Runaway epidemics are scary. The problem is, diseases don’t respect borders and can’t be added to the no-fly list.
Wave after wave of virulent diseases have struck around the globe in recent years. Last year, the threat of Zika made athletes question the trip to Rio de Janeiro for the Summer Olympics. Since 2005, the nasty viral disease Chikungunya has spread widely from Africa and Asia to the Americas, infecting an estimated 3 million people globally a year. Hundreds of thousands will die.
One way to minimize the human toll is to warn people, and that depends on knowing a disease threat is developing. Then health officials can alert vulnerable populations and help them take necessary precautions, such as simply washing their hands or avoiding social contact.
And yet, forecasting disease is a tricky business. Health records are an obvious source, but their built-in lag time in reporting limits their usefulness for disease forecasting. So a multidisciplinary team with expertise in epidemiology, ecology, mathematics, data science, computer programming and remote sensing at Los Alamos National Laboratory have turned to data from some surprisingly familiar sources. Look no farther than what’s under your thumb, plus extensive satellite imagery.
The lab is using the kind of data you generate when you tweet or look up something on your cellphone. By mining public data from Wikipedia, Twitter and Google, along with data from doctors’ offices and clinics, the lab has found a way to refine its disease-forecasting mathematical models. (All the data has been stripped of personal information before analysis.) Last year the team accurately predicted a late and mild flu season based on this data.
Now they are improving their models by including data about conditions on the ground that can influence the spread of disease. That’s where satellite imagery comes in. Santa Fe-based Descartes Labs, a Los Alamos National Laboratory spin-off, has developed a computer-based platform that collects public and commercial high-resolution satellite images and aggregates them into a database.
Los Alamos researchers use those pictures to detect disease-harboring environments. They look at satellite data on soil moisture, land use, standing water, temperature and deforestation — all factors affecting potentially contributing to mosquito-borne diseases, which include Zika and Chikungunya.
In a current project, the Los Alamos team combines the Descartes data with social media chatter, online search trends and multiyear clinical surveillance data on mosquito-borne diseases from approximately 5,500 Brazilian municipalities. By incorporating all this data into laboratory-developed algorithms and computer models, the team can more accurately predict the path of a disease through a population and through time.
The lab team’s approach is a finalist for a prestigious R&D 100 award. Known as the “Oscars of Invention,” the awards from R&D Magazine honor the best new innovations of the past year from industry, academia and government-sponsored research organizations.
In what’s known as retrospective analysis, the team has looked at past data to see if their models can predict what they know already happened. If the models square up with history, the team will then try to predict the future. They expect initial results soon and hope to answer several key questions: Do these viruses strike big cities first? Or are rural areas where they emerge? Are standing water and moisture the key? Whatever answers come, they will move science one step closer to understanding how diseases spread — and, ultimately, help save more lives.
Nick Generous is a digital epidemiologist in the Information Systems and Modeling group at Los Alamos National Laboratory.