Figure 2.

Adjusted associations between mobility change and COVID-19 incidence at the subcity level with varying time lags

IRRs (per log unit increase in mobility change) are shown with 95% CIs (error bars). Adjusted models included weekly COVID-19 incidence (outcome), mobility (primary exposure), weeks since second case, population density, residential overcrowding, educational attainment, and country for 1031 subcity areas in Latin America. Weekly mobility was lagged from 1–6 weeks before weekly COVID-19 incidence in successive models. IRR=incidence rate ratio.