A new EuroSCORE or something better?

The article by Miceli et al published in this issue of Heart (see page 362) describes a study designed ‘to assess the potential contribution of mild renal dysfunction as a predictor of operative mortality (after cardiac surgery) and then develop and validate a modified EuroSCORE model which would take this variable into account.’1 The authors note that patients with mild renal failure have worse outcomes after coronary surgery than those with normal renal function, and that the current EuroSCORE risk algorithm only attributes increased risk for those patients with a creatinine greater than 200. Using multivariate analysis, they show that preoperative mild renal dysfunction is an independent risk factor for mortality with an OR of 1.8. They conclude that the original EuroSCORE was able to predict mortality in their patients, but their modified logistic EuroSCORE had signifi-cantly better discriminatory power between patients of lower and higher risk. The finding of mild renal failure being associated with worse in-hospital mortality after cardiac surgery is not new, and it has also now been well established that the logistic EuroSCORE no longer accurately predicts hospital mortality.2e8 However, the limitation of existing risk models and what needs to be done about them are worthy of further consideration. The first cardiac surgery risk model to find widespread use was the Parsonnet score.9 This was produced from data collected in the USA in the 1980s, and the model was ‘additive ’ in that each risk factor (age, gender, left ventricle function, etc) was attributed a score, and these scores could be summed to give an indi-cation of the likely mortality for that patient. The ‘performance ’ of the score was initially thought to be good. We now accept that this performance should be assessed in three ways