For Terri Steinberg, population health isn’t just another project; it’s a passion. She has spent the last five years working to build a program that leverages analytics – and the use of embedded care managers – to manage the care of 200,000 patients. But the journey hasn’t always been easy. Christiana Care had to first create an infrastructure that would enable more sophisticated use of data, then demonstrate that it could be utilized to achieve improved outcomes. And then, of course, there’s the tweaking that needs to be done to help the program continue to thrive.
In this interview, Steinberg, who serves as Chief Health Information Officer and VP of Population Health Informatics, talks about what is has taken for Christiana Care to start the transition from fee-for-service to value, including securing a grant from CMS and having leadership that bought in to the vision. She also discusses the “evolution” required to align incentives, the pivotal role that care managers play in improving care and decreasing readmissions, and what it’s been like to build Carelink CareNow from the ground up.
- Expanding Carelink CareNow – “We grew very quickly”
- Redesigning clinical workflows
- Building from scratch – “Most of what we used is homegrown.”
- “Giant data buckets”
- Redesigning workflows to enable embedded care management
- Challenge of having 20 different EMRs across the system
- Utilizing CCDs, regional HIE
- “You have to be careful not to overtask people.”
The operations model for a single entity grant where all of the patients originate in your ED is quite different from a payer-based care management mix where some of the members are your providers’ patients. It’s a very different model, and while the changes are subtle, they are important changes.
It was very challenging, especially since the CMS Innovation Center (CMMI) grant expected us to have IT systems up and running on day one. The IT systems we needed did not exist. You couldn’t buy them. Most of what we used is homegrown.
The degree to which the providers can integrate gaps in care workflow with care management workflow helps as well. So we’ve designed an optional workflow that touches the practices.
What can easily happen with care managers is if you give them a system that’s sufficiently sophisticated with lots of tasks based on data, they can spend their whole day closing tasks and not managing care, so you have to be really careful.
Gamble: In terms of expanding Carelink CareNow’s capabilities to include multiple disease states, how you did manage the process so that it expanded in a deliberate way as opposed to growing too big, too fast?
Steinberg: That’s a great question. At the end of the day, Carelink CareNow was no different from any other business in that controlling the growth of the business can make or break it. We experienced some quick success early on, but we couldn’t get a great read on it because we couldn’t get our claims data from Medicare, so we never really knew for the first two years of our grant how we were performing. We had to use markers; we knew our patient satisfaction was very good, and we knew that we had less depression in our heart disease patients. But although we had some soft measures that it was making a difference, we really didn’t know how well we were doing. And yet, the hype was such that it was intuitively the right thing to do, and before we knew it, everybody wanted to be Carelinked. It became a verb.
Delaware is a very tightly integrated state where you’re two people (or sometimes even one person) away from knowing everybody, and the medical community has a very tight collaboration. And so the other health systems were very interested in working with us. Our chief financial officer was strongly supportive of our strategy, as were our CEO, CMO, and everyone in the C-suite.
The challenge was not growing too quickly — and we did, so there were some pitfalls. One of the pitfalls which we are addressing now is that the operations model for a single entity grant where all of the patients originate in your ED is quite different from a payer-based care management mix where some of the members are your providers’ patients. It’s a very different model, and while the changes are subtle, they are important changes. We’ve been running so fast for the past five years that we have not until now had an opportunity to stop and reimagine and redesign our clinical workflow.
So we’ve done that now. We’ve reviewed current state. We’ve reviewed what we need to change. We’ve implemented changes. We’re really in the optimization phase for a new business model. But yes, we grew fast. We like to say is we were driving down the road at 60 miles an hour and changing our clothes at the same time. It was very challenging, especially since the CMS Innovation Center (CMMI) grant expected us to have IT systems up and running on day one. The IT systems we needed did not exist. You couldn’t buy them. Most of what we used is homegrown and we’ve had to build from scratch. We’re a hospital IT shop. We don’t have teams of developers. We don’t have development tools. We don’t have UML tools. We don’t have processes. We had to do this with contract developers, and it’s been very challenging.
