Medical Data of Mothers and Babies Predict Complications of Prematurity, Study Led by Stanford Medicine Finds | Information Center

“There is an IT challenge in using electronic health records because they are longitudinal and contain a large amount of data from each patient,” Aghaeepour said. “A long-term and short-term memory neural network works in the same way as a person reading a book. When we read, we don’t remember every word, but we remember key concepts, read the next part, add other key concepts, and move on. The algorithm does not memorize every patient’s entire electronic health record, but it can memorize key concepts and carry them over to the point where we make a prediction. »

At the time of birth, the machine learning model provided strong predictions for which infants would develop various conditions, including bronchopulmonary dysplasia, a type of chronic lung disease; retinopathy of prematurity, a problem with the retina that can lead to loss of vision or blindness; premature anemia; and necrotizing enterocolitis, a serious gastrointestinal complication that is often not diagnosed until several weeks after birth, at which time interventions are complex and associated with poor outcomes.

The model also gave strong predictions one week before birth for multiple outcomes, including mortality and retinopathy of prematurity, which can lead to vision loss or blindness, as well as moderately strong predictions for 11 other conditions.

“I was surprised by the predictive power we have even before the baby is born, and right from birth,” Aghaeepour said. “I didn’t expect to see this. I had thought the accuracy would come days after birth, once we collected the baby’s data.

Some complications were not reliably predicted by the model, such as infants developing candidiasis or yeast infections; polycythemia, a high concentration of red blood cells in the blood; or meconium aspiration syndrome, in which the infant inhales meconium, a sticky substance expelled from the fetal intestine, during birth.

The researchers validated that the strength of the predictions had not changed over the years (by comparing births from 2014 to 2018 to those from 2019 to 2020); they also validated some of the findings using an independent panel of 12,258 mother-baby pairs from UC San Francisco.

The model’s predictions at birth provided more accurate information than currently used risk assessment tools such as Apgar scores and the National Institute of Child Health and Human Development’s risk score. These scores only take into account the baby’s condition at birth and do not incorporate any information from the mother’s medical history, the researchers noted. However, more studies in more diverse populations are needed before this machine learning tool is ready to replace existing bedside risk calculators, the researchers said.

The health of the mother matters

The model revealed unexpected links between certain health or social conditions in mothers and the health of their infants, the researchers said.

For example, mothers with anemia – a common complication of pregnancy – were more likely to have anemic newborns. These infants were also more likely to develop necrotizing enterocolitis, a bowel complication, according to the study.

“We need to explore what links explain these relationships at the biological level, as these could offer clues to how certain conditions occur,” Stevenson said. “This will allow us to better intervene to help these children.”

The new algorithm was also able to link specific types of socioeconomic disadvantage in mothers to certain complications of prematurity in their babies.

“If a mother was homeless, we found that the impact on the health of the baby would be different from the impact of incarceration, whereas according to traditional paradigms, these two socio-economic factors could have similar effects on the risk of prematurity,” Aghaeepour said.

The model’s predictions could help neonatologists better identify patients who will benefit from existing protocols to prevent birth complications, Stevenson said. For example, newborns who lack oxygen during birth can now receive cooling protocols early in life, which lower their body temperature for a few days to prevent brain damage. Predictive scores can help identify other infants who might be helped by cooling, he said.

The work needs to be replicated in larger, more diverse patient populations and integrated with other Stanford Medicine research that characterizes pregnancies based on thousands of biomarkers that change during gestation, the scientists said.

Scientists from UC San Francisco contributed to the study.

Funding for the research was provided by the National Institutes of Health (grants 1R01HL139844, 3P30AG066515, R35GM138353, 1R61NS114926, 1R01AG058417, R01HD105256, P01HD106414, T32GM007618, and T32GM067547), the Alfred E. Mann National Science Foundation, and the American National Science Foundation. Foundation.