EPIDEMIC FORECASTING MODELS APPLIED TO MALARIA TRANSMISSION DYNAMICS

Background to the Study

Ross-Macdonald model dIh/dt = mab Im (1-Ih) - r Ih mosquito dIm/dt = ac Ih (1-Im) - μ Im threshold R0 = ma²bc / rμ. SEIR-SEI extension adds exposed compartments latency human intrinsic incubation 10-14 days mosquito extrinsic incubation 10-12 days. Seven ODEs: human dSh/dt = Λh - λh Sh - μh Sh dEh/dt = λh Sh - (σh+μh) Eh dIh/dt = σh Eh - (γh+μh+δh) Ih dRh/dt = γh Ih - μh Rh mosquito dSm/dt = Λm - λm Sm - μm Sm dEm/dt = λm Sm - (σm+μm) Em dIm/dt = σm Em - μm Im forces of infection λh = b βh Im/N h λm = b βm Ih/N h. Baseline parameters representative of holoendemic: biting rate b 0.5 per day transmission probabilities βh βm 0.2-0.4 mosquito mortality μm 0.1 per day human recovery γh 0.01 per day.

R0 derived via next-generation matrix F V⁻¹ spectral radius at DFE. Stability via Jacobian eigenvalues: if all Re(λ)<0 DFE stable else unstable; here eigenvalue 0.0299 positive indicates unstable with R0 1.788 >1. Endemic equilibrium located via fsolve Newton-Krylov 43.3 infectious humans 76.8 infectious mosquitoes per reference cohort. Numerical methods RK4 fourth-order accuracy O(h⁴) Euler first-order O(h) convergence study empirical orders 0.99 Euler 4.06 RK4 as expected LSODA adaptive validates RK4 max discrepancy 8.88e-4.

For malaria burden data see WHO World Malaria Report 2023 and CDC malaria biology and NCBI malaria SEIR-SEI modelling literature. Related mathematical biology in mathematical modeling project topics on SCHOLARNESTHUB.

Statement of the Problem

Despite substantial investment in malaria control across Nigeria over past two decades including large-scale distribution of insecticide-treated nets and periodic indoor residual spraying campaigns malaria transmission remains endemic in most parts and progress toward elimination stagnated in several states. Central difficulty facing programme managers absence of quantitative model-based framework linking specific intervention coverage levels to expected epidemiological impact. Intervention planning frequently guided by available funding and logistical convenience rather than rigorous assessment which transmission parameters most strongly influence disease persistence. Furthermore many existing applications of epidemic models to malaria in Nigerian literature present either purely qualitative formulations without full numerical validation or purely statistical forecasts without underlying mechanistic transmission structure former limits practical applicability because policy-relevant quantities such as reduction in R0 achievable under given ITN coverage level not computed latter limits interpretability because purely statistical forecasts cannot distinguish between transmission driven by vector abundance human infectiousness or seasonal climatic forcing. Study addresses gaps by formulating fully specified numerically solved independently validated SEIR-SEI compartmental model complemented by formal sensitivity analysis ranking parameters most amenable to intervention and by seasonal regression forecasting layer fitted to incidence data. Problem stated concisely: how can mathematically rigorous computationally validated epidemic model be constructed and analysed to forecast malaria transmission dynamics and quantitatively evaluate comparative effectiveness of vector-control intervention strategies in manner directly useful for public health decision-making?

Aim and Objectives of the Study

Aim is to develop analyse and numerically validate compartmental SEIR-SEI epidemic model for malaria transmission dynamics and apply model to forecast disease incidence and evaluate intervention scenarios.

·         Formulate coupled SEIR human - SEI mosquito system of ordinary differential equations describing malaria transmission dynamics;

·         Derive basic reproduction number R0 analytically using next-generation matrix method;

·         Solve resulting nonlinear system numerically using fourth-order Runge-Kutta method and validate solution against independent adaptive-step solver;

·         Determine disease-free and endemic equilibria of model and establish local stability properties using Jacobian eigenvalue analysis;

·         Conduct normalized forward sensitivity analysis to rank model parameters by proportional influence on R0;

·         Compare empirical convergence behaviour of Euler method and RK4 method against fine-step reference solution;

·         Fit seasonal regression model to simulated surveillance-representative monthly incidence data and generate six-month forecast; and

·         Simulate and compare intervention scenarios representing varying levels of insecticide-treated net and indoor residual spraying coverage.

Research Questions

·         What is analytical form and numerical value of basic reproduction number for formulated SEIR-SEI malaria transmission model under baseline parameter conditions?

·         How accurately does fourth-order Runge-Kutta method reproduce trajectory of independent adaptive-step numerical solver and what is its empirical order of convergence relative to Euler method?

