Evaluation of the NICE Mini-GRACE Risk Scores for Acute Myocardial Infarction Using the Myocardial Ischaemia National Audit Project (MINAP) 2003-2009: National Institute for Cardiovascular Outcomes Research (NICOR)

Study Questions:

What is the performance of two variants of the Global Registry of Acute Coronary Events (GRACE) risk score—the mini-Global Registry of Acute Coronary Events (MG) and adjusted mini-GRACE (AMG) risk scores—in predicting 6-month mortality following acute myocardial infarction (AMI)? These revised risk scores may better reflect the data collected by national British registries such as the Myocardial Ischaemia National Audit Project (MINAP).


This was a retrospective observational study of patients discharged following AMI in acute hospitals in England and Wales. The AMG and MG variables were used to predict 6-month all-cause mortality risk using logistic regression with binomial distribution and log link. Compared to the GRACE variables, the MG risk score excludes creatinine and Killip class. The AMG risk score includes “prescription of a loop diuretic during admission” as a surrogate for Killip class and creatinine concentration. The main outcome measures were model performance indices of calibration accuracy (whether the observed mortality rates match the expected mortality rates), and discriminative and explanatory performance, including net reclassification index (NRI) and integrated discrimination improvement.


For AMI, AMG calibration was better than MG calibration (Hosmer-Lemeshow goodness of fit test: p = 0.33 vs. p < 0.05). Both MG and AMG risk scores had good predictive accuracy and discriminative ability (Brier score: 0.10 vs. 0.09; C-statistic: 0.92 and 0.84, respectively). Model performance was reduced in patients with non–ST-segment elevation MI, chronic heart failure, chronic renal failure, and the elderly (>85 years of age). Although performance indices were better for AMG, only 27.7% of cases of AMI had complete data to allow the calculation of the AMG score.


Two variants of the GRACE model—the MG and AMG scores derived from the MINAP data—perform well at predicting 6-month mortality following AMI in England and Wales, and may be relevant in this setting. Model performance is reduced in higher-risk subgroups.


The current study adds to the literature that indicates that surrogate data and risk scores that may account for missing GRACE predictors do not result in poor model performance in predicting mortality following AMI. While the AMG appears to have performed better, the amount of missing data may limit its use.

Keywords: Outcome Assessment (Health Care), Myocardial Infarction, Calibration, Binomial Distribution, Patient Discharge, England, Logistic Models

< Back to Listings