D. E. Go1, D. E. Abbott1, K. Wima1, D. J. Hanseman1, A. Ertle1, A. Chang1, J. J. Sussman1, S. A. Shah1, R. S. Hoehn1 1University Of Cincinnati,Department Of Surgery,Cincinnati, OH, USA
Introduction: Previous work has demonstrated significant outcome and cost disparities between hospitals with different safety-net burdens, though the implications of policy changes to limit operative interventions at specific hospital types are unclear. We aimed to understand how performance of pancreaticoduodectomy (PD) at different safety-net burden hospitals affects cost and outcomes, and how redistribution of patients from low performing hospitals to high performing hospitals could improve cost-effectiveness.
Methods: Hospitals performing PD were queried from the University HealthSystem Consortium database (UHC; 2009-2013) and grouped according to safety-net burden (proportion of Medicaid and uninsured patient charges, as previously described). A decision analytic model was then constructed, populated with UHC clinical and DRG-based cost data. Primary endpoints were perioperative mortality, readmission, and cost (index hospitalization plus readmission, when applicable). Sensitivity analyses were conducted to determine how primary endpoints were affected by alternative distribution of patients between hospital types and clinical outcomes.
Results: 15,090 patients populated the final dataset. Low (LBH), medium (MBH) and high (HBH) burden hospitals were comprised of 4,220 (28%), 9,505 (63%) and 1,365 (9%) patients, respectively. Perioperative mortality was twice as high at HBH (3.7%) than at LBH (1.6%) and MBH (1.7%) (p<0.001). In the base case, when all clinical and cost data were considered, PD at HBH hospitals cost $35,628/patient, 35% and 55% higher than MBH ($26,357) and LBH ($22988) hospitals, respectively. This increased cost at HBH hospitals was associated with a significantly higher readmission rate (23%) than at MBH (18%) and LBH (15%) hospitals (p<0.001. Patients at HBH hospitals were more likely to have extreme severity of illness (SOI) (18.8%) than at MBH (12.8%) and LBH (11.4%) but less likely to have moderate SOI (74.2% vs. 78% and 79.4%, respectively) (p<0.001). After patient SOI-adjusted equal redistribution of all HBH to LBH and MBH hospitals, per patient cost remained significantly lower at LBH ($25,594; 39.2% less) and MBH ($27,860; 27.9% less), with a readmission rate of 18% (vs. 23% at HBH in the base case).
Conclusion: HBH perform a minority of PDs, but have higher readmission and perioperative mortality rates at significantly higher costs. Redistribution of patients from HBH to LBH/MBH—adjusted for patient-specific risk—demonstrates significant improvement in clinical outcomes, with a potential for over $2M in annual cost savings. While patient-specific factors partially contribute to sub-optimal clinical outcomes, these data show that inherent hospital characteristics are important in optimizing cost-effective PD care, and that certain hospital types may not be optimally equipped for complex pancreatic surgery.