Abstract WP74: Predictors Of Acute Ischemic Stroke In Young Adults With Cannabis Use Disorder: An Artificial Neural Network Analysis Using A Nationwide Cohort (2019)

Abstract

Background: An increase in cannabis use disorder (CUD) has been associated with a rise in adverse cerebrovascular events in the United States. However, large-scale data on risk stratification of acute ischemic stroke (AIS) among young patients with CUD remains limited. We aim to determine predictors of AIS in the young CUD cohort using the Artificial Neural Network (ANN) model in a national cohort. Methods: We identified hospitalizations of young adults (18-44 years) with CUD (unweighted n=101094, weighted n=539125) using the National Inpatient Sample (2019). Later we grouped them into AIS and non-AIS cohorts. The neural network’s predictive factors were selected for AIS in CUD cohort. CUD admissions were randomly split between training (80%, n=80754) & testing datasets (20%, n=20340). Training data was used to calibrate ANN, while testing data was used to evaluate the algorithm's accuracy. We compared the frequency of incorrect prediction between training and testing data and measured the area under the curve (AUC) to determine ANN’s efficacy in predicting AIS with CUD. Results: The 2019 cohort consisted of a total of 539125 admissions (56.1% male, 52.9% white, 28.7% blacks, 12.4% Hispanic, 6.0% Asian or Pacific Islander) in young adults with CUD with a median age of 30 (IQR 24-36) years. The rate of AIS admissions in young CUD cohort was 0.5% (n=2535). Training data showed improved predictions with a significantly low error rate (0.5%), thereby depicting better accuracy. The normalized importance of independent predictors of AIS in the CUD cohort is displayed in Fig. 1a . History of PVD, prior stroke/TIA, hypertension and hyperlipidemia had the highest normalized importance. The AUC was 0.811 (Fig. 1b) which shows an excellent ANN model for the prediction of AIS in young CUD cohort. Conclusion: The ANN model successfully revealed the order of prevalent independent predictors of AIS that can be utilized to screen high-risk young adults with CUD and improve outcomes.