Machine Learning Models Identify Higher-Risk Pregnancies Earlier in Care

Although early risk stratification can guide targeted monitoring and timely intervention, its use remains inconsistent across care settings. A new multicenter study introduces machine learning models designed to identify higher-risk pregnancies earlier.

The machine learning–based first‑trimester antenatal risk prediction approach used information available in the first 14 weeks and analyzed records from more than half a million pregnancies in Sweden, Chile, and Singapore. The multicenter model development study was published in the Journal of Medical Internet Research. The models estimate individualized risks for adverse maternal and neonatal outcomes to inform care.

The models incorporated medical history together with social, demographic, and behavioral information available early in pregnancy. Social and demographic factors ranked among the most informative predictors in some populations. These tools are intended as decision support for clinicians and not as replacements for professional judgment.

The models generally outperformed existing early risk assessments. Discrimination improved substantially in Sweden and Chile, while gains in Singapore were smaller but still statistically significant. Calibration varied, with reasonable agreement between predicted and observed risks in Sweden and Singapore, and a less well‑calibrated model in Chile, underscoring the need for local tailoring and careful testing.