"Cluster countermeasures: Japan's response to the coronavirus pandemic and last hope"

Notes from reading a blog post by Kouhei Yoshimine.

The post is based on publicly available materials, such as meeting notes from the panel of experts and press conferences, and notes by Professor Minato Nakazawa, aimed to help and promote the public's understanding of where Japan stands. As the author says himself, he is a lawyer in Japan and not a subject matter expert.

The reader is expected to be familiar with background knowledge such as R₀ and the pathology of COVID-19.

March 20, 2020

Japan's coronavirus responses are driven by a panel of experts and the Cluster Response Section of Ministry of Health, Labor and Welfare.

1. “At least 100k people will die if we fail”

(1) What an overshoot will lead to

There are two broad established responses to infectious diseases: suppression and mitigation
Failing to suppress results in the disease spreading exponentially, ending with herd immunity.
Individual immunity required for herd immunity is determined by R₀. This is estimated as 40-70% in COVID-19's case.
We've already failed to suppress, haven't we? Which means exponential spread of the virus is inevitable.
That's how the story's been going so far.
But here, we have people who are making a strange claim that there's a way to keep the virus from spreading in Japan even then. They're the panel of experts and Cluster Response Section mentioned above. It's a strange strategy based on a strange feature of the novel coronavirus. One could call this quasi-suppression strategy.
Based on a naive calculation and an analysis of charts found in one of the meeting notes of the experts panel, approximately 120-200k people may die until we achieve herd immunity under conventional countermeasures.

(2) Japan's R₀ is strangely low

Keep in mind that Japan's R₀ has been lower than that of other countries. Otherwise, Japan would've been in an overshoot situation by February like them. The rate of growth of the number of cases and deaths in the country has been consistently low since January.
Is this because Japan isn't testing as much? Suppressing tests can only shift an exponential growth by a few days and wouldn't explain the consistently low growth rate.
The real reason is unknown right now. Japan hasn't done any major lockdown at the level of Singapore or Taiwan, and it's not clear why it's been so low even as early as January.
It may have to do with the fact that many people already had a habit of personal protection in the form of hand washing and masks and no cultural norm to be physically close to other people, such as shaking hands or hugging. There are also many otakus who stay at home, living most of their lives in the virtual world.
It wouldn't have been possible to even think of bringing R₀ below 1 through cluster countermeasures, had it been higher like other countries.

2. Theoretical background of cluster countemeasures

Grantz K, Metcalf CJE, Lessler J (15th Feb 2020) Dispersion vs. Control. provided the basic framework, and Nishiura H et al. “Closed environments facilitate secondary transmission of coronavirus disease 2019 (COVID-19) showed a concrete strategy based on that.

(1) How COVID-19 spreads: high variance in R₀ (Grantz)

Grantz's paper showed that the coronavirus has a high variance in its R₀ based on Wuhan's data. Only a minor fraction of infected people transmitted the virus to a large number of people, while the rest stayed isolated cases.
This is what we call cluster infection.
With influenza, the spread is more continuous, whereas with coronavirus, it's more discrete. It's harder to predict, but there's room for intervention.
In theory, if we can suppress or control cluster infection, we can achieve R₀<1.

(2) Determining how cluster countermeasures can be implemented (Nishiura)

Such theory is only a pie in the sky unless we can execute it. Nishiura's paper (not peer-reviewed yet) showed how. Some consider it a genius.
The paper analyzed 110 cases from 11 clusters that were identified in Japan and showed that there are certain conditions for a large scale secondary infection to happen.
One key condition is whether the spread happened in a closed environment. In an open environment, one infected person may spread it to about one or two people at most, while in an closed environment, the resulting cluster size can go up to ten or so. A closed environment carries 18.7 times more risk.
But, if a closed environment causes cluster infection, why hasn't the packed trains of Tokyo caused an outbreak?

(3) The three conditions of cluster infection: closed space, close quarters, close contact

The panel of experts published a memo on March 9th that lists these three conditions. It appears to be this is still in the realm of heuristics and hasn't been elaborated as a mathematical model yet.

Group infections we've seen so far have occurred in situations that meet these criteria: 1. Closed space with poor ventilation 2. Many people were gathered densely 3. Conversations or vocalizations within short distance (be able to touch each other if you reach out)

These three conditions explain why packed trains have not caused cluster infection (isolated infections are still possible): they satisfy the conditions of closed space and close quarters, but not close contact.
If every citizen can reliably avoid these conditions, which should be fairly easy compared to lockdowns and quarantines issued in other countries, large-scale secondary infections can be prevented.
By altering human behavior in a relatively easy manner, Japan may be able to prevent cluster infections (not all infections; this is the key difference) and suppress most of corona infections.

