Glenn D. Steele Jr., MD, president and chief executive officer of Geisinger Health System, says the integrated delivery network's pioneering population health programs depend on insightful use of data to drive behavior change.
Serving 43 counties over some 20,000 square miles of northeastern Pennsylvania, Geisinger provides care for more than 2.6 million people; its innovative efforts to combat diabetes, coronary disease, COPD, hypertension and other chronic ailments have been held up as examples for the rest of the country.
When Steele delivers the keynote at the HIMSS Media Healthcare Business Intelligence Forum, April 16-17 in Washington, D.C., he'll offer lessons from Geisinger about deploying analytics tools to improve patient care and the bottom line. And he'll stress the importance of making sure that data changes behavior and thought processes – not just of patients, but of providers and payers, too.
In a Q&A with Healthcare IT News, Steele discussed Geisinger's strategies for deploying analytics for care coordination, emphasized the "critical" importance of integrating clinical and financial data, pointed to specific cases where BI has borne fruit for Geisinger and offered advice for smaller providers looking to make the most of their data.
Q: We'll talk about Geisinger in a second, but first, from your point of view, how is healthcare in general doing these days when it comes to using business intelligence and analytics?
A: Well, we're probably about to enter the 19th century. (Laughs.)
Q: Baby steps, right? But why are we so far behind other industries? What's standing in the way of smarter use of data?
A: For all the reasons that you and I could talk about for a week. We have legitimate regulatory concerns, and I think they've always taken precedence over true innovation in terms of how we look at our data, how we analyze it, how we distribute it. How we use it to change behavior. I don't think the balancing act between innovation and regulation is correct in most areas of healthcare data.
We also have the intrinsic structural issue in healthcare, where it's been compartmentalized on both the payer and provider sides, and each of those seems to strive for an optimal function without actually any integrated structural aspirations. And that's changing as well, but obviously when you sell IT enabling into all those compartments, you're kind of handicapped right from the start.
We have a series of expectations, even in the biggest compartments that are most well-capitalized, the hospital-centrics; we have a series of expectations that up until recently were pretty simplistic: that if you put an electronic health record in it would automatically improve badly engineered and badly transacted systems. It doesn't work that way.
And then we have this great big divide between payer and provider where a huge amount of the data, which is collected on the transaction side, is kind of husbanded and treated as an intellectual property. And is either on-purpose not used, or inadvertently not-usable, whether because of a lack of timeliness or a lack of functionality.
We could go on for weeks talking about why we are where we are, but the good thing is, we're changing.
Q: What's finally bringing about that change? Improvements in the technology or shifts in attitudes and awareness?
A: Oh, I think it's both. I think it's both, and I think the more critical rate limiter is the attitudes and the intrinsic behavior, which, quite frankly, I think is amplified by a lot of the stakeholder business models, and it's also amplified by backward-looking regulatory issues. And I think we're going to have to blast through both of those things to get maximum functionality when you think about healthcare enabling through health healthcare information technology.
Q: Describe Geisinger's use of analytics. Is there a mission statement or overarching philosophy to how you use BI?
A: Well, our top strategic aim is innovation and quality. Our structural advantage that we talk about internally and externally over the past 15 years is this payer-provider sweet spot. We've tried to figure out how our version of vertical integration between payer and provider can really optimally mesh the information and the use and analysis of it from both side of the house. That's a 35,000-foot glib summary of what drives us, and continues to drive us.
But the end point is change in behavior. We realize that in order to capture the 30 to 40 percent of value which is now lost because of stuff in healthcare that doesn't help human beings, actually hurts them, in order to capture that, we have to fundamentally change behavior. And we have to change it on behalf of our providers as well as our patients. And on the insurance side of the house.
The way to change behavior is not just changing the incentives. Everybody has been – appropriately – whacking fee-for-service, and that's OK. I'm alright with that. You have to change the incentives but you also have to enable people with timely feedback about what they're doing, and the consequences of that, and what they should be doing. And that's all about data.
Again, we think that in order to achieve that continuing aspiration of innovation and quality, we've got to have, from both sides of the house, behavior change enabled by timely and usable data.
Q: Talk a bit about Geisinger's ProvenCare program, which puts evidence-based standards and patient engagement to work in the service of fixed-price procedures. How does data play into it?
A: We have come to believe that a huge amount of our value reengineering, whether it's hospital-based care episodes or whether it's taking care of patients with multiple chronic diseases, does two things. Number one, it gives better outcomes, both near-term as well as long-term. Number two, it decreases a lot of the cost.
There's a lot of things that realization has allowed us to do. One of them is it allows us to look at cost as a surrogate for bad outcome, so if on the insurance company side you see a cohort of high-cost patients, we pinpoint and target, and have both payer and provider engaging in a discussion of how can we improve those outcomes for that cohort of patients.
A good example would be use of erythropoietin in patients who have anemia that's associated with chronic renal disease. This is really a paradigm; it's the result of lots of high-price but very effective biologicals. And the questions we asked when we looked at this group of patients with anemia secondary to chronic renal disease, is are they getting the best outcomes in terms of alleviation of their anemia?
