COVID-19 vaccine coverage targets to inform reopening plans in a low incidence setting

1. Introduction

In early 2020, Australia adopted a strong suppression strategy in response to the COVID-19 pandemic, aiming for no community transmission of the SARS-CoV-2 virus [1]. As a result of a broadly successful implementation of this strategy, which included international and internal travel restrictions, much of the population lived in a COVID-19-free environment up until late 2021. Nonetheless, sporadic outbreaks and a number of major (but geographically isolated) waves of infection occurred, most notably the ‘second wave’ in the state of Victoria in June–October 2020 [2] and the Delta ‘third wave’ seeded into New South Wales in June 2021 [3], which quickly spread into neighbouring jurisdictions.

Like elsewhere, the availability of highly effective vaccines for COVID-19 from early 2021 provided a new opportunity to protect the population and reduce harms related to SARS-CoV-2 [4]. However, the transition from a state of minimal (daily and cumulative) incidence and low vaccine coverage to one of high vaccine coverage and established but manageable transmission presented unique challenges compared to vaccination roll-out in high incidence and high pre-existing immunity settings.

In much of Europe and North America, where COVID-19 has circulated widely since its emergence and both daily and cumulative incidence were high at the time of vaccine availability, vaccination provided a very clear, although challenging, ‘re-opening’ pathway. As vaccine campaigns were initiated, with ancestral and Alpha variants in circulation, increasing vaccine coverage reduced the ability of the virus to spread and assisted in bringing case incidence down [5]. While the emergence of the Delta variant led to a resurgence in cases, there was clear evidence that vaccination played an important role in ‘decoupling’ clinical caseloads and deaths from mild infections [6].

By contrast, for countries such as Australia, where pre-vaccination cumulative incidence was very low and the population remained largely susceptible, the relationship between vaccination, incidence and the public perception of COVID-19 was markedly different. Any plan to ‘re-open’ society and transition away from a strong suppression strategy would necessarily lead to an increase in cases, morbidity and mortality, with clear challenges for communication and decision making. The ‘National Plan to transition Australia's National COVID-19 Response’ (hereafter the National Plan) [7], released in July 2021, describes a transition from ‘phase A’ in which strong suppression and no community transmission is the goal, to ‘phase B’ whereby SARS-CoV-2 infection was allowed to establish in the broader population. This transition was enabled by vaccination. The aim was for COVID-19 to be manageable, from both a public health and clinical perspective, through continued but more targeted application of public health and social measures (PHSMs) and test–trace–isolate–quarantine (TTIQ) strategies. The targets in the National Plan were agreed by all Australian States and Territories at National Cabinet (forum for the prime minister, state premiers and chief ministers to meet) on 6 August 2021. The transition from phase A to phase B was the first step on the path to ‘re-opening’. Latter phases ‘C’ and ‘D’ in the National Plan were supported by higher vaccine coverage levels allowing for further reductions in whole-of-population pandemic responses. Exploring target coverage thresholds for vaccination to enable the transition from phase A to phase B, and the required level of supporting interventions including ongoing PHSMs and TTIQ capabilities was identified as a priority under the National Plan. More broadly, identifying such thresholds was a common challenge for low incidence settings other than Australia [8,9].

Here, we report on the dynamic modelling study used to identify priority age groups for vaccination, target coverage thresholds, and the anticipated requirements for continued PHSMs and TTIQ to support a transition to a more open society in which SARS-CoV-2 could circulate while the health and clinical impacts remain manageable. Findings from this research directly informed specific re-opening policies within the National Plan and in particular the threshold and conditions for the transition from phase A to phase B.

2. Modelling framework

The modelling framework comprises three distinct components (depicted in figure 1), each summarized here and described in detail in the electronic supplementary material.

