GRACE Score Calculator

The GRACE score converts eight bedside variables into an additive point total that maps onto a predicted probability of death after an acute coronary syndrome. The additive form is a practical approximation of a Cox or logistic model: each predictor contributes points roughly proportional to its regression coefficient, so age near 90 contributes about 100 points while elevated troponin contributes only 14. Three of the eight variables are haemodynamic. Heart rate and systolic pressure enter in opposite directions because tachycardia and hypotension both signal a failing pump, and Killip class adds up to 59 points because overt pulmonary oedema or cardiogenic shock dominates short-term prognosis. Serum creatinine is included as a marker of both chronic vascular burden and acute renal hypoperfusion. Totals run from roughly 0 to just over 370. The conventional bands are 108 or below for low risk, 109 to 140 for intermediate and above 140 for high risk, corresponding to in-hospital mortality of under 1 percent, 1 to 3 percent and over 3 percent respectively. In non-ST-elevation ACS these bands drive the timing of coronary angiography, with above 140 favouring an invasive approach within 24 hours. Discrimination is good, with a c-statistic near 0.83 in validation cohorts, but calibration is the weak point: the underlying registry ran from 1999 to 2009, and improvements in reperfusion, antiplatelet therapy and statin use mean current absolute mortality is generally lower than the original tables predict. The score also omits diabetes, ejection fraction and coronary anatomy.

The Global Registry of Acute Coronary Events was launched in 1999 as an observational, multinational registry of unselected patients admitted with a suspected acute coronary syndrome, eventually enrolling more than 100,000 patients from over 90 hospitals across 14 countries. Unlike randomised trial populations, GRACE deliberately included elderly patients, those with renal impairment and those who would have been excluded from trials, which is why the derived score generalises better than earlier trial-based models. Keith Fox, Christopher Granger, Joel Gore, Frans Van de Werf and colleagues led the analyses. The in-hospital death model was published in the Archives of Internal Medicine in 2003, and the widely used discharge-to-six-month model appeared in the BMJ in 2006. The score was adopted quickly into guidelines because it addressed a concrete decision rather than prognosis alone: which non-ST-elevation patients need early catheterisation. European Society of Cardiology guidelines on non-ST-elevation acute coronary syndromes have carried GRACE-based thresholds for invasive timing across successive editions, and it is generally preferred over the simpler TIMI risk score, which was derived from the TIMI 11B and ESSENCE trial populations and uses seven binary items. A recalibrated version, GRACE 2.0, was released in 2014 to extend predictions to one and three years and to model continuous variables with nonlinear terms rather than point bands, and later work has explored machine-learning successors trained on the same registry structure.