Abstract 3: Impact of Financial Incentives on Variation in Use of Cardiac Imaging Tests

Abstract

Background: Rapid growth and geographic variation in the provision of cardiac imaging tests have led to concerns about overuse due to fee-for-service (FFS) incentives. The degree to which FFS incentives may influence rates of cardiac imaging over and above patient characteristics and local practice styles is unknown. Objectives: To examine overall rates, degree of geographic variation, and correlation in use of echocardiography (ECHO) among veterans who primarily use services provided by the Veterans Health Administration (VA - a fixed budget health system without significant FFS incentives), versus veterans who use FFS Medicare. Design: We analyzed administrative claims from VA and Medicare of veterans with heart failure over the age of 65 from 2007-2010. Veterans were assigned to the VA or Medicare cohort according to the volume of services (procedures, hospitalizations, and visits) received within each system. The analysis was restricted to 34 major metropolitan service areas (MSAs). Rates of ECHO in the overall cohort and in a propensity-matched cohort were compared using multilevel mixed effects regression models adjusted for patient-level characteristics. Mean adjusted rates for each MSA according to cohort were tested for correlation and difference in variance. Results: The Medicare cohort included 364,413 veterans (mean age 77 years) and the VA cohort included 15,330 veterans (mean age 76 years). The Medicare cohort had a significantly higher adjusted rate of ECHO use compared to the VA cohort (1.09 versus 0.28 ECHOs per person-year, incidence rate ratio 4.23 [95% CI 4.12 to 4.34], p<.001). The higher rate persisted in the propensity-matched cohort of 14,889 pairs (Medicare incidence rate ratio 1.98 [95% CI 1.92 to 2.04], p<.001). Variance of the mean adjusted use of imaging across MSAs was greater in the Medicare cohort than the VA cohort (0.14 versus 0.02, p<.001). There was modest correlation in geographic variation between cohorts (r = 0.56, p<.001, Figure 1). Conclusions: ECHO rates and degree of variation were significantly higher in the Medicare cohort than the VA cohort in both overall and propensity-matched analyses, with modest regional correlation. ECHO utilization rates may be strongly influenced by payment system despite differences in patient characteristics and local practice styles.