Modeling COVID-19 transmission: Apply aerospace
May 18, 2020
While robots and infectious diseases may seem worlds apart, they actually are quite similar when you look at the underlying theory. – Dan Klein
A&A alum Dan Klein draws on his controls training to model the transmission of COVID-19 in King County.
Not many A&A alums pursue a career to track and reduce world disease as part of the United Nation’s Sustainable Development Goals. But Dan Klein (A&A Ph.D. 2008) has adapted his training in control theory to model transmission rates of diseases, including malaria, polio, tuberculosis, HIV, pneumonia and typhoid. Now, his group at the Institute for Disease Modeling, based in Bellevue, is tracking transmission rates of COVID-19 in King County, Washington.
We asked Klein about his career path into disease modeling and his research on COVID-19 transmission.
Were you interested in disease modeling when you started your studies in A&A?
When I started A&A, I had a keen interest in robots, and I still think robots are the coolest thing in the world. I had just completed a B.S. in mechanical engineering, but I wanted to work on the mathematics of robot brains - that’s what brought me to control theory. My Ph.D adviser, Professor Kristi Morgansen, was building these amazing free-swimming fish robots and the nonlinear theory to go with them, and I knew that’s what I wanted to do! The coolest part is that we had several of these fish robots, and so I got to play with the early mathematics of multi-agent control systems and learn about schooling behavior from biologists.
What was the inspiration for you to pursue this work?
In infectious disease modeling, we use mathematics, programming and supercomputing to improve lives around the world. These models inform data needs and policies, thereby leading to burden reductions and hopefully disease eradication. The most inspiring part of this work is visiting the field to see the impact of these recommendations on the lives of real people.
Can you tell us about a field experience that has really impacted you?
One of the most cost-effective and long-lasting interventions against HIV is voluntary medical male circumcision (VMMC). The efficacy of VMMC in reducing HIV incidence in men is about 60% (really good)! We modeled the impact of increased VMMC programs in Zimbabwe, Kenya, and Uganda, and I was in Kenya presenting the simulation results to the country decision makers when they asked if I’d like to do a field visit. SURE!
We traveled to a rural clinic where a group of about 20 school boys had volunteered to come in. The visit started with counseling on safe practices - wow, I hadn’t appreciated the impact of these messages. Long story short, I ended up being handed a surgical gown and hair net before being pushed into the operating room. I’m a doctor, but not that kind! Anyway, opportunities like these really bring the numbers in our simulations to life. These brave boys now have an increased chance of avoiding HIV due in small part to our modeling work.
What is unique about modeling for COVID-19?
Before COVID-19, IDM focused almost exclusively on low and middle-income (LMIC) settings in Africa and South Asia. While we’re modeling COVID in those settings today, we’re also focusing right here on Washington State. It’s likely that we won’t have a COVID vaccine for over a year, so the big question right now lies in balancing health and economic outcomes. As we learn about physical distancing, contact tracing and case isolation towards reopening the economy, we’re asking how these ideas could be deployed in LMIC settings.
One challenge in LMIC is that COVID-19 is just one of many infectious diseases. We have to take a holistic view to ensure the COVID response doesn’t trigger a massive reversal of decades-long progress in reducing the burden of malaria, HIV, polio and many other preventable diseases. So in many ways, we’ve come full circle and are using the math and models we’ve spent years developing to reinterpret the landscape in light of COVID.
What was your path to the Institute for Disease Modeling?
After A&A, I joined the Center for Control and Dynamical Systems at the University of California, Santa Barbara as a postdoctoral scholar. While there, I worked on a sensor network question in which information spread like a disease - that’s what led me to join IDM in late 2010. I stepped away from IDM for a one-year sabbatical at the Bill & Melinda Gates Foundation, where I led various data science and strategic initiatives in the Strategy, Data and Analytics team. Now I’m back at IDM leading the Computational Sciences Research team, although I spend most of my time on COVID modeling these days.
How did your research at A&A prepare you for this area?
While robots and infectious diseases may seem worlds apart, they actually are quite similar when you look at the underlying theory. Both are dynamical systems, and in each case we begin by building a simplified model, e.g. using ordinary differential equations (ODE). The foundation of mathematical models in epidemiology is the simple susceptible-infectious-recovered “SIR” ODE, much like the ones I had studied in A&A.
My initial work at the IDM was to build a model of HIV. In HIV, there’s within-host viral dynamics and between-host contact networks. It looks a lot like a school of robotic fish if you squint hard enough!
How would you advise students interested in this kind of modeling to make the transition?
It's likely that many facets from your work will translate, even if it's just knowing how to learn! So master your core discipline, but also show curiosity and seek answers more broadly. It might feel out of your comfort zone at first, and it does take time, but it's worth it. There are great textbooks and resources online, and the UW has some fantastic courses in mathematical biology. For those looking to take the next step, IDM has summer intern and postdoc programs.