Abstract 4367162: Risk of Gastrointestinal Bleeding by Mechanical Circulatory Support Device

Introduction: Gastrointestinal bleeding (GIB) is a common, potentially modifiable, complication in patients receiving mechanical circulatory support (MCS) but head-to-head data for single and multi-device MCS strategies remains sparse. Research Question: To compare the incidence of GIB, transfusion burden, and in-hospital mortality across single- and multi-device MCS strategies using the MIMIC-IV critical care database. Methods: This retrospective study analyzed ICU encounters (2009-2022) of adults at Beth Israel Deaconess Medical Center with documented Extracorporeal Membranous Oxygen (ECMO), Impella, or Intra-aortic Balloon Pump (IABP) use. Data was obtained using the MIMIC-IV database with 85,000 ICU encounters. Exposures were categorized as single-device (ECMO, Impella, IABP) or multi-device combinations. The primary endpoint was any GIB (ICD-9/10 diagnoses). Secondary endpoints were packed red-blood-cell (pRBC) transfusion, hemorrhagic anemia, and in-hospital mortality. Categorical variables were compared using Fisher’s exact test; relative risks (RR) with 95% confidence intervals (CI) were calculated in comparison to IABP. Further analyses were stratified by support duration (<7 vs. ≥7 days). Results: A total of 1,404 MCS encounters were analyzed. Among 1,280 single-device ICU admissions (IABP=1,006, Impella=141, ECMO=133) GIB occurred in 5.1%, 5.0%, and 11%, respectively. ECMO had 2.2x higher risk vs. IABP (p<.01) and 2.3x vs. Impella (p=0.07). When limited to patients supported for <7 days, the risk of GIB on ECMO vs. IABP narrowed to 1.8 (95% CI 0.85-3.9), but the risk of need for pRBC transfusion or hemorrhagic anemia remained 2.4 (95% CI 2.0-2.9) and 1.6 (95% CI 1.2-2.0) times higher, respectively. pRBC units per transfused patient was likewise greatest at both durations with ECMO (6.9, 20.4) vs. Impella (2.9, 4.4) and IABP (2.5, 5.5). Multi-device cases showed higher crude bleeding (19–25%) but small sample size and heterogeneous timing or sequencing limit inference. In-hospital mortality likely reflected underlying severity of illness, with single-device ECMO and Impella both exceeding 50% vs.18% for IABP. Conclusions: Gastrointestinal bleeding and transfusion burden vary significantly by MCS strategy and duration, with ECMO carrying the greatest risk. Bleeding risk should factor into MCS selection. Prospective studies are needed to evaluate individualized bleeding mitigation protocols.