Establishing and Validating Predictive Models for Coronary Atherosclerotic Heart Disease Risk by Utilizing SCI,TyG and AIP

CHEN Zhuo; YI Bulayin; GAO Ying; WANG Yu; PALIDA Abulaiti; XING Zhi; HALISHA Airikenjiang; LI Hui; SHAJIDAN Abudureyimu

doi:10.13730/j.issn.2097-2148.2023.11.004

Abstract：Objective To construct a predictive model for the risk of coronary atherosclerotic heart disease[coro-nary heart disease(CHD)for short]using systemic coagulation-inflammation index(SCI),triglyceride glucose(TyG)and atherogenic index of plasma(AIP),and to validate the predictive efficiency of the model.Methods A retrospective collection of patients was conducted from January 2014 to January 2021 in author's hospital.Patients were divided into control group(n=616)and CHD group(n=1056)based on the results of coronary angiography;the study population was then randomly assigned to training set(n=1170)and validation set(n=502)according to the ratio of 7:3.The clini-cal data such as blood routine and blood biochemistry of the patients were collected.In the training set,a stepwise backward regression analysis was performed to identify independent risk factors for developing CHD,the nomo-gram model was constructed,and the predictive efficiency and applicability of the model were internally validated u-sing the validation set.Results The results of the multi-factor Logistic regression analysis on the training set pop-ulation indicated that SCI,AIP,D-dimer(D-D)and high-density lipoprotein cholesterol(HDL-C)were independent protective factors for the occurrence of CHD,while age,TyG,PLR,Fib and ApoA were independent risk factors for the occurrence of CHD.Internal validation of the model showed that the area under the curve(AUC)was 0.739(95％CI:0.624-0.775)in the training set and 0.846(95％CI:0.697-0.886)in the validation set.The calibration curve results suggested good calibration of the model,and the clinical effec-tiveness of the model was tested by decision curve analysis(DCA),showing that the model has good clinical effectiveness when the threshold probabilities for the training set and validation set were within the range of 10％-50％and 10％-75％respectively,clinical impact curve(CIC)confirmed the model's effective predictive ability.Conclusion The predic-tive model demonstrates good discriminative ability and calibration,as well as a favorable net benefit,which indicating its potential for predicting the risk of CHD.