Modelling Malaria Mortality in Ghana Using Distributed Lag Nonlinear Frameworks
Background: Malaria mortality studies in Ghana and sub-Saharan Africa have commonly relied on traditional regression approaches or generalised additive models, which may not fully represent delayed transmission effects, nonlinear exposure-response patterns, or uncertainty in mortality estimates. This study modelled the delayed and nonlinear dynamics of malaria mortality in Ghana and examined incidence-related thresholds associated with cumulative mortality effects. Objectives: The study applied a Distributed Lag Nonlinear Model (DLNM) to assess temporal malaria mortality dynamics in Ghana using a climate-sensitive health modelling framework that incorporated uncertainty-adjusted estimation. Methods: An ecological time-series analysis was conducted using annual age-standardised malaria mortality and incidence data for Ghana from 1990 to 2023 obtained from the Global Burden of Disease database. DLNMs with nonlinear spline functions, distributed lag structures and inverse-variance weighting were fitted separately for males, females and the combined population. Model performance was assessed using adjusted R², deviance explained, Akaike Information Criterion and residual diagnostics. Results: Malaria mortality and incidence declined over the study period, while mortality remained consistently higher among males. The DLNM performed well, with adjusted R² values ranging between 0.9936 and 0.9945, and deviance explained over 99.6%. Nonlinear temporal effects were observed in all sex groups (p < 0.05). Delayed mortality amplification was mainly observed at lags 2-4, with higher cumulative delayed effects among males. The maximum cumulative mortality effects occurred at incidence thresholds around 21,000 cases per 100,000 population, with the highest cumulative effect estimated for males (49.28). Conclusions: The findings indicate that malaria mortality dynamics in Ghana are nonlinear, delayed and threshold-sensitive. The combined use of DLNM, sex-specific lag modelling and uncertainty-adjusted weighting provides a useful framework for strengthening malaria surveillance, early-warning modelling and targeted public health planning.