C19 A STATISTICAL MODEL TO IMPROVE HEART SCORE IN STRATIFYING THE RISK OF MAJOR ADVERSE CARDIAC EVENTS IN PATIENTS WITH CHEST PAIN

Abstract

Abstract Background Heart Score is a simple and reliable post–test score to predict chest pain patients at higher risk of major adverse cardiac events (MACE), but fails in 1.5 to 2.5% of cases. Aim of this study was to evaluate the effectiveness of the nomogram in improving Heart Score performance. Methods In this multicenter retrospective study we considered all the patients consecutively observed for chest pain from January to June 2021 in the Emergency Departments of three Northern Italian hospitals. Main outcome was the onset of MACE within a 3 months–follow–up. All data were carefully recorded and included in univariate analysis. The variables associated with outcome (significance level p<0.1) were subsequently evaluated in multivariate analysis by a Logistic Regression model and, when significant (accuracy level at least 0.5%), included in the final model, the nomogram. The statistical weight of each variable was numerically quantified and the total score corresponded to a risk probability. We further validate the model by internal bootstrap on 5000 patients re–sample. Validation and discrimination were assessed by the area under receiver operating characteristics curve (AUC). Results Overall 9837 patients (5863 males, 3974 females, mean age 63 ys) were eligible for the study. In the follow–up period 1671 patients (16.9%) developed a MACE (MACE+). Significant variables (p<0.001) after univariate analysis were: risk factors (median MACE+ 3 vs MACE– 1); Chest Pain Score (median 7 vs 4); lenght of chest pain (median 1 vs 4 hours), electrocardiogram findings and troponin levels (median 70 vs 5). All those were confirmed in the multivariate analysis (p<0.001): risk factors (OR 2.66); Chest Pain Score (OR 1.24); pain duration (OR 1.47); electrocardiogram findings (OR 1.51); troponin levels (OR 2.01). The Logistic Regression model reached a good likelihood level (R=0.685). The individual score of variables in the nomogram contributed to final score from 0 to 220, corresponding to a 3–months MACE risk rate (range 0.1–0.9). Discrimination level of nomogram (AUC 0.954) was higher than Heart Score (AUC 0.935) (p<0.005). In our series 1.6% of MACE+ were “low–risk patients” according to Heart Score, but those reported as low risk (risk rate < 0.2) in the nomogram were only the 0.2% (sensibility 99%, negative predictive value 99%). Conclusions In patients observed for chest pain nomogram achieved a more accurate stratification of low–risk ones than Heart Score.