From HIV to the Opioid Epidemic: O.R. at the Heart of Better Public Policy

Ashley Klimp:
Welcome to Resoundingly Human, the podcast by Informs. I’m Ashley, and in each episode, I talk with an Informs member to explore how operations research, analytics, AI, and more are transforming our world from healthcare and transportation to public policy and sustainability. This episode is extra special. It kicks off our refreshed season of Resoundingly Human, and I can’t think of a better guest to help us launch than one of the plenary speakers from the 2025 Informs annual meeting.

Joining me today is Dr. Margaret Brandeau, Coleman F. Fung Professor of Engineering and Professor of Health Policy at Stanford University. Margaret is a world renowned researcher whose work uses applied mathematical and economic models to support health policy decisions. Her research has spanned HIV and drug abuse prevention, the opioid epidemic and housing programs for individuals experiencing homelessness.

She’s also a fellow of INFORMS and has been honored with numerous awards, including the INFORMSs President’s Award and the Award for the Advancement of Women in ORMS. At this year’s annual meeting, she delivered a plenary address on how O.R. can better inform social policy decisions in areas ranging from housing to criminal justice.

I am so thrilled to welcome her today. Margaret, for those who may not be familiar with your work, could you start by sharing a bit about the kinds of challenges your research addresses and how operations research can support good decision-making in health and social policy?

Margaret Brandeau:

So my research focuses on public policy. Originally, I started working in the area of public health, which is things like how do you control diseases? Who should get vaccinated? So much of our original work was on HIV and policies for reducing the spread of HIV. Like, is it cost effective to give people pre-exposure prophylaxis? Basic public health decisions, a lot related to communicable diseases. And then as time has gone on, I’ve started thinking about broader classes of problems. We started thinking about the opioid crisis in the United States. Of course, everyone who works in this field certainly got involved in thinking about COVID, but we also started thinking about broader issues like criminal justice policy and social conditions such as housing policy. So originally we started off really doing rather focused, traditional, slightly traditional work in operations research applied to public health. But as I’ve become more wise, I’ve learned that in fact, we need to think about these bigger problems. A person who is homeless is not gonna be able to get treatment for opioid use disorder or very unlikely. So I just decided maybe we should start thinking about these bigger and even harder problems. So that’s broadly the kinds of problems we are working on.

Ashley Klimp:

That’s incredible. So your annual meeting plenary session focused on models that can help tackle incredibly complex social problems, as you mentioned from the opioid crisis to mass incarceration. What first inspired you to dedicate your research to these issues?

Margaret Brandeau:

So actually, an interesting fact you may not know about me. I know later you’re going to ask for an interesting fact, so I’ll have to come up with another one. But an interesting fact you may not know about me is back in the day,

I used to work on fairly traditional logistics and manufacturing problems. And the interesting fact is I actually have a patented Hewlett Packard on how to set up printed circuit board assembly machines. But anyway, so way back in the day, I was using my OR skills, for example, network modeling optimization, to focus on pretty traditional problems related to mostly to manufacturing and optimization, but some also to supply chains.

I just realized that for me, that was interesting. But the thing is, if you’re working for Hewlett Packard, they have the profit motive. They are gonna optimize their supply chain to within a very narrow percentage of what it could be. But when you think about public health, there are so many actors involved. Many of the problems are so complex. I just realized that people in public health need our help. And I thought, well, of course people at HP need our help, but I think to me it was people in public health need our help a lot more. So for example, when you think about the spread of communicable diseases and what we should do, who should we vaccinate first against COVID or what policy should we have? Should we have clean needle exchange programs for people with HIV?

So these public policy questions, if you’re thinking about a communicable disease, this is a nonlinear dynamic system. So epidemiologists are very good at saying, here’s how the disease is spreading. And public health officials, many of whom are doctors, do not understand dynamic systems really. They just want to think about policy. So I realized that there was a space in there for someone who knows dynamic systems modeling. Who could help inform good decisions in the area of public health. So for example, if you’re thinking about the spread of HIV and who you should give treatment to, maybe we should try to understand the dynamics because that will help us think about what we should do. So I just realized that people in public health need our help.

