Transcript
Speaker: Hi, welcome to the Health Data Ethics Podcast.
Speaker: I'm your host, Jenny Owens, and I'm here with another wonderful guest.
Speaker: Eric, would you like to introduce yourself?
Speaker: Yes.
Speaker: Hi, Jennifer.
Speaker: Thanks so much for having me.
Speaker: I'm Eric Swanson.
Speaker: I'm a senior vice president at a management consulting firm called Kauffman Hall & Associates.
Speaker: And there, I lead a data science and analytics team where we are focused on helping our clients, providers, and others solve some of the biggest and harriest challenges in healthcare using data science and analytics.
Speaker: So it's wonderful to be speaking with you today.
Speaker: Yes, I am so excited for this conversation because I actually pulled from your February Flash report for a couple of podcast episodes ago, looking at the overall health of the healthcare industry.
Speaker: And you had some comments there that really resonated with me.
Speaker: Can you summarize those for me?
Speaker: Do you mind?
Speaker: Yeah, absolutely.
Speaker: So as part of our work in evaluating the performance of healthcare systems nationwide, it's worth noting that we're beginning to see larger and larger discrepancies between organizations that are doing very well,
Speaker: and those who have stagnated to some degree.
Speaker: And as part of our work, one of the areas we wanted to explore was the use and application of artificial intelligence and data science or advanced analytics and solving those.
Speaker: And what we found was when looking at the high-performing organizations, they had a different set of characteristics than those who had stagnated relative to these spaces.
Speaker: And it was principally around having really well-defined AI and advanced analytics strategies.
Speaker: Contained within those strategies, we found a number of very interesting areas.
Speaker: So number one is that organizations who are deploying advanced analytics in really meaningful ways are doing so by focusing on the business problems first and then potentially finding AI ways in which to solve them rather than having interesting AI solutions by which they are trying to find a problem.
Speaker: And that is a very common thread we see too often.
Speaker: We are in a period where
Speaker: For many technologies, this included, we might be going through a little bit of a hype cycle.
Speaker: And what we find is that those who are using and looking at all of those solutions first, rather than the problems, tend to not be focused in the areas that they ultimately need to and are not seeing the type of impact and outcomes that the higher performing organizations certainly do.
Speaker: So number one, focusing on business problems first.
Speaker: That's the first key point.
Speaker: Second is,
Speaker: early wins and the ability to roll this out is very often focused on some of these more mundane tasks.
Speaker: And the way I often describe this is we think about how healthcare is being delivered.
Speaker: Not only do we have to work on the improvement and performance of the healthcare itself, but we also have to work on that science of delivery.
Speaker: And so this is why when I talk about it, I make a distinction between the science of healthcare delivery and the science of healthcare.
Speaker: And what you will find in the news generally, broader media, is we will hear a lot about the science of healthcare, artificial intelligence for reading mammograms, all kinds of predictive work.
Speaker: Sepsis is another very common area in which we think about it.
Speaker: How do we think about certain clinical pathways that someone needs to follow?
Speaker: And these are all really exciting, interesting areas of work, areas in which research ultimately needs to continue.
Speaker: But that science of healthcare delivery is where we can begin to see some early wins in organizations that really begin to impact them operationally, either from a financial side, an operational side, or in a patient outcome and experience side.
Speaker: So advancing that triple aim, quadruple aim, quintuple aim, whatever those organizations are focused on.
Speaker: So that's the second one.
Speaker: So focusing some of the early wins on the science of healthcare delivery.
Speaker: So that's number two.
Speaker: Third, and this one is so incredibly important, but there has to be a recognition that given some of the newness and even some of the complexity of which these types of advanced analytic models work under is that the adoption is only going to occur at the speed at which the stakeholders and others using these tools will ultimately trust them.
Speaker: And so I talk about this as being rolled out and adopted at the speed of trust.
Speaker: Why is this important?
Speaker: Well, this is really important because there are areas in which we may even find very interesting neural nets, as an example, that do really great predictions.
Speaker: But if we cannot explain why those predictions are occurring, the adoption and the ability for stakeholders to actually appreciate what they're doing becomes very, very challenging, particularly in a highly risked
Speaker: averse type industry.
Speaker: So this means that what we will likely see for speed of trust is, in many instances, simpler algorithms, those in which there is explainable AI type solutions, and importantly, always having a human somewhere in the loop here, such that these are not running fully autonomously.
Speaker: And then last, what I would just say, the very final point here is relative to accomplishing all that, there has to be an appropriate set of infrastructure.
Speaker: And this infrastructure is important not only for data security, but also data governance, ensuring decisions are being made across the right type of business problems, and last but not least, ensuring that we can track and manage the performance of these types of tools.
Speaker: So those are some of the key findings in talking with our clients about the rollout of this that's
Speaker: driving some of the higher performers and really understanding and accomplishing some value from their efforts.
Speaker: Yeah, absolutely.
Speaker: So you've just given us a lot of content to chew on.
Speaker: I want to call out a couple of things.
Speaker: One is going back to your moving at the speed of trust, the explainability of AI and having a human in the loop were two of the items that were called out on the joint AI Bill of Rights that the National Institute of Standards and Technology and the White House put forward.
Speaker: Oh my gosh, was it last year or was it 2022?
Speaker: I honestly cannot remember.
Speaker: I think that was last year.
Speaker: Yeah.
Speaker: So, so last year.
Speaker: So this is something that I think a lot of our bigger picture AI frameworks are starting to wrestle with.
Speaker: What does this really mean?
Speaker: You know, once you've implemented this technology.
Speaker: But I wanted to take you back for a second because you mentioned that the use of artificial intelligence and enhanced analytics was a real differentiator between health systems that were really financially healthy and were performing really well and health systems that were less financially healthy and were struggling with their margins.
