With machine learning, Mayo Clinic researchers found it is possible to predict how patterns of changes in pregnant patients who are in labor can help identify whether a vaginal delivery will occur with good outcomes for mom and baby.

WHY IT MATTERS
The ability to change the shape of and open the birth canal to make way for a baby to be born varies from patient to patient. When obstetricians analyze contractions, as well as fetal heartbeats, they assess the progress of labor and make recommendations on levels of care for the medically risky delivery process of birth.

Mayo Clinic researchers say these new models can predict a composite of medical outcomes and the probability of poor labor outcomes – cesarean delivery in active labor, postpartum hemorrhage, intra-amniotic infection, shoulder dystocia, neonatal morbidity and mortality – based on what machine learning can do with dilation data.

Use of the models could result in more individualized clinical decisions using the baseline characteristics of each patient, and they could also be a tool to help remote physicians and midwives transfer rural or remote patients to the appropriate level of care, Dr. Abimbola Famuyide, a Mayo Clinic OB-GYN and senior author of the study, said in a prepared statement.

"This is the first step to using algorithms in providing powerful guidance to physicians and midwives as they make critical decisions during the labor process," he said.

To create the baseline and multiple intrapartum prediction models, Mayo Clinic researchers used a dataset of pregnancy and labor characteristics from a dozen U.S. medical centers, including 19 hospitals located in all nine districts of the American College of Obstetricians and Gynecologists, known as the Consortium of Safe Labor.

The hospitals provided the Eunice Kennedy Shriver National Institute of Child Health and Human Development with de-identified electronic obstetric, labor and newborn data between 2002 and 2008.

According to the published study, of the 228,438 delivery episodes in the database, there were 779 antepartum, intrapartum and postpartum variables.

The algorithms analyzed data known at the time of admission in labor – patient baseline characteristics, the patient's most recent clinical assessment and labor progress from admission.

Researchers used 66,586 records in the prediction models, where 14,439 deliveries (21.68%) reported poor labor outcomes. 

The researchers noted in the study that although intrapartum fetal heart rate monitoring was considered, it was not included in the models due to a lack of documentation in the database.

THE LARGER TREND
Studies in recent years have detailed the high costs of maternal morbidity, which is driving maternal health investments.

Maternal and reproductive health startups brought in $424 million in funding during the first quarter of this year, as reported in MobiHealthNews.

The report's authors point to a Mathematica-led 2019 study that determined a total maternal morbidity cost of $32.3 billion from conception to the child at age 5, amounting to more than $8,600 for each maternal-child pair.

Community-based models and telemedicine are aiming to address rural access and racial disparities to fill gaps in care to improve maternal health and child outcomes. Telehealth has shown to have a positive impact on outcomes and costs.

However, the use of telehealth applications in the monitoring of high-risk pregnancies during COVID-19 had a positive impact on both mothers and babies, as well as costs, but was limited in addressing escalating maternal morbidity rates, according to a meta-analysis published in the Journal of Telemedicine and Telecare.

Artificial intelligence could help healthcare providers with diagnosis by effectively forecasting complications and predicting whether particular treatments would be effective for a patient based on their personal health history, according to risk-adjustment services and software experts.

And while digital health Investments in machine learning are high, most clinical use cases are in trial with impact, infrastructure and regulatory oversight yet to be figured out.

ON THE RECORD
"The AI algorithm’s ability to predict individualized risks during the labor process will not only help reduce adverse birth outcomes but it can also reduce healthcare costs associated with maternal morbidity in the U.S.," said Dr. Bijan Borah, endowed scientific director for health services and outcomes research at Mayo Clinic.