COVID-19 Model Release Notes
The COVID-19 dataset has been prepared by on a “best efforts” basis and is being made freely available as our contribution to the analytical effort that is going on across the UK. The analysis and data provided here comes with no guarantee as to its accuracy and we accept no liability whatsoever for its use by others.
The data is provided under the terms of the: Creative Commons Attribution-NonCommercial 4.0 International Licence.
We are making available a COVID-19 dataset that provides estimates of risk factors and COVID-19 infection rates at a range of local geographies (Ward, Parliamentary Constituency, Local Authority and Clinical Commission Group). The plan is to update the data weekly to enable the tracking of infection rates at a local level over time across the UK.
We have used existing and new datasets to create 20 risk measures that it considers to be relevant to COVID-19. As part of this, we have applied our disaggregation method where appropriate to estimate cumulative infection rates on an “as is” basis and “timeline adjusted” basis across the whole of the UK.
There are some important points to note about the use of this data
The background and details of our analysis and methods used to obtain our results are detailed in the documentation available from our website. Anyone considering using our data should spend time familiarising themselves with the documentation before using the data.
All source materials and attribution notices are included in the download.
Version 10 updated: 10/06/2020
The main changes since version 9.0 are:
- The Infection rate estimates have been calibrated to Antibody testing results for England in the general population. In this approach, we convert the reported cumulative confirmed cases to a cumulative infection rate that is in-line with published antibody test results. An adjustment is made to account for the reported lag between a positive swab test (pillar 1) and the generation of sufficient antibodies for a positive antibody test (circa 2 weeks). Calibration between nations within the UK has been based on death counts using England infection rate to death ratio data as the benchmark. This approach accounts for the different levels of swab testing across UK nations.
- Changes to the disaggregation modelling approach now mean that for all geographies including lower level and intermediary, the cumulative infection rates increase monotonically over time. This has been achieved by modelling changes in infection rates week to week and reconciling the aggregated cumulative values to the estimated cumulative values at reporting level (e.g. LTLA in England). In previous versions, we modelled cumulative infection rate estimates for each week that could result in “ups and downs” in the timeline progression, particularly at lower geographies (e.g. wards).
- Changes to the RARID calculation have been made to deal with low confirmed case counts. Where there is insufficient data to make an estimate RARID is now set to value zero.
Version 11 updated: 14/06/2020
Version 12 updated: 21/06/2020
- An adjustment has been made to the Scottish confirmed cases time-series used as an input to our models. This is to account for a step-change in reported cases as at 15th June when all retrospective UK tests were added on that date. The excess cases are deducted and apportioned across all prior dates on a pro-rata basis to correct for this.
- Changes have been made to the “FutureCasesToCurrentCasesRatio” calculation to ensure the estimates are consistent with the year to date “As Is” values.
Version 13 updated: 28/06/2020
- An adjustment has been made to the Northern Ireland confirmed cases time-series used as an input to our models. This is to account for a step change in reported cases as at 24th June when retrospective tests were added on that date. The excess cases are deducted and apportioned across all prior dates on a pro-rata basis to correct for this.
- A new “CovidToAllCauseDeathRatio” value has been calculated using death data for England and Wales between March and May, published by ONS at MSOA level