• Download figure
• Open in new tab
• Download PowerPoint

These three components are underpinned by an analysis of SARS-CoV-2 transmissibility in Australia, measured by the ‘transmission potential’ (TP). As described elsewhere [11], the TP draws on a number of behavioural data streams to estimate (at a state level) the reproduction number that could be expected during widespread transmission. The TP is distinct from the effective reproduction number in that it represents the expected reproduction number of a pathogen in the general population rather than the reproduction number among active cases (who may not be representative of the general population). The TP is included as an indicator in Australia's national COVID-19 surveillance plan [12] and thus routinely reported for all states and territories of Australia and made publicly available through the Australian Government's Common Operating Picture [13]. Under pre-pandemic conditions, the TP corresponds to the basic reproduction number, R₀. It varies through time due to changes in health system performance, different public health orders, and trends in population mixing and behaviours (such as limiting non-household contacts and adherence to cough etiquette and other infection control recommendations). Furthermore, because vaccines act to reduce transmission (through a reduction in the probability of acquisition and reduced contagiousness given breakthrough infection), additional reductions in TP due to vaccination can be estimated in the context of behavioural and public health response settings.

Using the TP model and time-series data on cases and population behaviours in Australia since March 2020, we first conducted a static analysis (described in [14]), to estimate TP for the Delta variant achieved under alternate vaccine allocation strategies, PHSMs and TTIQ capabilities. An example output from this analysis is displayed in figure S1 of the electronic material (adapted from Ryan et al. [14]). For the static analysis, we considered four ‘bundles’ of PHSMs: baseline, low, medium and high. Each bundle relates to a specific time and place in Australia's pandemic experience, thereby capturing both behavioural responses and the proportional reduction in TP achievable by PHSMs in the Australian context (table 1 for details).

The static analysis also considered the contribution of TTIQ in reducing virus spread, which is described in detail in [15]. Briefly, this included a detailed study of a limited time-series of case data from New South Wales between July 2020 and January 2021. During this time, caseloads in New South Wales were low and the impact of TTIQ was clear in suppressing transmission [11]. The empirical distribution of times from case detection to isolation was determined and used to evaluate the reduction in transmission due to TTIQ. This reduction—54%—defines an ‘optimal’ TTIQ effect, representing the maximum reduction in TP that we may expect due to TTIQ at any time during an epidemic. By assuming improvements in TTIQ are proportional to improvements in times to detection (i.e. times from symptom onset to test), the relative performance of TTIQ was measured during different periods of epidemic activity. This approach provided a distribution of times to isolation in epidemiological contexts where TTIQ was assessed to be partially effective. In particular, when calibrated against the data for the state of Victoria from 4 August 2020—the peak of daily locally acquired COVID-19 cases in Australia in 2020—this gives an estimated reduction of 43% in TP and defines the ‘partial’ TTIQ scenarios explored in the static analysis.

For all scenarios considered in this study, the dynamic transmission model uses a baseline reproduction number of 6.32, corresponding to a TP estimated for the Delta variant under baseline PHSM and after ‘backing out’ the effect of TTIQ (indicated by the dashed line ‘Baseline TP’ in figure S1 in the electronic supplementary material). The baseline TP is lower than the R₀ for SARS-CoV-2 due to the baseline behavioural changes in the Australian population. It is further reduced based on the assumed impact of (partial or optimal) TTIQ, enhanced (low, medium or high) PHSMs and vaccination.

Here, we examine epidemic dynamics and clinical consequences of infections following transition from phase A to phase B of the National Plan (table 2) at different vaccine coverage thresholds between 50% and 80%, for alternative age-based vaccine allocation strategies, and assuming continued application of PHSMs and TTIQ.

3. Results

A comparison of time-series for daily infections, by vaccine allocation strategy and vaccine coverage threshold for transition from phase A to phase B, is presented in figure 2. The corresponding time-series for ward and ICU occupancy, and deaths, are displayed in figures S7–S10 of the electronic supplementary material, and cumulative infections and deaths are displayed in figures S11 and S12 of the electronic supplementary material. Here, baseline PHSMs and partial TTIQ are assumed to be in place. These analyses demonstrate that the ‘All adults’ and ‘Transmission reducing’ strategies result in fewer infections (figures 2 and electronic supplementary material, figure S11) and afford greater clinical protection (electronic supplementary material, figures S8–S10 and S12) compared to the ‘Oldest first’ and ‘40+ years first’ strategies at all transition coverage levels. Furthermore, marked reductions in the time to epidemic peak and peak size are noted as vaccine coverage increases from 50% to 60% and beyond.

