Predicting what the COVID-19 pandemic will do to Russia and when it will give way to a new normal is unusually hard. The only way to make predictions on that scale is to use models that expand on highly incomplete data, and those models are typically the territory of epidemiological researchers. Western researchers, however, appear to be somewhat uninterested in figuring out Russia’s coronavirus prognosis, and Russian researchers (with extremely rare exceptions) either don’t work on models at all or don’t publish their results. Still, thanks to the efforts of infectious disease scholars worldwide, a number of open-source epidemiological models and improved data sources have become available in recent weeks. We used those tools to create our own, journalist-made model of how Russia’s pandemic may develop. The result: since Russian regions began introducing self-isolation rules in March, COVID-19 has spread through the country more slowly, but even that decrease won’t be anywhere near enough to mitigate the pandemic in many Russian regions unless even more quarantine regulations are added.

What we found, in brief

Even very rough models of COVID-19’s development in Russia have been difficult to build because the necessary data are hard to come by. The main problem here applies to almost any country: official counts of COVID-19 cases tend to be vast underestimates. To counterbalance that disparity, statistics on the number of people infected by the new coronavirus have to be corrected based on reported fatality numbers, which are significantly more accurate. That corrected data ultimately led us to two scenarios. In the worst case, current restrictions don’t work, no new ones are put in place, and mass testing fails to take hold. In a more likely and much better scenario, the Russian government would increase its limits on the population’s movements, the health care system would weather the storm, and no more than 6,000 people would die of COVID-19 throughout the country. However, such a positive outlook would require Russia’s current self-isolation measures to be tightened significantly and extended through the end of May.

What this model does

Our model, like most models being used to describe the progression of the COVID-19 pandemic today, is an SEIR model. This means it predicts how individuals move among four groups: susceptible (S), exposed (E), infectious (I), and resistant (R). “Susceptible” individuals have not yet been exposed to the virus, which means they lack immunity and could contract it in the future. “Exposed” individuals already have the coronavirus in their bodies, but the virus has an incubation period, so those individuals are not yet contagious. “Infectious” individuals are, well, infectious, and scientists believe they become “resistant” to the virus once they defeat it. How infectious individuals pass the virus on to susceptible ones depends on a vital parameter called R₀ (pronounced are-naught). This is the average number of people each infected person infects in turn before they become resistant. By predicting how many people are in the “infectious” category at a given time, we can predict whether enough of them will need hospital care or ventilators to overwhelm the health care system. We can also make a prediction about how many of them will die.

The problem is that R₀ and the initial numbers of people infected early in a pandemic are both hard to predict unless you have widespread testing. Fortunately, scientists have begun to find ways around this gap as COVID-19 has progressed. Every time the disease has spread to a new country, a new city, or a new cruise ship, scientists have used the resulting data to get a better sense of these and other key parameters in order to better fight the new coronavirus down the line.

For example, epidemiologists had previously floated many different estimates of how quickly the virus spreads if no measures are taken against it. They had also struggled to make an educated guess about the average fatality rate of COVID-19: when governments are only counting a fraction of those who contract the new coronavirus, death rates seem a lot higher than they actually are. Errors like these completely change the picture of the COVID-19 pandemic that is accessible to society. For example, when the number of cases in Moscow jumped on April 13, Mayor Sergey Sobyanin rushed to explain that the spike was a result of increased testing, not a massive increase in actual cases.

Scientists are now fairly sure that COVID-19 is deadly in slightly less than one percent of cases, depending on the demographics of the population at hand. They also have a better idea of how to compensate for cases that remain uncounted in official records. This model takes advantage of those changes in ways that models made just a few weeks ago couldn’t. Nonetheless, many questions about how the virus spreads remain open, and this model is therefore very limited even though it tries its best to make up for that lack of data.

