Original title: Do you really understand the daily epidemic update? “Digital games” are being played around the world
“Epidemiology is influenced by biological and sociological factors.
Each basic element is a scientific paper and
a research topic.”
The global “number game” in the new crown period
ends at 8 a.m. on April 7th. According to the New Crown Epidemic Data of Johns Hopkins University in the United States, more than 100,000 cases have been diagnosed in Germany, and the number in the UK is 52,000, which is not only lower than that in Germany. , Also lower than other European countries such as Italy and Spain. Looking at statistical data is the most important way for us to understand a country ’s epidemic situation, but just reading the conclusion that “the German epidemic situation is much more serious than that of the United Kingdom” from the digital surface, is this consistent with the real situation?
Many experts said that comparing the figures of countries in the epidemic is misleading and even dangerous. Sheila Bird, a professor at Cambridge University in the UK, is a well-known biostatistician. She said that the case detection and reporting system is inconsistent even within the country, let alone internationally. The rules about who is tested and when and where they are tested may change, and these changes may not be reflected in the data released by different governments in a timely and accurate manner.
In the UK, the government says that the current health system does not have enough capacity to detect every individual with new coronary symptoms, so only those who are seriously ill and need to go to the hospital for treatment will be tested. In this case, obviously many people with mild symptoms will Excluded from the statistics of infected people. In contrast, in Germany, anyone who has flu-like symptoms, someone who has been in contact with a confirmed case in the past 14 days, or who has been to a high-risk area can be tested.
Mark Lipsic, a professor at the Center for Infectious Disease Dynamics at the Harvard School of Public Health, wrote in the New England Journal of Medicine that if the scope of medical services or laboratory testing capacity bottlenecks limit our count of confirmed cases, or If only severe cases are tested, a simple count of confirmed cases may be a misleading indicator of the trajectory of the epidemic. People believe that numbers, like the reason it has always symbolized, are accurate, reliable, and undoubtedly more powerful than most text descriptions and personal experiences. But epidemiological statisticians and mathematicians reminded that when it comes to the new crown epidemic, looking at numbers alone can be very dangerous.
If the UK suddenly starts testing larger populations and discovers that new cases have increased significantly, this does not necessarily mean that the epidemic spreads faster; similarly, if a country reaches the upper limit of detection capacity, it may suddenly report new cases The decline, and these, are misleading figures. “The danger is that we rely on these data and assume that they can accurately calculate the total number of infections.” Mike Tilsley, associate professor at the University of Warwick, UK, said people need to pay attention to the bigger picture behind the numbers.
Schrodinger’s number The number that
people care about most every day — is the number of newly diagnosed cases, as expected, the number of people infected? What is “confirmed diagnosis”, is each country a definition? What if a human nucleic acid reagent test is negative but the kit sensitivity is only 60%? If a person gets the test result 5 days after the onset of illness today, on what day is the new case?
On March 27, Gao Ben’en, director of the Epidemiology and Biostatistics Division of the School of Public Health of the University of Hong Kong, and others published on the medRxiv website of the pre-printed medical paper platform the title “The Impact of Changing the Case Definition of COVID-19 on the Epidemiological Curve and Transmission Parameters in Mainland China “An article. The article sorted out several changes in the definition of “diagnostic criteria” for Chinese new coronary pneumonia, and found through modeling analysis that when this definition changed from the first to the second edition, from the second to the fourth edition, and from the fourth to the fifth edition At that time, the number of diagnoses increased by 7, 2.8, and 4.2 times, respectively. They estimate that if the fifth edition of the case definition is used throughout the outbreak, the number of people who are “confirmed” by February 20 will be as high as 232,000.
If this human factor is not taken into account, and only the apparent number changes, then Gao Benen and others pointed out that there will be a series of chain reactions in the calculation of propagation parameters, including overestimating the growth rate of the epidemic, and then overestimating the number of basic infections. The value of R0. For example, an article from the Guangdong Provincial Center for Disease Control and Prevention on February 13 calculated that the national new coronary pneumonia R0 value is as high as 4.5, but it is now generally believed that the R0 value should be around 3.
