Current multivariate risk scores in patients undergoing non-cardiac surgery

Abstract

Several indexes to predict perioperative cardiovascular risk have been proposed overtime. The most widely used is the Revised Cardiac Risk Index (RCRI) developed by Lee since 1999. It predicts major cardiac outcomes from five independent clinical determinants: history of ischemic heart disease, history of cardiovascular disease, heart failure, insulin-dependent diabetes mellitus, and chronic renal failure (i.e. serum creatinine >2 mg/dl). In external validation studies, the RCRI showed high negative predictive value in all groups of age, indicating that it may be used to identify people at low risk for perioperative adverse cardiovascular events in noncardiac surgery. However its accuracy is suboptimal in many clinical settings. More recently the National Surgical Quality Improvement Program database) (NSQIP) hasdeveloped a new index to predict perioperative myocardial infarction (MI) or cardiac arrest (MICA) from a cohort of 211,410 patients (the Gupta index) and afterwards a universal surgical risk estimation tool has been developed, using standardized clinical data from 393 ACSNSQIP hospitals in US (a cohort based on 1,414,006 patients), showing a good performance for mortality (C-statistic = 0.944) and morbidity (C-statistic =0.816) as compared with procedure-specific models. Other risk scores include the Vascular events In noncardiac Surgery patIents cOhort evaluatioN (VISION), which has evaluated cardiac complications in 15,065 patients, the Physiological and Operative Severity Score for the enUmeration of Mortality and Morbidity (POSSUM) and the large Preoperative Score to Predict Postoperative Mortality (POSPOM) that was built up from data collected in the National Hospital Discharge Data Base (NHDBB) including a cohort of 7.059.447 patients. In Italy a new risk index (the Orion score) builkt up from a cohort of 9000 patients generated four classes of major cardiovascular adverse events perioperative risk ranging from 1 (0.6%); 2 (2.4%); 3 (7.4%) and 4 (23.1%). The AUROC curves showed higher accuracy as compared to the RCRI score both in the derivation than in the validation cohort (AUROC= 0.872 ± 0.028 vs 0.807 ± 0.037). Thus, many risk indices are available nowadays. Despite the latest European guidelines recommended them for risk stratification (class I, level of evidence B), their use in clinical practice is still scarce.