As data scientists, we must be able to analyze and interpret data. And we were very concerned about the results of the data analysis regarding covid-19. The most at risk are the most vulnerable - the elderly and people with lower than average incomes - but to control the spread and impact of the disease, we all need to change our behavior. Wash your hands thoroughly and regularly, avoid crowds, cancel events, and avoid touching your face. In this message, we will explain the reason for our concern, and tell you why you should also be concerned. A summary of the key information can be found in Ethan Alley's post. (The author is the president of a non-profit organization that develops technologies to reduce the risk of pandemics).
Contents:
1. We need a working medical system
Just 2 years ago, one of us (Rachel) contracted an infection that attacks the brain and kills ¼ of those infected, and also leads to cognitive impairment in one in three of those infected. Many who have been ill receive permanent hearing and vision impairments. Rachel was delirious by the time she got to the hospital. She was lucky to receive timely medical care, diagnosis and treatment. Right before this event, she felt great, and her life was certainly saved due to quick access to the emergency room.
Now let's talk about covid-19 and what could happen to people in a situation like Rachel's in the coming weeks and months. The number of detected cases of covid-19 infection is doubling every 3-6 days. If we take this period equal to three days, the number of infected people will increase 100 times in three weeks (in fact, everything is not so simple, but let's not be distracted by technical details). One in 19 infected people requires long-term (many weeks) hospitalization, and most of these patients require oxygen. Although the spread of the virus is just beginning, in some regions hospitals are already overcrowded and people are unable to receive the treatment they need (in various conditions, not just in case of covid-16 infection). For example, in Italy, where just a week ago the authorities said that everything was fine, now 6 million people are in quarantine (upd: XNUMX hours after the publication, the entire country was closed). To help deal with the influx of patients, they set up tents like this one:
Dr. Antonio Pesenti, head of the regional crisis center in Italy's hardest-hit area, says: "We have to set up intensive care units in corridors, operating rooms and rehabilitation rooms... One of the best healthcare systems in the world, in Lombardy, is on the verge of from collapse."
2. It's not like the flu
The mortality rate for influenza is approximately 0,1%. Marc Lipsitch, director of the Center for Communicable Disease Dynamics at Harvard, gives for covid-19 at the level of 1-2%. gives a mortality rate of 1,6% for China in February, i.e. 16 times higher than for influenza1 (this can be a very conservative estimate, since mortality rises sharply when the health system fails to cope). The best estimates so far say that covid-19 will kill 10 times more people this year than the flu (and Elena Grewal, a former data science director at Airbnb, estimates 100 times more than the flu at worst.) And all this without taking into account the important influence of the health system, as mentioned above. One can understand why some people convince themselves that nothing new is happening and that it is a flu-like illness. It is very uncomfortable to realize that in fact they have not encountered this at all.
Our brains are not equipped to intuitively perceive the exponential growth in the number of infected people. Therefore, we will carry out the analysis as scientists, without relying on intuition.
Each person infected with influenza on average infects 1,3 other people. This indicator is called R0. If R0 is less than 1, the infection stops spreading, and if it is greater than 1, it continues to spread. For covid-19 outside of China, the R0 is now 2-3. The difference may seem insignificant, but after 20 "iterations" of infection in the case of R0=1,3, the number of infected people will be 146 people, and for R0=2,5 - 36 million! These are simplified calculations, but they serve as a reasonable illustration. relative difference between covid-19 and flu.
Note that R0 is not a fundamental disease characteristic. This indicator is largely dependent on the response [to the disease] and can change over time2. For example, in China, the R0 for covid-19 is declining rapidly and is now reaching 1! How is this possible, you ask? By applying measures the scale of which is hard to imagine in countries like the United States - for example, the complete lockdown of many giant cities and the development of diagnostic procedures that allow testing a million people a week.
On social media (including popular accounts like Elon Musk's) one can often see a misunderstanding of the difference between logistical and exponential growth. Logistic growth in practice corresponds to the S-shaped curve of the spread of the epidemic. Of course, exponential growth also cannot continue indefinitely, since the number of infected is always limited by the population of the Earth. As a consequence, the incidence rate should decrease, eventually leading to an S-curve (sigmoid) for growth rate versus time. However, the reduction is achieved in certain ways, not magically. Main ways:
- mass and effective reaction of the society;
- the proportion of cases is so high that there are too few people who are not sick for further spread of the infection.
Thus, it is unwise to refer to the logistic growth curve as a way to "control" the pandemic.
Another difficult aspect to intuitively understand the impact of covid-19 on the local community is the very significant delay between infection and hospitalization - usually around 11 days. It may not seem like a very long time, but such a period means that by the time all hospital beds are filled, the number of infected will be 5-10 times higher than the number of hospitalized.
Note that there are some early signs of climate influence on the spread of infection. In publication they say that so far the disease is spreading in temperate climates (unfortunately for us, the temperature in San Francisco, where we live, is just in the right range; densely populated regions of Europe, including London, fall here).
3. Don't Panic, Stay Calm Doesn't Work
On social media, people pointing out reasons for concern are often told to "don't panic" or "keep calm". This is at least useless. No one claims that panic is an acceptable response. But there are reasons why "keep calm" is a common response in certain circles (but not among epidemiologists, whose job it is to keep track of such things). Perhaps "keep calm" makes people more comfortable with their own inaction, or it makes them feel superior to those they think are running around like headless chickens.
But “keeping calm” can easily get in the way of preparing and responding appropriately. In China, tens of millions of citizens were isolated and two hospitals were built by the time the disease statistics reached the level now observed in the United States. Italy has waited too long and just today (March 8) 1492 new cases and 133 new deaths have been reported despite 16 million people in quarantine. Based on the best information available to us, just 2-3 weeks ago, Italy was at the same level as the US and UK in terms of incidence statistics.
Please note that we have little knowledge about covid-19 at this stage. In fact, we don't know what its rate of spread or mortality is, how long it persists on surfaces, whether it can survive and spread in warm conditions. All we have is speculation based on the best information we have gathered together. And remember that most of the information comes from China in Chinese. Right now, the best source for understanding the Chinese experience is the report based on collaborative work of 25 experts from China, Germany, Japan, Korea, Nigeria, Russia, Singapore, USA and WHO.
In the face of such uncertainty that there will be no global pandemic and that's all, perhaps, will do without the collapse of the health care system, inaction does not seem the right response. This would be extremely risky and sub-optimal in any simulated scenario. It also looks unlikely that countries such as Italy and China have actually shut down significant segments of their economies for no good reason. And the lack of action is inconsistent with the actual impact we are seeing in affected regions where the medical system is unable to cope (e.g. Italy uses 462 triage tents and still requires .
Instead, a considered and sensible response is to follow the steps that experts recommend to prevent the spread of infection:
- Avoid large events and crowds
- Cancel events
- Work from home whenever possible
- Wash your hands when you get home and when you go out and spend time outside the home
- Try not to touch your face, especially when you're out and about (it's not easy!)
- Disinfect surfaces and packaging (it is possible that the virus remains active on surfaces for up to 9 days, although this is not known for sure).
4. It's not just about you
If you are under 50 years of age and do not have risk factors such as a weakened immune system, cardiovascular disease, a history of smoking in the past, or chronic illness, you can reasonably be sure that COVID19 is unlikely to kill you. But your reaction to what is happening is still extremely important. You still have the same chance of getting infected as everyone else, and if you get infected, there is just as much chance of infecting others. On average, each infected person infects more than two people, and they become contagious before symptoms appear. If you have parents you care about or grandparents and plan to spend time with them and then find yourself responsible for infecting them with the COVID19 virus, it will be a heavy burden.
Even if you don't have contact with people over 50, chances are you have more colleagues and acquaintances with chronic diseases than you think. that few people disclose their health at work if they can avoid it, . We're both [Rachel and I] in the high-risk category, but many of the people we interact with on a regular basis may not have been aware of this.
And, of course, it's not just about the people in your immediate environment. This is a very important ethical issue. Every person who does everything possible to fight the spread of the virus helps the entire community as a whole to reduce the infection rate. As Zeynep Tufekci wrote in : "Preparing for the almost inevitable global spread of this virus... is one of the most pro-social, altruistic things you can do." She continues:
We need to prepare, not because we personally feel at risk, but to help reduce the risk for everyone. We must prepare not because we are facing a doomsday scenario beyond our control, but because we can change every aspect of this risk that we face as a society. That's right, you must prepare because your neighbors need your preparation - especially your elderly neighbors, neighbors who work in hospitals, neighbors with chronic illnesses, and your neighbors who may not have the means or time to prepare.
It affected us personally. The most extensive and most important course we've ever created on fast.ai, which represents the culmination of our years of work, was due to start at the University of San Francisco in a week. Last Wednesday (March 4) we made the decision to move it all online. We were one of the first major courses to go online. Why did we do it? Because we realized early last week that if we did this course, we would be indirectly encouraging hundreds of people to gather in an enclosed space several times over the course of a few weeks. The worst thing to do is to gather groups of people in a closed space, and it was our moral duty to avoid this. The decision was difficult, because our work with students every year was for us the greatest pleasure and the most productive period. And there were students who were going to fly in from abroad, whom we did not want to let down3.
But we knew that we were doing the right thing, because otherwise we would have contributed to the spread of the disease in our community4.
5. We need to make the curve flatter
This is critical because if we can reduce the rate of infection in the community, it will allow hospitals to handle both the influx of infected people and their regular patients. The illustration below clearly shows this:
Farzad Mostashari, former National Health IT Coordinator, explains: “New cases among non-travelers and non-contacts are detected every day, and we know this is just the tip of the iceberg due to delays in testing. This means an explosive increase in the number of infected in the next two weeks ... Trying to contain the exponential spread in society is like focusing on putting out sparks when the whole house is on fire. When this happens, we need to switch to easing - take protective measures to slow the spread and reduce the peak public health impact." If we keep the rate of spread low enough, hospitals can handle the load and patients get the treatment they need. Otherwise, those in need of hospitalization will not be hospitalized.
According to :
The United States has approximately 2,8 hospital beds per 1000 people. With a population of 330 million, this results in approximately 1 million beds, 65% of which are permanently occupied. Thus, a total of 330 thousand beds are available (perhaps a little less due to seasonal flu, etc.). Let's trust the experience of Italy and assume that approximately 10% of cases are severe enough to require hospitalization. And remember that often hospitalization lasts for weeks - in other words, beds with patients with COVID19 will be released very slowly. According to such estimates, all hospital beds will be occupied by May 8. And at the same time, we do not take into account the suitability of these beds for the maintenance of patients with viral diseases. If we are wrong about the proportion of severe cases by 2 times, this shifts the saturation time of hospitals by only 6 days in one direction or another. All this does not imply an increase in the need for places due to other reasons, which looks like a dubious assumption. With the growing pressure on the healthcare system and the emergence of a shortage of prescription drugs, people with chronic diseases may find themselves in a situation requiring care and hospitalization.
6. Public reaction matters
As discussed, these numbers are not certain - China has already demonstrated that extreme measures can reduce the spread of the disease. Another excellent example is Vietnam, where, among other things, a nationwide advertising campaign (including a haunting song!) quickly mobilized the population and brought about the required behavioral change.
These calculations are not hypothetical - everything was tested during the flu pandemic in 1918. In the US, two cities reacted completely differently: Philadelphia held a giant parade of 200 people to raise money for the war. But in St. Louis, social contacts were minimized to reduce the spread of the virus, and all mass events were canceled. This is what the number of deaths in each of the cities looked like according to the data :
The situation in Philadelphia has become extremely difficult, to cope with the huge number of deaths.
Richard Besser, who was the executive director of the Centers for Disease Control and Prevention during the H1N1 pandemic in 2009, says that in the US, “the risk of infection and the ability to protect oneself and one’s family depend, among other factors, on income , access to health care and immigration status". He claims:
The elderly and people with disabilities are at particular risk when their daily lives and support systems are disrupted. Those who do not have easy access to health care, including rural and indigenous populations, may be faced with the need to travel great distances when necessary. People living in cramped conditions—whether social housing, nursing homes, prisons, orphanages (or even the homeless on the streets)—can be hit by waves, as we've seen in Washington State. And vulnerable sectors of the economy with low wages, unpaid workers and unstable work schedules will be exposed to the public during this crisis. Ask the 60 percent of the U.S. workforce who get paid by the hour how easy it is to turn down a job when needed.
