Purpose: There is considerable institutional variation in the treatment of high-risk congenital diaphragmatic hernia (CDH) requiring extracorporeal membrane oxygenation (ECMO), especially in regards to the timing of repair relative to ECMO. Recent studies have suggested that timing of repair is not related to in-hospital mortality. We aimed to determine if institutional strategies defined by the timing of repair were associated with differences in preoperative risks, operative management, oxygen requirements at discharge and presumed drivers of cost.
Methods: 45 CDH-study-group hospitals provided ECMO during 2007-2010. Centers were divided into three groups: (1) always repair on ECMO ("always," 11-centers,n=127) (2) sometimes repair on ECMO ("sometimes," 20-centers,n=271) and (3) never repair on ECMO ("never," 14-centers,n=76). All unrepaired patients requiring ECMO were analyzed. Known risk factors of mortality, treatment strategies and outcomes were compared between hospital groups. Multivariate linear regression was used to compare adjusted length-of-stay (LOS) and ECMO duration.
Results: 474 previously unrepaired patients required ECMO at one of the 45 study hospitals. On crude analysis, only birth-weight, repair-type, proportion repaired, center volume and ECMO duration varied significantly across hospital groups. In hospitals that "always" "sometimes" and "never" repair on ECMO, 95.3%, 77.9%, and 63.2% of ECMO patients were repaired (p<.001). Only 31.7% of ECMO patients at "sometimes" centers were actually repaired on ECMO. Hospitals that "sometimes" repair on ECMO waited almost twice as long to repair on ECMO compared to "always" centers (9.9 vs 5.5 days, p<0.001). Patch repairs ranged from 91.7% in the "always" hospitals to 81.3% in the "never" hospitals. Oxygen was required at discharge in 65%, 51.5% and 40% of patients in hospitals that "always", "sometimes" and "never" repair on ECMO (p=.04). Crude survival of ECMO patients was 52%, 55.4% and 46.1% respectively in the "always," "sometimes" and "never" hospitals (p=0.34). On multivariate analysis, patients at hospitals that "always" and "sometimes" repair on ECMO had a longer adjusted mean duration of ECMO (2.2 days 95%CI 1.1-3.4, p<.001 and 2.2 days 95%CI 1.0-3.3, p<.001 respectively) compared to those at hospitals that "never" repair on ECMO. Patients in hospitals that "sometimes" repair on ECMO had on average 8.4 (CI 95%1.1-15.7, p=0.02) fewer hospital days when compared to those at hospitals who either "always" or "never" repair on ECMO.
Conclusions: When centers were defined by their timing of repair, there were significant institutional differences in the type of repair, proportion repaired, oxygenation requirements at discharge, LOS and ECMO-duration. Patients treated at hospitals that varied in their repair strategy had the lowest mean number of hospital days. As previous studies have demonstrated that timing of repair does not affect mortality, these noted differences in operative management, secondary outcomes and presumed drivers of cost should encourage a multi-institutional trial to determine an optimal treatment strategy.