Read on Twitter
I'm glad to announce that our Covid-19 epidemiology simulation, called JUNE, is finally out in medRxiv!
https://www.medrxiv.org/content/10.1101/2020.12.15.20248246v2.full.pdf
The whole code (100% pure Python) is openly available under an open source license in https://github.com/IDAS-Durham/JUNE Thread
https://www.medrxiv.org/content/10.1101/2020.12.15.20248246v2.full.pdf
The whole code (100% pure Python) is openly available under an open source license in https://github.com/IDAS-Durham/JUNE Thread
June is named after June Almeida ( https://en.wikipedia.org/wiki/June_Almeida) the Scottish virologist who first discovered the coronavirus family of viruses.
We build a very detailed digital twin of England from the census data, including multiple locations at the almost exact geographical coordinates: companies, schools, pubs, care homes, restaurants, grocery stores, universities, hospitals, train stations, etc.
The geographical resolution of the code is the census output area (~300 people)
The demography is also very detailed, including people’s age, sex, ethnicity, socio-economical status, work location and work sector. In the picture you can distinguish the age-sex distribution of Durham Uni college area (big area in green), and the one where HM prison is.
People’s residences are included in full detail, with all kind of household compositions (single parents, multi-generational families, communal housing, etc.) and care homes. The age and sex differences between the different members of families are also matched.
Schools are divided between primary, secondary, and mixed, and each school is further divided into year groups and classes. Companies are also initialised with a sector and a size, and are matched against workers’ location and sector to recreate the commute flow data.
On top of this England's digital twin, June operates an Agent Based Model, simulating the movement and interactions of almost 60 million people across time.
On a typical June simulation day, people start their morning by commuting to work from multiple locations and routes into the main English cities.
They then proceed to their primary activity: kids go to school, adults go to work and unis, and retired people go to the pub or visit friends. After an exhausting day, June’s virtual population has a few hours of leisure time, which can be spent in multiple locations.
The probability of getting infected in an interaction depends on who is interacting, where, and at what time. Each location in June (like a school) has different interacting patterns between groups (like students and teachers, or visitors and residents in a care home).
Once a person is infected, they follow a particular symptoms trajectory, which depends on their age, sex, and residence type. If they require hospitalisation, they will be moved to closest NHS trust based on their location, and can then be transferred to the ICU if needed.
To prevent the infection spread, June includes more than 20 different policies, including quarantine, closure of companies (by sector and type of worker) and social venues, shielding, reduction of visits to care homes, and more. Each policy is fully customizable.
June is able to reproduce the hospital, care home, and home deaths during the first wave period at a regional level without needing any region-specific parameters.
We are currently working on understanding how the disease has spreaded in the second wave
