Risk Prediction Model for Long-Term Mortality after Percutaneous Coronary Intervention

Introduction: There are many risk prediction models utilized to see effects of risk factors to dependent variables with final aims to aid clinicians and patients making decisions. This study mainly aimed to develop a risk prediction model for long-term mortality after percutaneous coronary intervention (PCI) for cardiovascular disease patients. Methodology: Data on 10,511 patients who underwent PCI procedure between 2008 to 2012 were obtained from a source data provider center for National Cardiovascular Disease Database (NCVD)-PCI registry. The data were randomly divided into development and validation datasets. After variable selection process, multiple logistic regression technique was used to develop a predictive model using the development dataset, then the model was validated using the validation samples. The goodness-of-fit and the performance of the models in both samples were evaluated by Hosmer-Lemeshow, and area under the receiver operating characteristic (ROC) curve. Result: Mortality rate in three years after PCI was 9.6%. Eight predictors were associated with the 3-year mortality of PCI and included in the final model. The area under the ROC curves were 0.7809 and 0.7780 in the development and validation dataset respectively. Conclusion: An accurate and reliable model was produced to predict three-year mortality after PCI procedure.