On May 17, 2026, the World Health Organization declared the Ebola outbreak caused by the Bundibugyo virus in the Democratic Republic of the Congo and Uganda a Public Health Emergency of International Concern, two days after the outbreak itself was first reported. Five months later, cases of Ebola have been recorded in 60 health zones across 6 provinces of the DRC.
Initially overwhelmed by the complexity and the scale of the epidemic, the response system has slowly been strengthening. For example, the proportion of contacts of confirmed cases followed up within 24 hours of their identification, a key indicator measuring the ability of the outbreak response system to get ahead of the epidemic rather than catching up on it, has climbed from 21% in the outbreak’s first week to over 82% by late August, according to reports from the Ministry of Health.
Key to being able to tackle the epidemic is the strengthening of the data system being used to track cases, understand the epidemic dynamics, orient response activities and predict the future of the outbreak. Collecting, managing, and analyzing the required data to accomplish this is a key challenge, especially in a constrained environment like conflict-affected eastern DRC.
In the last 5 months, Bluesquare has been growing its work to support the Incident Management System for the Ebola outbreak, developing solutions to improve the way data is collected, its quality checked and managed, and to bring cutting edge data analysis tools including AI capabilities to the frontline of the fight against Ebola.
Currently, there are no approved vaccines or treatments for the Bundibugyo Ebola strain. There are also reports that the clinical symptoms of the virus are milder than for other Ebola virus strains. Limited symptoms and no clear therapeutic options, in conjunction with reticence associated with Ebola Treatment Centers (ETC) in affected communities, mean sick people have little incentive to visit health centers, and are identified late, if at all, by the health system.
The control strategy for the outbreak is affected by this situation. The outbreak response system risks always running behind new infections, identifying them late. The need for effective contact tracing and case follow-up is then essential to control the epidemic. In a context affected by accessibility issues and violence, case investigation and home visits for contact tracing are exceptionally hard and rely on the courage and commitment of all involved health workers and partners.
To make contact tracing as efficient as possible, good data are key. Alerts for potential new cases have to be of good quality so that they can be reached in the community and investigated by response teams. Possible contacts need to be identified, reached, tested, and the results of tests need to be shared back. New cases need to be taken care of, while symptom development among contacts will be tracked. At every step, data is collected. This data can take many shapes or forms, from names and symptoms written on paper at the end of a mud road, to the daily number of alerts shared on WhatsApp over faulty phone networks, to testing results obtained in a laboratory. Ensuring all of these data are obtained, compiled and used is key to our work.
A key resource to collect and transmit data from the frontline of the epidemic is the DHIS2 tracker that the DSNIS and COUSP have put in place to support the outbreak response. But encoding data directly into DHIS2, while streamlined and simplified, requires a level of literacy to navigate and enter complex data that are not always present in remote health centers. Case and contact information is then registered on paper forms at the health center, then physically transported to the health zone office, where data managers digitize tens to hundreds of forms by hand. And even in health facilities where data entry is possible, connectivity, even facilitated by Starlink terminals, is not always possible.
Every step in that chain adds delay and data missingness, giving the response system a delayed and partial image of the outbreak, even when it can reach it. Even with powerful data analysis systems like the one Bluesquare has been developing with the DSE and COUSP existing data infrastructure in Ituri this patchy data is a fundamental bottleneck for outbreak intelligence and needs to be tackled.
To improve the quality and availability of primary data, Bluesquare is working in two directions.
Advanced matching to link treatment centre data back into DHIS2: Completeness of care and treatment data in DHIS2 has been a challenge since the start of the outbreak. DHIS2 rollout in Ebola Treatment Centres (ETCs) has been slower than in laboratories, but individual-level data has still been collected. Bluesquare is working on implementing OpenHEXA pipelines to extract, match and push this individual data back into DHIS2. The goal is a more complete official data collection tool for the Incident Management Team in Bunia.
AI agents on WhatsApp to improve DHIS2 Tracker promptness and identification: Early digitalisation is key to outbreak management. In the DRC, the notification form goes from the health centre to the health zone, the patient goes to the ETC, and the sample goes to the laboratory. Without a unique identifier, matching the form, the patient and the sample is difficult.