• Download figure
• Open in new tab
• Download PowerPoint

The ‘Transmission reducing’ strategy was designed as an implementable version of the ‘All adults’ strategy and is considered hereafter. Furthermore, when our initial results (shown in figure 2 and electronic supplementary material, figures S7–S10) were considered alongside a complementary analysis that specifically considered the time in lockdown and economic consequences for different vaccine thresholds (see [14] for a description of the modelling analysis and [23] for the government report), they indicated that adult vaccine coverage of least 70% was required to support a transition compatible with Australia's strategic road map [10]. We therefore now restrict our focus to vaccine coverage thresholds of 70% and above.

In recognition of the Delta variant becoming established in some Australian jurisdictions prior to reaching 70% coverage thresholds, we examine the consequences of an increased case load at the time of transition from phase A to phase B in the National Plan. Row one of figure 3 presents time-series of infections (asymptomatic and symptomatic combined) for three initial conditions following transition from phase A to phase B at either 70% (left) or 80% (right) coverage. Baseline PHSMs and partial TTIQ are applied. For a transition at 70% coverage, an increase from ‘low’ (tens) to ‘medium’ (hundreds) numbers of infections at the time of transition results in a marked leftward shift in the timing of the epidemic. This result is unsurprising given basic epidemic theory. However, if the transition occurs at a point with ‘high’ (thousands) infections, the epidemic is not only left-shifted, but peak and final size also increase. This is a result of dynamic ‘overshoot’, otherwise avoidable due to the continued roll-out of vaccines and hence increasing coverage level during the epidemic. If the transition from phase A to phase B is made at 80% coverage, the leftward shift in epidemic dynamics remains, but even for the ‘high’ (thousands) infections scenario, there is minimal impact on peak and overall size of the epidemic. This is due to two factors: (i) vaccine coverage is sufficiently high to prevent a large ‘overshoot’; and (ii) under the model for vaccine distribution, the rate of increase in coverage slows beyond 80% as we approach saturation of vaccine uptake in the eligible population (around 89.9% of the eligible population, reached at 56 weeks). Corresponding time-series for clinical outcomes are presented in rows two to four of figure 3.

• Download figure
• Open in new tab
• Download PowerPoint

Australia's National Plan envisages continued application of enhanced PHSMs to further suppress epidemic activity and minimize morbidity and mortality. Figure 4a,b demonstrates the marked benefit of continued application of low PHSMs, accompanied by partial TTIQ, from the point of transition from phase A to phase B. At both 70% (figure 4a) and 80% (figure 4b) the epidemic is strongly suppressed compared to under baseline PHSMs (figure 3). Similar beneficial outcomes can be achieved through application of optimal TTIQ under baseline PHSMs (electronic supplementary material, figure S13), although we note that maintaining optimal TTIQ was not considered feasible over the long term.

• Download figure
• Open in new tab
• Download PowerPoint

While all of the above simulations assume a fixed policy from the point of transition from phase A to phase B, substantial benefits may be realized by the imposition of increased restrictions for limited periods of time. Figure 4c shows how an adaptive strategy can support a transition at 70% even under ‘high’ numbers of infections. Medium PHSMs are applied during the period from 70% to 80% coverage, with low PHSMs enacted thereafter. Compared to the scenario in which low PHSMs are active during this transition period (figure 4a), infections are notably reduced. Overall epidemic impact is similar to when the transition from phase A to phase B is only made at 80% (figure 4b).

