Question: Does Layered "Coding" Really Raise the Standard of Epidemic Prevention?

The epidemic situation in Suzhou has recently become severe. Previously, entry to the company only required showing a green "Sukang Code" to proceed. Later, the prevention standards were raised, requiring the "Travel Code" to also be shown without any asterisks to gain entry. Even later, to restrict personnel from outside the region, an "Entry/Exit Code" was introduced. As a result, every time I entered the park, I needed a stack of credentials: employee ID, Sukang Code, Travel Code, and Entry/Exit Permit... This caused a line of cars to get stuck. Especially on the highway, many truck drivers had asterisks on their Travel Codes, so they couldn't exit the highway and had no choice but to park their trucks and improvise meals in the wild.

The most puzzling question is: Does adding an extra code really make things safer? Have the functions of some codes already been covered by others? And how does the system of multiple codes provide a pathway for people with uncertain risk levels to have their prevention strategies relaxed?

For example, having "Shanghai" on the Travel Code often implies a higher level of danger. However, viruses do not spread based on administrative divisions. There are safe places in Shanghai, just as there are dangerous places in Suzhou. For instance, Dishui Lake in Shanghai has had almost no cases. If a trip to Dishui Lake results in an asterisk on my Travel Code, preventing me from entering the park, that seems unreasonable.

Adding extra codes creates quite a few problems. For example, many people have two phones, or they can secretly borrow a phone to fraudulently obtain a green code or a pass to sneak into the community. This forces us to incur legal and moral costs to correct these issues.

Perhaps for grassroots workers, ensuring epidemic prevention safety, adding an extra code provides an additional layer of verification. However, in the spirit of precise epidemic prevention, adding an extra code often brings more obstacles without reflecting "precision"; instead, it reflects a "one-size-fits-all" approach.

To truly solve this problem, perhaps we need to rethink the health system.

Let's first discuss some simple matters:

Viruses Do Not Spread by Administrative Divisions

For a given case, if they have historically visited a location, we can use the horizontal axis to represent time and the vertical axis to represent location, thereby expressing a case event. This allows us to establish a two-dimensional coordinate system, where each action of each case can be described as a case event, represented by a red dot.

For example, if on April 2nd it is discovered that a case has previously visited Jinji Lake, we can generate the following image:

Obviously, based on the transmission patterns of the virus, for a period of time after the case visited a location, the surrounding area consists of spatiotemporal points with a higher probability of infection.

Therefore, it should clearly appear to the right on the time axis, distributed as a probability density cloud according to location. The closer to the case's actions, the higher the risk. It gradually diffuses into the surrounding space.

Consequently, the closer a person's actions on the timeline and at specific locations are to the yellow area, the more likely they are to become infected.

Based on a person's degree of proximity, their probability score of infection can be inferred.

Does a green health code guarantee that you are definitely not infected?

The most serious problem with traditional health codes is that it is far too easy to be green. As a result, nearly everyone has a green code, making it extremely difficult to gauge risk in special situations, such as when enhanced epidemic prevention measures are required.

For example, during an outbreak in Suzhou, residential communities may be sealed off. This might be because a large number of cases have been discovered in the surrounding area, which significantly increases the likelihood of cases appearing within the community as well. After sealing off a community, nucleic acid testing is required for all residents. This is indeed the best solution we currently know of, as it can rapidly identify infected cases.

However, there is a logical inconsistency: during nucleic acid testing, nearly all participants have green health codes. If health codes are supposed to effectively assess an individual's probability of infection, and a green code is only displayed when infection is almost impossible, then testing a group of people with green codes—essentially testing a group that the system has already deemed unlikely to be infected—is not very efficient.

On the other hand, if we argue that health codes do not reflect an individual's probability of infection, then using health codes as a criterion for entry and exit is no longer reasonable. A green health code implying that one is definitely not infected is not a valid claim. Consequently, communities would develop their own entry and exit permit systems. This ultimately leads to a layered addition of "codes."

