COVID-19 epidemic forecast for Moscow for years 2022-2023

Abstract

The paper analyzes the spread of COVID-19 in Moscow in 2020-2022 and gives its long-term forecast to the end of 2023. It describes the infection propagation model and the forecasting technique used. Agent-based modeling became the main tool of forecasting. Lethality statistics collected before 22.09.2021 in Moscow was used as initial data. The model simulated the states of 12 million individuals, who communicated the infection through daily social contacts. Their possible states were susceptible, infectious (symptomatic or asymptomatic), immune, and dead. The lethal outcome could only occur from the first symptomatic disease. Immunity duration is sampled individually for each agent; it is no longer than 180 days. In our model, the probability to die from the first symptomatic disease was determined to be 1.66% for all variants before omicron appeared. With the approach we developed, a forecast was made for the autumn-winter period of 2021-2022. Using retrospective data collected before 15.04.2022, we managed to evaluate the accuracy of our forecast and demonstrated lethality from omicron to be 2.66 times lower than from the variants that circulated earlier. A long-term forecast for the years 2022-2023 was also made. It shows much smaller values of daily lethality if omicron dominates. The effect of the seasonal factor in the forecast is seen to be weak because of the high transmissivity of the circulating variants. About 70% of individuals in the population remain immune; people continuously become infected as they lose immunity after past disease.