Figure 5 compares the cumulative number of symptomatic infections and clinical impacts by age group and vaccine status for outbreaks seeded at 50% and 80%, assuming baseline PHSMs and partial TTIQ. There is a significant overall reduced burden given establishment of community transmission at 80% compared to 50% coverage (figure 5a,b). In this context, a substantial fraction of symptomatic infections and severe outcomes are anticipated in vaccinated individuals within highly vaccinated age groups. At both 50% and 80% vaccine coverage, a decoupling of infections from clinical burden is evident, with infections concentrated in younger less vaccinated populations, yet much greater health impacts are anticipated in older more highly vaccinated age groups (figure 5c,d).

• Download figure
• Open in new tab
• Download PowerPoint

Results of our post hoc sensitivity analysis exploring the impact of more optimistic and pessimistic parameters for clinical severity and vaccine effectiveness, given epidemic seeding at 50%, 60%, 70% and 80% vaccine coverage thresholds, are shown in figures S14–S20 of the electronic supplementary material. The patterns in epidemic dynamics between coverage thresholds for alternate parameters sets are not markedly different from those seen in the primary analysis.

4. Discussion

To transition from a strategic goal of ‘no community transmission’ to one of ‘minimizing COVID-19 burden’ requires sufficient vaccine coverage to (i) suppress case incidence such that TTIQ remains an effective response to reduce transmission and (ii) ensure anticipated health and clinical impacts remain manageable. Here, through a model-based analysis, we have demonstrated the requirement for adult vaccine coverage of at least 60% using the ‘Transmission reducing’ allocation strategy to achieve these objectives. This result assumes the continued application of TTIQ (partial) and PHSMs (baseline) during the vaccine roll-out phase and beyond. These dynamic results were considered by policy makers alongside a complementary analysis that specifically considered the risks of escalating case numbers under different vaccine thresholds, requiring re-imposition of lockdown with associated economic consequences (see [14] for the modelling analysis and [34] for the government report). Based on the combined analysis of health and economic risks, a threshold adult vaccine coverage of least 70% was determined as the target to support a transition compatible with Australia's strategic road map [10]. Our analyses were conducted based on assumed circulation of the Delta variant of SARS-CoV-2 and contingent upon the Australian epidemiological, health system and societal context in which circulation of SARS-CoV-2 was strongly suppressed up to the point of transition.

If, as was the case in the state of Victoria, case incidence was high (thousands per day) at the time of reaching the 70% coverage threshold, then heightened but temporary public health and social measures (medium PHSMs) could help bridge the period to achieving 80% coverage, reducing the risk of a surge in transmission that may threaten capacity. With coverage at 80% or beyond, our analyses indicate that epidemic dynamics would likely be manageable within the constraints of the clinical system and only baseline PHSMs and partial TTIQ in place, compatible with Australia's strategic roadmap for managing COVID-19 in the vaccine era. Further gains—in terms of reduced infections and so reduced hospitalizations and deaths—can be afforded by maintenance of low PHSMs over the vaccine roll-out phase and beyond. Furthermore, those low PHSMs would support a strong and more effective TTIQ response, helping avoid escalation of local epidemic activity. If the transition to ‘living with COVID’ were to occur prior to reaching 70% coverage, case numbers would likely rise to such a level that TTIQ effectiveness was diminished and epidemic ‘overshoot’ would result in additional—and a priori avoidable—cases, hospitalizations and deaths.

Our findings are comparable to those from model-based studies for other low prevalence settings [8,9]. Nguyen et al. [8] and Steyn et al. [9] investigated the impacts of alternative age-based vaccine allocation strategies and coverage thresholds in the New Zealand context. Similar to Australia, New Zealand's vaccine roll-out was intended to support a shift in response strategy from elimination to border re-opening and virus circulation. Both studies concluded that high vaccine uptake (e.g. greater than 80% of the population aged 16+) and maintenance of other public health measures during the vaccine roll-out phase would be required to prevent serious adverse health impacts. Furthermore, studies from high prevalence settings, including the United Kingdom, also highlighted the potential adverse health and health system impacts of complete relaxation of social restrictions during the early phases of vaccine roll-out [4]. Of course, our specific findings on vaccine coverage thresholds are limited by the low level of transmission in Australia prior to vaccine roll-out and are not generalizable to higher prevalence settings.

