Inside Alvarez Business Podcast
Inside Alvarez Business is a podcast produced by the Carlos Alvarez College of Business at the University of Texas at San Antonio. It is dedicated to bringing you stories of our faculty, the real-world impact of their research and what led them to study these important topics.
Inside Alvarez Business Podcast
Using Data to Make Better Decisions
Use Left/Right to seek, Home/End to jump to start or end. Hold shift to jump forward or backward.
How does an overtaxed healthcare system deal with increasing demands and staffing shortages that disrupt the delivery of healthcare? In this episode of Inside Alvarez Business we’ll explore ways to optimize healthcare delivery systems to ultimately improve patient health.
Katherine Adams, a postdoctoral fellow in operations and analytics in the Alvarez College of Business, specializes in the operations of healthcare. An Alvarez research fellow, she completed her PhD in industrial and systems engineering from the University of Wisconsin. Quickly making an impact in the field, most recently she has published a paper in Operations Research.
Hear how data and operations research can be used to help companies make better decisions.
Stay connected with the UT San Antonio Carlos Alvarez College of Business to learn more about how we are empowering the next generation of business thinkers. Follow us on social media or visit us online at business.utsa.edu
So I always wanted to make an impact, you know, work on something that felt purposeful. And healthcare is, I think, a pretty natural decision for that. Why are people dying from something preventable and treatable?
Jonathon Halbesleben, PhDHow does an increasingly overtaxed health system deal with demands and staffing shortages without disrupting the delivery of care? We explore this issue and more in this episode of Inside Alvarez Business, a podcast dedicated to bringing you the stories behind the research. I'm your host, Jonathan Halbesleben, Dean of the Carlos Alvarez College of Business at The University of Texas at San Antonio. Our guest today is Katherine Adams, a postdoctoral fellow in operations and analytics, and an Alvarez Research Fellow in the Alvarez College of Business. Her research explores ways to optimize healthcare delivery systems to improve health outcomes. We had a great conversation about how we can merge concepts from operations research and machine learning to improve decision making. Hope you enjoy this discussion as much as I did. Katherine, welcome to Inside Alvarez Business. We're delighted to have you here, and I'm excited to hear the stories behind your research.
Adams, PhDThank you so much for having me.
Halbesleben, PhDOkay, well, we'll start before we dig into your research. I want to kind of get a little bit of background on you. So I'm always interested, I ask everybody the same question for the first question, why people became faculty members or how they became interested in this. Because as I tell people, I've never met a kid who said they wanted to be an operations professor when they were growing growing up. And so I'm curious for you, what what inspired you to kind of take this career path?
Adams, PhDYeah, so I would say from an early age in school, I really liked math. Surprise, surprise. And I was kind of an informal tutor for my friends who struggled with math a lot. So before exams, all of that, we would study together. I would explain things to them, go over problems. And you won't find this on my CV, at least not on the education section, but I did go to college in Brazil for two years before moving to the US. And while I was there, I worked on a volunteer project where I was teaching high school students from low-income, you know, families mostly, were trying to prepare to get into college in Brazil. So Brazil has tuition-free public colleges.
Halbesleben, PhDOkay.
Adams, PhDBut they're very competitive. And these students, they didn't go to the best schools, you know, for their education, which made them, made it harder to be able to get into those universities.
Halbesleben, PhDYeah.
Adams, PhDAnd they don't consider your educational history, it's only the entrance exam that matters.
Halbesleben, PhDOh, okay.
Adams, PhDYeah. So it's a lot of pressure on a single exam, a single day.
Halbesleben, PhDYeah.
Adams, PhDSo when I was a volunteer for a year and a half, I taught math and physics, and it was an amazing experience that showed me that while it was really hard for me because I was a pretty shy kid, I really enjoyed teaching them. They were great students who inspired me, helped me, you know, end up here in the US.
Halbesleben, PhDOkay.
Adams, PhDUh yeah, and studying here.
Halbesleben, PhDSo that's really cool. Uh if you don't mind me asking, what part of Brazil are you from?
Adams, PhDSo, I'm from the metropolitan area of São Paulo.
Halbesleben, PhDOkay. Very good. Very good. I've been there a couple of times. So I've enjoyed my visits there.
Adams, PhDGreat.
