Clinical Decision Support System for Post-Pregnancy Mental Health
Sep. 2024 – May. 2025
Data-driven modeling for postpartum depression risk prediction and clinical decision support
Overview
This project focuses on developing a clinical decision support system for post-pregnancy loss (PPL) mental health, with an emphasis on early risk prediction for postpartum depression (PPD). The goal is to bridge the gap between data-driven prediction and actionable clinical insights.
The work is conducted at Cornell University under the supervision of Dr. Clifford Whitcomb and Dr. Yiye Zhang.
Problem Motivation
Postpartum depression affects a significant portion of patients, yet early detection remains challenging in clinical practice. Existing approaches often lack integration of social support indicators and real-time predictive capability.
This project investigates how machine learning models can leverage large-scale healthcare data to provide early warnings and support personalized intervention strategies.
Methodology
- Analyzed PRAMS Phase 8 dataset (2016–2021), containing over 200,000 observations
- Engineered features including:
- social support indicators
- socioeconomic status
- mental health history
- Implemented multiple predictive models:
- regression models
- tree-based models (Random Forest)
- Bayesian methods
- deep neural networks
System Perspective
Beyond prediction, this project incorporates a systems engineering perspective by integrating model outputs into a decision-support framework for clinicians. The objective is to enable actionable, interpretable, and personalized recommendations in healthcare settings.
Poster
See our results in this Poster.