PREDICTIVE RISK MODELING OF MATERNAL MORTALITY IN GHANA: A MULTIVARIATE BIOSTATISTICAL FRAMEWORK

Maternal mortality in Ghana remains a critical public health challenge, with rural and low-income populations disproportionately affected despite modest progress toward global health targets. This study is vital as it addresses the lack of predictive, data-driven frameworks in existing interventions. The primary objective was to develop a multivariate biostatistical model that forecasts maternal mortality risk using socio-economic, demographic, antenatal care, and healthcare access variables. Utilizing secondary data from 2020 to 2024 covering 3,100 maternal death cases, the study employed chi-square tests, ANOVA, logistic regression, and ensemble machine learning models to assess predictors and model performance. Key findings revealed a strong relationship between low income and mortality (χ² = 36.22, p < 0.001), a significant influence of antenatal visit frequency on survival outcomes (F = 14.85, p < 0.01), and the predictive power of age, parity, and rural residence (Nagelkerke R² = 0.41, p < 0.001). The ensemble model achieved the highest accuracy (90%), sensitivity (88%), and specificity (91%), with an overall correlation coefficient of 0.99 between accuracy and sensitivity, and a final regression model R² of 0.68. These results affirm the model’s reliability in identifying high-risk cases. The findings have practical implications for deploying targeted mobile health services and integrating predictive analytics into national maternal health surveillance. It is recommended that Ghanaian policymakers adopt ensemble-based frameworks and enhance antenatal care outreach, especially in underserved regions, while empowering frontline health workers through data literacy training.