Systematic Review of Prediction Models for Preterm Birth Using CHARMS

This study sought to evaluate prediction models for preterm birth (PTB) and to explore predictors frequently used in PTB prediction models.

A systematic review was conducted. We selected studies according to the PRISMA, classified studies according to TRIPOD, appraised studies according to the PROBAST, and extracted and synthesized the data narratively according to the CHARMS. We classified the predictors in the models into socio-economic factors with demographic, psychosocial, biomedical, and health behavioral factors.

Twenty-one studies with 27 prediction models were selected for the analysis. Only 16 models (59.3%) defined PTB outcomes as 37 weeks or less, and seven models (25.9%) defined PTB as 32 weeks or less. The PTB rates varied according to whether high-risk pregnant women were included and according to the outcome definition used. The most frequently included predictors were age (among demographic factors), height, weight, body mass index, and chronic disease (among biomedical factors), and smoking (among behavioral factors).

When using the PTB prediction model, one must pay attention to the outcome definition and inclusion criteria to select a model that fits the case. Many studies use the sub-categories of PTB; however, some of these sub-categories are not correctly indicated, and they can be misunderstood as PTB (≤ 37 weeks). To develop further PTB prediction models, it is necessary to set the target population and identify the outcomes to predict.