Sleep Data from Wearable Device May Help Predict Preterm Birth

Although disrupted sleep is recognized as a predictor of preterm birth, defined as delivery before 37 weeks of gestation, the exact reasons for this link have remained uncertain due to the reliance on self-reported data. Now, researchers have discovered that fluctuations in sleep patterns during pregnancy can be a reliable indicator of preterm birth risk.

An interdisciplinary team of researchers at Washington University in St. Louis (St. Louis, MO, USA) utilized machine learning algorithms to assess sleep data collected from pregnant individuals. This team has previously used wearable data to address various health issues, including tracking COVID-19 exposure, forecasting surgical outcomes, and identifying signs of depression and anxiety. While many people report poor sleep during pregnancy—typically in the third trimester—the researchers specifically examined sleep behavior before 20 weeks of pregnancy to find early warning signals. They analyzed information from a 2014 cohort study involving 665 pregnant participants during their first and second trimesters, all of whom had documented delivery dates. Within this group, around 14% experienced preterm births. Participants wore a validated wrist-worn device, known as an actigraph, for about two weeks, which tracked body movements to monitor sleep activity.

This data enabled the researchers to identify daily sleep-related patterns, such as sleep duration, bedtime, and wake time, movement during sleep, and other relevant variables. Alongside this, participants also completed sleep surveys. The team combined the objective and subjective data and fed it into machine-learning models to evaluate the influence of sleep patterns on the likelihood of preterm birth. Their analysis revealed that sleep-related metrics are reasonably effective at predicting preterm birth. Notably, inconsistency in sleep schedules emerged as a more powerful predictor than average sleep quality or duration, indicating that maintaining a regular sleep routine is more crucial than simply getting more or better sleep overall. The researchers noted that they kept their models intentionally straightforward to highlight clinically relevant associations. Moving forward, the team intends to test their findings across different populations and academic medical institutions.

“Raw data from wearables can be very messy, but using a healthy combination of statistical methods, AI (artificial intelligence) and clinical knowledge, researchers can extract important clinical insights,” said Chenyang Lu, the Fullgraf Professor in the Department of Computer Science & Engineering in McKelvey Engineering. “Then AI scientists and clinicians work together to extract the insights from these very complex data from the real world and get meaningful insights from it.”