The unexpected dynamics of COVID-19 in Manaus, Brazil: Herd immunity versus interventions

He, D.; Artzy-Randrup, Y.; Musa, S. S.; Stone, L.

The arrival of SARS-COV-2 in late March 2020 in the state of Amazonas, Brazil, captured worldwide attention and concern. The rapid growth of the epidemic, a health system that had collapsed, and mass gravesites for coping with growing numbers of dead, were broadcast by the media around the world. Moreover, a majority of the local Amazonian indigenous communities were physically distant from appropriate medical services, to the point where warnings of genocide were issued. In a recent Science paper (December 2020), Buss et al. reported that some 76% of the residents of the city of Manaus, the capital of Amazonas, had been infected by October 2020. This estimate of the COVID-19 attack rate was based on a seroprevalence analysis of blood donor data, which despite its shortcomings was thought to be a sufficiently reliable proxy of the larger population. An attack rate of this magnitude (76%) implied that herd immunity had already been reached and the community was relatively protected from further infection. Yet in December 2020, a harsh second wave of COVID-19 struck Manaus, and currently appears to be even larger than the first wave. Here we use mathematical modelling of mortality data in Manaus, and in various states of Brazil, to understand why a second wave appeared against all expectations. Our analysis is based on estimating a "flexible" reproductive number R_0 (t) from the mortality data, as it changes in time over the epidemic.