Can a vaccine-led approach end NSW’s outbreak in 100 days, or at least substantially reduce morbidity and mortality?

Introduction

Australia currently has a policy goal of “zero community transmission of COVID-19”, which encourages early, sharp, lockdowns for new outbreaks as a key component of eliminating the virus. The recent roadmap approved by National Cabinet foresees a pivot to a strategy of increasingly “living with COVID-19”, particularly as 70% and 80% vaccination targets are met¹.

The current NSW outbreak, starting June 16, was not eliminated early through a short, sharp lockdown – and as of 11 August has a five-day rolling average of more than 300 daily cases and a 4% per day average increase in cases over the preceding fortnight. The NSW Government did not immediately implement aggressive containment strategies akin to ‘stage 4’-style city-wide (or even state-wide) lockdowns as used elsewhere. Instead, the strategy appeared to be a geographically varying intensity of restrictions based on caseloads - the first state-wide lockdown was implemented at 6pm on 14 August 2021. This was combined with increasing vaccination rates, while targeting younger, critical workers. 50% of adult Sydneysiders were vaccinated with their first dose on 14 August 2021, which the NSW government contends will provide strategic space to ease restrictions on 28 August, even though most people under 40 years of age will, at that time, remain unvaccinated.

To assist policymaking in NSW, and public understanding, our objectives were:

1. To estimate how long it would take to reach fewer than five cases per day, under 12 combinations of three strengths of lockdown (weak, moderate, and strong) by four vaccination roll-out scenarios: (a) as per original plan, b) prioritizing essential workers with Pfizer, c) b plus rapid vaccination of 25% of <60-year-old adults with AstraZeneca, and d) b plus rapid vaccination of 50% of <60-year-old adults with AstraZeneca.

2. To estimate the number of cases, hospitalizations, and deaths in the 100 days after 1 August 2021 for the 12 above scenarios.

We used the COVID-19 Policy Model², an agent-based model used to underpin Victoria’s exit from its second wave in 2020³ and used in peer-reviewed research on Victoria’s COVID-19 outbreak⁴. In applying this model to NSW and the latest outbreak, it was challenging to know exactly what levels of restrictions were being applied across Sydney and NSW, and how to capture all elements of heterogeneity that may matter (e.g., full heterogeneity of household size, essential workers, and social contact patterns in Western Sydney).

Thus, it is essential at the outset to emphasize that we are primarily attempting to forecast likely patterns of infection across the community, rather than predicting exact numbers of cases, hospitalisations, and deaths – although we do provide the latter by scenarios with wide uncertainty intervals. We use Monte Carlo simulations that include both stochastic (random) uncertainty and uncertainty in inputs, for example, uncertainty about vaccine efficacy.

Methods

To reflect the increased infectivity of Delta we updated the model with a shorter time from infection to being infectious⁵ and increased peak infectivity⁶. We calibrated our model’s transmissibility profile parameters to achieve R0s of 5.5 to 6.5, and then draw from within this range for each run. We doubled the infection fatality ratios and hospitalisation rates we used for 2020 modelling given accruing evidence about Delta.⁷ See Supplementary Table 1 for further details of model parameterisation.

Results

Figure 1 shows the median daily cases over the preceding week for the four vaccination strategies, for each of the three-stage settings. For our first objective, the fastest option is Stage 4 with the most aggressive vaccination strategy, taking 78 days (90% uncertainty interval 61 to 103 days) from 1 August to reach an average of five daily cases over a 14 day period (or median of 18 Oct, 95% UI 1 Oct to 12 Nov). This time increases to 207 days (168 to 250 days) for a weak lockdown with no rollout acceleration (or 24 Feb 2022, 90% UI 16 Jan to 8 Apr).

• Download figure
• Open in new tab

Figure 1:

Median daily cases (log scale; 7-day average) for each combination of Stage and vaccination strategy. Vaccination thresholds include a 10-day period for the second vaccine dose to be fully effective. Solid lines are medians across 4000 runs of the ABM, and bands show the 5^th to 95^th percentile across runs.

For the more likely or realistic scenario of Stage 3b, and essential workers vaccinated first and 25% of < 60-year-olds opting to get AstraZeneca rapidly, it is 124 days (78 to 175) until fewer than 5 cases per day is achieved (3 Dec, 90% UI 18 Oct to 23 Jan).

Superimposed on Figure 1 are solid vertical lines for when 50%, 70% and 80% vaccination coverage of the population over 16-years-old is achieved for each vaccination strategy. Except for fewer than half of the Stage 4 lockdown runs, the date at which 50% of adults are double vaccinated (plus a 10-day buffer to allow immune response) occurs before less than five cases per day is achieved.

Figure 2 below shows heatmaps of the numbers of estimated cases, hospitalisations, and deaths for each vaccination strategy by stage combination, for a timeline of 100 days. Hospitalisations and deaths are given for infections acquired within the first 100 days, avoiding the lag between infections to hospitalisations and deaths, to make for clearer comparisons. Deaths from infections caused with a two-week lag from infections, reflecting infections in days 0 to 86, are in Supplementary Table 3.

• Download figure
• Open in new tab

Figure 2:

Heatmaps of cases, in the first 100 days of each strategy (or cases are less than five per day is achieved first). Hospitalizations and death are those among infections in first 100 days, and therefore the total hospitalizations are those up to (roughly) 10 days beyond 100 days due to time lags, and deaths are two to three weeks after infections to allow for time lags. BAU: vaccine roll-out per the National Plan; INT: shifting all Pfizer to essential workers as a priority then back to BAU; AZ_25: prioritizing essential workers as in INT and vaccinating 25% of adults aged under 60 years with AstraZeneca; AZ_50: increasing to 50% the adults aged under 60 years vaccinated with AstraZeneca.

The most substantial reduction in cases, hospitalisations and deaths is generated through tightening restrictions, i.e., going vertically down the columns in the heatmaps, rather than faster vaccination strategies presented horizontally, across the rows. For example, for deaths under the BAU vaccine rollout, median deaths increase 33-fold from 79 under strong lockdown to 2570 under weak lockdown. Under a strong lockdown, median deaths only increase by 7%, from 74 under a rapid vaccine rollout (AZ50) to 79 under a BAU vaccine rollout. However, there is an interaction, in that under a weak lockdown a slower BAU vaccine rollout with 2,570 deaths is 75% greater than the 1,470 deaths with a faster AZ50 rollout.

For what might be the most likely vaccine rollout scenario, AZ25, 1,440 (90% UI 262 - 10,600) deaths occur over 100 days under a weak lockdown, 198 (50 to 654) under a moderate lockdown, and 71 (90% UI 26 - 178) under a strong lockdown scenario. The number of infections and AZ25 for a weak lockdown are 191,000 (or an average of 1910 per day [not implausible given nearly 400 cases per day on August 13], 90% uncertainty interval 42,000 to 1,150,000), for a moderate lockdown 26,400 (7,500 to 68,600) and for a strong lockdown 9,320 (or an average of 93 cases per day [seemingly implausible now, unless there is a very strong turn-around], 4,960 to 18,700).

Conclusion

NSW will likely achieve 70% vaccination of >16-year-olds before reaching fewer than five daily cases. Put another way, for most scenarios it took more than 100 days (and sometimes much longer) to ‘bend the curve’ and reach fewer than five cases per day. Accelerating the vaccine rollout is important for the medium-term in NSW (and the rest of Australia), as it will make the population more resilient to outbreaks later in 2021 and into 2022 when we open the borders. However, in the next few months, it is the strength of lockdown (or public health and social measures) that has the largest potential impact in reducing COVID-19 hospitalisations and deaths.

There were some modelling analyses performed on the Delta NSW outbreak in July 2021. The Burnet Institute⁹ analysed whether Victorian-style Stage 3 and Stage 4 restrictions would be sufficient to control the outbreak. They found it would take roughly 6 weeks to control the outbreak. Similarly, our previous modelling¹⁰ found that stage 4 restrictions would take around 5.8 weeks to get cases to five or less with longer times to exit restrictions for less stringent restrictions. However, at the time of publishing their study, restrictions in Sydney had recently been introduced and the impact was unknown. Also, since our previous modelling, we have updated the model for new data on Delta – the shift forward in time from infection to being infectious (which undermines in part the impact of contact tracing). Last, a nowcasting model¹¹ found that the level of social distancing in NSW around 25 July 2021 was inadequate to control the outbreak and 80% of agents need to comply with distancing regulations to control the outbreak in a pandemic in a month. All studies, including this one, concur that strong restrictions would be required to turn around cases and head towards re-elimination. The modelling in this paper improves on previous efforts by updating the analyses to the observed scenario in August 2021. It also incorporates emerging evidence on the twicer higher virulence of the Delta variant ⁷.

Epidemiologically, a sustained and hard lockdown for (unfortunately) several months is necessary to minimize the heath loss due to COVID-19. The ‘best’ policy options require balancing the direct health impacts of SARS-CoV-2 with the unintended health impacts of sustained and hard lockdowns, and the social and economic impacts of lockdowns.

In applying our model to NSW, we want to emphasise it is only a model. It is becoming increasingly challenging to model Delta, especially amid changing and inconsistently applied public health interventions. Nevertheless, it seems most unlikely the conclusion that strong public health and social restrictions remain more important in the short term for constraining case numbers than vaccination, would be overturned by any plausible adjustments to the model.