Christiana Care's Terri Steinberg: Analytics vital to successful population health

For Christiana Care Chief Health Information Officer Terri Steinberg, the key to successful population health management is balance in three areas: people to run the program, analytics to offer real-time updates and a robust electronic health record system. Simply put, without one area stabilized, the others are likely to falter.

"If you implement the analytics without a population health program, it doesn't work," she said. "If you implement the EHR without the people to run the program, it doesn't work."

In 2012, Steinberg, who also serves as vice president of population health analytics, helped Christiana win a $10 million Center for Medicare and Medicaid Innovation grant to launch a program to find and fill gaps in care for ischemic heart disease patients. The health system backed the effort with $6 million of its own money. She called the effort a technology play, saying it's important to integrate internal and external data.

In a recent interview with FierceHealthIT, Steinberg discussed details of the initial grant effort, as well as updates to Christiana's work. She also talked about challenges she and her colleagues continue to face.

FierceHealthIT: You say integrating both internal and outside data is important to population health. Why?

Terri Steinberg: Before the grant, we really thought of our systems as enterprise or departmental systems. For example, we have a visiting nurse business unit of Christiana Care. It was always thought of as a departmental system off to the side, but the data generated was very rich.

I started to think of our systems not as enterprise or departmental, but as strategic and non-strategic. And that was important from a business perspective. It really goes back to the notion of point-to-point interfaces. We have a statewide HIE, the Delaware Health Information Network. It's very well implemented and very robust. Ninety-eight percent of our results flow through it.

Years ago, we stopped doing point-to-point interfaces. We said, "System A, if you want to send data to System B at Christiana Care, go through DHIN." Everything is standard, the data is shared across the state. I was never really comfortable with that because we were really giving our data and connections to another business. It made me nervous. That's when I had the notion of strategic versus non-strategic. If it's not strategic, that's fine, but if it's strategic I want it all and I want to make those connections.

FHIT: EHRs and people are vital to your population health efforts. What role do analytics play?

Steinberg: The difference between what we do and lots of what everybody else does is that when you discover that a patient requires an intervention, you can't just put it on an Excel spreadsheet; you have to actually trigger the person who does the intervention. With that in mind, an important component of our strategy is the ability to take all of this data, which comes from many sources both within Christiana Care and from outside, and run analytics on it. So if you had a hospital admission down south, we know about it and we can use that data.

The analytics do a number of things. They segment the population. So care coordinators in Christiana CareLink--doctors, nurses, social workers, pharmacists--only need to react to those individuals who are high risk.

And we actually task the right person, saying "John Smith is within the top 15 percent of risk, he's had an ED admission to Christiana Care, and because he's high risk, we need to follow up." Somebody else, same scenario, low risk, doesn't need to follow up. That's the magic. We are able to scale.

FHIT: Are there any specific results that really stand out that you think made leadership say "this is what we should be doing?"

Steinberg: No. There's too much to learn. We couldn't do it as a standard ROI because CMS wouldn't give us our claims data. The first three years, this was CMS only and we had no data. Starting this year, we have lots of risk programs, lots of data, we're a Medicare Accountable Care Organization, we have a lot of independent risk contracts so now we can't get our data, but it's not really about the ROI. It's about the notion that, in the state of Delaware, we asked for risk.

FHIT: Are you looking to gather more patient-generated data?

Steinberg: We have the concept of a suite of tools; a bag of biometric devices. Today, if you're a patient, you call up and say you need something, you go to the ED or come into the office; those are your two options. We want another option--to deliver and set up a suite of biometric devices so the patient can stay home. Or maybe the patient will skip the ED and have a direct admission.

Biometric devices are really important. The data generated from biometric devices is gold, but I'm in the process of trying to figure out where in the pipe it goes. Does it go in the HIE? Does it go into the telemedicine platform? Does it come in here and become part of our predictive analytics? Or all of the above? We know it's important from a patient workflow perspective.

At the end of the day, many different pipes may be the answer, but the reason we're trying to find the one best answer is that it's bad hygiene to send the same data to multiple places.

FHIT: What challenges have you encountered?

Steinberg: Every day is a challenge. Here's one: In my Medicare ACO, I have 18 different EMRs. Gaps in care. How do I understand who has a quality measure that needs to be satisfied?

To manage quality in an ACO, I need that EMR data. I could do what everyone else does and do chart reviews, but I will not do it; it's stupid. It doesn't make sense because you can't scale it enough.

We have a lot on our plate. We're growing our size dramatically. We're taking on risk contracts. We're expanding our clinical programs and we're expanding our technology. Think about EMR integration on this scale, we'll have clinical data from labs and hospitals, EMR data from small practices, claims data, a prediction engine that sifts through all that data and tells us what we need to do--and then we'll deliver it to the right person to make it actionable. Each of those pieces is monumentally hard.

Plus, we have to keep the lights on and keep doing the day-to-day stuff. It's like driving down the highway doing 60 miles an hour while you're changing your clothes. 

It's also hard to get people to understand. The vendors are coming in and saying "let me show you what my population health program is" and it's all about their EMR. That's not population health.

Executing the plan is the biggest challenge of all. 

Editor's Note: This interview has been edited for clarity and length.