How this model is built

To prevent our model from falling into the specific pitfalls that run plentiful in Russia’s COVID-19 landscape, we used a special coefficient to represent infections that go unnoticed and uncounted in official state reports. Scientists around the world have used variations of this method to model the pandemic in a range of countries and regions. For example, researchers from the University of Melbourne in Australia used a coefficient that worked much like ours to develop a model that works internationally. In Russia’s case, this method has upsides and downsides:

• Unfortunately, it’s impossible to estimate this coefficient for the entire duration of the pandemic. This, in combination with the fact that the new coronavirus began spreading through Russia in earnest only recently, means that our coefficient system only works well for describing a short period of time.
• Fortunately, that period includes the days immediately preceding the introduction of Russia’s first serious “interventions.” Those included Vladimir Putin’s March 26 order calling for a week of paid work leave across the country, which was then extended to a month and supplemented with mitigative steps taken in individual cities and regions. Our model is also relatively strong for the 10 days following Putin’s initial order, when strong social distancing measures began to take effect. This allows us to estimate the consequences of Russia’s first major policy shift.

As mentioned above, we predicted how many cases of COVID-19 there actually were in Russia on various days using officially reported data on how many people had died of the disease. Death counts aren’t entirely accurate, either, but they are by far the most accurate data we have. If we know about how many people the new coronavirus has killed per day, about what percentage of those it infects tend to pass away, and how much time it takes for the virus to kill them, we can estimate the total number of cases at any given time.

We know about how long COVID-19 takes to become fatal thanks to scientific studies: on average, a person who dies from the disease was infected 21 or 22 days before, meaning they die 17 days after first displaying symptoms.

That said, there is a catch: not every patient who dies of COVID-19 passes away at the same point as the average victim of the disease. To take more patients into account, you need a “floating window” that spreads a given day’s deaths out over the series of a few days when those individuals were infected. We built that floating window into our model by putting together official fatality statistics and a statistics-based expansion of official case counts. The big assumption we made in that process was about how many days pass between when someone contracts the virus and when their case is counted. Given the time it takes for symptoms to show up, medical attention to be given, and tests to be evaluated and reported, we put that figure at 10 – 11 days.

The result was a very rough but functional model. We were able to calculate the coefficient for uncounted cases from March 17 through April 5 — the new infections that occurred on those dates have already yielded data on deaths and recoveries. This coefficient then allowed us to evaluate, however approximately, the effects of anti-pandemic measures like self-isolation, social distancing, good personal hygiene, and increased or sped-up testing. A number of other factors, from Russia’s age demographics to the average period when a patient is contagious, were already included in our base model, which was developed by the Neher group at the University of Basel (the rationale for our exact modifications is described in Russian here, and it’s similar to the reasoning described in English here).

By this point, we were ready to put together a prognosis for the future of Russia’s epidemic. You can see all our basic assumptions and our best-case scenario in an easily accessible, interactive format at this link. The results are analyzed below.

When will Russia’s peak be? When will its shutdown end? Two scenarios

Scenario one

The measures Russia has in place now to hold off the coronavirus epidemic are essentially the same ones that were put in place at the end of March. COVID-19 testing also remains in the same ballpark: the country hasn’t been able to organize mass testing for all demographic groups and has continued to test only people who show symptoms. Our model indicates that under these conditions, in early April, the average infected person passed their infection on to between 2.4 and 2.6 people (in other words, R₀ = between 2.4 and 2.6). This lines up with data provided by the Russian government: on April 13, Vice-Premiere Tatiana Golikova said case counts were increasing by 16 – 18 percent daily, which would point to an R₀ of between 2.65 and 2.91.

Long story short, all this means that what Russia’s federal and regional governments are doing right now isn’t enough. If the epidemic continues to hold to those numbers, it would probably peak in Russia in mid-June, after which herd immunity would start to set in. By mid-May, however, the country’s healthcare system would be overwhelmed, and by the time it was all over, more than one million people would have died from COVID-19 in Russia. It is plainly obvious that such a result is politically and socially unacceptable. This means the country must escalate what it’s doing to fight the COVID-19 pandemic.

