Data linkage methods used in maternally‐linked birth and infant death surveillance data sets from the United States (Georgia, Missouri, Utah and Washington State), Israel, Norway, Scotland and Western Australia

Abstract

In this paper we describe the methods used to link birth and infant mortality and morbidity surveillance data sets into sibships using deterministic or multistage probabilistic linkage methods. We describe nine linked data sets: four in the United States (Georgia, Missouri, Utah and Washington State), and four elsewhere (Scotland, Norway, Israel and Western Australia). Norway and Israel use deterministic methods to link births and deaths into sibships. The deterministic linkage is usually dependent on the availability of national identification numbers. In both countries they assign these numbers at birth. Deterministic linkage is usually highly successful, and the major problem is the validation of linkages. In the United States, Western Australia and UK linkage is multistage and probabilistic. This approach is usually dependent on the calculation linkage weights from sociodemographic variables. The success rates of probabilistic methods are above 80%. Maternally-linked perinatal data open new vistas for epidemiological research. Recurrence of poor perinatal outcomes is more appropriately studied using longitudinally-linked data sets. In addition, the emergence of risk factors and the recurrence of risk factors can be studied.