Mount Sinai Presents AI Models That Could Predict Serious Pregnancy Complications Before They Develop

Placenta accreta spectrum is a life-threatening complication in which the placenta attaches too deeply into the uterine wall. Its incidence has risen sharply alongside increasing cesarean delivery rates in the United States, and it often goes undetected until delivery - when hemorrhage can become catastrophic. Predicting which patients are at elevated risk before pregnancy even begins would fundamentally change how clinicians counsel and monitor high-risk individuals.

That is the premise behind one of ten studies presented by Mount Sinai maternal-fetal medicine specialists at the Society for Maternal-Fetal Medicine's 2026 Annual Pregnancy Meeting in Las Vegas. Taken together, the presentations reflect a broad push to apply artificial intelligence, machine learning, and population-level data analysis to complications that currently rely on identifying risk only after pregnancy is underway.

Predicting Placenta Accreta Before Conception

Maternal-fetal medicine attending physician Henri Mitchell Rosenberg presented a study showing that a preconception machine learning model using Electronic Medical Record data can identify patients at elevated risk for placenta accreta spectrum with high sensitivity and specificity. The model detected anemia as a novel predictor of risk - highlighting a potentially modifiable risk factor that clinicians could address before conception occurs.

If validated prospectively, such a tool could support risk counseling, specialist referral, and personalized care planning before a patient becomes pregnant - giving medical teams time to prepare rather than respond. A second Mount Sinai study, presented by OBGYN resident Tess Cersonsky, extended this work by examining how specific surgical techniques used during a prior cesarean - including suture type and closure method for abdominal layers - affect placenta accreta spectrum risk in a subsequent pregnancy. A recurrent neural network model was used to assess these surgical variations, opening a path toward more individualized preconception risk assessment based on a patient's surgical history.

AI for Fetal Cardiac Screening

Congenital heart defects are among the most serious fetal abnormalities, and prenatal detection is critical for planning the immediate postnatal care that can determine whether affected neonates survive. Jennifer Lam-Rachlin, a maternal-fetal medicine attending at Mount Sinai West, presented an evaluation of an AI-supported tool that flags suspicious findings for severe congenital heart defects during routine ultrasound at an AIUM-certified prenatal diagnostic center.

The study assessed whether AI assistance improved completion rates for fetal cardiac screening - a measure of whether the key diagnostic views were obtained during each scan. Early evidence from other settings shows strong performance; this work represents real-world validation in a busy clinical environment.

Corticosteroids, Weight, and Preterm Birth Outcomes

Antenatal corticosteroids are a standard intervention given to pregnant people at risk of preterm delivery to improve neonatal respiratory outcomes. The drug is administered as a fixed dose regardless of maternal size. Sara Edwards, a maternal-fetal medicine fellow, presented two analyses questioning whether that fixed-dose approach is equally effective across all maternal body sizes - specifically, whether obesity reduces corticosteroid efficacy for improving neonatal respiratory outcomes. The hypothesis that standard dosing may be insufficient for some patients has direct implications for how the intervention is administered in high-risk pregnancies.

Racial Disparities in Labor Management and Delivery Mode

Racial and ethnic disparities in labor and delivery management have been documented across US institutions, but how those disparities translate into differences in delivery mode - specifically rates of cesarean delivery - has been less studied. Clinical research program director Nicola Tavella presented findings on racial and ethnic disparities in labor management and their mediating effects on cesarean risk among a cohort of first-time mothers carrying a single full-term fetus in the head-down position. A separate Tavella-led study examined whether offering elective induction of labor at 39 weeks on a universal institutional basis can reduce disparities based on neighborhood-level social vulnerability.

Neighborhood Gun Violence and Adverse Pregnancy Outcomes

Prenatal stress is a recognized contributor to adverse pregnancy outcomes. Edwards presented findings examining whether neighborhood-level exposure to gun violence during pregnancy was associated with increased risk of such outcomes - extending the literature on environmental stress beyond traditional socioeconomic indices to a specific, measurable stressor with distinct geographic distribution.

Preeclampsia Prevention and Social Vulnerability

Two additional studies focused on preeclampsia prevention and social determinants. One examined universal implementation of low-dose aspirin (81mg daily) for pregnant individuals with high or multiple moderate risk factors - assessing whether broad implementation changes outcomes compared to more selective prescription. Another examined whether a composite social vulnerability index independently predicts adverse pregnancy outcomes, testing whether combined measures of socioenvironmental stress capture risk that individual indices miss.

Across all ten presentations, the common thread is an attempt to identify risk earlier, account for population-level factors that clinical protocols often ignore, and tailor interventions to individuals whose circumstances differ from the population averages on which current guidelines are based.