Forecasting the trajectory of the COVID-19 pandemic into 2023 under plausible variant and intervention scenarios: a global modelling study

Background The recent Omicron-related waves of the COVID-19 pandemic have resulted in unprecedented levels of population transmission due to the variant’s high level of infectiousness across most of the world. China, the last large country to end its “zero-COVID” policies, is currently facing its own massive Omicron-related wave, and the final impact of that wave remains uncertain. We have seen repeatedly that the epidemiological characteristics of new variants can have profound impacts on global health outcomes. While the characteristics of these new variants are difficult to predict ahead of their emergence, considering the impact of potential future scenarios is of central importance for prudent planning and policy making. This paper samples across a range of potential variant-level characteristics to provide global forecasts of infections, hospitalisations, and deaths in the face of ongoing Omicron-related transmission and waning levels of past immunity and evaluates a range of interventions that may diminish the impact of future waves. Methods We created a susceptible-exposed-infectious dynamic model that accounts for vaccine uptake and effectiveness, antiviral administration, the emergence of new variants, and waning protection from both infection- and vaccine-derived immunity. Using this model, we first estimated past infections, hospitalisations, and deaths by variant, location, and day. We used these findings to more fully understand the global progression of the COVID-19 pandemic through December 12, 2022. Second, we forecasted these same outcome measures under five potential variant emergence scenarios. Third, we evaluated three different interventions in isolation and in concert within each potential variant scenario, to assess the impact of available intervention strategies through June 30, 2023. Findings We estimated that from November 15, 2021, through December 12, 2022, there were 8.60 billion (95% uncertainty interval [UI] 6.37–11.7) SARS-CoV-2 infections, 13.1 million (10.6–16.5) hospitalisations, and 3.04 million (2.65–3.55) deaths, the majority of which were attributable to Omicron variants (98.5% [97.4–99.1] of infections, 82.6% [76.7–86.3] of hospitalisations, and 72.4% [66.4–76.0] of deaths). Compared to the pre-Omicron pandemic period from January 1, 2020, to November 15, 2021, we estimated that there were more than twice as many infections (214% [163–286]) globally from November 15, 2021, to December 12, 2022, but only 20.6% (19.8–21.4) of the estimated deaths. The massive Omicron waves and high vaccination rates in many high-income countries have together contributed to high levels of immunity against SARS-CoV-2 infection, leaving only 97.3% (96.3–98.2) of the global population with no protection as of December 1, 2022. Concurrently, however, China, where only 17.6% [5.28–34.8] of the population have ever experienced infection due to its zero-COVID policy, requires special attention over the next few months, as all our future scenarios predict substantial increases in transmission, hospitalisation, and death in China in now that zero-COVID policies have been relaxed. Under the future scenario we consider most plausible (a scenario with another new Omicron-like variant emerging and reference levels of the drivers of transmission), we estimated there will be an additional 5.19 billion (3.11–7.78) infections, 13.6 million (8.50–21.8) hospitalisations, and 2.74 million (1.40–5.68) deaths between December 12, 2022, and June 30, 2023, with the Western Pacific region projected to sustain the highest rates of additional deaths, driven primarily by the uncontained outbreak in China. By comparison, a baseline scenario in which no new variant emerges results in 3.54 billion (2.24–5.43) infections, 6.26 million (4.11–9.65) hospitalisations, and 1.58 million (0.829–3.95) deaths in the same forecast period. The ability for a new variant to break through past infection- and vaccine-derived immunity greatly influences future outcomes: we estimate a new variant with the high severity of Delta, but correspondingly moderate immunity breakthrough rates will have difficulty overtaking current variants and will result in similar outcomes to the Omicron-like variant scenario with 3.64 billion (2.26–5.83) new infections, 7.87 million (4.81–13.0) new hospitalisations, and 2.87 million (1.03–5.56) new deaths. Finally, if we consider a variant that combines the high infectiousness and breakthrough rates of Omicron with the high severity of Delta, we again estimate 5.19 billion (3.11–7.78) new infections, but due to the presumed increase in severe outcomes, we estimate 30.2 million (13.4–51.2) new hospitalisations and 15.9 million (4.31–35.9) deaths over the forecasted period. The impacts of interventions vary by variant characteristics and region of the world, with increased mask usage and reimplementation of some mandates having massive impact in some regions while having less impact in others. Finally, assuming variant spread was as rapid as observed for Omicron, we find almost no impact of a rapidly developed and deployed variant-targeted booster. Interpretation As