Improving completeness of ascertainment and quality of information for pregnancies through linkage of administrative and clinical data records

PurposeBirth cohorts are a common tool used in epidemiological studies about pregnancy; yet these datasets systematically miss pregnancies that are spontaneously lost or terminated. This study examined the feasibility of linking administrative and clinical datasets from Alberta Canada to identify a pregnancy cohort that includes spontaneous and medical pregnancy losses.MethodsDeterministic linkage was used to link data from twelve clinical and administrative datasets for women who conceived between November 2007 and February 2008. Descriptive statistics were used to characterize the relative contribution of each dataset to the overall dataset.ResultsOverall, 6,477 unique pregnancies were eligible for inclusion, resulting in a live birth rate of 94.1%, a stillbirth rate of 0.5%, a fetal death rate of 4.1%, and an estimated 1.3% of the cohort moving out of the study area. No single dataset could identify all pregnancies. Individual databases identified 2.0–99.1% of the cohort. Fetal deaths were most frequently identified in outpatient physician claims, emergency room visits, ultrasound data, or from the cytogenetic laboratory.ConclusionsLinkage of clinical and administrative databases to identify pregnancy is feasible and can overcome many limitations associated with the use of a single dataset; however, fetal deaths continue to be under-ascertained.