Modelling testing and response strategies for COVID-19 outbreaks in remote Australian Aboriginal communities

The SARS-CoV-2 pandemic continues to cause significant morbidity and mortality worldwide, disproportionately affecting vulnerable and disadvantaged groups such as those of lower socio-economic status, or with comorbidities [1]. Protecting such groups must be a priority. As of mid-2021, Australia remains in a favourable position compared with much of the world and although several outbreaks have led to regional lockdowns, COVID-19 case numbers since January 2020 have totalled only around 30 450, with 910 deaths [2]. No cases of community transmission have yet occurred in remote Australian Aboriginal and Torres Strait Islander communities.

Aboriginal and Torres Strait Islander peoples (hereafter respectfully referred to as ‘Aboriginal’), however, are significantly more vulnerable to severe COVID-19 than the non-Aboriginal population due to a high prevalence of comorbidities that are associated with more severe clinical outcomes [3]. The incidence of chronic respiratory diseases is 1.2 times higher than for non-Aboriginal Australians, type 2 diabetes 3.3 times higher, and chronic kidney disease 3.7 times higher [4]. SARS-CoV-2 transmission is likely to be even more intense within remote communities due to crowded housing, larger family sizes, inadequate hygiene facilities, and residence across multiple dwellings (4–7). These communities are also further from specialist health services, with SARS-CoV-2 tests needing to be transported, thereby resulting in delays to diagnosis and treatment. Previous influenza outbreaks in these communities have underscored their vulnerability. During the 2009 H1N1 pandemic, hospital and ICU admissions for Aboriginal people were 12 and 5 times higher, respectively, than for non-Aboriginal Australians [5]. Similarly, First Nations Americans have suffered from COVID-19 rates 3.5 times that of white Americans, and mortality rates that are almost double [6]. The consequences of overcrowding and disadvantage have also been demonstrated in Singapore, where migrant workers in overcrowded dormitories suffered from infection rates of up to 20% [7].

In Australia, protection of remote Aboriginal communities was prioritised early, with strict movement controls, within designated biosecurity zones, established in consultation with communities [8]. The Aboriginal and Torres Strait Islander Advisory Group on COVID-19 (IAG), co-chaired by the Department of Health and the National Aboriginal Community Controlled Health Organisation, provides evidence-based and culturally safe guidance for COVID-19 preparedness and response to the government and other key stakeholders, with a view to locally led adoption of recommendations within each community [9]. This group liaises with peak national health advisory bodies on COVID-19 and commissioned the work that we present here to help inform optimal public health response strategies in remote settings.

This study presents a novel exploration of COVID-19 control interventions in remote Aboriginal communities in Australia, which are vulnerable to COVID-19 due to the underlying comorbidities, and with infection expected to transmit quickly due to overcrowding and dynamic household structure that extends beyond single dwellings. The model output was also used to shape the COVID-19 outbreak response policy for these communities. A report outlining the key results of this work and recommendations is publicly available from the Australian Government Department of Health [10].

We compare plausibly implementable non-pharmaceutical-based strategies in a remote Aboriginal community, examining the impact of alternative scenarios in an outbreak response, including: initial delays with testing; differing definitions of case-contacts and consequent quarantine strategies; community-wide lockdowns; and exit testing strategies. We assume a wholly susceptible, unvaccinated population as the COVID-19 vaccine uptake rate in remote communities has so far been low [11].

A participatory approach was employed throughout this study. All SARS-CoV-2 outbreak response scenarios explored were designed through iterative engagement between the academic investigators, the IAG, and other public health end-users to ensure cultural sensitivity, and to maximise the relevance and uptake of findings.

An individual-based model, repurposed from a framework developed to examine dynamics of sexually transmitted infections in remote Australia, is used to explicitly represent each community member [12]. Community sizes comprising 100, 500, 1000 or 3500 people are modelled, with results presented here focusing on communities of 1000 people but noting key differences. The full detail of the model is provided in the Additional file 1: Appendix S1, with the key features and assumptions for this analysis highlighted in the following sections.

Prompt case finding is essential to prevent a SARS-CoV-2 outbreak in a remote Aboriginal community. A high transmission propensity, due to interconnected and often crowded households, means that in an unmitigated scenario the majority of the community would be rapidly infected. By the time early cases are identified, active infections in the community may be up to ten-fold higher. This is especially apparent for smaller communities due to the high degree of interconnectedness of individuals, as compared to larger communities where the number of households is larger and distinct but separate clusters of households are more likely to form. We assume only half of all infected patients will self-present to health services for testing, due to absent, unrecognised or minimal symptoms, fear, or stigma. This may be an overestimate, but there is evidence that pre-symptomatic transmission may contribute > 40% of SARS-CoV-2 transmission exists [15, 23]. This non-presenting proportion may not be detected using a passive case finding approach, although a high prevalence of other co-morbidities may result in non-COVID related presentations resulting in opportunistic case detection. Higher non-presenting proportions would lead to poorer mitigation in all scenarios, while interventions will have greater impact if this proportion is higher than assumed here (see Additional file 1: Appendix S1).

