Annual Maternal Mortality Prediction for Zambia Using The Multilayer Perceptron

Abstract

Despite global commitments to reducing preventable maternal deaths, maternal mortality remains unacceptably high in Sub-Saharan Africa, with Zambia exhibiting persistent disparities in reproductive health outcomes. As the 2030 deadline for achieving Sustainable Development Goal 3.1 approaches, robust predictive analytics are essential for evidence-based health systems strengthening and resource allocation. This study employs advanced machine learning techniques to model and forecast Zambia’s maternal mortality ratio (MMR) trajectory through 2035, thereby providing critical foresight for strategic health policy formulation and programmatic intervention. The study utilized annual time series data on Zambia’s MMR spanning 1985–2023, sourced from the World Bank database. A Multilayer Perceptron Artificial Neural Network (ANN) with a (12,12,1) architecture employing the hyperbolic tangent activation function was constructed. Model stability and predictive accuracy were rigorously evaluated using mean squared error, mean absolute error and residual diagnostics. Out-of-sample forecasts were generated for the 2024–2035 period. The ANN model demonstrated excellent stability (MSE = 70.16; MAE = 4.49), with residual analysis confirming no systematic bias. Forecasts indicate that Zambia’s MMR will plateau at approximately 95– 98 deaths per 100,000 live births throughout the projection horizon, with a marginal declining trend from 97.95 in 2024 to 94.71 by 2035. Critically, these projections suggest Zambia will substantially miss the SDG 3.1 target of 70 deaths per 100,000 live births by 2030, with the projected MMR remaining 35–40% above the aspirational threshold. The persistence of MMR above 94 per 100,000 live births over the next decade signals systemic failures in Zambia’s maternal health continuum of care that cannot be addressed through incremental improvements alone. Without transformative policy interventions, the nation will fail to meet its SDG commitments. The study proposes a five-pronged policy framework.