KEYWORD1:
Registry master patient index; match algorithms

BACKGROUND:
Immunization registries have great potential for interfacing with community-based databases, but only if the records can be matched. An efficient strategy for matching immunization registry records is critical to optimizing its linkages with other databases.

OBJECTIVE:
Compare algorithms for matching records for children identified through community based programs with records in a network immunization registry.

METHOD:
In April 2002, we searched for immunization registry records for 1,082 children <5 years old that had enrolled in a community immunization outreach program. The three identifiers were child’s first and last name and date of birth. These two parameters were used to create hierarchical algorithms systematically allowing for discrepancies between the identifiers in the two databases. Matches were classified as possible, probable, or exact. Manual verification was used to determine if possible matches were probable.

RESULT:
195 records (18% of 1082) had an exact match. The least restrictive algorithm (allows 3 day birthdate window and a two-letter partial name match) yielded 5,967 possible matches for 820 records. 7.3 possible matches per record were manually reviewed, of which 184 (0.3% of 5967) were verified as probable matches. The hierarchical algorithm (3 rounds of matching) yielded a total of 839 possible matches for 462 records, reducing manual matches to 1.8 per record. The hierarchical algorithm yielded a final match rate of 35%.

CONCLUSION:
Application of a hierarchical matching algorithm rather than a single algorithm is a more efficient strategy for matching patient records from different databases. It is an easily implemented strategy to improve match rates.
LEARNINGOBJECTIVES:
Compare hierarchical matching algorithms for improving registry match rates.