The Predictive Pandemic – How AI Maps Global Flight Data to Stop the Next Outbreak.

Reading the early reports of an unidentified disease spreading through a city that most people can’t locate on a map evokes a certain kind of dread. Most of us recall that feeling from late 2019: wondering if a cluster of pneumonia cases in Wuhan, China, was important as we scrolled past short, almost dismissive news articles about it. It already did, according to a small Toronto business.

On December 31, 2019, BlueDot’s algorithm detected the Wuhan outbreak. It would take the World Health Organization nine more days to issue a public warning. The gap was actually a speed issue rather than a bureaucratic failure. Additionally, speed is practically crucial when it comes to infectious diseases.

BlueDot — Company Profile & Key Information

BlueDot’s concept is simple, but its implementation is powerful. Its AI system simultaneously gathers disease-related signals from thousands of sources, including news articles, livestock health bulletins, local health agency statements, and even airline ticketing data. It then uses machine learning to determine what is noise and what could be the start of something serious. Most importantly, it goes beyond detection. Before the first confirmed international case, the system creates probabilistic maps of disease spread by using flight route data to predict where an infected traveler is likely to go next. Airports are the modern nervous system of global disease transmission, according to Dr. Kamran Khan, who founded the company after working through SARS in Toronto hospitals in 2003. That is difficult to dispute.

BlueDot has a texture that pure tech startups typically lack thanks to Khan’s career. During his infectious disease fellowship, he witnessed the arrival of the West Nile virus in New York City. The virus had traveled across the ocean in a human or possibly a bird before making an unexpected appearance in Queens. When SARS struck Toronto, he was treating patients at St. Michael’s Hospital while observing the city’s transformation. Your perspective on risk is shaped by such front-line exposure. It also appears to influence the questions you choose to pose with data.

The approach is truly novel because of the flight data component. Over the course of their lives, mosquitoes only travel about a kilometer. However, in just eighteen hours, a Singaporean with dengue fever could be sitting in an emergency room in Toronto. Khan has repeatedly noted that the front-line clinician treating that patient might not have seen dengue before, might not have thought to ask the appropriate questions, or might not be aware of what they’re looking at. Part of the purpose of BlueDot’s system is to alert that clinician. to connect the dots before anyone in the hospital room realizes they need to.

All of this is becoming more difficult and happening more quickly due to climate change. The dengue, Zika, chikungunya, and yellow fever-carrying Aedes mosquito was formerly found in tropical regions. With the exception of Antarctica, it is now present on every continent. In regions of Mediterranean Europe and the American South without a history of the illness, dengue outbreaks are being reported in people who have never left their homes rather than returning tourists. In ways that would have seemed unlikely twenty years ago, tick-borne illnesses are spreading northward across Canada. These changing disease patterns are occurring in the real world, in actual communities, and sometimes even in backyards; they are not speculative future scenarios.

Attempting to keep up presents a significant challenge for AI systems. The technology has shown real promise, according to a scoping review of 33 studies on AI-based epidemic early warning systems published in Sage Journals last year. However, the review also found ongoing issues with data quality, model bias, and what researchers refer to as “explainability,” or whether the system can truly explain why it flagged something. When a public health official is trying to defend the mobilization of resources, a black box that cries wolf is not very helpful. These issues are likely solvable, but they haven’t been resolved yet.

In the larger scheme of things, there is also something worthwhile to sit with. In essence, BlueDot and similar systems are trying to make chaos more readable by searching through the cacophony of millions of daily human interactions, travel patterns, and biological events for a sign of an impending pandemic. The systems might improve to the point where they consistently matter and reliably reduce the time interval between emergence and response. It’s also possible that a vector or mechanism that no one has thought to develop a model for will cause the next significant outbreak. The gaps are often found by nature.

It’s difficult to deny that something genuine has changed, though, given how fast AI identified the Wuhan outbreak in comparison to conventional channels. A novel type of map was the 2019-nCoV visualization from Healthmap, which showed China covered in red and yellow dots that gradually spread to Tokyo, Chicago, and Paris. Not precisely where the illness was, but where it was headed. It may not seem like much, but the difference matters.