Clinical Decision Support Systems for Predicting Patients Liable to Acquire Acute Myocardial Infarctions

Abstract

Acute myocardial infarction (AMI) is a major cause of death worldwide. There are around 0.8 million persons suffered from AMI annually in the US and the death rate reaches 27%. The risk factors of AMI were reported to include hypertension, family history, smoking habit, diabetes, serenity, obesity, cholesterol, alcoholism, coronary artery disease, etc. In this study, data acquired from a subset of the National Health Insurance Research Database (NHIRD) of Taiwan were used to develop the clinical decision support system (CDSS) for predicting AMI. Support vector machine integrated with genetic algorithm (IGS) was adopted to design the AMI prediction models. Data of 6087 AMI patients and 6087 non-AMI patients, each includes 50 features, were acquired for designing the predictive models. Tenfold cross validation and three objective functions were used for obtaining the optimal model with best prediction performance during training. The experimental results show that the CDSSs reach a prediction performance with accuracy, sensitivity, specificity, and area under ROC curve (AUC) of 81.47–84.11%, 75.46–80.94%, 86.48–88.21%, and 0.8602–0.8935, respectively. The IGS algorithm and comorbidity-related features are promising in designing strong CDSS models for predicting patients who may acquire AMI in the near future.