Five RCRI details that move a patient a whole class

The RCRI counts six preoperative predictors and reports how often major cardiac complications occurred in patients with that many predictors, after nonemergent major noncardiac surgery. The figure is an observed group rate for a defined endpoint, not an individual probability. Lee and colleagues derived it from 4,315 patients aged 50 or older having nonemergent major noncardiac surgery with an expected hospital stay of at least two days, at a single tertiary teaching hospital between July 1989 and February 1994. The index was built in a derivation cohort of 2,893 patients and tested prospectively in a validation cohort of 1,422.¹

The outcome is a composite of four events during the index hospital admission: myocardial infarction, pulmonary oedema, ventricular fibrillation or primary cardiac arrest, and complete heart block. Cardiac death is not one of the four; the paper reports it separately.¹

It is a points model. Each of six predictors adds one point, so the score is an integer from 0 to 6. The authors tested weighting the predictors differently and found no demonstrated gain in discrimination, so every item counts the same.¹ Scores of 3 and above are grouped together: no primary source publishes a separate rate for 4, 5 or 6.

Source: Lee 1999¹.

Ischaemic heart disease and heart failure each bundle several sub-criteria into one yes or no. A patient who meets three of the five ischaemic heart disease items still gets one point. The classes are the derivation paper's own labels. Lee 1999 did not attach "low risk" to Classes I or II, and Curie Calc shows the class without an adjective.¹

Source: Lee 1999¹. Definitions as signed off at clinical review of the Curie Calc RCRI calculator, October 2026.

Does diabetes count, or only insulin?

Only insulin. The criterion is "preoperative treatment with insulin", not a diagnosis of diabetes.¹ A patient whose diabetes is managed with diet or oral agents scores 0 on this item. Reading it as "any diabetes" adds a point for every diabetic patient not on insulin, which moves the patient up a class whenever the correct total is 0, 1 or 2.

Does a past bypass or stent count as ischaemic heart disease?

Not on its own. The ischaemic heart disease item is met by any one of five findings: prior myocardial infarction, a positive exercise test, current ischaemic chest pain, current nitrate therapy, or pathological Q waves.¹ A coronary bypass or angioplasty is not one of the five. A revascularised patient scores the point if one of the five findings is also present, such as a prior infarction, and scores 0 if none is.

Is a creatinine of exactly 2.0 mg/dL a point?

Not under the original definition. Lee 1999 scores creatinine only when it is greater than 2.0 mg/dL (176.8 µmol/L).¹ The 2022 ESC guideline restates the criterion as 2.0 mg/dL or above.² For a patient at exactly 2.0, the two wordings give different scores.

Curie Calc applies Lee's strict "greater than", because the rates it displays come from Lee's cohorts, and it shows a note whenever creatinine is between 1.9 and 2.1 mg/dL so the boundary is visible. Units matter at the edge too: 176.8 µmol/L converts to exactly 2.0 mg/dL and does not score; 177 µmol/L converts to just above it and does.

The item asks about a serum creatinine concentration. An eGFR or creatinine clearance answers a different question and cannot be substituted for it.

Which vascular surgery counts as high risk?

Suprainguinal vascular surgery. Lee's high-risk category is intraperitoneal, intrathoracic, or suprainguinal vascular surgery.¹ Carotid endarterectomy and infrainguinal procedures such as femoral-popliteal bypass do not count. Some later restatements shorten the category to "vascular" surgery, which reads more broadly than the original.³

There is a further caution for vascular patients. Within the derivation paper's own subgroup of abdominal aortic aneurysm repair, risk class showed no relationship to major cardiac complications, and across pooled studies the index discriminates less well in vascular surgery (AUC 0.64) than in mixed noncardiac surgery (AUC 0.75).¹⁶

Does carotid stenosis count as cerebrovascular disease?

Only if there has been a TIA or stroke. The cerebrovascular item is a history of TIA or stroke.¹ Carotid stenosis found on imaging, without a prior event, does not score. A TIA does count, with no look-back period stated.

This item is not the same question as the stroke items in other scores. CHA₂DS₂-VASc also counts systemic thromboembolism, and HAS-BLED's stroke item is a bleeding-risk marker with its own wording, so an answer recorded for one score should not be copied into another without checking. Read the CHA₂DS₂-VASc guide.

The number beside an RCRI class depends on which endpoint is counted, over what time window, and in which cohort. The same score has been published with very different risk figures, and each describes its own endpoint and cohort. At a score of 0, published figures run from 0.4% for Lee's in-hospital composite to 8.2% for VISION's 30-day composite including myocardial injury.¹⁵

Rates in different rows count different endpoints and should not be compared as if they measured the same thing. Each is an observed group average in its own cohort, not an individual forecast.

Three things drive the spread:

• How myocardial infarction was detected. Lee's cohort identified infarction by CK-MB. Later cohorts used routine troponin surveillance, which finds more events. This is why the Canadian Cardiovascular Society recalibrated its risk communication around later studies rather than the 1999 rates.⁴
• What the endpoint includes. Adding cardiac death, extending the window to 30 days, or counting myocardial injury after surgery each raises the rate. In VISION, 35% of events on the 30-day composite including myocardial injury occurred in patients with a score of 0.⁵
• Who was in the cohort. Performance varies across surgical populations. The index under-predicted cardiac complications in vascular surgery⁷ and over-predicted in patients with kidney failure.⁸

Why Curie Calc shows Lee's validation rates. They are the derivation paper's own prospectively validated figures, with a precisely defined endpoint and published counts for every class. Clinical review chose them on the condition that the endpoint and the era are always stated beside the number, which the calculator does on every result.

What the guidelines say. The 2024 ACC/AHA perioperative guideline defines elevated risk as a predicted risk of major adverse cardiovascular events of 1% or more, and notes that an RCRI above 1 has traditionally been used as the threshold, with the determination depending on the calculator used.³ In Lee's validation cohort, the observed rate of the in-hospital composite was 0.9% at a score of 1 and 6.6% at a score of 2. The two are different measures: the guideline's 1% refers to a predicted risk of major adverse cardiovascular events, not to Lee's composite. The 2022 ESC guideline lists the RCRI alongside other risk scores and does not recommend one specific score.² The 2017 CCS guideline states that its panel favoured the RCRI among available indices.⁴

Synthetic cases. Any criterion not mentioned is absent. Each result was checked on the live Curie Calc RCRI calculator.

Try these cases in the RCRI calculator

The RCRI was derived in adults aged 50 or older having nonemergent major noncardiac surgery with an expected stay of at least two days. It has no role in emergency surgery or cardiac surgery, and minor or day-case surgery, an expected stay under two days, and patients under 50 fall outside the cohort it was built on. Patients with active unstable disease, such as unstable angina, class IV heart failure, an active TIA or critical aortic stenosis, were poorly represented.¹ Performance is weaker in vascular surgery, and in patients on chronic dialysis or with a kidney transplant the creatinine criterion was not derived for them and the index over-predicts.⁶⁸ Curie Calc records dialysis or transplant as an optional question that scores nothing and raises a note.

Several factors that matter clinically are not in the score: age beyond the entry criterion, functional capacity, valvular disease including critical aortic stenosis, pulmonary hypertension, arrhythmia, frailty, anaemia, emergency status, and biomarkers such as BNP or troponin. Lee found that age, critical aortic stenosis and rhythm abnormalities were not retained by the model and cautioned that this may reflect who was in the cohort. Their absence is not evidence that they do not matter.¹

The index also does not predict mortality well: in a systematic review of 24 studies, the six that reported death had a median AUC of 0.62.⁶