Background and purpose
Veno-arterial extracorporeal membrane oxygenation (VA-ECMO) provides temporary circulatory and gas exchange support for patients with refractory cardiogenic shock when conventional therapy is insufficient. Indication and timing decisions require weighing expected benefit against procedural risk, resource use, and patient goals. The SAVE (Survival after Veno-Arterial ECMO) score was developed to stratify probability of survival to hospital discharge among adults treated with VA-ECMO for cardiogenic shock, using variables typically available around the time of cannulation.
The score was derived from a large international cohort in the Extracorporeal Life Support Organization (ELSO) registry. It is intended as a prognostic adjunct for communication, risk discussion, and research—not as a stand-alone triage rule. It does not estimate neurologic outcome, quality of life, or long-term functional status after hospital discharge.
Population and scope
The SAVE score applies to the context of its derivation: veno-arterial (VA) ECMO for cardiogenic shock. It should not be applied to veno-venous ECMO for isolated respiratory failure, nor extrapolated to populations that were not represented in the original model (for example, many protocols for extracorporeal cardiopulmonary resuscitation include patients who were not part of the same derivation framework).
When using the calculator, align inputs with the worst clinically relevant values in the window specified for each element (often within several hours before cannulation), consistent with how the original variables were intended to reflect pre-ECMO physiology and organ dysfunction.
Score structure
The model combines sixteen weighted items that capture demographics, shock etiology, renal and acid–base status, pre-ECMO ventilation, hemodynamics, and other organ failures. After all applicable points are summed, a constant of −6 is subtracted from every patient. The resulting SAVE score typically falls in a published range of roughly −35 to +17, with higher values associated with higher observed rates of in-hospital survival in the derivation cohort.
Demographics and body size
Age is grouped into bands reflecting the nonlinear relationship between advancing age and outcome in this critically ill population. Weight is categorized into three groups; body size may relate to reservoir tolerance, nutrition, and comorbidity patterns in registry data, though the precise mechanisms are not specified by the score itself.
Etiology of cardiogenic shock
Points reflect the primary shock etiology when it matches prespecified categories such as myocarditis, refractory ventricular tachycardia or fibrillation, post–heart or lung transplantation, or congenital heart disease. When shock is due to other or mixed mechanisms, no etiology points are assigned in the usual implementation—mirroring the idea that the listed categories capture distinct risk profiles in the registry.
Renal function and metabolic stress
Acute renal failure (for example, creatinine elevation beyond a defined threshold, with or without renal replacement therapy) and chronic kidney disease (reduced estimated glomerular filtration rate sustained over months) both contribute negative weight, reflecting multiorgan failure and baseline vulnerability. Severe metabolic acidosis is captured by a low serum bicarbonate threshold measured before ECMO, signaling profound circulatory compromise and tissue hypoperfusion.
Respiratory support before ECMO
Duration of mechanical ventilation prior to ECMO initiation is scored in three bands: shorter support, intermediate, and prolonged. Longer pre-ECMO ventilation often tracks illness severity, atelectasis, infection risk, and ventilator-induced lung injury. A lower peak inspiratory pressure threshold assigns favorable weight, consistent with less injurious ventilation or relatively preserved respiratory mechanics in some presentations—interpretation should remain physiology-based rather than automatic.
Cardiac arrest and hemodynamics
Pre-ECMO cardiac arrest is a negative contributor, reflecting the incremental risk carried by collapse before mechanical support. Diastolic blood pressure and pulse pressure encode vascular tone and cardiac output coupling in the pre-cannulation window: preserved diastolic pressure and wider pulse pressure associate with more favorable points, while very narrow pulse pressure marks low cardiac output states with higher observed risk.
Hepatic and central nervous system dysfunction
Liver injury may be reflected by hyperbilirubinemia and/or transaminase elevations beyond defined cutoffs, indicating hypoxic or congestive hepatopathy in the shock state. Central nervous system dysfunction includes conditions such as stroke, encephalopathy, or seizures that compound morbidity and may interact with anticoagulation and perfusion goals during ECMO.
Risk stratification and interpretation
The continuous SAVE score is grouped into five risk classes (I through V) that map to progressively lower approximate rates of in-hospital survival in the derivation cohort. The highest class corresponds to the strongest observed survival association; the lowest class corresponds to the weakest. These estimates are population-level and will not precisely predict any single patient’s outcome.
Clinicians should integrate the score with bedside examination, echocardiography, lactate trends, end-organ perfusion, bleeding risk, anticipated recovery trajectory of the myocardium or surgical options, and patient values. A higher score supports a more optimistic discussion of short-term survival probability in the original data; a lower score supports transparent discussion of higher expected hospital mortality, without dictating a specific management pathway.
Limitations and appropriate use
Performance may differ across eras, centers, and selection practices for ECMO. The SAVE score addresses hospital mortality only; it must not be read as a measure of cognitive or functional recovery. Any use for quality metrics or benchmarking should account for case-mix and unmeasured confounders. The tool is best used to structure prognosis discussions and documentation alongside—not instead of—multidisciplinary ECMO team judgment.