How data and AI help detect and fight Ebola outbreaks

Etien Luc Koua, WHO Africa’s Program Area Manager for Health Emergency Intelligence, Surveillance and Laboratory, explains how epidemic intelligence has reduced outbreak detection time from 7 days to 2, while forecasting models support the verification of clusters of unexplained deaths and rumors of illness.

What digital technologies are used in the current Ebola response in Africa?

Digital tools now support every stage of the response. At the detection level, the Epidemic Intelligence from Open Sources (EIOS) system scans news, social media, and, in some countries, local radio for early signals of unusual health events. It is now used in 44 of the 47 countries of the WHO African Region.

For the ongoing outbreak in the Democratic Republic of Congo (DRC), EIOS is used daily to detect new signals in affected and unaffected provinces to enhance alert management. Routine Integrated Disease Surveillance and Response (IDSR) data through the DHIS2 (District Health Information Software 2) platform and Early Warning, Alert and Response System (EWARS) also track new signals.

Like other countries in the region, DRC also reports weekly on priority diseases through the centralized regional integrated disease surveillance platform built on DHIS2, used by 43 countries in the region, and suspected events are tracked in a regional signals and events management system. Another support is an alert management platform: a centralized, mobile-phone-based system that stores, triages, and manages alerts from all sources, including the community.

At the diagnosis stage, laboratory results flow into a shared case database, so national and regional teams can see new confirmed cases the same day. Laboratory network expansion and new testing platforms, including RadiOne and standard RT-PCR using prioritized Altona kits, mobile testing capabilities, and improved access to diagnostics, help accelerate case detection and reporting.

For containment, case investigations, contact listing, and follow-up are recorded on tablets and phones in DHIS2 Tracker; dashboards, case mapping, and tools for tracking treatment centers, partners, and border crossings; and outbreak analytics tools track cases, deaths, and response indicators across affected areas.

Smartphones seem to be the most important tool to manage the outbreak. How do healthcare professionals use them to contain the epidemic?

Frontline teams use smartphones and tablets to register alerts, investigate suspected cases, and follow up contacts in the DHIS2 Tracker, which works offline and synchronizes when a network is available.

GPS coordinates captured during investigations and alerts let us map cases and alerts daily, see where new health zones are being affected, and position treatment centers, burial teams, and supplies. Mobile networks also provide free hotlines (151 in the DRC, 719 in Kenya) and messaging services so the public can report sick people or deaths and ask questions. Points of entry screen several hundred thousand travelers a day in the DRC, and the DRC is expanding the digitalization of screening data.

And how is AI being used?

Yes, in specific, practical ways, with experts always reviewing the outputs. The production of situation reports (SITREPs) has now been automated using AI. This has largely reduced the time needed for daily data consolidation and compilation, data verification, and producing daily reports.

AI/Machine-learning in EIOS sorts through hundreds of thousands of articles and flags the few that point to real signals. Speech-to-text tools transcribe local radio in several languages so that rumors and reports of illness in communities without internet coverage can be picked up.

Our team has integrated an ensemble of six forecasting models, developed in partnership, and produces weekly projections of cases and deaths, estimating how fast the virus is spreading in each health zone. Geospatial AI combines surveillance data with satellite and climate data to map higher-risk areas. These tools support decisions; they do not replace epidemiologists or field teams.

WHO Africa has also introduced the AI-enabled Preparedness Data Exchange.

The Preparedness Data Exchange brings together readiness assessments, mobility, border, and surveillance data to rank where the virus is most likely to spread next. In the current outbreak, its cross-border risk scoring identified districts in the Central African Republic along the eastern border with the DRC, notably Bangassou and Obo, as among the highest-risk places for undetected introduction: porous forest and river crossings, very limited detection, and insecurity.

As a result, surveillance and laboratory support were prioritized there: mobile laboratories were deployed to the Bangassou and Obo districts so suspected cases can be tested locally rather than waiting for samples to travel to the capital. The same risk ranking now guides which neighboring countries receive readiness support first.

What can epidemic-intelligence systems detect today that traditional surveillance would miss or detect too late?

Epidemic intelligence systems such as EIOS complement traditional, indicator-based surveillance with signals from online publicly available sources. Traditional surveillance depends on health facilities recognizing and reporting cases, which can take days or weeks, especially where people avoid health facilities or die at home. Epidemic intelligence adds signals from outside the health system: news reports, social media, local radio, community workers, and calls to hotlines. EIOS can pick up and verify clusters of unexplained deaths, rumors of illness in a village, or unusual reports from a border area before the first laboratory-confirmed case.

In 2025, for about half of public health events in the WHO AFRO region, EIOS was the first source of information to WHO AFRO before receiving official notification from Member States. Most of the documented signals verified by WHO in the African Region came through EIOS rather than routine reporting, and the median time from detection of a signal to its verification was one day. In countries using EIOS, the median time to detect an outbreak fell from about seven days in 2023 to two days in 2025, and about 76% of all public health events were detected within 7 days in 2025.

