Serologic Guidance for Extracorporeal Membrane Oxygenation Selection and Survival Using the Cosmos Dataset
Extracorporeal membrane oxygenation (ECMO) is a life-support therapy for patients with severe cardiac or respiratory failure. Despite its growing use, evidence-based guidelines for ECMO candidacy remain limited due to small, single-center studies and heterogeneous designs. This study leveraged Epic Cosmos, an electronic health record dataset encompassing over 300 million patient records, to analyze outcomes among 42,210 adults treated with ECMO between January 1, 2019, and February 15, 2026, involving 23,116 (54.8%) venoarterial (VA), 14,632 (34.7%) venovenous (VV), and 4,462 (10.6%) hybrid (VA and VV) cases. Hospital mortality varied from 37% (VV) to 50% (VA). Poisson models by ECMO type for in-hospital mortality adjusted for key demographic and clinical variables showed the admitting diagnosis group had the strongest association with subsequent mortality. Several blood chemistry markers measured within the 24 hours before ECMO initiation were also associated with increased mortality, including elevated lactate, elevated creatinine, and lower pH, though strength of association varied by ECMO type. Our findings support the use of large electronic health record (EHR) datasets for ECMO outcomes research and identify exploratory associations between preinitiation biomarkers and in-hospital mortality that require further validation.
Authors
- Luca Paoletti (ORCID: https://orcid.org/0000-0001-5122-6067)
- David Fitzgerald (ORCID: https://orcid.org/0000-0003-0980-7284)
- Kit N. Simpson (ORCID: https://orcid.org/0000-0002-1713-0632)
- Mary J. Dooley (ORCID: https://orcid.org/0000-0003-4454-8732)
- Ralph C Ward (ORCID: https://orcid.org/0000-0001-6193-2838)
- Alicha Gibbs
- Laura R Dell'Aiera
Institutions
- Medical University of South Carolina (US)
Publication Details
- Journal
- ASAIO Journal
- Published
- 2026-11-12
- DOI
- https://doi.org/10.1097/mat.0000000000002893
- Primary Topic
- Mechanical Circulatory Support Devices
- Type
- article
- Field-Weighted Citation Impact
- 0.00