The SARS-CoV-2, which causes COVID-19, rapidly spread from the first reported symptoms in Wuhan, China, in December 2019 through to the declaration of an International Public Health Emergency by the World Health Organization in January 2020 (1, 2). The nature of the virus has since changed through its evolution into new variants of concern, with differing potentials for infection, severe disease course, and negation of immunity from vaccination or prior infection (3, 4). The human response to the pandemic has also developed from the implementation of non-pharmacological countermeasures through to pharmacological and vaccination strategies, which have mitigated the impact of the pandemic with varying degrees of effectiveness (5, 6).

In the early pandemic, non-pharmacological interventions were implemented across many regions to ease the immediate disease burden of COVID-19 and prevent the overwhelming of public health systems. Although lockdowns and school/workplace closures were effective in reducing transmission and patient healthcare costs, the economic disruption, declining mental health, and social discontent caused by these interventions raised concerns (7–9). The development and wider rollout of effective vaccines has further changed the dynamics of the situation (10–13). In the context of limited vaccine supply, the WHO recommended administration strategies that aimed to reduce mortality from COVID-19 infection (14). Individuals at risk of infection and/or severe disease were targeted, namely, front-line healthcare workers and older adults, the latter group being at particular risk of morbidity and mortality from COVID-19, compared with younger age groups (6, 15, 16). As supplies have stabilized, campaigns to vaccinate the wider public have faced challenges in the form of hesitancy among certain populations to receive initial doses of the vaccine and/or subsequent boosters (17, 18). Additionally, from 2022, the progression from a prevalent Delta to Omicron variant has shifted the burden of disease to younger people, including pediatric patients in Asia (19–21).

Ongoing high infection rates, an evolving virus, vaccine fatigue, and the relaxation of pandemic countermeasures suggest that COVID-19 will pose ongoing challenges to public health with potential differential effects across age groups (22, 23). Owing to large and diverse populations with different cultural practices and healthcare systems, the development of the COVID-19 situation in Asian settings is likely to pose unique questions for healthcare professionals and policymakers in the region. The dynamics of vaccination against SARS-CoV-2 are complicated by age-dependent factors, type of vaccines used and population vaccine coverage, changing levels of infection and the prevailing strains of concern, and the relaxation of non-pharmaceutical interventions, hence, necessitating the use of mathematical modeling studies in Asia.

Throughout the crisis, epidemiological models have been widely developed and used by policymakers to estimate the impact of interventions on projected disease burden and demands on public healthcare systems (24, 25). Mitigation strategies, including the mandating of mask-wearing, social distancing, restrictions on movement and gatherings, through to pharmaceutical and vaccination strategies, have been implemented to control the spread of virus with varying degrees of effectiveness (5, 6). Although models cannot exactly predict key factors such as the basic reproduction number (R₀), these tools have been applied both globally and in Asia to guide potential care needs in terms of stratifying risk, directing limited resources, and planning for future outbreaks (26, 27). Models of increasing complexity have also been developed to account for the dynamics of transmission among different age groups in the context of different vaccination statuses and waning vaccine effectiveness over time or against emerging variants (28).

Mechanistic models are formulated to mimic the nature of spreading diseases, allowing the simulation of the complex dynamics and nonlinear feedback of COVID-19 transmission within a population (29). An example is the Susceptible-Infected-Exposed-Removed (SEIR) compartmental model, which considers interactions among cohorts of a population, categorized according to disease status. Individuals within a compartment are not differentiated from one another but can flow from one compartment to another at rates defined by parameter inputs. This approach can be effective for examining disease dynamics at a macro level. As a type of mechanistic model, individual-level models treat members of a population as unique agents; this may better capture phenomena such as super-spreader events. Statistically derived models (for example, distribution fitting and regression-type analysis) can also be applied to accurately characterize data from a small number of parameters and are suitable for short-term forecasts that are accurate, repeatable, and sensitive to momentary changes in a system (30). These tools include regression using least-squares or Bayesian estimations. These features make statistical models valuable for producing up-to-date estimates of COVID-19’s underlying nature as a disease (e.g., the basic reproduction rate) and the demographic patterns of the population (e.g., how long people self-isolate). Although statistical models can be highly accurate, they treat the situation as a black box and do not mimic the underlying nature of the disease or its consequences. Thus, these models are not well suited for long-term projections or for exploring hypothetical scenarios.

This review aims to examine published COVID-19 epidemiological modeling studies conducted in Asian settings and assess the models used and their findings regarding the impact of vaccination and non-pharmacological measures on COVID-19 infections, hospitalization, mortality, and policies. We discuss the implications of modeling studies in Asia in the context of the latest data on COVID-19 vaccines, the emergence of SARS-CoV-2 variants, and regional vaccination policies.

We examined modeling studies applied to the COVID-19 pandemic in Asia to identify patterns that might guide decision-making on appropriate and timely countermeasures for the ongoing endemic situation and management of future potential SARS-CoV-2 variants of interest. Although, global studies have raised concerns that modeling studies may not give accurate long-term forecasts of numbers of COVID-19 patients, hospitalizations, and deaths, the scenarios generated by such models are still useful for guiding decision-making and resource prioritization (90–92). Across many regions, modeling studies have also played roles in messaging to the public in order to explain and justify otherwise undesirable interventions (93). Our review showed that data-driven modeling studies play a critical role in the understanding of disease and various preventive policies in Asia.

We identified more reports from East Asian nations, particularly dominated by reports from China. However, the strong tendency to use mechanistic modeling was generally reflected across all regions. Despite the considerable differences in public health infrastructure and substantial variations in population densities, income levels, and sociocultural aspects (82), the common use of mechanistic cohort models across regions suggests that this approach is universally attractive for understanding general disease dynamics and the effectiveness of different interventions (94). This choice may also reflect the intuitive nature of this type of model and the ability to easily develop tools suitable for use by policymakers. Other key advantages of mechanistic models include the insights into the disease that might be gained, particularly from validation of the modeling against real-world evidence. Hence, mechanistic disease-spread models are often used as a basis for cost-effectiveness analyses (95). Additionally, the broad range of symptoms that manifest in COVID-19 introduces considerable uncertainty into national reporting of disease statistics; thus, more sophisticated models (i.e., agent-based models) may not offer particular benefits in accuracy based on available input data.

Particularly early in the pandemic, prior to the wider availability of vaccines, models examined the influence of non-pharmacological interventions on disease-related outcomes. Models set in South and South-East Asian regions were particularly focused on non-pharmacological interventions rather than vaccination, which may reflect the timing of this literature search and the slower vaccine approval, acquisition, and administration in these regions. The majority of models examined here looked at reducing contact through restrictions of movement and gatherings. This focus may reflect the economic disruption caused by these interventions. Although such measures are acknowledged to be effective in reducing disease transmission, their impact is difficult to assess in terms of financial burden and sustainability and compliance; thus, it remains challenging to gauge the accuracy of modeling applied to these measures (96). Few studies here included explicit cost-effectiveness assessments for managing COVID-19 with non-pharmacological interventions or vaccination strategies (47, 88, 89). Whereas vaccination is considered to be cost-effective in terms of treatment costs, strictly enforced countermeasures can pose considerable socio-economic and mental health burdens on populations, which are not fully considered in any of the models identified here (97, 98).

Following the introduction of vaccines from late 2020, reports more often incorporated the effects of vaccination into models. Although information on the specific vaccines used was often limited, actual administration patterns may be inferred from local approval schedules and government procurement policies. Whereas China has almost exclusively used inactivated vaccines (e.g., CoronaVac, Sinovac Biotech), Japan and South Korea granted early approval for mRNA vaccines (e.g., Comirnaty, Pfizer–BioNTech, Pfizer Inc.; Spikevax, Moderna Inc), which have been widely administered. In India, a viral vector vaccine (Vaxzevria, Oxford-AstraZeneca) was used extensively; and across other South and South-East Asian countries, a variety of vaccine types have been deployed due to varying availability and access.

Early recommendations on vaccination strategies were broadly similar, regardless of the vaccination type and local situation, emphasizing the need for rapid distribution and high overall population coverage. Models examining vaccination identified here underline the importance of rapidly achieving 67–83% population coverage with effective vaccines to reduce infection, hospitalization, and/or mortality rates (99). Direct comparisons of modeling results with actual real-world data are complicated by various factors. The methods of diagnosing infection have developed from polymerase chain reaction (PCR) testing initially, through to antigen rapid testing later in the pandemic. Additionally, the implementation of home-based recovery programs in some regions may have underestimated the actual number of infections and hospital admissions (100). Nevertheless, predictions of vaccination outcomes from Asian models have been widely confirmed by lower rates of infection and hospitalization among vaccinated individuals and higher excess mortality among unvaccinated adults across these regions (101, 102).

