UCLA Anderson’s Healthcare Analytics Course Untangles a Nuanced Industry

Assistant Professor of Decisions, Operations and Technology Management Fernanda Bravo joined the UCLA Anderson faculty in 2015. One of the courses Bravo teaches meshes perfectly with her own scholarship: Healthcare Analytics. The course is part of the school’s Master of Science in Business Analytics curriculum, though it is open to Anderson’s MBA students as well. The course teaches students to apply advanced analytic techniques to a wide array of issues and problems in the U.S. healthcare industry.

Healthcare as a field has long been a topic of research and analysis. But one of the goals of Bravo’s course is to consider the issues in this space from new perspectives. She says there is much more data available today, and high-quality analysis of such data leads to better health care decisions. According to Bravo, students who take her course are preparing for careers in the industry, many in consulting or with insurance companies. Humana, for example, has hired several Anderson MSBAs who took Healthcare Analytics.

Q: What are the goals of your Healthcare Analytics course?

The main goal for the students is to understand some of the nuances of the U.S. healthcare system. The second big goal is to apply some of the technical skills that they have learned in the core classes in the context of healthcare, and maybe a third one is for the students to explore their own kind of healthcare problem by working on an applied project in a group setting using health care data. Typically, I also ask them to participate in a case competition at the national level.

Q: Could you describe an example of the type of health care issues you’re talking about?

So, for example, in one of the classes we discussed COVID analytics. Students were presented with a problem regarding where you could place vaccine access points so you could get the most people vaccinated. They had to do a demand prediction component to understand how people seek vaccinations, based on how hard or easy access points were to reach. From there, the students had to decide, using optimization techniques, where to open vaccination centers.

Q: For a student or, for that matter, a professional in the field, what are the challenges in acquiring the necessary data for analysis?

Healthcare, I think, is a unique industry, in the sense that to solve a problem you need to tap into multiple data sources. With the COVID example I gave you, you might need demographic data from the census bureau and you might collect your own data — based, for example, on where access points could be located. Or the data might relate to vaccination rates that are available at the state level.

So, you start with that question and then you think about the best data sources. You need to think about multiple data sources because you’re not going to find your answers all in one place. You have to think about combining these data sources, and then how you’re going to use the data.

My goal is for students to always think about making better decisions, not just the data. I want them to use these data to solve a problem and make better decisions.

Q: Are there any prerequisites for the course?

The course is open to both MSBA and MBA students. They definitely need to have some coding skills that they can learn in the core classes. They work in groups, so sometimes someone in the group might be stronger in coding than others and they combine their skills. They have to have basic knowledge of regression analysis and some understanding of optimization models.

Q: What source material do you use for the course?

The course has evolved over the years, and I’d say it’s a combination of sources. There’s some background information like the structure of the healthcare system (stakeholders and their dynamics) and the performance of the U.S. system. Then we use case studies based on research. I bring in cases that I’m very familiar with and some technical papers that I think are good examples. For instance, I use a paper on identifying providers’ upcoding behavior in Medicare claims data.

Q: How is the class structured?

It’s a short class, five sessions with a three-hour model.

Typically, in the first half, we go over some of the more theory-based aspects of the material. Then in the second half, I will have some kind of lab experience for the students to play with the data a little bit and try to answer some questions related to the theory part, and in some cases apply the technique themselves. Then I ask them to submit something at the end of the class, like answers to 5 to 10 questions about what we discussed.

Then there’s also a “B” component, which is this project that I asked them to do. I typically ask them to be part of the Humana-Mays Healthcare Analytics Case Competition, which is a case competition that has been running for over eight years. At least two times teams from my class have won it. This is a big competition, the main prize is something like $60,000, so it’s a big deal.

Q: What kinds of challenges does the Humana-Mays Healthcare Analytics Case Competition pose the teams?

The first time I offered the class, we looked at ways of predicting opioid abuse. The competition provided all the data, millions of records. The teams had to build their variables, build their models and try different approaches to predict who was going to become a long-term opioid user based on the history of the patient. And that team won the competition.

But something interesting about this competition is that they don’t only care about the model performance, like achieving 99% accuracy. They also care about the implications of the teams’ work, and I think that’s why our students do really well. They go the extra mile to explain the implications. What do we do with this model? How can we make better decisions? How can you identify these people at risk and then intervene?

Another competition a couple of years ago was about predicting who is at risk of becoming homeless — also using health care data — and still another about predicting adverse reactions to a specific cancer drug.

Q: What kinds of knowledge do these hands-on components of the course cement?

After taking the class, you will definitely understand at a high level how the U.S. healthcare system operates. I think something that is unique is that it is not a system, per se, it’s more like a group of separate organizations that come together to deliver care to the U.S. population — all these organizations with different objectives and priorities. So, in the class, you will learn the tradeoffs that all these kinds of stakeholders are negotiating with each other.

You will understand the role of data, how data and models can be used to make better decisions in specific settings. There’s a lot of value in this skill because it can be used in other industries, too.

To successfully inform better decisions, you cannot just plug and play existing, off-the-shelf methods. In healthcare — I would say, more than in other settings — you will have to critically think about data sources, how to combine them and how to develop interpretable features and models to make them useful.