Mary Ann Evans^1,, Vivian Siu², Krista Markwardt³, Sarah Hargand⁴

Self-reported payer data from the birth certificate is frequently used to identify Medicaid births, but evaluation of validity and reliability is lacking due to the challenges of gaining access to Medicaid data and data complexity. Oregon birth certificate and Medicaid paid claims were linked and compared to assess the validity and reliability of birth certificate reported payer data. Birth certificate data underestimated Medicaid paid births and underestimation grew from 18% in 2008 to 25% by 2014. Birth certificate data had moderate to substantial agreement with Medicaid claims, but reliability declined over time from a Kappa score of .82 to .75. Reliability was substantially lower for younger, less educated, Hispanic, American Indian or Alaskan Native (AIAN), Black or African American (BAA), and Native Hawaiian or Pacific Islander (NHPI) women. Specificity and positive predictive value were above 90% over time and by age, ethnicity, race and education. Sensitivity and negative predictive values declined from 84% and 87% in 2008 to 78% in 2014 and were lower than specificity and positive predictive values. Sensitivity values were higher for younger, less educated, Hispanic, AIAN, BAA and NHPI women. Negative predictive values were lower for younger, less educated, Hispanic, AIAN, BAA and NHPI women. The results document the limitations of birth certificate and strengths of Medicaid data for counting Medicaid paid births and allocating resources for community programs.

birth certificate payer, medicaid births, medicaid paid claims, validity, reliability