Limitations

There are a number of limitations to our current approach, mostly due to limitations in the fidelity and detail of the publicly available data.

  • The major limitation is that our method is based upon analysis of counts of positive tests, as opposed to true infections. There is an unclear relationship between true incidences and the community testing in Pillars 1+2, and there can be substantial selection bias in the population who (successfully) get tested. As such the predictions of our method cannot be thought of as reflecting true incidences of Covid-19 in the community, but rather as predictions of positive test results.
  • Testing capacity and regimes in the UK has changed and continues to change over time. This is particularly salient in the difference between Pillar 1 and Pillar 2 tests, and in the widening application of Pillar 2 tests, initially in healthcare workers, then to key workers, and now to the wider community. In September 2020, bottlenecks in the testing system has meant that test positivity has been increasing, indicating that under-testing is increasing.
  • Further, delays between infections and test results are unknown, different across Pillars 1 and 2, and across time.
  • NHS Test and Trace and outbreak investigations mean that testing is now more targeted, and, from the perspective of understanding rates of transmissions from case counts, can introduce biases and noise.
  • Uncertainties in $R_t$ in local authorities are high and sensitive to random local events.
  • In future it would be useful to incorporate data on daily death counts as well as seroprevalence and symptom surveys.

Making the following data publicly accessible will help the epidemic monitoring research we and others are carrying out:

  • Daily case count data from local authorities in Northern Ireland.
  • Likewise, data at the level of individual local authorities in Scotland.
  • Separating out Pillars 1 and 2 data.
  • Number of tests carried out per local authority.
  • Number of individuals tested and whether they are tested regularly.
  • Some information about circumstances under which tests were taken, e.g. whether these are part of outbreak investigations, whether the individual has symptoms.
  • Daily data segregated by age and/or at MSOA spatial resolution.