Predicting functional outcome after stroke by modelling baseline clinical and CT variables.

Area of Science:

• Neurology
• Medical Imaging
• Clinical Prediction Models

Background:

• Stroke outcome prediction models are crucial for patient management.
• Assessing the added value of complex clinical variables and CT scan data to existing models is important.

Purpose of the Study:

• To evaluate if incorporating complex clinical variables and CT scan data enhances stroke outcome prediction models.
• To develop and validate a reliable stroke outcome prediction model.

Main Methods:

• 538 acute stroke patients were analyzed for independent survival at 6 months using the modified Rankin scale.
• Multivariate logistic regression was used to develop prediction models based on clinical and radiological variables.
• Models were compared using the area under the receiver operating characteristic curve (AUC) and externally validated.

Main Results:

• Model II, using simple clinical variables (age, pre-stroke independence, Glasgow coma score, arm power, ambulation), showed strong predictive performance (AUC=0.876).
• Model III, including CT scan data and complex clinical variables, was not statistically superior to Model II (AUC=0.901, P=0.12).
• Model II demonstrated robust external validation with AUCs of 0.773 and 0.787 in independent datasets.

Conclusions:

• An externally validated stroke outcome prediction model utilizing simple clinical variables was developed.
• The addition of CT-derived radiological variables or more complex clinical data did not significantly improve prediction accuracy.
• Simple clinical variables are sufficient for effective stroke outcome prediction.