A data visualization map initiated and maintained by Dong Ensheng, a first-year doctoral student at the Center for Systems Science and Engineering (CSSE) at Johns Hopkins University in the United States, is a simple integration of the latest simple data such as confirmed diagnosis, death, and recovery cases in countries around the world. It attracts more than 1 billion pairs of eyes every day. Dong Ensheng told China News Weekly that although most of the time they quoted official data from various countries, they also encountered the problem of adopting a standard to count. For example, the early “confirmed cases” counted by the US Centers for Disease Control required a positive nucleic acid test plus CDC determination to be a confirmed diagnosis, but they used the “presumed positive case” estimation method, which is the first Cases for a while, but not confirmed by CDC, so the number is sometimes larger than CDC.
William Schaffner, an infectious disease expert at the Vanderbilt University Medical Center in Tennessee, said on March 9 that we are still in constant understanding of new coronary pneumonia, and the statistics are still in the preliminary stage. “(Existing data) are all estimates. What does your country think is a new crown case? The definition of each case is very different.”
Many experts pointed out that COVID-19 is called new crown pneumonia in China and abroad. Most of them are called new crown infections, so in China, asymptomatic infections are not classified as confirmed cases, but counted separately. According to the WHO’s definition, “whether or not there are clinical signs and symptoms, laboratory-diagnosed new coronavirus infections” are confirmed cases, and this standard is also used in Europe.
The data for “asymptomatic infected people” is another mystery. For example, a frequently cited data comes from an article published in the International Journal of Infectious Diseases on February 13. The paper conducted a study of 565 Japanese citizens evacuated from a charter flight in Wuhan, and found that 13 of the evacuees were infected, of which 4 were asymptomatic infections, the proportion was 30.8%. Jiang Qingwu, former dean of the School of Public Health of Fudan University, pointed out that this sample size is too small. If there is one more person who is asymptomatic, the proportion will increase to 38%.
Is the number of deaths certain? According to the British BBC report, the Ministry of Health of the country reported daily how many people who tested positive for new coronavirus died on that day, including the number of people who were infected with new coronavirus but may have died from other diseases. It contains the direct causes of death and possible causes of the deceased, as long as they are suspected to be related to the new crown, they will all be statistical.
In the United States, as of April 5th, public information showed that more than 9,400 people died of new crown infections, but experts in various fields believe this is an underestimated number. Because there is no unified system to report deaths related to the new crown, and there is a continuing shortage of detection capabilities, some state and county statistics on the number of deaths are ambiguous. According to the New York Times, the US federal government does not expect to know the final number of deaths due to the new crown until 2021, because the United States will not release an annual compilation of the main causes of death until then.
“Epidemiology is affected by both biological and sociological factors. Each basic element is a scientific paper and a research topic.” Jiang Qingwu said. Rather than simply counting, Gao Benn thinks we need to find a better way to track the progress of infectious diseases.
Case fatality rate: not a simple division.
Everyone wants to know how deadly the disease caused by this coronavirus is. Its term is the case fatality rate. In simple terms, it is the number of people who died of new coronary pneumonia divided by the total number of people infected with the disease . John Allen Paul, a professor of mathematics at Temple University, told China News Weekly that the number of deaths is relatively clear, but due to insufficient testing, especially the lack of random testing, the number of infected people is still unknown. He called this- “The mystery of the denominator.”
Italy has attracted media attention due to its high case fatality rate. Researchers from the Italian Higher School of Health published an article in the Journal of the American Medical Association on March 23 that analyzed three reasons for the country ’s high case fatality rate. One of the reasons is related to the molecule: Italy ’s case fatality statistics include the occurrence of deaths of positive patients in nucleic acid tests, regardless of whether his death was caused by an underlying disease. The scientific adviser of the Italian Ministry of Health recently pointed out that among the death certificates issued, only 12% of the causes of death have a direct causal relationship with the infection of the new coronavirus.
