Purpose

Prolonged ECMO has become more common during the COVID-19 pandemic but is associated with high resource utilization and poor outcomes. This study aimed to identify predictors of prolonged ECMO and explore prognostic indicators in patients who underwent prolonged ECMO.

Methods

This multicenter retrospective study analyzed patients who underwent venovenous ECMO for severe COVID-19 at three high-volume ECMO centers in Japan from January 2020 to December 2021. Patients requiring ECMO for ≥ 21 days were classified as the prolonged ECMO group. The study was structured in four steps: [1] comparison of baseline characteristics between survivors and non-survivors [2], identification of pre-ECMO predictive factors for prolonged ECMO [3], determination of prognostic factors among prolonged ECMO patients, and [4] comparison with established prognostic scoring systems.

Results

Among 121 patients, 32 (26%) required prolonged ECMO. Lower positive end-expiratory pressure (PEEP) before ECMO was identified as an independent predictor of prolonged ECMO (P < 0.001), with an optimal cutoff of 12 cmH₂O (area under the curve [AUC]: 0.70). Among prolonged ECMO patients, the Sequential Organ Failure Assessment (SOFA) score on ECMO day 21 was the only independent predictor of in-hospital survival (P = 0.002), with an optimal cutoff value of 12 (AUC: 0.82). The SOFA score on day 21 outperformed established prognostic scoring systems.

Conclusion

Lower PEEP before ECMO is a predictor of prolonged ECMO. In patients undergoing prolonged ECMO, the SOFA score on day 21 is an independent predictor of survival. Continuous assessment of organ dysfunction during ECMO may enhance prognostic evaluation and support clinical decision-making.