenient for the patients because it will save the patients a trip to the ophthalmologist. This particular technology was cleared by the FDA back in 2018. In fact, when it was cleared by the FDA in 2018, it was the first autonomous AI system in any medical field that was approved by the fda. So we've been deploying this system at Johns Hopkins Medicine since 2020. And we have been able to look back at the data and we saw some very interesting patterns. We noticed that after the deployment of this technology at scale, we really improved the compliance rate of our patients with diabetes significantly in terms of them being screened for diabetic retinopathy every year. Perhaps even more encouragingly, we also notice improved access for patients who have historically been disadvantaged, such as patients covered by Medicaid or African Americans. The second example I'd like to give falls into the operational domain and specifically it is the use of generative AI for the purposes of revenue cycle management, which is very relevant to our conference today. In general, RCM is a very critical part of how healthcare is delivered in the US and RCM accounts for about $200 billion of expenditure each year. We have been deploying Gen AI in the RCI management space in the past couple of years, especially for obtaining power authorization. And I think across industry we've seen GEN being deployed across the different stages of rcm, including a front end which typically involves prior authorization, mid cycle, which involves coding, and cdi which is clinical documentation integrity, on the back end, which is denials management.
- [05:00]
B
And as virtual care expands from AI enabled tools and remote monitoring to broader digital health platforms, introducing new technology does bring challenges. So what advice do you have for leaders navigating everything from governance to patient engagement? And can you share an example of how your organization has balanced innovation with operational constraints?
- [05:21]
C
On one hand, it's certainly very exciting that we're seeing a lot of innovations when it comes to AI. On the other hand, it is really quite a chaotic environment where from a health system perspective, we see a lot of pitches when it comes to AI products and AI vendors. So I think in order to navigate this chaotic environment, it's critical to have a robust AI governance structure. What this means is that a centralized leadership team or a person will be in charge of governing the interaction between health systems and AI vendors. And at Johns Hopkins Medicine we established the AI oversight team for this particular purpose about a year ago. We perform different functions, but Here are some high level summary with we have established a standardized protocol that governs the interaction between Johns Hopkins Medicine and each AI vendor. For example, if an AI vendor is interested in selling a service to Johns Hopkins, the vendor will have to find an internal champion from Johns Hopkins. First, after undergoing the routine intake process which involves evaluation of cybersecurity protocols or requirement for IT integration, then the vendor advances to the next step which involves answering a list of standardized questions which include questions such as the kind of problem they're trying to solve, how they think about return on investment, the data cards, the model cards, and some of the safety guardrails. And then the next stage is for this particular application to be evaluated internally by Johns Hopkins by domain experts. So I think it's very critical to have a similar standardized AI governance structure that will provide some guardrails in terms of deploying AI responsibly.
- [07:22]
B
And how are you seeing recent legislation, both state and federal, affect healthcare organizations and healthcare IT specifically. And have you adjusted strategies in response?
- [07:33]
C
I would say is an evolving landscape. In fact, earlier today, in addition to this particular backers meeting, I was at a meeting that just five minutes away that's hosted by the American Bar Association. So at that conference we talked a lot about the uncertain legal landscape when it comes to healthcare AI regulations. So I think it's an open question. It's a shift in landscape. No one really has the answers to all the questions. However, if ultimately we're guided by some immutable North Star such as putting patients interest in safety first, we will be okay and we will figure out how to navigate this shifting legal landscape down the road.
- [08:16]
B
And as we wrap our conversation up, I'd love to know your top piece of advice for healthcare leaders as they prepare for further advancements in technology and rising demands for care.
- [08:26]
C
I'm a big fan of having a top down approach in terms of setting strategic decisions on the highest level. And AI, just like many aspects of healthcare, is highly complicated. So in order to establish a top down structure and establish strategies that make sense, my advice is to assemble a team that involve talents and expertise from different domains. Specifically, you certainly want someone with clinical expertise, understands the delivery of healthcare, you want someone with operational experience, you want someone who understands finances. And last but not least, you certainly want to have AI experts in the room so that those people can identify problems that can and should be solved by AI.
- [09:21]
B
Wonderful. Well Dr. Liu, thanks so much for joining me today on the Becker's Healthcare Podcast. Again we are recording live at the 2025 Health IT Digital Health and RCM Conference.
- [09:31]
C
Thank you so much for having me.