Objective

Many multinational risk-scoring systems have been produced for cardiac surgery. Current research is suggesting that smaller individual risk scores, modelled upon certain populations are more accurate when applied to those populations.

Methods

Over 10,000 patients were included who had heart surgery from 2001-2011. The primary analysis was performed using the forward selection multiple logistic regression method. This allowed the formulation of a model, which could be then used to predict percentage mortality for patients. The data was split into a working and a testing sub-set. This allowed the model created to be validated using a dataset different from the one used to create it. The primary outcome was status at discharge.

Results

Comparing the profile of the patients with that of the recently created EuroSCORE II risk score, the Sheffield patients had a lower mortality (2.3% versus 4.8%), and generally better risk profiles at face value. (81.7% non-diabetic versus 75%; active endocarditis 1% versus 2.2%; 0.5% on dialysis versus 1.1%). SheffScore was found to have good calibration and discrimination during its testing, with an area under the ROC curve of 0.826, superior to that of any risk score available to date.

Conclusion

Over the past ten years, mortality from cardiac surgery has reduced, despite the operative risk profile of patients becoming greater. SheffScore is a well-calibrated risk stratification tool that can help to assess surgical outcomes, and also provide patients with a quantifiable risk of operative mortality. SheffScore is the first risk score produced from a single UK centre, and is recommended to be implemented throughout the UK for risk stratification of all adult cardiac surgery.