infection-derived and vaccine-conferred protection wanes, we expect infections to rise, but as most of the world’s population has some level of immunity to SARS-CoV-2 as of December 12, 2022, all but the most pessimistic forecasts in this analysis do not predict a massive global surge by June 30, 2023. Paradoxically, China, due to its lower levels of population immunity and effective vaccination will likely experience substantial numbers of infections and deaths that, due to its large population size, will adversely affect the global toll. This could be substantially mitigated by existing intervention options including masking, vaccination, health-care preparedness, and effective antiviral compounds for those at most at risk of poor outcomes. While still resulting in morbidity and mortality, this endemic transmission provides protection from less transmissible variants and particularly protects against sub-lineages of the more severe pre-Omicron variants. In the scenarios where a new variant does emerge and spread globally, however, the speed of this spread may be too fast to rely on even the most quickly developed mRNA vaccines to provide protection soon enough. Existing vaccines and boosters have played an important role in increasing immunity worldwide, but the continued contribution of mask usage (both past and future) in the prevention of infection and death cannot be understated. The characteristics of future COVID-19 variants are inherently difficult to predict, and our forecasts do show considerable differences in outcomes as a function of these variant properties. Given the uncertainty surrounding what type of variant will next emerge, the world would be wise to remain vigilant in 2023 as we move to the next phase of the COVID-19 pandemic. Funding Bill & Melinda Gates Foundation, J. Stanton, T. Gillespie, and J. and E. Nordstrom. Evidence before this study Since the beginning of the COVID-19 pandemic, there have been a plethora of COVID-19 models developed; most were designed to focus on a specific location (or small set of locations) and a short time horizon (usually less than a month). A number of modelling consortiums were created to develop ensemble predictions across models of this sort (e.g., the COVID-19 Forecast Hub [maintained by the Reich Lab of the University of Massachusetts Amherst in collaboration with USA CDC (Centers for Disease Control)] or the European COVID-19 Forecast Hub [created by a multitude of infectious disease modelling teams and coordinated by ECDC (European Centre for Disease Prevention and Control)]), and the final results typically predicted four weeks, and at most, six weeks forward. The models combined for these ensembles ran the spectrum from transmission dynamic models that incorporated complex mixing patterns between individuals, to machine learning models that were agnostic of the fact that the input and output were associated with infectious diseases. Moreover, most of these models were designed to predict the most likely outcome as opposed to evaluate potential future scenarios. A small subset of these models were created with this sort of flexibility, though they have primarily been applied to limited global regions (e.g., USA CDC scenarios) and they typically do not evaluate multiple potential scenarios three to six months into the future. The Institute for Health Metrics and Evaluation (IHME) COVID-19 model has been generating and publishing forecasts of SARS-CoV-2 infections and COVID-19 deaths globally with four-month time horizons and making these available at mostly weekly intervals on its website since March 26, 2020 (). The cadence has now slowed to monthly updates as in many parts of the world, data needed to support the modelling of COVID-19 have reduced and/or ceased to be collected as the attention of policy makers and funders is drawn elsewhere. Several epidemiological scenarios have been evaluated in these online estimates, but the outcomes have not been formally compared across these scenarios globally into 2023. This article is also the first full formal documentation of the IHME-SEI model incorporating foundation work on infection–fatality ratio, more robust cumulative infection calculations, as well as more recently developed work that allows for waning immunity. Added value of this study To our knowledge, this study is the first to forecast multiple future COVID-19 scenarios of variant emergence against a background of high rates of past SARS-CoV-2 exposure globally, nationally, and for a set of subnational locations, six months into the future. It is also one of the first to forecast the impact of China relaxing its zero-COVID policy. The scenarios considered were selected to represent a range of realistic potential futures and are directly comparable by region, country, and territory (and in many instances subnational units), to identify future risk as well as inform on the effectiveness of potential intervention strategies. In particular, we directly compared scenarios where a future variant is presumed to be similar to Omicron (high infectiousness, low severity, high immune-breakthrough), Delta (moderate infectiousness, high severity, moderate immune-breakthrough), a Delta with increased immune escape (moderate infectiousness, high severity, high