What we found, in looking at the data and analyzing the data – and then looking at every step along the way at how this care was given to this group of patients throughout our system – we found that about 20 percent of those patients actually didn't need to be receiving the high-cost biological that is EPO. That by standard, off-the-shelf indications, they could just as well have been treated with iron, at about one one-hundredth of the cost.
And, by the way, EPO has side-effects, and those are cardiovascular toxicities, so stripping out the use of that high-expense biological was not only a cost-savings but a savings in terms of avoiding toxicity for a huge number of these patients.
So that's about a 20 percent benefit right there in extracting that value by doing away either with ambiguous indications or essentially no indications for the use.
The second thing we noticed in looking at how these patients were cared for was the efficiency. Inpatients that actually had hard indications for the particular biologic EPO, what was the efficiency of the actual treatment protocol, and how often did the patients have optimized red blood cell levels that were maintained and sustained. And what could we do in terms of changing the information that went back to the places that were giving that EPO treatment that could optimize their red blood cell levels and sustainability?
The long and the short of it is, we changed the venues where they were treated. We changed the expertise of the people who were essentially doing the treatment so that they were people who were much more attuned to doing that treatment as a focus of their job – as opposed to only one out of 100 or 150 things. And we supported them with best-practice algorithms that can be very dynamic.
Now, I think that's a perfect example of how data – both in terms of analytics and in terms of fundamentally effecting a change in the behavior of how these folks were treated on the provider side – really is to be married.
Q: And what about Geisinger's Center for Clinical Innovation – one whose main goals is to "leverage IT and advanced analytics to support population health." What are some gratifying things that have come out of there, recently?
A: We have changed expectations with regard to how we treat patients with serious chronic disease. Type 2 diabetes was just the first one in a list. We've now expanded it to coronary artery disease, to COPD, to hypertension.
But probably our longest-lived commitment was to treat type 2 diabetes patients – there were probably 27,000 or so that we started with, almost eight years ago – to create a set of aspirations, a set of data feedback loops and to change our incentives, in terms of how we pay our primary care physicians. And really commit to a bundle of nine best-practice goals being achieved in all of these type 2 diabetics that we took off the shelf from all of these endocrinologists and diabetologists but had never previously been engineered into the expectations for trying to achieve 100 percent of those goals in 100 percent of the diabetic patients.
We have an article coming out in May in the American Journal of Managed Care that shows that after three years of improvement in these 27,000 patients, getting all of them up to and optimal for all nine of these expectations, it took only three years before we saw a significant diminution in their risk for heart attacks, for strokes, for amputations or for diabetic retinopathy.
For me as a clinician, someone who actually used to take care of patients, I mean, that's amazing. To actually see, after only three years, a beneficial effect on diabetes-related secondary disease.
Now we haven't even done the economic analysis of that, but presumably, if you're avoiding heart attacks and strokes, you're probably avoiding a lot of hospitalization – high-cost and high aggravation treatment.
Q: IBM has helped Geisinger to consolidate its Clinical Decision Intelligence System. What has that meant, bringing all those different data sources – EHR, insurance claims, even patient satisfaction scores – together?
A: It's critical. It's absolutely critical. I can't quite remember when we started that CDIS, but it was around 2001 or 2002, and we were just finishing the electrification of our system. We'd started the electronic health record in the ambulatory aspects of our system, and then only by 2002 or 2003 had it completely transacted throughout all of our hospitals.
But we made the commitment to do the data warehousing because we wanted to ask questions about how many of the people we were responsible for – in our 41- or 42-county service area at that time – how many people had osteoporosis? How many people had been tested with DEXA (bone density scans) for osteoporosis? And how many of those had actually had a primary care physician list osteoporosis as one of the entities on the problem list? And then how many of those folks had been treated for osteoporosis appropriately?
Now you can't ask any of those questions without a data warehouse. And then, obviously, in order to try to get a best-practice treatment algorithm, you've got to have the feedback through CDIS in order to get it out to our 70-some sites. So it's been critical. Absolutely critical.
Q: You guys are so far ahead of the game compared to so many other providers. But what have been some challenges or stumbling blocks for you along the way?
A: Oh, there's a huge number of challenges. And again, we could talk for a week about it. Again, as I said, it's really all about behavior change. And a lot of times in healthcare – particularly among docs, when we have a good idea, we just assume everybody else thinks it's a good idea. And so we've designed solutions in a number of our really important problem areas that have been duds, because the people that are actually transacting the solutions don't think they're solutions.
Another really good example: When we, as a group of providers, whether it's docs or PAs or nurses or pharmacists, when we hand an individual who's got a chronic disease a prescription, we assume that they, number one, agree with our recommendations for the treatment, and we assume they're going to get the prescription filled. Those assumptions are wrong between one-third and 50 percent of the time. Isn't that amazing?
And so we've done what I consider to be a predicated for really activating the human beings who should be partners in what we do – our insurance company members or our patients – through this effort called OpenNotes. We've opened up our progress notes throughout our entire system to all of our patients. They can do a Web-based connection and see what their doctors have said about them – even see what our trainees have written in the charts about them – in real time. With two exceptions: We haven't opened up our notes in psychiatry and psychology, and we haven't opened up our notes in the emergency room for some specific and real reasons.
