Subscribe
Share
Share
Embed
340B Insight provides members and supporters of 340B Health with timely updates and discussions about the 340B drug pricing program. The podcast helps listeners stay current with and learn more about 340B to help them serve their patients and communities and remain compliant. We publish new episodes twice a month, with news reports and in-depth interviews with leading health care practitioners, policy and legal experts, public policymakers, and our expert staff.
Speaker 1 (00:04):
Welcome to 340B Insight from 340B Health.
Monica Forero (00:12):
Hello from Washington DC and welcome back to 340B Insight, the podcast about the 340B drug pricing program. I'm Monica Forero with 340B Health filling in today for our host, David Glendinning. Our guest for this episode is Elizabeth Gibson, the Enterprise 340B Program Coordinator at West Virginia University Medicine based in Morgantown, West Virginia. Elizabeth's presentation at the most recent 340B Coalition Summer Conference focused on using artificial intelligence to assist with 340B operations and compliance, and we wanted to learn a bit more about what this entails.
(00:46):
Before we get into that conversation, I'd like to acknowledge that this is the 100th episode of 340B Insight. We'd like to thank our loyal listeners for their support and look forward to continuing to bring the latest insights in education about 340B. Now for our feature interview with Elizabeth Gibson. As developments in AI continue to spark curiosity across all industries, we wanted to hear about how one health system is using it to automate aspects of their 340B program. David caught up with Elizabeth in between conference sessions to learn more. Here's that conversation.
David Glendinning (01:20):
I'm here with Elizabeth Gibson, Enterprise 340B Program Coordinator at WVU Medicine based in West Virginia. Elizabeth, welcome to 340B Insight.
Elizabeth Gibson (01:31):
Thank you for having me.
David Glendinning (01:33):
We are here to speak about some artificial intelligence, machine learning in 340B and specifically a bot that WVU Medicine developed. Before we get into that, could you first tell us a little bit about WVU Medicine and the patients you serve?
Elizabeth Gibson (01:51):
There are about 22 hospitals in our system across West Virginia. We also have hospitals in Pennsylvania, Ohio, and Maryland, but the majority of those are in West Virginia. Most are 340B. We have 19 covered entities that are dish critical access hospitals and one rural referral center. We're stretched pretty far across the state. It's a small state, but it has arms that go out into random directions, so we cover a lot of space. Probably one of the few things people know about West Virginia is quite a poor state, so we do have a lot of un and underinsured patients that we're taking care of.
David Glendinning (02:32):
We are, as I said, speaking about artificial intelligence. I will forego the usual references to Skynet and The Terminator, and we'll just get right into asking about this particular use of it. Where did the idea originate for using artificial intelligence in your 340B program?
Elizabeth Gibson (02:51):
The original idea I think came from an outside vendor who works successfully with other departments in our system and it didn't work out. It was just bad timing. We were a very new team. We are an enterprise team and at the beginning there were about seven of us. We're now up to, I think 24, 25. Early on when our system was in expansion mode, we just didn't have the resources to take on something like that. Even a product off the shelf was too much to implement at the time.
(03:29):
Later on, our system had a department for robotics. Robotics Process Automation. It's difficult. We just call them the Bot Team because that is easier, but once they were established, that was brought up again of, "Hey, we know this could be a resource for 340B, would they be able to build something for you?" That's where it took off.
David Glendinning (03:55):
Through that Bot Team, you proceeded to build it basically from the ground up, it sounds like. How did you go about that process? How did you go about implementing the bot as it were?
Elizabeth Gibson (04:06):
I had no idea what I was doing from the beginning, so it was very much steep learning curve for I think everyone involved because 340B is a complex topic. For the Bot Team to learn 340B and for me to learn what they needed from me to make this happen just took a lot of conversations at the beginning. We started with the audit process that our audit analysts already did. We took the template that they use and the Bot Team looked at ways that they could fill that in in a non-manual way.
