Modelling the spread of Covid-19 in the UK

June Dalziel Almeida

JUNE

Joseph Bullock, Carolina Cuesta-Lazaro, Arnau Quera-Bofarull, Miguel Icaza-Lizaola, Aidan Sedgewick, Henry Truong, Aoife Curran, Edward Elliott, Richard Hayes, James Nightingale, Tristan Caulfield, Kevin Fong, Ian Vernon, Julian Williams, Richard Bower and Frank Krauss

Outline

  • Motivation for new Covid-19 model
  • Main model components:
    • Virtual population
    • What can residents do in JUNE?
    • Modelling and mitigating infection
  • Tuning the model
  • How does JUNE differ from other models?
  • Preliminary results

The questions JUNE would like to answer

  • Understanding the geographical spread of the disease
  • What measures need to be taken to end national lockdown situations safely, and how effective are local/regional lockdowns?
  • What are the most effective measures to be taken to not overwhelm NHS capacity?

We need an accurate model of the UK's geography and demography to answer these questions!

Virtual population: a "digital twin"

  • Use 2011 Census information
  • What defines a JUNE "resident"?
    • age¬† (27)
    • sex (f)
    • ethnic group (Caribbean)
    • deprivation index 2 (1-10)
    • household composition
    • work sector / subsector
      (healthcare/doctor)
    • mode of transport (public)
    • area of residence
    • area of work

(From Nomis census service)

London

North East

June's Demography

male

female

male

female

What can residents in JUNE do?

  • Residence
    • Care Home
    • Household
  • Primary activity
    • Company
    • Hospital
    • School
    • Care Home
    • University
  • Travel
    • Commute
    • National Travel
  • Leisure
    • Shopping
    • Pubs/restaurants
    • Cinema
    • Residence visits

Modelling and mitigation infection

  • Model the spread of the infection as related to:
    • strength of interaction in a given venue
    • proportion of infected venue occupants, and how infectious they are
    • number of contacts experienced in a given venue, and length of time
  • 'Policy' interventions to reduce the spread of the infection
    • Close schools, except for children of key workers

Tuning the model

  • Tunable parameters?
    • ~10 interaction strengths
    • Infection seeding strength
    • Proportion of asymptomatic cases
    • Policy "interaction strength" reduction
    • ...
  • Very expensive model to evaluate!
  • Use emulators to estimate the behaviour model output between known data points¬†

How does JUNE differ from other models?

  • Granularity - demographic based on individual census OAs
  • Data outputs
  • Precise interventions to quantify the effectiveness of different policies

Results: matching the first peak

Results: Isolated regions do not capture the infection spread

City of London workers' usual residence

Results: prevalence trends by demography

Rates

Infection probability

Intensive care

Asypmtomatic

Mild

Severe

Exposed

Hospitalised

Dead at home

Dead in hospital

Dead in ICU

Recovered

JUNE201102

By aidansedgewick

JUNE201102

  • 273