Premium
Linkage of mother–baby pairs in the German Pharmacoepidemiological Research Database
Author(s) -
Garbe Edeltraut,
Suling Marc,
Kloss Sebastian,
Lindemann Christina,
Schmid Ulrike
Publication year - 2011
Publication title -
pharmacoepidemiology and drug safety
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.023
H-Index - 96
eISSN - 1099-1557
pISSN - 1053-8569
DOI - 10.1002/pds.2038
Subject(s) - medicine , linkage (software) , record linkage , pharmacoepidemiology , pregnancy , german , family medicine , health care , pediatrics , database , medical prescription , environmental health , population , nursing , genetics , history , archaeology , gene , computer science , biology , economics , economic growth
Abstract Purpose Administrative healthcare databases are increasingly being used to investigate potential drug risks in pregnancy. Our study aimed to develop an algorithm for linkage of mother–baby pairs (MBPs) in the German Pharmacoepidemiological Research Database (GePaRD) as a prerequisite for such studies. Methods GePaRD contains sociodemographic data, drug dispensations, ambulatory, and hospital information on more than 14 million insurants from four German statutory health insurances (SHIs) covering all regions in Germany. Linkage was based on co‐insurance information of the newborn with the potential mother (direct linkage) or of both potential mother and newborn with the potential father (indirect linkage). Linkage is not possible if the baby is co‐insured with the father and the mother is self‐insured. Further information on birth or childbed was used to validate the potential mother as true mother in MBP. Descriptive comparisons between linked and unlinked mothers were conducted. Results Of 323 993 newborns identified between 2004 and 2006, 250 355 (77.3%) could be linked in MBP. Of those, 189 702 (75.8%) MBP were based on direct linkage. Mean age was similar in linked (31.1 years, standard deviation (SD = 5.4) and unlinked (31.8 years, SD = 5.5) mothers as was the proportion of caesarean sections in both groups (28.9% vs. 29.3%). Conclusions The developed algorithm permits linkage of a great number of newborns with their mothers and creates a potential data source for investigation of drug risks in pregnancy. Further validation studies are needed also including information on pregnancies not resulting in live births. Copyright © 2010 John Wiley & Sons, Ltd.