(04:44):
Instead of having someone go in and look at our EMR, ways we could pull reporting to make that more streamlined. It took us over a year before we got this up and running where it could actually be utilized for audits. In that time we learned a lot and because neither of us really knew exactly what we were doing. There was a lot of trial and error.
David Glendinning (05:08):
Okay. Interesting that you were teaching the Robotics Team a little bit about 340B and they were probably teaching you a little bit about automation. Once you got this up and running to the point where it could be doing some tasks for you, what were some of those first tasks that you assigned to the new system?
Elizabeth Gibson (05:26):
The first thing we really built was our mixed use audit bot, and that involves looking at things like mapping, so NDC mapping in our TPA, checking patient status to validate that what's being sent to the TPA is correct. The other things would be looking at Medicaid claims to see if correct modifiers are applied. Since again, we're spread all over the place, there are going to be different rules for different states. Really, I guess the first task is to go into the TPA and download the audit reports and start that process.
David Glendinning (06:04):
It sounds like it took a pretty prominent role in your internal auditing process. How well did it work when you put the spot into place?
Elizabeth Gibson (06:13):
At first there were errors, there were problems, so it wasn't straight out of the gate fantastic. We've really seen improvement in our overall audit experience, I would say. We are having an increase in number of claims we're able to look at simply because to start with, all of the, I guess, administrative tasks of setting up an audit are taken care of. All the auditors have to do is go in and actually look at the data.
(06:44):
It's also flagging things that are potential problems or likely to be problems. We have an idea of where we need to look. The overall number of claims we're able to audit is higher, and then those are more targeted to what would be potential problems. Previously we were finding a very small percentage of audit findings, so auditors were checking boxes all day saying, "This is fantastic, everything looks good." Now they're looking at things that are more likely to be potential problems and really diving into those.
(07:18):
I think they've learned a lot from it too because they have increased visibility to those things that actually are problems. Instead of being a hypothetical example that we're giving them, they're able to see them. It also allowed us to be more prepared for HRSA audits because we were more aware of the data that we had, what it looked like, potential problems. We had already made corrections and we're very familiar with what we had.
David Glendinning (07:45):
I'm curious about not just the bot itself and how it works, but the way you went about installing it and the process that was required for that. Were there aspects of installing this system that you thought worked very well?
Elizabeth Gibson (08:02):
One of the things I liked the most was being able to work with the RPA or Bot Team. Having that relationship with them has been really beneficial to us overall. We also were able to make quick customizations relatively quick because it was a small team and the way they built this was for iterative changes. As soon as we realized something wasn't working, we just changed it up, tried it again.
(08:33):
It was a very trial and error approach, but it made it so that we could get what we really wanted out of it, and it also allowed us to focus on the build when we could instead of times when instead of saying, "This is the timeline, you have a vendor you have to work with." We were able to say, "Okay, we have a big project right now that we need to put this on hold for a moment," and then move forward with it.
David Glendinning (08:56):
You mentioned before the, I guess we could call them growing pains, with installing this bot. Could you go into a little bit more detail, anything that did not go as smoothly?
Elizabeth Gibson (09:07):
I think it took a lot longer than any of us anticipated. Part of that was because as I said, we took those breaks to work on other projects as needed, but it was very time-consuming. Also, there was a lot of education and communication that had to happen between the two teams, which is a negative, but it was also really beneficial I think in the long run because now we have a better relationship and we've built that foundation that we can move on with further projects in the future.
(09:36):
The product is a bit clunkier than something we probably would've purchased for example, because of the way it was constructed to allow us those customizations and modifications as we needed them. It's not as sleek or as smooth. From my end that doesn't matter because I just get an output file and I have no idea what's happening on the backend. From their perspective, from the Bot Team's perspective, I think it does take a little more time and it's a bit clunkier for them at this point.
David Glendinning (10:11):
It sounds like you were pioneering in a number of ways with this process and installing this bot. Is there anything else you learned from going through that long time before you were able to bring the bot live?
