Congo Ebola Response Turns to Mobile Data to Predict Spread

Health officials fighting the Ebola outbreak in eastern Democratic Republic of Congo have begun using anonymised mobile-phone data to map how people move across the country, in an effort to predict where the virus could appear next.

The technique, which has not previously been used in an Ebola response, is being deployed as authorities struggle to contain what the World Health Organization has called the fastest-growing outbreak of the disease on record. Government figures show more than 6,250 people infected and 3,039 dead since the outbreak was declared on May 15. Cases have been confirmed in six provinces.

This is Congo's 17th Ebola outbreak. It is caused by the Bundibugyo species of the virus, for which no vaccines or treatments have been approved.

Mapping movement to target resources

As the virus spreads, responders are looking for ways to anticipate new cases and direct limited surveillance and response resources. Olivier Le Polain, who heads epidemiology and analytics for response at the WHO Health Emergencies Programme, said population movements have always mattered in outbreaks but are especially significant in the current epidemic.

Conventional risk assessments focus on outbreak hotspots and assume the virus is most likely to move into surrounding areas. Mobility data allows responders to spot risks in places that are geographically distant but closely linked by travel patterns.

"It gives us a more nuanced understanding of risk," Le Polain said, pointing to Ituri province in the northeast, where the virus is believed to have begun circulating. Significant population movements there are tied to mining and trade, and understanding them is crucial to understanding how the outbreak spreads, he said.

The analysis is carried out by Flowminder, a Swedish non-profit, using anonymised records generated when mobile phone users connect to telecommunications networks. As people move, their phones connect to different antennas, producing records that indicate where those connections occurred. Researchers use the records to estimate travel between areas.

The data is supplied free of charge by Vodacom, Congo's largest operator. Flowminder says it receives aggregated, anonymised records and has no access to information identifying individual subscribers.

"We seek to provide estimates of how people move, which in turn is a predictor of how infectious people move," Flowminder founder Linus Bengtsson said.

Early signals and a transport hub

Flowminder's first analysis, published in early June, focused on the three areas where the outbreak is believed to have started: Bunia, Mongbwalu and Rwampara. Researchers tracked where people who had spent time in those areas between April 3 and April 23 later travelled. The largest flows were to nearby areas in Ituri and North Kivu provinces. By the end of June, all 10 destinations receiving the largest flows from the original outbreak areas had reported confirmed Ebola cases.

As the virus spread beyond its initial epicentre, the analysis was expanded to additional affected areas. One location that drew particular attention was Kisangani, a major transport hub on the Congo River linking eastern and western Congo.

"The fact that Kisangani was at risk was not a huge surprise, even if it was not bordering the affected areas," Le Polain said. By mid-June, the data showed the risk was significant and that the outbreak could spread there faster than initially thought, prompting strengthened operations before the first cases were detected, he added.

More recent studies have highlighted links between Kisangani and Kinshasa, as well as settlements along the Congo River and areas near the borders with the Central African Republic and South Sudan. Fifteen of the 50 areas most connected to Kisangani were in Kinshasa province, underscoring the potential risk of spread to Congo's capital and largest city.

The findings come as the WHO continues to warn of the risk of the outbreak reaching Kinshasa and neighbouring countries. In a modelling exercise presented in late June, the agency estimated a 70% probability of a case being detected in South Sudan by the end of September.

Limits of the data

One limitation is that the analysis relies on Congolese mobile-phone networks and therefore offers only a partial picture of movements across international borders. It also currently depends on data from a single operator in Congo, Vodacom. While the company is the country's largest operator, with more than 26 million active subscriptions and a market share of nearly 35%, the analysis cannot capture the movements of people using rival networks.

Flowminder and U.N. officials involved in the response say efforts are under way to secure similar agreements with other operators to improve coverage and obtain more detailed data.