Increased Payments to Medicare Advantage Plans for Dually Eligible Beneficiaries

ABSTRACT

Objective: To address concerns about payment adequacy in Medicare Advantage (MA) plans, a 2017 federal policy change increased risk scores and associated capitated payments for community-dwelling dually eligible beneficiaries with full Medicaid benefits. This study examined whether this payment change was associated with changes in health care utilization or mortality for dually eligible beneficiaries.

Study Design: Difference-in-differences analysis comparing dually eligible beneficiaries who qualified for risk score increases (full Medicaid enrollees) vs those who did not (partial Medicaid enrollees).

Methods: CMS plan payment files for 2014-2022 provided plan-level information on mean risk scores. We linked 2013-2019 Medicare data and Minimum Data Set nursing home assessments to analyze inpatient use, nursing home use, and mortality for community-dwelling dually eligible beneficiaries. We also investigated hospital readmissions, stays lasting longer than 100 days, and mortality within 365 days among dually eligible beneficiaries with skilled nursing facility (SNF) use.

Results: Among plans in which more than half of members had full Medicaid, plan-level risk scores increased 8.9% from 2014-2016 to 2017-2022 relative to the change observed in other plans. The payment change was associated with small declines in mortality, inpatient use, and nursing home use among beneficiaries 65 years and older, but these findings were not clinically significant or robust in sensitivity analyses. No significant changes were observed for dually eligible beneficiaries younger than 65 years or among SNF users in either age group.

Conclusions: These results raise questions about whether payment increases to MA plans led to meaningful improvements in quality of care for dually eligible members. As MA participation increases among dually eligible beneficiaries, policy makers should pay attention to whether higher MA payment levels for these beneficiaries translate to improved outcomes.

Am J Manag Care. 2026;32(2):In Press

Takeaway Points

After a 2017 federal policy distinguished between full and partial Medicaid status in the risk calculation and increased risk scores and associated capitated payments for community-dwelling dually eligible beneficiaries with full Medicaid, Medicare Advantage (MA) plans that primarily served this population experienced large increases in mean risk scores.

• Our analysis comparing dually eligible beneficiaries who qualified for risk score increases (full Medicaid enrollees) vs those who did not (partial Medicaid enrollees) found little evidence that this policy change was associated with changes in inpatient or nursing home use among full Medicaid enrollees or improved outcomes among full Medicaid enrollees with skilled nursing facility use.
• These results raise questions about whether payment increases improved quality of care for dually eligible MA members.
• Cutting MA payments may be feasible without major disruptions for dually eligible members.

Medicare Advantage (MA) participation has soared in the past decade among dually eligible beneficiaries with full Medicaid and Medicare, increasing from 18% in 2012¹ to 59% in 2024.² MA plans receive a fixed monthly payment per beneficiary from CMS that is risk adjusted for health care conditions and other factors. Medicaid participation is one factor that increases MA payment rates, reflecting the fact that dually eligible beneficiaries’ spending levels in traditional Medicare are more than double the levels of non–dually eligible beneficiaries.³

Without adequate payment rates for dually eligible beneficiaries, MA plans may reduce coverage generosity or exit markets that predominantly serve this population. At the same time, Medicare’s financial sustainability is at risk unless CMS addresses the substantially higher costs of covering beneficiaries through MA plans compared with traditional Medicare.^4-6 This concern prompts debate about whether it is feasible to reduce MA payment rates without adversely impacting health outcomes, especially for dually eligible beneficiaries with more complex health care needs.

A 2017 policy change provides an opportunity to understand how an MA payment rate increase affected outcomes for dually eligible beneficiaries. Prior to 2017, CMS adjusted MA payments to be higher for dually eligible beneficiaries without considering beneficiaries’ level of Medicaid benefits. Starting in 2017, CMS updated its risk adjustment approach to differentiate between level of Medicaid benefits.⁷ In practice, plans started receiving larger monthly payments for community-dwelling beneficiaries with full vs partial Medicaid. Full Medicaid eligibility is based on state-specific rules regarding income, assets, need for long-term services, or out-of-pocket spending. Partial Medicaid eligibility is based on income and assets limits that are higher than resource limits for full Medicaid.

