Fast-track failure after cardiac surgery: development of a prediction model.

Objective: Risk factors for unsuccessful fast-tracking of cardiac surgery patients have not been collectively defined in the literature. The aim of this study was to determine risk factors for fast-track failure and incorporate them into a predictive fast-track failure score. Design: Prospective observational study. Setting: Cardiothoracic Department of St Mary's Hospital, London. Patients: Data were collected from April 2003 to April 2005 including 1,084 patients undergoing heart surgery who were admitted into the fast-track unit. Interventions: Multifactorial logistic regression was used to develop a propensity score for estimating the likelihood of fast-track failure. Measurements and Main Results: One hundred and sixty-nine patients failed fast-track management (15.6%). Independent predictors for fast-track failure were impaired left ventricular function with or without recent acute coronary syndrome (odds ratios 2.89 and 1.65 respectively), re-do operation (one, two, or more vs. none, odds ratio 1.75, 7.98), extracardiac arteriopathy (odds ratio 2.63), preoperative intra-aortic balloon pump (odds ratio 3.09), raised serum creatinine in μmol/L (120-150, >150 vs. <120, odds ratio1.57, 11.24), and nonelective (odds ratio 3.43) and complex surgery (odds ratio 2.70). Model validation showed very good discrimination (area under the curve = 0.815) and calibration (c statistic = 8.527, p =.129). Conclusions: The fast-track failure score incorporates several preoperative factors and has been successfully internally validated; after undergoing external validation and possible recalibration it may be used as a tool to facilitate planning and flow of cardiac surgery patients, based on the predicted probability of failure. Application of this score may limit fast-track failure rates and help to reduce morbidity and cost.