Five-Feature Models to Predict Preeclampsia Onset Time From Electronic Health Record Data: Development and Validation...

Area of Science:

• Obstetrics and Gynecology
• Medical Informatics
• Predictive Analytics

Background:

• Preeclampsia is a dangerous pregnancy complication characterized by high blood pressure and proteinuria.
• Accurate prediction of preeclampsia onset remains a significant clinical challenge.

Purpose of the Study:

• To develop quantitative models for predicting preeclampsia onset gestational age.
• Utilize electronic health records (EHRs) for predictive modeling.

Main Methods:

• Retrospective collection of 1178 (discovery) and 881 (validation) preeclamptic pregnancy records.
• Construction of two Cox-proportional hazards models: a baseline model and a full model including lab findings, vitals, and medications.
• Model development using 80% of discovery data, testing on 20%, and validation with external data.

Main Results:

• The full model demonstrated improved predictive performance (Concordance index of 0.69) compared to the baseline model (0.64) in discovery data.
• Area Under the Curve (AUC) values for predicting preeclampsia at 34 and 37 weeks were 0.70 for the full model.
• Key predictors included number of fetuses, hypertension, parity, and early pregnancy diastolic blood pressure.

Conclusions:

• EHR data are valuable for predicting preeclampsia gestational age.
• Cox-proportional hazards models with 5 predictors offer practical tools for clinicians to assess preeclampsia onset risk.