If MA plans spent the 2017 payment increases on better coverage, then plans could have devoted more funding to high-quality providers, expanded supplemental benefits, or engaged in more robust care coordination that may have improved outcomes for beneficiaries with full Medicaid. This study’s difference-in-differences (DID) analysis compares changes in inpatient use, nursing home use, and mortality among community-dwelling dually eligible beneficiaries with full vs partial Medicaid before and after 2017 to determine whether the policy change led to meaningful improvements for this population. Because many Medicaid programs aim to delay nursing home entry among dually eligible beneficiaries, we also investigated quality-of-care outcomes for dually eligible MA enrollees with skilled nursing facility (SNF) use.

METHODS

Data Sources and Study Population

Publicly available CMS plan payment files for years 2014-2022 provided plan-level information on mean risk scores; data on plan composition came from the Medicare Beneficiary Summary File (MBSF) data and Minimum Data Set (MDS) nursing home assessments. For the main study analysis, we linked 2013-2019 MBSF data, Medicare Provider Analysis and Review hospitalization records, and MDS data to analyze outcomes for community-dwelling dually eligible MA enrollees. We excluded beneficiaries with long-term nursing home use in the past year because they were not affected by the payment change. Because many states have focused on delaying nursing home entry among dually eligible beneficiaries, we also investigated outcomes for dually eligible MA enrollees with an SNF admission within 7 days of hospital discharge. This analysis excluded beneficiaries with any nursing home use within the prior year.

We selected dually eligible beneficiaries who qualified for Medicaid with income at or below the federal poverty limit (FPL), which we identified based on Medicaid participation via the Qualified Medicare Beneficiary (QMB) program for individuals with income less than 100% of the FPL. We focused on this subset because we expected their income and asset levels to be less variable than in the entire dually eligible population. All states must offer at least partial Medicaid to beneficiaries with income less than 100% of the FPL, and federal guidelines further require most states to offer full Medicaid to beneficiaries with income below 75% FPL.³ Depending on their state’s eligibility rules, beneficiaries may also qualify for full Medicaid with income up to 100% of the FPL or because of long-term care needs or high out-of-pocket costs. Partial Medicaid assists only with Medicare out-of-pocket costs, whereas full Medicaid also provides additional Medicaid coverage of services that Medicare does not cover, such as long-term services and supports.

Because states that offer full Medicaid to all beneficiaries with income under 100% of the FPL have very few QMB enrollees with partial Medicaid, our main analysis focused on the 33 states with full Medicaid income limits lower than 100% of the FPL (listed in the eAppendix [available at ajmc.com]). Sensitivity analyses report results for all states. Additional sensitivity analyses examined whether results differed for plans that operated every year of our study period (2014-2019), referred to as established plans. Relative to plans entering or exiting the market, established plans may have been better positioned to translate payment increases into improved benefits.⁸

Main Outcomes

If MA plans invested in resources to improve outcomes for dually eligible beneficiaries, then these beneficiaries may have less need for acute services and institutional postacute or long-term care. Among all dually eligible beneficiaries, we measured annual total number of inpatient discharges, inpatient days, nursing home admissions (for skilled or long-term care), and associated nursing home days per year. We also measured annual mortality rates.

With higher payment rates, MA plans may have been able to make investments, such as better care coordination, that would improve the following outcomes associated with SNF stays: percentage of SNF admissions with stays that lasted more than 100 days, hospital readmission within 30 days following SNF admission, and mortality within 365 days following SNF admission.

Analysis Approach

Plan-level risk scores. We compared the postpolicy change in mean plan-level risk scores based on the share of plan members who were community-dwelling beneficiaries with full Medicaid. These risk scores represent how much payment rates are adjusted due to members’ age, health conditions, receiving Medicare due to disability, institutional residence, new enrollment in Medicare, and Medicaid status. We expected mean risk scores to increase more in plans in which at least 50% of members were community-dwelling beneficiaries with full Medicaid.

Health care use and quality outcomes. A DID analysis compared dually eligible beneficiaries with full Medicaid (the population that qualified for larger capitated payment rates) vs dually eligible beneficiaries with partial Medicaid. Our linear regression models included an indicator variable identifying whether a beneficiary had full Medicaid, indicator variables for each study year from 2014 to 2019, and interaction terms for these 2 sets of variables. For postperiod study years (2017-2019), these interaction terms identified a difference in outcomes between beneficiaries with full vs partial Medicaid after the policy change relative to the year prior to the policy change (2016). We report the mean of all postpolicy interaction terms as our estimate of whether the payment change was associated with outcome changes. All models were stratified by age group (< 65 years and ≥ 65 years) and included covariates for age and sex. Models for the overall population included county fixed effects. Models for the SNF cohort included covariates based on the beneficiary’s preceding hospital stay: length of stay and fixed effects for the diagnosis related group. Given the smaller sample size for the SNF cohort, we used state rather than county fixed effects. SEs were clustered at the beneficiary level in all models.

