Predictive Models for Very Preterm Birth: Developing a Point-of-Care Tool.

Area of Science:

• Maternal-fetal medicine
• Predictive analytics in healthcare
• Public health research

Background:

• Very preterm birth (VPTB) poses significant health risks.
• Accurate prediction of VPTB is crucial for timely intervention.
• Existing predictive models may lack applicability at various pregnancy stages.

Purpose of the Study:

• To develop and validate point-of-care predictive models for very preterm birth.
• To create models utilizing data available at three distinct time points: pre-pregnancy, first-trimester end, and mid-pregnancy.
• To assess the predictive performance of these models.

Main Methods:

• Retrospective cohort study of 359,396 Ohio Medicaid mothers (2008-2015).
• Multivariable logistic regression used to develop three predictive models.
• Models validated on a separate dataset to assess performance.

Main Results:

• The study included 359,396 live births, with 1.81% being very preterm births.
• All developed models exhibited excellent calibration and strong goodness-of-fit.
• The mid-pregnancy model showed acceptable discrimination (AUC ≈ 0.75).

Conclusions:

• Point-of-care predictive models for very preterm birth were successfully developed.
• These models, usable at different pregnancy stages, can estimate VPTB probability.
• Further research is needed to integrate these models into interventions for VPTB prevention.