A key limitation of our work is that we considered a single, large population (24 million) in which the virus spreads. This was a deliberate and necessary choice designed to support the Australian national (whole-of-country) re-opening plan. However, and particularly in the early establishment phase of a country wide epidemic, transmission was expected to be highly focal. Jurisdictions where SARS-CoV-2 transmission established prior to the 70% coverage threshold, such as New South Wales and Victoria, began a transition from a state of ‘medium’ and ‘high’ case incidence, respectively (electronic supplementary material, figures S21 and S22). Other jurisdictions maintained zero case incidence well beyond vaccine coverage thresholds of 70% and 80%. Furthermore, at a sub-jurisdictional level, we would expect systematic differences in vaccine coverage, behavioural patterns and TTIQ capabilities. These considerations emphasized the need for small-area assessment of TP (to anticipate risk) and other real-time epidemiological metrics.

More broadly, the goal of scenario analyses such as those reported here is to provide insights on potential patterns in epidemic activity given different assumptions about the future, including what intervention options are chosen. Scenario analyses are not forecasts or predictions of the future course of the epidemic, not least because of uncertainty in key model inputs, including vaccine effectiveness, the duration of immunity, intrinsic severity of the Delta variant and future population behaviour. When we reported the findings documented in this manuscript to government in mid-2021 to support Australia's COVID response strategy, ongoing situational assessment (as described elsewhere [11,35]) was acknowledged as critical to the success of the National Plan. That is, monitoring of local data was anticipated to allow benchmarking of the scenarios to guide real-time policy decision making on the transition to phase B of the National Plan. Likewise, a summary of the scenario modelling on vaccination and the easing of restrictions in the United Kingdom, published in February 2021, articulated the need for measures to be relaxed ‘based on data and the situation at the time, rather than at pre-determined dates' [4]. The degree of PHSMs needed for disease control prior to the vaccine threshold being reached in Australia and during the transition period would require reference to near-real-time estimates of the effective reproduction number, and forecasts of cases and clinical burden, at a sub-national level. Outcomes were anticipated to be highly situation specific—related to the actual starting number of cases, the population characteristics where transmission is concentrated (e.g. vaccine coverage, age, co-morbidities, access to health services, ability to adhere to personal protective measures, etc.), the rate of vaccination, and the level of epidemic suppression achieved.

In mid-2021 transmission of the Delta variant became established prior to the 70% coverage threshold in the jurisdictions of New South Wales, Victoria and the Australian Capital Territory [36]. Informed by the model-based analyses presented here, state governments imposed strict stay-at-home measures (corresponding to high PHSMs) before relaxing those measures (to settings corresponding to approximately medium and then low PHSMs) upon reaching the 70% and 80% coverage thresholds, respectively. Assisted by the ongoing application of PHSMs and TTIQ, and greater than 80% vaccine coverage, the initial waves of Delta infection had peaked and were in decline by November 2021, with epidemic activity stabilizing at levels manageable within health system capacity—as anticipated by the scenarios explored here.