A DID analysis assumes that the intervention and control groups would have experienced the same changes in outcomes during the post period in the absence of the policy change. It is not possible to directly test this parallel trends assumption, but one indicator of whether this assumption is plausible is whether the intervention and control groups had similar trends in outcomes prior to the policy change. We examined the coefficients for the interaction terms of the indicator for full Medicaid participation with indicators for each prepolicy year. An F test assessed whether these prepolicy interaction terms equaled 0 to formally test whether each outcome had parallel trends.

Given our large sample size, some outcomes did not meet the parallel trends test even when a difference in trends was small in magnitude. To address this issue, we report the original DID estimates and also used the sensitivity approach developed by Rambachan and Roth to generate estimate bounds.^9,10 This approach provides an empirically derived estimate of how much differences in parallel trends might impact findings. Larger differences in pretrends would lead to more uncertainty about DID estimates and thus wider bounds relative to the original CIs. After examining different thresholds to determine when results were no longer significant, we report bounded estimates based on the assumption that the intervention and control groups experienced a postperiod difference in trends that was 0.2 times as large as the largest difference observed during the preperiod. Statistical analyses used Stata 18 (StataCorp LLC).

RESULTS

Relative to MA plans that served fewer beneficiaries with full Medicaid, MA plans in which at least 50% of members were community-dwelling beneficiaries with full Medicaid had larger increases in mean plan risk scores after the policy change (Figure 1). Risk scores averaged 1.41 points for the 2014-2016 period for plans in which the majority of members had full Medicaid and increased to 1.55 points over the 2017-2022 period. In contrast, risk scores averaged 1.07 and 1.08 points before and after the policy change, respectively, among plans with fewer full Medicaid beneficiaries. The postpolicy increase in risk scores was 0.13 points (95% CI, 0.10-0.15) higher, or 8.9% greater in relative terms, in plans with more full Medicaid beneficiaries.

In our main cohort, more than half (58%) of dually eligible MA enrollees 65 years and older had full Medicaid benefits (eAppendix Table 1), so their plans received higher capitated payment rates starting in 2017. Relative to partial Medicaid enrollees, beneficiaries with full Medicaid were older (mean age, 75.5 vs 74.5 years) and more likely to be Hispanic (42.9% vs 27.5%) or Asian American or other race/ethnicity (13.5% vs 5.3%). Among dually eligible beneficiaries younger than 65 years, 59% had full Medicaid. In this age group, beneficiaries with full Medicaid were younger (mean age, 49.6 vs 53.0 years) and slightly more likely to be Hispanic (19.9% vs 15.6%) or Asian American or other race/ethnicity (3.1% vs 1.9%) than their counterparts with partial Medicaid. In both age groups, the distribution of beneficiary characteristics was similar for beneficiaries who belonged to established MA plans. Demographic differences persisted but were less pronounced among SNF users with full and partial Medicaid (eAppendix Table 2).

Dually eligible beneficiaries with full Medicaid had moderately greater use of inpatient and SNF services than beneficiaries with partial Medicaid (eAppendix Table 1). For example, among beneficiaries younger than 65 years, the unadjusted number of inpatient days per beneficiary was 2.0 for beneficiaries with full Medicaid vs 1.9 for beneficiaries with partial Medicaid in 2017 and stayed fairly consistent throughout our study period (Figure 2).