immune-breakthrough), or the worst of both (high infectiousness, high severity, high immune-breakthrough) to a scenario where no new variant emerges. We then evaluated several interventions against each potential variant future, each in isolation and in concert. In addition to providing timely predictions for China as they remove restrictions, we provide insight into which locations may be at highest risk for future COVID-19 infections, hospitalisations, and deaths, and what they might do to mitigate the worst possible outcomes. Implications of all the available evidence The Omicron waves have already resulted in an estimated 8.60 billion (95% uncertainty interval [UI] 6.37–11.7) infections in the past 13 months globally (from November 15, 2021, through December 12, 2022). Previous exposure to other variants and vaccination have together resulted in 97.3% (96.3–98.2) of the global population being estimated to have some immunity to SARS-CoV-2 as of December 12, 2022. While infection- and vaccine-derived immunity has and will continue to wane, this protection and ongoing transmission of currently circulating variants will mitigate the scale of the next COVID-19 wave. The scale of mitigation possible is highly dependent on the characteristics of the next variant. To assess the potential for a COVID-19 surge in early 2023, we evaluated several future variant scenarios, as well as the unlikely baseline scenario of no new variant emerging. In the absence of any new variant, our baseline model predicts 1.58 million (0.829–3.95) deaths globally between December 12, 2022, and June 30, 2023. If a variant with similar characteristics to Omicron (eg, high infectiousness and low severity) emerges on January 15, 2023, our model predicts 2.74 million (1.40–5.68) additional deaths over the same period. A variant with the characteristics of Delta is predicted to have difficulty overtaking current variants and past immunity, and despite its substantial severity, our model predicts a number of deaths similar to an Omicron-like new variant (2.87 million [1.03–5.56]). In the worst-case scenario considered, a variant with the transmission and breakthrough characteristics of Omicron and the severity of Delta would result in 15.9 million (4.31–35.9) deaths, 14.3 million (3.33–32.7) more than a scenario where no new variant emerges. In China, the potential morbidity and mortality in 2023 is high, due to a combination of pandemic history and policy that has kept levels of population immunity to COVID-19 low. In our “worst case” variant scenario, we estimate initiatives to return mask use to 80% of the population (or the location-specific current level, if higher) as well as the reimplementation of moderate mandates would avert 32.8% (18.7–51.3) of the predicted deaths, with maximal impact occurring in the European region (44.8% [28.7–61.6]). In every variant scenario, given the estimated speed of global spread, we predict that variant-targeted mRNA boosters are not able to be deployed soon enough to have a substantial impact. While there is considerable uncertainty in the future of COVID-19 variant characteristics, this study demonstrates a range of plausible outcomes expected across a spectrum of future realities. Although it would require nations to react quickly to newly detected threats, our predictions show that increased mask use and (where necessary) reimplementation of moderate social distancing mandates can mitigate much of any future challenge. ### Competing Interest Statement The authors have declared no competing interest. ### Funding Statement Funding was provided by the Bill & Melinda Gates Foundation, J. Stanton, T. Gillespie, and J. and E. Nordstrom. The funders of the study had no role in the study design, data collection, data analysis, data interpretation, or the writing of the report. Members of the core research team for this topic area had full access to the underlying data used to generate estimates presented in this paper. All other authors had access to, and reviewed, estimates as part of the research evaluation process. ### Author Declarations I confirm all relevant ethical guidelines have been followed, and any necessary IRB and/or ethics committee approvals have been obtained. Yes I confirm that all necessary patient/participant consent has been obtained and the appropriate institutional forms have been archived, and that any patient/participant/sample identifiers included were not known to anyone (e.g., hospital staff, patients or participants themselves) outside the research group so cannot be used to identify individuals. Yes I understand that all clinical trials and any other prospective interventional studies must be registered with an ICMJE-approved registry, such as ClinicalTrials.gov. I confirm that any such study reported in the manuscript has been registered and the trial registration ID is provided (note: if posting a prospective study registered retrospectively, please provide a statement in the trial ID field explaining why the study was not registered in advance). Yes I have followed all appropriate research reporting guidelines, such as any relevant EQUATOR Network research reporting checklist(s) and other pertinent material, if applicable. Yes All code used in the analysis can be found online . .