January 3: Today on TownHall Jake Lancaster, Chief Medical Information Officer at Baptist Memorial Health Care talks with Marc Tobias, MD, Founder & CEO of Phrase Health about connecting clinical decision support to outcomes. First Marc walks through his journey and how it led to the formation of Phrase Health. How do you ensure interventions are actually driving change? How does their CDS connect to outcomes and what does that look like for the customer? What value propositions would he use to sell the ROI to a CFO? What is the future direction of Phrase Health?
Join our next 2 webinars exploring the macro challenges facing healthcare and how technology will play a role in addressing these challenges:
This transcription is provided by artificial intelligence. We believe in technology but understand that even the smartest robots can sometimes get speech recognition wrong.
Today on This Week Health.
it shows you that if you just kind of let people build stuff and you're not tracking it, it's just like the analogy I give is imagine.
your a Doctor and you give an antibiotic to your patient and then just don't follow up. Like, did it help? Are they getting sick? And just like with this intervention with clinicians, we gotta make sure that we're putting these interventions in and ensuring that they're actually driving the change that we want.
Welcome to this week, health Community Town Hall is our show hosted by leaders on the front lines with interviews of people making things happen in healthcare with technology. My name is Bill Russell. I'm creator of this Week Health, A set of channels dedicated to keeping health IT staff current and engaged. For five years now, we've been making podcasts that amplify great thinking to propel Healthcare forward. We wanna thank our show partners for investing in our mission to develop the next generation of health leaders now onto today's.
Hey everybody. I'm Jake Lancaster. I'm an internal medicine physician and the Chief Medical Information Officer for Baptist Memorial Healthcare based outta Memphis, Tennessee. And today I'm excited to have a conversation with Dr. Mark Devas about his company phrase Health. Mark, welcome to the program.
Yeah, thanks for inviting me. Excited to chat with you.
So for the audience, give us just a little bit of your background and how you got to where you are today.
Yeah, so. On the clinical side, my background is as an emergency physician, so still work clinically , in the emergency department.
Also background growing up was in computer science and software development and ultimately got very interested in entrepreneurship and kind of delivering that type of value and got my undergrad degree in economics and. Long story of how I got into medicine, but ultimately, after I was finishing my residency ACGME came up with the Clinical Informatics Board certification and that was a really nice way of marrying my interest in software development.
As well as medicine. So, brought those two things together. Did a clinical informatics fellowship and my research which I'm sure will end up talking about was ultimately spun out as a company. So, kind of marrying a lot of different interests and, experiences to get to, where I am today.
It was great. You know, I had a very similar story except for I didn't have a software background, but the interest in, healthcare technology is what drove me to the fellowships as well. But let's talk a little bit more about your research and what it's become. Take me through that.
Yeah, so I did my informatics fellowship at Chop Children's Hospital of Philadelphia, and I got very interested in clinical decision support and it actually all started with Health 2.0 had a contest that they put on, sponsored by the De Beaumont Foundation about connecting.
The primary care setting with public health. Mm-hmm. , and following in the footsteps of some other great informaticians, ended up kind of creating a system where public health officials could build rules in the cloud and the EHR would kind of query off of them, pull that down so that the EHR builders maintainers weren't responsible for building.
At that time it was Zika building a Zika alert or a local measles mumps outbreak where these things are moving really quickly. So that was where my research started. But when I was talking to folks and, this is where you start getting into what it's like to, run a startup and, how to start small and learn from your users they didn't have trouble with the content.
They had trouble with actually understanding how to make the content work and how to make people adhere to decision. So ultimately started looking more at the performance of decision support, and most recently we got an NIH grant with some of my former co-fellows to actually start connecting the decision support to outcomes.
So when you change a workflow like an alert, how does that impact your process measure, and ultimately, how does that impact the outcome measure that you're looking?
No, it's great. And then, you know, I get asked for new alerts all the time from, could be from revenue cycle, could be from quality or, somebody else.
And very seldom do we follow up and ask. Did it have the intended effect is this actually, working to, you know, whether it's our, we have several, a couple for c diff. Did our c diff rates actually improve? Did our MRSA rates improve with this new alert that we put in place? So that's the sort of I guess reporting that your research led you to uncover.
Yeah, it's really focused on that and actually as part of my learning about the space, we spoke with 20 or 30 health systems and actually collected their intake forms around a BPA request, or could be more of a general decision support request. And some of them were actually really good in, requiring that you define, well, what is the.
Process measure that we're trying to drive. And then we would ask them, are you actually holding these requesters accountable after you build it? And there were actually a couple of places that took a very draconian approach of saying, If you don't report back to our decision support committee in six months and tell us is this actually working, we're just gonna.
