Dengue is an infection most frequently spread through the bite of mosquitoes. It is not new, but it is growing.
Since the 1960s, there has been a 30‐fold increase in global dengue incidence. Currently, almost half of the world's population live in areas with a risk of dengue. This peaked in 2023, which had the highest number of dengue cases to date, with over 6.5 million cases and more than 7,300 dengue-related deaths reported.
These numbers are increasing, driven by climate change and the mosquito being highly sensitive to climate variability.
Warmer climate and changing rainfall patterns create more regions with suitable habitats for the mosquito, thereby increasing the transmission potential to new areas. The introduction of dengue to previously immunologically naive areas has been the main driving force for transmission, leading to epidemic spikes in the past couple of years.
We must do something to prevent exponential increases in dengue prevalence.
Whilst there are many prevention strategies for dengue, it is unclear which are the most effective and there is real risk of policy makers, local leaders, and people living in dengue-endemic areas investing in ineffective interventions.
Filling in the gaps
At the start of 2025, we embarked on a gap maps project that aimed to do three things.
Firstly, we wanted Cochrane to produce its very first gap map introducing a new evidence synthesis product to our repertoire. Gap maps often encompass a large evidence base, much bigger than a systematic review would take on. They help researchers, guideline developers, and funders to identify where research exists and where more research is needed.
Did we achieve this? Almost, we hope so. The first version of the map with an accompanying manuscript will be submitted imminently and published this year.
Secondly, we wanted to pilot artificial intelligence (AI) in this work to help make this map a ‘living’ map. There are two main areas where we feel some level of automation is possible: identification of the potentially relevant studies, and the extraction of information about those studies. Gap maps are a synthesis of the presence of evidence, but they do not analyze what that evidence concludes. So for each included study, the data extraction and analysis required is lighter.
Did we achieve this? In this first iteration of the gap map, we’ve used humans to identify the eligible studies. This involved creating an exciting new task on Cochrane Crowd called Dengue Detect. Over 100 people helped to assess thousands of studies. The core team then categorized and coded the studies for the gap map .All this manual work has created gold-standard data sets which we can now use to help build, train and validate some AI capability to help semi-automate future iterations of this map. This has potential to be a living gap map updated with minimal human effort. If successful, it sets a precedent for future living gap maps in areas of clinical development.
Finally, we wanted to work with stakeholders who have a lived experience of dengue in designing our map by prioritizing the interventions and outcomes we included, as well as interpreting our map helping us identify the gaps that needed filling most urgently.
Did we achieve this? Absolutely!
Early stakeholder engagement – a game-changer
Stakeholder engagement is key to making evidence gap maps truly useful. An evidence gap map is meant to support decisions, not just to organize research. For that to happen, it must reflect real needs, everyday challenges, and local realities. The experience of developing an evidence gap map on dengue prevention, learning from stakeholders in Brazil, shows clearly why involving policy-makers, health and environmental workers, researchers, civil society representatives, and citizens throughout the process is so important.
Through a partnership with Instituto Veredas and the Brazilian Coalition for Evidence, we carried out surveys and interviews, allowing participants to identify which dengue prevention actions and outcomes mattered most to them. They also suggested new topics that were not initially included. As a result, community education, environmental actions, vaccines, behavior change, and reductions in dengue cases emerged as key priorities – areas closely linked to what people see and do in their daily work and communities.
In a later stage, stakeholders interacted with the first version of the map and shared feedback. They found the map easy to use but suggested practical improvements, such as filters by country or income level. More importantly, they helped identify which evidence gaps should be addressed first. These included gaps related to equity, environmental racism, possible harms of interventions, long-term effects, and the impact of environmental actions on dengue outcomes.
An equity lens to evidence gap maps
Stakeholder engagement also helped capture important local knowledge that is often missing from research papers. Participants explained how dengue affects urban areas more strongly, especially neighborhoods facing poverty and poor sanitation. They highlighted the higher risks faced by people living in settlements, traditional communities, and peripheral areas, as well as workers exposed to mosquitoes in forests and near water. Stakeholders also pointed to concrete barriers to prevention, such as lack of sanitation, weak waste collection, shortage of supplies, limited staff and transport, and low community awareness.
These insights helped ensure that the map reflects real conditions, social inequalities, and regional differences, particularly in the Amazon region. By involving the people who work on dengue prevention and those affected by it, the map becomes more relevant, equitable, easier to use, and more likely to inform future research, policies, and actions that improve public health.
Next steps
Once the map and manuscript are submitted and published, the focus will be on filling the gaps. Cochrane will work to conduct systematic reviews in gaps where primary evidence is present but not yet synthesized to a high quality. We hope that by publishing our work, funders and researchers will fill gaps where primary evidence is needed.
Future iterations will be updated using AI, setting a replicable models which can be used for other areas of clinical importance.
