Investigating Protective and Risk Factors and Predictive Insights for Aboriginal Perinatal Mental Health: Explainable...

Area of Science:

• Artificial Intelligence in Healthcare
• Perinatal Mental Health
• Indigenous Health Research

Background:

• Perinatal depression and anxiety pose significant risks to maternal and infant well-being, with Aboriginal women facing heightened vulnerability due to colonization's impacts.
• The Baby Coming You Ready (BCYR) model offers a strengths-based, digitized approach, but reliance on traditional risk scores can overlook contextual factors for Aboriginal mothers.
• Improved clinical decision-making tools are needed to better support Aboriginal mothers' perinatal mental health, considering cultural strengths and protective factors.

Purpose of the Study:

• To explore explainable artificial intelligence (XAI) powered machine learning for culturally informed, strengths-based prediction of perinatal psychological distress in Aboriginal mothers.
• To develop models that identify and evaluate influential risk and protective factors.
• To ensure AI-driven decisions are transparent and interpretable for healthcare professionals.

Main Methods:

• Utilized deidentified data from 293 Aboriginal mothers in the BCYR program (Perth, WA, 2021-2023).
• Selected 20 key variables from a dataset including cultural strengths, life events, and psychosocial factors, using Kessler-5 for distress.
• Developed and compared machine learning models (EBM, RF, XGBoost, etc.), applying XAI techniques (SHAP, LIME) for interpretability.

Main Results:

• The Explainable Boosting Machine (EBM) demonstrated superior predictive performance (accuracy 0.849, AUC 0.821).
• XAI techniques consistently identified key factors like "Feeling Lonely," "Blaming Herself," and "Makes Family Proud."
• Individual-level insights were provided, visually distinguishing protective and risk factors' impact on predictions.

Conclusions:

• XAI-driven models show significant potential for predicting psychological distress in Aboriginal mothers.
• These models offer clear, human-interpretable explanations of influential factors, aiding clinical decision-making.
• The findings suggest XAI can facilitate more informed, equitable perinatal mental health screenings for Aboriginal women.