New model predicts PPD risk after hospital discharge

A simple machine-learning model demonstrates its capacity to stratify the risk for postpartum depression (PPD) before discharge following an individual’s delivery, reports a study.

“This tool could help identify patients within a practice at the highest risk and facilitate individualized postpartum care planning for the prevention of, screening for, and management of PPD at the start of the postpartum period and potentially the onset of symptoms,” the investigators said.

A total of 29,168 individuals were included in the analysis, of whom 2,696 (9.2 percent) met the criteria for PPD in the 6 months following delivery. [Am J Psychiatry 2025;doi:10.1176/appi.ajp.20240381]

The model demonstrated good discrimination and remained well calibrated in external validation. The area under the receiver operating characteristic curve (AUC) was 0.721 (95 percent confidence interval [CI], 0.709‒0.736), while the Brier calibration score was 0.087 (95 percent CI, 0.083‒0.091).

Moreover, the specificity of the model was 90 percent, with a positive predictive value of 28.8 percent (95 percent CI, 26.7‒30.8) and a negative predictive value of 92.2 percent (95 percent CI, 91.8‒92.7).

“The model demonstrated reasonable calibration, distinguishing between higher- and lower-risk populations,” the investigators said. “The risk model also had similar performance across patient subgroups, suggesting that it could equitably be applied in a diverse population.”

These findings support the feasibility of the model in assisting clinical care team when stratifying the risk for PPD and when directing resources and support services to prevent and treat PPD. The tool may be used in settings and practices with limited resources for patients in need of ongoing postpartum and psychiatric care and support, according to the investigators.

Depression scale

In validation studies of the Edinburgh Postnatal Depression Scale (EPDS), results showed the ability of this instrument to screen for depression during pregnancy and the postpartum period. [BMJ 2020;371:m4022]

Unfortunately, the combination of prenatal EPDS score with other patient characteristics has not been fully studied as a tool for predicting PPD development. [BMC Pregnancy Childbirth 2022;22:527; Front Psychiatry 2022;13:1082762]

“We hypothesized that prenatal EPDS score would be an important feature in a PPD risk stratification model and thus limited our study population to those with an available score,” the investigators said.

“When comparing the same models with and without the maximum prenatal EPDS score, the model with the score had higher discrimination (AUC, 0.72 vs 0.68) in the external validation set,” they added.

Based on these findings, prenatal administration of the EPDS might have some benefits, such as screening for depression during pregnancy and as a component of PPD risk stratification at the end of pregnancy, according to the investigators.

The present retrospective cohort study included individuals who delivered between 2017 and 2022 in one of two large academic medical centres and six community hospitals in Boston, Massachusetts, US.

A team of investigators developed an elastic net model and validated it externally to predict PPD, defined as having a mood disorder, an antidepressant prescription, or a positive screen on the postpartum EPDS. Predictors used were as follows: medical history, sociodemographic factors, and prenatal depression screening information.