BACKGROUND:
The Michigan Childhood Immunization Registry (MCIR) is a statewide registry which has captured immunization information for 2.8 million children since 1997. MCIR is used widely by public and private providers for assessment and reminder / recall and therefore data completeness is an important concern.

OBJECTIVE:
Assess the completeness of MCIR vaccination records using Medicaid administrative claims data.

METHOD:
MCIR records for primary series vaccinations were obtained for children 0-6 years of age in 2002 who were Medicaid enrollees. Medicaid enrollment data and administrative claims data were obtained and merged by Medicaid identifier with MCIR records. Univariate and bivariate analyses were conducted at the vaccination event level (defined as one or more vaccination for a child on a given date) to describe the concordance between MCIR vaccination records and Medicaid claims.

RESULT:
Overall, there were 185,425 vaccination events; 80.5% of events were found in MCIR,69.5% were found in Medicaid claims, and 50.0% were found in both data sources. Of the 149,187 vaccination events in MCIR, 33% had one or more dose with an invalid code which is intended to distinguish vaccination history from administered doses; 30% of events had all missing codes for this field. Significant variations in concordance between the two data sources was observed by child's age, race, gender, county of residence, and health plan (p<.001).

CONCLUSION:
Missing vaccination data for Medicaid enrollees in MCIR impacts its usefulness for assessment and patient reminder / recall. Accurate determination of provider identity for doses administered is hampered by missing data, limiting the degree to which MCIR data can be reliably used for provider profile analyses. These findings point to aspects of immunization registries that warrant further attention.

LEARNING OBJECTIVES:
Participants will learn how administrative claims data can be used to verify immunization registry completeness and the potential implications of disconcordance between these data sources.