Background Critically ill patients with cardiogenic shock could benefit from ventricular assist device support using the Impella microaxial blood pump. However, recent studies suggested Impella not to improve outcomes. We, therefore, evaluated outcomes and predictors in a real-world scenario. Methods In this retrospective single-center trial, 125 patients suffering from cardiac arrest/cardiogenic shock between 2008 and 2018 were analyzed. 93 Patients had a prior successful cardiopulmonary resuscitation. The primary endpoint was hospital mortality. Associations of covariates with the primary endpoint were assessed by univariable and multivariable logistic regression. Adjusted odds ratios (aOR) and optimal cut-offs (using Youden index) were obtained. Results Hospital mortality was high (81%). Baseline lactate was 4.7mmol/L [IQR = 7.1mmol/L]. In multivariable logistic regression, only age (aOR 1.13 95%CI 1.06-1.20; p3.3mmol/L and age 66 years. Patients were retrospectively stratified into three risk groups: Patients aged [less than or equal to]66 years and lactate [less than or equal to]3.3mmol (low-risk; n = 22); patients aged 66 years or lactate 3.3mmol/L (medium-risk; n = 52); and patients both aged 66 years and lactate 3.3mmol/L (high-risk, n = 51). Risk of death increased from 41% in the low-risk group, to 79% in the medium risk group and 100% in the high-risk group. The predictive abilities of this model were high (AUC 0.84; 95% 0.77-0.92). Conclusion Mortality was high in this real-world collective of severely ill cardiogenic shock patients. Better patient selection is warranted to avoid unethical use of Impella. Age and lactate might help to improve patient selection.