K. L. Bailey1, Y. Sanaiha1, E. Aguayo1, Y. Seo1, V. Dobaria1, R. J. Shemin1, P. Benharash1 1David Geffen School Of Medicine, University Of California At Los Angeles,Division Of Cardiac Surgery,Los Angeles, CA, USA
Introduction:
With calls for value-based healthcare delivery, high-risk procedures are increasingly being performed at more experienced centers. Recent studies produced conflicting associations of hospital volume and outcomes for critically-ill patients on extracorporeal membrane oxygenation (ECMO), with some suggesting higher mortality at high-volume centers. We aimed to describe the relationship of institutional volume and mortality in ECMO patients and assess causes of discrepant outcomes.
Methods:
Adult patients receiving ECMO from 2008-2014 were identified from the National Inpatient Sample (NIS). Volume was calculated as tertiles of total institutional discharges for each year independently. Statistical analyses included multivariable logistic regression and propensity matching to adjust for patient demographics and comorbidities as measured by the Elixhauser Index.
Results:
Of the estimated 18,684 ECMO patients, 14%, 28% and 58% were admitted to low-, medium-, and high-volume centers, respectively. Large centers had the greatest relative fraction of respiratory failure cases (45%), while medium and small centers commonly treated postcardiotomy syndrome (44% and 56%, respectively). Mortality at low-volume hospitals (43.7%) was lower compared to the rate at medium (50.3%, P=0.03) and high-volume (55.6%, P=0.002). On multivariate regression, high-volume, respiratory failure, and cardiogenic shock were predictors of mortality. Length of stay was shorter at low-volume hospitals than at medium- (14.8 vs 21.1 days, P<0.001) and high-volume (14.8 vs 25.2 days, P<0.001). Similarly, cost was lower at small centers ($142,803) compared to medium ($166,458, P<0.001) and large ($176,397, P<0.001). In sub-group analysis of high-volume institutions, 4,183 (39%) patients were transferred to the reporting hospital and 6,676 (62%) patients were initial admissions. The majority of transferred patients received ECMO for respiratory failure (56%). After propensity matching, the transferred cohort had higher mortality (58.5% vs 53.7%, P=0.045) and greater cost ($190,300 vs $168,970, P=0.009) compared to patients admitted directly to the index hospital.
Conclusion:
Our findings in this contemporary ECMO experience depict an association between high-volume institutions and greater mortality in the overall sample and in patients transferred to larger centers. Whether this phenomenon represents selection bias or transfer from another facility deserves further investigation and will aid with the identification of surrogate markers for quality of high-risk interventions. Improved selection criteria and the prediction of futile care are essential for the future growth of ECMO technology.
