Machine learning can predict dengue fever with 80% accuracy

Northeastern University researchers can predict the appearance of dengue fever with 80% accuracy. This is a breakthrough for public health officials tasked with preparing care workers to handle disease spikes.

Almost half of the world's population lives in places where mosquito-borne dengue fever may occur, with an uptick in cases worldwide doubled between 2023 and 2024.

About 40,000 people die from the virus each year, according to US national data.

“We wanted to reduce the cognitive load of decision makers who wanted to extract the best predictions from multiple mathematical models,” says Mauricio Santirana, professor at the Mechanical Intelligence Group in Northeastern for improved health and environment. “There's a computational model called an ensemble method to do this.”

Machine learning is used to analyze existing dengue prediction models, and researchers identify the most accurate predictions for a particular region. They see how they perform accurately over three months and decide they are the most accurate over the next three months, says Santillana.

Another approach is to find consensus among the available predictions. Different ensemble models perform better under different conditions, he says.

Tracking and predicting the outbreak of disease is difficult, says Santilana. Because countries have different ways of reporting cases. In some countries, there is no reliable funding for testing for diagnostic verification. However, even when data reporting was delayed or incomplete, the ensemble method was consistently ranked among the top three models when tested positively in 180 locations around the world for over a year.

This study was published in Proceedings of the National Academy of Sciences. This approach has been tested in parts of Brazil, Malaysia, Mexico, Thailand, Peru and Puerto Rico.

The virus thrives in tropical and subtropical regions, says Michael Johansson, professor of public health research in Northeastern. Johansson recently completed a study on how dengue outbreaks spread in North and South America. Areas like Puerto Rico can quickly move from thousands of reported cases to 20,000 cases, he says.

Previous studies conducted in Southeast Asia found that dengue epidemics tend to occur simultaneously in eight local countries. Johansson wanted to know if the same pattern applies to America.

Johansson's study examined data from 14 US countries between 1985 and 2018. It was published in the journal Science Translation Medicine.

“We were able to see peaks occurring at different times across the region,” he says. “It travels north in orbit early in the Southern Hemisphere calendar year throughout the year.”

Health officials in areas prone to dengue fever will find clues about upcoming outbreaks by looking at what's going on nearby, Johansson says.

“It's a network of places that really matter,” he says. “We don't just look at temperatures and what's going on in El Niño, for example, but what's going on in our neighbouring countries.”

He says that driving these patterns remains unknown, but it could be a combination of climate and human mobility. These factors can also alter socioeconomic factors and contribute to the rise in dengue fever in recent years. The outbreak of dengue fever reflects the way humans and mosquitoes interact, and he reflects changes in garbage collection or water storage.

Furthermore, a major change in human travel patterns is one of the main ways that dengue fever spreads. If a traveler becomes infected and is bitten by a mosquito in a new location, the mosquito can be infected and biting others.

“People are moving around a lot more than they used to,” says Johansson. “This means the virus is moving around with far more people than before.”

This story has been republished courtesy of Northeastern Global News news.northeastern.edu.