In May 2025, LDI Senior Fellows Eric T. Roberts, PhD and Aaron Schwartz, MD, PhD, and LDI Executive Director Rachel M. Werner, MD, PhD wrote a research memo on the loss of subsidized drug coverage and mortality following Medicaid disenrollment, based off of a recent study, which found that dually eligible individuals who lost Medicaid, and consequently the Medicare Part-D Low-Income Subsidy, experienced higher mortality. 

The research memo delves further into the study findings, considers how the results can inform projections about the effects of proposed Medicaid policies for low-income Medicare beneficiaries, and compares the findings to other recent studies.

---

MEMO

Re: Request for Technical Assistance: Loss of subsidized drug coverage and mortality following Medicaid disenrollment: translating our findings and implications for Medicaid policy

From: Eric T. Roberts, PhD; José F. Figueroa, MD, MPH; Aaron Schwartz, MD, PhD; and Rachel M. Werner, MD, PhD

Date: May 23, 2025

Summary: The Congressional Budget Office (CBO) projects that the House budget reconciliation bill would cause 1.38 million low-income Medicare beneficiaries to lose Medicaid coverage. Based on our new study in the New England Journal of Medicine, we estimate that Medicaid coverage losses of this magnitude could result in 18,200 additional deaths among Medicare enrollees every year. This projected increase in mortality stems from the fact that losing Medicaid also leads to the loss of prescription drug assistance through the Low-Income Subsidy (LIS) program.

Introduction

Congress is considering major changes to Medicaid policy and financing that could have widespread effects on coverage. CBO forecasts that the proposed changes will affect Medicaid coverage for low-income Medicare beneficiaries by reducing enrollment in the Medicare Savings Programs (MSPs). The MSPs are Medicaid benefits that help cover the out-of-pocket costs of Parts A and B for Medicare beneficiaries with limited income and assets.

Importantly, enrolling in Medicaid also confers automatic eligibility for a separate program—the Low-Income Subsidy (LIS)—that substantially reduces out-of-pocket drug costs in Medicare Part D. Lower Medicaid enrollment is expected to result in fewer low-income Medicare beneficiaries receiving the LIS.

In this memo, we delve further into the findings from a recent study we co-authored, published in the New England Journal of Medicine. Our study examined what happened to low-income Medicare beneficiaries after they lost Medicaid, and with it, enrollment in the LIS.¹ We consider how our results can inform projections about the effects of proposed Medicaid policies for low-income Medicare beneficiaries and compare our findings to other recent studies.

Overview of our study and findings

Our study examined the impacts of losing the LIS due to Medicaid loss, focusing on a rule that determines when LIS coverage ends based only on when in the calendar year a person loses Medicaid. Because of this rule, LIS coverage ends much sooner when Medicaid loss occurs slightly earlier in a calendar year. This creates two similar groups who lose LIS coverage at different times relative to Medicaid loss, enabling us to isolate what happens when LIS coverage is lost. We find that when Medicare beneficiaries lose LIS² after Medicaid disenrollment, they fill fewer prescriptions,³ and were more like to die.⁴

Applying our estimates to project potential effects of Medicaid policy changes

To approximate the mortality effects of the reconciliation bill, we scale our estimates to match the projected coverage losses from the bill. Our study found that people who lost Medicaid earlier and had 2 fewer months of LIS coverage had higher mortality. To scale our estimates to the effect of not receiving the LIS for a full year, we multiply by a factor of 6 (i.e., scaling from 2 fewer to 12 fewer months of LIS coverage).⁵ This extrapolation implies that losing LIS for a full year results in:

• Individuals filling an average of 7.2 fewer prescriptions per year
• An additional 18.0 deaths/1,000 individuals per year overall. This includes a weighted average of:⁶
  • 13.2 additional deaths/1,000 per year, among those eligible for the MSPs
  • 27.0 additional deaths/1,000 per year, among those eligible for full Medicaid⁷

The overall increase in mortality (18.0 deaths/1,000 individuals) is a weighted average of the effects among those eligible for the MSPs and those eligible for full Medicaid.

Section 44101 of the House of Representatives’ budget reconciliation bill proposes to delay until 2035 the implementation of a Medicaid eligibility and enrollment rule that streamlines MSP enrollment. CBO estimates this provision would reduce Medicaid enrollment among Medicare beneficiaries by 1.38 million people in 2034.⁸ Owing to the close link between Medicaid and the LIS, lower Medicaid enrollment is expected to result in fewer people receiving LIS.

Using our findings and CBO’s estimates, we project the potential increase in deaths attributable to individuals not receiving the LIS because they do not have Medicaid (Table 1). Since the Medicaid eligibility and enrollment rule primarily affects enrollment in MSPs, we use our mortality estimate for MSP enrollees (13.2 additional deaths/1,000). We estimate that 18,200 additional deaths would occur each year.

Table 1: Projected increase in annual mortality among low-income Medicare beneficiaries due to the delayed implementation of the Medicaid eligibility and enrollment final rule

The vast majority of Medicare beneficiaries enrolled in the MSPs live in the community. Therefore, we expect this increase in deaths to occur primarily among community-dwelling Medicare beneficiaries.

Comparing our findings to the prior literature

Several studies have estimated the effects of public insurance, or cost sharing within public insurance, on mortality. In Table 2, we summarize and scale those studies’ estimates to show implied effects on annual mortality per 1,000 individuals.

