The development of a prediction score of in-hospital death in coronary care unit patients

Abstract

Objective To identify the risk factors of in-hospital death in coronary care unit patients and to develop a score for relevant risks. Methods Totally 1 067 patients from the coronary care unit between February 2013 and December 2016 were retrospectively selected from the medical record system of He'nan Provincial People's Hospital. The univariate and polychotomous Logistic regression analysis was used to identify the independent risk factors of death in coronary care unit patients. A Logistic regression model was built, and its performance in prediction and identification was tested by the area under the curve (AUC) of receiver operating characteristic (ROC) , Hosmer-Lemeshow goodness-of-fit (GOF) and O/E value. Results According to the polychotomous Logistic regression analysis, finally six risk independent risks factors were included in the score model, and they were scored as follows: age (1) , history of coronary heart disease (CHD, 1) , admission cardiac function level (2) , consciousness (1) , systolic pressure (2) and oxygen saturation (2) . The test suggested that the score's identification and calibration for the occurrence of in-hospital death among coronary care unit patients was high (AUC=0.97; Hosmer-Lemeshow GOF: P=0.995; O/E value=1.13) . Conclusions The final death risk score includes six risk factors, each scored between 1 and 6, and the total score is 21. The patients scored lower than 7 is defined as the low-risk group; the patients scored between 7 and 10 is defined as the medium-risk group; the patients scored between 11 and 14 is defined as high-risk group; and the patients scored higher than 15 is defined as the extremely high-risk group. This death risk score can predict the occurrence of in-hospital death among coronary care unit patients accurately. Key words: Death; Patients; Coronary care unit; Risk prediction