• AI for Good Simulator

    Modeling the spread of COVID-19 within refugee camps for timely interventions

  • About this project

    Together with the AI for Good community, we participated in HackfromHome in April 2020 and built the COVID19 Simulator for refugee camps and other high density & low resource settings. The project is led by a team of international volunteers with experience in epidemiology modelling, data science, humanitarian work, social science, user research and AI ethics.

     

    The Simulator is a web tool for NGOs and local authorities to model COVID-19 outbreak inside refugee camps and prepare timely and proportionate response measures needed to flatten the curve and reduce the number of fatalities. This tool helps to predict the possible outbreak scenarios and their potential outcomes and help NGOs design an optimal intervention strategy. Read more

  • Problem

    Refugee camps lack a strategy to combat COVID-19 because:

    1

    High density

    Moria's population density is 204K people / km2 vs. London's 4.5K people / km2

    2

    Sanitation

    1 toilette per 167 people, 1 tap per 1300 people, waiting lines are 2-3 hours

    3

    Malnutrition

    Food distributions fall short of daily nutritional requirements

    4

    Poor Health

    Not enough ICU bed, chronic and respiratory diseases, PTSD

    5

    Legal Status

    Asylum seekers are not included in national plans and basic healthcare

  • Solution

    To address this problem we use multiple epidemiological modeling approaches:

    Agent-based Model

    Based on: ABM for cholera spread in Daabab refugee camp (George Mason University)

    • Digital prototype of Moria Camp
    • 500 agents act based on rules informed by research; UNHCR and experts outline behaviour and demographics
    • Model implemented in Netlogo
    • Health inputs: Transmission probability, Incubation time, Infection and recovery probabilities

    Compartment Model

    Based on: Epidemic calculator for COVID19 (Gabriel Goh et al)

    • “Macro” method models population as a whole 
    • Model repurposed in python’s streamlit module → our own mobile dashboard 
    • Key inputs are aligned with parameters estimate from Princess Diamond cruise ship (R0, Tinc, Tinf)
  • Modeling Conclusions

    Based on our modelling, an example intervention strategy for Moria could be:

    Long-term strategy

    Invest in a prolonged programme in improving personal hygiene for the camp residents in the whole duration of the pandemic

    Reducing transmission

    Ask the residents who are infected to minimise their contacts and movement and make sure they have access to the necessary treatment (including ICU)

    Urgent action for the high-risk

    Moving elderly and other high-risk individuals with pre-existing conditions swiftly away from the camp at the very start of the epidemic

  • Impact

    Real time insights from our models will have impact in the following areas:

    Empower NGOs

    Essential supply forecasting

    Serve as a tool for NGOs and grassroots organisations to mobilise action and coordinate efforts to prepare and contain COVID-19

    Inform Policy Making

    Decision making on the ground

    Inform governments, camp management and local authorities and help them make responsible decisions in a timely manner

    Save People's Lives

    Reducing the spread and fatalities

    Identify the most effective COVID-19 interventions designed for refugee camps and help avoid severe casualties

  • We need your help!

    We are looking for partners and volunteers - please get in touch if you are interested

    All Posts
    ×