Malaria transmission modelling and conditional prevention allocation in Uganda

Malaria transmission models can generate precise intervention rankings even when their parameters, observation mappings and programme costs remain uncertain. I evaluated an age-stratified SEIR–SEI framework in a standardised population of one million, using an unchanged 36-month historical synthetic target. The reconciled implementation maintained virtual age-group totals, fitted the transmission composite K = β₀M, integrated infectious onsets and disease deaths, and compared interventions from one phase-aligned baseline state over 720 days. An aggregate vaccine effect entered child infection before simulation, allowing transmission feedback.

Five calibration seeds gave total-series R² 0.9537; subgroup fits were weaker. The closed-form unforced reproduction index, 6.3787, matched the six-state next-generation matrix. The baseline produced 528,041 expected reported episodes. ITN coverage of 50% averted 36,692 episodes; vaccine coverage of 70% averted 2,123 under the assumed aggregate mechanism. At a hypothetical US$2 million additional prevention budget, five allocation seeds and a 10,201-point boundary grid selected 71.43% ITN coverage, averting 52,482.83 reported episodes. The same corner won within a 115-portfolio pool for all 300 assumed parameter draws and remained selected after 40 repeated noisy-target recalibrations. Halving the IRS round price and using two rounds changed the selected mix to 40.07% ITN and 21.95% IRS. All 17 mathematical and numerical assertions passed.

The study demonstrates reproducible conditional calculations and identifies important structural confounding. It does not establish empirical Uganda programme effectiveness, a clinical treatment allocation or a national cost-effectiveness recommendation. Empirical use requires compatible surveillance observations, a cohort-based vaccine mechanism, treatment-specific clinical effects and locally measured programme resources.