Aneurysmal Subarachnoid Haemorrhage

Aneurysmal Subarachnoid Haemorrhage (aSAH) is a severe type of stroke caused by the rupture of an intracranial aneurysm, leading to bleeding in the subarachnoid space. Despite accounting for only a small percentage of strokes globally, aSAH has a disproportionate impact due to its high mortality and morbidity rates, particularly affecting individuals at a relatively young age (average 55 years). Early detection and prevention are vital but remain challenging due to limitations in current predictive tools and a lack of understanding of key risk factors. The thesis leverages advanced data-driven approaches, including machine learning and statistical models, to address these gaps in aSAH research. The thesis has three overarching objectives. 1) Enhancing aSAH Risk Prediction. Several chapters focus on developing and improving clinical prediction models using electronic health record (EHR) data and population studies. Chapter 2 highlights the importance of data preparation for model performance, demonstrating how choices such as outcome definitions and handling of missing values impact predictions. Chapter 3 builds on this foundation to predict cardiovascular risks in younger individuals using routine care data, providing insights into sex-specific factors. Chapter 4 introduces a model that differentiates aSAH risk from other stroke types, identifying unique predictive markers. Chapter 5 overcomes data limitations by using systematically collected UK Biobank data to create and validate the SMA2SH2ERS risk prediction model for aSAH. 2) Identifying Novel Risk Factors: Machine learning techniques are applied to uncover previously unidentified risk factors for aSAH. Chapter 6 demonstrates how combining data-driven methods with traditional statistics enables the discovery of new predictors, including lifestyle and comorbidity factors, which could aid in earlier detection and prevention strategies. 3) Exploring Drug-Based Prevention: The thesis investigates the potential of pharmacological treatments as non-invasive interventions for aSAH. Chapter 7 employs a drug-wide association study (DWAS) to analyze primary care data, identifying medications associated with reduced aSAH incidence. Chapter 8 further explores the antihypertensive drug lisinopril, revealing its potential for lowering aSAH risk compared to other drugs in the same class, suggesting a targeted, preventive approach for at-risk populations. The thesis concludes with a general discussion (Chapter 9) summarizing the implications of the findings, including their relevance for clinical practice and public health. The integration of machine learning with population-based datasets offers powerful tools to refine risk prediction and identify actionable prevention strategies. However, challenges such as data heterogeneity, misclassification, and missing values in EHR systems highlight the need for standardized, high-quality data collection methods. The thesis represents a step towards better risk stratification, early detection, and prevention of aneurysmal subarachnoid haemorrhage, ultimately improving patient outcomes through innovative, data-driven methods.