Continuous variables were reported as medians and 25th and 75th percentiles, and categorical variables were reported as percentages. To test for independence of transfusion status and in-hospital outcomes, Wilcoxon rank sum tests were used for continuous variables and chi-square tests were used for categorical variables. Multivariate models were used to determine the factors associated with transfusion, and relationship between transfusion and in-hospital outcomes. For the first analysis, a stepwise approach, including a list of variables, was used to establish the factors that were associated with blood transfusion. For the second analysis, the model controlled for a standard list of factors. Candidate variables included in the model included patient demographics, (such as age, gender, body mass index, race), cardiac risk factors (such as family history of coronary artery disease, hypertension, diabetes, current/recent smoker, hypercholesterolemia), medical comorbidities (such as renal insufficiency, previous myocardial infarction [MI], previous percutaneous coronary intervention, previous CABG, previous CHF, previous stroke), presenting characteristics (such as ST-segment depression, ST-segment elevation, positive cardiac marker, signs of CHF at presentation, heart rate, systolic blood pressure), and socioeconomic status (such as insurance status). Because patients within a hospital are more likely to be similar, generalized estimating equations were used to adjust for correlations among clustered responses (i.e., within hospital correlations) (15).