Scenario two

Now, imagine if a number of regional governments (or the federal government) made the decision to introduce new interventions that would decrease the number of “dangerous contacts” — contacts between infectious people and susceptible people — by 80 percent from what they were before the pandemic started. That would mean cutting many social contacts down to a fifth of their previous frequency. In such a case, the pandemic would immediately begin to decline, though we would only find out about that decline from official case counts 10 days later. Fewer than 6,000 people would ultimately die of COVID-19, and the Russian healthcare system would not be overwhelmed. By early July, the number of people spreading the virus at any given time would be around a thousand, and officials could gradually lift their restrictions on the public.

The question that remains is whether the government could ever make this scenario happen given social and technological constraints. In Moscow, for example, city officials introduced a mandatory digital pass system for most people planning to use private or public transport in the city. Not only did the system see major glitches; the process of checking each pass led crowds of passengers to wait in close quarters at the entrances to major metro stations. Even theoretically effective interventions can sometimes make the situation worse in practice, and if the population doesn’t trust its government’s choices, that can decrease the effectiveness of any restrictions at all.

The main limits of this model

• The data on fatalities that we used could be significantly lower than the actual number of COVID-19 fatalities in Russia. If this is the case, the government would have to take even harsher measures to meet our benchmarks.
• The data available from the period we looked at focuses on a small number of cities and highly populated regions in Russia because that’s where the most cases were in late March and early April. As case counts in those places ultimately decline, they could rise simultaneously in less populated areas, and governments would have to take that and other location-based differences into account.
• R₀ could be significantly lower or higher than what we’ve predicted here. This is because recent studies have found that the time between when a person contracts the new coronavirus and when they first pass it on to someone else could be up to 30 percent shorter than previously thought (according to a Chinese study) or much longer (according to a British study).
• If the British study is right and people actually take longer to pass on the virus than we think, that would bring both good news and bad news. The good news is that R₀ would be lower, so new isolation measures wouldn’t have to be as strict to get the total number of COVID-19 cases to stop growing. The bad news is that beyond that, it would be significantly harder to detect new cases and stop the epidemic entirely. Government and medical officials would have to use repeated mass testing or contact tracing to pin down newly infected individuals before they even show symptoms, and everyone who had been in contact with each of those individuals would have to be isolated in as little as two or three days. If that contact tracing strategy were to fail and R₀ were to go above 2.0 or so once again, then it would again become impossible to track each individual case, potentially bringing about a new full-on quarantine.

So Russia will just have to endure a lengthy shutdown? Are there any alternatives?

The only examples we can look to on this count are the countries in East Asia that have successfully suppressed the virus. China, for example, used a range of strategies to draw down the pandemic, but it’s hard to say which of them was most effective. Some researchers believe that contact tracing was China’s real skeleton key, with its quarantine doing half as much to slow the spread. Finding and isolating infected individuals and their contacts also served Singapore very well: because doctors and police officers were able to coordinate their actions immediately and track down patients stemming from eight main clusters, R₀ never went above 1.5 in that country.

In Russia’s case, however, such a strategy is likely unrealistic. Contact tracing would certainly serve its purpose in stemming the tide of the pandemic, but outside regions where COVID-19 has just arrived, things may be too far along for officials to track down every infected individual and everyone they’ve recently met. The problem is not so much that an individual approach wouldn’t be useful; it’s that any country would struggle to summon the resources and sudden coordination needed to make it work at this stage.

South Korea, meanwhile, beat the bug by testing throughout its population and isolating all carriers. It did so early on in the epidemic’s development: South Korea’s R₀, like Singapore’s, never went above 1.5. However, mass testing can be an important part of fighting COVID-19 even at more advanced stages. Germany, for example is holding out well using relatively mild social distancing measures and mass testing of all population groups, including asymptomatic individuals.

According to official data, Russia is conducting a massive number of tests, but it’s only begun to shift toward testing asymptomatic people quite recently. The result is that officials say up to 50 percent of the infected individuals they’re now finding don’t show symptoms. Further mass testing will be a powerful resource that should help slow down the virus’s spread even without a complete national lockdown.