To the Editor:
We read with interest the article by Glance et al.1 comparing the three cardiac risk calculators recommended by the American College of Cardiology/American Heart Association Perioperative Guidelines,2 and we commend the authors for their study using a large sample from the National Surgical Quality Improvement Program. We published a similar comparison3 using a much smaller sample of patients seen and evaluated in our preoperative clinic rather than information from an administrative database and came to somewhat different conclusions—that all three calculators were similar in their classification of low versus elevated risk and performed reasonably well when used as intended but not as well when used to predict different outcomes in different timeframes. Fleisher’s editorial4 touched on issues of varying complexity and completeness of risk factors and outcomes assessed by the calculators. We feel it is important to further highlight differences in patients included, risk factor definitions, and outcomes and their timeframes, which may explain differences in performance of these tools.
Although the authors did note differences in complications studied, they did not mention that the Revised Cardiac Risk Index5 only included patients aged 50 yr and older with an expected in-hospital length of stay of at least 2 days or more. The myocardial infarction or cardiac arrest6 and American College of Surgeons National Surgical Quality Improvement Program7 calculators included patients aged 16 yr and above and many more surgical procedures, including low-risk and ambulatory surgeries. The Revised Cardiac Risk Index included more cardiac complications as outcomes with heart failure being the main addition and only included in-hospital complications rather than at 30 days postoperative as in the National Surgical Quality Improvement Program calculators. The Revised Cardiac Risk Index routinely screened patients for postoperative myocardial infarction with electrocardiograms and creatine phosphokinase, whereas the National Surgical Quality Improvement Program database had different definitions of myocardial infarction, used the more sensitive troponin, but did not screen routinely. Because of these differences, direct valid comparisons cannot really be made. The Revised Cardiac Risk Index can be improved by the addition of an age variable as well as by changing from serum creatinine more than 2 mg/dl to creatinine clearance less than 30 ml/min.8 In another study, as age increased, complications and positive predictive value also increased within each Revised Cardiac Risk Index class.9 A recently published geriatric-sensitive Revised Cardiac Risk Index10 performed better on patients over age 65 than the original Revised Cardiac Risk Index. The addition of angina within the past 6 months to only a history of myocardial infarction improved National Surgical Quality Improvement Program predictors.11
As noted by Glance et al., evaluation in real time allows the physician access to more information than the National Surgical Quality Improvement Program database. A complete history and physical can be performed, medical records including cardiac test results can be reviewed, medications are reviewed and changed as needed, and a better evaluation of the patient is likely compared with incomplete database records. Additionally, postoperative follow-up and surveillance may differ. Although we agree that there is wide variability in the performance of these calculators, much of it can be explained by the differences noted above. Because these risk predictions may alter physician behavior in terms of ordering further tests, it is important that these calculators are fully understood and used as intended. Further adjustments and comparison on a large scale in real time where more information can be obtained may help define whether one calculator is better than the others.
The authors declare no competing interests.