Let’s look at the denominator again. John Ioannidis, a professor of epidemiology at Stanford University in the United States, recently wrote that the data so far about how many people are infected and how the epidemic has progressed are “completely unreliable.” “We don’t know if the number of infected people who have not been caught is three times or 300 times, because three months after the new crown outbreak, most countries, including the United States, lack the ability to detect large-scale populations, and no country has Reliable data on the prevalence of the virus in a representative random sample. ”
According to public information, as of April 1, the total number of testers in Iceland has exceeded 19,500, and the total number of testers is close to 5% of the total population. Far higher than any country in the world. In other words, as of April 1, no country in the world has more than 5% of its population.
In the early days of the United States, there were many delays in the detection problem, which made the number of infected persons announced by the United States very few. As of 8 a.m. on April 7, the number of known diagnoses in the United States is 367,500, the country with the highest number of infections in the world. But how far this is from the real situation is still unknown. He Dayi, a professor at Columbia University School of Medicine and founder of AIDS Cocktail Therapy, said at a new crown epidemic sharing meeting held by the Brookings Institution, an American think tank, on April 3 that about two-thirds of the results of nucleic acid tests in the United States Positive results indicate that the current number of tests is far from sufficient.
Chen Jian, founder and CEO of Xinfeng Technology, also participated in this sharing session. During the epidemic, he has been making relevant model predictions. He explained to “China News Weekly” that the “two thirds” mentioned by He Dayi is called the diagnosis rate. Although this is not a basic indicator of epidemiology, it can be used to measure the degree of undertesting in a region. Using data from outside Hubei Province of China and South Korea as a reference, he believes that one of 25 to 30 people diagnosed (3% to 4%) is more reasonable data, lower than this number, such as 80: 1 before Hong Kong, It means that the scale of the test is a bit too large; above this number, it means that the test is inadequate. There are few states that can reach this reasonable rate in the United States, including New York State, which has the largest number of testers, because the number of infections is too large, so the rate is still very high.
In Africa, as of the afternoon of April 3, 49 countries reported a total of 7123 confirmed cases of new crowns and 289 deaths. However, these countries have little ability to produce new corona virus detection equipment. The total number of confirmed diagnoses is still unknown.
Even if the numerator and denominator are well controlled, there are still other calculation “traps” to be wary of. Liang Zhuowei, Dean of the Li Ka-shing School of Medicine at the University of Hong Kong, pointed out that it is a completely wrong calculation method to roughly divide the cumulative deaths of the new crown by the cumulative number of diagnosed patients. “Because many denominators are recently diagnosed cases, their course At the beginning, it is impossible to judge whether he will recover or die after the condition deteriorates. “The results of a study published by his team on March 6 believe that the upper limit of the mortality rate of new coronary pneumonia is 1.4%.
“Based on what we have learned so far, the mortality rate of new coronary pneumonia does not seem to be as good as other coronavirus infections and diseases that have evolved into major epidemics in recent decades. But as more data arrives, our understanding of the situation may also require Change. “John Allen Paul said.
wary of simple comparisons.
Be “Currently, data from different countries have great deviations, so these data cannot be directly compared.” Dietrich Rothenbacher, director of the Institute of Epidemiology and Medical Biostatistics at Ulm University in Germany, pointed out in an interview with the media that The real and comparable figures come from a well-defined, systematic and representative sample.
But when reading the following paragraph, we still can’t restrain the “temptation” of comparison. According to a report by Time Weekly on March 30, Germany ’s latest number of diagnoses exceeded 63,000, one of the top five countries hit hardest by the epidemic, but only 560 people died of the disease, meaning that its mortality rate was only At 0.9%, it is the country with the lowest mortality rate in the world-especially when the mortality rate of Italian diagnoses is as high as 11%, this data looks a bit unusually low. As of the morning of April 6, Germany ’s figure remained at 1.6%, which was lower than Italy ’s 12%, Spain, France and the UK ’s 10%, and the US ’s 3%.
