Medical research is the cornerstone of evidence-based healthcare, yet its global production remains strikingly uneven. Although 80% of the world's population lives in low- and middle-income countries (LMICs), the majority of medical knowledge is still generated in high-income settings. This disconnect between where diseases impose the greatest burden and where scientific attention is focused raises concerns about the external validity of research, equitable access to innovations, and the global distribution of research capacity [1]. In this paper, we ofer a fine-grained global view of how medical research is distributed across space, time, and disease, and how closely that distribution tracks the geography of health needs [2]. We construct a geographyto (i) the diseases they study, (ii) the countries or territories whose data or patients they analyse, and (iii) the institutional homes of their authors. The underlying pipeline combines two large-language-model prompts: one that extracts biomedical relations and another that parses study context, mapping the resulting free-text strings to controlled vocabularies via neural embeddings. A new crosswalk aligns medical subject headings with the Global Burden of Disease taxonomy, and a companion module classifies tens of thousands of funders into public, corporate, philanthropic, and hybrid categories.
CEUR Workshop
ISSN1613-0073 The authors used large language models (LLMs) to assist with grammar and style editing, as well as code generation and refactoring. LLMs were also employed within the research methodology for text analysis supporting the construction of the medical knowledge graph. No figures were generated using generative AI. All AI-assisted outputs were reviewed and verified by the authors, who take full responsibility for the content of this publication.