Halbesleben, PhDSo, kind of taking the next step in terms of your career, what got you thinking specifically about some of the topics that you study? So what like what got you interested in healthcare and some of the sort of operations types of issues that you you study?
Adams, PhDYeah, so I always wanted to make an impact, you know, work on something that felt purposeful. And healthcare is, I think, a pretty natural decision for that. Meanwhile, there's this optimization challenge because you have, you know, very limited resources. Even in rich countries like the US, it's still a challenge, especially now with the aging population. You have an increase in demand for services. Meanwhile, the supply is not keeping up. You could even argue it's going down in terms of the workforce, people leaving the healthcare workforce due to burnout, or just people retiring. So it's a very challenging problem, you know, operationally that has a human element to it. So I think that's what attracted me to the field. And everyone needs healthcare services at some point in their lives. So it's a pretty universal impact in that way.
Halbesleben, PhDA lot of your work seems to be focused on sort of lower socioeconomic status countries. Or, you know, I don't - I hate the term "less developed countries", but like, you know, countries that maybe don't have quite the same healthcare system that the US has. Has that been inspired by your previous experiences growing up in Brazil, that like that kind of focus, or has that been more just a coincidence of your work?
Adams, PhDYeah, I would say so. I'd say growing up in Brazil had an impact, you know, it made me - those inequalities - you see them day- to- day in your life. And I guess, I felt passionate that while there are problems such as diabetes, right, that I've worked on, where it's gonna have negative consequences for someone living in the US, for example.
Halbesleben, PhDYeah.
Adams, PhDRight. It's hard to deal with, but there are ways to manage it. There are resources available. Whereas in a developing country, it could be a death sentence. Right. So that's something that really bothered me. Like, why are people dying from something preventable and treatable?
Halbesleben, PhDYeah, yeah. Well, let's let's talk about that. That was actually the next thing I wanted to talk about. That's a perfect segue. So kind of getting into some of your specific papers that you've worked on. And I'm curious to kind of hear what inspired them and the impact you were hoping to make with them. So let's start with that operations research paper I mentioned earlier. That's specifically about diabetes care. What problem were you trying to solve with that particular paper?
Adams, PhDYeah, so we were trying to use community care to be able to screen and treat patients with diabetes. Why community care? Because you don't have enough hospitals, enough healthcare workers, so you need to be able to do something that doesn't require, you know, so many years of training, for example.
Halbesleben, PhDRight.
Adams, PhDSo you can train community health workers to serve their local communities. And in the case of our intervention, they're just going door to door. They're in very densely populated areas, so they can easily walk from one home to the next to screen, you know, and treat patients who have enrolled in treatment.
Halbesleben, PhDOkay. What kind of screening? They just gotta do standard screenings that just write, you know, face-to-face, you know, kind of a checklist of kind of screening types of things, or are they doing lab work or what type of screenings are they doing?
Adams, PhDYeah, so they they're not nurses, they're not doctors, right? So they cannot draw blood and take it to a lab or something.
Halbesleben, PhDOkay.
Adams, PhDSo they're taking your measurements, you know, and they can measure, for example, your waist, your weight. They can do with a little prick of the finger, measure the blood glucose. That is a noisier measurement, so it's gonna vary more day to day than something such as HbA1c, which is the standard measurement used, you know, to diagnose diabetes.
Halbesleben, PhDOkay. So the the issue you were trying to work with is figuring out how best to implement that intervention, like in the most effective way to impact the most people, presumably.
Adams, PhDYes, exactly. With limited resources.
Halbesleben, PhDOkay. So you developed this model based on data from India that you found where you could reduce blood glucose by about 25%, which is impressive. I mean that's, that's terrific. Can you kind of talk through in terms of the - like terms that an audience can follow? How do you, how do you construct one of these models? Because if you look at the papers, they're pretty, pretty overwhelming. But how does one go about kind of figuring out these models and putting things together to say if now if we just do it this way, we're gonna have the best outcome?
Adams, PhDYeah, that can be a misleading thing about a paper. The way it's written, it makes it very obvious. Oh yeah, this is the right framework to solve it.
Halbesleben, PhDOkay.
Adams, PhDThat's not clear at first, right? Especially when you're trying to do something truly innovative that hasn't been done before. It can take a while to get on the right path.
