Research in Progress

A Prescription for Manipulation? Ex Ante Hospital Strategic Behavior and the 340B Drug Discount Program. (with Melinda Buntin and Rena Conti)

The 340B program entitles certain public, non-profit hospitals to discounted outpatient drugs, and substantial price-cost margins when discounts are not passed onto payers. Previous work suggests 340B increases ex post strategic behavior. Using hospital administrative data from 1996-2016, we show that 340B also results in ex ante strategic behavior through manipulation of eligibility criteria. Non-parametric density tests show marked discontinuities in the eligibility criterion around the eligibility cutoff. The non-parametric hazard of manipulating eligibility for 340B is higher among hospitals with larger returns to participation. Finally, within-hospital trends in the 340B eligibility criterion break after hospitals meet the minimum eligibility criteria.
Draft available upon request.

Impact of Switching Health Insurance Plans on Cost and Utilization: Evidence from Medicaid Managed Care. (Zhang S, Dusetzina S.)

Changing health insurance plans in the presence of selective contracting may disrupt continuity of care. Yet estimating the effect of switching plans is difficult because consumers infrequently switch, and observed switching may be endogenous. We use the random assignment of Medicaid beneficiaries to new managed care plans in one state in 2015 to estimate the effect of switching to a new plan on spending and utilization. Comparing outcomes of individuals who were randomly enrolled in a new plan to those who remained in their existing plan before and after randomization, we find those who were switched to a new plan use less outpatient care and were slightly more likely to have an inpatient stay in the first months after the randomization. However, differences between those randomly enrolled in a new plan and those who remained in the same plan disappeared after one year.
Draft available upon request.

The Medicaid Windfall: Medicaid Expansions and the Target Efficiency of Hospital Safety-Net Subsidies

Many Americans seeking hospital care lack health insurance or are unable to afford out-of-pocket payments. To assist hospitals with this $38B burden of unreimbursed care, the Federal government subsidizes certain hospitals that provide large amounts of uncompensated care to uninsured and low-income patients, including unrestricted, lump-sum payments through the Medicare or Medicaid disproportionate share program and outpatient drug acquisition discounts though the 340B program. Ideally receipt of these subsidies would be proportional to the amount of unreimbursed care that hospitals provide, relative to their total operating budget (“safety-net burden”). However, existing methods for allocating safety-net subsidies depend wholly or in -part on the Medicaid share of hospital inpatient care. As Medicaid has become more ubiquitous, more than tripling between the 1970’s and 2014 participation in safety-net subsidies has grown, even as uncompensated care to uninsured patients has fallen. Using hospitals administrative data, I leverage two natural experiments to provide quasi-experimental variation in Medicaid coverage unrelated to hospital need for safety-net subsidies: the Affordable Care Act's 2014 Medicaid expansion and the 2005 disenrollment of 170,000 on Medicaid in Tennessee. I find that the ACA’s Medicaid expansion increased Medicaid-covered hospital inpatient stays by 15%, which increased Medicaid-based eligibility criteria by 1.8 pp. The increase in eligibility corresponds to a 10% increase in Medicare DSH payments, and a 6 pp increase in the probability of receive Medicaid DSH payments, and a 5-pp increase participating in 340B. These increases occurred despite a 40% reduction in uncompensated care and a 60% drop in uninsured patients among hospitals in expansion states relative to non-expansion states. I find that the TennCare disenrollment reduced Medicaid inpatient stays by 7% and therefore also reduced Medicaid-based eligibility criteria by 1.8 pp. This reduction in eligibility corresponded to an 8% reduction in Medicare DSH payments, a 7% reduction in 340B participation, and a 4-pp reduction in Medicaid DSH payments. These reductions in safety-net subsidies occurred despite a 46% increase in uninsured discharges.
Draft available upon request.

Measurement Error in Hospital Cost Reports.

Hospitals occupy a central role in the US health care system. One-third of all health care spending occurs in hospitals and two-thirds of uncompensated medical care is provided by hospitals. Public subsidies such as tax exemptions, disproportionate share payments, and the 340B program target hospitals. Numerous recent health policies, including the Affordable Care Act, focus on hospitals. For these reasons, studies of hospitals operation, particularly hospital financial characteristics, are both important and a necessary aspect of policy analysis. However, the only source of publicly available, longitudinal hospital financial data, the Hospital Cost Reports are self-reported and can differ significantly from audited statements. Although discrepancies between audited statements and hospital cost reports have been documented, no research considers whether measurement error is random or correlated with key hospital characteristics and policies. I link cost report data from 7 states (AZ, FL, MA, ME, OR, PA, WA) to audited hospital financial data. My analysis illustrates that errors in two key financial measures, the operating margin and uncompensated care, are correlated with several important hospital characteristics. I also propose solutions to minimize bias from non-classical measurement error in analyses of hospital cost reports.
Draft available upon request.