And as operations researchers, we have unique skills that we can bring to the table. So I just, there’s a niche for us, I think. And I will also say it’s not even clear that people in public health thought they needed our help, but I thought they needed our help. So I just started working in this area. So there is a certain element of, it’s not like people call you up and say, I’m thinking about public health policy for the state of California. Can you help me?

It’s more you think about, know, here’s a really difficult public health problem that people are grappling with. Can I make models that will give insight? So that’s how I got into this area. I think we have a unique role that we can play. And when you think about epidemiologists, they generally just say, here’s what’s happening. You know, here’s where who’s getting infected with this disease, with what frequency. But they don’t optimize. They don’t think about, well, OK.

What should we do? I mean, they obviously implicit is how you should respond, but they don’t think about the whole system and allocation of resources, and we do. So I just think we have a unique role that we can play. I agree. I like the idea of you may not know it yet, but you need my help and here’s how.

Social policy questions are often influenced by limited resources, incomplete information, and political considerations. When you’re building models in this space, what’s the biggest challenge you face? I think the biggest challenge is how to frame the problem and then how to understand how the work you have done can help address the problem, but what its limitations are. So to give an example, one of our studies that we recently published about providing housing to individuals with opioid use disorder. So we made a dynamic model to look at, well, many, many people who are homeless on the street with opioid use, have opioid use disorder and aren’t being treated. It’s just almost impossible to get treated when you don’t have a fixed abode, you don’t have a way to get to treatment. So what if we would provide housing? What would be the health and economic consequences of that?

So a person who’s housed is more likely to get into treatment for opioid use disorder, less likely to die from violence and exposure. But on the other hand, people who are housed are likely to, if they are using drugs, less likely for the overdose to be witnessed. So slightly higher chance of overdosing. Okay, so it’s not a complete win-win. And of course, housing costs money. So.

We made a model that took all this into consideration and we basically showed that this is a good use of money from a healthcare perspective. You spend money, but for not too many dollars per quality adjusted life you’re gained, you can improve life. So a good use of money. But now let’s think about what I’ve done. We just made a dynamic model of that problem. Now we have to think about the broader context.

So the broader context is, well, okay, did other costs change that you didn’t measure? So it turns out that probably criminal justice costs go down. When people are homeless, sadly, they do tend to get arrested pretty frequently for sort of misdemeanor type things, but this costs money in the criminal justice system. And if they’re housed, these costs will go down. So after we did our study, we thought, okay, where will you spend money and where do you save money? So you spend money on housing programs for sure. That’s like, I don’t know, $24,000 here was our estimate of how much you would have to subsidize a person. They spend money, but you save money on criminal justice and you save money on healthcare because they’re not going to the emergency room so much. So that’s the second part of our problem is thinking about different parts of the system that are gonna be affected. And then the third part though is thinking about affordability. Do we have houses for these people? Have we even built them? So we did a little back of the envelope model looking at San Jose, California, one of the cities near me, about how much it would cost to build housing and to frame what we said. So we said it’s a good use of housing, but now you have to think about the bigger picture.

I think that the biggest challenge is to understand what your model can and cannot do and to put it in the backdrop of the broader social picture. So I think that’s one of the biggest challenges is to be really clear about what your model can and can’t do.

and how people should act on this insight. I have to say, I’ve worked her in forms for almost 10 years now. just whenever I talk to a member that talks about all the factors that go into these decision-makings, gives me goosebumps. It’s still so impressive. So thank you. I think it’s just a really exciting area. And I really find it very rewarding. And no slur on Hewlett-Packard, a great company. For me, it just didn’t seem exciting and impactful.

Ashley Klimp:

So quick recap, you’ve studied areas like opioid treatment programs and housing for people experiencing homelessness. Could you walk us through maybe one example where OR-based modeling provided surprising or even especially valuable insights for policymakers?

Margaret Brandeau:

A few years back, one of my students was just thinking broadly at a rather high level of health interventions that are good for you but bad for others. Okay? So what might this be? A good example is antibiotics. You go in, you have a sore throat, it may be viral, it may be bacterial, but you convince your doctor to give you antibiotics, Well, if you did have a bacterial infection, that’s good, right? So it’s good for you. But on the other hand, that might lead to the rise of drug resistant strains in the population if everyone takes antibiotics for things they don’t need. So that’s one example. And then we were thinking and we realized opioids for pain relief are another example.

