Oral Concurrent Session 5 - Morbidity and Mortality & Public/Global Health
60 - A prediction model for postpartum readmission for hypertension or preeclampsia
Friday, January 29, 2021
5:15 PM – 5:30 PM EST
Objective: To develop a model to predict a woman’s probability of postpartum readmission for hypertension or preeclampsia prior to hospital discharge after delivery.
Study Design: A predictive model for postpartum readmission for hypertension and/or preeclampsia at a tertiary care medical system. To identify candidate risk factors, we conducted a 1:4 matched case-control study (N=135 women: 27 readmissions and 108 time-matched controls) from 09/01/2012-03/25/2014. To develop a prediction model, we conducted a retrospective cohort study (N=10,100 women: 32 readmissions and 10,068 delivery discharges) from 03/26/2014-09/30/2015. We assessed statistical (generalized linear models with a logistic function) and machine learning methods (random forest and gradient-boosted trees) to identify the best model. We measured model discrimination with the C-statistic (i.e., concordance index), which was internally validated. We then plotted calibration and decision curves to assess the accuracy of predicted probabilities for readmission and whether the model provided clinical decision-making benefit.
Results: Of 10,100 women at hospital discharge after delivery, the postpartum readmission rate for hypertension or preeclampsia was 0.35% (95% CI: 0.29-0.42). The statistical logistic regression model had the best fit, and included seven clinical predictors of postpartum readmission available at discharge: maternal age, parity, chronic hypertension, gestational hypertension, preeclampsia before discharge, infant birthweight, and maximum postpartum diastolic blood pressure. The model had excellent discrimination (C-statistic: 0.94, 95% CI: 0.92-0.96). Calibration demonstrated that the model was accurate up to 20% when it began to overpredict the risk of postpartum readmission. The model provided superior net benefit when clinical decision-making thresholds for postpartum readmission were between 0 to 20%. The final model is available as an online calculator here.
Conclusion: Postpartum readmission for hypertension or preeclampsia can be accurately predicted at hospital discharge after delivery.