Table 2: Mortality estimates from other recent studies

Chandra, Flack, and Obermeyer’s 2024 study is most closely related to ours as it focuses on prescription drug cost assistance through Medicare Part D. Their study analyzed the Part D “donut hole” (coverage gap)—a feature of Part D’s original benefit design in which individuals faced 100% cost-sharing after reaching an initial coverage limit (until reaching a catastrophic coverage threshold). They analyzed individuals who were likely to have spending in the coverage gap and used variation in the timing of when new Medicare enrollees reached this gap during their first year of Part D coverage.¹¹

Annualized estimates from Chandra et al.’s paper imply that each $100 increase in out-of-pocket costs led to an increase of 2.0 deaths/1,000 per $100 increase in out of-pocket costs.¹⁰ To compare to our findings, we scale Chandra et al.’s estimate for the median level of cost-Part D sharing subsidies provided to LIS enrollees ($600 in the year before LIS loss), yielding an estimate of 12.0 additional deaths/1,000 per year. This scaled estimate is very close to the increase in mortality due to LIS loss that we estimated among MSP-eligible Medicare beneficiaries (13.2 additional deaths/1,000 per year).

Other studies consistently show that public insurance saves lives among low-income Americans, confirming likely mortality effects of proposed Medicaid policy changes. For example, Miller, Johnson, and Wherry (2021) examined the effects of Medicaid expansion on low-income adults aged 55-64 at the time of expansion, whereas Wyse and Meyer (2025) and Sommers (2017) analyzed broader populations of working-age adults with low incomes. These studies estimated that Medicaid expansions reduced annual mortality by between 1 to 4 deaths/1,000 working-age adults. However, because the populations in these studies were not Medicare-eligible, direct comparisons to our findings are more limited.

Conclusion

Our research adds to the growing body of evidence that access to affordable insurance is essential for protecting the health of low-income Americans. We show that losing prescription drug subsidies tied to Medicaid leads to increased mortality among low-income Medicare beneficiaries. Policies that cause low-income Medicare beneficiaries to lose Medicaid—and, as a result, the LIS—could lead to thousands of preventable deaths.

Authors

Eric T. Roberts, PhD is an Associate Professor of General Internal Medicine and Senior Fellow at the Leonard Davis Institute of Health Economics at the University of Pennsylvania

José F. Figueroa, MD, MPH is an Associate Professor of Health Policy and Management at the Harvard T.H. Chan School of Public Health and Assistant Professor of Medicine at Harvard Medical School

Aaron Schwartz, MD, PhD is an Assistant Professor of Medical Ethics and Health Policy and Senior Fellow at the Leonard Davis Institute of Health Economics at the University of Pennsylvania

Rachel M. Werner, MD, PhD is the Executive Director of the Leonard Davis Institute of Health Economics, Professor of Medicine at the University of Pennsylvania Perelman School of Medicine, and the Robert D. Eilers Memorial – William Maul Measey Professor in Health Care Management and Economics at the Wharton School

---

Endnotes:

1. For 85% of LIS enrollees, receiving this subsidy is linked to having Medicaid. Therefore, losing Medicaid often results in LIS loss.
2. An average of about 2 fewer months of cumulative LIS enrollment.
3. An average of 1.2 fewer prescriptions through the Medicare Part D program.
4. The cumulative mortality difference was 3.0 deaths/1,000 beneficiaries. Mortality increases varied across beneficiary subpopulations. For example, cumulative mortality was higher by 4.5 deaths/1,000 among those who lost full Medicaid (which covers the poorest and sickest individuals); higher by 2.2 deaths per 1,000 among those who lost partial Medicaid (for those with slightly higher incomes); and amplified among people who, at baseline, had high drug spending and used medications for conditions such as HIV.
5. This extrapolation uses the intuition of instrumental variables (IV), a method used by economists to estimate treatment effects in settings where an exogenous factor has a meaningful, yet “incomplete,” effect on treatment. In our study, the month of Medicaid disenrollment (exogenous factor) leads to cumulative differences in LIS coverage (treatment). The extrapolations assume that differences in outcomes (i.e., medication filling and mortality) between groups losing Medicaid earlier vs. later were due only to differences in those groups’ subsequent LIS enrollment. We examined these assumptions in our study’s appendix.
6. Similar person-months of LIS loss occurred for individuals losing full and partial Medicaid, implying that a similar extrapolation can be applied to both subpopulations.
7. Not used in our Table 1 projections because the majority of individuals affected by the Medicaid eligibility and enrollment final rule qualified for partial Medicaid.
8. Reflects a May 7, 2025 CBO estimate that “in 2034, 2.3 million people would no longer be enrolled in Medicaid under this option. Roughly 60 percent of the people who would lose Medicaid coverage would be dual-benefit enrollees who would retain their Medicare coverage.” Sixty percent of 2.3 million is 1.38 million individuals.
9. Wyse and Meyer estimated a 25% relative reduction in annual mortality hazard from gaining Medicaid due to expansion (treatment-on-treated effect). On a baseline annual mortality rate of 3.0 deaths/1,000, this implies that gaining Medicaid coverage leads to 0.75 fewer deaths/1,000 per year.
10. Chandra et al. estimated that that each $100 increase in out-of-pocket drug costs (due to reaching the Part D coverage gap sooner) increased 1-month mortality by 0.164 deaths/1,000 population. Annualized, this implies an increase of 2.0 deaths/1,000 per $100 increase in out-of-pocket costs.
11. Chandra et al. excluded dual Medicare-Medicaid eligibles and persons enrolled in the LIS, who did not face the Part D coverage gap. They also focused on 65-year-olds in their first year of Medicare coverage, a group with lower mortality than other subpopulations of Medicare beneficiaries.