Artificial intelligence is drastically changing clinical and operational capabilities to support better patient outcomes and more efficient processes.
Given the pace of transformation, what will be possible in healthcare three to five years from now that isn’t today? Three physicians who are also IT leaders within their organizations shared their perspective at Becker’s 15th Annual Meeting in April.
Note: The responses below are lightly edited for clarity.
Chris Longhurst, MD. Chief Medical Officer, Chief Digital Officer and Executive director of the Jacobs Center for Health Innovation at UC San Diego Health: I’m going to give three areas where I’m excited that AI and agents are going to help us make progress we haven’t been able to make before. The first is operational. Our processes stink, and every patient knows that. I think about our back office processes, prior auth and revenue cycle and call centers, and there’s tremendous opportunities to automate workflows and help make our staff more productive. That’s going to be a huge area to both improve our experience for patients and lower costs per unit.
The second area that I’m really excited about is patient safety. It was 25 years ago that the Institute of Medicine released the landmark report “To Err is Human” which acknowledged that we killed 100,000 Americans per year through medical errors, and the sad truth is we haven’t improved. Dr. David Bates published data last year suggesting one in four patients discharged from a Boston hospital had a preventable medical error, and we’re seeing continued poor outcomes. This is the worst kept secret in the industry. Every health system has cases they settle that never go to court, that never should have happened because processes that haven’t been improved despite a lot of good efforts.
I’m really excited about things like computer vision that are going to help us provide that 24/7 safety net to ensure patients never get these preventable blood clots or bed sores or pressure ulcers. Patient safety is the second area we’re going to make a lot of progress.
The third area I want to call out is empowerment of our patients. We don’t talk about this a lot at healthcare conferences, but our patients are using these tools too. It turns out they don’t really care if their PHI is going into a cloud server, because they’re looking for a diagnosis for their child or treatment options for cancer. But what we’ve seen this doing already anecdotally through stories, is amazing. It’s the old maxim of, the doctor that uses AI will replace the doctor that doesn’t use AI. The same is true for patients, and it’s really going to be the care team plus the patient plus the AI that’s going to change the way healthcare is delivered.
Michael Pfeffer, MD. Chief Information and Digital Officer, Stanford Health Care: Healthcare is about relationships. How many of us dial our bank or airline or credit care company and press zero repeatedly until you get an agent? That’s how I function. Because a lot of the stuff you want to do is online through an app and you can complete a lot of the work there, but when you want to speak to someone, when you really have to call, you want to speak to someone and you don’t want to be on prompt after prompt after prompt. My hope over the next three to five years, and I think we’ll actually get there, is to eliminate all this stuff that healthcare workers are doing that are not about the relationships with the patients and open up space so that you don’t have to keep pressing zero.
There are times you want to speak to a person but we have people doing our non-high yield stuff. If you think about the revenue cycle, that’s a great example that could be fully automated in three to five years. And the list goes on and on.
Tools like computer vision to help clinicians with safety issues and monitoring patients in real time are going to be really helpful as well. I’m just thinking about all of these tools to help us get back to this relationship-based medicine I think is really, really powerful and we’re going to get there in three to five years.
I think medical knowledge is going to be more democratized than it is today. AI will be able to ingest multimodal data and make diagnoses for any physician. Traveling to different centers to get a particular expert is probably going to happen less and less because your community physician is now going to have access to incredible tools that are going to be able to help with the differential diagnosis.
The number one thing, when we look at lawsuits, is patient safety issues: a misdiagnosis or delayed diagnosis in patient care. We have to eliminate that as we think about the next three to five years in AI.
Gurmeet Sran, MD. Chief Clinical Data Science Officer of CommonSpirit Health (Chicago): To build on what everyone else has said, a lot of this needs to underpin the agentic AI workforce, often colloquial the middleware layer, and right now you have many disparate systems that don’t API, or don’t talk to each other. In theory, you can use an agent to create an ability to pass through the information and it’s still very clunky. What you’re seeing is a lot of hyperscalers and the open source community beginning to focus on what could be those agentic protocols. We speak on the internet using [Transmission Control Protocol / Internet Protocol], that’s where we know how data is transferred in bits and bytes across cellular technology. Now you’re starting to see that movement of people talking about protocolization of these middlewares so these agents can become more functional.
Google just announced at Google Next that they have a protocol called a Number Two, an agent to agent. Some of the other large, large language companies are building their own protocol. You’re going to see a convergence of that, and what it will do, that middleware layer, will underpin a lot of the core technology that we’re going to see.
I’m very optimistic that, as much as we’ve been focusing on the physician and scribe technology, that in two to three years you’re going to see that move into the nursing space. With that, you’ll get more real time information pushed into the chart. You’ll get the ability to then be able to learn on that and do more real time intervention so the harm event doesn’t occur that we know has immense consequences both for the patients as well as the financial burden of the hospitals.
I certainly think that we in the U.S. health system used to think about the brokenness and complexity of these edges and these nodes, whereas when you look at virtual first, we may have to look outside of the United States of America and look at what could potentially work outside in other countries that aren’t beholden to the same restrictions, both from a regulatory standpoint and from an economic standpoint, and see what plays and what opportunities we can learn from them. Then it would have to be pushed to the U.S. – there would have to be regulatory and financial change – but I see immense opportunity with how flat the world is now to learn from those other advanced countries.
