Identifying novel risk factors for aneurysmal subarachnoid haemorrhage using machine learning

Area of Science:

• Neurology
• Genetics
• Epidemiology

Background:

• Aneurysmal subarachnoid haemorrhage (aSAH) presents significant mortality and morbidity challenges.
• Identifying novel risk factors is crucial for improving patient outcomes and preventative strategies.

Purpose of the Study:

• To uncover new predictors of aSAH by integrating machine learning (ML) with conventional statistical analysis.
• To analyze a comprehensive dataset for potential novel associations with aSAH risk.

Main Methods:

• Utilized the UK Biobank dataset, identifying aSAH cases through ICD codes.
• Employed the CatBoost ML algorithm and SHAP for variable importance analysis.
• Conducted logistic regression on top identified variables, adjusting for known risk factors.

Main Results:

• Analysis of 501,847 participants revealed 893 aSAH cases.
• Machine learning identified 214 variables with predictive influence.
• Four potential novel risk factors emerged: increased risk with mean sphered cell volume and tea intake; decreased risk with peak expiratory flow and haematocrit percentage.

Conclusions:

• Mean sphered cell volume, tea intake, peak expiratory flow, and haematocrit percentage represent potential novel risk factors for aSAH.
• Further research is warranted to validate these findings and explore complex ML-indicated relationships.