About the model
The model is a hybrid model with a transmission model feeding its output to a healthcare demand model. The transmission model is a dynamic infectious disease transmission model that takes into account the age-group profile of local populations, social mixing patterns between people of different ages as well as transport data and interactions between locations to represent how the epidemic is likely to spread across the region. The model outputs are then used to explore the probable spread of SARS-CoV-2 and the demands on services caring for people with COVID-19.
The technical classification of the model is a deterministic SEIR (Susceptible, Exposed, Infectious, Recovered) metapopulation model. The population is divided up into LAD geographies and 5-year age/sex groups, using the 2011 census to estimate the number of people in each age/sex group by LAD. The structure of contacts between LAD residents that may lead to SARS-CoV-2 transmission is based on previously measured social mixing information (the POLYMOD survey data) – this is a quantitative, evidence-based way of considering how contacts between people may transmit infectious diseases spread by breathing or close contact. To emulate the effect of school closures, the information from POLYMOD was modified to exclude contact between school-age children outside of the household after 23/03/20. Transmission of the virus between LADs is assumed to follow commuting patterns, derived from 2011 census commuting data, which are down-scaled each day during the epidemic period, using transport data, relative to February, which showed a marked decline leading up to and following 23/03/20.
The healthcare demand model is a discrete event simulation that simulates the individual trajectories of the infected, using output from the transmission model (number of new symptomatic infections per day, by age group and location). The model calculates the following: –
Modelling parameters and assumptions
A summary of the model parameters is given below: –
Table 1: Lancaster-Liverpool COVID19 Demand Model main parameters and assumptions
|Latent period (time from being infected to being infectious)||4 days||Assumed|
|The hospitalisation rate (% of infections needing hospital admission)||8% (varying by age from <50 – 2%; >80 – 44%)||PHE Joint Modelling Cell|
|Intensive Care Unit [ICU] rate (% of hospitalisations needing a CCU bed)||25 % (varying by age – peak rate at age 60 reflecting current age distribution of COVID19 patients in ICU)||25% admitted to ICU from Covid-19 NHS dashboard applying the age distribution from ICNARC|
|Onset to attendance at hospital Emergency Department [ED]||Gamma distribution (shape=2, rate=3.4), with median 4 days||Estimated from COCIN UK patient data|
|ED to Admission||Same Day||Assumed|
|Admission to ICU||Poisson distribution, mean 1 day||Assumed|
|Admission (general) to discharge||Uniform distribution, 4-8 days||Zhou 2020|
|ICU to discharge||Uniform distribution, 10-14 days||Guan 2020|
Table 2: Items in the output file (values are the medians of 200 simulations – the file does not yet include measures of uncertainty, future versions will)
|ED_0.5||Number of Attendances at A&E|
|bed_norm_0.5||Number of Normal Beds occupied (levels 1-3)|
|bed_icu_0.5||Number of ICU Beds occupied (levels 3)|
|discharge_norm_0.5||Discharges from normal beds|
|discharge_icu_0.5||Discharges from ICU beds|
|death_norm_0.5||Deaths from normal beds|
|death_icu_0.5||Deaths from ICU beds|
|inf||Daily new infections|
Comparing Lancaster-Liverpool COVID-19 models, with the output from applying the national Reasonable Worst Case Scenario model to the local population, the latter gives 18% lower number of infections, 35% lower number of people needing hospitalisation or intensive care, 36% lower peak demand on intensive care beds.
The current model does not fully consider the impact on transmission of recent control measures. It should therefore be understood and interpreted as a worst-case scenario projection. There is considerable uncertainty around the estimates, particularly for smaller geographic areas.
Find out more
For more information on the Lancaster-Liverpool COVID-19 Model, including the associated code please see https://github.com/cipha-uk
Alex Alexiou, Matt Ashton, Ben Barr, Iain Buchan, Martin O’Flaherty, Chris Jewell, Rachel Joynes, Chris Kypridemos, Roberta Piroddi, Jonathan Read and Sally Sheard on behalf of MRF Health Intelligence Cell.