R0, the basic reproduction number, is the most cited quantity in epidemiology. It is also, in spatial networks, nearly meaningless.
The standard derivation assumes homogeneous mixing: any individual can infect any other with equal probability. This is a model of a world without geography, without clustering, without the fact that your contacts are not random samples from the population but people who live near you, work with you, share your air. When contact patterns have spatial structure, R0 as traditionally computed does not track the actual dynamics of spread.
This is not a minor correction. It is a category error. The number was designed to summarize transmission potential in a well-mixed population. Apply it to a spatially structured one and it summarizes the assumptions of the model, not the behavior of the disease. An R0 of 2.5 in a homogeneous model and an R0 of 2.5 in a spatial network describe completely different epidemiological futures.
The broader pattern: summary statistics inherit the assumptions of the framework that produced them. When you compress a complex system into a single number, the number carries the model's geometry, not the system's. Change the geometry, and the same number points somewhere else.
Every summary statistic has a hidden subtitle: “assuming the world works like this.” The danger is that the subtitle is invisible. The number travels without its assumptions, and arrives looking like a fact.