Five months since COVID-19 exploded in Wuhan, researchers still don’t know exactly how many people caught the disease in the city during those crucial early days of the outbreak. At the heart of this problem is the fact that testing in the city of 11 million people was very limited in those early days of the outbreak. By January 23, the first day of the city’s lockdown, just 495 cases had been officially confirmed.
As one researcher based in Hong Kong puts it, Chinese authorities were clearly aware that they had a much bigger problem on their hands than that number suggested.
“They decided to build the hospitals, they sent doctors to Wuhan,” says Daihai He, a mathematical epidemiologist at Hong Kong Polytechnic University, referring to two hospitals with capacity for more than 2,500 people that were constructed in early February in just 12 days.
Months later, estimates vary widely for how many people got COVID-19 in Wuhan. This is, frustratingly, a characteristic of any emerging disease or fresh outbreak of a known one. Unless testing captures cases very quickly and comprehensively, there will always be uncertainty about how things unfolded in those first few days or weeks. And yet, knowing how many people caught the disease in Wuhan in January could inform predictions of how the COVID-19 will spread elsewhere or even reveal clues as to why it affected the population of Wuhan so badly in the first place.
The confirmed case numbers being wrong is often seen as a failure or dismissive behavior from a government. Generally speaking, that’s not the problem.—Timothy Russell, London School of Hygiene and Tropical Medicine
He and colleagues published a preprint on medRxiv on March 2 in which they tried to estimate the number of infections in Wuhan by January 23. To do this, they counted cases exported from Wuhan to other Chinese cities, where, they assumed, more focused screening and testing was able to catch a greater proportion of positive cases than in Wuhan. They also used location data gathered from cell phones by Chinese tech giants Tencent and Baidu, allowing them to calculate the average number of people who traveled out of Wuhan each day in January.
The researchers’ estimate for the number of cases in the city ranges from about 1,000 to 5,000—well above the official confirmed tally of 495. This is a prediction of the number of cases in Wuhan that would have been picked up had testing there been as good as testing at travel points outside Wuhan.
While the range is in line with an analysis done early on by epidemiologists at Imperial College London, who reckoned there were about 4,000 symptomatic cases in Wuhan by January 18, other researchers have come up with significantly larger figures for the true number of cases. One paper in The Lancet published January 31 analyzed data on cases that were exported from Wuhan on international flights. The authors estimated that there were more than 75,000 cases in the city by January 25. To date, there have only been 83,000 confirmed cases reported across the whole of China. Another modeling study published as a preprint in late January pegged the number of cases closer to 11,000 by January 21.
Unfortunately, it’s difficult to favor any one estimate over another, says Peter Byass, a professor of global health at Umeå University in Sweden. There’s no consensus yet as to what happened in Wuhan and, using modelling alone, there may never be.
Nonetheless, there is still value in probing the question. Another paper, published in The Lancet in May by Timothy Russell, an infectious disease epidemiologist, and colleagues at the London School of Hygiene and Tropical Medicine takes a slightly different approach. The authors were interested in determining Wuhan’s Rt—the number of people who, on average, catch the disease from an infected individual at a particular time. This figure varies depending on things such as lockdown measures. Based on data about cases within Wuhan, infections detected elsewhere in China, cases exported to other countries from Wuhan, and the proportion of infected people on special evacuation flights, the authors estimated that the Rt was 2.35 a week before lockdown on January 23, but fell to 1.05 a week after lockdown was put in place.
They also used this analysis to estimate the exponential growth of cases in Wuhan. For January 23, they got a bigger figure than He’s team did, around 15,000 infections.
Estimates that indicate cases amounted to 5,000 or less by this point seem small, says Russell. He also notes that in the early stages of an outbreak it will always be difficult to understand the true nature of the situation. But retrospective analyses are, in part, valuable because they back up the idea that rolling out diagnostic testing really is essential in the early days of an outbreak, adds Russell.
Antibody testing may give researchers some idea of past prevalence later on but people may not retain antibodies long-term and antibody results also can’t tell you exactly when a person caught the disease.
Both Byass and Russell say that people should hold off on being overly skeptical of the official Chinese statistics. Public figures have openly questioned whether Chinese authorities have been honest about how many people contracted COVID-19—and how many died.
Probing the veracity of statistics is worthwhile but, Byass and Russell say, there is a danger of unfairly politicizing scientific results. For one thing, says Russell, COVID-19 is not an easy disease to capture with tests, even when those tests are widely available. Large numbers of people may never be tested, for example, because they may only have very mild—or no—symptoms.
“The confirmed case numbers being wrong is often seen as a failure or dismissive behavior from a government,” he says. “Generally speaking, that’s not the problem, the problem is a natural problem of how difficult this disease is.”
Other researchers have analyzed satellite images of hospital parking lots in Wuhan and reported a large increase in vehicles between August and December—months before COVID-19 is thought to have spread widely in the city. However, flaws have been found with the analysis, including the fact that some images, taken at sharp angles from orbiting satellites, don’t show the car parks in full.
Some have used information about crematoria, which were open around the clock at the height of Wuhan’s outbreak, to suggest that tens of thousands more people died in the city than reported. But there is no proof that the crematoria worked at capacity, burning many more bodies in January than suggested by official death figures. For instance, one Reuters article notes that one crematorium only dedicated one of eight furnaces to COVID-19 victims.
Analyzing proxy data such as travel records, satellite imagery, and others is valuable, but a better understanding of what happened in Wuhan will hopefully come from painstaking analysis of biological samples gathered in the city’s hospitals during December and January, says Peter Daszak, the president of EcoHealth Alliance, a US-based health research organization. Ideally, evidence of SARS-CoV-2, the coronavirus that causes COVID-19, will be found in such samples, enabling researchers to improve estimates of COVID-19’s prevalence in the city in those early days.
“The real answer will come out after that detailed, hard work has been done,” says Daszak.