Halbesleben, PhDOkay.
Adams, PhDSo that was the case for us. It took us one to two years to truly pick the right methodology. We tried a few different ones before settling on approximate dynamic programming. But I guess one thing that I've that I learned while working on that paper, which was my first you know, paper from my PhD, is that you just have to break down the problem into smaller pieces.
Halbesleben, PhDOkay.
Adams, PhDSo while you're trying to solve the problem from the provider's perspective, on what is the best way to, you know, schedule these visits? Which patients should you assign them to? In what order, what should be the interval between consecutive visits for each patient to have all these decisions. So we first had to look at the patient because I think this is where the human element comes in when you have a behavioral intervention. You have to consider the element of, you know, people. How do they make their decisions to enroll in an intervention or not? And I would think about this a lot and think, okay, even let's say takeout, right? People know that it's not always healthy to get takeout, it's not cheap either. Yet a lot of people we just do it, right?
Halbesleben, PhDRight.
Adams, PhDAlmost on a daily basis. Even why is that? Because there are trade-offs that we are always considering, right? We have to take into consideration the day-to-day context of each patient. There, they have they're overwhelmed with work, with their family obligations, they're trying to learn all these things about diabetes at the same time. It can be overwhelming, right? So that's what we were trying to look at. What are the trade-offs the patient is examining when deciding to enroll or not enroll in treatment, and then to stay enrolled if they have already enrolled. And for the provider, going when we went back to the provider perspective, we looked at a single patient problem first.
Halbesleben, PhDOkay.
Adams, PhDObviously, that's not where you're trying to solve. You're trying to solve it for an entire community, right?
Halbesleben, PhDRight, right.
Adams, PhDBut it does provide insights into how you can, you know, tackle the problem you actually want to solve, which is planning visits for this whole community, like an urban slum.
Halbesleben, PhDOkay. So you so you take the the different aspects of the patient, you take the different like aspects of the provider and the the kind of services they would provide to the patient, and then you kind of from there you can kind of put these things together. And I know the, the goal of your models is to kind of optimize things. What does it mean to optimize in that kind of context?
Adams, PhDThat's another tricky thing about healthcare, you know, picking the right metric to optimize is gonna depend a lot on the context. So in this case, we picked glycemic control. So trying to keep that blood glucose, you know, not too high, which tends to be the problem with type 2 diabetes, is that it goes way too high. So we wanted to keep it below a threshold.
Halbesleben, PhDYeah. Perhaps because of the takeout thing that you mentioned earlier. That's certainly not helping matters, but anyway. So, you know, it was interesting. You kind of hinted at this a moment ago, but one of the themes that I noticed across several of your studies is personalization of treatment and trying to trying to get set up the treatment so that it is ideal for that particular patient, which makes sense that that would work better than just a one-size-fits-all approach. And you focus a lot on receptivity of treatment. Are they willing to accept treatment? What advantage does that focus bring to your models, like as you think about the -that human element?
Adams, PhDYes, that's a great question. I think the key thing to keep in mind is that you could visit and enroll someone in treatment and they could drop out right away. So that, you know, we don't want to say it was a wasted resource, because it could still impact the patient at some point, but there might be another patient who you could have visited instead on that period who would have actually, you know, been impacted and would have made a difference for them. And maybe it's just the intervention wasn't the right one for that patient. But you, when you have such limited resources, you do want to allocate them as efficiently as possible. So that is why unfortunately that requires sometimes trading off the needs of different patients and measuring what would be the effect on this patient if I visit them versus if I didn't visit them, what is that net difference?
Halbesleben, PhDYeah. How do you, how do you assess the receptivity to treatment in a case like that? I mean, is there is it just a question like, are you interested in treatment? Are you you willing to follow up, or how do you how do you assess that?
Adams, PhDSo we had to work with the data that was available to us.
Halbesleben, PhDRight, right.
Adams, PhDYeah, which was a key driver in choosing the the modeling approach, which is something I think healthcare providers don't always understand the importance of data. It really guides a lot of our modeling decisions. But basically, all we had was some historical data of fasting blood glucose measurements and enrollment. Were they enrolled at this period? Yes or no.
Halbesleben, PhDOh, okay.