To a large extent, this undermines the original purpose of health codes: serving as a standard for determining whether an individual is infected during an epidemic.

In reality, what we truly need is a quantifiable metric. It should genuinely reflect the probability of infection, be expressed as a numerical value, and be capable of quantitative statistical analysis.

Health Codes Should Display the Numerical Value of Infection Probability

Based on a straightforward model that matches the spatiotemporal intersections of individuals and confirmed cases, we can easily determine whether a person may have been infected.

However, rather than simply assigning a yellow status code and manually investigating their travel history and close contacts, it is more reasonable to use a metric that quantifies their level of risk. Especially once such a metric is established, we can clearly and accurately assess our own health status, and it also facilitates the management of entry and exit personnel in residential communities and companies.

For example: A safety score app for each individual allows them to view their probability of infection. This value may range from 1% to 100%. The method of calculating it involves summing the vectors of this person’s recent travel history with those of other individuals.

Confirmed cases are also people, and they also have a safety score. However, since they may be infected, their probability of infection is 100%. If you have ever been at a location close to them, this will be added to your probability of infection.

Nucleic Acid Testing - Significantly Altering the Infection Probability in the System

In the model above, we used a numerical value to describe a person's probability of infection.

So when a community is in danger, there may be many people inside the community with a high probability of being infected: for example, some might have a 0.01% chance, some 1%, and others 10%.

For those with a 0.01% probability of infection, conducting nucleic acid tests on them is almost meaningless. On the other hand, for those with a 10% probability of infection, the situation is very dangerous. Many of them may already be cases.

Regarding nucleic acid testing, some aspects can be fully automated. On one hand, nucleic acid testing can be selectively executed based on the probability of infection. On the other hand, nucleic acid testing can alter a person's probability of infection.

For example, if a person initially has a 10% probability of infection. After each nucleic acid test, if the result is negative, this probability should be reduced. For instance, after one test it might drop to 5%, and after a second test to 2.5%. However, if the result is positive, the probability immediately becomes 100%.

Recursion Safety Score - Fully Automated Epidemiological Analysis

Now that we are able to determine the probability of infection for each individual based on nucleic acid testing, the model we establish should be recursive.

To put it simply: suppose your initial probability of being infected is 10%. Recently, people you have been in close contact with have undergone multiple nucleic acid tests, and nearly all of them have tested negative, reducing their probability of infection to a very low value, such as 0.01%. Since their movements have heavily overlapped with yours, being in contact with so many negative individuals should also lower your infection score. This is scientifically sound: if the people you interact with daily are negative, it is highly probable that you are also negative.

Conversely, if a person is suddenly found to be positive, their previous actions are considered highly dangerous. Anyone who has been in close proximity to them could potentially be positive. This would directly and significantly alter the infection probability of everyone who has previously crossed paths with them. Instantly, this entire group would have a very high probability of being infected.

From this perspective, the behavior of any individual in a city can potentially change the safety score of someone located very far away. Although they may not have interacted with you directly, they might indirectly affect your score through four or five levels of recursion.

This type of calculation is difficult for humans, requiring a large number of medical professionals to conduct epidemiological investigations. However, for a computer, it involves only simple polynomial multiplication and addition. Each person's health status is an expression that recursively involves the entire city's population. Representing the movements of the entire city's population as millions of matrices and then performing matrix transformations is not a significant computational burden for a computer.

This algorithm is essentially a form of "bionics," mimicking real-world epidemiological investigations, but it uses real-world data for calibration.

Once this algorithm is implemented, it will make testing in residential communities much more convenient. Normally, communities can require a 99.9% health probability for entry. During special periods, they can raise the requirement to 99.99% or even 99.999% health probability. Since everyone knows that undergoing additional nucleic acid tests can significantly increase this score, individuals at high risk will voluntarily complete their nucleic acid tests outside first, facilitating their entry and exit from the community.

Based on AI?

We can think of an algorithm that estimates each person's probability of infection based on citywide travel data and individual nucleic acid test results, and it is not particularly complex.

