Skip to main content

Table 2 Comparison of linkage strategies for the bespoke algorithm and ReclinkIII

From: Evaluation of record linkage of two large administrative databases in a middle income country: stillbirths and notifications of dengue during pregnancy in Brazil

  Bespoke algorithm ReclinkIII
Manipulation of names • Multiple variables created for first name, second name, and last name • Variables created for first and last name
Blocking • Municipality • Soundex for name + municipality
Calculation of m and u probabilities m-probability: calculated using true-matches in gold-standard
u- probability: calculated using non-matches in SINAN
m-probabilities = 0.9
u-probabilities = 0.1
Match weight calculation • Separate weights calculated for the five most common names
• Agreement on name classified using Jaro-Winkler string comparator
• Different weights calculated according to closeness of age.
• Did not account for common names
• Levenshtein string comparator
• Did not account for timing issues