Natural Infection Disconnects Communities
How a False Assumption Might Lead to Policy Disaster
A bit of a short post today, inspired by a CSPI article/podcast episode by Philippe Lemoine. His core observation is that a fundamental assumption of pandemic models - that population interaction is uniform - is not only blatantly false, but likely extremely consequential. There are many excellent observations and predictions that stem from this idea, many of which Philippe discusses. However, one vital observation that isn’t (at least explicitly) discussed is the idea of bridges.
Most models involving interactions between individuals and networks operate off of a specific field of mathematics called graph theory (conveniently, my field of study). The basic building blocks of graph theory are nodes, in this case representing individuals, and edges, representing interactions between individuals. A graph might look like this:
Take the above example that could represent a hypothetical group of people. Here, the letters A-H are all people, while the edges (lines) between them mean that a pair of people are in contact with each other. Visually, you can tell that it looks like there are two communities with little interaction between them, namely, {ABCDE} and {FGH}. Between these groups, only E and F interact. The formal definition of these communities vary, but the idea they capture is the same: there are groups of people that are much more connected with each other than with anyone outside of the group. Philippe provides examples of such communities - cities, religious groups, university campuses, social circles - that are incredibly intuitive.
I want to draw attention particularly to the connections between different groups. They are relatively few in number and very crucially, are more likely to be infected naturally during the pandemic. This is true for several reasons. Firstly is that in real life, the people connecting different groups just tend to have many interactions. Think frequent flyers between cities, truckers, people with interaction-heavy service jobs, etc. More interactions roughly correlates with higher likelihood to be infected. Additionally, if some disease infects a group, at some point it must pass through someone (and in practice, a number of people) with connections to the outside. This is exacerbated by both top-down and natural changes in behaviour. People may become more wary in a group once the disease is detected or more widespread in that group, leading to those who have fewer connections outside to be more likely to take precautions (willingly or not) and possibly avoid being infected.
The consequence of this pattern is that as pandemics progress further in, the inter-group spread becomes much less than intra-group spread. While there will not be a complete separation of groups (as there will always remain a lower, but nonzero percentage of inter-group infections), the likelihood of inter-group spread will be lower. In many countries in which vaccine requirements are required for travel or even regular interactions, this dynamic is increased. Then, we can expect COVID to become more regional, more unequal, and more of a slow burn. Consider this map of US cases:
Moreover, this increasing regionalization means that not just vaccination immunity but natural immunity will vary greatly depending on region. In other words, it can be simultaneously true in many areas, namely those with high rates of vaccination and prior infection, COVID is over, while other areas remain susceptible and growing. Of course, this isn’t a single-issue explanation. Most notably attitudes towards vaccination can greatly affect when and to what degree waves occur in different regions. Also note that by the nature of real-life community structure, these communities won’t necessarily be states or even cities, but more like small social groups and local networks.
Regionalization lines up perfectly with COVID becoming endemic. Endemic viruses, such as influenzas or rhinoviruses, typically spread in pockets, changing seasonally, and “migrating” across different regions as time progresses. This is indicative of greater immunity within a population in general, and likely indicative of greater immunity within inter-group people in specific. This also lines up poorly with most COVID plans and guidelines, which use fairly large (relative to communities of spread) areas of attack, to implement policies which mostly attack inter-group, not intra-group spread with significant social and economic costs, and that only go away when lower rates of infection are reached, which is unlikely to happen quickly as intra-group spread is not as explosive as initial waves of the pandemic. Following the trend of Philippe Lemoine’s article, this sets up a scenario in which the consequences of endemic COVID are low and mostly unavoidable, while the consequences of policies will be elongated and amplified by a combination of misuse of metrics and specific epidemiological conditions. In other words, if we aren’t careful, we might be setting up the circumstances to maximize harm done by bad policy.