Identifying prenatal risk factors of postpartum depression with machine learning

Area of Science:

• Reproductive Psychiatry
• Computational Psychiatry
• Maternal Mental Health

Background:

• Postpartum depression (PPD) affects maternal and child well-being.
• Early intervention is crucial but challenging due to difficulties in identifying at-risk women.
• Predictive timing and key prenatal risk factors for PPD remain unclear.

Purpose of the Study:

• To determine the earliest point at which PPD can be predicted during pregnancy.
• To identify specific risk factors for PPD that are informative at different pregnancy stages.
• To leverage machine learning for enhanced PPD prediction and risk factor identification.

Main Methods:

• Applied machine learning algorithms, specifically Lasso-regularized linear regression.
• Utilized data from 2865 mothers in the Brabant Study, including 233 variables.
• Analyzed data across all pregnancy trimesters to identify predictive factors.

Main Results:

• A Lasso model identified depressive symptoms, negative affectivity, neuroticism, BMI, and mental health history as key predictors.
• The model achieved high specificity but moderate sensitivity, effectively ruling out low-risk cases.
• Consistent predictive performance across trimesters suggests early identification (12 weeks) is feasible.

Conclusions:

• Low-risk postpartum depression status can be identified as early as 12 weeks of gestation.
• Machine learning models can aid clinicians in efficiently allocating preventive resources by identifying low-risk individuals.
• Early identification of low-risk pregnancies allows for targeted support for women at higher risk of PPD.