WHO’s Collaboratory Bets on Shared Tools and Trust to Speed Epidemic Intelligence

The twelfth session of the WHO Pandemic and Epidemic Intelligence Innovation Forum, held on March 24th 2025, turned its attention to a question that has long haunted global health: how do you get analysts, software developers, national public health agencies and field epidemiologists to actually build things together, rather than in parallel silos? The session, reported in BMC Proceedings, examined the WHO Collaboratory, a digital platform and convening mechanism that now hosts more than seventeen active communities of practice spanning dengue forecasting, mpox analytics, influenza surveillance and epidemic intelligence from open sources. Through short presentations and a panel discussion, contributors from the WHO Hub for Pandemic and Epidemic Intelligence, EcoHealth Alliance, the London School of Hygiene & Tropical Medicine, the Public Health Agency of Canada and data.org laid out both the promise of this collaborative model and the considerable institutional friction that comes with it.

The Collaboratory is designed to attack two persistent gaps in the global health landscape. Operational teams in national and local public health settings often hold rich, context-specific data but lack in-house capacity for advanced analytics. Analytics communities and technology developers, conversely, bring cutting-edge methods but frequently misunderstand the day-to-day constraints faced by practitioners and decision-makers. The Collaboratory’s stated core goal is co-creation: stakeholders shape the design, implementation and adaptation of global public health tools from the outset, so that analytical innovations are not developed in isolation but refined through continuous feedback from the communities they are meant to serve. Its dengue analytics community of practice, for example, brings surveillance experts, climate scientists, modelers and public health officials together to co-develop approaches for forecasting dengue outbreaks, with the explicit aim that models be usable and meaningful for those managing health responses on the ground, not merely scientifically rigorous.

The flagship example of this integrative function presented at the session is the Global Repository of Epidemiological Parameters, or grEPI, developed through collaboration between WHO, Imperial College London, the London School of Hygiene & Tropical Medicine and the Public Health Agency of Canada. grEPI provides an open-access, centralized source for key epidemiological parameters such as reproduction numbers, incubation periods and serial intervals, the quantities that feed directly into disease modelling and risk assessment during outbreaks. What distinguishes the repository, according to the meeting report, is not just its technical architecture but the fact that it is shaped by the inputs and needs of modelers, epidemiologists, public health practitioners and policymakers across multiple geographies, an attempt to prevent the curation of parameters from becoming a top-down exercise disconnected from end users.

Sitting upstream of grEPI is an AI-enabled evidence synthesis pipeline led by the Public Health Agency of Canada in collaboration with WHO. Where grEPI curates and centralizes validated parameters, the pipeline addresses the labor-intensive problem of systematically gathering, reviewing and synthesizing the research evidence that informs those parameters. By combining artificial intelligence with human oversight, the pipeline is intended to accelerate the production of systematic reviews and meta-analyses, so that the data feeding into grEPI and other decision-making tools is timely and rigorously assessed. The pairing illustrates a theme that ran through the entire session: individual tools gain practical value when they are chained together into a workflow, in this case automated evidence review flowing directly into parameter repositories used in real-world modelling.

A third initiative, the Epiverse toolkit, co-developed by the London School of Hygiene & Tropical Medicine and other organizations through the data.org Epiverse-TRACE initiative, offers a modular suite of R packages for data cleaning, outbreak modelling and intervention assessment. It is released under open-source licensing with open development, training materials and user guides designed to make advanced analytics accessible to experts and non-experts across varied settings. Complementing this, data.org presented a semantic search and mapping tool that uses an AI-driven system to organize more than 130 epidemiological R packages into a searchable, thematic space, allowing practitioners to identify relevant packages even without deep technical knowledge and, in principle, reducing duplication of effort across the global data analysis ecosystem.

The panel discussion confronted the harder question: what keeps any of this alive after the initial funding and enthusiasm fade? Panelists argued that sustainability is not solely a matter of money but is strongly tied to building networks of communities in which no single institution becomes a critical dependency. The Epiverse initiative, they noted, deliberately connects to multiple partner institutions and draws on diverse streams of expertise and funding to avoid dependence on any one actor. A second principle was that tools must generate routine, practical value: when software demonstrably saves time, simplifies workflows or improves decisions, it is far more likely to be adopted and maintained. A third was local capacity and ownership. Rather than delivering one-size-fits-all solutions, Collaboratory initiatives are framed as sustainable when they enable local adaptation, training and co-development, whether through modular software design, human-in-the-loop AI or flexible governance models.

