Heterogenous treatment effects of blood transfusion in hospitalized patients with congestive heart failure

Background: Anemia is nearly ubiquitous in hospitalized patients with congestive heart failure (CHF), yet little data informs the decision to transfuse blood in this population. Objectives: To determine average and heterogenous effects of blood transfusion in hospitalized patients with CHF. Methods: We performed a multicenter retrospective cohort study with individual treatment effect analysis using the Premier Healthcare Database (2022-2024). Adult patients with CHF hemoglobin concentrations between 6.5-7.4 g/dL were included. The exposure was blood transfusion based on hemoglobin concentration threshold of <7.0 g/dL. The primary outcome was hospital free days by day 28 (HFDs). We determined the average effect of transfusion using instrumental variable analysis based on a hemoglobin threshold of 7.0 g/dL and estimated the predicted effects of transfusions on HFDs (i.e., conditional average treatment effects) for individual patients using causal forest machine learning models. Results: We included 31,408 patients in a derivation cohort and 30,677 in a validation cohort, of which 13,437 (42.8%) and 13,334 (43.5%) received transfusions, respectively. The average association between transfusion and HFDs suggested harm (derivation mean difference -1.8 [95% CI -2.3, -1.3] days; validation mean difference -1.5 [95% CI -2.0, -1.0] days). However, the effects of transfusion were heterogenous (p=0.001) with the strongest drivers of transfusion benefit being transfusion on hospital day 1 and low serum bicarbonate concentration. Conclusions: The average association between transfusion and HFDs in hospitalized patients with CHF suggested harm; however, there were potential small benefits early in hospitalization and in those with low serum bicarbonate concentrations.