Determining the optimal COVID-19 policy response using agent-based modelling linked to health and cost modelling: Case study for Victoria, Australia

Abstract Importance Determining the best policy on social restrictions and lockdowns for the COVID-19 pandemic is challenging. Objective To determine the optimal policy response ranging from aggressive and moderate elimination, tight suppression (aiming for 1 to 5 cases per million per day) and loose suppression (5 to 25 cases per million per day). Design Two simulation models in series: an agent-based model to estimate daily SARS-CoV-2 infection rates and time in four stages of social restrictions; a proportional multistate lifetable model to estimate long-run health impacts (health adjusted life years (HALYs) arising from SARS-CoV-2) and costs (health systems, and health system plus GDP). The net monetary benefit (NMB) of each policy option at varying willingness to pay (WTP) per HALY was calculated: NMB = HALYs × WTP – cost. The optimal policy response was that with the highest NMB. Setting and participants The State of Victoria, Australia, using simulation modeling of all residents. Main Outcome and Measures SARS-CoV-2 infection rates, time under various stages of restrictions, HALYs, health expenditure and GDP losses. Results Aggressive elimination resulted in the highest percentage of days with the lowest level of restrictions (median 31.7%, 90% simulation interval 6.6% to 64.4%). However, days in hard lockdown were similar across all four strategies (medians 27.5% to 36.1%). HALY losses (compared to a no-COVID-19 scenario) were similar for aggressive elimination (286, 219 to 389) and moderate elimination (314, 228 to 413), and nearly eight and 40-times higher for tight and loose suppression. The median GDP loss was least for moderate elimination ($US41.7 billion, $29.0 to $63.6 billion), but there was substantial overlap in simulation intervals between the four strategies. From a health system perspective aggressive elimination was optimal in 64% of simulations above a willingness to pay of $15,000 per HALY, followed by moderate elimination in 35% of simulations. Moderate elimination was optimal from a partial societal perspective in half the simulations followed by aggressive elimination in a quarter. Shortening the pandemic duration to 6 months saw loose suppression become preferable under a partial societal perspective. Conclusions and Relevance Elimination strategies were preferable over a 1-year pandemic duration. Funding Anonymous philanthropic donation to the University of Melbourne. Key points Question To determine the optimal of four policy responses to COVID-19 in the State of Victoria, Australia (aggressive and moderate elimination, tight suppression (aiming for 1 to 5 cases per million per day) and loose suppression (5 to 25 cases per million per day), based on estimated future health loss and costs from both a health system and partial societal perspective. Findings From a health system perspective aggressive elimination was optimal in 64% of simulations above a willingness to pay of $15,000 per HALY, followed by moderate elimination in 35% of simulations. Moderate elimination was optimal from a partial societal perspective (i.e., including GDP losses) in half the simulations followed by aggressive elimination in a quarter. Meaning Whilst there is considerable uncertainty in outcomes for all the four policy options, the two elimination options are usually optimal from both a health system and a partial societal (health expenditure plus GDP cost) perspective.