Mike McDermott, MD, is president and CEO of Mary Washington Healthcare in Fredericksburg, Va. He is a diagnostic and interventional radiologist who served in a variety of leadership and governance roles at the system prior to becoming its CEO in 2015. He sat down with Phil Lindemann, Epic's vice president of business intelligence and analytics, to talk about Epic's Cosmos and the future of healthcare research.
Dr. Mike McDermott: How many health systems are participating in Cosmos right now?
Phil Lindemann: More than 150 organizations are live, which includes about 800 hospitals and 10,000 clinics with their data in Cosmos right now. In total, it's currently around 120 million patients. It's the largest single set of EHR data in the U.S., potentially the world. But what's really amazing is that Cosmos mirrors the demographics of the U.S. population. Oftentimes, research is done using data pooled from academic medical centers, which skews toward urban populations. Epic organizations are across the country and in all sorts of places, so Cosmos includes people from rural areas and different socioeconomic backgrounds that other data sets don't usually have access to for their studies.
MM: How do we get insights from Cosmos into the hands of clinicians who don't have the time to deep-dive into the tool themselves? How do you picture it fitting into their workflow?
PL: Today, when you go to a physician's office, some treatment decisions lack evidence or are based on a very small set of data. Many times, they're based on a physician's personal experience, education or what they think might be true. Our goal is that when a physician reviews treatment options with a patient, the patient can ask the question, “Well, what happened to patients like me? What were the actual outcomes?” and the physician can look to Cosmos and pull up the actual results of other patients with a similar condition who tried this treatment, as well as communicate what the long-term outcomes were.
MM: I think there's a core set of problems that are a good place to start using this, like treating hypertension, diabetes and depression. Could this make a difference in the current mental health crisis we're seeing across the U.S. that has been exacerbated by the pandemic?
PL: That's a huge area where we currently rely on "try and see" for treatment, but a physician could instead go to Cosmos and say, “Which treatment had the best outcome for other patients in this situation?” and give the patient options. Patients can see those results and then work with their clinicians to decide on the best treatment based on whether patients like them had more success controlling their depression with cognitive behavioral therapy, medications or both. There's a lot of potential with Cosmos in many areas of health.
MM: How close are we to having this in physicians' offices?
PL: We're targeting the first use cases — common diagnoses like hypertension and diabetes — within the next two years, and then we'll expand from there. We plan to implement this in offices a bit more gradually than other applications and features we've developed so that we can learn as we go and make sure that we're getting it right, because this has the potential to change the way practice is done completely.
MM: Researchers in academic medicine must love this — I know I would. Is there anyone already looking at outcomes of patients who took certain medications long-term compared to patients who didn't?
PL: We actually just opened up Cosmos for broad research across all participants in the last few months, so at this point we've only just begun. One example of an early success was that researchers found a higher incidence of myocarditis in children who have had COVID-19 compared to those who haven't.1 This and related work using Cosmos data helped the FDA and CDC during their review to decide to approve the COVID-19 vaccine for younger children.
MM: You have to have a healthy bit of skepticism in such a consumer-driven market for pharmaceuticals. But Cosmos has the potential to really remove that by showing what the long-term outcomes were for people who took a medication. Has there been any pushback from pharmaceutical companies?
PL: Cosmos is certainly going to add a new variable to drug development. I think this is going to continue pushing life science companies to develop new drugs that are tailored to more specific populations, rather than a lot of the generalization we see today.
MM: We've talked about drugs, but this could also be used for medical devices, right?
PL: Right, orthopedic surgeons could look at Cosmos prior to a joint replacement to say, “Which implant had the lowest rate of failure over time?” They would then be able to decide using information directly from patients, such as patient-reported outcomes, which implants had a lower chance of needing to do another surgery later.
MM: Give me a high-level look under the hood at the technical side: How does Cosmos work?
PL: First, the code for every Epic system in the world is the same at the database level. We save allergies, providers, medications and everything else in the same place for each organization. We can go element by element and match them up to each other easily. Another early focus was privacy. Many data aggregators send your data somewhere else to remove identifying information, but then protected health information has left your site and is sitting somewhere else, which was a risk we did not want to take. Instead, our approach is to remove 16 categories of direct identifiers before data even leaves your servers. Additionally, it was important for us to give patients the ability to opt out at any time and be able to handle state and local regulations that go above and beyond HIPAA standards.
MM: How can organizations get started with Cosmos, and what do they get from joining?
PL: It's free to sign up, and we onboard organizations quarterly and have sped up the install process to take only two months. After we finish the install, you can choose 50 users to give access to at your organization and they can immediately start exploring the data and using it to publish research.
MM: So what's coming next? What is right around the corner?
PL: There are two big things coming next year. First, a data science toolkit for researchers. Right now, the interface is fun and easy to use, but it doesn't include advanced statistical tools. You'll be able to run regression tests or other advanced algorithms with this toolkit. The other important feature is called Look-Alikes. It will be an embedded tool for providers who are in the office with a patient experiencing a rare constellation of symptoms. As the physician treating this complex patient, you will be able to ping Cosmos for patients who have had this same rare circumstance, and Cosmos will return physicians who have treated a patient like yours and allow you to connect with them so you can find out what they tried, what worked and what didn't work.
1 Oster M. (2021). mRNA COVID-19 Vaccine-Associated Myocarditis. CDC COVID-19 Vaccine Task Force.