New model forecasts birth dates

The model takes account of individual factors related to the mother to calculate the most likely date on which she will give birth.

t was 16 years ago, when he himself was about to become a parent, that Mikael Elinder began to wonder about the accuracy of the ultrasound prediction of birth date, a method based the size of the foetus. But how significant is this in reality?

“If one is at all interested in statistics, as I am, one can’t help but wonder what the date is based on. Is it a mean value? Is the baby equally likely to born earlier or later? This frustrated me greatly and I thought that this was something parents should be able to find out,” says Elinder.

The very fact that first-time mothers are told that they are likely to give birth post-term suggests that there is room for improvement in the form of more reliable and more comprehensible forecasts. It would, however, be several years before Elinder was able to make his own forecasts.

Valuable answers in the birth register

Together with colleague Oscar Erixson, in early 2020 Elinder began to look at figures from the National Medical Birth Register. This provided them with data on 1.7 million Swedish pregnancies, allowing them to study how factors such as the mother’s age and height and whether she had previously given birth affect the length of pregnancy.

“This revealed that in nine out of ten cases a more likely birth date could be forecast than the one predicted by ultrasound.”

Thanks to the National Medical Birth Register, they were able to develop a model for preparing individual forecasts and in October 2021 the website Pregspect.com was launched. The service allows curious expectant parents to obtain a forecast of the date they are most likely to give birth.

“What we do is look at a combination of factors, such as whether it’s a first-time mother, whether the foetus is male, whether the woman is young, has a low BMI or is a smoker. The model then calculates the most likely date on which that individual baby will be born.”

A more accurate model

Even if the service offered by the researchers can provide a more accurate forecast of birth date than one is likely to receive during prenatal care, the length of a pregnancy is impossible to predict with certainty. So, why should someone use their model? Mikael Elinder draws parallels with other forecasting tools available on the market.

“No one wants to use the next best weather app or the next best election forecast. Why should expectant parents be any different? Why simply trust the date given at an ultrasound examination when there is a more accurate model?”

He goes on to explain that women who have used the service have been reassured to be given a different date than the one they received at the ultrasound.

“Much of the feedback we receive suggests that they feel more relaxed when they are not fixated on a given date. When the first date passes, it doesn’t feel that they are post-term.

The result of economics research

Interest in the service has been enormous. Since its launch, some 18,000 pregnant women have entered their details. Now the plan is to market the forecasts internationally. Elinder also sees considerable opportunities for collaboration with pregnancy apps and for developing models to assist healthcare providers, especially when forecasting staffing needs on maternity wards.

While Elinder and Erixson have started a company alongside their regular duties at the Department of Economics and The Institute for Housing and Urban Research, the model used to forecast birth dates is very much a product of their work at Uppsala University. Both have conducted research into health economics and their focus is currently on healthcare resources, healthcare planning and birth outcomes.

“As an economist, one naturally has an interest in statistics. Health economics is also a major interest at our department. We may not be midwives, but our methods are very useful. So, we are delighted to have found something that we can bee of assistance with.”