Emory Using Real-World Evidence for Medication Formulary Design
Atlanta-based Emory Healthcare is in the early stages of partnering with Stanford Medicine spinout Atropos Health on real-world evidence generation. The Atropos Evidence Network, with over 300 million patient records, is working with Emory users to run studies on both large national data sets and local Emory data. Emory’s Collin Lee, PharmD, director of clinical pharmacy services, recently spoke with Healthcare Innovation about some use cases around medication formulary design and protocols.
Healthcare Innovation: Before we talk about Atropos, could you describe your role as director of clinical pharmacy services at Emory and some of the issues you deal with day to day?
Lee: Sure. At Emory Healthcare we have 10 hospitals. We have multiple infusion centers and hundreds of clinics out there. Probably the main responsibility that I have that fits in with Atropos is that I am responsible for formulary management. That means that I'm responsible for reading the literature, making sure that there's evidence to support bringing a medication within the hospital. Do we need to restrict it? Are there any specific things that we need to build into the computer system to help facilitate it? On the opposite side of that is looking at utilization and asking if there is an opportunity for us to use something else at a lower cost without compromising patient care. And I do a lot of looking at quality. How well are we doing with our glycemic control, our glucose control? Or how well are we doing with our anticoagulants? And then trying to see if there's something that we can change to improve quality for our our patients.
HCI: I've interviewed execs from Atropos before, so I understand a little bit about their concept. I saw an interesting quote from Alistair Erskine, M.D., your chief information and digital officer, who said, “Atropos surfaces practice-based evidence where evidence-based medicine does not exist.” Do you see it that way, too? Could this have a huge impact on the speed at which you can adapt new clinical practices, based on observations of what's happening in real-world settings?
Lee: Yes, absolutely. I can give you a good example of that. There is a medication called Alvimopan that we use in our colorectal surgery patients. Those patients need to get an opioid. This particular medication is an opioid receptor antagonist. It helps the bowel recover, so you don't get stasis that you get from opioids. There's a new medication called Naloxegol. It works exactly the same way, but it's not studied in colorectal surgery patients, but it is a lot less expensive than Alvimopan. So my question is: Could we use Naloxegol in that patient population, even though there's very little evidence for it in the literature? I can use Atropos to ask that question across many other institutions in the real world: Is anybody else doing this? And if they are, I can look at their outcomes compared to my outcomes and say whether or not I think that this is something that maybe Emory would want to try. I can get that answer where there's not anything in the literature to support that.
HCI: Is there a conservative mindset among some clinicians who believe that kind of evidence isn't as good as longer, more traditional studies?
Lee: I would say it probably depends on what we're treating and what the seriousness of the outcome is. There are some physicians who would say that if Stanford or Yale has been doing this for the past three years and they haven't had any issues with it, they get some comfort from that. But if it involved things related to something like treatment of stroke, probably they might be less inclined to do it unless there was some specific data that would support doing it.
HCI: I read that by accessing large, historically untapped local clinical data, Emory and Atropos hope to make impactful changes in medication formulary design and protocols. Can you describe an example of that?
Lee: One of the things that we were looking at from a formulary decision is about whether we were going to use Methylene Blue or Isosulfan Blue for sentinel node biopsies for women who have breast cancer. There's a very, very small incidence of tissue necrosis with Methylene Blue. Even though there wasn't anything necessarily out there about it happening with Isosulfan Blue, I wanted to look across this huge database and ask if there is any documentation of Isosulfan Blue causing tissue necrosis. And in fact, there wasn't. So that was a really big key for me to say, this is not apples to apples. If we were to use Methylene Blue, there are going to be some patients who might get tissue necrosis, and I need to take that into account when I make this decision.
There are some clinical quality things we can do as well. I just had Atropos run a query for me looking at patients who get hypoglycemic — they have a blood glucose level that drops too low. Based on some of the things that we've seen from our own patients, we think we’ve identified some predictive factors about the patients who have these low blood glucose levels. We’re wondering if that holds true. Are these things that we could use to predict people who are at risk for having low glucose levels? So I've asked Atropos to take those factors and run a study on that. They can take everybody who had a blood glucose less than 50 and everybody who didn't, and see if there's a statistical difference between those two groups in these factors, to see whether or not they could be predictive. The nice thing about Atropos, too, is that it does some propensity score matching, so it makes it a little more apples to apples.
HCI: Dd you have pharmacists who are working for you who have lots of ideas of hypotheses or questions to ask of this data?
Lee: Yes. We're always interested to see what other people are seeing, too. We could probably take all of Atropos’ time and just keep making queries. We're curious people, right?
HCI: Are you hearing of other areas of Emory Healthcare where people are interested in looking at the data in similar ways, like surgeons or chronic disease managers?
Lee: It seems like there are lots of ways this could be used. Respiratory therapy could use this. I know that we've had ophthalmology doing it. We've had our lab doing things. I think it opens up a huge number of thought-generating or hypothesis-generating possibilities for people. Even if we were going to do a study, we could look to see if there is something there before we actually go and put in an IRB request.
HCI: How long have you been working with Atropos?
Lee: I would say at least three to four months for me personally. Not super long, but the turnaround time is quick. And one thing I like for the particular kind of questions that I've been putting in is that there's a medical person who's looking at that before it comes back to me, just to validate that what I'm getting back is really answering the question that I'm asking, so that gives me a lot of faith in it, too.
HCI: Do you sometimes ask questions of the local Emory data and compare that to results from the larger de-identified data set?
Lee: Yes, I’ve done both. Because I have both questions, right? Most of my questions involve what we are seeing out in the world, and how does that compare to Emory. So I'll run those two reports, and then I can look between the two.
