Sara Del Valle is a computational epidemiologist at Los Alamos National Laboratory who uses mathematical models to predict the spread of diseases.
What is the biggest question facing your field?
The next frontier in epidemiology is forecasting diseases. We’ve spent the last several decades working to understand disease dynamics—what causes them, how they’re transmitted, how to treat them. And while there are still gaps in that knowledge, we have made a lot of progress. Now our biggest challenge is predicting outbreaks early enough to better mitigate their impacts.
Why is it significant?
Until we can forecast when and where diseases will spread, we’re significantly limited in our ability to respond. Just like hurricane forecasting helps policymakers deploy resources and develop plans for evacuation and shelters, disease forecasting will allow us to prepare for deadly outbreaks. The 2014 Ebola epidemic in West Africa that killed more than 11,000 people is the perfect example of why we need forecasting models. Controls that were put in place to contain the virus were often too late or minimally effective, because the disease had already spread to the area. If we had been able to detect the earliest cases before the disease took hold, medical equipment could have been staged and preventative practices could have been put in place to significantly lessen the devastation.
Where is the answer likely to come from?
At Los Alamos National Laboratory, we’re working toward being able to forecast diseases by integrating different data streams, including satellite data, social media trends, climate data, demographic information, and regional clinical data. Each of these heterogeneous data streams provide epidemiologists with a piece of the puzzle to answer the question: When and where will a disease outbreak happen?
For example, satellite data can tell us where environmental conditions most favor an epidemic (such as standing water that would harbor mosquitoes and perhaps increase the incidence of Zika). Similarly, many diseases are seasonal, so climate data such as temperature and precipitation can be indicative of mosquito risk.
Also, social media trends can tell us about human behavior in response to an ongoing outbreak, such as wearing protective face masks, hand washing, vaccinations, or changing travel plans. By looking for key words and hashtags in Twitter feeds, as well as Wikipedia and Google searches, it gives us an idea of where a disease might already be present in a population and how people are changing their behavior (or not) to protect themselves from getting infected. Prior to social media, we weren’t able to capture behavioral changes until after the fact. Now we can see them in real time and infer disease outbreaks and behavioral changes based on that.
Add to that regional clinical data, which can tell us what illnesses are precipitating visits to a doctor’s office, and we get a more complete picture of the path of an epidemic.
In addition to these data streams, computational power is critical in helping to answer the question of how to forecast diseases. Access to cloud services allows us to collect data. Plus, at Los Alamos, we have some of the most powerful supercomputers in the world to process big data at lightning-fast speeds.
We use machine-learning approaches to analyze satellite and internet data streams. We then feed the results—along with climate, demographic, and clinical data—into our mathematical models to predict disease spread, much the same way data assimilation and climate models are used to predict the weather. Currently our models are “backcasting” (using older data to see if they accurately predict what’s already happened), “nowcasting” (monitoring the present state of disease incidence), and even “forecasting” (predicting the incidence of influenza in the United States). We’ll use these outcomes to improve our models and then soon begin applying them to forecasting other diseases.
Success in disease forecasting could literally change the world. In our increasingly mobile society where diseases only need a short plane flight to spread to an entirely different continent, it’s easy to see how public health impacts national security. That’s why Los Alamos is working on this problem. Knowing the potential path of a disease could give communities time to prepare for an outbreak and keep the disease at bay; it could help us determine when and where to travel; it could help us create timely vaccination programs. It has the potential to save thousands of lives. In short, it will make the world a safer place.
For more information about Sara Del Valle, visit her Los Alamos page. If you’re interested in epidemiological forecasting, Los Alamos’ mathematical computational epidemiology project site offers a solid start point.