Database Matching Process

The process of database matching involves:

Evaluating databases for linkage in terms of validity, logic and completeness.

Identifying "matching fields" - those fields occurring in two or more databases which are likely to be identical/similar if records pertain to the same individual/event/service.

Preparing databases for linkage, standardizing matching fields using the same coding system for all databases, and creating additional fields for blocking and matching data.

Creating and running match programs that:

• weight the matching fields according to the probability that data in all databases will be identical/similar;
• assign scores to the fields in each potential pair of records based upon weight and degree of similarity;
• sum field scores to create a total match score for each pair;
• set cutoff scores above which pairs will be matched and below which they will not be matched;
• segregate cases with borderline match scores for manual review;
• extract data from matching software; and
• import in the desired format (e.g., DBF, SAS, ASCII).

Assessing linked record pairs, performing case mix validation, and analyzing match scores.

Implementing use of linked data.