Computational Approach For Epidemic Response.
The increasing trend for international travel and population mobility have led to increased global interconnectivity and along with that global pandemics. The recent Ebola outbreak – the largest recorded in history – illustrated how these features along with a community unprepared for a sudden outbreak can turn a situation into a crisis, one that costs thousands of lives.
UNICEF Innovation is using technology to tackle this threat. In recent years, the availability of detailed data on human behaviour combined with environmental measurements have led to great advances in computational modeling of these epidemics.
The work we do.
UNICEF is developing an open-source platform to collect, combine, analyze, and display real-time information based on contributions from academic, private sector and open source data. Data sources include, among others: high-resolution population estimates, air travel, regional mobility estimated using CDRs, temperature data, and case data from WHO reports, such as Zika, Dengue and Ebola.
This data is used as input for epidemiological models running on the platform in real-time and generating forecast scenarios about the spatio-temporal spread of a given disease so that UNICEF field offices and other stakeholders can take informed and data-driven decisions about prevention and containment.
A closer look.
Data and infrastructure
We are collecting and integrating data from different partners and open sources, creating pipelines that allow for automatic ingestion of data as soon as it becomes available. Data sources include among others:
- Number of people travelling by plane on weekly basis by Amadeus
- Temperature for certain regions on weekly basis
- Population estimates at high resolution
- Estimates of Aedes Aegypti mosquito prevalence (the vector responsible for the transmission of tropical diseases like Yellow Fever and Zika)
- Zika cases, from WHO reports
Building tools for epidemic modeling, monitoring, and validation
By combining those data sources, we are building a monitoring tool that can run and validate different epidemic models on pseudo real-time data (weekly basis). Our research aims to answer the following questions. Where is the disease under investigation more likely to spread next? Therefore, where should the response efforts be mostly focused to prevent the spread?
Current models are probabilistic and use as input the number of new reported cases at a certain week (currently Zika cases), the number of air travelers from/to each country, and other factors such as population, population density or Aedes Aegypti mosquito prevalence to provide a relative measure of risk of new cases at a country level for the following weeks. Same models can be applied at a subnational level for countries with enough air travel traffic.
A practical example: Risk of Ebola spread in DRC and CAR
Influenced by academic research on the spread of infectious diseases, and on models of human mobility in absence of ground truth data, UNICEF Innovation developed a computational model to simulate the spread of Ebola in DRC and CAR in May 2017. Following the simulation with cases as of May 29th (reported by WHO) the epidemic was predicted to remain confined to Bas-Uele up to 12 weeks later, as later confirmed by the authorities. Further computer simulations were performed to investigate the probability of spread to the Central African Republic and analyze potential worst case scenarios.
Find out more about:
Email us via: eomodei [at] unicef [dot] org