Medicare Audits
Introduction
Medicare is the federal health insurance program for elderly and disabled persons in the United States. Medicare reimburses hospitals with a prospective payment for each inpatient stay. The amount of the payment depends on several factors, but not the length of stay. One concern with this set-up is that there is incentive for hospitals to unnecessarily admit patients for extremely short stays, as this yields a beneficial payment-to-cost ratio. To create deterrence for this behavior and to recoup some of the costs of unnecessary hospital admittances, Medicare contracts private auditing firms to manually review individual Medicare claims through the Recovery Audit Contractor (RAC) program.
Shi (2024) uses administrative data and an difference-in-differences design to study trade-offs between preventing and recovering wasteful Medicare spending and the potential social costs of additional monitoring. The paper finds that, for each dollar Medicare spent on monitoring, there was $24-29 in government savings with no evidence of harm to patients.
In a companion online appendix, Shi (2024b) uses the paper’s estimates of the effects on admissions and compliance costs to estimate the MVPF of an increase in the 2011 RAC audit rate. To account for the effect of future deterrence on total benefits and costs, the appendix calculates the MVPF over different time horizons, using the cumulative costs and benefits between 2011 and each year of 2011-2018. The calculation for 2011-2013 is reported here as the baseline.
MVPF = 1.5
Net Cost
The paper calculates the net cost to the government as the sum of deterred and refunded payments less the change in monitoring costs (i.e., the payments to the private auditing firms).
The saved cost due to deterred admissions is $818,581 as discussed above.
The total refunded payments is the reclaimed payments (after fees to the auditing firms) minus the amount refunded to hospitals as a result of lawsuits. This is equal to (1-0.1075-0.68) * (314,115 + ((1/1.02) * 179,550) + ((1/1.02)^2 * 133,512) = $131,425.
The paper estimates the effect on government monitoring costs by multiplying the reclaimed payments by a contingency fee of 10.75% (the average of the range of RAC contingency fees, 9% to 12.5%). The estimated reclaimed payments is 314,115 in 2011, 179,550 in 2012, and 133,512 in 2013. Assuming a 2% discount rate, this is $618,473 * 10.75% = $66,486 in 2010 dollars.
The baseline net cost is then $818,581 + $131,425 – $66,486 = $883,521.
Willingness to Pay
The paper calculates the societal willingness to pay to avoid an increase in the RAC audit rate as the sum of four things: (1) the change in hospital revenue, (2) the change in hospital compliance costs, (3) the change in treatment costs, and (4) the change in patient health.
(1) The estimated effect on hospital revenue is equal to the sum of the change in revenue from inpatient admissions plus the share of reclaimed payments that were later refunded to hospitals.
The estimated change in log inpatient revenue is -0.0102 in 2011, -0.0177 in 2012, and -0.0280 in 2013. Assuming a 2% discount rate and using the median inpatient revenue value from 2010 ($15,029,306), the estimated willingness to pay to not experience that decrease in revenue from inpatient admissions is $15,029,306 * (0.0102 + ((1/1.02) * 0.0177) + ((1/1.02)^2 * 0.0280) = $818,581.
The amount directly reclaimed from a 1 percentage point increase in audit rate is estimated to be $314,115 in 2011, $178,550 in 2012, and $133,512 in 2013. Some amount of reclaimed payments is ultimately refunded back to hospitals as a result of lawsuits. Based on the details from a 2014 lawsuit, the paper assumes this number is 68% (that is, 68% of reclaimed payments are refunded back to hospitals). As a result, the total estimated revenue lost to hospitals as a result of payments reclaimed by Medicare is equal to (1-0.68) * (314,115 + ((1/1.02) * 179,550) + ((1/1.02)^2 * 133,512) = $197,911.
(2) The estimated effect on log hospital compliance costs is 0.0154 in 2011, 0.0068 in 2012, and 0.0034 in 2013. Assuming a 2% discount rate, the estimated change in hospital administrative costs is $12,822,887 * (0.0154 + ((1/1.02) * 0.0068) + ((1/1.02)^2 * 0.0034)) = $324,863 (where $12,822,887 is the median administrative cost value from 2010).
(3) In the baseline calculation, the paper assumes that the change in treatment costs is equal to 0. The paper notes that this would be true if, for example, hospitals substituted marginal admissions with other case with the same cost. The paper notes that this is likely a lower bound on savings from changes in treatment cost (i.e., the baseline willingness to pay is likely an overestimate).
(4) In the baseline calculation, the paper assumes that the value for changes in patient health is 0. This is based on the result in the paper that there is no effect on health outcomes for the marginal patient.
The baseline WTP is then $818,581 + $197,911 + $324,863 + 0 + 0 = $1,341,354.
MVPF Summary
The MVPF of an increase in the 2011 RAC audit rate, assuming two years of deterrence effects, is $1,341,356/$883,521 = 1.52. The baseline MVPF in the first year (2011) is about 2.4; assuming seven years of deterrence effects, the MVPF is below 1.2.
The paper also notes that the MVPF assuming two years of deterrence effects ranges from 0.67 to 8.05 depending on alternative assumptions (e.g., different assumptions about the impact on patient health or excluding compliance costs).
References
Maggie Shi (2024). Monitoring for Waste: Evidence from Medicare Audits. The Quarterly Journal of Economics 139(2): 993-1049. https://doi.org/10.1093/qje/qjad049
Maggie Shi (2024b). Online MVPF Appendix for “Monitoring for Waste: Evidence from Medicare Audits.” https://mshi311.github.io/website2/Shi_MedicareAudits_MVPFappendix.pdf
Policy Features
- Category
- Social Insurance
- Sub-Category
- Health Adult
- Beneficiary Type(s)
- Adults
- Country of Implementation
- United States
- Empirical Method
- Difference in Differences
- Research Type
- Primary
- Peer Reviewed
- Yes
- MVPF Publication Link
- doi.org/10.1093/qje/qjad049