DID Results: Main Cohort

The payment change was not associated with changes in any outcomes for beneficiaries younger than 65 years, including in sensitivity analyses for beneficiaries enrolled in established plans (Table and eAppendix Figure 1). Among beneficiaries 65 years and older in our main cohort, the payment change was associated with small decreases in inpatient days (–0.042 days; 95% CI, –0.077 to –0.007; P = .018), nursing home admissions (–0.003 admissions; 95% CI, –0.005 to –0.001; P = .005), nursing home days (–0.152 days; 95% CI, –0.286 to –0.018; P = .026), and mortality (–0.001 percentage points; 95% CI, –0.002 to 0.000; P = .014). Compared with the unadjusted means of these outcomes prior to the policy change for beneficiaries with full Medicaid, these differences represent less than a 3% change in relative terms. The DID analysis that focused on older beneficiaries in established plans found that the payment change was associated with small decreases in inpatient discharges (–0.007 discharges; 95% CI, –0.011 to –0.002; P = .006), inpatient days (–0.049 days; 95% CI, –0.093 to –0.005; P = .029), nursing home admissions (–0.004 admissions; 95% CI, –0.006 to –0.002; P = .001), nursing home days (–0.205 days; 95% CI, –0.363 to –0.046; P = .011), and mortality (–0.002 percentage points; 95% CI, –0.003 to 0.000; P = .008).

However, these results should be interpreted with caution given the small size of these effect estimates and preperiod trend differences in some outcomes between beneficiaries with full vs partial Medicaid (Table and eAppendix Figure 2). To understand how sensitive our results are to violating the parallel trend assumption, we bounded each DID estimate based on the magnitude of trend differences observed in the preperiod (Figure 3 and eAppendix Figure 3). If any concurrent differential change in the post period was only 0.2 times as large as the largest difference in trends observed during the preperiod, then these bounds suggest that the policy change was not associated with changes in inpatient or nursing home use among older beneficiaries in the main cohort or any outcomes among older beneficiaries in established plans. Findings for reduced mortality among older beneficiaries in the main cohort were slightly more robust: The differential change in the post period had to be 0.3 times as large as the largest difference in trends observed during the preperiod to suggest a null association (data not shown).

DID Results: SNF Cohort and Sensitivity Analyses

Among SNF users (eAppendix Table 3 and eAppendix Figures 4-6), none of the outcomes—mortality within 365 days of SNF admission, hospital readmission within 30 days of SNF admission, or SNF length of stay greater than 100 days—had a significant association with the payment change. When we repeated our analyses among beneficiaries living in all states, including those that offer full Medicaid to all beneficiaries with income below 100% of the FPL, there was a greater number of significant differences in trends during the preperiod between beneficiaries with full and partial Medicaid, so our findings were null after taking into account these pretrend differences (eAppendix Tables 4 and 5).

DISCUSSION

A new approach to MA risk adjustment increased capitated payment rates by boosting risk scores for beneficiaries with full Medicaid. In plans in which a majority of beneficiaries had full Medicaid, mean risk scores increased by 8.9% more after the policy change relative to the change observed in other plans. In DID analyses comparing beneficiaries with full and partial Medicaid, we found some evidence that this policy change was associated with small decreases in inpatient use, nursing home use, and mortality for beneficiaries 65 years and older. However, these results were not robust in sensitivity analyses that examined whether findings would hold when we loosened the assumption that beneficiaries with full and partial Medicaid experienced similar trends in outcomes during the post period. We found no evidence of significant changes for beneficiaries younger than 65 years or among SNF users.

These findings come amid debates about whether the federal government should curb MA overpayments and strengthen Medicare’s financial stability. CMS pays MA plans 22% more than it would cost to cover the same beneficiaries in traditional Medicare due to several factors, including CMS payment policies, favorable selection into MA, and plans’ coding practices that maximize payments.⁴ One barrier to reducing MA payments is the risk of unintended consequences for beneficiaries. Most prior studies examining the effects of payment changes do not focus on dually eligible beneficiaries, although some evidence suggests that MA insurers are more likely to offer plans targeted to dually eligible beneficiaries in areas with higher payment rates.¹¹ More broadly, evidence from other MA payment policies suggests that insurers partially use payment increases to offer features that make MA plans more attractive, including reduced premiums and, to a lesser extent, reduced cost-sharing and supplemental benefits such as hearing aid coverage.^8,12-14 Under payment decreases, the generosity of these features could decline.

Other work in the broader Medicare population suggests that there are opportunities to reduce MA payments without significantly reducing benefits or access for beneficiaries.^15,16 Reducing payments to MA plans under the Affordable Care Act was not associated with changes in health care access or affordability issues for MA enrollees,¹⁷ nor was it associated with declines in MA enrollment.¹⁸ Without evidence of improved outcomes with increased payments, it is reasonable to forecast that decreased payments may have minimal effect on dually eligible beneficiaries. Such optimism should be tempered with awareness that coverage may change in other ways that are harder to measure, including breadth of provider networks, availability of high-quality providers or care coordination, and use of prior authorization or step therapy. More transparent, explicit requirements regarding how MA plans allocate payment increases tied to risk adjustment for dually eligible beneficiaries could improve confidence that these funds are used to improve health outcomes.

