Dense phenotyping from EHR enables machine learning-based prediction of PTB

The study developed prediction models for preterm birth by applying machine learning methods to diverse data from Vanderbilt’s EHRs with 35,282 deliveries. The trained models (gradient-boosted decision trees) that combined demographic factors, clinical history, laboratory tests, and genetic risk with billing codes were established and compared in detailed patterns. Findings: The models based on billing codes alone can predict preterm birth risk at various gestational ages and outperform comparable models trained using known risk factors. The machine learning approach also predicts preterm birth subtypes (spontaneous vs. indicated), mode of delivery, and recurrent preterm birth. The generalizability of billing code-based models was also validated in a large, independent UCSF cohort (5978 deliveries) and resulted in the similar accuracy.