| Summary: | Cardiovascular Diseases (CVD) rank as the primary factor responsible for non-traumatic incidents of Out-of-Hospital Cardiac Arrest (OHCA) among adults and stands out as the predominant contributor to both mortality and morbidity rates in Malaysia. Consequently, it is crucial to promptly identify individuals afflicted by CVD or those at an elevated risk of experiencing cardiac arrest. This proactive identification is essential for effectively addressing and mitigating this issue. Given the elevated mortality rate, identifying variables potentially leading to cardiac arrest presents a significant challenge, underscoring the necessity for primary preventive measures and comprehensive risk assessment strategies. In this study, machine learning models; Support Vector Machines (SVM), Logistic Regression (LR) and Random Forest (RF) are used to build a prognostic model that can predict the probability of having cardiac arrest as an outcome. The data used to train and test the models merge five heart disease datasets collected from UCI Machine Learning Repository with 11 selected common attributes. The data was pre-processed, analysed and fit to the models to determine the patient attributes that are important for the models. The study suggested that older patients and patients with higher ST depression levels are more likely to get cardiac arrest. The slope of the peak exercise ST segment is found to be the most important attribute influencing the prediction of the models. After evaluation by cross-validation, RF model achieved the highest accuracy and performed best with n_estimators parameter tuned to 40. The model was then deployed on a web application to enable live prediction by user input of patient attributes, hoping to contribute to the early detection of OHCA. © 2023 IEEE.
|
|---|