A paper, published in The Lancet Public Health, has looked at the effectiveness of several different COVID-19 contact tracing strategies and the impact of speed of testing on their effectiveness.
Prof Paul Hunter, Professor in Medicine, UEA, said:
“The paper by Kretzschmar and colleagues is a useful contribution to the debate about how countries can ease restrictions whilst maintaining public safety by keeping the R value for COVID-19 less than one. The model that they describe seems appropriate for giving insights into the questions set. Their conclusions that for a test track and trace system to work well it has to be able to identify a high proportion of new cases soon after exposure are supported by their analyses and consistent with common sense. However, the paper makes several assumptions that may or may not be valid and that would impact on the outputs of the model. For example:
“1. “Symptomatic and asymptomatic cases were assumed to be equally infectious” is a big assumption that could have a big impact. The paper does not distinguish between the truly asymptomatic and the pre-symptomatic individual. We know that viral load may be roughly similar in people who are symptomatic and those not symptomatic but this does not necessarily mean that they are equally infectious.
“2. As far as I can tell the model assumes that individuals are equally infectious on each day of the illness which is almost certainly false, partly because culturable viral load declines rapidly after symptom onset and partly because exposure behaviour may change as people become ill/increasingly unwell, though there was account taken of the later point in the paper.
“3. “We assumed that 80% of all infected people develop symptoms at some time during their infectious period” This is another big assumption. We know from the Diamond Princess that the estimated asymptomatic proportion was 17.9% (95% credible interval (CrI): 15.5–20.2%). However, the average age of the population on the Diamond Princess was over 60 years. Given that much of the transmission associated with COVID-19 is now in Europe is now in much younger age groups it is not certain whether and 80% symptomatic rate is still a valid assumption.
“In conclusion, this remains an important paper and the conclusions that it is vital to minimise testing delays in order for test, track and trace to have a beneficial impact are almost certainly sound. However we should be careful not to assume that the mathematical associations reported are necessarily valid in the real world. As George E. P. Box, the late British Statistician is often quoted as saying ‘essentially, all models are wrong but some are useful’.”
Dr David Bonsall, senior scientist and clinician at Oxford University’s Nuffield Department of Medicine (and epidemiological advisor on the NHSx app programme & test and trace programme), said:
“We support the findings of this Lancet Public Health manuscript demonstrating the critical importance of speed in the effectiveness of contact tracing, both app-based contact tracing and manual contact tracing. The findings are both consistent with orthogonal modelling approaches we took in our study published in Science (Ferretti et al.), and in our agent-based simulations commissioned for the NHSx digital contact-tracing development programme.
“Contact tracing efforts need to be integrated with other important public health measures, including social distancing, shielding of vulnerable groups, and face masks. If combined with rapid contact tracing they are capable of lowering the R value below 1, preventing and containing local outbreaks. However, contact tracing must notify people before they infect others; high rates of pre-symptomatic transmission requires that the total delay, from symptom onset to contact-tracing, does not exceed 48 hours.
“In our simulations, twice as many infections were prevented by a digital system that notified contacts instantly, when people requested a test, compared to a system that delayed tracing by 48 hours after testing. We agree that manual tracing efforts also should aim to return test results within 24 hours, and trace people rapidly thereafter.
“The UK faces a specific problem: the majority of community testing relies on the postal service to send out and collect testing kits from peoples homes, this might take 3-5 days from request to result. This testing strategy will not be fast enough to stop many onward infections, and may fail to prevent second waves.”
Mr Charles Radclyffe, Industrial Fellow in the School of Computer Science, Electrical and Electronic Engineering and Engineering Maths at the University of Bristol, said:
“If fighting COVID is likened to fighting a war, then face-masks are the infantry, and social-distancing is the cavalry. In this analogy, contract-tracing might be likened to the security services. If it does its role effectively, then many lives on the battle-field will be saved, as intelligence about where to deploy troops will be more targeted. If it is inept or relies on false intelligence then vital resources will be wasted.
“Fighting COVID requires a complete arsenal of weaponry, and we need to use all at our disposal and ensure that there are no weaknesses in the chain. Currently, a critical weakness is the public’s trust in technology, technology firms, and the government’s ability to execute a digital strategy. Given the prior success of initiatives such as GOV.uk, some of this mistrust is misplaced – but still it needs to be recognised and addressed. Only a multi-disciplinary approach to solving the trust problems behind contract tracing will be effective, and unless they are – we have a key gap in our offensive against this pandemic.
“What is missing in this study is an objective measure of trust and how this impacts the assumptions behind the model. What we need is a study on this question, and urgently.”
Dr Daniel Lawson, Lecturer in Statistical Science, School of Mathematics, University of Bristol, said:
“This study explores some harsh mathematical truths about contact tracing. Simply put, if the tracing system is not fast enough then it becomes useless, as onward transmission has already occurred when contacts are isolated.
“The numbers in the study are heavily dependent on assumptions, and should not be over-interpreted. It uses generic estimates for the reproduction number, overly simplifies the transmission network, and does not address population variation. However, the key message is robust: it is a mathematical truth that contact tracing must be fast to be effective.
“Whether trace time is key to bring R below 1 in the UK is dependent on many other factors that are not considered here – for example, what was R before contact tracing? How soon do people seek testing? How many people share a household? How compliant are people? Yet the UK should heed the importance of response time, which emphasises the potential benefit of a properly implemented digital contact tracing system.”
‘Impact of delays on effectiveness of contact tracing strategies for COVID-19: a modelling study’ by Mirjam E Kretzschmar et al. was published in the Lancet Public Health on Thursday 16 July 2020.
All our previous output on this subject can be seen at this weblink: