Risk stratification for stroke in acute persistent vertigo: development and internal validation of a multivariable...

Area of Science:

• Neurology
• Cardiovascular Medicine
• Epidemiology

Background:

• Acute persistent vertigo is a challenging stroke symptom for early bedside assessment.
• Existing risk scores have limitations in predicting stroke in these patients.

Purpose of the Study:

• Develop and internally validate a multivariable prediction model for clinically detected strokes in patients with acute persistent vertigo.
• Compare the model's performance against routinely used scores like ABCD2, CNS, and Triage-Plus.

Main Methods:

• Retrospective analysis of 689 patients with acute persistent vertigo (vertigo > 24 hours).
• LASSO-regularized logistic regression for predictor selection, followed by multivariable logistic modeling.
• Internal validation using 5-fold cross-validation and bootstrap resampling for confidence intervals.

Main Results:

• The multivariable model demonstrated excellent discrimination (AUC 0.902) and good calibration.
• Predictors included older age, smoking, hypertension, hyperlipidemia, diabetes, coronary heart disease, atrial fibrillation, higher CNS score, and nausea/vomiting; tinnitus was inversely associated.
• The model outperformed ABCD2 (AUC 0.642) and Triage-Plus (AUC 0.514) and was superior to CNS alone (AUC 0.846).

Conclusions:

• A multivariable prediction model can significantly aid in risk stratification and diagnostic decision-making for acute persistent vertigo.
• The model offers tailored stroke risk assessment, potentially improving early detection.
• Further impact studies and external validation are recommended.