With 60 health zones now affected, and data stemming from fragile collection systems, making sense of the growing list of patients (line list) gets harder by the week. As the number of alerts rises, the Ministry of Health has to work harder to maintain and verify the quality of a database that keeps expanding, and maintaining the quality of the line list is essential to ensuring a clear understanding of the outbreak.
With multiple teams working simultaneously on the quality assurance of complex data streams, it has proven difficult for the SGI to enforce standardized data audit processes across data managers. The tools available have been mainly applying Excel filters to identify issues, with no standardized method or approach. Enforcing a standardized data audit routine by developing a dedicated tool seemed like a difficult task for an emergency management team, and using readily available tools seemed easier.
An AI-assisted development process within the OpenHEXA data integration platform of the COUSP allowed the rapid build of a web application tailored to the exact needs of surveillance teams working in incident management. The tool gives these teams fast visualization of the line lists entered in DHIS2, with automated and verified computation of key indicators, including confirmed cases and confirmed deaths, alongside flags for data quality issues such as inconsistent notification dates or incorrect final diagnostic classifications. It ensures every data manager applies standardized filters for data screening and uses the same tool and visuals.
The app was developed with very quick turnaround time (less than a week from specification to validation), ensuring the solutions we bring are very close to the needs of the field users. Additionally, the app is hosted on the SGI’s own resources, making it transparent, and easy to evolve if new specifications emerge.
In a context of emergency, this AI-assisted, locally auditable and managed development is in the truest sense a game changer. Teams involved in emergency management, working alongside accessible developer resources, can rapidly create data management tools, iterate, test and abandon ideas with no cost. This makes it much more acceptable to ask for some time from these teams to ask for specifications, as we will be able to provide solutions to their demands in time frames that actually make sense for them.
We think that building and embedding small development capability empowered by contextually relevant AI agents in emergency response teams is a clear path forward in the future.
A fast-spreading outbreak, limited knowledge of the specific virus strain, and the absence of population immunity or vaccine coverage all make it essential for outbreak response teams to have a clear and rich analytical toolbox at their disposal to orient their strategies and actions. We are currently working to develop this capacity in three directions.
The first need we need to answer is to provide key figures on the outbreaks. These key figures are what define what everybody should know about the outbreak. Key, standardized, reproducible indicators that can be shared with partners, funders, the media and the wider public to communicate on the outbreak in a trustable way.
Bluesquare has supported the automation of the Situation Report of the SGI to provide such figures, as well as the development of standardized, internal dashboards. These completely deterministic data products are essential for the coordination and strategy definition of the outbreak.
For operational needs, the SGI often needs specific information and results to be communicated in a clear way. Weekly surveillance meetings tend to start with key epidemiological updates but are followed by findings specific to the situation: eg., how is Ebola spreading in communities since the back-to-school period in September?
These analyses do not only rely on preexisting standardized indicators, but require innovative, complex analysis to pinpoint exactly the information needed in this moment of time. AI agents within OpenHEXA have been optimized to generate dashboards through simple prompting, while using current operational definitions and conventions developed by the data management and analysis teams of the SGI. This makes it possible for teams to get a fast, comprehensive view of where the outbreak is spreading, which populations are affected, and where risk is concentrated, and to answer rapidly emerging questions that may arise during the epidemic, even when the data comes in unruly formats as is typically the case during emergency situations.
The more advanced analytical needs are complex questions requiring key epidemiological insights: what are the most affected populations and key risk factors? What are the most probable outbreak spreading routes in the coming days? Is there an impact of the outbreak on the mortality associated with other diseases?
For this, Bluesquare collaborates with key academic teams to develop a direct connection between field data, additional data and cutting edge epidemiological thinking. Under funding from CEPI, Bluesquare is part of the B-React consortium led by the DRC INRB and INSP, and bringing together the Kinshasa School of Public Health, University of Oxford, Imperial College, the Pasteur Institute and Flowminder to develop a platform facilitating complex epidemiological analysis, to facilitate the BDBV vaccine development. This is the top tier of the platform.
Across these three stages, from a paper form at a health center to a dashboard used by the Ministry of Health, the throughline is the same: the Public Health Emergency Operations Center (COUSP) and its epidemiologists remain in full operational control of the response, and Bluesquare’s role is to keep the underlying data usable at each step, testing where AI genuinely removes friction rather than adding a new point of failure.
To learn more about how Bluesquare supports epidemic surveillance more broadly, visit the One Health Surveillance page.
