Figure 2.

Derivation of a propensity score to test the robustness of association of baseline hemoglobin concentration with study outcomes. (A) Sequential logistic regression models were used to ensure the balance of the predictors of baseline anemia across groups of patients with and without anemia. The standardized differences before and after adjustment are shown in blue and red circles, respectively. All the variables were adequately balanced after adjustment. (B) Kernel density of the propensity score across patients with and without anemia.