Probabilistic record linkage and an automated procedure to minimize the undecided-matched pair problem

Probabilistic record linkage allows the assembling of information from different data sources. In this article, we present a procedure in case a one-to-one relationship between records in different files is expected but is not found only by applying the probabilistic record linkage methodology. Our data were births and infant deaths from the 1998-birth cohort whose mother's place of residence was the City of São Paulo at the time of birth. Our assumption was that pairs for which a one-to-one relationship was obtained, and a best-link was found with the highest combined weight would be considered as uniquevocally matched pairs or gold-standard and should then provide information in order to decide about pairs in which such a relationship could not be established. For example, we observed that the for the unequivocally matched pairs a clear and expected relationship between differences in dates of death and birth registration could be assessed. As a result, such a relationship was used to help solving the remaining pairs for which a one-to-one relationship could not be found. Indeed, we reduced the number of non-uniquely matched records and even though we could not establish a one-to-one relationship for every single death we reduced the number of uncertain. We suggest that future research using record linkage should use combined strategies from results from first record linkage runs before a full clerical review (the standard procedure under uncertainty) in order to most efficiently (and less costly), retrieve record matches.