Abstract Right ventricular (RV) injury occurs in about 21% of patients with acute respiratory distress syndrome (ARDS) and significantly increases mortality. While veno-venous extracorporeal membrane oxygenation (VV-ECMO) mitigates many of the factors causing RV injury, nearly 50% of ARDS patients on VV-ECMO paradoxically experience new or worsening RV injury, tripling their mortality risk. The mechanisms behind RV injury on VV-ECMO remain unclear. This study hypothesized that RV function on VV-ECMO is dynamic and influenced by factors beyond respiratory failure severity. We conducted a single-center retrospective study on adult COVID-19 ARDS patients treated with VV-ECMO from April 2020 to December 2021. Patients with at least one echocardiogram during VV-ECMO were included, excluding those initially on venoarterial ECMO or with left ventricular systolic dysfunction. Disease states were categorized as normal, RV injury, and RV failure. RV injury was defined by abnormal systolic function parameters or moderate RV dilation, and RV failure was characterized by RV injury with concurrent inotropic agent use. Data collected during each echocardiogram included vital signs, vasopressor use, intake/output balance, ventilator settings, and blood gas values. The primary outcome of the study was the probability of disease state transitions during VV-ECMO. To assess the probability of changes in RV function, we used a multi-state Markov model (MSM), which predicted state transitions at days 3, 7, 14, and 28 of VV-ECMO, with death and decannulation as competing states. Among 60 included patients, RV injury and failure were observed in 23-33% and 6.7-13% of patients, respectively, across time points. Mortality rate on VV-ECMO was 51.7%. 117 disease state transitions were observed. Patients with RV failure had a 60% chance of improving to RV injury, but a 40% chance of death. Patients with normal RV function were equally likely to transition to death or decannulation (37.5% each), with roughly 25% chance of transition to RV injury. Relationship between covariates and disease state transitions could not be assessed due to failure of model convergence due to low sample size. In conclusion, RV injury on VV-ECMO is a dynamic process, and patients frequently shift between states. Potential implications include recognizing the patient's position on the RV function spectrum and employment of mitigation strategies to prevent progression as well as providing prognostic clinical data. This study highlights the need for further research into the mechanisms behind RV injury on VV-ECMO and the development of strategies to mitigate these factors.