Phylogenetic analyses

The full sequence dataset was aligned with MAFFT³⁶ and filtered to contain only the earliest sequence per individual, in order to focus phylogenetic inference on between-host evolution and avoid within-host inflation of lineage-level diversification rates. Surveillance drug resistance mutations were masked before maximum likelihood inference of 100 bootstrap phylogenies in IQ-TREE³⁷ under a GTR+F+R10 model, as determined with ModelFinder.³⁸ Transmission clusters were determined using a minimum size threshold of five, a 90% bootstrap threshold, and a phylogenetic distance threshold of 0.02 substitutions/site, as defined by Poon et al. .³⁹ Cluster growth, branching events and lineage-level diversification rates were assessed in three 60-day time periods, including “pre-lockdown” (January 22–March 21, 2020), “lockdown” (March 22–May 20, 2020), and “post-lockdown” (May 21–July 19, 2020). In each time period, the three phylogenetic measures of transmission were evaluated across clusters and compared between groups of clusters associated with different risk factors. Each cluster was assigned a risk composition score based on the breakdown of risk factors reported by cluster members. The total number of sequences collected from newly sampled individuals during each of the time periods under study is shown in Figure 1C.