Artificial intelligence-oriented predictive model for the risk of postpartum depression: a systematic review.

Area of Science:

• Computational psychiatry
• Maternal mental health
• Predictive modeling in healthcare

Background:

• Postpartum depression (PPD) affects millions of mothers globally, posing risks to maternal and infant well-being.
• Early prediction and intervention are crucial for mitigating PPD's adverse outcomes.
• AI and ML offer novel approaches for identifying PPD risk.

Purpose of the Study:

• To systematically review studies employing AI/ML algorithms for PPD prediction.
• To assess the performance and identify key predictors of PPD from existing research.
• To evaluate the potential and limitations of AI/ML models for PPD risk assessment.

Main Methods:

• Systematic review of studies published until October 31, 2024.
• Inclusion of studies using algorithms for PPD prediction.
• Quality assessment using the Prediction Model Risk Of Bias Assessment Tool (PROBAST).

Main Results:

• Eleven studies were included, with algorithms like random forest, SVM, and logistic regression showing high predictive performance (AUROC > 0.9).
• Key predictors identified include maternal age, pregnancy stress, mental health history, education, marital status, and sleep.
• Models demonstrated excellent overall performance but faced limitations in generalizability and potential bias.

Conclusions:

• AI/ML models show promise for early PPD risk prediction and intervention support.
• Future research should focus on optimizing models, enhancing accuracy, and ensuring cross-cultural applicability.
• Addressing data quality and algorithm interpretability is essential for widespread adoption.