Of the contact tracing strategies assessed, quarantining extended household members (residents of all dwellings used by the case) is the most effective strategy for constraining the initial outbreak, reducing peak prevalence from 60 to 70% to ~ 10% (Fig. 3) for a community of 1000 individuals. However, large numbers of people must be quarantined for extended periods and outbreaks are predicted to resurge when community mixing resumes, with overall community attack rates exceeding 80% (Table 3). Clearance testing modestly reduces this attack rate to 65%. Lockdown of all non-quarantined households for 14 days, concurrent with this quarantine strategy, results in the greatest likelihood of definitive outbreak control. Peak prevalence of the initial outbreak is less than 5%, and the overall attack rate less than 10%. Clearance testing from lockdown further improves control, preventing subsequent waves of infection following the release of individuals with undetected infection (Fig. 4): overall infections are constrained to < 5% with clearance testing, versus > 80% without. In contrast, without clearance testing and/or additional changes to response strategies, subsequent waves of infection are highly likely following the release of large numbers of susceptible individuals from quarantine and/or lockdown restrictions resulting in recrudescence of infection among susceptible individuals. Clearance testing from the lockdown strategy also results in fewer tests due to prompt suppression, fewer person-days in quarantine, and is effective in mitigating outbreaks with delays of up to 6 days (Fig. 5). Larger communities benefit most from lockdown, with the effect dampened in smaller communities (100–500) by the large proportion already in quarantine. Compliance with lockdown must be at least 80–90%, or epidemic control will be lost. While this is an ambitious target given the lockdown compliance rate has been observed to be closer to 50–60% in Australian cities [24], our consultation with stakeholders suggested that a higher compliance rate is possible for remote communities due to smaller population size and physical isolation. Furthermore, with COVID-19 vaccination now available in remote communities of Australia, it is likely that a lower compliance rate in lockdown will be required to control outbreaks.

Our findings are consistent with recent guidelines for a ‘contain and test’ strategy developed by Central Australian health organisations [8], which acknowledge that symptom-based case identification will be insufficient, and endorse active case finding and lockdown with multiple rounds of voluntary testing. Analyses of SARS-CoV-2 outbreaks overseas also support the effectiveness of lockdowns. In the Italian town of Vo, researchers concluded that a 14-day lockdown reduced transmissibility of infections (including asymptomatic) by 82–98% [25]. Lockdowns in Wuhan contributed to a significant decrease in spread [26], and an analysis of French data suggested that over 80% of potential COVID-19 deaths were averted by their lockdowns [27]. Recent modelling from the UK, examining the impact of delays with testing and contact tracing, suggests that if cumulative delays exceed 3 days for these processes, control of an outbreak is unlikely [28].

The participatory process employed between this study’s investigators, the IAG, and other public health end-users throughout, has allowed for direct feedback of our findings and incorporation into IAG guidelines [9], and collaborative development of plain-language messaging for health providers and community members. Our findings support the suggestion that prompt case finding and a rapid public health response upon first diagnosis will be critical for effective control [29], which in our context, can be facilitated by having access to decentralised point-of-care testing (e.g., GeneXpert). Local planning and preparation should occur in advance, and must involve community members to ensure cultural appropriateness, local support and community control. Early patient presentation should be encouraged, and testing, contact tracing and isolation/quarantine guidelines and facilities clarified. The extensive public health response required to achieve best outcomes necessitates prior preparedness planning to ensure that the significant logistical and human resources support needed can be rapidly mobilised. Throughout an outbreak, community trust must be preserved in order to maximise compliance; in particular, the historical context and consequent sensitivities regarding enforced lockdowns in remote Aboriginal communities must be kept foremost in mind in the design and implementation of such strategies.

In developing our model, simplifying assumptions due to limited observational data regarding population structure and mixing were necessary. Other than household structure, ‘real-world’ mixing opportunities such as schools and workplaces have not been explicitly included. Assumptions regarding transmission dynamics are derived from non-Aboriginal populations, but where possible we have erred on the side of caution. For example, the high R₀ to which the model is calibrated is based on early estimates from Wuhan and amplified to reflect the propensity for intense transmission in remote households. However, as shown in the additional analyses described in Sect. 7.4 of the Additional file 1: Appendix S1, majority of our findings still apply for smaller R₀ of 2 and 3. We also note that at the time of writing, the Delta variant of SARS-CoV-2, which is believed to have an effective reproduction number close to double that of the previous variants [30], has reached Australia and has sparked outbreaks in several jurisdictions resulting in strict lockdowns in Sydney, Melbourne and Adelaide [31]. It is therefore entirely possible that our assumed R₀ of 5 for remote communities is close to or even underestimates the real R₀ should the Delta variant reach these communities. We also assume perfect sensitivity and specificity of testing throughout the infectious period. Morbidity and mortality outcomes have not been estimated in this model, or the anticipated demand on health resources (testing requirements aside).

We assume that cases in isolation and contacts in quarantine will have no contact with others (i.e., will not transmit SARS-CoV-2). This may not be possible to achieve for every remote communities, but this assumption was based on stakeholder input from the Northern Territory, Australia, where suitable facilities have been made available for use to many communities by the resources sector. By representing this ideal we assess the maximum effectiveness of these measures and demonstrate the added value of lockdown. We also assume a wholly susceptible, unvaccinated population, as less than 5% of the total Australian population was fully vaccinated as of mid-2021[32], with this proportion likely to be lower in remote communities [11]. However, the expansion of Australia’s vaccine rollout, with prioritisation of remote areas and at-risk populations, will increase the proportion of the population that is vaccinated over time.

Remote Australian Aboriginal and Torres Strait Islander communities have the potential to be severely impacted by COVID-19, due to factors favouring increased transmission and disease severity. Our modelling affirms the need for early case detection as multiple secondary infections are likely already present by the time an index case is identified. Quarantining of extended household contacts, together with 14-day community-wide lockdown with clearance testing, are the most effective strategies in limiting the outbreak.