The US Bureau of Labor Statistics shows that with the lowest income have access to paid sick leave:
7. We in the US are ill-informed.
One of the big problems in the US is that there are very few tests for coronavirus, and the results of the tests are not properly distributed, and we do not really know what is really happening. Scott Gottlieb, a former Food and Drug Administration (FDA) commissioner, explained that Seattle had better testing and that's why we're seeing the infection there: "The reason we knew about the COVID-19 outbreak early in Seattle, consisted in the work of sanitary-epidemiological surveillance [sentinel surveillance] by independent scientists. Such oversight has never been carried out on a similar scale in other cities. So other US hotspots may not yet be fully discovered.” According to the message , Vice President Mike Pence promised "roughly 1.5 million tests" would be available this week, but fewer than 2000 people have been tested in the US so far. Based on results , Robinson Meyer and Alexis Madrigal of The Atlantic state:
The evidence we have collected suggests that the US response to the COVID-19 virus and the disease it causes has been extremely sluggish, especially when compared to other developed countries. Eight days ago, the Centers for Disease Control and Prevention (CDC) confirmed that in the United States the virus is spreading among people, namely that it infected Americans who did not travel abroad and did not communicate with those who did. In South Korea, more than 66 people were tested within a week of the first case of infection, and it quickly became possible to test 650 people a day.
Part of the problem is that it has become a political issue. President Donald Trump has made it clear that he wants to keep the number of people infected in the US low. This is an example of how metrics optimization gets in the way of getting good results in practice (for more on this issue, see the Data Science Ethics article - ). Google AI CEO Jeff Dean tweeted their concerns about politicized disinformation:
When I worked for the World Health Organization (WHO), I was involved with the Global AIDS Program (now UNAIDS), created to help the world deal with the HIV/AIDS pandemic. Conscientious doctors and scientists worked there, focused on helping to overcome this crisis. In times of crisis, clear and reliable information is important to help you make an informed decision about how to respond (at all levels: country, state, local government, company, NGO, school, family, and individual). With access to the right information and advice from the best medical and scientific experts, we will get through the challenges, whether it be HIV/AIDS or COVID-19. But in the case of disinformation driven by political interests, there is a huge risk of seriously exacerbating the situation, if not act quickly and decisively in the face of a growing pandemic, but rather actively contribute to the faster spread of the disease. It's very painful to watch all of this happening right now.
It does not seem that there will be political forces interested in the transparency of the situation around COVID-19. Health and Human Services Secretary Alex Azar, , “began discussing the tests healthcare workers use to determine if one of them is infected with the new coronavirus. But the lack of such tests means a lack of information about the spread and severity of the epidemiological disease in the US, exacerbated by lack of transparency on the part of the government. Azar mentioned that the new tests are now undergoing quality control.” But further, according to Wired:
Trump then interrupted Azar. “I think the most important thing is that anyone who needs to be tested gets it. There are tests, and they are good. Anyone who needs to be checked is checked, ”Trump said. It is not true. On Thursday, Vice President Pence told reporters that the U.S. does not have enough COVID-19 tests to meet demand.
Other countries are reacting much faster than the US. Many Southeast Asian countries are doing a good job of containing the spread of the virus. For example, Taiwan, where R0 has now dropped to 0.3, or Singapore, which how the state should respond to COVID-19. This is not just about Asia; in France, for example, any event with 1000 or more participants is banned, and schools are currently closed in three districts.
8. Заключение
COVID-19 is an important social issue and we all not only can, but must do our best to reduce the spread of the disease. For this:
- Avoid large events and crowds of people (social distancing)
- Cancel cultural and other public events
- Work from home whenever possible
- Wash your hands when you get home and when you go out and spend time outside the home
- Try not to touch your face, especially when you are away from home
Note: Due to the need to publish this message as soon as possible, we have been a little less than usual rigorous in citing the sources of information we rely on. Please let us know if we have missed something.
Thanks to Sylvain Gugger and Alexis Gallagher for their valuable comments.
Notes:
1 Epidemiologists are people who study the spread of disease. It turns out that estimating things like mortality and R0 is actually quite difficult, which is why there is a whole field that specializes in this. Beware of people who use simple ratios and statistics to tell you how covid-19 behaves. Instead, look at the modeling done by epidemiologists.
2 This is technically incorrect. Strictly speaking, R0 refers to the level of spread of infection in the absence of a response. But since this is not really what we care about, we allow ourselves to be a little sloppy in definitions.
3 After this decision, we worked hard to find a way to launch a virtual course that we hope will be even better than the face-to-face version. We have been able to open it up to everyone in the world and will be working with virtual study and project teams every day.
4 We also made many other small changes to our lifestyle, including exercising at home instead of going to the gym, replacing all of our appointments with video conferencing, and skipping the nightly activities we looked forward to.
A. Ogurtsov, Yu. Kashnitsky and T. Gabruseva worked on the translation.
Source: habr.com