Prediction of the risk of mortality using risk score in patients with coronary heart disease

// Qian Chen 1, 2 , Ding Ding 1 , Yuan Zhang 3 , Yunou Yang 1 , Qing Li 1 , Xuechen Chen 1 , Gang Hu 2 , Wenhua Ling 1 1 Guangdong Provincial Key Laboratory of Food, Nutrition and Health, Department of Nutrition, School of Public Health, Sun Yat-sen University, Guangdong, China 2 Chronic Disease Epidemiology Laboratory, Pennington Biomedical Research Center, Baton Rouge, Louisiana, USA 3 Department of Cardiology, General Hospital of Guangzhou Military Command of People’s Liberation Army, Guangdong, China Correspondence to: Wenhua Ling, email: lingwh@mail.edu.cn Gang Hu, email: gang.hu@pbrc.edu Keywords: coronary heart disease, risk score, mortality, cohort study Received: August 23, 2016 Accepted: October 17, 2016 Published: November 07, 2016 ABSTRACT Background: The aim of the study is to develop risk scores with traditional factors for all-cause and cardiovascular mortality among coronary heart disease (CHD) patients. Methods and Results: We performed a prospective cohort study of 1911 CHD patients aged 40 and older. Cox models were used to estimate the association of traditional factors [sex, age, fasting blood glucose (FBG), high-density lipoprotein cholesterol (HDL-C), low-density lipoprotein cholesterol (LDL-C), blood pressure (BP), and cigarette use] and risk scores with all-cause and cardiovascular mortality. During a mean follow-up of 4.9 years, 232 deaths were identified, 159 of which were cardiovascular-related. Both 4-year and whole follow-up data showed age, sex, HDL-C, LDL-C, and FBG were significantly associated with the risk of mortality, while BP and smoking were not significant predictors in all models. We incorporated age, sex, FBG, HDL-C, and LDL-C to establish risk scores for all-cause and cardiovascular mortality in the 4-year and whole follow-up study. These risk scores were positively associated with the risk of death as quartiles and continuous variables. Assessed by the area under the receiver operating characteristic curves (AUROCs), these risk scores demonstrated strong discriminatory capacity, from 0.744 to 0.763; and the utility of these scores was confirmed with AUROCs from 0.736 to 0.756 (all P