Abstract P73: Comparison of Hospital-Acquired Anemia and TIMI Bleeding for Mortality Prediction in Patients with Acute Myocardial Infarction

Abstract

Background: In-hospital bleeding and new onset, hospital acquired anemia (HAA) are both associated with higher mortality in acute myocardial infarction (AMI). Since bleeding is variably defined and often poorly documented, HAA could be a better method to identify at-risk patients, if its prognostic ability were at least as good as documented bleeding. We directly compared the association of HAA and TIMI bleeding with 1-year mortality. Methods: Among 2,803 AMI patients who were not anemic at admission in the 24-center TRIUMPH registry, the presence and severity of HAA and TIMI bleeding were prospectively collected to identify their relative discrimination of 1-year mortality. Logistic regression models, accounting for clustering using generalized estimating equations, were fit for 1) no bleeding, TIMI minimal, minor and major bleeding and 2) no HAA, mild (hemoglobin (Hgb) > 11 g/dl), moderate (Hgb 9 - 11 g/dl) and severe HAA (Hgb < 9 g/dl). Discrimination was compared using c-statistics and reclassification was assessed using the integrated discrimination improvement (IDI), which measures a model's improvement in average sensitivity without sacrificing average specificity vs. another model, and the continuous net reclassification improvement (NRI), to identify the proportion of patients correctly reclassified by the HAA model. Results: HAA was more common (mild: 33%, moderate: 10%, severe 2%) than TIMI bleeding (minimal: 5%, minor: 3%, major 1%). Over 1-year follow-up, 111 patients (4%) died. The HAA model was superior to TIMI bleeding model for 1-year mortality prediction (c-statistic 0.60 vs. 0.51, p<0.001). The IDI of the HAA vs. the bleeding model was 0.009 (95% CI 0.005 - 0.014) and the relative IDI was 0.26 (26% better average discrimination), with a NRI of 0.32 (0.13-0.50) - 17% of patients with events were correctly reclassified to a higher risk while 14% of patients without events were correctly reclassified to a lower risk by the HAA model. Conclusions: HAA is better than TIMI bleeding for identifying 1-year mortality after AMI hospitalization, and may better identify patients without recognized bleeding who are also at risk for poor outcomes. HAA may be useful to identify high-risk patients and as a quality assessment tool.