Hospital quality and payment variation for pancreaticoduodenectomy in michigan

Abstract

Presenter: Adrian Diaz MD, MPH | University of Michigan Background: Previous studies have documented cost variation for inpatient surgical procedures suggesting opportunities for savings, however assessment by claims data is often criticized for lacking clinical granularity and validation. By leveraging linkage between two unique datasets, a clinical registry and a claims based registry, we sought to characterize variation in payments for pancreaticoduodenectomy (PD) across hospitals in Michigan. By linking outcomes to cost this study is able help differentiate between quality problems (e.g. complications) and efficiency problems (e.g. overutilization). Variation in outcomes and utilization may highlight opportunities for collaborative quality improvement and practice standardization. Methods: We linked Michigan Surgical Quality Collaborative (MSQC) clinical registry data for PD performed at 8 Michigan hospitals to the Michigan Value Collaborative (MVC) registry from 2014 to 2018. The MSQC is a regional collaborative that maintains a robust and well-established state-wide clinical registry focused on outcomes and health care use. The MVC is a statewide collaborative focused on improving the value of care in the state of Michigan which maintains a claims-based registry with 30-day price-standardized episodes of care from multiple payers. Payment components between hospitals were compared with identified drivers of variation (e.g., index hospitalization/procedure, readmissions, post-acute care, and professional fees). Results: Among 1,535 PD episodes identified, the 30-day episode payment by hospital ranged between $34,680 and $56,973 with a median of $40,063. Index hospitalization payments were the primary driver of this variation (46%), followed by post-acute discharge payment (34%). Readmissions by hospital ranged from 5% to 22% with a median of 17%; post-operative morbidity ranged from 11% to 35% with a median of 28%. Notably, higher payments were not associated with improved clinical performance (Figure 1). As such, individual hospitals may look at their own complication/cost profiles to understand if they are managing complications as efficiently as possible. Alternatively, hospitals with low morbidity but high costs may be over-utilizing services (e.g. post-acute care). Conclusion: In this analysis, the linkage of clinical registry data with claims-based registry data provides a unique opportunity to study drivers of cost and variation. For patients undergoing PD in Michigan, we found a wide variation in surgical episode spending across hospitals without major difference in clinical performance. Hospital leaders may seek to better understand variation in practices between their hospitals in order to standardize care and reduce variation in post-acute care and other utilization. Similar procedure-specific analyses should be performed for other clinical service lines in order to understand whether variations in quality (e.g., complications) or efficiency (e.g., post-acute care) drive observed differences in spending.