Speaker: And as I was reading, I found myself kind of wondering, is this a chicken and the egg situation?
Speaker: And I'm choosing my words very carefully here, right?
Speaker: Is this a leading indicator that this system is financially healthy, that they have resources
Speaker: and capacity to be focusing on these things?
Speaker: Or is it that the AI and the analytics are actually bringing some value here?
Speaker: Or is it both, right?
Speaker: Is it chicken and the egg rather than chicken or the egg?
Speaker: Yeah, well, to continue that analogy, I think it is both the chicken and egg in this instance.
Speaker: And it's really interesting that you're asking these questions.
Speaker: And I think, again, very apropos for this discussion today.
Speaker: So number one, what I'll note is relative to that defining of some of those business problems and finding ways in which to apply AI to them.
Speaker: These are areas in which we are finding very clear correlation with the adoption and rollout of AI actually driving financial performance.
Speaker: And again, if we think about some of these these broad business problems that exist today,
Speaker: These are things like managing the workforce more effectively, doing better resource planning, thinking around how do we optimize certain objectives and outcomes relative to multivariate type analysis.
Speaker: So financial planning is a good example of this or service line analysis.
Speaker: These things all very much can lead to improved outcomes.
Speaker: Again, financially is the way we looked at it in our
Speaker: a flash report, but very often operationally and patient outcomes can improve as well.
Speaker: It is also worth noting though, that organizations that are very healthy, many of them have already started down this pathway and are exploring not only those types of solutions that I just talked about that are driving bottom line performance,
Speaker: but also doing more exploration and maybe even that science of healthcare and more innovative areas in which they have the capital to be able to deploy and invest in these types of resources, enhanced digital health, other types of programs along those lines.
Speaker: So we see it from both angles in that it can be done in judicious ways such that it is driving performance for smaller organizations.
Speaker: And those, by the way, they don't even need to build all these competencies in-house.
Speaker: There are ways that some of these competencies can be procured from vendors.
Speaker: But then for some of the larger organizations, really academic areas too, there's a lot of focus on just the advancement of this research and broader, more interesting ways that their strength allows them to be able to do.
Speaker: So you mentioned focusing your efforts on the science of health care delivery.
Speaker: Can you give me one or two examples of areas where you might suggest a health system focus if they're thinking, OK, this is a place where I might like to invest.
Speaker: I might like to actually use this to drive some efficiencies.
Speaker: Are there places you would suggest that people start thinking about if I'm a healthcare executive, if I'm an IT employee, and let's say because you do have to be careful about your term on this, right?
Speaker: Am I looking for something to pay off in the next few weeks or in the next like 10 years, let's say short to medium term, six to 12 months.
Speaker: Yep, absolutely.
Speaker: So again, if we look at the major challenges that organizations are facing and the areas in which there is a lot of controllable ability to influence that, a lot of this right now is around labor.
Speaker: And so we think about persistent labor shortages that exist primarily on the caregiver side.
Speaker: We've had a resetting of expectations of where those labor rates may be.
Speaker: And for many organizations still challenged with high levels of contract labor,
Speaker: usage, which is very, very expensive.
Speaker: These types of problems broadly fall into employee scheduling type problems.
Speaker: So how do you effectively deploy your resources?
Speaker: And again, I'll be very careful here in noting that when we talk about labor optimization, so using predictive and prescriptive analytics for labor optimization, this does not simply mean just cutting.
Speaker: This is really about effective, efficient deployment of those staff.
Speaker: So ways in which we see this manifest itself.
Speaker: One is appropriately right-sized float pools.
Speaker: So how do you build a float pool by which you have resources that can be shared across an organization very effectively such that that can cut back on some of the amount of overtime or potential contract labor that's being used?
Speaker: And in addition, just understanding relative to the ebbs and flows of volume,
Speaker: throughout the course of the day, a week, a month, a year.
Speaker: How do we build plans that best meet those coverage needs?
Speaker: And so that is at least one example where we're seeing tremendous amount of benefit, typically in the form of savings, but we also see improved employee satisfaction, improved retention, patient falls tends to decrease because you have more periods of appropriate coverage of the patients and burnout tends to decline.
Speaker: So a lot of really interesting indicators when that's done well, which you wouldn't otherwise get if you just went to try to cut labor, try to cut costs.
Speaker: And so it's the way in which that can be accomplished.
Speaker: There's one very good example.
Speaker: That's a terrific example.
Speaker: Thank you so much.
Speaker: Eric, I cannot thank you enough for taking time out to chat with me today.
Speaker: Before we close, are there any questions you want to answer or any other information that you want to convey that I didn't ask you about today?
Speaker: No, I think you've really covered it.
Speaker: And again, where I just keep going back to is focus on business problems first.
Speaker: That is the way in which to really drive much of this forward.
Speaker: And don't be afraid if those business problems are a little mundane or boring, perhaps, because those are often some of the areas in which those quick wins and ability to accomplish some real improvement here can occur and so
Speaker: You know, if I had one through line with every client that I speak with, it's about focusing those efforts from that direction rather than a purely technology driven perspective, trying to find business problems to solve.
Speaker: And so that's really my few key takeaways there.
Speaker: I love it.
Speaker: As much as I love a cool technology, and I do deeply love a cool technology, finding a good business problem and finding a couple of good indicators.
Speaker: What are we going to measure?
Speaker: How are we going to know that it's better?
Speaker: It seems to be like the best place to start.
Speaker: Eric, thank you so much for coming on the podcast today.
Speaker: This has been so, so enjoyable.
Speaker: I really hope that you will come back soon.
Speaker: Well, you're very welcome.
Speaker: Thank you so much for having me.
Speaker: And I hope to join again.
