Development of Atrial Fibrillation Prediction Models in Ischemic Stroke Patients Using Electronic Medical Records and Text Mining Techniques

Cerebrovascular disease, which is also known as stroke, is the second largest reason of deaths of human worldwide and the third largest reason of disability. Atrial fibrillation (AF) is the potential factor to cause ischemic stroke, and it is strongly related to ischemic stroke as well. The incidence of ischemic stroke and cardiovascular mortality in patients with AF were five times and two times higher than those in patients without AF, respectively. However, AF is difficult to detect, and there are often paroxysmal episodes that are misdiagnosed as asymptomatic and cannot be properly treated. When AF is detected in an ischemic stroke patient, strategies for secondary prevention of stroke usually change accordingly. The aim of this study is to use electronic medical records and the machine learning algorithms to build early AF prediction models for ischemic stroke inpatients. In addition, this study also examines whether the performance of the prediction models established by structured data and unstructured data is different. It is hoped that the model established in this study can assist doctors to make more effective decisions, thereby allowing medical resources to be effectively utilized.