3. How it'll look like when cluster countermeasures are in effect

There are two broad pillars of cluster countermeasures:
1. Quickly detecting and suppressing cluster infections that have already occurred
2. Preventing new cluster infections from occurring through behavioral measures
By suppressing cluster infections, R₀ can be pushed down to far below 1. Isolated infections will not pose a threat.
The risk lies in asymptomatic carriers accidentally causing cluster infections. They need to be quickly detected and suppressed before they cause further cluster infections.

4. Detecting clusters and operating with a one-month delay

(1) The role of PCR testing

We detect clusters by testing severe cases with PCR. Severe cases mostly consist of elderly people.
Clusters made up of young people are 'invisible' and may cause a chain reaction of cluster infections, leading to an uncontrollable outbreak.
Once a cluster causes a severe case, it'll be caught by the Cluster Response Section, who will trace and analyze the formation of the cluster, feeding back to the countermeasures put in place. This forms a cycle of observing and acting.

(2) Operating with a one-month delay

As the coronavirus can stay dormant for two weeks, there will be a lag between a cluster forming and it becoming detectable.
Once an outbreak occurs, the number will continue to rise for at least two more weeks because of this delay, even if a lockdown is issued.
It's a very difficult and terrifying thing to handle an exponential phenomenon that you can only observe its past from two weeks ago. Whatever action we take, its effect will be only known two weeks from now.
Therefore, the flow of information needs to be sped up: expand capabilities to perform PCR testing earlier, streamline the reporting process, and trace contacts as quickly as possible. We need manpower here.

5. Can we detect invisible clusters by testing more?

It's not a question of whether to test more. We need to consider the purpose of testing and our strategy in how to go about testing.

(1) Possible strategies for testing

(a) Test patients who have developed symptoms that match COVID-19 (b) Test contacts (epidemiological survey) (c) Massive testing of the younger population

The Japanese government has adopted a) and b). There are opinions that push for c), so that we can detect invisible clusters.

(2) Why massive testing of younger population is untenable

(A) There will be a large amount of false positives, making cluster detection impossible (B) False positive cases will overwhelm the healthcare system or quarantine facilities (C) We either do not have the resources to execute on the strategy, or it will be too straining to do so

All of this boils down to avoiding testing when the prior probability is low.

(3) Why South Korea succeeded

South Korea implemented drive-thru / massive testing and has succeeded in suppressing the spread. Why is this?
South Korea had already expanded on its testing capabilities due to its experience combating (C)SARS.
There was a special cluster of younger population within a religious group. This led to high prior probability.

(4) Expand testing capabilities but do not change strategy

This will allow the country to detect invisible clusters earlier

6. What we should do

(1) Worrisome factors

Whether the spread in Hyogo prefecture can be contained
Whether the air of collective self-restraint will dissipate

Hokkaido experienced a spread without known links after the snow festival but seems to have succeeded in suppression by announcing an emergency and coordinating with the Cluster Response Section. (Hokkaido model)
Hyogo is in a similar situation.
Prime Minister Abe's decision to close schools was made without consulting the panel of experts and the Cluster Response Section.
At first, the author thought it may have had its value as a signal that the situation is serious, but the follow-up decision to not extend the closure seems to have made people think that they're okay now.
This may lead to new clusters forming. We don't know yet how closely the three conditions should be followed. We lack experience and data.

(2) Bi-directional risk communication

Summary so far: the essence of the strategy that the panel of experts and the Cluster Response Section have come up with lies in the three conditions of closed space, close quarters, and close contact, and thorough behavioral measures backed by citizens' understanding of these conditions.
The officials are moving fast compared to expectations, but this is a strange strategy based on a strange characteristic of the virus, one that defies common sense. It's crucial for everyone to understand what they must do and why, and the communication channels set up so far may not be enough.
The risk we face may be dire. Without the knowledge of what the worst future entails, it may be hard to change one's behavior.

We will find ourselves in a nightmare if we fail. Yet there is hope. With the Japan model, we're trying to find a sustainable measure that can maintain a basic level of economy and societal functions, all the while protecting the vulnerable population. I believe the optimum answer lies there, but it might not be an easy compromise for everyone to make. We need a process for building consensus. (Paraphrased Prof. Nishiura's words from a press conference held on March 19th)

If you've read this long winding blog post so far and can understand such tedious topics as R₀ or false positives even somewhat, you definitely have a better understanding of cluster countermeasures than the average population. It's fine if you don't agree with what's written here. You may be able to cause a 'cluster infection of knowledge' and contribute to risk communication and this fight against the virus.

(3) What we can do as a common citizen

We're in for the long haul. What we can do:
1. Protect yourself: gargle and wash your hands.
2. Avoid high-risk behaviors that involve 'closed space, close quarters, close contact' situations, but balance it between keeping the economy running.
Keep in mind that you may be infected without symptoms and can cause cluster infection, especially if you're young. We need to operate under the assumption that any person can be already infected.

Referenced materials

Publicly available materials from the panel of experts
Videos of press conferences and interviews of Prof. Yamanaka
Prof. Minato Nakazawa's website (Japanese) – essential for understanding the current situation