Clinical Research Directory

Risk Analysis of Intracranial Aneurysm Rupture Through Social Determinants of Health

Intracranial aneurysms (IA) are arterial malformations affecting about 3% of the overall population. Rupture is the most severe complication, as it is associated with nearly 30% of death or severe disability. The available scores to assess rupture risk are mainly based on usual modifiable and non-modifiable risk factors from the literature, but they appear insufficient to predict rupture. Emerging factors, such as sleep apnea syndrome and the use of certain medications, seem to influence the risk of rupture. The study of social determinants of health (SDOH) is highly relevant, given numerous reports showing the impact of SDOH, in addition to vascular risk factors, on vascular diseases like ischemic stroke or myocardial infarction. It is therefore reasonable to study the interaction between rupture risk factors and SDOH on the rupture risk of IA. Several initiatives have been undertaken to assess rupture risk, but few have included SDH. Limitations were often raised, especially regarding data accessibility. However, it is now possible, thanks to artificial intelligence (AI) algorithms, particularly natural language processing (NLP), to reuse large-scale health data to address longstanding issues, such as those posed by SDH. The use of health data warehouses (HDWs) offers an opportunity to collect and analyze accurate, real-world data, particularly through AI and NLP to extract information from medical reports. However, various challenges limit the use of NLP models, notably the dominance of models trained on English medical texts and privacy-related legislative restrictions. Therefore, alongside leveraging these models for clinical research, it is essential to continue efforts to develop transparent French-language models that comply with legislation. Thus, the ARAMISS project proposes to study the interaction between SDH and known risk factors for IA rupture by comparing control populations and rupture cases. This study will be based on a certified health data warehouse (HDW) and an NLP algorithm previously developed by the team. In parallel, the project plans two FAIR-compliant knowledge-sharing approaches to disseminate the algorithm and training corpus to the broader community.