You have broken your leg. You’re in a lot of pain. You get some opioids. So that’s good for you. But it’s kind of bad because maybe the pills are sitting around on the shelf. Maybe you personally become addicted. maybe that’s bad for you. But it could be bad for others if other people can get access to those pills. Right. And in fact, we have a huge opioid epidemic. Anyway, so my student, Allison Pitt, we were talking about this problem. And then we realized that nobody had made a model of the opioid epidemic in the United States and how things we might do to mitigate it. So we decided to just make a relatively simple spreadsheet model where we looked at people who do and don’t have pain, people who do and don’t have opioid prescriptions, people who do and don’t have opioid use disorder, in other words, addiction to opioids. So then we met with a colleague, we worked with a colleague from Stanford who had actually worked in the office of National Drug Control Policy to ask him what are policies that people are thinking about to mitigate the US opioid epidemic. And so one example is safe disposal. You have some pills left over, you can go give them back somewhere. And actually at the time of our study, there were not very many places.

So anyway, we came up with a number of policies and we used our model to examine it. another thing that people, not surprisingly, that people are doing is cutting back on prescriptions, right? You go to get, you have a broken leg, you go in, you often only now going to get a seven day supply of pills. So cutting back on prescriptions and if you want to refill, maybe you won’t get it. So, but the interesting and unusual finding of our model was when the pill supply is cut back, people who are relying on diverted pills, people with opioid use disorder, they’re going to turn to something else to satisfy their opiate cravings, and that’s going to be heroin. And that’s drug injection, which is far more deadly.

So what our model showed is in the short term, well, in the long term, of course we have to cut back on opioid prescriptions, but in the short term, unless you’re very aggressive about scaling up treatment for people with opioid use disorder, there are to be more deaths, And in fact, this study made the front page of the New York Times saying, a new study shows unintended consequences of cutting back on opioid prescriptions.

So that was just a really good example that I hadn’t even really thought about before we made the model, but there are some people when the opioid pills are no longer available, they will turn to heroin. And in fact, we’ve seen this in various places in the US. So I thought that was a very valuable insight for policymakers. This paper was published in American Journal of Public Health. And interestingly, this is in our previous administration, not the current administration, the Assistant Secretary for Public Health called me up and said, I read your paper. I have been assigned the job of, you know, looking into the opioid crisis in the US. And as far as I can tell, you’re the only research group that’s actually made a model and provided insights. I would like you to come to Washington and I’d like to set up some working groups. and I just thought that was super cool. This all started with a spreadsheet model, which I mean, to be fair, we did it very, very carefully and very thoughtfully.

But it provided insights that we did not have before we began the research. And insights that it sounds like went on to save lives, which is always incredible. Well, we hope so. I mean, I think one of the key findings was that you have to scale up treatment. And that’s been a little iffy because then we had the COVID pandemic and treatment programs were shut down or there was less access. So lots of things happen in this world.

Ashley Klimp:

Yes, yes, they do. Speaking of looking at today’s headlines, know, opioid overdoses do remain devastating in communities across the country. Housing affordability is a growing crisis. How do you see O.R. helping us respond more effectively to these urgent issues?

Margaret Brandeau:

I think what O.R. can help you do is these are many of these public problems are very complex problems. It can help us understand the consequences of different actions, can help us understand if you don’t house these people, here’s what will happen. If you don’t start stemming the spread of fentanyl, here’s what will happen. Well, that’s one thing. The second thing, though, is thinking about resource allocation. Should I do this or should I do that? And I think operations research is really helpful for trying to think about alternative allocations of limited resources.

At a very high level, I think that’s how our modeling can help. of course, as similar to the previous example I just mentioned, it’s about understanding, getting insights into complex systems. So like in the opioid model in advance, I just hadn’t really thought about that unintended consequence, but once you finish the analysis and look at it, even under conservative assumptions, it’s plain as day that that’s what will happen.

So I think that by modeling these dynamic systems, we can get insights and we can help think about how to allocate resources.

