AI analysis uncovers hidden pregnancy risks, offering insights for personalized prenatal care.
A new study examining nearly 10,000 pregnancies has uncovered previously unknown risk factors that can contribute to severe complications, including stillbirth. Researchers at the University of Utah used artificial intelligence to analyze extensive pregnancy data, revealing patterns that had not been detected before. These findings highlight differences in how risk factors combine, sometimes leading to a tenfold variation in risk for infants who are otherwise treated the same under standard guidelines.
The study looked at a broad range of data, including social and medical details such as support systems, blood pressure, medical history, fetal growth, and pregnancy outcomes. Through AI-driven analysis, researchers pinpointed combinations of maternal and fetal factors that increase the likelihood of negative outcomes. One surprising discovery was that female fetuses, typically considered lower risk than males, faced higher danger if the pregnant person had pre-existing diabetes. This pattern had gone unnoticed in previous research.
Doctors often rely on experience to assess risk, but even the most skilled professionals may not recognize every possible combination of risk factors. The AI model, however, was able to detect subtle patterns that human judgment alone might miss. One area of focus was on fetuses in the lower weight range—specifically, those in the bottom ten percent but not the absolute lowest three percent. These babies often fall into a gray zone where doctors must decide whether to recommend intensive monitoring or allow the pregnancy to proceed as usual. Current guidelines call for close monitoring of all cases, but the study found that risks within this category vary greatly. Some pregnancies posed no greater danger than average, while others had a much higher likelihood of complications. Factors such as fetal sex, diabetes, and structural issues like heart defects played a role in determining risk.

Although the study does not prove cause and effect, it does provide valuable insights into how risk factors interact. Experienced doctors often have an intuitive sense that some pregnancies in this category are safer than others, but AI can offer a more objective, measurable way to assess risk. By incorporating a vast amount of data, AI-based models can help guide medical decisions in a way that is more consistent and less influenced by individual judgment.
The research team used an “explainable AI” model, which differs from traditional black-box AI systems that produce conclusions without revealing how they were reached. Instead, this system allows doctors to see which factors contributed to the final risk assessment, making it easier to understand and trust the results. This transparency is particularly important in medicine, where every decision carries significant consequences. By showing its reasoning, the AI model provides a level of accountability and allows for adjustments if biases or errors are identified.
Pregnancy risk assessment is complex, involving countless variables that interact in unpredictable ways. While experienced doctors are skilled at weighing multiple factors at once, even the best clinicians may struggle to articulate exactly how they arrive at their decisions. Factors like fatigue, personal biases, and limited experience with rare cases can also affect judgment. AI does not replace doctors but acts as a tool to supplement their expertise, offering a second opinion that is based on patterns detected across thousands of cases.
The researchers believe their model is particularly useful for assessing rare pregnancy scenarios, where limited data makes it difficult to predict outcomes with confidence. By analyzing large datasets, the AI system can recognize risks that might not be apparent in smaller studies or individual clinical experiences. The next step is to test the model on new populations to ensure it performs well in real-world settings. If successful, this kind of AI-driven risk assessment could help personalize pregnancy care, providing expectant parents with more precise information about their individual situations.
Dr. Nathan Blue, a lead researcher on the project, hopes that explainable AI will help make risk assessments more reliable and tailored to each pregnancy. Instead of applying the same guidelines to every patient, this approach allows for a more customized evaluation, ensuring that care is both effective and equitable. By improving how risks are identified and understood, this technology has the potential to reshape prenatal care, giving doctors and patients better tools to navigate the challenges of pregnancy.
Sources:
AI-based model detects unseen risk combinations in pregnancies
Join the conversation!