Geisinger Health President and CEO Dr. Terry Gilliland is drawing on his background in medicine and artificial intelligence to shore up the system's operations.
Nearly a year ago, Geisinger became part of Risant Health, a Washington, D.C.-based nonprofit organization created by Kaiser Permanente to acquire health systems and set up a national value-based care network.
Related: Risant Health emerges as Kaiser-Geisinger deal closes
Since then, Geisinger has implemented various tools and programs to smooth workflows and keep costs down. It has two major bed tower projects and is starting to collaborate with Cone Health, which Risant acquired in December.
"This is not different from getting married. In the first six months to 12, you're trying to figure out who's going to squeeze the toothpaste in the right way," Gilliland said.
It's been about a year since Gilliland took over at Geisinger, which serves 1.2 million patients across Pennsylvania with 10 hospital campuses and a 600,000-member health plan. He was the chief medical officer and chief science officer for healthcare AI company Cogitativo and also held leadership roles at Blue Cross and Blue Shield of California, Sentara Healthcare and Kaiser Permanente.
In an interview, Gilliland detailed Geisinger's financial challenges and how the system's rollout of a value-based platform and efforts to drive innovation with AI can make a difference in its operations.
The interview has been edited for length and clarity.
How was your financial performance last year?
We ran a -5.9% margin last year in our health plan. As a system, in aggregate, we booked a $141 million loss last year, just Geisinger. Half of that $141 million was from one-time items. Otherwise, our clinical enterprise did pretty well, but the health plan pulled it back. For a roughly $10 billion company, a $70 million loss is not huge and aggregates to about -1%, but this year we've got to turn that around. We are going to turn it around. We’re looking to get to a little over 2% operating cash flow margin.
How do you turn it around?
We have to get better at how we approach it from the plan side by getting people as healthy as they possibly can be, and then on the clinical enterprise side, getting the very most out of inefficiency. I'm optimistic that we have a bunch to do on the efficiency part. Are we getting people in and out in an efficient way? Are we using our operating rooms effectively? Are we using our surgeons effectively? That's my sliver of hope, that if you can get a whole lot better at how you're doing things, then you might be able to mitigate some of this offset that we could feel from the payer side.
How would the proposed cuts to Medicaid affect the health system and health plan?
On the health system side, you're going to get less for having people in the hospital. We don't make money on it now, so it'd be more of a loss. On the health plan side, we don't figure that the federal government's going to show up and say, “We have more money for you.”
The federal government and the states are interdigitated on Medicaid. My experience with the Commonwealth [of Pennsylvania] is that they've been kids in a candy store. You can put in waivers and get extra stuff from the federal government for certain programs. The benefits are very rich in Pennsylvania. What I'm worried about is whether the federal government comes through and decides two things: No. 1, less money; and No. 2, our 65-35 match is going to go down to 50-50, like it is in California, as an example. I’m really concerned about the ability of our administrative officials to then reduce their aperture on benefits.
What does Risant's value-based care platform look like?
I think of the value-based platform as a tool chest, and we're pulling stuff out of the tool chest and making sure that it works. At the top of the list is ambient documentation. The second thing inside the tool chest is value-based care guides. For example, say there are 12 common conditions in oncology. We're going to put this guide in a workflow with the primary care physician or the referring physician that says, “If you have lung cancer, here are the things that you will need to work up in advance of seeing the oncologist.” It'll automatically populate your order set, and then you can sign it at the end. It’s basically a forcing function on physician autonomy. It’s not to take away their autonomy. It’s to make them better.
We launched intelligent triage. It’s a symptom checker. We have inbox management, which is important for easing the burden on our practices. We've also got on-demand virtual care. In central Pennsylvania, our patient population is trained to go to the emergency department, which is a huge amount of waste. We're going to try to retrain them to do the on-demand part.
How are you thinking about AI in healthcare?
AI and technology is the way through the conundrum we're in. We're in uncertain times for payers, uncertain times for patient populations, and if we just keep on applying the traditional logic to it, we aren't going to make it based on cost structure and ability to meet the demands of people. You can get out in front of it a little too far, and then you're chasing your tail. Where we try to be is right behind the front edge, where you're taking things that are new and applicable, but not so far out that you're not sure if the stuff works.
My worry is that we don't have enough bandwidth to do as much as we need to do, fast enough, to get out in front of our cost structure and make it so you can translate artificial intelligence or augmented intelligence into your operations in a cohesive way. I was talking to our people that do AI and said, “Show me your plan for this year.” It was not ambitious enough.
What is the biggest impediment to AI in healthcare?
The biggest impediment right now is the reliability of your artificial intelligence. In healthcare, our problem is that we're trying to work out into that 95-plus area of the curve. The area of the curve is the probability that your prediction is correct. How do we go in and validate the machine learning models? We’ve got to get them higher out on that curve, so that there's more reliability.
We've also got to move away from the word artificial. At my old company, we stopped calling it artificial. We called it augmented because it means you're always connecting it to a human.
