Evaluating Outcome Prediction Models in Endovascular Stroke Treatment Using Baseline, Treatment, and Posttreatment...

Area of Science:

• Neurology
• Medical Statistics
• Interventional Cardiology

Background:

• Current statistical models for predicting outcomes after endovascular therapy for acute ischemic stroke primarily use baseline variables.
• There is a need to assess the performance of these models by iteratively incorporating treatment-related and posttreatment variables.

Purpose of the Study:

• To evaluate the performance of stroke outcome prediction models for endovascular therapy by iteratively adding baseline, treatment-related, and posttreatment variables.
• To determine the contribution of different variable types to outcome prediction accuracy.

Main Methods:

• Developed four multivariable logistic regression models using data from the ESCAPE-NA1 trial.
• Model 1: baseline variables; Model 2: baseline + treatment variables; Model 3: Model 2 + early posttreatment variables; Model 4: Model 3 + late posttreatment variables.
• Compared model performance using the area under the receiver operating characteristic curve (AUC) and Shapley values to assess variable contributions.

Main Results:

• The area under the receiver operating characteristic curve (AUC) improved iteratively from 0.74 (baseline only) to 0.82 (all variables).
• Even with the most comprehensive model, 19.8% of patients achieving functional independence were misclassified.
• Patient age, severe adverse events (pneumonia), and early intracranial hemorrhage were significant contributors to outcome variance.

Conclusions:

• Factors beyond baseline variables significantly contribute to outcomes after endovascular therapy for acute ischemic stroke.
• Current outcome prediction models have limited accuracy, with substantial misclassification rates.
• Caution is advised when applying existing models in clinical practice due to their limitations.