The reports ‘Network modelling of elimination strategy pillars: Prepare for it, Stamp it out’; ‘Alert Level 2.5 is insufficient for suppression or elimination of Covid-19 community outbreak’; and ‘Contagion network modelling of effectiveness for a range of non-pharmaceutical interventions for Covid-19 elimination in Aotearoa New Zealand’ were sent to the New Zealand government on 9 December 2020, 15 February 2021, and 16 November 2020 respectively. The reports are presented here as a connected trilogy that are intended to be read in the order above.

Here is some background and context that might be of use to readers:

All three reports deal with the wild type variant of SARS-CoV2 that was common around the world (and in many cases in Aotearoa New Zealand) in the latter half of 2020. The transmissibility of SARS-CoV2 is commonly communicated via an R0 value. This value was around 2-3 for the wild type variant considered in this report.1 More recently a number of more transmissible variants of concern have emerged. The alpha and delta variants that were first noted in the United Kingdom and India, respectively, have R0 values that are around two and four times higher again than the wild type strain of SARS-CoV2. Hence, any control measures and results discussed in these reports will typically no longer apply with the current prevalence of these more transmissible strains.

The ‘Elimination strategy’ report considers the policy settings and possible interventions that can be used to prepare for the emergence of Covid-19 in the community in Aotearoa New Zealand. It then looks at the effectiveness of different Alert Level 3-like interventions in eliminating community transmission once it is detected. Pre-detection, the report focuses primarily on testing rates of symptomatic individuals in the community. Post-detection, a range of measures that could reduce transmission (e.g. closing workplaces, mask wearing) are considered, along with the effect of elevated community testing rates and contact tracing of exposed individuals. Many of the parameters for this report were estimated from the behaviour observed during the August 2020 outbreak in Auckland.

At the end of August 2020, at the tail end of the outbreak, Auckland dropped from Alert Level 3 (AL3) to Alert Level 2.5 (AL2.5), while there were still new cases of Covid-19 being found in the community.2 Despite this, the outbreak remained under control and was eventually eliminated. This raises the question of whether or not the less restrictive AL2.5 could have been used to eliminate a similar outbreak in future, rather than the more disruptive AL3.

The ‘Alert Level 2.5’ report addresses precisely this question. It finds that for a similar community outbreak, AL2.5 is most likely to result in suppression, but not elimination-like behaviour. The probability of elimination under AL2.5 is strongly linked to the outbreak size at initial detection. Outbreaks with ten or fewer total cases (including unknown cases) at the time of alert level elevation have an approximately 60% chance of being eliminated within 150 days of detection, while if the outbreak size is 11 or more at the point when alert levels are elevated, the probability of elimination falls to under 12%.

The fact that a period of AL2.5 did lead to elimination while community cases were still being discovered at the tail-end of the Auckland August outbreak may be explained either by good luck (i.e. a very low probability but successful elimination event) or, more likely, by the preceding period of AL3. During AL3, high levels of contact tracing and testing may have essentially ring-fenced the outbreak. In such a scenario, individuals in the vicinity of the existing cluster would effectively remain at a higher Alert Level while the rest of Auckland moved to level 2.5.

Both the ‘Elimination strategy’ and the ‘Alert Level 2.5’ reports consider a number of different variations of the AL3 and AL2.5 interventions, such as improved contact tracing processes due to increased rates of QR code scanning or adoption of Bluetooth contact tracing.

The final report of this trio, ‘Modelling of effectiveness for a range of non-pharmaceutical interventions’, looks at a large number of combinations of transmission reduction interventions at four levels of strictness (none, partial, increased, and strict) for workplace and community, and three levels (none, partial, and closure) for schools. For example, closing schools but not workplaces, while mandating mask-wearing indoors would approximately match the closure of schools, and ‘partial’ control in workplaces and communities.

The purpose of this report is not so much to design alternative Alert Levels, but rather to provide a sensitivity analysis. This analysis can be used to infer some coarse bounds on the types of behaviour that might be expected in the previous two reports if parameters were adjusted, or if assumptions were modified. Even though this report was produced in advance of its two companions, it is presented here as an add-on that may help to quantify any uncertainty or variation in the more realistic scenarios of the first two reports.


  1. We do not directly use R0 to parameterise infectivity or transmissibility in our contagion model. Rather, we characterise transmissibility with the parameter beta – the average number of infections per person per unit time, which is calibrated against the generation time. R0 is related to beta by: R0 = beta/gamma, where gamma is the rate per unit time at which individuals are removed from the infectious state(s).
  2. Covid-19: What happened in New Zealand on 31 August – Radio New Zealand

Read the reports