Background
Selection, outcome and publication biases are well described in case reports and case series but may be less of a problem early in the appearance of a new disease when all cases might appear to be worth publishing.
Objective
To use a prospectively collected database of primary sources to compare the reporting of COVID-19 in pregnancy in case reports, case series and in registries over the first 8 months of the pandemic.
Study Design
:
MEDLINE, Embase and Maternity and Infant Care databases were searched from 22 March to 5 November 2020, to create a curated list of primary sources. Duplicate reports were excluded. Case reports, case series and registry studies of pregnant women with confirmed COVID-19, where neonatal outcomes were reported, were selected and data extracted on neonatal infection status, neonatal death, neonatal intensive care unit admission, preterm birth, stillbirth, maternal critical care unit admission and maternal death.
Results
149 studies comprising 41,658 mothers and 8,854 neonates were included. All complications were more common in case reports, and in retrospective series compared with presumably prospective registry studies. Extensive overlap is likely in registry studies, with cases from seven countries reported by multiple registries. The UK Obstetric Surveillance System was the only registry to explicitly report identification and removal of duplicate cases, although five other registries reported collection of patient identifiable data which would facilitate identification of duplicates.
Conclusions
Since it is likely that registries provide the least biased estimates, the higher rates seen in the other two study designs are probably due to selection or publication bias. However even some registry studies include self- or doctor-reported cases, so might be biased, and we could not completely exclude overlap of cases in some registries.