Ashley Klimp:

With research that touches on such important and often personal challenges, is there a particular moment in the course of your career that really underscored for you, this is why I do this work, this is why I’m doing what I’m doing?

Margaret Brandeau:

About 10 years ago, I had a student in my class on health policy modeling. I asked the students to do a project, make a simple model of a policy and analyze it.

So this student came to me and said, I’m working with a group at Stanford called the Asian Liver Center. And we are thinking about what should we do about hepatitis B among Asian and Pacific Islanders in the US? Well, I, of course, didn’t know anything about this. So the professor he was working with cold called me. So this is a transplant surgeon at Stanford. And he cold called me. never met the man in my life. He came to my office. He said, You know, I’m interested in in hepatitis B. There’s a very high burden among Asian and Pacific Islanders in the US, and I’m like, what’s hepatitis B and? Anyway, he said, you know, we should maybe do make some models and I understand you know how to make models. I said yes, I do. So we did a model looking at screening and vaccination policies for Asian and Pacific Islanders in the US. And after.

It turns out that ⁓ for adults, you would screen them and treat them because we have very good antivirals, but you wouldn’t vaccinate because the chance of them not having yet had it and getting it in the US is quite low. anyway, we did some analysis and the CDC changed its recommendations. But the more exciting part was we worked on, we took our model or same ideas.

And applied it to look at catch-up vaccination for hepatitis B in China. So hepatitis B at the time, I think about nearly 10 % of the Chinese population had hepatitis B infection. Very often it’s a silent infection. We’ve had a vaccine available since the 1980s and it’s very, very cheap, but it was not available in China until later. China’s a very large geographically dispersed country.

And there was some thought about, should we do catch-up vaccination? You didn’t get vaccinated as a baby. Should we vaccinate you? So we did some analysis. And in fact, we showed that catch-up vaccination in China for young people is cost saving, because the vaccine costs $3. And even if only a few people get treated, it’s just a great use of money. So after this study, my colleague, Professor So from hepatology is very well connected with individuals in ⁓ China, CDC and other places. China changed its recommendation to offer free catch-up vaccination to the 120 million children in China who didn’t receive the full vaccination dose. That’s incredible. It’s got to myself, you know, we had a model that we did carefully with the best data we could. And that led to many, many, many lives being potentially saved. So that just crystallized it for me, thinking, yeah, of course not everything you do has incredible impact, but this is a great example of enormous impact. So that was, think to me, it just kind of crystallized like, yeah.

Yeah, this is really exciting and worthwhile thing to be doing.

Margaret Brandeau:

Absolutely. So Margaret, you’ve been recognized for your contributions to teaching and mentoring, including receiving the award for the advancement of women in ORMS, as well as the SolGast award, which was just presented at this year’s annual meeting. First of all, congratulations. What excites you most about the future generation of OR researchers stepping into these problems?

Margaret Brandeau:

I think the most exciting thing is that there are going to be new problems. There will always be new problems. And I think if we, if young people are excited about trying to tackle these problems, to try to develop good insights for decision makers, that’s what’s exciting is I think having an open mind and looking to see what are the exciting problems that are out there, right?

So to me, that’s the most exciting thing is just young people with new ideas who want to work on these problems, whatever they may be.

Ashley Klimp:

All right, for this next section, this is something new. This is our Ask AI segment where obviously AI is everywhere now. Our members are contributing to the development of it. The informed staff is using it, explaining to our family at holidays what AI is, or at least trying to. So I wanted to give us the opportunity to ask AI a question and then maybe have a conversation about it. And it can be something groundbreaking or just something fun and silly. So I will go first with my question. I asked AI, there are a lot of different types of pizza out there based on what you know, what is officially the best pizza style? I know this is a very controversial topic. And AI recommended several types, Neapolitan pizza, New York style pizza, as well as Chicago pizza.

And so that’s great information AI, but I want to know what AI would eat if it was craving pizza. What is its go-to? So it answered, if it absolutely had to pick one, it would go with Neapolitan pizza because it’s original. It’s the foundation every pizza style was built on. It’s the only style with a protected status like champagne or Parmesan Reggiano.

So it has an official claim to being the best. Also, and when AI is hungry late at night, that is the pizza that it craves. So.

