Validation of a Predictive Model of Postpartum Readmission for Hypertension and Pre-eclampsia

A new study aimed to create a predictive model for postpartum readmission due to hypertension and pre-eclampsia at delivery discharge. 

The model's external validation and transportability across clinical sites were assessed using an internal-external cross-validation approach. Data from two tertiary care health systems in the Southern (2014–2015) and Northeastern USA (2017–2019), involving 28,201 postpartum individuals, were utilized.

The developed model, employing penalized logistic regression, incorporated six variables and demonstrated adequate discrimination during internal validation at both health systems. However, discrimination varied between sites in the internal-external cross-validation, with improved performance for the Northeastern model on the Southern cohort. However, the calibration was deemed inadequate. Subsequent model updating using a combined dataset resulted in a new model with satisfactory discrimination, moderate calibration, and superior net benefit – for clinical decision-making thresholds.

From the results, it was inferred that an accurate prediction of postpartum readmission for hypertension and pre-eclampsia is feasible. Nevertheless, the predictive model required further validation. Model updating with data from diverse clinical sites is crucial before widespread implementation. The suggested prospective validation and updating in additional cohorts should focus on establishing a robust and transportable predictive model for clinical use, potentially enhancing risk stratification, resource allocation, and overall postpartum care.

Source: Venkatesh KK, Jelovsek JE, Hoffman M, Beckham AJ, Bitar G, Friedman AM, Boggess KA, Stamilio DM. BJOG: An International Journal of Obstetrics & Gynaecology. 2023 Nov;130(12):1531-40.