In fact, Germany is one of the fastest countries in finding infected people. Dietrich Rothenbacher, director of the Institute of Epidemiology and Medical Biostatistics at Ulm University in Germany, believes that the low mortality rate in Germany is related to its extensive testing. As early as January, Germany was one of the first countries to develop a reliable method for detecting new coronaviruses. Because the German health system is managed at the state level, it is not restricted by the central regulatory agency, and private companies quickly start mass production of testing equipment. Now Germany can complete 160,000 samples every week, much higher than other European countries. Detecting people with virus infections as much as possible is equivalent to increasing the “denominator” and statistically “decreasing” the case fatality rate.
In contrast to Italy, in the early stages of the epidemic, Italy first implemented an extensive testing strategy. People who have been in contact with the infected person will be tested regardless of whether they have symptoms, but the country ’s Ministry of Health decided on February 25 to narrow this test pool: Priority is given to testing suspected patients who require hospitalization and have more severe clinical symptoms. This test policy has resulted in a high percentage of positive results. As of mid-March, this figure was 19%, and the case fatality rate increased significantly-because those with milder symptoms, that is, those with lower mortality rates, were not included Into the denominator.
Even if we do not discuss the “outside statistics” figures, simply dividing the known number of deaths and the number of diagnoses, this figure is still not comparable. “There is no indicator that allows us to compare the situation in different countries.” Mike Tillsley, an associate professor at the University of Warwick, UK, said the mortality rate can tell the severity of the situation in each country, but the international comparison is very complicated. No two countries are the same in terms of access to medical resources, basic health status or even the age structure of the population.
In response, Mike Ryan, executive director of the WHO Health Emergencies Program, said that the new crown outbreak in South Korea is related to a church, so the age distribution of infected people is lower than in northern Italy; and outbreaks in long-term care centers At times, the average age of infected people will be much higher.
In 2019, nearly a quarter of Italy’s population is over 65 years old, while China’s age structure accounts for only 11%. As of mid-March, Italy’s total case fatality rate was 7.2%, much higher than China’s 2.3% over the same period. Researchers at the Italian Higher School of Health pointed out that if the age is stratified, from 0 to 69 years old, the fatality rate in these two countries is comparable. For example, according to the report of the WHO-China Joint Inspection Mission, among the laboratory-confirmed cases in China, the mortality rate of the elderly aged 80 or above was 21.9%, and that in Italy was 20.2%.
In South Korea, about one-third of confirmed cases are 30 years of age or younger, and the mortality rate is 1.6%. Although the German and Italian populations have the same average age, a large number of tests have lowered the average age of infected persons in Germany to 46 years, compared with 63 in Italy.
Chen Jian said that in addition to statistical reasons and age structure, he pointed out that the first cause of high mortality in Italy is that the serious medical conditions cannot keep up. In northern Italy, there were once more than 1,200 patients pouring into the intensive care unit, which is an alarming number. Although the number of ICU beds in Italy is not small in Europe, about 12 per 100,000 people, but because the outbreak is concentrated in a certain place, the number of beds in the country cannot be used. In comparison, Germany has a figure of 28, the Netherlands has a figure of 7, and the figure in India is about 6.5. Chen Jian analyzed that the figure of China is only 3.6, which is not high, but the rapid establishment of Vulcan Mountain and Thor Mountain makes up for this defect.
When asked about the mathematician John Allen Paul, in this epidemic, scientists should continue to work on the most valuable data, he listed a bunch of data, including diseases for different ages, economic status, The influence of gender, the relationship between access to medical facilities and disease … “No matter how much we want to know all the facts about coronavirus, we still have a lot of unknowns, and we should accept the discomfort caused by this uncertainty. ”
Journalist / Peng Danni