Changes in data management contribute to temporal variation in gestational duration distribution in the Swedish Medical Birth Registry

Journal article, 2020

Multiple factors contribute to gestational duration variability. Understanding the sources of variability allows to design better association studies and assess public health measures. Here, we aimed to assess geographical and temporal changes in the determination of gestational duration and its reporting in Sweden between 1973 and 2012. Singleton live births between 1973 and 2012 were retrieved from the Swedish Medical Birth Register. Gestational duration trends in percentiles and rates of pre- and post-term deliveries were analyzed by plotting the values over time. Temporal changes in gestational duration based on ultrasound and last menstrual period (LMP) estimation methods were compared. Intervals between LMP date and LMP-based due date were analyzed to assess changes in expected gestational duration. In total, 3 940 577 pregnancies were included. From 1973 until 1985, the median of gestational duration estimated based on LMP or ultrasound decreased from 283 to 278 days, and remained stable until 2012. The distribution was relatively stable when ultrasound-based estimates were used. Until the mid-1990s, there was a higher incidence than expected of births occurring on every seventh gestational day from day 157 onward. On an average, these gestational durations were reported 1.8 times more often than adjacent durations. Until 1989, the most common expected gestational duration was 280 days, and thereafter, it was 279 days. The expected gestational duration varied from 279 to 281 days across different Swedish counties. During leap years, the expected gestational duration was one day longer. Consequently, leap years were also associated with significantly higher preterm and lower post-term delivery rates than non-leap years. Changes in data handling and obstetrical practices over the years contribute to gestational duration variation. The resulting increase in variability might reduce precision in association studies and hamper the assessment of public health measures aimed to improve pregnancy outcomes.