Developing an AI Model to Identify and Support Women at Risk of Postpartum Depression in Pakistan: A Longitudinal Study Protocol

Authors:
Open Access
Publisher website
Google Scholar
Abstract
Background Postpartum depression (PPD) is a major global health concern, particularly in low- and middle-income countries such as Pakistan, where prevalence ranges from 28%–63%. PPD affects maternal well-being and mother–child bonding. Perinatal mental health services are limited by stigma, cultural barriers, and a lack of care. Current approaches rely primarily on symptom-based screening and reactive treatment. Therefore, the study aims to develop and validate an AI-based model and digital application to identify pregnant women at risk of perinatal anxiety and PPD and to facilitate timely intervention. Methods This study adopts a three-phase translational research design to develop, validate, and implement an AI-enabled digital application, Sehatmand Maa (Healthy Mother), for early identification and management of PPD in Pakistan. Phase 1, a prospective longitudinal cohort already underway, will enroll 1,581 pregnant women from three Maternal and Child Health facilities in Karachi between April 2024 and December 2026. Participants will be assessed at four time points: two during the prenatal period (T 0 and T 1 ) and two during the postnatal period (T 3 and T 4 ), using clinical and psychosocial measures, including EPDS, PSAS, PSQI, and MSPSS. Phase 2, currently in progress, uses machine-learning algorithms, including CatBoost, to develop and validate separate predictive models for perinatal anxiety and PPD. Phase 3, not yet started, will integrate validated models into Sehatmand Maa and pilot the application across the three facilities to assess feasibility, usability, and integration into routine maternal healthcare. Expected Results The study will generate context-specific evidence on predictors of perinatal anxiety and PPD and establish validated AI-based models for early risk identification. The application is expected to support risk stratification, clinical decision-making, and timely referral of high-risk women. Pilot findings will inform refinement before broader implementation, with the potential to provide a scalable, AI-enabled Tool for maternal mental health care in low-resource settings.
Funding Information
  • Foreign, Commonwealth and Development Office (53488)
  • International Development Research Centre (53488)

This publication has 27 references indexed in Scilit: