There’s a lot of excitement around predictive analytics, and for good reason. Organizations like University of Mississippi MC are seeing positive results in areas like reducing pressure ulcers. But, as with any advance in technology, there are hurdles that must be overcome, most notably resistance from clinicians about how workflow will be affected. In this interview, John Showalter, MD, talks about his team’s approach to change management and how they’re working to quell clinician fears. He also discusses the pros and cons of going big-bang, what he believes will be the next wave of predictive analytics, what it was like to go from the “well-defined” CMIO role to the more nebulous CHIO role, and his advice on how to communicate more effectively with physician leaders.
Chapter 1
- UMMC’s 6-hospital system
- Defining the CHIO role
- Pros and cons of going big bang — “It’s a lot of focus on alignment & getting clinicians on board.”
- Leveraging predictive analytics to reduce pressure ulcers
- Integrating analytics into workflow
- Clinician concerns — “There’s some skepticism”
- Best practices: “Start with an agreed-upon need.”
LISTEN NOW USING THE PLAYER BELOW OR CLICK HERE TO SUBSCRIBE TO OUR iTUNES PODCAST FEED
Podcast: Play in new window | Download (Duration: 11:49 — 10.8MB)
Subscribe: Apple Podcasts | Spotify | Android | Pandora | iHeartRadio | Podchaser | Podcast Index | Email | TuneIn | RSS
Bold Statements
It’s a lot of focus on alignment and getting the clinicians all on board that this is going to be good for the patients at the end of the day, and rallying around the transformation that you’re trying to make.
We really felt that they needed to get pulled back into the clinical workflow, so we worked with Epic and Jvion to integrate the two systems so that we would be able to actually put the risk into clinicians workflows and adjust care based upon what the analytics were telling us.
There’s great concern from the clinicians. There’s concern about whether or not they are going to need to do extra work to gather additional information for the analytics. There is concern about the validity of the predictions. There’s some skepticism towards how accurate they are, and then there’s always the fear of change.
We clearly communicated that it was an addition; that we weren’t asking them to replace anything else. We were just giving them another tool in their arsenal to combat this problem. So coalescing the team around what’s best for the patient and what’s in the greatest interest of care is the first step.
Gamble: Hi John, let’s get things started with an overview of University of Mississippi Medical Center — what you have in terms of hospitals, physician practices, the schools, things like that.
Showalter: UMMC is an academic medical center. We have six hospitals with just over 900 licensed beds, 100 clinic locations and five schools on campus: a school of pharmacy, school of nursing, a medical school and a school of health related professions, and a school of dentistry.
Gamble: And you’re located at the University of Mississippi. Are you affiliated with the university?
Showalter: Yes, the main campus is in Oxford, Mississippi and the medical center campus is in Jackson.
Gamble: And you are the Chief Health Information Officer there, and how long have you had that role?
Showalter: I’ve been the Chief Health Information Officer for a little over three years now.
Gamble: And in total, how long have you been with the organization?
Showalter: Almost five and a half years.
Gamble: Can you give of a brief overview of what the CHIO role entails and what are your main focuses in that role?
Showalter: The Chief Health Information Officer role is really about getting a return on investment from the data assets that are created across the healthcare system. So whether it’s your EHR, your accounting system, or your HR system, it’s about bringing in that data together to drive change and drive a return on investment by improving care and improving the revenue cycle of the institution.
Gamble: Who do you report to in your role?
Showalter: I report to the Associate Vice Chancellor of Research.
Gamble: And in terms of the EHR, you’re using Epic?
Showalter: Yes, we use Epic.
Gamble: And just to get some background, you were the physician executive when the medical center did the big bang Epic roll out. How long ago did that take place?
Showalter: It was just over four years ago that we went live with Epic. We brought five hospitals live, 90 clinic locations, and about 10,000 employees, all in one day. We implemented 23 Epic applications.
Gamble: And you were a physician executive during this?
Showalter: Yeah, at that time I was the Chief Medical Information Officer.
Gamble: We’ve spoken to CIOs about what it was like to do a big bang, but this is a bit of a different perspective. Can you just give some thoughts on what that experience was like to do it all at once, and whether it was something that you would have done again or if you would have approached it differently?
Showalter: Actually, I would definitely do the implementation the way we did it again. Everything has its pros and cons. One of the big cons with the big-bang roll out is that you can’t get the specially-specific training that you want for the nurses and physicians, but the big pro is that your system is consistent, and you have one platform to use and train on immediately. I think it worked out really well for us and we made the transition well.
