Suppression and Mitigation Strategies for Control of COVID-19 in New Zealand

Alex James; Shaun C Hendy; Michael J Plank; Nicholas Steyn

We simulated two types of control strategy: suppression and mitigation. Both types of strategy were modelled by reduction in the transmission coefficient, resulting in a proportional reduction in the reproduction number with control (Rc). This is a very simple control model which assumes that transmission rates from pre-symptomatic, unconfirmed symptomatic and confirmed symptomatic infections are all reduced by a constant factor. This describes society-wide control interventions, such as social distancing, hygiene measures, and lockdowns. It does not account for control measures that specifically target confirmed cases, such as case isolation. These could be modelled by a larger reduction in transmission rates for confirmed cases; we did not attempt this but it could be included in future model refinements to investigate the dependence of control efficacy on testing. The magnitude of the reduction in Rc was calibrated by comparisons with international data on case trajectories ( Fig. 2 ) and modelling studies for the UK and US outbreaks (Ferguson et al 2020) . The latter study assumed that interventions would have a similar impacts as they do with seasonal influenza. We then simulated suppression strategies by a fixed reduction in Rc for a period of 400 days. We simulated mitigation strategies by dynamic changes in Rc aimed at keeping the demand on the healthcare system under capacity (i.e. current infections under 40,000). In all cases, control began at t = 42 days after exposure of the initial seed infections. Table 1 . Estimates for the hospitalisation rate, ICU admission rate and case fatality rate (CFR) for New Zealand. These were calculated from the age-specific rates published by CDC (2020) combined with the age structure of New Zealand's population (StatsNZ). These values assume that NZ has a similar healthcare system, rates of underlying health conditions, and Covid-19 testing rates as the US. Note that our model uses lower values than these estimates, which is reasonable because the model parameters represent rates per infection, which will be lower than rates per confirmed case if some infections are asymptomatic, subclinical or otherwise undiagnosed.