Model aggregating for epidemics

People provide information about their health using technology

Catalan-based company Mitiga Solutions uses supercomputers to provide early warnings to governments and business about natural hazards. Its recent foray into providing such information about infectious diseases has come at a time when the world needs it most, and it is now using PRACE resources to refine its methods for use on a global scale.

In 2018, a group of employees from the Barcelona Supercomputing Centre founded a spin-off company, Mitiga Solutions, aiming to harness the power of supercomputers to provide early warnings about natural hazards. Originally providing information about volcanic activity to the aviation industry, Mitiga has since expanded to cover sandstorms, wildfires, and – more recently – infectious diseases.

Alejandro Marti, CEO and co-founder of Mitiga, explains the company’s goals. “Our primary objective in all of our work is facilitating risk transfer,” he says. “Our deterministic models help first responders and businesses make better decisions by providing them with data about the risks they may face ahead of time.”

Alejandro Marti, CEO and co-founder of Mitiga

Alejandro Marti

When Mitiga began developing their epidemic early warning system, Epi-EWS, they had no idea that the worst pandemic for 100 years was looming just around the corner. Now partnered with the Barcelona Supercomputing Centre, various national health institutes worldwide, and a startup, Pebble Analytics, they are aiming to roll out their technology in time for a second wave of the pandemic.

Mitiga’s strategy for providing early warnings about epidemics consists of three phases. The first, known as digital participatory surveillance, developed almost ten years ago as part of a European project on influenza. It involves gathering information from people about how they feel and symptoms they may be experiencing, which is then cross-referenced with information from hospitals and doctors on the ground.

This data is then inputted into agent-based models, a class of computational models that can be used to simulate the actions and interactions of individuals. Data for these models comes from a number of sources. Governments can provide some information on the economic activity of their people. Mobile phone networks are also a valuable resource, providing detailed information on people’s movements. However, in some areas such as sub-Saharan Africa, this kind of data can be confounded by shared ownership of mobile phones. In these cases, gamification can be used, asking representative selections of the population to play simple games where they provide details about what they do and where they go in a typical week.

The Mitiga team working with the Red Cross to incentivise people to provide information about their health using technology.

The Mitiga team is working with the Red Cross to incentivise people to provide information about their health using technology.

The final part of the strategy adds a layer of business intelligence that helps to reduce the financial impact of any disease outbreak, providing information about how supply and demand, transport, financial markets and other institutions might be affected. “Our mission is to stop epidemics from turning into pandemics while reducing negative social and financial impacts,” says Dr Marti. “We’re never going to be able to avoid infectious diseases from arising in certain places, but what we can do is detect these outbreaks two to four weeks early and stop them from spreading and causing widespread disruption.”

Despite the success of this approach in the past, issues arise when applying agent-based models to different populations, as Dr Marti explains: “Human behaviour is very important when looking at infectious diseases – people do not behave the same in sub-Saharan African countries as they do in European countries, in the same way that older people do not behave the same as younger people, and men do not behave the same as women. These factors substantially alter the probability of encounters, patterns of exposure, and the likelihood of disease propagation.” The current lack of understanding of behavioural differences between people in terms of country, gender and age is a significant issue when trying to predict outbreaks of infectious disease. To get around this problem, Dr Marti and his team are creating an agent-based modelling platform that compares the different models and benchmarks them for different users, ensuring that they use the most appropriate model for any given situation.

The proposed model aggregation platform will enable Mitiga to compare various agent-based models and see which ones work better in different situations and environments. Ultimately, this will help to build a global platform that uses parts of all of these models. “To carry out this work, we need to do stress modelling with each of them to see which are best-suited for predicting the spread of COVID-19. This is a computationally intense activity and is the reason we applied for computing hours with PRACE.”

The Mitiga team

The Mitiga team.

The secret to success with these models is using large numbers of individuals and including as much behavioural information as possible. Alongside more general behaviour, specific situational factors such as whether lockdown restrictions have been introduced and whether people are interacting indoors are outdoors all have to be taken into account. Culture also plays a huge role in these models, as Dr Marti explains: “Although we have similar regulations across Europe, we see that people in Spain have closer physical interactions than people in countries such as Germany. This tacit social interpretation of what confinement means is important when looking at the spread of infection.”

The nature of agent-based models – simulating large numbers of individuals, each with different behaviours under different scenarios – mean that they require huge computational resources. That, along with the fact that the simulations have to be run many times over in order to create accurate ensemble forecasts, means that multicore processing in parallel is needed to get answers within a reasonable timeframe.

A new consortium started by the company is now bringing together a number of African countries in the fight against the pandemic. For the first time ever, these countries will be collecting, sharing, analysing and acting upon this kind of data in a uniform way. With 10 countries already engaging in the consortium and a digital participatory surveillance pilot programme already underway in South Africa, Mitiga is now ready to input the results of the PRACE project into their work to create a more accurate picture of the spread of COVID-19 in the continent.

Looking beyond the project, Mitiga is working on promoting a programme in Africa with the Red Cross that aims to incentivise people to provide information about their health. “This concept, known as a community inclusion currency, is based on blockchain technology and provides people with tokens that can be exchanged for certain good such as food, transport or education when they give information about their health status,” says Dr Marti. “An economy is thus generated that helps people and also gives us vital information that can help us stop the spread of disease.”

This article was also published in PRACE Digest 2020.

More information:

Eesources awarded:
5 million core hours on the Beskow hosted by KTH-PDC, Sweden


For all questions about PRACE Communications, promotional and press materials, social media, and publications, email: or phone us on +32 2 613 09 28.

Share This

Share on twitter
Share on linkedin
Share on facebook
Share on email


For all questions about PRACE Communications, promotional and press materials, social media, and publications, email: or phone us on +32 2 613 09 28.

Latest News

Recent Tweets

Let's Stay Connected

Stay up to date with all the latest PRACE news and HPC access call announcements by subscribing to our newsletter: