Background: The National Surgical Quality Improvement Program Myocardial Infarction & Cardiac Arrest (NSQIP MICA) calculator and the Revised Cardiac Risk Index (RCRI) were derived using currently outdated methods of diagnosing perioperative myocardial infarctions. We tested the external validity of these tools in a setting of a systematic perioperative cardiac biomarker measurement.
Methods: Analysis of routinely collected data nested in the Vascular Events In Noncardiac Surgery Patients Cohort Evaluation Study. A consecutive sample of patients ≥45 yr old undergoing in-hospital noncardiac surgery in a single tertiary care centre was enrolled. The predictive performance of the models was tested in terms of the occurrence of major cardiac complications defined as a composite of a nonfatal myocardial infarction, a nonfatal cardiac arrest, or a cardiac death within 30 days after surgery. The plasma concentration of high-sensitivity troponin T was measured before surgery, 6-12 h after operation, and on the first, second, and third days after surgery. Myocardial infarction was diagnosed according to the Third Universal Definition.
Results: The median age was 65 (59-72) yr, and 704/870 (80.9%) subjects were male. The primary outcome occurred in 76/870 (8.7%; 95% confidence interval [CI], 6.9-10.8%) patients. The c-statistic was 0.64 (95% CI, 0.57-0.70) for the NSQIP MICA and 0.60 (95% CI, 0.54-0.65) for the RCRI. Predicted risks were systematically underestimated in calibration belts (P<0.001). The RCRI and the NSQIP MICA showed no clinical utility before recalibration.
Conclusions: The NSQIP and RCRI models had limited predictive performance in this at-risk population. The recently updated version of the RCRI was more reliable than the original index.
Keywords: cardiac arrest; myocardial infarction; perioperative care; risk assessment; vascular surgical procedures.
Copyright © 2019 British Journal of Anaesthesia. Published by Elsevier Ltd. All rights reserved.