Impact of Increased Cost Sharing on Utilization of Low Value Services

Monday, June 13, 2016: 8:30 AM
G55 (Huntsman Hall)

Author(s): Jonathan Gruber; Catherine Maclean; Kevin G. Volpp; Bill Wright; Eric Wilkinson

Discussant: Haizhen Lin

Given rising healthcare costs in the United States and the relatively poor health of Americans, research on cost-effective benefit designs that do not impede patient outcomes is critical.  Value-based insurance design (VBID) programs hold great promise as they align patient out-of-pocket spending with service value, and can thereby direct health dollars away from wasteful services, and improve patient outcomes through better alignment of treatment and patient need. 

The existent VBID literature focuses primarily on reducing patient cost sharing for high value care and has tended to be limited to medications.  There is scant information on the effectiveness of raising patient cost sharing for low value care (e.g., unnecessary imaging services).  Moreover, insight from behavioral economics suggests that the latter benefit design may be more effective in changing behavior as it penalizes those patients using services of low value.  Better evidence on VBID programs is needed, given that such programs are embedded in the Patient Protection and Affordable Care Act and numerous employers across the U.S. are implementing such programs. 

 In this study we evaluated two innovative VBID programs implemented by public employers in the state of Oregon between 2010 and 2012.  The VBID programs we evaluated increased cost-sharing for low value healthcare services.  Specifically, patient co-payments were increased by $100 to $500, representing a much more substantial rise in out-of-pocket spending than has typically been evaluated in the past.  Targeted services included a number of services believed to offer limited benefits to patients (for example, inpatient sleep studies, upper GI endoscopies that are not viewed as medically necessary, and over-used imaging services and surgeries), and emergency room episodes. 

To examine the effect of these VBID programs we estimate differences-in-differences (DD) models coupled with administrative claims data spanning 2009-2013 on approximately 300,000 public sector employees and their dependents.  Specially, we exploit programmatic changes that occurred across time, organization, and service, leveraging a quasi-experimental design to estimate the impact of the VBID programs.  To further strengthen our design, we utilize comparable data from three public employers in Oregon that did not implement a VBID program over the same time period as a comparison group.  Our outcomes include healthcare utilization (both on the extensive and intensive margins), healthcare costs, and patient well-being.  We also examine heterogeneity in treatment effects by patient characteristics.

Our findings suggest that increasing cost-sharing for low value healthcare services reduces utilization of some, but not all, targeted services.  However, we find that increased cost-sharing may lead to modest increases in overall healthcare spending.  Finally, we identify heterogeneity by patient characteristics.  These findings suggest that although VBID programs may steer patients away from targeted low value services, these programs may not lead to substantial reductions in healthcare spending overall, at least in the short run.  Future studies should examine longer term impacts of VBID programs and different types of VBID program designs on utilization and costs.