ABSTRACT
BACKGROUND: Venovenous extracorporeal membrane oxygenation (ECMO) has had encouraging evidence suggesting efficacy and acceptable safety in trauma patients with refractory respiratory failure. Given the obstacles of accruing adequate quality prospective data for a resource-intensive modality, it is unclear what is indicative of survival to discharge. We investigate pre-ECMO characteristics (age, Injury Severity Score [ISS], time from admission to cannulation, P:F ratio) in trauma patients to determine correlation with survival. METHODS: To address these challenges, we use Bayesian inference and patients from a level I trauma center and Extracorporeal Life Support Organization-designated Gold Center of Excellence (N = 12), published literature, and Markov chain Monte Carlo simulation to determine if there is strong predictive probability regarding survival to discharge. RESULTS: Bayesian inference probabilities expressed as odds ratios with 95% credible intervals (CrIs) were as follows: age (e = 0.981; CrI, 0.976-0.985), ISS (e = 0.996; CrI, 0.980-1.012), P:F ratio (e = 1.000; CrI, 0.996-1.003), and time from admission to ECMO (e = 0.988; CrI, 0.974-1.004). Bayes factors (BF) were as follows: BFage = 3.151, BFISS = 3.564 × 10, BFpf = 0.463, and BFtime = 913.758. CONCLUSION: Age was the only pre-ECMO factor that demonstrated the most certain effect on hospital mortality for trauma patients placed on venovenous ECMO. The ISS and time to ECMO initiation had some appreciable impact on survival although less certain than age; P:F ratio likely had none. However, the pre-ECMO factors that were found to have any impact on mortality were relatively diminutive. More studies are necessary to update prior distributions and enhance accuracy. LEVEL OF EVIDENCE: Prognostic, Level IV.
