Meet the Hong Kong forecaster who predicted spread of coronavirus cases in US

A computer engineer who designed a model to forecast the number of new Covid-19 cases has been hailed by Chinese internet users for the accuracy of his predictions.

Jonathan Li Zhibin said he had been following the epidemic closely from the start because of his memories of severe acute respiratory syndrome (Sars) in his hometown of Hong Kong and because of his close connections to Wuhan, which he visits regularly to see friends and family.

Forecasts published on Li’s blog at the end of March for the number of new coronavirus cases in the United States and European Union in April turned out to be more than 90 per cent accurate.

He estimated on March 27 that the number of cumulative cases for the US on April 3 would be 253,273. The reported number according to the US Centres for Disease Control and Prevention for that day was 277,205, giving his forecast a 91 per cent degree of accuracy.

“In January, only Wuhan and Hong Kong had coronavirus data. Hong Kong had a bad experience with Sars, all Hongkongers are very sensitive to outbreaks like this, including me. So I wanted to find out why that was early on,” Li said.

Working by himself, Li started collecting and monitoring data on the epidemic when Hong Kong’s Department of Health started releasing daily figures in January. He then began to add figures from mainland China and the rest of the world as it became available.

“It was a painful process to collect data, quite tiring. Slowly I got used to it after several months,” he said.

China cancels coronavirus clinical trials due to shortage of patients

The computer engineer – who graduated from Tsinghua, one of China’s top universities, in 1985 – specialises in data analysis and forecasting and works for Hong Kong-based e-jing Technologies

He credits the accuracy of his modelling to the range of data he collected, factoring in up to 100 other variables in his forecasts.

These included 30 or so key factors, such as population density, demographics, the age of a city’s infrastructure, sewage, plumbing and weather.

“Forecasting is based on history, based on the past to tell your future,” he said.

Does air con help spread coronavirus? Chinese study of 3 families in a restaurant suggests so

His predictions have been widely shared across Chinese social media sites, with some blog posts receiving more than a million views on Weibo, and many internet users praised the accuracy of his model.

In January, Li predicted at that time that Covid-19 would peak in February and mostly die down by March or April in China. However, he did not publish estimates for mainland China publicly because of the sensitivity of the topic. But he shared them with his former classmates in a chat group.

Li noticed from his modelling that several factors were common to Covid-19 hotspots such as Wuhan, Daegu in South Korea, Milan in Italy and New York in the US.

All these cities have an elderly population, ageing infrastructure and are wet and cold during the winter – with average temperatures ranging from 5 to 15 degrees Celsius (41-59 Fahrenheit).

Li supports the South Korean model for fighting the epidemic by conducing mass testing and quickly isolating and treating those infected with the coronavirus that causes Covid-19.

China’s initial coronavirus outbreak in Wuhan spread twice as fast as we thought, new study suggests

He also used South Korean data to correctly forecast that the US would see the rate of growth of new cases slowing on April 6.

“Based on South Korea’s data, when the US has done 2 million coronavirus tests it would see a dramatic slowdown [in the growth rate]. Today, the US has done almost 1.8 million tests, it’s in line with what I predicted before,” he wrote in a March 31 blog post.

The growth rate for April 4-5 was 12.43 per cent, but dropped to 8.13 per cent on April 5-6. It had been between 12 and 20 per cent before that, according to Li.

He said the world had underestimated the severity of Covid-19 and argued that one reason was a lack of understanding of the mathematical modelling that predicted the course of the outbreak when case numbers were low.

He called for more data analysis and forecasting experts to be involved in public health policymaking and to build on the expertise of epidemiologists to provide modelling that can inform better control measures.

“Forecasting is not just a crystal ball, it’s not as simple as telling you something. It’s about helping society to prepare as a whole and save lives,” Li said.