Limitations

Our study has limitations, the greatest being differences in preperiod trends for some outcomes. We describe our findings as an association, not the effect of the payment change, given these limitations. However, DID results can still be informative even without parallel trends during the preperiod.^19,20 Other limitations include our focus on QMB enrollees because we expected this population to be less heterogeneous than the entire dually eligible population. Our results might not generalize to dually eligible beneficiaries with higher income. Second, some dually eligible beneficiaries may have qualified for full vs partial Medicaid through alternative pathways, such as waiver programs, that we cannot observe in our data. Third, we identified SNF users based on a nursing home admission in MDS data within 7 days of hospital discharge. MDS data cannot confirm whether MA plans paid for SNF use. Fourth, we had limited data available to control for differences in underlying health among beneficiaries. However, our SNF analyses incorporated more detailed data on the reason for the prior hospital stay and length of the hospital stay.

Finally, beneficiary-level data to examine how the policy change increased risk scores (and thus payments) were not available, so we present plan-level changes in risk scores. We could not examine whether risk scores differentially changed within plans. Increases in plan-level risk scores may reflect the direct impact of the policy (higher risk scores for beneficiaries who would have lower scores under the prior risk-adjustment approach) and potential indirect effects, such as more enrollment among beneficiaries with greater comorbidity or increased coding of comorbidities among beneficiaries with full Medicaid. Future research with individual-level risk score data should identify potential heterogeneous effects within plans and investigate long-term impacts of this policy change.

CONCLUSIONS

Despite a substantial adjustment in how CMS paid MA plans to cover dually eligible beneficiaries with full Medicaid, we found little evidence of changes in health care use or outcomes for this population. In assessing whether payment levels for MA plans are appropriate, policy makers should consider whether higher payments are linked to better outcomes.

Author Affiliations: Department of Health Policy, Vanderbilt University School of Medicine, Vanderbilt University Medical Center (LMK, EMA), Nashville, TN; Leonard Davis Institute of Health Economics, University of Pennsylvania (EMA), Philadelphia, PA; Department of Health Services, Policy and Practice, Brown University School of Public Health (DK, DJM, HV, ANT), Providence, RI; Transformative Health Systems Research to Improve Veteran Equity and Independence Center of Innovation, Department of Veterans Affairs Providence Health Care System (ANT), Providence, RI.

Source of Funding: This study was supported by the National Institute on Aging under grant P01AG027296. The views expressed in this article are those of the authors and do not necessarily reflect the position or policy of the Department of Veterans Affairs or the United States government.

Author Disclosures: Dr Keohane has received payments from Health Affairs Forefront for blog posts. Dr Achola has received grant funding from the National Institute on Aging. Dr Meyers has received personal fees as a senior advisor to the Center for Medicare and Medicaid Innovation and grant funding from the National Institute on Aging. The remaining authors report no relationship or financial interest with any entity that would pose a conflict of interest with the subject matter of this article.

Authorship Information: Concept and design (LMK, EMA, DJM, HV, ANT); acquisition of data (ANT); analysis and interpretation of data (LMK, EMA, DK, DJM, HV, ANT); drafting of the manuscript (LMK, DK); critical revision of the manuscript for important intellectual content (EMA, DK, DJM, ANT); statistical analysis (LMK); obtaining funding (ANT); and administrative, technical, or logistic support (HV).

Address Correspondence to: Laura M. Keohane, PhD, MS, Department of Health Policy, Vanderbilt University School of Medicine, 2525 West End Ave, Ste 1200, Nashville, TN 37203. Email: laura.keohane@vanderbilt.edu.

