Innovative Machine Learning Model Enhances Detection of Postpartum Depression Risk

Postpartum depression (PPD) remains a significant challenge affecting up to 15% of new mothers following childbirth. Early identification is crucial to provide timely mental health support and improve outcomes. Researchers at Mass General Brigham have developed a sophisticated machine learning model designed to assess the risk of PPD using data readily available during hospital discharge. This model incorporates clinical, demographic, and visit history factors extracted from electronic health records (EHR), enabling healthcare providers to identify at-risk individuals earlier in the postpartum period.

The study pooled data from 29,168 pregnant patients across two academic medical centers and six community hospitals within the Mass General Brigham system, spanning from 2017 to 2022. Approximately 9% of these patients experienced PPD within six months after delivery. The researchers trained the model using half of the cohort and tested its predictive accuracy on the remaining patients. Results showed the model effectively ruled out PPD in 90% of cases, while accurately predicting nearly 30% of those flagged as high risk would develop PPD, making it two to three times more precise than general population risk estimates.

Importantly, the model performed consistently across different racial, ethnic, and age groups, even among patients without prior psychiatric history. Incorporating prenatal Edinburgh Postnatal Depression Scale scores further enhanced the model's predictive power, suggesting existing screening tools can be integrated into automated risk assessments.

The team is now conducting prospective validation studies and collaborating with clinicians and stakeholders to integrate this tool into clinical workflows. The goal is to facilitate earlier intervention and improve maternal mental health outcomes. According to lead author Dr. Mark Clapp, this advancement represents a significant step toward a predictive approach that, alongside clinician expertise, can better support postpartum patients.

This innovative model underscores the potential of machine learning to transform postpartum care by enabling proactive mental health management, ultimately aiming to reduce the impact of depression during this vulnerable period.