Response plans, and the modelling work supporting them, should be adaptable to new phases of the pandemic, including the emergence of new variants and intervention options. All findings from our scenario analysis were made in the context of the Delta variant (specifically, what was known in June 2021) and the Australian healthcare system and society under conditions of low prevalence. In the latter half of 2021, the modelling framework described here was adapted to investigate new epidemic dynamics and policy needs in response to emerging information on the increased severity of Delta (relative to ancestral strains), the emergence of Omicron, and the roll-out of third vaccine doses. The Omicron variant (BA.1) was first detected in Australia in late November 2021 and rapidly became the dominant circulating variant. At the time, daily case incidence in five of Australia's eight states/territories was either zero or fewer than tens of cases per day (electronic supplementary material, figures S21 and S22). Consequently, when widespread transmission became established at various time points beyond the 70% coverage threshold in late 2021/early 2022, the first ever SARS-CoV-2 epidemics managed by these jurisdictions were dominated by Omicron. Additional scenario analyses required adjustment to consider emerging evidence of Omicron's increased intrinsic transmissibility, higher propensity for immune-evasion, and decreased clinical severity relative to the Delta variant. These analyses also incorporated a data-driven model of the relationship between neutralizing antibody levels (either infection or vaccine-induced) and protection against a range of outcomes over time since infection/vaccine administration, which exhibits waning of protection. Furthermore, our vaccine allocation scenarios were restricted to two-dose vaccination of the 16+ adult population. With approval of vaccines for those 5–16 years of age and third doses for the adult population by late 2021, additional research was required to assess the anticipated additional benefits of these programmes. The scenario modelling described here informed a stepwise and agile approach to the relaxation of COVID-19 response measures in Australia in the vaccine era.

Ethics

The study was undertaken as urgent public health action to support Australia's COVID-19 pandemic response. The study used data from the Australian Immunization Register (AIR) provided to the Australian Government Department of Health under the National Health Security Agreement for the purposes of national immunization recording. Data from the AIR were supplied after de-identification to the investigator team for the purposes of provision of epidemiological advice to government. Contractual obligations established strict data protection protocols agreed between the University of Melbourne and sub-contractors and the Australian Government Department of Health, with oversight and approval for use in supporting Australia's pandemic response and for publication provided by the data custodians represented by the Communicable Diseases Network of Australia. The ethics of the use of these data for these purposes, including publication, was agreed by the Department of Health with the Communicable Diseases Network of Australia.

Data accessibility

All code and input datasets required to perform the analyses are available at https://github.com/aus-covid-modelling/NationalCabinetModelling and https://zenodo.org/record/8191103 [37], respectively, except for the raw Australian COVID-19 vaccination data. These data were provided by the Australian Government Department of Health under a data sharing agreement and are not publicly available. However, we provide aggregated datasets containing the daily proportion of people at a national level by dose number, vaccine product and age group for each scenario.

The data are provided in electronic supplementary material [38].

Declaration of AI use

We have not used AI-assisted technologies in creating this article.

Authors' contributions

E.C.: conceptualization, formal analysis, investigation, methodology, software, writing—review and editing; C.R.W.: conceptualization, formal analysis, investigation, methodology, software, writing—review and editing; C.B.: investigation, methodology, writing—review and editing; M.J.L.: investigation, methodology, software, writing—review and editing; G.E.R.: formal analysis, investigation, methodology, writing—review and editing; T.C.: methodology, writing—review and editing; J.C.M.: methodology, writing—review and editing; N.R.: data curation, investigation, writing—review and editing; M.Y.: formal analysis, methodology, writing—review and editing; G.K.: methodology, supervision, writing—review and editing; N.G.: methodology, supervision, writing—review and editing; J.W.: conceptualization, investigation, writing—review and editing; J.M.M.: conceptualization, investigation, methodology, supervision, writing—original draft; J.M.: conceptualization, funding acquisition, investigation, methodology, supervision, writing—review and editing; N.G.: conceptualization, investigation, methodology, supervision, writing—review and editing; D.J.P.: conceptualization, investigation, methodology, supervision, visualization, writing—review and editing; F.M.S.: conceptualization, investigation, methodology, supervision, visualization, writing—original draft.

All authors gave final approval for publication and agreed to be held accountable for the work performed therein.

Conflict of interest declaration

We declare we have no competing interests.

Funding

This work was directly funded by the Australian Government Department of Health Office of Health Protection. Additional support was provided by the Australian Research Council (N.G. DECRA fellowship DE180100635) and the National Health and Medical Research Council of Australia through its Centres of Research Excellence (SPECTRUM, GNT1170960) and Investigator Grant Schemes (J.M. Principal Research Fellowship, GNT1117140; F.M.S. Emerging Leader Fellowship, 2021/GNT2010051).