Coronavirus COVID-19, dengue and Zika are all viruses that have caused havoc in various parts of the world in the recent past. But what if we could predict where and when an infectious disease will break out next? Experts like AAAS Member Michael Johansson, a research biologist for the U.S. Centers for Disease Control and Prevention (CDC), are working toward exactly that.
"It's very hard for human beings to make really good predictions about the future. There are a lot of pieces to the puzzle," says Johansson.
Yet, advance warnings of outbreaks could help in a number of ways. Accurate predictions could pinpoint ideal times to promote vaccination in a threatened population or give insight into when to launch campaigns to curb mosquitoes or other vectors (organisms that don’t cause a disease but carry and transmit it from one host to another).
If scientists could better predict disease spread, pharmaceutical companies could also ramp up production of drugs to combat it. Hospitals could train personnel and make other preparations for treating patients.
Johansson and colleagues have found that analytical tools like computational models show great promise in predicting disease outbreaks. "A quantitative model allows you to develop a much more complex system" for analyzing all the data and components that contribute to risk, he says. This was his message to attendees during a talk he gave at the 2020 AAAS Annual Meeting in Seattle, Washington.
“There's a lot of research out in the community that could potentially be applied to some of these problems, but it's often really hard to see how valuable that research is,” Johansson says, because it has often been done in different ways, using different data. A good model allows researchers to identify useful research, understand why it's valuable and put it to work tracking disease in real time, he says.
As the Epidemic Analytics Team lead in the CDC’s Dengue Branch, based in San Juan, Puerto Rico, Johansson ordinarily works with "arboviral," or insect-borne, viruses like dengue, a painful and potentially fatal disease transmitted by Aedes aegypti and albopictus mosquitoes. Lately though, he's been part of the CDC Modeling Team investigating the developing outbreak of the coronavirus COVID-19, a respiratory disease first identified last year in Wuhan, China.
COVID-19 has virtually shut down parts of China, where it has caused more than 2,000 deaths; cases have been confirmed in more than 24 other countries and the U.S. International Health Regulations Emergency Committee of the World Health Organization (WHO) declared the outbreak a “public health emergency of international concern.”
The round viruses of the coronavirus family appear to have a halo or crown of spikes when viewed under a microscope. ("Corona" in Latin means "crown.") This family of viruses was first identified as a human pathogen in the 1960s, and COVID-19 is an example of an emerging, or previously unknown, disease. Severe Acute Respiratory Syndrome (SARS) and Middle East Respiratory Syndrome (MERS) are two dangerous coronaviruses that previously put the world on edge when they burst onto the scene in 2003 and 2012, respectively.
Modeling for disease outbreaks—like coronaviruses COVID-19, SARS and MERS—uses some techniques developed for weather forecasting, like probabilistic forecasting. However, unlike the weather, the course and severity of a communicable disease like COVID-19 or the flu can be affected by human intervention. Impact can be lessened by a successful vaccination effort, for example, or it could be made worse by infected people traveling to a new location with a vulnerable population. Some diseases, like Zika, can be transmitted not only by a vector like a mosquito, but also through human-to-human contact, adding to the variables.
In 2019, Johansson worked with 16 teams that had their own methods and compilations of data in an attempt to forecast peak incidence, the week of the peak, and total incidence of dengue over eight disease seasons in Peru and Puerto Rico. That study identified strengths and weaknesses of the modeling effort and some techniques that showed real promise.
One of those techniques was the use of the disparate teams themselves, an approach called ensemble modeling. It turns out that models of disease outbreaks, like those of weather events, give more accurate predictions if they reflect the pooled results of a number of teams working separately, employing the methods they're most comfortable with. Ensemble modeling has worked with every disease his group has tried it with, Johansson says.
Ultimately, Johansson hopes to "see arborviral diseases controlled, and to see a more regular use of analytics and data science in infectious diseases,” he says. “I think they can give us new insights into those diseases."
Johansson says he values being an AAAS member as "a way to stay engaged with the broader world of science." The work he and his colleagues are doing to improve probabilistic forecasting for communicable disease will often be conducted behind the scenes, he says, adding, "but when successful, it will be behind important public health decisions that affect the average person."