We've done this with BI Deaconess in Boston, and with a cohort from Harborview in Washington State, and what we've found from the beta test, which was a relatively small cohort of primary care physicians and patients, to this big, big opening, is that patients and their surrogates are very interested in diving into those notes.
First, the frequency of interaction is extraordinarily high. Number two, we unearthed problems – either assumptions on the part of our providers that were not correct, or in terms of patients either understanding what's going on or agreeing to the treatment. And we can correct those, we can change our assumptions, we can actually get closer to a true, meaningful partnership, and that's a start on that second area of behavior change: understanding what turns patients on, what their sweet spot is.
This is a cliche, but it's a journey; it's not an epiphany. But I'm really excited about that fundamental change in the relationship between the people we're responsible for and us.
Q: There are going to be a lot of people at Healthcare Business Intelligence Forum looking to learn, looking for tips on how smaller providers could some of this stuff to work in their own hospitals and practices. What are some doable, practicable tips you could offer for putting BI tools to work?
A: I think the biggest piece of advice is to look at the functionality you want first. There are going to be people there who have more technical capabilities than I'll ever have. And understand the nuance of the various systems, whether they're hospital-based or ambulatory, or what have you.
But I think what drives us is, "What is the functionality you want? What is the end-point in the relationship between us as insurers and us as insurance company members?" Or, "What is the relationship we want between us as caregivers and as someone with chronic heart disease?"
And then work backwards, from: "What is it going to take to get a different functional outcome than the one I'm getting now?" And there's almost no outcome that can't be improved.
The second thing to consider, which I think is very important, is: "What is my capacity, based on my structure, based on my size, based on my access to capital, what is realistic to me? And if I can't achieve the optimal function, how am I going to have to change?"
One of the thing we've learned – and I'm always nervous about extrapolating to others, because you have to change from what we've done well and had success in at Geisinger, to "What works in Philadelphia, or what works in Northern Virginia/Bon Secours?" or what have you.
But basically, it's "What can I do that has a great probability of success, early on?"
I learned this from Jim Walker, who had been our CMIO for a period of about 12 years. A great partnership. He's now moved on to Siemens. He would ask, "What is it that's doable, even if it's not perfect?"
Y'know, "How can we begin to enable small practices with a non-proprietary connection to our version of Epic. And it won't be perfect, but it'll help them. And then they'll start to demand more intricacies and more complexity to move toward better, if not perfect function."
But so often, what happens is everybody wants the perfect solution immediately, and it ends up impeding taking the first step. Which is always interesting.
Q: If healthcare is just now entering the 19th century, as you said, what will it take to get us into the 21st? Will we have to wait 200 years, or will it come faster than that?
A: Oh no, I think we're in the midst of just huge change right now, and I think what it's going to take is winners and losers. We're going to have organizations that figure it out and also have a viable business model, and they're going to be different than what we see now, with this compartmentalized, fragmented, mom-and-pop industry.
That change could be painful. It's kind of like the grocery business 120 or 130 years ago – the supply chain issues and the consolidation and what have you put almost all of the small grocery stores out of business. On the other hand, the amount of household resources that use to go toward groceries – it used to be up to about 30 percent of discretionary money for a household – has gone down to low single digits now. And guess what's happening to healthcare.
I think this change is going to be extraordinarily fast. I think it's going to be quite disruptive, and I think there will be winners and losers. And my only wish would be that I was younger. If I had my experience and my knowledge now, and I were just beginning, in the midst of this transformation, I could do some real damage.
Q: What's immediately in Geisinger's sights, going forward? What do you hope to accomplish in the next two years or so?
A: Well, we're doing a lot of scaling and generalizing experiments. We basically feel pretty comfortable, knock on wood, that a lot of our value reengineering here that's been enabled by information technology is sustainable within our market and within our particular fiduciary structure. But the question is how much of it is scalable and generalizable, outside of Pennsylvania.
We're doing that experiment in three was. Number one, our insurance company is going outside of Pennsylvania and working with other, non-Geisinger providers. Number two, we're consolidating on the provider side, so when you move to Scranton, when you move to Harrisburg, when you move to New Jersey with AtlantiCare, the question is whether we'll be able to do in those markets what we've been able to do here.
And the third is xG. We've created a very interesting joint venture, for-profit with Oak Investment Partners (which invested $40 million in the initiative), and we're attempting to take our data analytics, our care management and a lot of our predictive modeling out to lots of other entities on both the provider and payer side.
I guess one more thing that we're really committed to – and we've recruited new leadership into the information technology area, on both the provider and payer side – we're interested in getting distributed data out there, we're interested in the functionality aspects. I would love to see Geisinger set the pace on applications, and to move to a much more open-source platform, as opposed to this incredibly rigid, incredibly proprietary pipe that we're now dependent upon for our EHR.
Those would be some of my visions for the future.
[See also: Geisinger adds BI platform to labs]
[See also: Geisinger and IBM collaborate on new IT infrastructure]
