A simple predictive model of prolonged intensive care unit stay after surgery for acquired heart valve disease.

BACKGROUND AND AIM OF THE STUDY: The study aim was to construct a simple model (the Fuwai risk score) to predict prolonged intensive care unit (ICU) stay after surgery to treat acquired heart valve disease. METHODS: Data were collected retrospectively from 2,218 consecutive patients who underwent surgery for acquired heart valve disease. Prolonged ICU stay was defined as > or =5 days. A simple logistic score was calculated using the logistic coefficient, and the additive score by odds ratio. The Fuwai risk score, EuroSCORE and Parsonnet score were applied to predict a prolonged ICU stay and mortality. A C statistic (receiver operating characteristic curve) was used to test discrimination of the models. Calibration was assessed with a Hosmer-Lemeshow goodness-of-fit statistic. RESULTS: The simple logistic model of the Fuwai risk score showed very good discriminatory ability (C statistic 0.76) and calibration (Hosmer-Lemeshow, p = 0.25) in predicting prolonged ICU stay, while the additive algorithm had good discriminatory ability (C statistic 0.75) but poor calibration (p <0.001). The additive algorithm greatly underestimated the risk for high-risk patients. The Fuwai risk score showed good discriminatory ability, but poor calibration in predicting mortality. Neither the EuroSCORE nor Parsonnet score was superior to the Fuwai risk score. CONCLUSION: The logistic algorithm of the Fuwai risk score is a simple, objective, convenient and accurate scoring system which may be used to predict prolonged ICU stay after surgery to treat acquired heart valve disease.