Machine learning-based scoring model for predicting mortality in ICU-admitted ischemic stroke patients with moderate...

Area of Science:

• Neurology
• Artificial Intelligence in Medicine
• Critical Care Medicine

Background:

• Stroke is a major global cause of death and disability.
• Ischemic stroke patients with moderate to severe consciousness disorders represent a high-risk group.
• Predictive models are crucial for managing this patient population.

Purpose of the Study:

• To develop and validate an automated machine learning scoring system.
• To predict short-term (3, 7-day) and long-term (30, 90-day) mortality.
• To aid clinical decision-making for high-risk stroke patients.

Main Methods:

• Retrospective analysis of 648 ischemic stroke patients from the MIMIC-IV database (GCS ≤12).
• Exclusion of patients with speech dysfunction but clear consciousness.
• Utilized the AutoScore framework to identify top predictors and build machine learning models.

Main Results:

• Mortality rates observed: 8.02% (3-day), 18.67% (7-day), 33.49% (30-day), 38.89% (90-day).
• Area Under the Curve (AUC) for mortality prediction models ranged from 0.678 to 0.730.
• The developed models demonstrated effective predictive performance.

Conclusions:

• A novel machine learning scoring tool was developed and validated.
• The tool accurately predicts mortality in ischemic stroke patients with consciousness disorders.
• This system can improve clinical decision-making and resource allocation in ICUs.