Machine learning to predict mortality for aneurysmal subarachnoid hemorrhage (aSAH) using a large nationwide EHR...

Area of Science:

• Neurology
• Medical Informatics
• Computational Biology

Background:

• Aneurysmal subarachnoid hemorrhage (aSAH) is a life-threatening condition requiring rapid clinical assessment.
• Early prediction of mortality in aSAH patients is crucial for timely intervention and resource allocation.

Purpose of the Study:

• To develop and evaluate machine learning models for predicting in-hospital mortality in aSAH patients.
• To identify early risk factors associated with aSAH mortality.

Main Methods:

• Utilized a large cohort (6728 patients) from the Cerner Health Facts EHR database (2000-2018).
• Applied various machine learning methods to predict in-hospital mortality using the initial 24 hours of EHR data.
• Evaluated model performance using the area under the receiver operating characteristic curve (AUC).
• Employed logistic regression to identify significant mortality risk factors.

Main Results:

• Machine learning models achieved an average AUC of 0.805 for predicting aSAH mortality.
• Identified 42 risk factors, including age and serum glucose, significantly correlated with mortality.
• Logistic regression maintained prediction power while identifying key risk factors.

Conclusions:

• Machine learning offers a powerful and practical approach for early aSAH mortality prediction.
• These prognostic tools can support bedside clinical decision-making for aSAH patients.
• Early identification of risk factors can guide targeted management strategies.