One of the things that made it a little easier for us is that we were mostly going from paper. We went from 95 percent paper to 95 percent electronic when we went live, and I think that’s actually a little more straightforward than going from 70 to 80 percent electronic to 95 percent electronic. It’s a lot of focus on alignment and getting the clinicians — the physicians, nurses, respiratory therapists, and allied healthcare — all on board that this is going to be good for the patients at the end of the day, and rallying around the transformation that you’re trying to make in quality. Because a lot of parts are moving and ultimately the implementation is not going to be perfect and people are going to have to work together while you get it to be in a better place.
Gamble: It’s kind of surprising that the organization had still been on paper at that point.
Showalter: Yes. I’m not sure if you’re familiar with the HIMSS Analytics EHR adoption model, but we went from a Level 2 to a Level 6 in one day.
Gamble: Wow. That’s a pretty big jump. So after that point, when the smoke cleared, what were the first priorities?
Showalter: The priorities really deviate into two, one of which is usability and efficiency, and that’s really the role of the Chief Medical Information Officer and the Chief Nursing Information Officer to focus on application usability, acceptance, and efficiency. And then a parallel priority is leveraging the data that’s created from the system to actually improve care and improve processes, and that’s where my role splits. About a year after the dust cleared, I transitioned to the Chief Health Information Officer role, and began to focus on how we were going to use the data that was being created from Epic to improve our system.
Gamble: And that’s where things can get really interesting, where you get into analytics and, as you said, seeing what this data can do. Can you talk about the predictive analytics work your organization is doing?
Showalter: We are definitely getting into the predictive analytics arena, trying to identify our patients that are highest risk for poor outcomes where we need to intervene. Our best example so far has been pressure ulcers. We were using the Braden Score, and then transitioned to using a deep machine learning risk score that significantly outperforms the Braden Score and allowed us to target those high risk patients we were missing. We saw a 66 percent reduction in stage 3 and stage 4 pressure ulcers, which are the worst ones. We’re expanding that to readmissions, and we should begin incorporating the predicted readmission risk into our workflows next month.
Gamble: That’s a really practical application for it. Is that something that you think could be applied by other organizations as well?
Showalter: I do. One of the things that we did early on working with our vendor was to not have the analytics exist on its own. We really felt that they needed to get pulled back into the clinical workflow, so we worked with Epic and the other vendor, Jvion, to integrate the two systems so that we would be able to actually put the risk into clinicians workflows and adjust care based upon what the analytics were telling us.
Gamble: You mentioned workflow. When you talk about the biggest challenges with predictive analytics, I imagine one of them is the fear or maybe hesitation by some of the caregivers about how their workflow will be affected by it. Does that make sense?
Showalter: It does. There’s great concern from the clinicians. There’s concern about whether or not they are going to need to do extra work to gather additional information for the analytics. There is concern about the validity of the predictions. Since the deep machine learning modules are not published in peer reviewed journals, there aren’t typically large studies about their effectiveness. There’s some skepticism towards how accurate they are, and then there’s just always the fear of change, with ‘have I not been doing this correctly in the past,’ or ‘what’s this going to change in my world’ — those types of things.
Gamble: How have you and the other leaders worked to address some of those challenges?
Showalter: As far as workflow, we have been making our predictions with available data so we aren’t requiring the nurses to do any additional work to gather information. We’re just using the data that’s produced in their regular workflow. We have done validation studies of the analytics so we can say this is how accurate they are on our patient population. We did the analysis, and then we followed the patient to see whether or not it was predicting outcomes accurately, and then we’re doing a lot of education about where to go with the next steps — for example, if you see someone’s at a high risk, this is what you should do. And we’re trying to engage our nurses and care teams right at the ground level and right at the point of care to make the best decisions for the patients.
Gamble: So change management definitely comes into play and just really educating and talking through how the fact that this is a different way of doing things.
Showalter: Yes.
Gamble: Coming from your background, are there any best practices you can share as far when something is being introduced like deep machine learning that is going to be a big change, how to approach that with clinicians?
Showalter: I think one of the best things you can do is start with an agreed-upon need. For us, pressure ulcers was really straightforward. Nobody on any care team wants a patient to get a pressure ulcer, so putting a new tool out there to help prevent them was easily accepted. We clearly communicated that it was an addition; that we weren’t asking them to replace anything else. We were just giving them another tool in their arsenal to combat this problem that they’re already trying to come back. So coalescing the team around what’s best for the patient and what’s in the greatest interest of care is the first step. And then I haven’t always followed that myself so I’ve had missteps where we did an analysis because it was cool or we thought it would be very cutting edge, and those are harder to get care implemented around because you don’t have the team that’s engaged with actually coming up with the treatment of plan once somebody’s high risk.
Share Your Thoughts
You must be logged in to post a comment.