Abstract
BACKGROUND: The CREST model is a prediction model, quantitating the risk of circulatory-etiology death (CED) after cardiac arrest based on variables available at hospital admission, and intend to guide the triage of comatose patients without ST-segment-elevation myocardial infarction after successful cardiopulmonary resuscitation. This study assessed performance of the CREST model in the Target Temperature Management (TTM) trial cohort. METHODS: We retrospectively analyzed data from resuscitated out-of-hospital cardiac arrest (OHCA) patients in the TTM-trial. Demographics, clinical characteristics, and CREST variables (history of coronary artery disease, initial heart rhythm, initial ejection fraction, shock at admission and ischemic time > 25 min) were assessed in univariate and multivariable analysis. The primary outcome was CED. The discriminatory power of the logistic regression model was assessed using the C-statistic and goodness of fit was tested according to Hosmer-Lemeshow. RESULTS: Among 329 patients eligible for final analysis, 71 (22%) had CED. History of ischemic heart disease, previous arrhythmia, older age, initial non-shockable rhythm, shock at admission, ischemic time > 25 min and severe left ventricular dysfunction were variables associated with CED in univariate analysis. CREST variables were entered into a logistic regression model and the area under the curve for the model was 0.73 with adequate calibration according to Hosmer-Lemeshow test (p = 0.602). CONCLUSIONS: The CREST model had good validity and a discrimination capability for predicting circulatory-etiology death after resuscitation from cardiac arrest without ST-segment elevation myocardial infarction. Application of this model could help to triage high-risk patients for transfer to specialized cardiac centers.