ELLIS against Covid-19

Quantifying SARS-CoV-2 transmission suggests epidemic control with digital contact tracing

ellis 15 April 2020 - 15 April 2020
no image

15 April 2020 • 14:55 - 15:10

(University of Oxford)


Questions & Answers

Link to the recording of the live Questions & Discussion session for this talk. 


  • Q: Are you aware of any publicly available dataset, which would provide information on behaviour of people (such as frequency of use of public transport, outside exercise, shopping in-store/internet) and whether that particular person was diagnosed with COVID or not? Ideally, this dataset would cover individuals, but we would be interested also in an aggregated data by country or region (schools open/max size of public gatherings...) and then a ratio between tested and infected. The intended use of this data is for learning associations/contrastive explanations/recommendations.

    • A: Some EU countries have questionnaires, but not publicly available yet.

  • Q: Given low infectiousness of asymptomatic cases, and that children are mostly asymptomatic, could you estimate the effect of re-opening schools?

    • A: Not clear yet - our simulations suggest a strong dependence on parameters that are not well characterised at present.

  • Q: How do you come to the assumption that the asymptomatic cases would have low infectiousness?

    • A: Best current direct evidence is in https://doi.org/10.1101/2020.03.24.20042606.

  • Q: What is the evidence for the low infection of the asymptomatic? Does this means school closure has a low effect?

    • A: See for example https://doi.org/10.1101/2020.03.24.20042606. Not clear yet about schools, but yes, it's a possibility.

    Q: Do pre-symptomatic cases mean that the symptoms will be severe? Or are future mild cases included in this?

    • A: Mild as well. (Note that mild cases include mild pneumonia as well, not only fever/cough.)

    Q: Do we know if South Korea / Singapore that did extensive test + contact trace achieve the level of speed/coverage the model suggests?

    • A: To my understanding, Singapore didn't. Not sure about South Korea (that has a more complex strategy).

  • Q: Do you think that online learning of the contact tracing algorithm would be feasible/helpful?

    • A: Definitely. But planned implementations are strongly privacy-preserving so it's not clear if it is feasible within their framework.

  • Q: How was the app for tracing implemented in South Korea (in terms of technology and privacy)?

    • A: AFAIK geolocalisation + little privacy for infected cases. It's really a very different model, although possibly an effective one.

  • Q: Is there at all the possibility to experiment with different strategies in order to better estimate the parameters of your model, given the ethical implications? Or is parameter updating limited to historic data?

    • A: Parameters must be tunable, because our understanding will improve. But not clear if it's possible to experiment with strategies in vivo, or just infer parameters.

  • Q: It seems that digital contact tracing is an scalable/popular option, specially in the developing world. Do you think other type of massive testing (image based, rapid) could become a better alternative than tracing?

    • A: No. For developing countries, the idea of using contact tracing + symptom tracking, both by app (together with old-school contact tracing) sounds like the most reasonable one.


no image
(University of Oxford)