Census Bureau Response to ACS 30-day FRN Health Insurance Comment

Census Bureau Response to the American Community Survey 30-Day Federal Register Notice Health Insurance Comment

The Census Bureau and our interagency subcommittee for health insurance thank the Urban Institute for the feedback provided in their letter dated April 25, 2024. After careful consideration, we have concluded that the long-term gain of implementing the revised health insurance coverage question in the 2025 American Community Survey (ACS) outweighs any potential short-term need for one additional year of continuity to track the Medicaid unwinding. Our key Federal agency stakeholders agree that improved estimates of the uninsured population and a reduction in misreports of multiple types of coverage are critical benefits of Test Version 2.

The ACS is the principal source of detailed health insurance coverage information for state and sub-state geographies due to its large sample size. The ACS added a question on health insurance in 2008 to provide states and localities with reliable data on insurance coverage so that they could develop outreach strategies for the uninsured population who may be eligible for enrollment in certain insurance programs, but who do not currently have that coverage. The results of the recent 2022 ACS Content Test found that Test Version 2 provided a more reliable measurement of uninsurance because it provides an option for respondents to select “no insurance.”¹ It also reduced instances of multiple coverage types, which may reflect a reduction in respondent confusion about how to categorize their coverage.

The purpose of testing the revised health insurance question was to enhance question reliability and validity overall. While we strive for continuous improvement in all our measures, there is no specific mandate for ACS to measure Medicaid enrollment and improving the measurement of public coverage was not defined as a higher rate of Medicaid coverage alone. The results of the content test showed that the overall estimate for public coverage was not statistically different between treatments. Results of the test also found that Medicaid-only responses were also not significantly different between the test and control versions. Allowing respondents to report their type of coverage without misreporting one coverage as two types is considered an improvement in measurement. The changes in response order and wording were intended to reduce the reporting of Marketplace plans as Medicaid or as both Medicaid and direct purchase.

We recognize that the impact of the Medicaid unwinding will continue into and beyond 2025. Some of the largest impacts of the unwinding are occurring in 2024, not 2025, so those will be captured in the 2024 ACS, before the transition to the updated instrument. This includes the mandatory continuous enrollment for children, which many states began to implement prior to 2024. Studies to date have also shown that survey respondents during this period may not have known they were still enrolled in Medicaid because states were not required to inform them, therefore, they may have responded accurately to the question from their perspective.^2,3 This will be the case in either version of the question, hampering our ability to clearly understand the unwinding process using survey data.

The next important step we plan to take is additional research into the consistency of Medicaid reporting in surveys with administrative records from the Centers for Medicare and Medicaid Services (CMS). We plan to conduct this match study with the Test 1 and Test 2 panels from the 2022 ACS Content Test once the 2022 Medicaid records needed for the project are available, sometime around mid-summer 2024.

1. Viver-Hernandez A, Keisler-Starkey K, Stern SM. 2022 American Community Survey Content Test Evaluation Report: Health Insurance Coverage. November 20, 2023.

2. McIntyre A, Smith RB, Sommers BD. Survey-Reported Coverage in 2019-2022 and Implications for Unwinding Medicaid Continuous Eligibility. JAMA Health Forum. 2024 Apr 5;5(4):e240430.

3. Hest, Robert, Lukanen Elizabeth, Blewett, Lynn. Medicaid Undercount Doubles, Likely Tied to Enrollee Misreporting of Coverage. December 2022. Medicaid_Undercount 2020-2021.pdf (shadac.org)