An Evaluation of Competing Risks in Studies of Perinatal Mortality and Birth Defects

Abstract

Globally, each year over 4 million infants die during the perinatal period, which includes stillbirth and neonatal death, yet little investment has been made in research or preventative efforts. Birth defects are common among perinatal deaths: major birth defects are identified among 15 to 20% of stillbirths and 20% infant deaths. Due to the interconnections between perinatal outcomes, studies of both perinatal deaths and of risk factors for birth defects may be biased by the occurrence of competing events which may deplete the susceptible population at risk, alter the category of a death without changing the risk of mortality, or may lead to biased risk factor assessments when analyses are conducted among survivors. We therefore sought to evaluate the impact of competing risks in studies of perinatal mortality and birth defects. First, we described the timing and etiology of stillbirths and neonatal deaths in the US during 2014. Then we examined the risk of stillbirth among fetuses with major birth defects and evaluated the effect of termination for pregnancy as a competing risk on observed stillbirth risk estimates. Finally, we evaluated whether risk factor studies of birth defects are biased when conducted among live births only. We found that the risk of stillbirth and first day mortality were higher than late neonatal deaths at all gestational ages and that stillbirth and first day deaths share have a greater etiological overlap than first day deaths and later neonatal deaths. Fetuses with major birth defects were found to have high risks of stillbirth, and that termination of pregnancy may lead to underestimates of stillbirth risk by depleting the susceptible population at risk. Risk factor studies of birth defects conducted only among live births were not found to be meaningfully biased in most situations, but studies of high mortality defects where the exposure is also very strongly associated with termination of pregnancy and stillbirth may lead to biased results. In studies of perinatal mortality and birth defects, competing risks may act to induce bias under certain circumstances.