Structural barriers hinder knowledge co-production during public health crises

During a public-health emergency, the most valuable evidence is often the evidence that can move quickly from laboratories and hospitals into government decisions. Yet a new qualitative study of South Korea’s COVID-19 research response suggests that the central obstacle is not simply poor communication between scientists and policymakers. Instead, the researchers argue that the systems designed to connect evidence with action can prevent meaningful collaboration before discussion even begins.

The study, by Ji Eun Park and Myounghee Kim, examines why “knowledge co-production”—the collaborative creation of knowledge by researchers, policymakers and practitioners—so often fails in practice. Co-production has become a prominent strategy for narrowing the research–policy gap because it promises to make research more relevant, usable and responsive to real-world needs. Rather than producing findings in isolation and attempting to translate them later, researchers and decision-makers are expected to define problems, interpret evidence and shape solutions together. The authors’ findings indicate that this ideal is undermined by institutional structures that determine whose knowledge counts, which questions receive funding and how decisions are made under pressure.

Park and Kim interviewed 15 researchers who had direct experience with South Korea’s COVID-19 research response. Ten participants worked at universities, while five were researchers at government-funded institutes. The interviews were semi-structured and in-depth, allowing participants to describe how evidence was generated, exchanged and used during the crisis. The researchers analyzed the transcripts using framework analysis, following Spencer’s five-stage approach. This method involves systematically familiarizing investigators with the data, developing a thematic framework, indexing relevant passages, charting the material and interpreting patterns across cases. The analysis was guided by the Knowledge-to-Action model, which divides knowledge translation into four connected stages: knowledge production, knowledge exchange, knowledge utilization and conversion back into knowledge production.

Across these stages, the researchers identified five structural mechanisms that reinforced one another. The first was what they describe as epistemological incommensurability, meaning that different groups approached evidence according to fundamentally different assumptions about what qualifies as reliable knowledge. Epidemiologists may prioritize statistical associations, randomized controlled trials or systematic reviews, while policymakers may need rapid estimates, operational intelligence and judgments about feasibility. Community organizations and frontline practitioners may contribute experiential knowledge about unequal access, public trust or unintended consequences. These forms of evidence are not automatically interchangeable. Without deliberate “boundary work”—the effort required to connect distinct knowledge systems—participants can speak about the same emergency while using incompatible standards of proof.

A second barrier arose from institutional fragmentation and misaligned incentives. Universities, government agencies, hospitals and publicly funded research institutes operate on different timelines and reward different outputs. Academic researchers may be evaluated through publications, grants and disciplinary recognition, whereas policymakers are judged by whether programs are implemented and crises are contained. A scientist who spends months attending policy meetings, revising questions with practitioners or building relationships with communities may perform essential co-production work without receiving equivalent professional credit. Meanwhile, agencies facing urgent decisions may not have the time or resources to sustain prolonged collaboration. The result is a system in which the people expected to work together are often rewarded for doing something else.

The study also found that biomedical ways of knowing dominated the response, systematically narrowing which questions were treated as urgent or legitimate. Biomedical evidence is indispensable during an infectious-disease outbreak, but a virus does not spread through biology alone. Transmission is shaped by housing, employment, transportation, education, poverty, disability, discrimination and public confidence. Participants described a research environment in which uncomfortable knowledge—particularly evidence that might challenge existing policies, expose inequities or complicate a technically attractive intervention—could be excluded or deprioritized. When advisory structures are heavily weighted toward biomedical expertise, social, behavioral and community perspectives may be invited symbolically but remain unable to influence decisions substantively.

A related mechanism was bureaucratic expert dependency and path-dependent decision-making. In a fast-moving crisis, governments often rely on a small circle of established experts because familiar advisers can be consulted quickly and their credentials are already recognized. This creates a feedback loop: officials repeatedly turn to the same institutions, those institutions become even more central, and alternative perspectives have fewer opportunities to enter the process. Path dependence means that earlier choices constrain later ones, even when circumstances change. A committee assembled around clinical and epidemiological expertise may continue to use the same interpretive framework after social disruption, economic hardship or public resistance becomes central to the emergency. Speed, in this setting, can preserve old decision pathways rather than encourage adaptive collaboration.

The fifth barrier was the absence of equity and power redistribution. Co-production is sometimes presented as a neutral partnership, but participants do not enter such partnerships with equal authority. Senior officials, prominent scientists and well-funded institutions generally control access to data, meeting rooms, research budgets and policy channels. Practitioners and members of affected communities may possess essential knowledge yet lack the status to define the problem or challenge dominant interpretations. The authors argue that collaboration cannot be considered genuinely co-produced if less powerful participants are merely consulted after the major decisions have already been made. Equity therefore requires more than adding diverse voices to a panel; it requires changing who controls agendas, resources and the interpretation of evidence.

Together, these mechanisms formed what the authors characterize as a self-reinforcing cycle. Fragmented institutions encourage narrow disciplinary work. Narrow disciplinary work strengthens biomedical dominance. Biomedical dominance concentrates authority among established experts. Concentrated authority preserves bureaucratic routines and limits the redistribution of power. Those routines then make it harder to build the relationships, incentives and infrastructure needed for broader knowledge production. The cycle can make co-production appear to fail because participants did not communicate effectively, when the deeper problem is that the system never supplied the conditions for meaningful exchange. In this interpretation, a workshop, consultation or one-off advisory meeting cannot compensate for structural exclusion.

The proposed remedy is therefore institutional rather than rhetorical. Park and Kim call for research funding systems that mandate interdisciplinary participation instead of leaving collaboration to individual goodwill. They recommend reforming advisory committee membership and deliberative processes so that non-biomedical expertise can shape decisions rather than simply provide background commentary. They also argue for permanent knowledge-brokering infrastructure. Knowledge brokers are people or organizations that help translate between research, policy and practice by clarifying concepts, identifying usable evidence, coordinating relationships and maintaining continuity across projects. Permanent infrastructure would be less vulnerable to the short time horizons of emergency grants and could preserve institutional memory between crises. The study does not suggest that co-production will eliminate disagreement or make decisions straightforward. Its more consequential claim is that disagreement becomes productive only when institutions recognize multiple forms of expertise, align incentives with collaboration and give affected groups genuine influence. Without those changes, evidence-informed policymaking may continue to treat knowledge translation as a final communication step, even though the most important barriers occur much earlier—when research priorities are set, advisory networks are formed and authority is distributed.