Rip it out, and if you do report back and it's not working, we're gonna kind of work with you to figure out, well, what change do we need to make? I'm gonna write that down. We gotta bring that. Yeah. And, and obviously I think that is a very draconian approach, but it shows you that if you just kind of let people build stuff and you're not tracking it, it's just like the analogy I give is imagine.
your a Doctor and you give an antibiotic to your patient and then just don't follow up. Like, did it help? Are they getting sick? Are they getting side effects? And just like with this intervention with clinicians, we gotta make sure that we're putting these interventions in and ensuring that they're actually driving the change that we want.
No, that's perfect. You know, We just had a long, over the last two years, we're able to decrease our PPAs about about 75% percent to all physicians and nurses. And, after diving into each one that was firing a ton, realizing that either one, they're no longer needed or, two, it was not having the intended effect we need to change the behavior.
So it, sounds like y'all are doing some great work in that area and that would've certainly been helpful to know. Without having to do that extensive research ourselves, asking, going about, just being able to look at a report and say, Is this driving a process metric or not?
But you've also done some work with order sets and outcomes as well, right? Can you tell us about that?
Yeah, I was gonna say, I mean, alerts are a great example and I think are really important when you think about clinician inefficiency. When we define the field of clinical decision support that spans the spectrum of alerts, order sets documentation templates.
I mean, there's all different forms of decision support that. Are built. And, and when I think about alerts, I really think about kind of the last ditch effort cuz you're really asking somebody to walk back and kind of make a different decision. Mm-hmm. at a point in their workflow that may not be ideal, where something like an order set, you can really more upstream get the decision making at the point of admission or at the point of driving them down some pathway.
So starting to nudge. People in the right direction, not just with the orderable that are there, but you can embed instructions around your orders to say, Hey, don't order this. Which is one example that I know some of my colleagues at CHOP published around bronchiolitis, where they left off albuterol from the bronchiolitis order set and they saw people were ordering it.
So they put albuterol on the bronchiolitis order set and just had instructions of do not order for bronchiolitis. So, There are different ways of driving decision support with these tools, and at the end of the day, they're all meant to drive some quality improvement effort. So again, looking at the process measure that we're trying to impact by increasing adherence to order sets, I think these are all really important concepts to focus on. 📍th, priorities for:
Yeah, no that bronchiolitis example is, is perfect. Tell us how your company, your product, is able to, I guess, connect those, that CDs to the outcomes. What does that look like for your customers?
Yeah, so, when we rolled out our software, I'D initially intended it to be for informatics teams. I was in an informatics fellowship.
I sat on a clinical decision. Committee and it was very informatics heavy. But when you think about who these interventions are built for, it's invariably for quality and safety in other operational teams. So I really view informatics to quality as a continuum. And we saw a lot of quality folks using our software, which was intriguing.
Hmm. . So as I got more in depth with the quality approach and quality methods we started adopting some of their tooling to look at the end to end spectrum. So as an example, we've adopted the key driver diagram as a self-service approach in our software to say, Well, this piece of decision support is the intervention.
We're gonna select our process measure, and we're gonna select our outcome measure. And connect it in a interactive way. And then what our software does is we kind of look at each one of those elements in isolation, but then each one of those connections we stratify the data to actually look as their correlation.
So as an example, if I think discharge order set is gonna reduce readmissions, if people. When people are using the discharge order set, is our readmission rate lower for a certain population of patients versus when the order set is not used? Is the readmission rate higher or is it lower, or is it.
There are no difference. So clearly we can't say causation, a lot of confounders in this, but if we are gonna try to find these proximal measures that it's impacting, we should see relationships between those measures, between the intervention and the outcome.
No, that's perfect. And just a lot of information that.
I wish on the informatics side we had, but I, think our quality leaders would want to know as well. What can you tell us about outcomes? Do you study that, Have you worked with any of your customers to, find whether, these tools are able to improve, on readmissions, for example, or, quality related options?
So I'll kind of start off by saying in the informatics domain, like we're really pushing the envelope for these teams to start thinking outside of their traditional silo of typically within the, it, under the CIO informatics will do their build work and really thinking about how to push them to think about, well, let's incorporate what our measures or outcomes or processes are gonna be.
Similarly pushing quality on the other end to kind of think About the informatics side and ultimately on the informatics side that data is pretty cut and dry, where when you start thinking about outcomes, there's a lot of variability about how you define an outcome. So are we gonna define, and just as an analogy in the cohort side, if we think about the sepsis cohort, are we gonna define it simply just by.
saying The billing code of sepsis, or are we gonna say phenotypically? The patient had IV antibiotics, they had a low blood pressure, they had an elevated lactate. How you define those things are very different and different organizations are gonna do it differently. So we have just recently started getting into, starting to prove out some of these outcomes and, as it relates to doing this in a more robust, scalable way, we also partnered with Vizient Who just really, really great partners and leaders in this space and are just starting to do pilot work where we're actually connecting the EHR workflow up through their risk adjusted outcome measures and actually seeing, well, we identified this opportunity with, Vizient how do we explain that at the workflow level?
