Enhancing stroke-associated pneumonia prediction in ischemic stroke: An interpretable Bayesian network approach.

Area of Science:

• Neurology
• Medical Informatics
• Predictive Modeling

Background:

• Stroke-associated pneumonia (SAP) is a leading cause of mortality after ischemic stroke (IS).
• Current predictive models for SAP often lack transparency and interpretability, hindering clinical application.
• There is a need for accurate and interpretable tools to assess SAP risk in IS patients.

Purpose of the Study:

• To develop an interpretable Bayesian network (BN) model for predicting SAP in IS patients.
• To enhance both the predictive accuracy and clinical interpretability of SAP risk assessment.
• To compare the BN model's performance against existing methods.

Main Methods:

• A retrospective study of 1252 IS patients admitted between January and December 2019.
• Clinical data within 48 hours of admission were analyzed for SAP occurrence within 7 days.
• A BN model was developed using a hill-climbing algorithm, with dimensionality reduction and data balancing techniques employed.

Main Results:

• The BN model identified age, risk of pressure injury (PI), National Institutes of Health Stroke Scale (NIHSS) score, and C-reactive protein (CRP) as significant predictors of SAP.
• The BN model achieved an AUC of 0.85 on the test set, outperforming other models.
• Decision curve analysis indicated a greater net benefit for clinical decision-making compared to existing systems.

Conclusions:

• Age, PI risk, NIHSS score, and CRP are significant prognostic factors for SAP in IS patients.
• The developed interpretable BN model shows superior predictive performance and interpretability.
• The BN model holds potential as an effective clinical decision support tool for SAP risk assessment.