Prediction of prolonged length of stay in the intensive care unit after cardiac surgery: the need for a multi-institutional risk scoring system.

BACKGROUND AND AIM OF THE STUDY: Predictive models for the length of stay (LOS) in the intensive care unit (ICU) following cardiac surgery have been developed in the last decade. These risk models use different endpoint and risk factor definitions. This review discusses the need for a uniform multi-institutional risk scoring system for a prolonged ICU LOS. METHODS: The MEDLINE database was searched for studies assessing the prognostic value of clinical variables predicting ICU LOS. Information on study design, patient population, extended ICU LOS definition, and predictors was retrieved. RESULTS: There is no consensus on the definition of a prolonged ICU LOS. This is mainly because some studies take the continuous variables of "days in the intensive care unit" and try to make it dichotomous when actually the LOS should be analyzed as a "continuous variable." We also report a cardiac surgeon-related component. The most important risk factors were: increased age, no elective surgery, type of cardiac surgery, low left ventricular ejection fraction, recent myocardial infarction, history of pulmonary disease, history of renal disease, and reoperation/reexploration. CONCLUSIONS: There is a need for the development of a multi-institutional risk scoring system for prolonged ICU LOS following cardiac surgery. This predictive model could aid in quality assessment, practice improvement, patient counseling, and decision making. In order to develop this risk model, uniformed and standardized definitions are needed.

Preoperative prediction of intensive care unit stay following cardiac surgery.

OBJECTIVE: Following cardiac surgery, a great variety in intensive care unit (ICU) stay is observed, making it often difficult to adequately predict ICU stay preoperatively. Therefore, a study was conducted to investigate, which preoperative variables are independent risk factors for a prolonged ICU stay and whether a patient's risk of experiencing an extended ICU stay can be estimated from these predictors. METHODS: The records of 1566 consecutive adult patients who underwent cardiac surgery at our institution were analysed retrospectively over a 2-year period. Procedures included in the analyses were coronary artery bypass grafting, valve replacement or repair, ascending and aortic arch surgery, ventricular rupture and aneurysm repair, septal myectomy and cardiac tumour surgery. For this patient group, ICU stay was registered and 57 preoperative variables were collected for analysis. Descriptives and log-rank tests were calculated and Kaplan-Meier curves drawn for all variables. Significant predictors in the univariate analyses were included in a Cox proportional hazards model. The definitive model was validated on an independent sample of 395 consecutive adult patients who underwent cardiac surgery at our institution over an additional 6-month period. In this patient group, the accuracy and discriminative abilities of the model were evaluated. RESULTS: Twelve independent preoperative predictors of prolonged ICU stay were identified: age at surgery>75 years, female gender, dyspnoea status>New York Heart Association class II (NYHA II), unstable symptoms, impaired kidney function (estimated glomerular filtration rate (eGFR)<60 ml min(-1)), extracardiac arterial disease, presence of arrhythmias, mitral insufficiency>colour flow mapping (CFM) grade II, inotropic support, intra-aortic balloon pumping (IABP), non-elective procedures and aortic surgery. The individual effect of every predictor on ICU stay was quantified and inserted into a mathematical algorithm (called the Morbidity Defining Cardiosurgical (MDC) index), making it possible to calculate a patient's risk of having an extended ICU stay. The model showed very good calibration and very good to excellent discriminative ability in predicting ICU stay >2, >5 and >7 days (C-statistic of 0.78; 0.82 and 0.85, respectively). CONCLUSIONS: Twelve independent preoperative risk factors for a prolonged ICU stay following cardiac surgery were identified and constructed into a proportional hazards model. Using this risk model, one can predict whether a patient will have a prolonged ICU stay or not.

Is EuroSCORE useful in the prediction of extended intensive care unit stay after cardiac surgery?

OBJECTIVE: Risk stratification allows preoperative assessment of cardiac surgical risk faced by individual patients and permits retrospective analysis of postoperative complications in the intensive care unit (ICU). The aim of this single-center study was to investigate the prediction of extended ICU stay after cardiac surgery using both the additive and logistic model of the European System for Cardiac Operative Risk Evaluation (EuroSCORE). METHODS: A retrospective observational study was conducted. We collected clinical data of 1562 consecutive patients undergoing cardiac surgery over a 2-year period at the Antwerp University Hospital, Belgium. EuroSCORE values of all patients were obtained. The outcome measure was the duration of ICU stay in days. The predictive performance of EuroSCORE was analyzed by the discriminatory power of a receiver operating characteristic (ROC) curve. Each EuroSCORE value was used as a theoretical cut-off point to predict duration of ICU stay. Three subsequent ICU stays were defined as prolonged: more than 2, 5 and 7 days. ROC curves were constructed for both the additive and logistic model. RESULTS: Patients had a median ICU stay of 2 days and a mean ICU stay of 5.5 days. Median additive EuroSCORE was 5 (range, 0-22) and logistic EuroSCORE was 3.94% (range, 0.00-87.00). In the additive EuroSCORE model, a predictive value of 0.76 for an ICU stay of >7 days, 0.72 for >5 days and 0.67 for >2 days was found. The logistic EuroSCORE model yielded an area under the ROC curve of 0.77, 0.75 and 0.68 for each ICU length of stay, respectively. CONCLUSIONS: In our patient database, prolonged length of stay in the ICU correlated positively with EuroSCORE. The logistic model was more discriminatory than the additive in tracing extended ICU stay. The overall predictive performance of EuroSCORE is acceptable and most likely based on the presence of variables that are risk factors for both mortality and extended ICU stay. Hence, EuroSCORE is a useful predicting tool and provides both surgeons and intensivists with a good estimate of patient risk in terms of ICU stay.