Artificial intelligence (AI) and its impact on healthcare is a pressing topic for professionals across the health information (HI) spectrum, raising countless questions about benefits, risks, and applications.
One leader at the forefront of the discussion is Maia Hightower, MD, MPH, MBA, former chief digital transformation officer at the University of Chicago Medicine, and the CEO and founder of Equality AI, a healthcare tech startup that focuses on responsible AI tools and strategies to develop algorithms that are fair and free from bias. Hightower will be speaking about the impact of AI on healthcare at the AHIMA23 Virtual Conference on Oct. 26.
AHIMA spoke with Hightower about how AI will change the healthcare landscape and what HI professionals should know about the evolving technology.
Q. How do you think AI will impact healthcare?
A. Overall, AI has and will continue to provide incredible value for healthcare. One way that I describe value is using the healthcare value equation, which is quality over cost times experience. That's a really easy formula to [use to] think about, ‘OK, where is the value that AI is providing for healthcare?’ Right now, it's a combination of clinical applications as well as administrative.
In the administrative realm, there's just such a huge administrative burden when it comes to the delivery of healthcare. There's incredible opportunity for alleviating some of that administrative burden whether [through] automation or leveraging AI technologies for addressing that administrative burden. For physicians, it's often documentation.
There's a slew of technologies that have already started to assist physicians in documentation, whether it's writing notes or communicating with the patients. Epic and its partnership with Microsoft and OpenAI has already started to deploy ways to simplify communication between providers and patients, decreasing some of that burden [and] increasing productivity. And when it comes to billing, there's a lot of opportunity for improving the billing or the revenue cycle component of healthcare.
Where it gets a little tricky is definitely in the clinical space. It's a far more complex process than with administrative tasks. And so, the developers of AI models for clinical care have been slower to adoption. But at the same time, I think they've done an OK job to date on taking that time to really ensure that the models are working as intended. In some very specific use cases, AI is showing promise. For example, sepsis has been identified for a very long time as an opportunity for AI, and there are some pretty nice sepsis models out there that are well into their deployment and have been integrated within workflow.
Q. What are some primary challenges associated with AI in healthcare?
A. It's always helpful to think of a problem in terms of a framework, and in terms of AI, within the AI lifecycle. There are pain points along that whole lifecycle journey, from problem formulation to development to deployment to retiring a model. In the problem formulation, the question is: Are we asking the right questions? Those who have the power to ask questions of our data sets generally are motivated by profit versus providing quality of care or improved experience. So, making sure that we're asking the right questions of our data sets and developing the right models that are going to solve our tougher problems. Within the development process itself, the second [challenge] is data in the electronic health record. Healthcare data, claims data is really messy. It's real-world data, and real-world data has real-world bias. Because of that, there are techniques that can be deployed to mitigate some of that bias. But unless they are uniformly adopted, we are at risk of perpetuating bias through the development of models because it's just the nature of the data sets.
And then, even if you have the perfect model, how is it deployed within the healthcare system? It should be deployed within the workflow and in a place that clinicians, care teams, administrators, and healthcare workers can actually use that information or use that prediction in a way that's of value. That takes time. The healthcare delivery system is not cookie cutter. What may work for deployment at one health system, may not work in another health system because workflows are different, staffing is different. There's just a whole slew of variables. Each health system has to evaluate when is the appropriate time to use a model, where within workflow, [and] make sure that it has that buy-in by the clinical care team or the administrative staff that's using the model.
Q. How is AI affecting health information professionals? What should they know about the growing use of the technology?
A. Depending on whether you're on the data side, the application side, or the security side, there are a lot of very role-specific areas that HI professionals should be aware of and be the master of. There's an upskilling component. Let's say you're on the data side. How do we ensure that the data is accessible and that there is tooling available to create data products and AI data products? If you're on the application side, generally, you're going to have your clinical applications and your administrative applications. Partnership is so important that you're developing relationships with your subject matter experts, both from the workflow perspective and the care delivery or the administrative part of healthcare, but also the experts on your team that are part of healthcare IT.
Many health systems have data science teams that can help with the technology component. [HI professionals] are generally in this middle spot because they often have the closest relationships with the care delivery or the administrative side of healthcare. Many of them have access to the experts with deeper technical expertise as well. When it comes to deployment of AI models, it's really important in understanding what makes an effective team. It’s fusion teams. It's a subject matter expert. It's the informaticist who understands both workflow and can speak a little bit of lingua franca when it comes to technology. It's the application developers.
Then it gets even deeper to the people who understand the platform where it’s being deployed. Maybe it's Epic, maybe it's Cerner, or maybe it’s one of your administrative platforms. Often, the [HI professional] has very specific knowledge about a platform. Then there's your data science team, your data architects, your engineers, your cloud engineers.
Having that fusion team approach is going to be so important in making sure that each team member understands their own role and really owns it, but also is working well with other team members to create a product or experience that is applicable to their healthcare system.
Q. Some have raised concerns that the use of AI will eliminate jobs for HI professionals. What do you think?
A. Right now, I think if there's a worry, it's about keeping up with the changes. So less so about having one's job replaced, but more about upskilling or making sure that one is keeping up with the advances in AI technology.
Let's say you're an expert in a particular application. Let's say it's Epic. Well, there is a whole slew of advances that Epic is deploying that are AI-related. It’s making sure that you understand the change] and you're that steward for your healthcare system of what advances are occurring within your platform, being that evangelist. For example, let’s say your skills are based on 2020, and the application has moved on to 2023 and now has all these advanced features. If you haven't kept up, not only does that not provide value for your healthcare system, but if you were to change roles in another healthcare system, it really decreases the value that you bring. At this point in time, AI really hasn't been decreasing the demand for health information professionals, but it has been increasing the demand for the skill set.
Q. How can HI professionals best prepare for AI and ensure they’re meeting the challenges?
A. Be curious. Take advantage of the opportunities that your health system provides for ongoing skill development. Typically, there are career ladders, and sometimes you don't have to be so vertical. Say, I start off as an analyst for an application, but now I'm curious about data science or data. How can I work with the data team and understand what they are doing to learn more about the technical aspects of data architecture and cloud computing? It’s about being curious about what's next to you or what the team member next to you is doing. That will always keep the job interesting and fresh.
Q. What else is important for HI professionals to keep in mind about AI?
A. Well, one is that we are all accountable. We all have a role to play. And when it comes to the deployment of AI, ensuring that it is ethical, fair, and really providing the healthcare value that it [should]. Is it really serving the mission, vision, and values of your healthcare system? Make sure that you're that advocate or that champion of technology being deployed in a people-process way that ensures that there is that strategic and mission-driven benefit to your healthcare system and, ultimately, to our patients. We are all here for one reason, and that true [purpose] really is our patient population, that we're improving the lives and the communities that we serve.
Alicia Gallegos is a freelance healthcare writer based in the Midwest.