UK-Israeli scientist’s algorithm can predict Covid outbreaks 17 days in advance

An algorithm developed by a British-Israeli can now predict which English towns are headed for a COVID-19 outbreak, more than 2 weeks before it happens.

The complex technology, built by epidemiologist, Professor Michael Edelstein, monitors Google searches to help English health officials detect such outbreaks 17 days ahead.

The journal Nature Digital Medicine said that England’s success in getting ahead was by using the algorithm, which has been taught what terms people search for when they begin to feel symptoms of the virus. Yet the data analysis doesn’t compromise privacy, as no personal search information is delivered to researchers. The tech has been tested and proved to work in Italy, Australia and South Africa.

Professor Edelstein, who began developing the tech alongside a team from University College London, before moving to Israel to take up an academic post at Bar Ilan University’s medicine faculty, said that “our best chance of tackling health emergencies such as the COVID-19 pandemic is to detect them early in order to act early.”

Edelstein, who was a senior official at Public Health England when he started the project, told The Times of Israel that the algorithm works by taking “clinical reports from cases of COVID and see what symptoms people report. You look at what symptoms they are concerned about, and see how exactly they describe the symptoms when searching. You can then use those search terms to see when COVID is likely to rise.”

“We then found ways to remove what we call the ‘noise,’ like searches that are related to news coverage rather than people feeling unwell. We applied the algorithm in countries that are ahead of others in outbreak progression and compared them to those that are behind, which gave us more information. What we got in the end is a very good predictor of places where COVID cases are set to rise in about two weeks’ time.”

He believes that the algorithm’s success shows that “using innovative approaches to disease detection such as analysing internet search activity to complement established approaches is the best way to identify outbreaks early.”