That’s impressive. How much faster can Ebola now be diagnosed? What solutions can shorten the time between testing and results?

During the 2014–2016 West Africa epidemic, many samples had to travel long distances to a few reference laboratories, and results often took several days. In the current outbreak, the DRC has around 25 functional laboratories, including laboratories close to affected areas, with a combined capacity of several thousand tests a day.

Once a suspected case is notified, the median time to a laboratory result is about one day, and most results come back the same or the next day after sampling. In Kenya, results are returned within 6 to 12 hours, including from mobile laboratories on the border with Uganda. What makes the difference: decentralized PCR testing, near-point-of-care platforms, mobile laboratories, faster sample transport, and electronic transmission of results directly into the case database.

Are genomic sequencing and epidemiological data now being combined in near real time to reconstruct chains of transmission?

Increasingly, yes, although not yet everywhere in real time. Sequencing capacity now exists in almost every country in the African Region, and hundreds of genomes have been sequenced in this outbreak. Sequencing confirmed that the virus in Kenya’s imported case is closely related to strains circulating in the DRC.

Epidemiological data, such as contact links and symptom-onset dates, are being used to reconstruct chains and estimate how quickly the disease spreads from one person to the next. The next step is to routinely combine genomic and epidemiological data so that unexplained cases can be linked to known chains within days, and that work is under way.

Digital contact tracing was a major technological advance during previous Ebola outbreaks. Does it work?

It works, but only as well as the people and systems around it. Digital tools make it possible to list tens of thousands of contacts, follow them daily, and see in real time who has been missed. In the DRC, around eight in ten contacts under follow-up are now seen every day, up from about four in ten in June, and in some provinces more than 90%.

The limits are not mainly technological: insecurity that prevents teams from reaching people; contacts who hide because 21 days of follow-up means lost income; poor connectivity; and the difficulty of linking contact records to confirmed cases.  Where those barriers are lower, contact tracing finds a much larger share of new cases among people already being followed. Kenya’s second case, for example, was detected while already in quarantine as a contact.

Can digital systems identify misinformation or rumors likely to delay treatment, drive vaccine refusal or resistance?

Yes. Social listening, EIOS media monitoring, radio monitoring, hotline calls, and community feedback collected by community workers are analyzed together to spot rumors early. Examples from this outbreak include beliefs that treatment centers are places where people go to die, and claims that hand-washing water or thermometers at checkpoints carry the virus.

Can authorities intervene before such rumors spread widely?

When a rumor is identified, teams respond quickly through trusted local voices, community dialogue, radio programs, and targeted messages, and most community feedback is now addressed within 48 hours. The technology helps us hear concerns sooner; changing minds still depends on trust and on people seeing that care works. Africa now has far more digital surveillance, AI, sequencing and rapid diagnostics than in 2014–16, yet the outbreak still spreads faster than authorities can contain it.

What is technology still failing to solve?

Technology helps us see the outbreak faster. However, it cannot, on its own, make people seek care early, reach people in insecure areas, or replace trust. In the DRC, about seven in ten deaths still occur in the community, before patients reach treatment, and many cases are only identified after death. That is driven by fear of treatment centers, distance, cost, insecurity, and community resistance to safe burials, not by a lack of data.

There are also gaps that technology has not yet closed. For Bundibugyo virus, there is no licensed vaccine or specific treatment; infection-prevention-control remains limited in some areas; and data systems are not yet fully connected, so cases, contacts, laboratory results, and patient outcomes are not always linked as we would like. But these are challenges being addressed.

What is the biggest challenge regarding the current outbreak?

The biggest challenge is reaching people early enough, detecting cases, and getting patients into care before they die at home and before they infect their families and carers. Insecurity in parts of Nord-Kivu and Ituri, community mistrust, and health-worker infections make that harder.

Technologies that would make the biggest difference are – already mentioned before – a licensed vaccine and effective treatments for Bundibugyo virus; validated rapid tests that can be used at the point of care in remote areas; fully interoperable digital systems that link each case to their contacts, laboratory results and outcome; and offline-first tools for teams working in areas without connectivity.

Are other surveillance systems or solutions planned or in development to strengthen preparedness and response?

Yes, the priority is to connect national disease surveillance reporting with epidemic intelligence so that signals and suspected cases are seen in one place; expanding AI tools such as automated signal classification and radio monitoring; scaling up collaborative surveillance with countries, including the DRC; wastewater and environmental surveillance in pilot countries including the DRC, Senegal, Uganda and Zambia; and digitalizing screening at points of entry.

WHO is also building an AI infrastructure for modeling and surveillance and a regional modeling network with African institutions at its Dakar hub, a center of excellence for public health intelligence and data innovation, so that countries can share data securely, run their own models and anticipate outbreaks rather than react to them.