Extracorporeal membrane oxygenation (ECMO) for adult respiratory failure has significantly increased, with longer duration ECMO support required in severe hypoxemia. We sought to examine independent predictors of survival of adult respiratory failure patients requiring prolonged (≥14 days) ECMO. We reviewed Extracorporeal Life Support Organization Registry data on all adult (≥18 years) patients who required P- ECMO (n = 4,361) over 10 years (2009–2018). Hospital survival was 51.3%, increased from 45.4% in our prior report of 974 patients (1989–2013). Univariate analysis confirmed factors associated with decreased mortality: younger age, white race, increased body weight, viral/bacterial pneumonia, higher positive end expiratory pressure, neuromuscular blockade, VV-ECMO mode, and decreased time from intubation to ECMO. For Pre-ECLS support, most vasopressor/inotropic drugs and nitric oxide had no association with mortality, but steroids (22% vs. 15%, p < 0.001), epinephrine (15% vs. 12%, p = 0.039), and bicarbonate (9% vs. 7%, p = 0.049) were more common in non-survivors. Extracorporeal membrane oxygenation complications (gastrointestinal hemorrhage, neurologic complications, and CPR) were associated with increased mortality. The RESP score was higher in survivors (−0.31 ± 3.36 vs. −0.83 ± 3.34, P < 0.001); however, discrimination was poor (c-statistic = 0.540 ± 0.009); it did not remain in the final model. A multivariable prediction model based on all information at ECMO initiation was fair (c-statistic = 0.670 + 0.012), but discrimination improved with the addition of ECMO complications (c-statistic = 0.675 + 0.012). These findings suggest that reducing ECMO-related complications will improve survival. We have identified predictors of mortality in prolonged ECMO patients, and inclusion of ECMO complications in a new predictive model improved discrimination.