posted by Sara Khor
We are starting a series of interviews with graduates from the CHOICE Institute, and are very excited to have Nathaniel Hendrix, who graduated from the PhD program in 2020, share his experience with us.
Nathaniel in the grad room, before the pandemic
“Grad school is like training a pet, where the pet is your own mind… You have to be intentional about figuring out what you want to teach yourself and how you’re going to do it.”
Nathaniel Hendrix
- Why did you choose to do a PhD? How did you choose health economics?
I entered the PhD program directly after graduating with my PharmD. I went into my clinical training imagining that almost all of our decisions were grounded in a solid basis of evidence. But once I learned to read studies for myself, I began to see how complex the processes of gathering and evaluating evidence really were. This made me a bit hung up on how clinicians make decisions and how they can learn to make them better.
I had always been pretty indulgent with myself about taking electives during my PharmD training. I’d explored computer science and philosophy of science, but when I started taking classes in health economics, that gave me a vocabulary for talking and thinking about decision-making that I hadn’t been able to find before. At the same time, it’s a relatively young field and a very interdisciplinary one, so it felt to me like I could impact the direction of the field and that my curiosity about exploring different ideas would be rewarded.
2. What was the topic of your PhD?
My dissertation was on using health economics tools to solve translational issues in artificial intelligence (AI). It had two separate aims. First, I conducted a discrete choice experiment to see what primary care providers see as valuable in how AI could be used for breast cancer screening. Most women who take part in breast cancer screening make the decision to start with their primary care providers, so these providers have an outsized role to play in determining how AI will be used for breast cancer screening. We found that there were multiple ways that developers can appeal to primary care providers with their AI products: by improving sensitivity, by having a well-thought-out workflow that includes radiologists, and by having highly diverse training data.
The second aim had to do with using cost-effectiveness analysis to inform how AI algorithms are used for breast cancer screening. We used a real set of AI algorithms that had been submitted to a contest and used different methods of selecting a sensitivity/specificity threshold at which they could operate. Our model ended up selecting very similar thresholds as other heuristic methods, but it taught me a lot about the challenges of using cost-effectiveness analysis on AI.
“Write down every research idea that you have, and revisit that list frequently. Half of the ideas will be terrible in retrospect, but that’s okay.”
3. What are you currently working on? Does this align with your training or your research interests?
After my PhD, I started a postdoc at the Harvard T.H. Chan School of Public Health, where I’m mostly studying cost-effectiveness methodology. I have three main projects right now. First, I’m working on developing methods for integrating financial risk protection into decisions about prioritizing healthcare interventions in low- and middle-income countries. Healthcare plays a major role in keeping people out of poverty, but this isn’t acknowledged in conventional cost-effectiveness analysis.
Next, I’m working on methods for assessing the cost-effectiveness of health system strengthening interventions, such as building new facilities, training personnel, or developing a new informatics infrastructure. This is challenging because these long-term projects, which almost everyone acknowledges as important, have to compete for funding against urgent needs like making more medicines available or expanding vaccinations.
Finally, I’m completing a project about how to use Deep Learning-based breast cancer risk scores to personalize screening. Even though breast cancer screening reduces cancer mortality, there are many downsides because the false-positive rate is pretty high in screening. So we’re hoping we can use AI to determine who is most likely to benefit from screening and who might have a higher risk of being harmed.
I see this last project as most connected to my PhD research, because I’m using many of the same techniques that I used in the second aim of my dissertation. But I’m also learning many methods like constrained optimization that aren’t emphasized at UW but that are important when thinking about operating a healthcare system efficiently. Of course, I’m also wrapping up several projects I started during my time at UW, so there’s some continuity there too.
4. What are some of the things you wish you’d known when you started your PhD?
I wish I had had a better system for keeping track of the things I was reading. Ultimately, I missed the opportunity for making connections because it took me until I was well into my dissertation to discover the tools I needed for taking notes and helping myself to rediscover ideas I’d read in the past. For me, this ended up being the Zettelkasten system, which I use on the website Roam Research (read Sönke Ahrens’s book, “How to take smart notes,” for more info on this!).
5. Can you share one of your favorite or proudest moments during the PhD years?
Probably my proudest moment was when I found out that I had gotten a PhRMA Foundation fellowship for my dissertation. I was on vacation in San Francisco, waiting out an afternoon rainstorm in a café when I got the call, and it was really a pivotal moment for me. I had been feeling a bit of imposter syndrome about my dissertation, but to know that these other researchers had found my ideas exciting enough to fund was a huge boost of confidence.
“I’m convinced that you can’t be a good researcher without learning to be a good reader.”
6. What do you think are the “secret sauces” of a successful PhD experience?
Completing a PhD program is a great chance to explore. I didn’t quite understand this when I went into it and felt pressured (by myself, really!) to focus more on one area or another. Once I started just saying yes to every experience I had time for, I started to learn a lot more. This meant not just learning more about different technical matters, but also learning about what sort of work styles and communication styles I’m most comfortable with. That’s super important, because our work is usually so collaborative.
As you go through a PhD program and learn more, what’s expected from you also changes. In the beginning, you’re soaking up new methods and concepts, so you end up doing a lot of work on tasks like data cleaning or literature review that are really time consuming but necessary. With more experience, you should be able to start making suggestions to your collaborators about the directions for your work. And then finally, with your dissertation, you get a lot more independence about its direction.
Part of this development process is having good reading habits. That means keeping track of what journals are publishing so that you have a sense of what current conversations are happening. And then, again, finding ways to make connections between what you read. Ultimately, I’m convinced that you can’t be a good researcher without learning to be a good reader.
7. What do you think is the best metaphor for your grad school experience? How would you complete this sentence: “Grad school is like…”
Grad school is like training a pet, where the pet is your own mind. During grad school, you’ll have to teach yourself a lot of new skills, so it’s important to think about how to keep yourself motivated. It’s also vital to learn to recognize when you’re tired and need to do something entertaining. But overall, you have to be intentional about figuring out what you want to teach yourself and how you’re going to do it.
8. Do you have any other advice for current/future PhD students?
Write down every research idea that you have, and revisit that list frequently. Half of the ideas will be terrible in retrospect, but that’s okay. It’s easier to come up with good ideas if you come up with lots of ideas. And, again, to come up with ideas, it’s important to cultivate a very intentional habit of reading academic literature.
Also, I would encourage students to publish broadly. Look at the areas where they’re doing projects, compare it to the major areas in the CHOICE curriculum, and see if there are any weaknesses. You’ll have a lot more flexibility with your job search if you publish broadly. One way to do this is to collaborate with a lot of different people who are working in different areas.
Incremental Thoughts 