Skip to main content
SMFM 2024 Pregnancy Meeting Main banner
Final Program


Icon Legend
This session is not in your schedule.
This session is in your schedule. Click again to remove it.
Presentation Icons
Additional registration fee required
Additional registration fee required
Faculty have requested this content not be shared on social media
Faculty have requested this content not be shared on social media
CME Credit Offered
CME Credit Offered
Poster Icons
Foundation Awardee
Foundation Awardee
No Social Media
No Social Media
Back
Q&A

Health Equity/Community Health

Poster Session 1

(254) Social Determinants of Health Improves Prediction for Postpartum Readmissions Due to Preeclampsia

Tuesday, February 13, 2024
10:30 AM – 12:00 PM  
Location: Prince George's Exhibit Hall CDE, Lower Atrium Level
Foundation Awardee
Objective: Strong associations exist between social determinants of health (SDOH) and preeclampsia. We hypothesize that socially disadvantaged patients are more likely to be readmitted for worsening preeclampsia in the postpartum period. The objective of this study is to compare the performance of 2 machine learning (ML) models where zip code related SDOH variables are added to the basic model.

Study Design: This is a retrospective analysis of 146 patients who had the diagnosis of hypertensive disorder of pregnancy (HDP) in a single tertiary medical center from 2020-2023. Two supervised ML models were created using four different ML methods (Random Forest, XGBoost, Logistic Regression, and Support Vector Machines). Model 1 only utilized clinically available variables. Model 2 uses the same variables as Model 1 and the zip codes related Area Deprivation Index (ADI) and SDOH variables from Agency of Healthcare Research and Quality database.


Results: Among the HDP cohort of 73 readmissions and 73 non-readmissions, 51 EMR variables and 175 zip code related variables were included in the models. The best performance was achieved using a two-sample Kolmogorov-Smirnov and correlation reduction-based feature selection method and Random Forest ML method in Model 2 (Table). Adding zip code related variables (Model 2) was able to improve the sensitivity by 65% (Model 1: 0.43; Model 2: 0.71), and area under the curve (AUC) by 30% (Model 1: 0.56; Model 2: 0.73). The top features associated with readmissions are the increase of body mass index in pregnancy, the distance from a tertiary center, the unemployment rate in the neighborhood, the percentage of non-English speakers in the neighborhood, and whether an electrocardiogram was ordered in the pregnancy.

Conclusion: Adding zip code related ADI and SODH variables can improve the ML model prediction. The performance of models should be further validated by a multicenter or state database. With greater insights regarding the geographic hot spots for preeclampsia readmissions, new guidelines for targeted intervention and resource allocation may be implemented.