ORLANDO, Fla. – The COVID-19 pandemic demonstrated the importance of ready access to data. In a sea of unknowns about the virus, analytics on a population-wide level helped provide insights for public health officials, providers and patients.  

Now, said Dr. Jackie Gerhart, a physician at Epic who works in clinical informatics, the electronic health vendor is seeking to help clinicians use that data to shape patient care on an individualized level.  

Gerhart sat down with Healthcare IT News at HIMSS22 to discuss Epic's wide-ranging patient database, Epic Research findings and how its forthcoming Best Care for My Patient tool can show clinicians others' most effective practices around the country.  

Q. How do you help to shape Epic's offerings as a physician?

A. Our role is to, number one, help with research and development in the software; number two, work on any new futures project, so things like data or any kind of analytics that could be used at the point of care or for future research.   

And then we also work with any of our current healthcare customers to try to make sure that their experience is really good. So, for example, you may have read in different headlines that clinicians don't like their EHR, or you know, "Oh, my God, I have to do it at night."   

Part of our job is to try to decrease that in many different ways from within the software, but also by engaging with the government and by making sure that internally, people's billing processes are as efficient as possible so that we aren't stuck writing more in our notes than we actually have to. So that's kind of the main role of a physician at Epic.

Q. We've been hearing a lot about Epic's Cosmos database, which draws on de-identified patient records from its customers. How can public health officials and clinicians use it to gain population-wide insights? 

A. Cosmos, right now, contains records from 140 million patients and 2.2 billion clinician visits. The goal is, unlike a lot of other healthcare databases, you have to contribute data into it in order to be able to use the data from it – meaning all the different customers that are part of the community can choose to be part of Cosmos. If they do, we take de-identified data from their patients down to what's called a limited data set. And then that limited data set is then used, along with all of the other organizations that are participating.  

We use that data for research, so anybody that's at an organization that contributes to Cosmos can do a query of the data.   

There are three differentiators for Cosmos as a database. Number one, you can search it really fast. The second piece – which the CDC tells us when we work with them – is with regard to how representative it is of the U.S. population. The goal is to try to get as clear and as close to a representation of the U.S. population as we can. We've been doing a lot of work with the CDC through COVID, and particularly working on larger data sets that the CDC may either not have access to right away, or they might not have a representative sample.   

For example, in the last month, one of their MMWRs had to do with breakthrough infections for COVID. They have data from New York Public Health, and they have data from California Public Health. But again, those are two specific populations that don't necessarily represent the entire United States. So we've been partnering together to look through Cosmos. Our research team and our data scientists will design a study along with the CDC, we'll run the data, and then we'll hopefully work together to put it out into a record. That's the third differentiator: the size of the database.

Q. What does the review process of these findings look like?  

A. There are three kinds. The first is internal, and it's a two-team process. So on one team, there's a clinician, a data scientist and usually some kind of public health or research background person. And then there's the same thing on the other team. They both independently design a way to answer a broad question. They meet, they look at each other's methods, they work together to figure out which makes the most sense – then they go back apart and recode everything in the specific method. And then they come together and look at validations for their results. Again, this is more of an accuracy kind of check. 

After that two-team process is done, if it's done internally, then we post it on our EpicResearch.org website. It doesn't necessarily draw any specific causal relationships. The goal is to say, "We investigated this. Here are the preliminary findings, now, everyone else in the community go and study this even more and do some more rigorous statistical analysis."

Option two is: We do collaboration. That's a little bit like what I talked about with the CDC. We've also done it with Kaiser Family Foundation and a few other different organizations like academic centers. In those cases, it can go one of two ways. It can either be similar, where we do an internal review process, or we submit it to a journal, and then the journal does [its] own peer-review process.

We've also had members of the media who want to be validating a certain story. That is sort of the third category, which is not peer-reviewed, it's literally just answering a single question and sending them the data, and then having them and their internal analytics folks do it.   

Q. So these are all population-wide insights. But can you talk to me a bit about Best Care for My Patient, and how that draws on the Epic database?  

A. There's two places that we see this data set going: One is research, which is what I just talked to you about. And then there's Best Care, which is focusing on the one patient that is in front of me in my office as a clinician.  

Let's say that you are a 30-year-old female who is coming in with hypertension, and you also have allergies, and you are on medication for your cholesterol. In this case, I want to find out how to treat your blood pressure. The idea is to try to do personalized and individualized medicine based on observational data from the rest of the community.

So there's 140 million patients in the database. I then put your criteria in – let's say it brings it down to 100,000 patients, and then I decide that I want to also include your hyperlipidemia. So now we're at like maybe 40,000 patients. In all of those 40,000 patients, when their clinician has been faced with that exact question of how to treat hypertension, what did those clinicians use? And what were their outcomes? Best Care for My Patient will display that.  

Q. As a clinician, do you see clinicians having time to do this, given the intense pressure many already face to see as many patients as possible?   

A. How I would address that is: It's one more tool in my tool belt to use if I want it. I might know a patient for five years, 10 years, 15 years, and have an idea of what I want to do to treat them. But there might be a different patient who I am less familiar with, or has other conditions that I just haven't read the literature on – like, what do you do for this condition if they have these three others?   

And so it just gives me a little bit more information. It's meant to be very simple and easy to play with, such that you can use it at the point of care. If you want to dive deeper, then you can explore the data further. But for the normal average everyday clinician, the goal is to have minimal intervention and just present you with the data and then you choose how to use it.  

Q. So this implementation is not live yet, correct? When will this go live? 

A. We're aiming for about a year to a year and a half. 

Kat Jercich is senior editor of Healthcare IT News.
Twitter: @kjercich
Email: kjercich@himss.org
Healthcare IT News is a HIMSS Media publication.