This modeling study argues that undocumented cases are very common and lead to most of the spread of SARS-CoV2.

One of the biggest concerns of public health officials is that people can be infected with SARS-CoV2, not know it, and infect others. We don’t know how many such people exist, and how contagious they are.

The authors of this study developed a mathematical model that simulates how infections might spread in 375 Chinese cities (see supplementary materials). They divided infections into two classes (1) documented infected individuals with symptoms severe enough to be confirmed, and (2) undocumented infected individuals. They each had a different level of contagiousness. They then used mobility data to simulate how people moved in key time periods and coupled that with Bayesian inference to infer characteristics about how SARS-CoV2 infects others.

They estimated that 86% of infections were undocumented prior to the travel restrictions put in place on January 23. Although undocumented cases were less contagious (their transmission rate was 55% of documented cases), because there were so many of them, transmission from undocumented cases accounted for 79% of documented cases.

This is a mathematical model, and it will need to be confirmed with real data, but it argues that containing this is going to be hard, and we are going to need to pay more attention to undocumented cases.