Margaret Brandeau:

Okay. Well, you’re making me hungry.

Ashley Klimp:

I was a little hungry when I was pulling some of these. And I agree with that. Neapolitan is the only way to talk. I think so too. Yeah. I’m sure there’ll be some folks that disagree with that, but you know, AI doesn’t lie.

Margaret Brandeau:

No, AI doesn’t lie.

Ashley Klimp:

Then Margaret, I asked you what question you would like to ask. So you said, what is a good place for relaxing vacation? And apparently AI has quite the vacation budget because it recommended if you’re looking for a beach or a warm weather escape, Maui, Hawaii, the Amalfi coast in Italy or Tulum, Mexico. If you’re looking for a quiet nature retreat, Sedona, Arizona or Lake Louise in Alberta, Canada.

If you’re looking for perhaps a slower pace and exploring the culture, Kyoto, Japan and Lisbon, Portugal are what it recommends. And then if you’re looking for a wellness retreat, Bali, Indonesia and Costa Rica. So if you want, I am more than happy to ask it to narrow down where AI would go on vacation.

Margaret Brandeau:

Ask where they would go on vacation. Budget is no object. I mean, we might as well think big here.

Ashley Klimp:

I agree. Oh my, two options, both sound incredible. If budget were no option for a once in a lifetime, soul soothing vacation. Number one choice is the Maldives. Next option would be a private chalet in the Swiss Alps.

Ashley Klimp:

I think I’d go with the Swiss Alps personally.

Margaret Brandeau:

I’d go Maldives, I think. Just I love the ocean and the water,

Ashley Klimp:

And that’s the beautiful ocean there. Well, thank you for indulging me with AI fun. Very enjoyable. People take AI very seriously. I think it’s fun to take a moment just to appreciate that it also, you know, it’s a fun and informative tool.

Margaret Brandeau:

It can give you ideas.

Ashley Klimp:

Yeah, exactly. The Maldives. All right.

So back to the more traditional interview. For INFORMS members, whether they’re in academia industry or still students who are interested in exploring social policy problems, what advice would you give them to just get started?

Margaret Brandeau:

One thing I always tell people is just read the news every day. Talk to people. See what’s happening. Keep your eyes open. See what you think is interesting. And then for example, on the homelessness, I had actually been thinking about it for a couple of years while doing other things. And I just had a very strong feeling that this is something I wanted to do research on. Then I finished doing some various projects about a year ago, and I thought, OK, during the summer, I will learn everything I can about homelessness. So I just read every single article I could about homelessness.

And I worked with a student to develop the model that we developed. But along the way, just, well, the two of us, in fact, is my student, Isabel Rao, who’s a professor at University of Toronto, my former student. We just read everything we could to learn all about it. This is something I didn’t know a lot about. So I think it’s to identify a problem you think is interesting and then just immerse yourself in the topic and see what you can learn about it. I think the second piece of advice depending on the problem is to work with a domain expert. So for example, when we looked at our model of the opioid epidemic, we worked with someone who had worked in the office of national drug control policy, because we didn’t know enough about what are the current policy issues that people care about. So I think those two things, learn as much as you can and as appropriate work with domain experts. And I think pick something that you think is interesting.

I strongly believe that you should not do research on a topic just because someone else thinks you should. I sometimes meet assistant professors who say, well, my department chair told me to do X, Y, No. I think that is not necessarily a recipe for success. I think people need to find out what problems they find exciting, important, and motivating, and work on those. Hopefully, in a way, they get some tenure, but yeah.

Ashley Klimp:

I think that’s excellent advice. So for our listeners who aren’t involved in academia that may be working in public policy, healthcare, or even in the nonprofits, if they wanted to start to incorporate OR into their own organizations, how would you suggest they get started?