However, I know that while such a simple algorithm may be reliable, it will inevitably have loopholes: for example, there might be a case where the algorithm assesses an extremely low probability of infection, but a stay-at-home individual who rarely goes out decides to visit a positive case one day without carrying a phone. Or, a person assessed with a very high probability of infection might remain negative because their protective measures are genuinely effective.

To truly achieve precise epidemic prevention, we might need more complex algorithms. Perhaps AI-based ones, where the AI also factors in your daily routine, number of phones, type of work, age, medical history, and so on. A taxi driver's probability of infection is significantly higher than that of a white-collar worker sitting in an office. People with weaker immune systems should obviously have a higher probability than young people.

I can only claim to have proposed a very simple model, suggesting a method that can quantify the probability of infection and facilitate precise prevention and control. I look forward to seeing papers and algorithms that use machine learning and artificial intelligence to make more accurate predictions about each person's health status.

Theoretical Security vs. Practical Security

The points mentioned above can only remain at the theoretical level. For efficient epidemic prevention, we must not separate theoretical safety from practical safety. Although isolation has historically been identified as the most effective method, in some situations, by combining algorithmic computation results, actual safety measures should often be more flexible.

For example: When the epidemic was at its peak in our residential community, takeout deliveries were prohibited. However, every time someone entered or exited the community, they still had to press a button to open the gate. I firmly believe that this button is covered with viruses from the hands of everyone in the community. They should have long since removed this entry button and kept the gate permanently open.

Another example: Suzhou recently implemented a policy banning dine-in service at restaurants. However, this has only led to people buying food and crouching at the restaurant entrance to eat it. Although the intent is for people to take the food home, humans are practical; they will always find the most convenient method for themselves.

Banning dine-in service is essentially aimed at reducing cross-infection during meals. But if the restaurant is already empty, who are you going to cross-infect while eating?

I still believe this policy should be changed to: prohibit multiple people from eating together at the same time. In practice, this means each restaurant should only provide one chair.

Strict epidemic prevention is necessary; only efficiency needs to be improved

I believe that genuine epidemic prevention efforts must not be relaxed. Many people argue that we should fully open up and simply lie flat. They claim that Omicron is just a big flu, and say things like: the cost of prevention has already exceeded the cost of getting infected.

I think this is unscientific. Strictly speaking, what we are defending against is not Omicron itself. Omicron may indeed have a low severe case rate, and the symptoms after infection are indeed similar to a cold. However, I believe the real threat is absolutely not Omicron. It is simply because Omicron’s transmissibility is so strong that our mainstream media has focused exclusively on Omicron.

Previous variants, such as Delta, have not disappeared just because Omicron spreads so easily. They are simply no longer in the spotlight. If we fully open up, sooner or later we will be infected by an even more dangerous disease.

Perhaps it is precisely because of strict epidemic prevention that we have altered the timeline. In another timeline where China has already fully opened up, Omicron has already produced a next-generation variant with a 50% mortality rate. We cannot rule out this possibility.

For me personally, strict epidemic prevention has protected me from more than just Omicron. I have almost not had a cold or fever in nearly three years. I have almost forgotten that humanity had not yet conquered the common cold. Perhaps these measures could further eliminate more epidemics in the future, allowing humans to live longer.

However, what we truly need to improve is the efficiency of implementation. Efficiency is about reducing our costs. I can also hear from many friends in Shanghai who are starving because they cannot buy vegetables, panicking because they cannot see a doctor, and furious because they cannot afford rent...

Perhaps what I am discussing is just keyboard warriors debating politics from their rooms, but there are clearly better methods to improve efficiency: for example, using AI to assist in epidemiological investigations, monitoring individuals with a high probability of infection precisely rather than adding layered restrictions, and adding a safety assessment feature to health codes, etc.

Health codes have been in operation for several years since the outbreak of the pandemic. The times have changed quite a bit. The virus is not as scary as before, but dangers are still lurking everywhere. What we can do is use a "Health Code 2.0" to improve efficiency, achieving precise zero-COVID while reducing this "cost".