Adams, PhDSo you could see, okay, we visited these patients, let's say, many times in a row, and then they dropped out. Okay, maybe we visited them too often and we overwhelmed them.
Halbesleben, PhDOkay.
Adams, PhDOr on the contrary, maybe the spacing between the visits is too long, and so they didn't see any actual benefit, and they dropped out. So it's about trying to find the sweet spot for the patient if they are you know open to this intervention.
Halbesleben, PhDOkay. So you you have to kind of infer receptivity from based on some of the behavioral patterns then, as opposed to just straight up asking them, which I'm not sure how as a psychologist, I'm not sure how reliable that would be anyway, if we just asked them, sure, I'll go to treatment. So, kind of elaborating on that point a little bit, that seems like, and you actually I noticed in one of your - the start of one of your papers, you mentioned that human behavior is kind of a central challenge to developing the kinds of optimization models that you you talk about. And again, as a psychologist, I kind of understand because I mean that's what I try to study is why they do different things when they do them.
Adams, PhDYeah.
Halbesleben, PhDHow, like given you have the constraints inherent in the data that you're working with, how do you kind of factor in the kind of human element to these models? Is it just - is that sort of noise and ends up kind of being part of the, the you know imperfection of the model? Or is there some other different techniques you can use to kind of model the human noise, I guess?
Adams, PhDYeah, there are different techniques. I wouldn't say I'm an expert. We just found one approach, you know, from the literature from Bandura, I think was the author.
Halbesleben, PhDOkay.
Adams, PhDSo that's what we use to build our models and this kind of scale, if you will, of the pros and cons of being enrolled in treatment. We're assuming the patient is making their decision based on a utility maximizing framework.
Halbesleben, PhDYeah.
Adams, PhDSo trying to maximize, you know, by measuring the pros and cons. Okay, if I enroll, am I gonna be better off than if I didn't enroll? And that's how they decide, is what we assume.
Halbesleben, PhDOkay. So, you know, I've had a couple of colleagues of yours on the podcast, but folks you've worked with or that are doing similar kind of work. And there's a question I've always wanted to ask and I'm gonna ask you.
Adams, PhDOkay. (laughter)
Halbesleben, PhDBecause I've always kind of wondered. So I'm reading your paper, in the papers, and there's these, you know, really complex equations. There's these beautifully colored graphs, okay, in the papers. And even like really interesting terms like combinatorial bandits with recovery and habituation, and which you shorten to Cobra. How do you translate that stuff in a way that then gets like health systems or healthcare providers to adopt the models that you're talking about? Because that, that seems like a like fundamental challenge there, right? You, you come up with this wonderful model that for your colleagues, you can describe it with great precision, but for the, the doctor or the like the healthcare administrator reading it, like, what am I looking at here? How do you translate to that in order to convince them that this is the model? This is how you need to do this. Like for the the work you've done with anesthesiologists and scheduling, I think that you know, at some point you have to convince them, do it this way.
Adams, PhDYeah.
Halbesleben, PhDHow do you do that?
Adams, PhDOh, if you'll find out, let me know.
Halbesleben, PhDMaybe you need to partner with the psychologist, right?
Adams, PhDI think so. We need to. No, but in all seriousness, I think coming from an engineering department, that's something that I'm still working on. Okay. Myself personally, it's a challenge. That's one thing that I like about business schools, is they try to keep it everything, even in classes, try to keep it very applied with very clear examples, you know, of why this matters, why these tools are useful. So that's something that I have been trying to do more of, both in research and teaching. Yeah, I think oftentimes it comes to some analogies or small examples that you can use to help illustrate, like this takeout example.
Halbesleben, PhDYeah.
Adams, PhDIt's a simple example, but everyone gets it, right?
Halbesleben, PhDRight.
Adams, PhDWhy at the end of the day, even if you rationally know what's the best decision for your health and your budget or whatever your goal is, it's not easy to implement it. So I think just having relatable examples like that is helpful. But sometimes they're just skeptical, you know, even with this anesthesiologist's work. Some of them were skeptical at first, and then we showed them some results, you know, using historical data of how they made anesthesiologist assignments previously, and a half of data. And we were able to use machine learning to show, hey, these are how the decisions are being made. And then they were surprised. They're like, oh, it seems to actually have done quite well, your model.
