The Effects of Supplemental Insurance on Health Care Spending for Disabled Medicare Beneficiaries: A Regression Discontinuity Approach

Background:   Medicare’s primary mechanism for controlling health care spending is beneficiary cost sharing.  Many beneficiaries, however, purchase private, supplemental insurance (Medigap) that covers most of the cost sharing of the standard benefit. A key concern for the efficiency of Medicare spending is that supplemental insurance increases spending on covered services and that much of that incremental spending is borne by the publicly funded program rather than private insurers. Estimating the effect of supplemental insurance on Medicare spending is challenging, however, since the purchase of supplemental coverage is voluntary and is likely related to unobserved characteristics of people that influence their health care use.       

Data and Methods.  In this paper, we use a “fuzzy” regression discontinuity design to estimate the effects of Medigap coverage on health care spending for disabled beneficiaries. Medicare has a 6-month open enrollment period for 65-year-olds newly entering the program during which guaranteed issue and community rating applies to all insurers. For disabled beneficiaries, this creates a discontinuity at age 65 in underwriting practices for Medigap coverage.  Prior to age 65, they are subject of insurer underwriting practices and face premiums reflecting their individual risk.  For 6 months after turning 65, they are eligible for guaranteed issue and a pooled premium.  

Using Medicare claims from 2006 to 2010, we study a sample of approximately 90,000 Medicare beneficiaries who became eligible for Medicare based on disability prior to age 55.  We test whether the change in underwriting practices generates a discontinuity in coverage by estimating a model of the relationship between monthly Medigap coverage and beneficiary age.  To estimate the effect of coverage on spending, we use an intent-to-treat approach, dividing the estimate of the effect by the change in coverage to determine the effect on the treated.  We also use instrumental variables, using the age of age 65 as the instrument.  In all models, we control the effect of age on expenditures by including both age and age-squared as well as their interaction with age >65 as control variables.  We estimate two-part models to account for the large number of zeros and skewness in the dependent variables.  For the IV estimates, we use 2-stage residual inclusion methods.   

Results.  Prior to age 65, about 10% of disabled beneficiaries have Medigap coverage.  We find that the change in underwriting practices that disabled Medicare beneficiaries experience at age 65 result in an approximately 7 percentage point increase in coverage.  The change is discrete – rates of coverage are constant in the 12 months prior to turning 65 and increase during the 6 open enrollment period after age 65.  Our preliminary estimate of the effects of coverage on spending indicate that supplemental coverage increases Medicare Part B, but not Part A, spending.  We estimate the Part B spending increases by 35-40%. 

Conclusions.  Medigap supplemental insurance generates a substantive externality on the publicly-financed Medicare program.  For disabled beneficiaries, demand for outpatient services is more price responsive than demand for inpatient services.