Related Experiment Video
Updated: Aug 6, 2026
Surfactant Depletion Combined with Injurious Ventilation Results in a Reproducible Model of the Acute Respiratory Distress Syndrome (ARDS)
Published on: April 7, 2021
Data-driven subphenotyping of severe ARDS patients requiring VV-ECMO
Micha Landoll1,2,3, Stephan Strassmann1, Wolfram Windisch1,4
1Department of Pneumology and Critical Care Medicine, ARDS and ECMO Centre, Cologne-Merheim Hospital, Cologne, Germany.
Background:
Acute respiratory distress syndrome is a heterogeneous syndrome that complicates risk stratification, therapy monitoring and personalised treatment during veno-venous extracorporeal membrane oxygenation. This study aims to identify discrete acute respiratory distress syndrome subphenotypes among veno-venous extracorporeal membrane oxygenation (VV-ECMO) patients using high-resolution electronic health record clustering, and to assess differences in clinical outcomes.
Materials And Methods:
We conducted a study of 598 adult patients with acute respiratory distress syndrome treated with VV-ECMO. Twenty-six clinically relevant parameters spanning inflammation, coagulation, kidney/liver function, mechanical ventilation, and ECMO parameters were analysed using K-means clustering. Shapley Additive Explanations models were used to identify key differentiating parameters between clusters and independent survival factors. Clinical outcomes were compared between clusters, including treatment duration and survival rates in intensive care.
Results:
Cluster analysis revealed distinct subphenotypes primarily driven by differences in inflammation (procalcitonin and C-reactive protein), kidney/liver function (creatinine and urea), coagulation (fibrinogen and D-dimer), and mechanical ventilation (positive end-expiratory pressure). Survival rates varied considerably between clusters, most notably within the kidney/liver function (32%) and combined parameter categories (21%). Subphenotypes defined solely by ECMO or ventilator settings showed smaller differences. Intensive care unit length of stay was longer in clusters with multi-organ dysfunction.
Conclusions:
Early data-driven clustering of electronic health record parameters identifies clinically meaningful acute respiratory distress syndrome subphenotypes among veno-venous extracorporeal membrane oxygenation patients. Renal, hepatic, and inflammatory dysfunctions are critical determinants of survival. Subphenotype-based stratification may refine risk stratification and management in severe acute respiratory distress syndrome treated with VV-ECMO.
