SSA post-analysis of COVID-19 incidence magnitude in Novosibirsk Region

The article is aimed at conducting post-analysis of COVID-19 epidemiological data in Novosibirsk for the period from 2020 to 2023. The study emphasizes the importance of data post-analysis for understanding the dynamics of SARS-CoV-2 spread and the characteristics of its impact on public health. The study results make possible to assess of how population susceptibility to different virus strains has changed, what is the difference between an outbreak of diseases associated with the emergence of a new virus strain and its seasonal infection spread. There was applied the SSA method for analyzing time series to separate them into components as well as studying key indicators such as the number of new infections, deaths, critical cases, hospitalizations, and ventilator-dependent patients in the Novosibirsk region. Three main components have been identified for the described data sets: a general trend that reflects changes in the rate of virus spread related to spread of new strains, as well as periodic phenomena associated with virus strains and seasonality. The results show that a significant part of the changes in disease dynamics is accounted for by the emergence of new strains, but also due to «chronicity» epidemic with seasonal fluctuations. The observed relationships and time lags between the number of critically-ill patients and the number of recorded deaths due to COVID-19, as well as between the number of hospitalized patients and ventilator-dependent patients are shown. Thus, it is concluded that the identified trend depicting a change between the number of infected people and development of virus strains can be useful for refining the parameters of mathematical models for COVID-19 spread. The SEIR-HCD differential model, which was previously used to simulate the disease spread in the Novosibirsk region, was chosen as an illustrative example. It is shown that the parameter of the virus spread rate, restored through the selected trend, when introduced into the model, provides a smaller modeling error than the if it was generated using the solution of the inverse issue.