Froedtert CTO talks about AI scaling strategy, from ‘small improvements’ to ‘hardest’ last mile
Editor’s Note: This is part two of our two-part interview with Dr. Melek Somai. To read part one, click here.
Froedtert & Medical College of Wisconsin Health Network is an academic healthcare system located in eastern Wisconsin. It embarked on a journey to drive disruptive innovations by establishing Inception Health as an independent vehicle to drive innovation and digital transformation, with a focus on digital health technology.
Dr. Melek Somai is vice president and chief technology and product officer at Inception Health and Froedtert & Medical College of Wisconsin Health Network, and assistant professor of medicine at the Medical College of Wisconsin.
Yesterday he discussed artificial intelligence in healthcare in general, focusing mainly on generative AI (the kind found in the popular ChatGPT). Today he talks about the AI work being done at Inception Health and Froedtert & Medical College of Wisconsin Health Network.
V. Inception Health and Froedtert are implementing AI to streamline certain processes, such as scheduling, so that patients can identify their best options based on their individual preferences, with a simple AI interface, removing the complexity of the traditional decision tree. Describe how this system works and what results you expect.
A. We see that AI has three phases. I’m simplifying it here. There is a phase of small improvements. What I mean by this is that while we have great hope that the value of generative AI to transform industries is great, the path will not be linear. It will require greater capabilities, greater resilience and, most importantly, greater preparedness from healthcare systems to deliver demonstrable and verifiable value for AI.
And that last kilometer will not be an easy kilometer, but one of the most difficult. So in the meantime, we see that AI can have some immediate benefits that we can safely implement. Today, AI has great potential in a number of areas.
You mentioned navigating our healthcare system and helping patients with scheduling. We know that navigating a healthcare system today isn’t easy. And as we continue to improve the experience of patients navigating our healthcare systems, AI, and especially its combination, will continue to improve large language models with API integrations can help us provide a more intuitive and personalized interface for patients.
This has less to do with providing the information; it’s more of an interface for patients to interact with our services that we already have. AI implementation at that layer can help patients check in for their upcoming appointments; For example, by helping them identify the best option for rescheduling an appointment, taking their preferences into account, or by actually mining and synthesizing the enormous value of data.
For example, we need to provide a summary of specific questions for the patient; about, for example, the laboratory results or their preventive care visit when necessary. So at this level, I want to be clear here: AI does not contribute to patient care, but is instead used as a more intuitive user interface for our back-end services that we already have.
This type of innovation has the ability to streamline certain processes and help patients better navigate the healthcare system. However, AI at this level is not transformative, and honestly, we can see these use cases as incompatible with the cost of building these AI models in the first place. However, integrating AI at this level can help create use cases and build even more capabilities as we evolve in a safe way.
For example, if we look at this from the beginning, we ensure that the building of these models happens across the board, using really appropriate frameworks that ensure governance. data safety and data security. And again, those models are built to help with navigation, so more of an interface layer. As I said, there’s another level of integration, and this is probably the level, what I call level two, that helps scale healthcare recommendations. This is still experimental today.
V. Inception Health is reviewing AI systems that provide personalized preventive care recommendations, integrating patient medical records, US Preventive Services Taskforce guidelines, and rules-based recommendations. What do you want to do with AI here, and what do you hope the outcome will be?
A. This is another level I mentioned about integrating AI and how it can help and scale healthcare recommendations. This is still experimental, but it brings AI much closer to the realm of clinical care practice.
This is the area where the implementation of a The AI governance framework will be critical, most likely nationally. At this level, you can think of AI as a co-pilot for a patient, helping them understand their options and care recommendations. This work covers a number of aspects.
The work we’re currently doing includes making advances in retrieval, enhanced generation, integration, embedding, and rapid engineering to help us build personalized and adaptable copilots for our patients. There are a few fundamental points that we’re currently modeling on how we build these types of experimental copilots.
One of the most important elements here is that the patient data is stateless at this stage. What I mean by this is that the data is not used by the model to train itself and is never stored. It is transient and provided to answer the question in real time. This model means that AI has no ability to learn from the data or even be aware that the data even exists.
Today, this type of integration ensures security, privacy and transparency. And this is a common approach that we’ll likely see in the near future, using it to develop more robust AI frameworks and infrastructure in the same way we do.
I’m really glad we did that the recent Executive Order (from President Biden) on the Safe, Secure, and Trustworthy Development and Use of AI, which is a prime example of such a commitment. But there is certainly one of many steps that private sector healthcare systems will need to develop over the coming months and decades to ensure that we can use this in a safe, secure and reliable way.
And this capability will be important for our patients. If you think about having the ability to synthesize data from the best recommendations, to be able to recommend but also guide the patient through that experience, under the supervision and guidance of the physician, this will still help that relationship streamline and improve patient capabilities. our healthcare systems.
V. Inception Health is developing a digital-first healthcare model that puts AI at the heart of scaling services and improving convenience. You are shifting the healthcare model where AI helps with the first mile of healthcare provision. Please explain this model and why you think it is the best approach to AI.
A. This is what I call level three, the next level where AI helps reshape us. For me personally, this is one of the most exciting areas that my team at Inception Health and at Froedtert are really working on. Not only impacting generative AI in the short term, but also in the medium to long term, with the opportunity to redesign the healthcare model to use generative AI and technology as a platform, rather than just a back office solution.
The approach we’re taking today is to think creatively and more fundamentally about the value of AI and generative AI technology in shaping the healthcare experience. And the way we’ve approached this as a team is we’ve built a user research group and developed a product mindset by thinking about where a generative AI technology can help us reimagine the model that we deliver.
So one of our active activities at Inception Health is leveraging our insights into healthcare technology with our capabilities as a healthcare system to think differently and ask the question: what if?
What if we could provide an experience that wasn’t based on a patient visit to the provider’s office? Take the example of the COVID era, where we have seen a boom in telemedicine. One of the reasons why telemedicine use has declined post-COVID is that we haven’t fundamentally changed the workflow and the patient and provider experience.
So changing the medium is not enough. And what we’re working on is what we can bring with the power of AI to completely reimagine the healthcare delivery experience. And one of the approaches we’re thinking about is, what if we get results? a digital-first experience where AI is not just a co-pilot, but also the driver to help patients communicate more clearly and actually more easily with their healthcare provider and the healthcare system?
This is an area that is really exciting because it is going to help us create value that will not add value to the current healthcare system, but will be a catalyst for value. Some values around lowering healthcare costs, improving the convenience of our healthcare systems, and improving the experience for our patients will be deeply impacted by this approach.
This is an area that is very important and where I think we are leading the wave. We’re working with industry leaders in this area, and this is probably going to be one of the most fundamental approaches that we’re going to learn a lot about: how to bring those experiences to our patients.
This is an area where, in my opinion, technology is combining with clinical care to solve the problem. This will be critical for us to be successful in realizing the promise of technology and generative AI in healthcare in the coming decades.
Click here to watch a video of this interview that includes bonus content not included in this story.
Editor’s note: This is the sixth in a series of articles from top voices in healthcare IT discussing the use of artificial intelligence in healthcare. Click here to read the first article about Dr. John Halamka of the Mayo Clinic. To complete the second interview with Dr. To read Aalpen Patel at Geisinger, click here. To read the third, with Helen Waters of Meditech, click here. To read the fourth, with Epic’s Sumit Rana, click here. And to read the fifth, with General Brigham’s Dr. Rebecca G. Mishuris, click here.
Follow Bill’s HIT coverage on LinkedIn: Bill Siwicki
Email him: bsiwicki@himss.org
Healthcare IT News is a HIMSS Media publication.