Tom Kai Ming Wang1, David Hyun-Min Choi2, David Haydock2, Greg Gamble3, Ralph Stewart4, Peter Ruygrok4. 1. Green Lane Cardiovascular Service, Auckland City Hospital, Auckland, New Zealand. Electronic address: TWang@adhb.govt.nz. 2. Green Lane Cardiovascular Service, Auckland City Hospital, Auckland, New Zealand. 3. Department of Medicine, University of Auckland, Auckland, New Zealand. 4. Green Lane Cardiovascular Service, Auckland City Hospital, Auckland, New Zealand; Department of Medicine, University of Auckland, Auckland, New Zealand.
Abstract
BACKGROUND: Risk models play an important role in stratification of patients for cardiac surgery, but their prognostic utilities for post-operative complications are rarely studied. We compared the EuroSCORE, EuroSCORE II, Society of Thoracic Surgeon's (STS) Score and an Australasian model (Aus-AVR Score) for predicting morbidities after aortic valve replacement (AVR), and also evaluated seven STS complications models in this context. METHODS: We retrospectively calculated risk scores for 620 consecutive patients undergoing isolated AVR at Auckland City Hospital during 2005-2012, assessing their discrimination and calibration for post-operative complications. RESULTS: Amongst mortality scores, the EuroSCORE was the best at discriminating stroke (c-statistic 0.845); the EuroSCORE II at deep sternal wound infection (c=0.748); and the STS Score at composite morbidity or mortality (c=0.666), renal failure (c=0.634), ventilation>24 hours (c=0.732), return to theatre (c=0.577) and prolonged hospital stay >14 days post-operatively (c=0.707). The individual STS complications models had a marginally higher c-statistic (c=0.634-0.846) for all complications except mediastinitis, and had good calibration (Hosmer-Lemeshow test P-value 0.123-0.915) for all complications. CONCLUSION: The STS Score was best overall at discriminating post-operative complications and their composite for AVR. All STS complications models except for deep sternal wound infection had good discrimination and calibration for post-operative complications.
BACKGROUND: Risk models play an important role in stratification of patients for cardiac surgery, but their prognostic utilities for post-operative complications are rarely studied. We compared the EuroSCORE, EuroSCORE II, Society of Thoracic Surgeon's (STS) Score and an Australasian model (Aus-AVR Score) for predicting morbidities after aortic valve replacement (AVR), and also evaluated seven STS complications models in this context. METHODS: We retrospectively calculated risk scores for 620 consecutive patients undergoing isolated AVR at Auckland City Hospital during 2005-2012, assessing their discrimination and calibration for post-operative complications. RESULTS: Amongst mortality scores, the EuroSCORE was the best at discriminating stroke (c-statistic 0.845); the EuroSCORE II at deep sternal wound infection (c=0.748); and the STS Score at composite morbidity or mortality (c=0.666), renal failure (c=0.634), ventilation>24 hours (c=0.732), return to theatre (c=0.577) and prolonged hospital stay >14 days post-operatively (c=0.707). The individual STS complications models had a marginally higher c-statistic (c=0.634-0.846) for all complications except mediastinitis, and had good calibration (Hosmer-Lemeshow test P-value 0.123-0.915) for all complications. CONCLUSION: The STS Score was best overall at discriminating post-operative complications and their composite for AVR. All STS complications models except for deep sternal wound infection had good discrimination and calibration for post-operative complications.