Reports diverged in terms of specific recommendations for prioritizing vaccination of different age groups. In China and South Korea, where strict social measures were implemented alongside widespread use of inactivated virus or mRNA vaccines, respectively, the majority of models recommended priority vaccination of older adults, a groups that has being prioritized for vaccination since the beginning of the pandemic (103). This is consistent with results from early global modeling studies that also prioritized older adults (90, 104, 105).

In Japan, which implemented voluntary social distancing and predominantly administered mRNA vaccines, one study noted a potential for benefits in terms of indirectly reducing mortality by targeting younger age groups to reduce overall transmission, depending on the strength of the lockdown (89). The actual vaccination strategy implemented in Japan, however, prioritized older adults, which may have helped reduce the hospital burden but could have been suboptimal in reducing transmission, particularly in the context of weakly enforced social measures. However, attitudes towards COVID-19 suggested remarkable concern among the Japanese public, which contributed to good compliance with government advice to wear face masks, social distance, and accept vaccination/boosters (106). Globally, the attitudes of populations towards the disease, trust in policy makers, and enforcement of interventions are factors that may need to be accounted for in modeling studies (107).

The evolution of new viral strains may also influence vaccination strategies. The main variants of SARS-CoV-2 circulating in different regions have changed, from the original strain prevalent in 2020, to four main variants of concern – the Alpha, Beta, and Delta strains, common across Asia in 2021 – through to the Omicron variants from 2022 (3, 4, 108). Notably, the changes from the original variant to the Delta variant and then later to the Omicron variant were accompanied by increases in R₀ values, from 2.79 to 5.08 to 8.20, respectively (109, 110). The recently emerged Omicron subvariants, such as XBB and XBF, appear to be key drivers of infection waves, although they appear to pose lower risks of mortality (111). The emergence of new variants presents challenges for modeling in terms of shifting profiles for infectiousness, severe disease and mortality, and breakthrough infection against different vaccines and vaccination statuses (i.e., number/type of vaccination and time since last booster) (112). For example, individuals with hybrid immunity (developed through both infection and vaccination) might experience greater and more sustained protection, which might allow for longer intervals between boosters (113).

The outcomes of many models have anticipated that sufficiently high vaccination coverage in a population might ‘break’ transmission, allowing for the relaxation of non-pharmacological interventions (71, 114). In reality, vaccinated populations in Asia have continued to fall short of the high proportions predicted to be necessary to confer ‘herd immunity’ despite large numbers having received initial vaccinations and boosters (115). The waning effectiveness of mRNA vaccination/booster shots over time (after approximately 20 weeks) – particularly among older adults, who typically have weakened immune systems – may increase rates of infection (116). The potential impact of immune imprinting also motivates further studies to understand the effects and limitations of boosters (117). Reduced effectiveness of vaccines against Omicron strains might also shift initial targets for vaccination coverage in a population upwards, obstructing the relaxation of non-pharmacological interventions (118). Ongoing viral transmission from viral shedding from convalescent individuals, in addition to cessation of mask usage, may contribute to future surges of the disease.

There is also a need to deeply understand the importance of using vaccines that confer broad protection in the Omicron era, such as bivalent vaccines (119). Recent real-world evidence on bivalent vaccines points to additional protection, primarily against hospitalization, among older adults with monovalent vaccines; however, there may also be some additional benefits against symptomatic disease in other age groups during Omicron subvariant circulation (120). The high-priority populations are defined as older adults, adults with significant comorbidities, children or adults with immunocompromising conditions, pregnant women, and frontline health workers.

We noted that few studies (32, 38, 55) examined here modeled the vaccination of pediatric populations in Asia. Modeling of the impact of vaccinating children (5–11 years) in Europe has been proposed as a route to the relaxation of restrictions in schools (121). Modeling studies have also suggested that extending vaccination to children may reduce hospitalization and mortality across all age groups in developing countries (122). The latest update from the WHO’s Strategic Advisory Group of Experts on Immunization (SAGE) on COVID-19 vaccination continues to prioritize the populations at greatest risk of mortality or severe disease to safeguard healthcare systems. As a low priority group, children are not routinely recommended for vaccination and boosters, but are instead determined by individual countries’ evaluation of the burden, cost, and effectiveness (123). Although data on the impact of COVID-19 among children are lacking overall, there appears to be greater risk of mortality and hospitalization in low and middle-income countries (124). Thus, there is a need for further studies focused on the local burden of COVID-19 disease among younger populations in Asia, where future modeling studies might also guide decision-making on the impact of vaccinations (121, 122, 125).

