CT15
RISK STRATIFICATION MODELS FOR HEART VALVE SURGERY

Purpose  Risk stratification models may be useful in aiding surgical decision‐making, preoperative informed consent, quality assurance and healthcare management. While several overseas models exist, no model has been well‐validated for use in Australia. We aimed to assess the performance of two valve surgery risk stratification models in an Australian patient cohort.Method  The Society of Cardiothoracic Surgeons of Great Britain and Ireland (SCTS) and Northern New England (NNE) models were applied to all patients undergoing valvular heart surgery at St Vincent’s Hospital Melbourne and The Geelong Hospital between June 2001 and November 2006. Observed and predicted early mortalities were compared using the chi‐square test. Model discrimination was assessed by the area under the receiver operating characteristic (ROC) curve. Model calibration was tested by applying the chi‐square test to risk tertiles.Results  SCTS model (n = 1095) performed well. Observed mortality was 4.84%, expected mortality 6.64% (chi‐square p = 0.20). Model discrimination (area under ROC curve 0.835) and calibration was good (chi‐square p = 0.9). the NNE model (n = 1015) over‐predicted mortality. Observed mortality 4.83% and expected 7.54% (chi‐square p < 0.02). Model discrimination (area under ROC curve 0.835) and calibration was good (chi‐square p = 0.9).Conclusion  Both models showed good model discrimination and calibration. The NNE model over‐predicted early mortality whilst the SCTS model performed well in our cohort of patients. The SCTS model may be useful for use in Australia for risk stratification.