Gamble: So I can imagine you’ve had to continually tweak the system and make adjustments.
Steinberg: We have. And actually, we’re on our second generation of systems. Five years is an eternity in IT. We’ve outgrown our first generation systems and we are either rewriting our systems or moving to commercial software in the few cases it exists. As we rewrite our software, we’re considering writing it in a way that is portable to other organizations. For example, the front end that the care managers use is a commercial software package. Most commercial care management applications started out in the provider’s space, which is insufficient for real-time data notification.
And so one important component of what we do is we have a giant data bucket. Based on the arrival of pieces of data that are basically logic-based data paradigms, we trigger care managers in real time. So they’re notified if somebody is in an ED in another part of the state, or if a hemoglobin A1c is abnormal, or if a smart scale has just read a 5-pound change from yesterday, and they can immediately jump on it. The commercial software is the worklist where they see the patient’s name and document, but the backend is the smart data. It’s basically a web services-based data platform that plugs into commercial software. As we rewrite that suite of features and functions on a commercial data platform, we’re considering selling it, because nothing like that exists today, and we think we’re not the only healthcare organization that would like to use these functions.
Gamble: You talked about reimagining and redesigning workflows — how are you doing that in a way that it helps clinicians to be as efficient as possible and prevents them from being overburdened by all the data?
Steinberg: That part of the workflow we have down. We use an embedded care management model where a single care manager is assigned to a practice across all the lines of business. So if we have six contracts that we’re care managing, we have one care manager, and that person is really a part of the team. We make sure social workers, pharmacists, nutritionists and other services available to that practice as needed. And so the workflow that touches the doctor is pretty good. Could it be better? Sure. We could message in and out of the EMRs, but there are more than 20 different EMRs that our care managers use. Obviously the care manager is going to use the EMRs in the practice, but we are across the entire state of Delaware and the surrounding states — Maryland, New Jersey, etc., and so you can imagine it’s a lot of EMRs. When you have to do duplicate data entry, it’s not so good.
The degree to which CCDs are sent to our regional HIE does help the workflow significantly. The degree to which the providers can integrate gaps in care workflow with care management workflow helps as well. So we’ve designed an optional workflow that touches the practices. The limiting factor is the EMRs, because they’re all over the map in terms of the sophistication, or even whether an EMR is implemented at all. It’s a whole story unto itself how you integrate the system. In terms of FHIR-based integration, we’re not there yet. We use CCDs, and it works. We have big dreams about how an integrated EMR data platform would make everything we do much better and easier.
We also make significant use of our regional HIE, the DHIN, and that really saves us. Our regional HIE communicates with exchanges in other states, and that helps as well. The better way to do it is to have contact sharing or direct API from one system to another in the way that Dr. Don Rucker from ONC is imagining. That’s much better, but nobody’s there yet. None of the vendors are there. It’s a direction that’s been promoted, but as far as the degree of adoption, we’ll have to wait and see.
In the meantime, we have a business to run, so we do duplicate data entry when necessary. The workflows I’m really referring to are the data flows. Do we pick up all the right ADT transactions? Do we pick up all the right fields in the transaction? How many times are we updating the transaction? Are we being most efficient and effective? Because what can easily happen with care managers is if you give them a system that’s sufficiently sophisticated with lots of tasks based on data, they can spend their whole day closing tasks and not managing care, so you have to be really careful. If you put people in front of a machine, they’re going to do whatever the machine tells them to do. You have to be really careful not to overtask people and to make sure that you put the right triggers for care management and you don’t turn them into task-closing machines.
And that’s what we’re working on right now. We’re taking a step back to look at all of this, because managing 3,000 people across one disease state is very different from managing 120,000 people across multiple lines of business and 60 disease states.
Chapter 3 Coming Soon…