In-Database Fuzzy and Partial Matching for Unmapped Codes — suggest_source

Uses DuckDB's native string similarity functions (jaro_winkler_similarity) and substring/token matching to suggest candidate standard OMOP concepts for unmapped source values (concept_id = 0).

Usage

suggest_source_mappings(
  con,
  unmapped_table = NULL,
  domain = "Drug",
  min_similarity = 0.85,
  vocab_db = NULL,
  limit_per_code = 3,
  source_column = NULL,
  output_table = NULL
)

Arguments

con: Active DuckDB DBI connection.
unmapped_table: Table name or data.frame with unmapped entries.
domain: Domain ID ('Drug', 'Condition', 'Measurement', 'Procedure').
min_similarity: Minimum similarity threshold (default 0.85).
vocab_db: Optional attached vocabulary schema/database name.
limit_per_code: Maximum number of suggestions per source code (default 3).
source_column: Optional explicit column name for source value.
output_table: Optional table name to persist results.

Value

A data.frame of ranked mapping suggestions.