Predicting outcome in ischemic stroke: external validation of predictive risk models

Area of Science:

• Neurology
• Clinical Prediction Models
• Stroke Research

Background:

• Six multivariable models for predicting 3-month outcomes of acute ischemic stroke were previously developed and internally validated.
• External validation in an independent dataset was necessary to assess generalizability.

Purpose of the Study:

• To externally validate six previously developed multivariable models for predicting 3-month outcomes in acute ischemic stroke patients.
• To assess the performance of these models in an independent dataset.

Main Methods:

• The study utilized data from 299 ischemic stroke patients who received placebo in the National Institute of Neurological Disorders and Stroke rt-PA trial.
• Model equations incorporated 6 acute clinical variables and head CT infarct volume to predict 3-month National Institutes of Health Stroke Scale, Barthel Index, and Glasgow Outcome Scale.
• Discrimination was measured using the area under the receiver operator characteristic curve (AUC), and calibration was assessed using calibration charts.

Main Results:

• The validation dataset included patients with more severe stroke (higher NIHSS and infarct volume) than the model development cohort.
• External validation demonstrated minimal degradation in model performance, with AUC values ranging from 0.75 to 0.89.
• Calibration curves indicated fair to good calibration of the models in the external dataset.

Conclusions:

• The previously developed models exhibited excellent discrimination and acceptable calibration when validated on an independent dataset.
• Further development and validation of models using only acutely available variables are needed for improved clinical application.