Margaret Brandeau:

Typically, it’s to work with someone who does simple OR. So we have done a number of projects at Lucille Packard Children’s Hospital at Stanford. We have both an adult and a children’s hospital, 550 bed children’s hospitals. So it’s a pretty good size and it’s a regional trauma center or something, know, so a pretty sophisticated hospital. So we have done some work with various physicians who say things like, I am wondering if you could help me better schedule my 10 operating rooms. Well, actually we have 10 operating rooms and 10 procedure rooms. So they have questions, you know, they understand we can do things. And what we found out is that

The best thing to do is start simple. just so, and same thing if you work in an organization, start simple. Make a dashboard. Make a dashboard showing my operating rooms were busy at this time. So we had this much overtime. The post anesthesia care unit beds were filled this much percent of the time. So we started with things that were easy wins. So you could just have all this data here, then the next step is maybe to do some analysis. So that is generally my advice is start with a simple thing or another example. Here’s a heuristic that I can use to schedule the chairs for outpatient chemotherapy. You know, not some fancy algorithm, just something simple. So I think that’s how you do it. You start with something that is small, understandable, that you can have a win with. Then you can move on to bigger things. But I think if you try to solve all the world’s problems with your model and your approach, it’s just, that’s not a recipe for success.

Ashley Klimp: 

As much as I hate to have our conversation draw to a close, ⁓ we’re getting near the end. But before we wrap up, I’d love to close with a few quick fire questions and just say the first thing that comes into your mind.  A book you think that everyone working in policy should read.

Margaret Brandeau:

Ethics in the Real World by Peter Singer. I just read that book and it’s 90 essays on things that matter.

Very thoughtful and interesting. Really like that book.

Ashley Klimp:

Now what’s your favorite analytics tool or approach to teach?

Margaret Brandeau:

I think to me the most exciting thing to teach is the art of mathematical modeling. How do you take a complex problem and frame it and develop an appropriate model? So the art of modeling. We are model agnostic in our research group. We don’t say I must always use simulation. We think about the problem and what’s the best model. So that I think is fun to teach.

Ashley Klimp: 

Who inspires you most in the OR and analytics community? I think there are a lot of really inspiring people, some people doing theory that’s just amazing. I would say in the policy area, a couple of people that I’ve always respected and admired is Ed Kaplan from Yale, my colleague Larry Wine from Stanford. I think they’ve both worked on really interesting high impact public problems. So a lot of inspiring people in the OR community.

Ashley Klimp:

If you had to describe the future of operations research in one word, what would it be?

Margaret Brandeau:

Far reaching. I think operations research can, especially with the rise of data and our incredible skills in data analytics, think we can, OR is going to creep further and further into many areas in a really good way. What’s one professional tip you wish you had known at the very start of your career?

Margaret Brandeau:

I think it’s that you have to be your own best judge. I remember I wrote a paper with a colleague. said, no one’s ever going to read this. And I was like, I don’t care. I think it’s interesting. And I’ve written a few papers like that. One of them I just noticed was published, think, in Operations Research. It a lot of citations. So I think it’s to be your own best judge and not think about, someone told me I should do this.

You have to think what’s interesting and important and follow that.

Ashley Klimp:

Now, do you have a hobby or interest that might surprise people? I do. I’m a beekeeper and I make beauty products and with the wax and the honey or personal care products with the wax and the honey. So that’s pretty cool. I’ve even got it branded with with I don’t sell it. I just give it. But I even have it branded with labels and everything.

It’s really fun. My bees were just making so much honey one summer I got three gallons out of three hives. my gosh, what am I going to do? And so I just started making like honey, but sugar honey body scrubs and wax lip balm and all kinds of products. Yeah, that’s amazing. And then I made labels and, you know, got beautiful containers and made fancy labels and branded everything.

Ashley Klimp:

Do you have a bucket list? And if so, what’s the top item on it?

Margaret Brandeau:

Well, I just got an item. It’s to go to the Maldives. Yeah, I told me. That’s on my bucket list now. Top of the list.

Ashley Klimp:

To sum up, today’s conversation reminded us that operations research isn’t just about algorithms and equation. It’s really about people. The models that Dr. Brando and her colleagues are creating are helping address some society’s most pressing challenges from addiction to homelessness and showing us how science can make a difference in human lives.

If you’d like to learn more about Dr. Brando’s work, you can find links in the show notes section of this episode on resoundingthehuman.com. Margaret, thank you so much for joining me today. And thank you to our listeners for being a part of this first episode of our refresh season of Resounding the Human. Be sure to subscribe, share the podcast with a colleague and leave us a review. We’ll see you next time.