Halbesleben, PhDYeah.
Adams, PhDYou know, so sometimes you just have to have some kind of preliminary results to kind of help with that skepticism that they have at first.
Halbesleben, PhDYeah. I've noticed on on many of your papers that there's a co-author or co-authors that are sort of embedded in that context as well. And I imagine that might help because you almost have somebody like, like a liaison between the, the operations or engineering group and the like physician group, for example. I can see how that might, you know, sort of like, oh, don't worry. I , I trust what they're doing. I was part of it. It seems like that could help as well.
Adams, PhDYes, very true. I think having such a liaison, it's the difference between a successful, you know, research project in healthcare for us, because they're gonna be the intermediary, you know, making that connection.
Halbesleben, PhDYeah.
Adams, PhDAnd sometimes translating the language because the terms use are completely different, right?
Halbesleben, PhDYes, yeah. Yeah, for my own work in healthcare. Yes, I'm still not sure what they're talking about sometimes (laughter) when I'm working with them. So moving a little bit out of the healthcare space, I was curious in, in as we were kind of doing some of the prep for this, you mentioned another study that you're kind of working on that is, is more about different allocation of different types of resources in Alaska. Could you talk a little bit about that? Are you okay to talk about that project? You talk a little bit? Because I it, it struck me as really interesting, but I wasn't, I only had a little bit of context, so I'm kind of curious to hear more.
Adams, PhDYeah. So that's a project we're working on in partnership with Professor Arka Roy and his wife, Kasey Faust, who's a professor at UT Austin. So she, she actually has some connections in Alaska. So it's a project in a very rural city called Bethel.
Jonathon Halbesleben, PhDOkay.
Adams, PhDSo basically it's so cold there that they don't have piped water. So they actually have to use trucks to deliver clean water to the homes and then trucks to retrieve the sewage.
Halbesleben, PhDWow. Okay.
Adams, PhDYes.
Halbesleben, PhDThat is a unique challenge, certainly.
Adams, PhDYeah, it is indeed. So while it's not exactly you know close to other works I've done, it is a marginalized population in a way, right?
Halbesleben, PhDYeah. Well, with impact on health too. I mean, obviously, if you can't get water or remove sewage, that, that would be a problem for health. So I guess there's a it's sort of adjacent to health as well.
Adams, PhDYes, maybe more public health.
Halbesleben, PhDYeah, yeah.
Adams, PhDThe work of routing has been studied extensively, you know, in operations research. But I think there's some unique things about the problem that make it interesting, such as the weather challenges that they have and limited drivers. So they have said that their shifts sometimes can be as long as 12 to 13 hours during the winter because it takes that much longer to hook up, you know, the, the what do you call them?
Halbesleben, PhDLike the connections as you're trying to connect to get the water to the the place.
Adams, PhDExactly. Yeah, and they can freeze and get stuck. I, I remember being sitting in Dr. Foss's students. No, now - now she's a professor too, Michaela. Her defense, and she gave an example of one of these drivers who I think was kneeling for a few you know minutes to plug something in.
Jonathon Halbesleben, PhDWe can see where this is going.
Adams, PhDAnd basically his knees like froze in that position.
Halbesleben, PhDRight, right.
Adams, PhDSo I think that's a good example of how extreme the weather conditions are and how these drivers are impacted, you know.
Halbesleben, PhDYeah, yeah. And something that it seems like it'd be really difficult to optimize in a model as well. different factors. I mean that, that probably isn't something you would include, but it kind of interesting different anecdotes there. So that's, that's really cool. And, and I though I know that that's a particular situation in a, in a particular place, certainly the the things you learn from that can you know maybe be applied to the getting services to other rural areas that you know similar there's similar connectivity issues with other rural areas even here in Texas that certainly are not worried about freezing, but you know the, the concepts are similar.
Adams, PhDYes. And I think it goes into this theme, which I think we need to study more of climate resiliency. So being able to deal with these extreme weather events.
Halbesleben, PhDYeah.
Adams, PhDLike here we have issues like floods, right? So some disruptions that can occur.
Halbesleben, PhDVery good, very good. Well and I know the the work there in Alaska is kind of work in progress but, but beyond that what's what's up next for you? What are you working on that you find really interesting that - that others might find interesting as well?
