Pandemic evolution prediction has started to become particularly complex, as a result of the various parameters that affect the epidemiological course, including (a) the non-pharmacological interventions (such as social distancing, extensive testing and self-testing), (b) the pharmacological interventions (vaccines), (c) the effect of seasonality (d) the different transmissibility, disease severity and reinfection capacity of the various strains.

All the above parameters, have a critical effect on the course of the pandemic, rendering it practically impossible to successfully track it with the simple SIR and SEIR models. On the contrary, all of the above have to be described dynamically in time by more complex equilibriums, having just as a starting point the SEIR models. Accounting for all of the above, and having in mind the decision making support to the national health system over the COVID-19 pandemic, a multi-modal computational tool (called CORE: COVID Risk Evaluation model) for evaluating the COVID-19 epidemic health risk in Greece, Italy and USA has been developed, able to access the impact of various pharmacological and non-pharmacological interventions. CORE also describes in detail the various states of the disease severity by capturing reliably its dynamics, including the impact of risk reduction measures, as well as the impact of the new strains. The CORE model has already been included in the OECD COVID-19 computational toolbox. For defining the contact matrix among the various population groups, accounting for their sociodemographic profiles (i.e. age, occupation etc), as well as the impact of targeted social distancing measures an Agent-Based Modelling (ABM) (Chapizanis et al., 2021) has been employed.

Of particular interest in the latest months, is the introduction in the model of the 2 new dominant strains, i.e. Omicron (by the end of November 2021) and Omicron-2 (by February 2022). The greatest challenge in the modelling effort was the description of the susceptible population equilibrium, since this is affected by (a) the extent to which natural immunity from one strain ensures immunity of the other strain and (b) the differences in the immunity and the efficiency against infection of the people vaccinated with 2 and 3 doses and (c) the gradual loss of immunity, either natural and/or acquired. This was resolved by describing the full set of all the equilibriums of each strain (Δ, Ο, Ο-2) for the various disease states independently, interacting at the level of the susceptible equilibrium; different fraction of the ones that had acquired immunity by one strain, where also deducted from the available susceptible population of the other strains. At the same time, differences in the transmissibility among the strains, as well as the differences of re-infection capacity between strains were explaining very well the pandemic evolution in the number of overall infections on a daily basis. Similarly, the differences in the severity between Δ, Ο, Ο-2 strains have been accounted for, reflected in the successful prediction of the people that were going to require hospitalization in critical condition (intubated in ICU).

The evolution of the COVID-19 dynamics for Greece, as well as the long-term prediction till the end of June 2022, accounting for the current existing measures, vaccination and self-testing rate, is illustrated in Figures 1a (7d moving average of new infections per day) and Figure 1b (daily number of people intubated in ICU). The implementation of the Omicron strain in the beginning of the December in Greece was critical, because in this way we have been able to identify at a very early stage the already significant dispersion of this highly contagious strain and to warn timely the public authorities. In fact, the increase on the rate of change of the new daily cases by mid-December (Figure 1a) could not be explained by the current social distancing measures and the extent of population vaccination, but it could be very well explained only by accounting for a higher extent of Omicron dispersion in the community, than the one that had been identified by the sequencing of random samples. This allowed us to identify the timing and the height of the pandemics peak and the very rapid decline following that. However, the rapid decline was slowed down retarded after 10 days, as a result of the delay in the third dose of vaccination and the significant difference of the efficiency against infection between the second and the third dose.

Similarly, an early warning has been provided for the presence of Omicron 2 by the end of February, in contrast to the common belief that the pandemic was finally going to an end (Figure 1a).

The number of new infections observed nowadays, is significantly higher than the expected when the de-escalation of the 5^th wave began (4^th of January) and indicates loss of immunity. The main reason for this loss of immunity is the delay in the application of the booster dose, given the significant difference in the effectiveness of protection between vaccinated with 2 doses and with 3. Also, a significant loss of immunity exists due to the increased rate of re-infections even in short period of previous infection, which is characteristic of the new subtype of Omicron 2, which is assumed to be even more contagious than variant C by 33% mean (Lyngse et al., 2022). Therefore, the combination of delay of the 3^rd dose and the presence of the new strain, led to a new wave has already further delayed the faster de-escalation.

In general, in the last 3 months, the theory of dynamic equilibrium that has already been formulated by our team has been confirmed. In this case, the relaxation of the social distancing measures by mid-February, allowed the room to the most contagious Omicron 2 mutation to disperse and to result in an additional peak (practically forming a 6th wave), further delaying the rapid de-escalation of the pandemic. However, a very high 6^th wave due to Omicron 2 has been avoided by the additional natural immunity that has been obtained (and added to the overall acquired and already existing natural immunity), as a result of this fluctuation. An additional aggravating (and of course uncertainty) factor that has to be taken into account, is the significant number of re-infections (accounting for almost 10% of the number of daily cases at the moment), as well as the impact of reduced immunity in people who were vaccinated with the 3^rd dose more than 4 months ago. Given the above, loosening existing restriction measures should be done with great attention, with special care.

Although the prediction for the next months is optimistic (also supported by the seasonality effect), given the gradual decline of natural and acquired immunity, as well as the high rate of re-infections related to Omicron 2, in order to reduce the risk of extensive flare-up of the pandemic and to return and sustain de-escalation, it is important to:

(a) continue at a high rate both the booster vaccination and the rapid completion of school-age vaccination; and

(b) continue to responsibly comply with the measures currently in force until de-escalation of the dispersion below 5000 per day is observed, including the high number of the self-tests.

Figure 1. Evolution of the COVID-19 dynamics for Greece, as well as the long-term prediction including (a) the number of new infections per day and (b) the daily number of infected in critical condition (intubated in ICU)

References

CHAPIZANIS, D., KARAKITSIOS, S., GOTTI, A. & SARIGIANNIS, D. A. 2021. Assessing personal exposure using Agent Based Modelling informed by sensors technology. Environmental Research, 192, 110141.

LYNGSE, F. P., KIRKEBY, C. T., DENWOOD, M., CHRISTIANSEN, L. E., MØLBAK, K., MØLLER, C. H., SKOV, R. L., KRAUSE, T. G., RASMUSSEN, M., SIEBER, R. N., JOHANNESEN, T. B., LILLEBAEK, T., FONAGER, J., FOMSGAARD, A., MØLLER, F. T., STEGGER, M., OVERVAD, M., SPIESS, K. & MORTENSEN, L. H. 2022. Transmission of SARS-CoV-2 Omicron VOC subvariants BA.1 and BA.2: Evidence from Danish Households. medRxiv, 2022.01.28.22270044.