What would other emergency stroke teams do? Using explainable machine learning to understand variation in thrombolysis practice

Machine Learning
Explainable AI
Monte Carlo Simulation Model
Authors
Affiliation

Mike Allen

PenCHORD

Anna Laws

PenCHORD

Kerry Pearn

PenCHORD

The multi-year SAMueL and SAMueL-2 projects have looked at between-hospital variation in the use of thrombolysis across hospitals.

Using XGBoost machine learning models and SHAP values to provide explainability, they have been able to look at how patient features and hospitals affect the likelihood of receiving thrombolysis.

https://arc-swp.nihr.ac.uk/publications/what-would-other-emergency-stroke-teams-do-using-explainable-machine-learning-to-understand-variation-in-thrombolysis-practice-2/

Clinical pathway simulation (using Monte Carlo simulation methods) could then be applied to look at the potential changes in hospital pathways on the percentage of people being thrombolysed.

https://arc-swp.nihr.ac.uk/publications/clinical-pathway-simulation-thrombolysis-acute-stroke/

https://www.ahajournals.org/doi/10.1161/STROKEAHA.121.038454

https://samuel-book.github.io/samuel-1/introduction/intro.html