Adams, PhDSo we have another project that we haven't started yet. We're working on trying to get access to the data.
Halbesleben, PhDOkay.
Adams, PhDBut it's about using RTLS real-time location systems data from UT Southwestern radiology to try to optimize their treatments for patients with you know cancer
Halbesleben, PhDOkay. So in that case talk to me a little bit about what what are you optimizing in that particular instance? So is it like you're trying to figure out where the patients are or is it more like where the cancer is within the the person?
Adams, PhDYeah the, the spatial data they have is of the patients moving through the facility.
Halbesleben, PhDOh wow okay.
Adams, PhDYeah so it's pretty detailed how long they're spending in each step of the process and the, the purpose is to try to schedule a lot the right time for each appointment for each patient.
Halbesleben, PhDYeah. How, how do you, how do they track that within the facility like that? I mean is it is it more just like check-in points like you know when you when you check in at the desk and when they call you back or I mean it sounds a little big brother otherwise.
Adams, PhDThey have this I guess it's RFID technology. I don't know if it's a badge that the patients wear once they arrive.
Halbesleben, PhDOh, okay .
Adams, PhDBut that's how they get the checkpoints of where they are.
Jonathon Halbesleben, PhDOh that's really fascinating. I guess I've not really encountered it in my own sort of treatment for things I've never really encountered having a badge, that at least that I knew had that type of thing. So that's really interesting. So they're kind of keeping track of them as they go. And then, wow I could see how that'd be really rich data to try to figure out how you can optimize the, the scheduling part of it if they're - if there's yeah somebody's like sitting waiting at a long period of time at one spot for example that's probably not the best outcome there.
Adams, PhDYes and some of them you know it can be a very debilitating disease obviously they could have mobility challenges you know which would affect their transit time from one you know room or step of the process to the next.
Halbesleben, PhDOkay well that's pretty cool. I think that'd be really in kind of an interesting set of problems to to try to figure out and making it as, as easy as you can for them for the patient in a otherwise relatively difficult situation. So that's - that's really great work as well. Well clearly you have been very busy with your research you've been working on a lot of different problems from India to Alaska (laughter). And so outside of your work here at the Carlos Alvarez College of Business what are some things that you do to just for fun or to to kind of keep you busy?
Adams, PhDYeah so I enjoy cooking a lot. So I've been cooking if I could I'd cook daily I don't quite have the time for that so I've been doing more meal prep recently you know just spending time with my cats and I started CrossFit about a year ago trying to focus more on my own health.
Halbesleben, PhDYeah
Adams, PhDBecause I, I do work in healthcare right so I should try to I guess be an example in some way.
Halbesleben, PhDGreat well what kind of foods do you like to cook?
Adams, PhDI actually funny enough I do like to cook a lot of Indian dishes.
Halbesleben, PhDOh really interesting.
Adams, PhDYes because I, I'm vegan so I eat mostly you know I eat plant based dishes and of course Indians have been eating you know predominantly vegetarian foods for centuries so they're experts in you know making the vegetables the center of the dish and bringing in lots of flavor so definitely take inspiration from that.
Halbesleben, PhDOh that's really cool. Very good well Katherine it has been such a joy to talk with you. I'm so glad I had a chance to kind of learn a little bit more about your work and about you. That's one of the the really fun things for me in doing the podcast is I get to kind of connect with our faculty a little bit more. And so I really appreciate that. And I'm so glad to know that we've got amazing researchers like yourself that are working on some of these really important issues that impact our health and the the delivery of healthcare thank goodness we've got folks like you working on these things. So thank you for that. Thank you for being here today and more importantly thank you for all the great work that you're doing.
Adams, PhDThank you so much for the invitation, it's been great.
Halbesleben, PhDThank you for listening to the Inside Alvarez Business podcast. Special thanks to our producer Brittney Johnson and for the support of Wendy Frost and Melissa Lackey to help make this podcast possible. Stay connected with the Carlos Alvarez College of Business at The University of Texas in San Antonio to learn more about how we are empowering the next generation of business thinkers and conducting groundbreaking research to ensure their success. Follow us on social media or visit us online at business.utsa.edu until next time I'm Jonathon Halbesleben thank you for listening
