Artificial Inteligence-Based Decision for the Prediction of Cardioembolism in Patients with Chagas Disease and Ischemic Stroke

Chagas disease (CD) and ischemic stroke (IS) have a close, but poorly understood, association. There is paucity of evidence on the ideal secondary prophylaxis and etiological determination, with few cardioembolic patients being identified. This study aimed to describe a multicenter cohort of patients with concomitant CD and IS admitted in tertiary centers and to create a predictive model for cardioembolic embolism in CD and IS. We retrospectively studied data obtained from electronic medical and regular medical records of patients with CD and IS in several academic, hospital-based, and university hospitals across Brazil. Descriptive analyses of cardioembolic and non-cardioembolic patients were performed. A prediction model for cardioembolism was proposed with 70% of the sample as the derivation sample, and the model was validated in 30% of the sample. A total of 499 patients were analyzed. The median age was similar in both groups; however, patients with cardioembolic embolism were younger and tended to have higher alcoholism, smoking, and death rates. The predictive model for the etiological classification showed close relation with the number of abnormalities detected on echocardiography and electrocardiography as well as with vascular risk factors. Our results replicate in part those previously published, with a higher prevalence of vascular risk factors and lower median age in patients with cardioembolic etiology. Our new model for predicting cardioembolic etiology can help identify patients with higher recurrence rate and therefore allow an optimized strategy for secondary prophylaxis.