Reduced sleep and physical activity in early and mid-pregnancy is associated with a significant increase in the risk for preterm birth, a new Artificial Intelligence (AI) model created by researchers at the March of Dimes Prematurity Research Center (PRC) at Stanford has found.
The findings, published recently in npj Digital Medicine, showed a nearly 50% increase in risk for those who slept and moved less. The risk of preterm birth was reduced by more than half for women who slept and moved more in later pregnancy.
The model used sleep and physical activity data from a wearable device on a pregnant woman to predict her gestational age and her risk of preterm birth.
It could do both these things because of something it did at the very start: it used wearable data from 1,500 pregnant study participants to create the first ever standard, or “clock,” for what is considered a normal amount of sleep and physical activity during pregnancy.
The study’s most interesting moments came when the model substantially over or underestimated a woman’s gestational age—in those cases, that woman saw her risk of preterm birth either dramatically increase or decrease.
“Based on this clock and data from the wearable, we showed that the AI model can tell us how pregnant people are,” said lead study author Dr. Nima Aghaeepour, Associate Professor of Anesthesiology, Perioperative, and Pain Medicine, and of Pediatrics, at Stanford Medicine.
“And we showed that when women diverge from this normal clock, when they move in a worse direction, when they are in early pregnancy, but the AI thinks they are in late pregnancy because of less movement and sleep, that is when bad things can happen.”
The model found that women who aren’t moving well and sleeping a lot have a 44% increased risk of preterm birth.
“We also showed the opposite,” he said. “We showed that women who the AI thinks are in early pregnancy because they are moving and sleeping a lot but are actually in later pregnancy are protected from early labor risk.”
The model found that women who are moving well and sleeping a lot have a 52% reduced risk of preterm birth.
While the figures are striking, Dr. Aghaeepour said the devil is in the details. And those details are yet to be cracked open by his team.
What exactly the normal “clock” of physical activity and sleep looks like for pregnant women, and what “more active,” “sleep better,” “less active,” and “sleep worse” (the terms the AI is using) really mean are still unknowns.
To decode the AI’s complex mathematical logic into something humans can understand and act on, Dr. Aghaeepour’s team is working on translating the data into a consumer grade wearable device like an Apple Watch.
“Once we understand how the AI works, we can give guidance,” he said. “Something like, OK, at this stage of pregnancy, you need to take this many steps and sleep this many hours.”
“And that’s really valuable for all women, but especially for African American women in low- and middle-income populations disproportionately affected by preterm birth—we’re giving them really cost-effective interventions that have the capability to dramatically alter the course of their pregnancies.”
Dr. Aghaeepour’s study, which still has to be validated in a larger, more diverse group, was a collaborative effort with Washington University in St. Louis.
It collected more than 180,000 hours of data from clinical grade wearables given to 1,500 pregnant women enrolled in the university over the course of their pregnancies. More than 90% of the study cohort was African American, an understudied yet overburdened group when it comes to preterm birth.
This work has allowed the study team to gain “valuable insights into a diverse population with an elevated risk for preterm birth,” said Dr. Erik Herzog, a Professor of Biology at Washington University and one of the study authors, adding that 14% of the study participants delivered early.
“Because of prior studies in shift workers, we hypothesized that disruption of the moms’ daily schedules puts a pregnancy at risk,” he said. “We are excited to learn that wrist actigraphy and light exposure are easily measured indicators of a mom’s sleep-wake schedule and risk for poor pregnancy outcomes.”
While the high correlative values—of 44% and 52%—from the study were striking even for Dr. Aghaeepour, the link between preterm birth and decreased sleep and exercise was not a surprise.
“We’ve known for some time that preterm birth is driven by inflammation, and that sleep and physical activity can modulate inflammation,” he said. “What we didn’t realize was just how strong this relationship was.”
“Our next step is to investigate the possibility of using physical activity and sleep to regulate inflammation—because that’s a completely free and natural way to lower preterm birth risk.”