Late Medicaid expansion was associated with higher Medicaid coverage: source-specific evidence from Oklahoma,...

Evidence remains limited on Medicaid expansions implemented after 2020 and on whether increases in Medicaid coverage were accompanied by changes in specific insurance sources.

This repeated cross-sectional study used 2015-2024 American Community Survey's 1-year public use microdata sample data for adults ages 18-64. Weighted state-year difference-in-differences models compared the 2021 expansions in Oklahoma and Missouri and the 2023 expansion in North Carolina with ten states that had not implemented full ACA Medicaid expansion through 2024. Models included state and year fixed effects. Conventional state-clustered 95% confidence intervals were supplemented with exact state-level randomization inference. Medicaid, uninsurance, employer-sponsored insurance, direct-purchase coverage, Medicare, other public coverage, and multiple coverage were estimated separately.

Medicaid coverage increased by 3.389 percentage points in Oklahoma-Missouri (95% CI, 1.024 to 5.754; exact randomization-inference P = 0.030) and by 3.332 points in North Carolina (95% CI, 2.823 to 3.840; P = 0.091). The largest estimates occurred among adults below 138% of the federal poverty level. In Oklahoma-Missouri, direct-purchase coverage declined by 1.968 points (95% CI, -2.693 to -1.244; P = 0.015), while employer-sponsored insurance and multiple coverage showed no corresponding assignment-based evidence of change. The North Carolina direct-purchase estimate was not supported by full-period exact randomization inference (P = 0.182). The Oklahoma-Missouri Medicaid estimate increased from 1.880 points in 2021-2022 to 4.887 points in 2023-2024 and changed little after adjustment for measured Medicaid-unwinding indicators. Synthetic estimates were similar to the conventional DID estimates, although pre-expansion diagnostics and treatment-specific-trend models indicated sensitivity to the assumed counterfactual trend.

Late Medicaid expansion was associated with increased Medicaid coverage, particularly among adults below 138% of the federal poverty level. The Oklahoma and Missouri findings were consistent with partial substitution away from direct purchase coverage and a larger Medicaid coverage association in later years. The limited number of treated states and concerns about trends before expansion warrant cautious interpretation.