The role of artificial intelligence drew pointed scrutiny. Speakers acknowledged that AI can accelerate evidence synthesis, semantic search and outbreak forecasting, but insisted it must be applied with rigorous validation, contextual adaptation and human oversight. The Public Health Agency of Canada highlighted the use of minimal, purpose-driven modelling approaches wherever possible to reduce unnecessary energy consumption, aligning the initiative with environmental sustainability goals. Panelists also discussed agentic AI approaches in which a system assesses its own confidence and passes decisions to human reviewers when confidence thresholds are not met. Such human-in-the-loop designs, they argued, allow experts to review, interpret and adapt AI outputs for local use, avoiding so-called black box risks and reinforcing trust. Reliability, the panel concluded, is not just a technical property but part of a broader ecosystem of ethical governance, accountability and user engagement: AI innovations must be transparent, auditable and designed in partnership with the communities that will use them.

Integration into national public health workflows emerged as perhaps the most consequential challenge. Many national agencies face mounting external pressure from media, political actors and the public to deliver rapid analytic insights during emergencies, precisely when the evolving evidence landscape makes tools like grEPI and the AI pipeline most relevant. Panelists stressed that such tools must complement rather than disrupt existing workflows, offering flexible points of entry, clear interfaces and robust human support. In some settings, public health teams may prefer modular components or packaged outputs; in others, direct access to databases or analytic code. The stated ambition is to optimize the evidence-to-action loop, allowing agencies to move from data to insight to intervention more efficiently, including rapid access to validated parameters and updated modelling inputs in near real time during outbreak response.

But panelists were emphatic that integration is an institutional problem as much as a technical one. Agencies are more likely to adopt external innovations when they are involved early in co-creation, when tools are transparent and well-documented, and when outputs can be tailored to local policy questions. The Collaboratory projects therefore provide training, documentation and ongoing engagement so that local teams understand how to use tools, adapt them and interpret outputs in ways that fit their operational and policy contexts. This relational, capacity-focused approach, participants argued, is what separates innovation as proof of concept from innovation as embedded practice in public health systems.

The session’s forward-looking agenda was shaped in part by the 2025 Collaboratory Summit, held in Berlin days before the Innovation Forum, which brought together eighty-five participants across four days of interactive workshops on governance, culture and innovation. Both at the Summit and the Forum session, participants insisted that governance is not an administrative afterthought but a central determinant of whether collaborative innovation succeeds or falters. Good governance, in their formulation, requires clear roles, transparent processes, fair and inclusive decision-making and the flexibility to adapt as challenges evolve. The launch of an Innovation Forum community space on the Collaboratory was presented as a concrete step, offering a digital venue where members can exchange lessons, propose initiatives and co-shape the Forum’s future. As the session closed, participants converged on several priorities: strengthening co-creation at every stage of innovation, sustaining governance models that balance openness with accountability and distribute leadership across the network, and expanding capacity-building so that both well-resourced and under-resourced public health teams can engage meaningfully with tools, data and collaborative projects. The recurring message of the session was that advancing epidemic intelligence requires not only new tools but sustained collective action, trust and shared ownership across sectors, disciplines and regions.

Subject of Research: Cross-community collaboration within the WHO Collaboratory to advance pandemic and epidemic intelligence tools and practices

Article Title: Fostering cross-community collaboration to advance pandemic and epidemic intelligence

Article References: Fitzner, J., Tornimbene, B., Gruson, H., Kucharski, A., Olival, K. J., Rice, C., & Waddell, L. (2026). Fostering cross-community collaboration to advance pandemic and epidemic intelligence. BMC Proceedings, 19(S4), Article 42. https://doi.org/10.1186/s12919-026-00399-9

Image Credits: AI Generated

DOI: 10.1186/s12919-026-00399-9

Keywords: WHO, epidemic intelligence, pandemic preparedness, Collaboratory, grEPI, Epiverse, artificial intelligence, evidence synthesis, public health, co-creation, outbreak modelling, governance