So we're just starting to dive into that in the past couple of months and really excited for, some of the stories that are gonna come out of that partnership.
No, that's, that's perfect. So you're Vizient we work with them as well, will tell us what our opportunities are, and then it's always up to us on the operational end or the informatics end to design the CDs or whatever to help drive that intervention.
But with, phrase, you're able to kind of bridge that gap. We can have a little bit more of a blueprint of what we need to do to make those interventions. Is that what you're saying?
Yeah. So essentially to bubble up. Exactly. If we have this opportunity around a lab utilization is that due to some order set adoption or lack of order set adoption?
Are there alerts that could be explaining some of the variation compared to, your peer organizations? Are there certain documentation templates that are not being adopted that may be explaining some of the quality outcomes that you're getting credit for? These are all these pieces in the workflow that explain the outcomes, but historically have been a big lift for organizations to draw those connections. 📍📍 In:
No, that's perfect. Yeah, and everybody tends to recreate the wheel themselves and build something different. So you mentioned the value to informatics and obviously I see it there who make my job a lot easier. And then quality as well. How do you sell the ROI for the company to a CFO who yes, wants to see readmissions go down, or quality go down.
But a lot of that is, reduction in penalties that are far off as opposed to, more revenue or something coming in that.
Yeah, so there's three core value props that we've identified and we currently work with over 40 hospitals around the country. And as a founder and just kind of the person who considers this their baby since I, I developed it from day one.
Thing I've been most proud of is that we've had no health systems kind of cancel with us and have all renewed. And I think that demonstrates that these relationships are, fruitful and valuable to our partners. And when we really dug into what is the ROI that we're driving, boils down to three things.
Clinical variation. So this is a lot of what we've just been talking about reducing waste in the system. So that's a lot of the. That were I just mentioned with vi. Clinician inefficiency, which I think probably resonates with you as the cmio around how do we get these workflows better so that physicians aren't kind of leaving early or kind of leaving medicine or finding different jobs and the costs associated with.
Kind of recruiting clinicians and staffing more generally in this environment. And then finally, the one that we really focus on with a more direct ROI is analytics step. and what this is referring to is there, there's no organization we've talked to where they just. They have analysts just sitting around twiddling their thumbs with no work to do.
We've heard cues of, data requests going six months even to a year out. I heard at one organization, and when you think about the time that people are sitting on these requests, that could be some cost that is building up for the organization cuz this person can't act to cut this cost or to improve some outcome for the organization.
So every day, That's going by that this clinician can't embark or this operational person can't embark on their improvement work is lost money for the organization and these analysts will stand up these grandiose BI tools they might be used. Once they might go into disrepair when that developer leaves.
There may be duplicative dashboards that are being built over time as different people try to attack similar projects in different places. And these are all just huge. Drags on an organization's analytics team. So what we are not saying that you adopt phrase and you can fire an FTE in your analytics group, but these are really valuable resources that can be reallocated to other initiatives.
So you can be more efficient about tackling some of these data requests. So across all those spectrums, we've come up with ROI analyses of how we drive value to our partner.
No, that makes perfect sense. I work with our analysts all the time on, requests related to what you just, you described, and, and we're not quite up to 12 months or longer, but, six months on some occasions for sure.
Yeah. So tell us about future directions. Where do you take the company from here?
Yeah, so we are really excited about scaling this outcomes work that we're doing and really starting to look at additional data sets. So how do we start working with external partners to, additionally deliver value to health systems?
So this includes, Kind of exploring, working with reference companies. So reference companies have amazing libraries of content that they're kind of working with health systems. But the question comes up is the content that these reference companies have, is it actually delivering the value. And if it's not delivering the value, how do you iterate on that content to make sure it is so kind of looking at partnerships there to use data around these workflows to ensure best practices are being adhered to.
And then additionally data sources like claims data and data from other places to really start marrying this so you can have a more complete picture. Of all the way down to the workflow of what clinicians are doing and understanding the repercussions that go outside of the walls of the hospital and beyond.
That's very nice. Certainly, you know, a different company, a different idea that I've never seen before. So, very excited about your work. Well, Mark, thank you so much for coming on, and thank you everybody for listening. Please join us again next week.
All right, thanks.
I really love this show. I love hearing from the people and the leaders on the front lines. We wanna thank our host who continue to support the community by developing this great content. If you wanna support this Week Health, let someone know about our channels. We have three this week, health conference this week, health Newsroom. And finally, the channel that you just listened to this week, health Community. Check them out today. You can find them wherever you listen to podcasts. Apple, Google Overcast, Spotify. You get the picture. We are everywhere. And you can also subscribe to our ongoing newsletter. It's at this week, health.com. Go ahead and subscribe today. We also want to thank our show partners for investing in our mission to develop the next generation of health leaders. Thanks for listening. That's all for now.