Nowcasting and Forecasting of the COVID-19 Pandemic
Real-time tracking of a pandemic, as data accumulate over time, is an essential component of a public health response to a new outbreak. A team of statistical modellers at the Medical Research Council Biostatistics Unit regularly nowcast and forecast COVID-19 infections and deaths. This information feeds directly to SAGE sub-group, Scientific Pandemic Influenza sub-group on Modelling (SPI-M) and to regional PHE teams.
The work uses a statistical programming model called a transmission model (Birrell et al. 2020), data on daily COVID-19 confirmed deaths from PHE (by NHS region and age group), and published information on the risk of dying and the time from infection to death, to reconstruct the number of new COVID-19 infections over time; estimate a measure of ongoing transmission (R); and predict the number of new COVID-19 deaths in different regions and age groups.
For more information and to view the latest published reports, go to the Nowcasting and Forecasting page
Nowcasting COVID-19 Deaths
Every day the UK Government publishes figures for the numbers of deaths within 28 days of a positive coronavirus test: https://coronavirus.data.gov.uk/. These numbers come from Public Health England (PHE). Both the number of deaths reported on each day and the number of deaths that occurred on each day are presented. The notification of a COVID-19 death to PHE is subject to a reporting delay, which is typically several days, but can be as long as several weeks. Consequently, the number of deaths that occurred on each day is incomplete, particularly for the most recent days. The process of estimating the number of deaths that occurred on each day from the numbers of deaths that have so far been reported is referred to as `nowcasting’. For details of our approach to such a nowcast see Seaman S et al. 2020.
Severity of COVID-19
For the COVID-19 pandemic, estimates of the infection- and case severity risks, i.e. the probabilities of experiencing severe events such as hospitalisations, admission to intensive care and death, are crucial to understand and predict the burden and impact on healthcare services.
No single dataset can provide enough information on its own to estimate severity, but estimation is feasible by synthesising multiple datasets. Work at the BSU is investigating use of both individual- and aggregate-level data, from clinical cohorts and population registries, and a combination of survival analysis techniques, to estimate severity as data accumulates over the course of the epidemic.
- Kirwan et al (pre-print, currently under revision): https://arxiv.org/abs/2103.04867
- Nyberg et al (pre-print, accepted): https://arxiv.org/abs/2104.05560
- Grosso, Presanis et al (pre-print, under review): https://assets.researchsquare.com/files/rs-288193/v1/3c97a17f-199c-4a11-9a88-c96a81d67418.pdf
- Presanis, Kunzmann et al (pre-print, under revision): https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3792155
Analysis of COVID-19 transmission in hospital
We have produced an app to detect potential transmission events from patient and HCW data. The new software tool, called A2B-Covid, will help doctors identify where cases of COVID-19 were caused by transmission within a hospital, helping them to prevent further spread of the disease.
A2B-Covid will be available for free to doctors and clinicians across the UK and worldwide.
Details of the package have been published in a pre-print article in MedRxiv: https://www.medrxiv.org/content/10.1101/2020.10.26.20219642v1.