COVID-19 is a highly infectious novel coronavirus whose spread has rapidly led to a global pandemic since the first cases were identified in China in late 2019 (Huang et al. 2020). In the absence of effective treatment or a vaccine, control of an emergent pathogen typically relies on effective case-finding (supported by testing), case isolation, and contact tracing, which is traditionally carried out manually using information provided by cases (Ferretti et al., 2020). Despite relatively advanced public health systems in many high-income countries, these methods have so far proved insufficient on their own to prevent the establishment of local epidemics, particularly without automation of contact tracing processes and where testing capacity is limited. To mitigate the impact of COVID-19 on health systems and reduce the number of deaths from the disease, many governments have resorted to measures including widespread social distancing, school and business closures, and mass quarantines (or ‘lock-downs’). However, these interventions impose heavy societal costs, including impacts on mental health, education, domestic violence and job losses; these are increased the longer lock-downs continue. Proposed exit strategies have emphasised the role of contact tracing in conjunction with substantially increased testing capacity (Ferretti et al., 2020).

Contact tracing is the process of finding individuals who are at increased risk of developing an infectious disease due to their exposure to a known case. This process is not conventionally automated, is time and resource intensive, and relies on cases’ accurate recall of information about their recent movements and interactions. Moreover, for it to have an impact case identification, diagnosis and contact tracing must all occur during the incubation period before a contact becomes infectious themselves. Because of this, contact tracing is rarely done for rapidly infectious respiratory infections, such as influenza. COVID-19 presents particular challenges for contact tracing (Hellewell et al. 2020): it is moderately transmissible, with the basic reproduction number (R₀) estimated at 2-3 (Liu et al. 2020), its symptoms can be difficult to distinguish from other respiratory tract infections, it has a relatively short incubation period of approximately 5 days (Lauer et al. 2020), and a substantial proportion of individuals have asymptomatic clinical courses or mild symptoms that go relatively unnoticed (Bai et al. 2020). Moreover, a large majority of individuals are currently susceptible, such that large outbreaks of COVID-19 can develop rapidly.

To address challenges including imperfect recall by cases and the time delays often introduced by manual contact tracing (potentially enabling more complete, accurate and timely contact identification and notification), as well as to enable rapid scaling and reduce their resource-intensiveness, a number of novel automated or semi-automated approaches to contact tracing have been developed in the wake of the COVID-19 pandemic (e.g. Wang et al. 2020; TraceTogether; Private Automated Contact Tracing [PACT]). However, there are a number of potential pitfalls to such approaches, including concerns regarding privacy, security and other ethical considerations, and the fact that high levels of uptake by populations may be required for effectiveness (Ferretti et al. 2020; Hellewell et al. 2020). A review of the evidence regarding the effectiveness and implications of attempts to automate contact tracing to date may offer valuable learning to inform potential exit strategies from mass quarantines and technology development efforts, so that risks are addressed pre-emptively within any proposed strategy.