Comparing the ability of GRACE 2.0 scores for predicting mortality in patients with and without diabetes

Abstract Background The GRACE 2.0 (G2) was an improvement on GRACE. Neither contains DM in the model. In the GRACE, DM independently predicted in-hospital but not the 6-month post-discharge mortality. Similar analysis for GRACE 2.0 score is unavailable. Purpose Assess the ability of the GRACE 2.0 score in predicting mortality in patients with and without diabetes. Methods Retrospective cohort analysis of data on non-STEMI and UA patients admitted in two hospitals between 2010 and 2019 entered into the Myocardial Infarction National Audit Project database. All baseline characteristics of the population and G2 scores were recorded. Death at 1 year was recorded from hospital and general practice records. Follow up was censored at the last recorded contact of the patient with a medical professional. The discriminative power of the G2 score in patients with and without diabetes in predicting mortality was determined by comparing the area under the receiver operating characteristic curve (c-statistics). Accuracy of calibration was evaluated by plotting the mean predicted vs observed mortality in each decile of predicted risk. Results Of 2841 patients entered, complete data was available for 2768 patients (age 69±14yrs, 62.3% male, 877(31.7%) with diabetes). G2 risk was higher in DM (8.3, IQR 3.9 to 19.0 v 5.5, IQR 2.6 to 13.0; p<0.0001). 337 patients died, 156 with DM. Higher deaths occurred in DM group (17.8% vs 16.3; adjusted OR 1.81, 95% CI 1.41-2.33, p<0.0001) (fig 1D). G2 discriminated between events well in whole cohort (AUC 0.81, p<0.001) with no difference between DM and non-DM patients (AUC DM: 0.78, 95% CI 0.75 to 0.81; non-DM 0.81, 0.79 to 0.83; δAUC 0.0301, p=0.242) (fig1E). Calibration of predictions by G2 was excellent across all deciles of risk for the whole cohort (fig 1A). However, G2 underestimated mortality in DM versus non-DM (fig1B,C), especially in the higher deciles of risk (fig 1 F,G). Conclusion(s) Discrimination of the GRACE 2.0 model was excellent for mortality at 1-year with no difference between patient with and without diabetes in our cohort of NSTEMI and UA patients. However, calibration plots suggest GRACE 2.0 underestimates mortality in diabetic compare to non-diabetic patients. This may suggest that risk stratification with GRACE 2.0 in diabetic patients may have to be considered with caution.Figure 1.