Elizabeth Gibson (10:25):
One thing it did was really open our eyes to potential automation opportunities, so things we would've never thought about. Now, anytime we're doing a task that takes a long time or seems redundant, we're able to look at it and say, "Okay, well, how could we automate this to make it easier for us?" It's not just me or the people who worked directly on building this, but our entire team is now looking for ways to automate their daily processes.
(10:56):
It's been really great to see our audit analysts expand their Excel skills, for example, because they see that they can save time by doing this. I think it's also because our organization doesn't show automation as a way to get rid of people, that it's a way to make you do your job better and for you to have a better balance in your job, that you're not doing menial labor because you're an expert. We have people on our team because they're experts and we want to use their expertise. This has really allowed them to do that, and it's been fantastic to watch them grow.
David Glendinning (11:37):
Eliminating the TDM and automating some of those processes and AI not coming to put people out of their jobs?
Elizabeth Gibson (11:45):
Yes.
David Glendinning (11:47):
Are there plans at WVU Medicine to expand this bot perhaps into other parts of your 340B program?
Elizabeth Gibson (11:56):
Yeah, so we're currently working on expanding, doing a spinoff version of the audit bot to make it look specifically at claims. Medicaid claims are not something that's easy to examine manually. Nobody's going to be able to look at all of those, but we want the bot to be able to look at all of them, not just those that go to our TPA, but also anything in clean and clinic sites.
(12:22):
They're also working on a retail audit bot. It's a little more complex because it's harder to find those encounters and match everything up. I think something with pricing would probably be another thing we would be looking at with making sure that all of our pricing is accurate, but I think we've got them busy. They have other departments to work with for now.
David Glendinning (12:47):
Well, I'm hoping... You've piqued some interest here among other hospitals and health systems, perhaps other covered entities that are looking to bring in more automation, machine learning, AI. For those health systems, what advice might you have for them considering all of these potential improvements to their 340B program?
Elizabeth Gibson (13:07):
I know that not everyone is going to have a Robotics Team that's able to do this, so being able to either find a company that is going to offer this to you, I think is a viable option. I also just hope people think of automation in other terms and that there are other departments probably in your organization where somebody has a skill that you can pick up on and that they would be happy to teach you that could be used for 340B.
(13:40):
I think sometimes we get stuck in our 340B mindset and, "Oh, this is too hard to explain to anybody. I can't tell anybody what this is. Nobody knows. We're like the secret organization. Nobody's ever heard of us before." If you reach out to someone in finance or procurement or any other IT, somebody probably has some automation experience that they can help you. You just need to be able to tell them what you need. That's really the thing we've taken away, is that collaboration and looking for others in our system that can help us.
David Glendinning (14:13):
Elizabeth, thank you for taking on that collaboration and helping to take WVU Medicine into this fascinating new automated future being on that cutting edge. We very much appreciate you sharing this story, so thank you for being here today.
Elizabeth Gibson (14:27):
Thank you for giving me the opportunity.
Monica Forero (14:30):
Our thanks again to Elizabeth Gibson for insight into ways healthcare organizations use AI to improve their 340B operations and compliance efforts. It is inspiring to hear about the network of humans responsible for the technology that helps providers care for their most underserved patients. What tech innovations is your hospital or health system using to support your 340B program? We would love to hear about it. You can contact us at Podcast@340Bhealth.org. We will be back in a few weeks with our next episode. In the meantime, as always, thanks for listening and be well.
Speaker 1 (15:10):
Thanks for listening to 340B Insight. Subscribe and rate us on Apple Podcasts, Google Play, Spotify, or wherever you listen to podcasts. For more information, visit our website at 340Bpodcast.org. You can also follow us on Twitter @340Bhealth and submit a question or idea to the show by emailing us at Podcast@340Bhealth.org.
Speaker 5 (15:34):
Voxtopica.