Stroke is a leading cause of death and disability in the UK. Much of stroke management revolves around addressing risk factors with medications and lifestyle modification. However, 30% of those who suffer stroke have no known risk factors. Hence, there is need to better identify individuals who are at high risk of stroke, and particularly those where the benefit of treatment outweighs the risk.
ABSTRACT is a three phase study that looks to address this issue by (1) using artificial intelligence (AI) to predict stroke risk from routine hospital data, (2) to validate this model on external datasets, and (3) validate the ability to improve outcome by guiding clinical decision making.
Phase 1 has shown promising results, with Xgboost achieving an AUC of 94% when using routine blood test data, medical history and CT/MRI head imaging data to predict future stroke risk from a cohort of 9155 stroke cases and 109,875 controls in Southwest England. The model also identified several novel risk factors for stroke, such as liver function tests and C-reactive protein.
We now look to commence phase 2 of our project and validate these findings on an external dataset. At this time we ask the following research questions:
1. Can AI accurately predict stroke risk from routine hospital data?
2. How well does our model generalise to UK Biobank participants?
3. Does model performance vary across demographic and clinical subgroups?
Based on these questions, our research aims and objectives are therefore as follows:
1. To perform external validation of our stroke risk model using UKbiobank data.
2. To validate the novel risk factors for stroke identified by our models in phase 1.
3. To Assess model performance and calibration across different demographic subgroups
Following the results of this external validation we will then look to commence phase 3 of our project and assess the effectiveness of models in guiding clinical decision making and stroke risk management.