Abstract
What we set out to do
The project aimed to create a comprehensive and actionable resource of research outputs for the Health Foundation from a decade of funded projects. It was to synthesise findings and generate insights from historical Health Foundation funded research to:
• better understand and utilize findings to direct future work, inform strategy and funding
• provide greater confidence and insight to draw on research findings in internal and external work
• inform decisions on the methodological approaches of our work
• underpin and support external communications where appropriate
• develop a comprehensive record of outputs from Health Foundation funded research to populate a new research cataloguing system.
To achieve these aims, the project sought to align identified outputs with the Health Foundation’s strategic priorities, including improving health and reducing inequalities, fostering innovation, and informing health and care policy. Key stakeholders included the Health Foundation’s thematic groups, policy experts, and operational teams, all of whom contributed to shaping the project’s direction and focus. The intended impact was to enhance strategic planning, support evidence-based decision-making, and explore the potential of Artificial Intelligence (AI) in improving evidence synthesis
processes.
What we achieved
The project successfully delivered a verified knowledge catalogue encompassing Health Foundation-funded outputs, providing a centralised and searchable repository for retrospective analysis. Thematic evidence summaries were produced, offering synthesised insights aligned with the Health Foundation’s priorities. These summaries were validated by subject matter experts, assessing their accuracy and relevance. Furthermore, the project demonstrated a proof of concept for integrating tools such as Apache SOLR, which is an open-source search platform to search and retrieve results from large document repositories, and Large Language Models (LLMs) into evidence synthesis workflows.
What we learned
The project underscored the value of leveraging AI to streamline evidence synthesis, particularly in managing large datasets and producing initial summaries. However, it also revealed critical challenges, including the limitations of AI tools in handling nuanced findings, the selection biases introduced by rule-based systems like SOLR, and the difficulties in aligning AI-generated outputs with broader evidence bases. Organisational readiness for adopting AI solutions emerged as a key consideration, as the project highlighted the need for dedicated training, cultural adaptation, and robust change management to support the integration of new technologies and practices. The dual-tool approach, involving both the knowledge catalogue and SOLR, raised issues of duplication and inefficiency, emphasising the importance of cohesive system design.
Our recommendations
Building on these findings, the Health Foundation should integrate the knowledge catalogue into its strategic planning processes while complementing it with broader evidence synthesis to mitigate selection biases. Future projects should adopt a hybrid model that combines AI capabilities with human expertise, ensuring accuracy and contextual relevance. Investments in user-friendly tool design, comprehensive training programmes, and transparent cost models for AI tools will be essential for scaling these innovations. The Health Foundation is encouraged to focus on bridging the gap between retrospective analysis and forward-looking strategic priorities, leveraging its learnings to enhance its contributions to health and care policy and practice.
The project aimed to create a comprehensive and actionable resource of research outputs for the Health Foundation from a decade of funded projects. It was to synthesise findings and generate insights from historical Health Foundation funded research to:
• better understand and utilize findings to direct future work, inform strategy and funding
• provide greater confidence and insight to draw on research findings in internal and external work
• inform decisions on the methodological approaches of our work
• underpin and support external communications where appropriate
• develop a comprehensive record of outputs from Health Foundation funded research to populate a new research cataloguing system.
To achieve these aims, the project sought to align identified outputs with the Health Foundation’s strategic priorities, including improving health and reducing inequalities, fostering innovation, and informing health and care policy. Key stakeholders included the Health Foundation’s thematic groups, policy experts, and operational teams, all of whom contributed to shaping the project’s direction and focus. The intended impact was to enhance strategic planning, support evidence-based decision-making, and explore the potential of Artificial Intelligence (AI) in improving evidence synthesis
processes.
What we achieved
The project successfully delivered a verified knowledge catalogue encompassing Health Foundation-funded outputs, providing a centralised and searchable repository for retrospective analysis. Thematic evidence summaries were produced, offering synthesised insights aligned with the Health Foundation’s priorities. These summaries were validated by subject matter experts, assessing their accuracy and relevance. Furthermore, the project demonstrated a proof of concept for integrating tools such as Apache SOLR, which is an open-source search platform to search and retrieve results from large document repositories, and Large Language Models (LLMs) into evidence synthesis workflows.
What we learned
The project underscored the value of leveraging AI to streamline evidence synthesis, particularly in managing large datasets and producing initial summaries. However, it also revealed critical challenges, including the limitations of AI tools in handling nuanced findings, the selection biases introduced by rule-based systems like SOLR, and the difficulties in aligning AI-generated outputs with broader evidence bases. Organisational readiness for adopting AI solutions emerged as a key consideration, as the project highlighted the need for dedicated training, cultural adaptation, and robust change management to support the integration of new technologies and practices. The dual-tool approach, involving both the knowledge catalogue and SOLR, raised issues of duplication and inefficiency, emphasising the importance of cohesive system design.
Our recommendations
Building on these findings, the Health Foundation should integrate the knowledge catalogue into its strategic planning processes while complementing it with broader evidence synthesis to mitigate selection biases. Future projects should adopt a hybrid model that combines AI capabilities with human expertise, ensuring accuracy and contextual relevance. Investments in user-friendly tool design, comprehensive training programmes, and transparent cost models for AI tools will be essential for scaling these innovations. The Health Foundation is encouraged to focus on bridging the gap between retrospective analysis and forward-looking strategic priorities, leveraging its learnings to enhance its contributions to health and care policy and practice.
| Original language | English |
|---|---|
| Publisher | Edge Hill University |
| Commissioning body | The Health Foundation |
| Number of pages | 159 |
| Publication status | Published - 23 Apr 2025 |
Keywords
- health
- artificial intelligence
- Report
- Language model
- Synthesis
