Dr. Nathan Blue, senior author of the study and assistant professor of maternal fetal medicine in the department of obstetrics and gynecology at U of U Health, said one of the most interesting findings is that if a woman has preexisting diabetes while pregnant, it puts female fetuses at a higher risk than males for experiencing complications.

“This was a very unexpected finding and I can’t tell you why for sure but there are other studies that are looking at the ways babies are affected by blood sugars between the sexes,” Blue said.

The study, which included nearly 10,000 pregnancies across the nation, also showed combinations of maternal and fetal characteristics linked to stillbirth. For babies in the bottom 10% for weight, but not the bottom 3%, the risk of unhealthy outcomes varied from no riskier than an average pregnancy to nearly tenfold. This was based on several different factors such as sex of the baby, preexisting diabetes in the mother and fetal anomalies such as heart defects.

“We do all of the medical history and testing, but there are a lot of very complex factors and even the best clinical brains may not be able to quantify exactly how they arrived at their final decision,” Blue said.

Bias, mood and loss of sleep can subtly skew judgement calls, Blue said, which is why he called the “explainable AI” an important tool that provides an estimated risk.

“The AI-based prediction models are transparent, so you can see how it arrived at its conclusion. We can tell a patient this could be a very serious thing, or it can also be nothing and everything will be fine. It’s a bit disconcerting and can cause a lot of emotional and financial stress, and sometimes for no reason at all. So we set out to use a new approach to identify the risks.”

Blue said the AI model could provide important advantages over what recommendations physicians can make.

All of this information gathered over the past three years is exciting, he added, because it shows the AI model can assist researchers and help them learn more about pregnancy health.

Blue said estimating pregnancy risks involves taking a very large number of variables into account, such as ultrasound data and maternal health. The variables can be weighed to make decisions for each individual.

The study is over, but Blue said researchers will still test new populations.

“In the state, there are about 1 in 175 pregnancies that end in stillbirth,” he said. “It’s more frequent than you would think. I will say, I guess this is not so much about a focus on this research but, more importantly, people need to know that every pregnancy has a risk of stillbirth and it’s important to attend all prenatal visits and talk about your risk so your physician can do their best to make sure you have a healthy, happy outcome.”

While the study was only able to detect correlations between variables, it doesn’t provide information about what the actual causes are for negative outcomes.

Blue said this shouldn’t alarm people in his line of work but instead should be looked at as a great assistant.

“A lot of people are asking if this is going to replace or take our job, and I would say that’s an understandable question, but I think it’s important to think about a doctor or clinician who is assisted with something that will help improve the way we take care of our patients so they can have the healthiest possible outcomes,” he said. “AI models can essentially estimate a risk that is specific to a given person’s context and they can do it transparently and reproducibly, which is what our brains can’t do.”