Assessment and Quantification of Maternal Morbidity Using a Synthetic Ghanaian Cohort

Abstract

The causes of psychological birth trauma have received little attention from researchers, who primarily concentrate on the effects of postpartum mental health conditions like postpartum PTSD. The main objective of this study was to develop an integrated statistical and machine-learning framework to identify salient features of childbirth-related complications and to statistical model for early prediction of side-effects. Analysis was done using synthetic data from Ghana Health Centre. Using demographic, obstetric, clinical, laboratory, economic, dietary, and functional-status variables, we constructed a synthetic Ghana-specific cohort of 10,000 women of reproductive age. Logistic regression and Cox proportional hazards models were applied alongside decision trees, random forests, k-nearest neighbours, naïve Bayes, and neural networks to assess the consistency and relative importance of predictors. Across models, salient factors included maternal age, antenatal care attendance, prior obstetric complications, postpartum hemorrhage, pre-eclampsia, gestational diabetes, infections, hemoglobin levels, National Health Insurance Scheme (NHIS) status, delivery and transport costs, food expenditure, and limitations in activities of daily living. Women with a history of complications or postpartum hemorrhage experience adverse outcomes during childbirth. Overall, logistic regression performed robustly and offers interpretable, policy-relevant insights to support maternal health planning and resource allocation in Ghana.