Skip to main content

Can Machine Learning Predict Postpartum Depression?

May 21, 2019

SAN FRANCISCO—Machine learning shows potential for predicting postpartum depression using clinical data routinely collected in electronic health records, according to a study presented during a poster session at the American Psychiatric Association’s annual meeting.

“Postpartum depression is considered to be one of the most frequent maternal morbidities after delivery with serious implications on the mother and children,” wrote poster presenter Shuojia Wang and coauthors Jyotishman Pathak and Yiye Zhang. “The ability to predict postpartum depression in women could enable the implementation of effective mental and behavioral health interventions.”

The researchers used electronic health records from Weill Cornell Medicine and New York-Presbyterian Hospital in New York City covering 27,716 singleton pregnancies from 24,627 women as a data source. Five different machine learning algorithms were created to predict postpartum depression.

“We compared models with demographic only, medication information only, diagnostic information only, and their combinations,” the authors wrote. “Models stratified by different trimester and their combinations were also used to predict postpartum depression.”


The study identified the following as the most significant predictors of postpartum depression: age; ethnicity; gestational week; prenatal mental health; diagnosis of threatened abortion in the first trimester; diagnosis of asthma in second trimester; diagnosis of infectious disease, abdominal pain, or pelvic pain in third trimester; and drug exposures including hyperosmotic laxatives, anti-infectives, antihistamines/antitussives/analgesics, and vitamins in third trimester.


“Clinical information including disease classifications and drug exposures during prenatal care procedure may assist in forecasting postpartum depression,” researchers wrote. “These results may facilitate effective detection and primary prevention of postpartum depression as clinical decision support.”

—Jolynn Tumolo



“Using electronic health records and machine learning to predict postpartum depression.” Abstract presented at: the American Psychiatric Association Annual Meeting; May 20, 2019; San Francisco, CA.

Back to Top