In Frontiers in cardiovascular medicine
BACKGROUND : Ischemic Heart Disease (IHD) is the leading cause of death from cardiovascular disease. Currently, most studies have focused on factors influencing IDH or mortality risk, while few predictive models have been used for mortality risk in IHD patients. In this study, we constructed an effective nomogram prediction model to predict the risk of death in IHD patients by machine learning.
METHODS : We conducted a retrospective study of 1,663 patients with IHD. The data were divided into training and validation sets in a 3:1 ratio. The least absolute shrinkage and selection operator (LASSO) regression method was used to screen the variables to test the accuracy of the risk prediction model. Data from the training and validation sets were used to calculate receiver operating characteristic (ROC) curves, C-index, calibration plots, and dynamic component analysis (DCA), respectively.
RESULTS : Using LASSO regression, we selected six representative features, age, uric acid, serum total bilirubin, albumin, alkaline phosphatase, and left ventricular ejection fraction, from 31 variables to predict the risk of death at 1, 3, and 5 years in patients with IHD, and constructed the nomogram model. In the reliability of the validated model, the C-index at 1, 3, and 5 years was 0.705 (0.658-0.751), 0.705 (0.671-0.739), and 0.694 (0.656-0.733) for the training set, respectively; the C-index at 1, 3, and 5 years based on the validation set was 0.720 (0.654-0.786), 0.708 (0.650-0.765), and 0.683 (0.613-0.754), respectively. Both the calibration plot and the DCA curve are well-behaved.
CONCLUSION : Age, uric acid, total serum bilirubin, serum albumin, alkaline phosphatase, and left ventricular ejection fraction were significantly associated with the risk of death in patients with IHD. We constructed a simple nomogram model to predict the risk of death at 1, 3, and 5 years for patients with IHD. Clinicians can use this simple model to assess the prognosis of patients at the time of admission to make better clinical decisions in tertiary prevention of the disease.
Yang Long, Dong Xia, Abuduaini Baiheremujiang, Jiamali Nueraihemaiti, Seyiti Zulihuma, Shan Xue-Feng, Gao Xiao-Ming
2023
LASSO, ischemic heart disease, machine learning, mortality, nomogram
