Evaluating the differential impact of what we have done, as we prioritize what to do next: a multi-provincial intervention modeling study using population-based data [Added supplements: COVID-19 Variant Supplement; COVID-19 Variant Network]

In Canada, as elsewhere, the COVID-19 epidemic has spread at varying speeds and amplitudes across people, places, and time. Early model predictions were dire across the board largely because of limited local data. So early models had to assume that we were all at equal risk, regardless of conditions that can lead to differential risks of transmission (e.g. living in shelters or long-term care facilities) and of severe outcomes (e.g. age, health conditions). Thus, an assumption of homogeneity was at the heart of the "hammer" part of the public health response. Public health measures (interventions) also varied between provinces. As we enter the "dance" phase and prepare for future waves of the epidemic, we have an opportunity to be more specific with our interventions if we can quickly learn from how well our public health measures worked or did not work for different subgroups and between provinces, using the wealth of data now available. Our team will use an integrated surveillance and health-administrative data infrastructure and mathematical models that were built over the last 2 months in Quebec, Ontario, Manitoba, Alberta, and British Columbia to answer the following questions: 1. Who, where, when, and under what conditions are subsets of the population most at risk? 2. What led to differences in the trajectory and size of COVID-19 sub-epidemics within and between provinces? 3. What types of population- and facility-specific strategies that could stop these sub-epidemics and prevent their re-emergence, while allowing us to relax universal physical distancing measures? Our team of epidemiologists, mathematical modelers, statisticians, clinicians, microbiologists, and public health officials will work together to rapidly provide answers in way that embraces data-driven heterogeneity in risks so that we can better inform decisions on what to implement, when, for whom, and for how long, to minimize the need for universal stay-at-home strategies.