The limitations of using Medicaid administrative data in abortion research

To identify limitations of abortion data in national Medicaid claims files by comparing abortion counts in Medicaid claims data with state abortion estimates.

We used procedure (Current Procedural Terminology/Healthcare Common Procedure Coding System) and drug (National Drug Code) codes to identify abortion claims in 2009 and 2010 Medicaid Analytic eXtract (MAX) and 2020 Transformed Medicaid Statistical Information System Analytic File (TAF) data. We compared the number of abortions in MAX and TAF to the number of expected abortions covered by Medicaid overall and by state. Based on recent published research, we estimated expected Medicaid-covered abortions as 62% of total abortions in states using state funds to cover abortion services for Medicaid enrollees and 0.9% in states that follow Hyde restrictions.

MAX data identified 11% (38,668/345,480) of expected Medicaid-covered abortions in 2009 and 13% (44,528/330,801) of expected Medicaid-covered abortions in 2010. In 2020 TAF data, we found 25% (69,728/279,048) of the expected Medicaid-covered abortions. Among the 16 states that used state funds to cover abortions for Medicaid enrollees in 2020, the majority had <10% of expected Medicaid-covered abortions (n = 8). Three states had between 10% and 50% of expected abortions. Four states had between 51% and 75% of expected abortions. One state had insufficient data for reporting.

Abortion claims in MAX/TAF are an undercount of abortions covered by Medicaid, and this undercount varies across states. Variation in reporting across states and across time likely introduces bias into research trying to use MAX/TAF abortion claims across states and time. Researchers should use extreme caution when using MAX/TAF for abortion-related research.

Researchers should use caution when using the Medicaid Analytic eXtract and Transformed Medicaid Statistical Information System Analytic Files for abortion-related research questions.