Researchers at Mass General Brigham have developed a machine learning model that identifies patients at risk of postpartum depression (PPD) using easily accessible clinical and demographic data. Published in the American Journal of Psychiatry, this tool could enable earlier mental health interventions, addressing a condition that affects up to 15% of individuals after childbirth.
The study, led by Dr. Mark Clapp and his team, analyzed data from 29,168 pregnant patients across eight hospitals within the Mass General Brigham system. The model was trained on health records from half of the patients and tested on the other half, achieving 90% accuracy in ruling out PPD. Notably, it identified 30% of high-risk patients who later developed PPD, outperforming general population risk estimates by two to three times.
The model relies on electronic health record (EHR) data available at delivery, such as demographics, medical history, and visit patterns, eliminating the need for additional tests. Importantly, it performed consistently across diverse racial, ethnic, and age groups. The inclusion of prenatal Edinburgh Postnatal Depression Scale scores further enhanced its predictive power.
Dr. Clapp emphasized the potential of this tool to transform maternal mental health care: “Our goal is to facilitate earlier support for parents during a vulnerable period marked by significant life changes.”
While further validation is underway, this model represents a promising step toward proactive PPD management. By integrating it into clinical practice, healthcare providers could offer timely interventions, improving outcomes for postpartum patients. Future efforts will focus on refining the tool and ensuring its ethical implementation in real-world settings.
Add comment