RCRI calculator (Revised Cardiac Risk Index)

How this RCRI calculator works

This RCRI calculator adds 1 point for each of six preoperative predictors that Lee and colleagues identified in Circulation in 1999, so the Revised Cardiac Risk Index runs from 0 to 6 points. The index was derived and validated in 4,315 patients aged 50 years or older who had elective major noncardiac surgery, with an expected stay of at least 2 days, at Brigham and Women's Hospital in Boston between 1989 and 1994. Two thirds of them, 2,893 patients, formed the derivation cohort and the other 1,422 the validation cohort. Logistic regression found six independent predictors of major cardiac complications, with adjusted odds ratios from 1.9 to 3.0. A version that weighted each factor by its regression coefficient discriminated no better than one that gave every factor 1 point (ROC area 0.765 versus 0.759), so the authors kept equal weights. Major cardiac complications meant myocardial infarction, pulmonary edema, ventricular fibrillation or primary cardiac arrest, and complete heart block.

The six predictors as Lee 1999 defined them:

Lee and colleagues tested several definitions of ischemic heart disease and kept the one most strongly linked with complications. Prior revascularization alone was left out because complication rates were the same with or without it: 2% after angioplasty and 3% after bypass surgery, against 2% in everyone else.

RCRI score results and what they mean

An RCRI score result places the patient in one of four classes, and each class carries an observed rate of major cardiac complications. In Lee's 1999 paper, Table 3, the rates in the derivation cohort were 0.5% for class I, 1.3% for class II, 3.6% for class III and 9.1% for class IV. In the prospective validation cohort of 1,422 patients they were 0.4% (2 of 488), 0.9% (5 of 567), 6.6% (17 of 258) and 11.0% (12 of 109). The ROC area was 0.759 in derivation and 0.806 in validation, better than the Original and Modified Cardiac Risk Indexes and ASA class. Those rates came from a single hospital that detected myocardial infarction with CK-MB. The 2017 Canadian Cardiovascular Society (CCS) guideline pooled later external validation studies that measured troponin after surgery and found higher risks of myocardial infarction, cardiac arrest or death at 30 days, shown in the last column.

A worked example: a 67-year-old with a myocardial infarction 4 years ago, diabetes treated with insulin and a creatinine of 1.3 mg/dL, booked for an open colectomy. Intraperitoneal surgery, ischemic heart disease and insulin give 3 points, class IV: 11.0% in Lee's validation cohort and 15.0% in the CCS pooled estimate. The CCS guideline also uses an RCRI of 1 or more, alongside age and known cardiovascular disease, to pick the patients who should have BNP or NT-proBNP measured before surgery.

When the RCRI is less reliable

The RCRI is less reliable outside the patients it was built on, and its authors said so. Lee's cohort came from one teaching hospital, included only patients aged 50 or older having nonemergent major surgery, and the paper calls the index of uncertain generalizability for minor procedures and for emergency operations. There was no relationship between RCRI class and complications after abdominal aortic aneurysm repair, where the ROC area was 0.543. In the validation cohort, insulin treatment and creatinine above 2.0 mg/dL were no longer independent predictors. Ford and colleagues' 2010 systematic review of 24 studies and 792,740 patients found the RCRI discriminated moderately well after mixed noncardiac surgery (pooled AUC 0.75) but poorly after vascular surgery (AUC 0.64), and it did not predict death well. Finally, the original rates rest on CK-MB, so they understate the myocardial injury that routine troponin testing detects, which is why the CCS guideline's pooled risks from troponin-monitored studies are higher.

The RCRI also leaves out factors that matter but were rare in the cohort, such as critical aortic stenosis, which only 5 patients in the derivation cohort had. For how the RCRI compares with the newer NSQIP-based calculator, read RCRI vs Gupta.