1Dr. Arsalan Raziq, 2Dr Ansa Batool, 3Dr Khalid Rafiq, 4Dr Zafar Khan, 5Dr Muhammad
Tahir

Machine Learning Prediction of Mortality
Following Acute

Abstract:

Despite improvements in reperfusion and medical treatment, acute myocardial
infarction (AMI) continues to be a significant cause of cardiovascular morbidity and mortality. For proper monitoring and treatment, it is crucial to identify individuals who are more likely to
die early. By uncovering intricate correlations between demographic, clinical, laboratory, and
treatment-related factors, machine learning (ML) techniques may enhance mortality
prediction.The goal is to identify the clinical and laboratory factors with the highest predictive
significance and to create and assess machine-learning models for predicting death after acute
myocardial infarction.Over the course of a year, a prospective observational study was carried
out at the Peshawar Institute of Cardiology’s Department of Cardiology. We included adult
individuals who had an acute myocardial infarction. Cardiovascular risk factors, Killip class, presenting clinical symptoms, vital signs, electrocardiographic results, cardiac biomarkers, complete blood count, renal function, glucose levels, echocardiographic parameters, and
treatment-related variables were all noted. Mortality was recorded, and patients were monitored
while they were in the hospital. For the purpose of predicting mortality, algorithms for logistic
regression, random forest, support vector machines, and XGBoost were created and contrasted. Area under the receiver operating characteristic curve (AUC), sensitivity, specificity, accuracy, precision, and F1-score were used to assess the model’s performance. The most significant
predictors were found using feature-importance analysis.A clinically significant percentage of
patients in the study experienced mortality. Increased mortality was linked to older age, higher
Killip class, lower systolic blood pressure, elevated heart rate, higher serum creatinine, higher
GlobalHealth & Medicine. 2026; 8(3):254-262. Original Article
cardiac troponin levels, hyperglycemia, anaemia, and a lower left ventricular ejection fraction. The best prediction performance was shown by XGBoost, which was followed by logistic
regression, random forest, and support vector machines. Killip class, age, serum creatinine, systolic blood pressure, troponin level, and left ventricular ejection fraction were found to be
significant predictors of death by feature-importance analysis.Death after an acute myocardial
infarction can be accurately predicted using machine-learning algorithms. When compared to the
other models that were assessed, XGBoost showed better discriminatory performance. Individualised therapeutic decision-making and early identification of high-risk AMI patients
may be supported by integrating ML-based prediction with routinely accessible clinical and
laboratory information.

Keywords: Random forest, XGBoost, logistic regression, acute myocardial infarction, machine

full text