Clinical value of Logistic regression model in differential diagnosis of coronary heart disease

Dai Fenghua; Yang Linfen; Dai Hongbin; Huang Benlin

doi:10.3969/j.issn.1674-7151.2023.04.006

Abstract：Objective To investigate the clinical value of Logistic regression model based on general clinical data and routine laboratory tests in the differential diagnosis of coronary heart disease.Methods A total of 253 patients with chest tightness and chest pain admitted in Qujing Second People's Hospital from March 2019 to November 2020 were selected as study objects,and they were divided into coronary heart disease group(128 cases)and non-coronary heart disease group(125 cases)according to the coronary angiography results.The past history(including smoking history,diabetes mellitus history and hypertension history),blood cell analysis[including white blood cell count(WBC),red blood cell count(RBC),platelet count(PLT),neutrophil count(NEU),lymphocyte count(LYM),monocyte count(MON),hemoglobin(Hb),platelet distribution width(PDW)and red blood cell distribution width(RDW)],blood lipid indexes[including total cholesterol(TC),triglycerides(TG),high-density lipoprotein cholesterol(HDL-C),low-density lipoprotein cholesterol(LDL-C)and serum creatinine(SCr)],and the monocyte count to high-density lipoprotein cholesterol ratio(MON/HDL-C)was calculated.A Logistic regression model was established to analyze the independent risk factors for diagnosing coronary heart disease.The receiver operator characteristic(ROC)curve was drawn,and a calibration curve was used to evaluate the model's discrimination and calibration.Results Compared with non-coronary heart disease group,there were were statistically significant differences in gender,hypertension history,diabetes mellitus history,age,body mass index(BMI),systolic blood pressure,TC,TG,HDL-C,SCr,WBC,RBC,PDW,NEU,MON and MON/HDL-C in coronary heart disease group.The results of multivariate Logistic regression analysis showed that BMI[odds ratio(OR)was 1.33,95%confidence interval(95%CI)was 1.18-1.51,P<0.001],smoking history(OR value was 3.27,95%CI was 1.74-6.24,P<0.001),diabetes mellitus history(OR value was 2.57,95%CI was 1.38-4.89,P = 0.003)and hypertension history(OR value was 1.99,95%CI was 1.07-3.71,P = 0.03)were the independent risk factors for coronary heart disease.MON/HDL-C(OR value was 3.67,95%CI was 1.02-14.14,P = 0.05)was a strong independent risk factor for predicting coronary heart disease.Based on the above indicators,a Logistic regression model was established to predict coronary heart disease.The ROC curve and calibration curve showed that the model had good discrimination(AUC was 0.838,95%CI was 0.790-0.857)and calibration(mean absolute error was 0.017).Conclusion Logistic regression model has good identification and prediction value for the diagnosis of coronary heart disease.