REFERENCES

1. Data analysis brief: comparing managed care enrollment trends among dually eligible individuals to Medicare-only beneficiaries, 2012 through 2021. CMS. October 2022. Accessed January 15, 2025. https://www.cms.gov/files/document/managedcareenrollmenttrendsdatabrief2012-2021.pdf

2. 2024 Medicare Advantage enrollment trends. ATI Advisory. September 2024. Accessed January 15, 2025. https://atiadvisory.com/resources/wp-content/uploads/2024/09/2024-Dual-Eligible-MA-Enrollment-Trends.pdf

3. Data Book: Beneficiaries Dually Eligible for Medicare and Medicaid. Medicare Payment Advisory Commission; Medicaid and CHIP Payment and Access Commission. February 2022. Accessed January 15, 2025. https://www.macpac.gov/wp-content/uploads/2022/02/Beneficiaries-Dually-Eligible-for-Medicare-and-Medicaid-February-2022.pdf

4. Medicare Payment Advisory Commission. Chapter 12: the Medicare Advantage program: status report. In: Report to the Congress: Medicare Payment Policy. Medicare Payment Advisory Commission. March 2024. Accessed January 15, 2025. https://www.medpac.gov/wp-content/uploads/2024/03/Mar24_Ch12_MedPAC_Report_To_Congress_SEC-1.pdf

5. Jacobs PD, Kronick R. The effects of coding intensity in Medicare Advantage on plan benefits and finances. Health Serv Res. 2021;56(2):178-187. doi:10.1111/1475-6773.13591

6. Chernew ME. Getting the most from payments to Medicare Advantage health plans—thoughts on the controversy. JAMA Health Forum. 2022;3(8):e222833. doi:10.1001/jamahealthforum.2022.2833

7. Keohane LM, Stevenson DG, Stewart L, Thapa S, Freed S, Buntin MB. Risk adjusting for Medicaid participation in Medicare Advantage. Am J Manag Care. 2020;26(8):e258-e263. doi:10.37765/ajmc.2020.44076

8. Duggan M, Starc A, Vabson B. Who benefits when the government pays more? pass-through in the Medicare Advantage program. J Public Econ. 2016;141:50-67. doi:10.1016/j.jpubeco.2016.07.003

9. Rambachan A, Roth J. A more credible approach to parallel trends. Rev Econ Stud. 2023;90(5):2555-2591. doi:10.1093/restud/rdad018

10. Bravo MC, Roth J, Rambachan A. HONESTDID: Stata module implementing the HonestDiD R package. EconPapers. Updated March 22, 2024. Accessed January 15, 2025. https://EconPapers.repec.org/RePEc:boc:bocode:s459138

11. McGarry BE, Layton TJ, Grabowski DC. The effects of plan payment rates on the market for Medicare Advantage Dual-Eligible Special Needs Plans. Health Serv Res. 2019;54(5):1137-1145. doi:10.1111/1475-6773.13170

12. Cabral M, Geruso M, Mahoney N. Do larger health insurance subsidies benefit patients or producers? evidence from Medicare Advantage. Am Econ Rev. 2018;108(8):2048-2087.

13. Murray RC, Meyers DJ, Fuse Brown EC, Williams TC, Ryan AM. Association between the Medicare Advantage quartile adjustment system and plan behavior and enrollment. JAMA Health Forum. 2024;5(1):e234822. doi:10.1001/jamahealthforum.2023.4822

14. Pelech D, Song Z. Private Medicare plans’ responses to benchmark changes and competition before and after the Affordable Care Act’s payment cuts. Health Serv Res. 2025;60(2):e14392. doi:10.1111/1475-6773.14392

15. Song Z. What will cuts to Medicare Advantage payments do to enrollment? JAMA Health Forum. 2023;4(6):e231693. doi:10.1001/jamahealthforum.2023.1693

16. Chernew ME, Miller K, Petrin A, Town RJ. Reducing Medicare Advantage benchmarks will decrease plan generosity, but those effects will likely be modest. Health Aff (Millwood). 2023;42(4):479-487. doi:10.1377/hlthaff.2022.01031

17. Skopec L, Aarons J, Zuckerman S. Did Medicare Advantage payment cuts affect beneficiary access and affordability? Am J Manag Care. 2019;25(9):e261-e266.

18. Schwartz AL, Kim S, Navathe AS, Gupta A. Growth of Medicare Advantage after plan payment reductions. JAMA Health Forum. 2023;4(6):e231744. doi:10.1001/jamahealthforum.2023.1744

19. Roth J, Sant’Anna PHC, Bilinski A, Poe J. What’s trending in difference-in-differences? a synthesis of the recent econometrics literature. J Econom. 2023;235(2):2218-2244. doi:10.1016/j.jeconom.2023.03.008

20. Bilinski A, Hatfield LA. Nothing to see here? non-inferiority approaches to parallel trends and other model assumptions. arXiv. Preprint posted online May 8, 2018. doi:10.48550/arXiv.1805.03273