Stillbirth: are we making more progress than we think? A retrospective cohort study

Objective To quantify how the changing stillbirth risk profile of women is affecting the interpretation of the stillbirth rate. Design A retrospective, population-based cohort study from 1983 to 2018. Setting Victoria, Australia. Population A total of 2 419 923 births at >= 28 weeks of gestation. Methods Changes in maternal characteristics over time were assessed. A multivariable logistic regression model was developed for stillbirth, based on maternal characteristics in 1983-1987, and used to calculate individual predictive probabilities of stillbirth from the regression equation. The number of expected stillbirths per year as a result of the change in maternal demographics was then calculated, assuming no changes in care and in the associations between maternal characteristics and stillbirth over time. Main outcome measure Stillbirth. Results Compared with 1983-1987, there were more women in older age groups giving birth, more nulliparous women, more indigenous women and women born in Oceania, Asia and Africa, more multiple pregnancies and more women with pre-existing diabetes in 2014-2018. Despite this, the rate of stillbirth fell from 5.42 per 1000 births in 1983 to 1.72 per 1000 births in 2018 (P < 0.001). Applying the multivariable logistic regression equation, derived from the 1983-87 data, to each year, had there been no changes in care or in the associations between maternal characteristics and stillbirth, the rate of stillbirth would have increased by 12%, from 4.94 per 1000 in 1983 to 5.54 per 1000 in 2018, as a result of the change in maternal characteristics. Conclusions Population rates of stillbirth are falling faster than is generally appreciated. Tweetable abstract Population reductions in stillbirth have been underestimated as a result of changing maternal characteristics.