Moving into a period where more vaccines are approved and made accessible, the hesitancy of uptake is likely to pose an ongoing barrier to booster strategies and sustainable protection (126). Among children and their caregivers, common reasons for fear and hesitancy to receive a vaccine include assumptions that the vaccine has side effects and may not be safe (18). Adults may also be complacent that the COVID-19 situation is no longer severe, and/or a negative perception of the vaccine, government, and/or pharmaceutical company may discourage them from receiving initial vaccination and/or boosters (17, 18). General fatigue over the topics of COVID-19 and vaccines and a wider circulation of misinformation on vaccines may contribute to these attitudes (127). Some reports examined here accounted for vaccine hesitancy, but this factor will likely be more important in future COVID-19 models, where the majority of a population have been vaccinated but coverage falls short of requirements for herd immunity. Modeling may also need to account for the impacts of educational interventions to address gaps in knowledge and attitudes among hesitant groups. Therefore, from a public health perspective, it is crucial to understand the drivers of vaccine uptake and vaccine hesitancy; this might help to identify groups that might have lower than average uptake and plan accordingly. Such pockets of immunity gaps and high susceptibility in the population could result in small-scale outbreaks that reduce the effect of population immunity. We predict that relaxation of control measures might be associated with new waves of infection and associated deaths; however, these outcomes will be reduced by increased levels of vaccine-derived immunity as well as hybrid immunity from infections in vaccinated individuals in the population.

The evolving nature of the COVID-19 situation especially motivates a dynamic vaccine-development pipeline to deliver more effective options against new strains of SARS-CoV-2 (128, 129). Furthermore, in an endemic scenario, non-vaccine pharmacological interventions are likely to become important for the treatment of severe disease. Here, two studies were identified that considered the impact of antiviral treatments, which may reflect the regional timeline for approval of such medicines. The introduction of monoclonal antibodies may help to manage cases of severe COVID-19 and improve these outcomes (10).

There are several limitations to this study. First, the scope of the literature search, particularly the focus on journal articles published in the English language, may have omitted more recent pre-print and local-language documents. The specific search terms may also have introduced bias into the types of modeling studies captured here. Furthermore, the large diversity of regional situations, model designs, interventions, and outcome targets limit the ability to systematically extract relevant quantitative data from these reports. Additionally, few studies examined the influence of pharmacological interventions other than vaccines. As common limitations of all modeling studies, there is considerable uncertainty in many parameters such as population immunity, infectious rate, and contact and the level of health-related behaviors in study populations (130). Thus, modeling studies are subject to unreliable predictions on the impact on outcomes of interest.

Computational modeling is an important tool to address a limited understanding of SARS-CoV-2 epidemiology, and the quantitative estimates of the duration of protection from infection and vaccinations. Whereas non-pharmacological interventions have had an early role in managing health outcomes of the COVID-19 pandemic, modeling studies underline the importance of vaccines. High population coverage and vaccine effectiveness are key to mitigating the outcomes of the disease and supporting the relaxation of disruptive restrictions on movement. Compartmental mechanistic modeling offers an adequate approach to projecting disease outcomes across large populations; however, future models may need to account for complications from evolving variants and vaccination statuses/inclinations within populations, age group stratification, especially including pediatric populations. Deeper consideration of the socioeconomic burdens associated with strict non-pharmacological interventions may also be useful for policymakers and further underscore the cost-effectiveness and social benefits of vaccination programs. In the near future, increasing vaccine coverage, particularly among at-risk populations and through outreach to vaccine-adverse groups, may help to ease infection and severe disease. The rapidly evolving nature of the virus also motivates the development of new vaccines and alternative therapies. Future administration strategies may also be nuanced by competing needs to reduce overall transmissibility or severe disease burden in high-risk groups.

KT, JS, EO, JY, and MK: conceptualization and methodology. KT, JS, JY, MK, and CM: formal analysis and data curation. KT, JS, JY, MK, EO, CM, and HO: validation, writing, reviewing, and editing of manuscript. All authors contributed to the article and approved the submitted version.