HITTING A BULL'S EYE? SPILLOVER EFFECTS OF A VALUE-BASED INSURANCE DESIGN PLAN ON NON-TARGETED MEDICAL SERVICES

Objective: Value-based insurance design (V-BID) is a type of health benefit structure in which patient out-of-pocket costs are varied based on the clinical value of the service. These designs have been used and studied for about a decade and have been shown to induce expected negative own-price elasticities for targeted services; e.g. reducing or eliminating medication cost sharing increases utilization. Yet, there is limited evidence on whether these programs affect the use of other medical products and services. This study evaluates the cross-price effects of an employer-sponsored V-BID program on non-targeted outpatient services. It looks at whether the inducement of demand on targeted services (shown in previous work) has spillover effects into other non-targeted services. If so, depending on the cost and clinical value of the non-targeted services, these spillovers may moderate the ability of V-BID plans to constrain short-run spending and utilization.

Methods: We analyze a V-BID program implemented by Connecticut for its state employees in 2011. The program removed financial barriers for selected high-value health care services and coupled that with enrollee accountability requirements and a surcharge for noncompliance. It induced positive changes in utilization for targeted services, with some evidence of substitution away from emergency department visits. With detailed claims data from all Connecticut state employees from one year before through two years after the benefit design change, we compare utilization in Connecticut with trends in six other state employee populations in the Truven MarketScan database using a difference-in-differences research design.  We examine services that are similar to those targeted by the intervention, including preventive screenings (both recommended and not recommended), imaging studies and other primary care services.

Results: We found heterogeneous effects across non-targeted services. Some services show significant increases after the intervention, such as prostate-specific antigen screening (15.2 ppt in year 1, 2.6 ppt in year 2; p<0.01 ) and screening for Hepatitis C in people 45 and older (0.5 ppt in year 1; 1.5 ppt in year; 2; p<0.01)  while others show no or slightly negative effects. Imaging for low back pain without complications, for example, shows a statistically significant decrease of -0.45 ppt in year 1 and -0.37 ppt in year 2 (p<0.01).

Conclusions: Overall, there is evidence that targeted insurance benefit design changes can have substantial spillover to non-targeted services. We find that certain services, including some not widely supported through clinical evidence, increase after the implementation of V-BID in a state employee population. Our findings suggest that demand-side benefit structure changes could have unintended consequences. While the welfare effects are ambiguous and likely depend on the specific medical service, the results suggest that the effects of targeted interventions on cost and value may be mitigated by these spillover effects. Achieving more specific targeting may require additional interventions such as supply-side incentive modification.