·         Is disease-free equilibrium of model locally stable under baseline parameter conditions and what is location of endemic equilibrium?

·         Which model parameters exert greatest proportional influence on basic reproduction number as measured by normalized forward sensitivity indices?

·         What level of combined vector-control intervention coverage is required according to model to reduce basic reproduction number below elimination threshold of unity?

·         How well does seasonal regression model fit simulated surveillance-representative malaria incidence data and what does it forecast for subsequent six months?

Significance of the Study

To public health policymakers and malaria control programme managers provides quantitative sensitivity-ranked basis for allocating limited intervention resources toward parameters strategies that most effectively reduce transmission rather than uniform historically determined allocation sensitivity indices biting rate 1.000 mosquito death rate -0.708 justify vector control. To academic mathematics community demonstrates complete reproducible workflow for compartmental epidemic modelling from analytical derivation through numerical validation to statistical forecasting template for similar studies on other vector-borne diseases dengue yellow fever lymphatic filariasis prevalent in Nigeria. To students and researchers in applied mathematics mathematical biology biostatistics illustrates practical application of ODE theory numerical analysis linear algebra through next-generation matrix and Jacobian eigenvalue methods and statistical regression within single coherent problem reflecting increasingly interdisciplinary nature. To broader Nigerian public contributes toward national goal malaria elimination by strengthening analytical tools available for evidence-based intervention planning. Additional tools in public health project topics on SCHOLARNESTHUB.

Scope of the Study

Restricted to mathematical formulation numerical solution stability analysis sensitivity analysis and statistical forecasting of deterministic compartmental model of malaria transmission dynamics. Human population modelled using four compartments susceptible exposed infectious recovered and mosquito vector population using three compartments susceptible exposed infectious yielding seven-dimensional system nonlinear ODEs. Parameter values drawn from ranges reported in published malaria modelling literature for holoendemic Anopheles-borne Plasmodium falciparum transmission and not derived from primary field data. Numerical simulation covers two-year time horizon at daily resolution sufficient to capture initial epidemic wave and subsequent approach toward equilibrium. Forecasting component uses simulated 24-month monthly incidence series constructed to reflect seasonal pattern and reporting scale typical of Nigeria Malaria Elimination Programme NMEP and WHO surveillance summaries rather than unpublished primary dataset obtained through fieldwork. Study does not address genetic immunological or within-host parasite dynamics nor does it consider spatial heterogeneity in transmission across geographic zones.

Limitations of the Study

·         Model parameters drawn from published literature rather than primary entomological or clinical surveys conducted specifically for this study therefore subject to uncertainty inherent in such secondary sources.

·         Model deterministic and does not incorporate stochastic effects which may be significant when infectious populations small such as near epidemic threshold or during approach to elimination.

·         Model assumes homogeneously mixing population and does not account for spatial heterogeneity age structure or behavioural heterogeneity in exposure to mosquito bites.

·         Forecasting component fitted to simulated incidence series constructed to reflect realistic seasonal surveillance patterns rather than officially released primary dataset given constraints on access to disaggregated NMEP facility-level records within scope of undergraduate project.

·         Computational results dependent on fixed step size chosen for RK4 scheme although convergence formally verified extremely fine-scale short-term dynamics below daily resolution not resolved.

Operational Definition of Terms

Compartmental model: Mathematical model dividing population into discrete classes compartments based on disease status with flows between compartments governed by differential equations; here seven-dimensional SEIR human plus SEI mosquito.

Basic reproduction number R0: Expected number of secondary infections produced by single infectious individual introduced into entirely susceptible population derived via next-generation matrix evaluated at 1.788 baseline indicating persistence threshold unity.

Disease-free equilibrium DFE: Steady state of model in which no infection present in either human or mosquito population; stability via Jacobian eigenvalues under baseline one eigenvalue positive real part 0.0299 confirming instability when R0>1.

Endemic equilibrium EE: Steady state in which infection persists at constant non-zero level located via Newton-Krylov fsolve approximately 43.3 infectious humans and 76.8 infectious mosquitoes per reference cohort.

Force of infection: Per-capita rate at which susceptible individuals acquire infection expressed as function of number of infectious individuals in interacting population λh = b βh Im/N h λm = b βm Ih/N h.

Sensitivity index: Normalized measure of proportional change in model output here R0 resulting from proportional change in given parameter biting rate 1.000 greatest positive mosquito natural death rate -0.708 greatest negative magnitude guiding vector control.

Vector control: Public health interventions principally insecticide-treated nets and indoor residual spraying aimed at reducing mosquito-human contact or mosquito survival; scenario simulations ITN 60 percent reduces R0 to 0.983 below unity combined ITN IRS 80 percent reduces to 0.536 eliminating sustained transmission.

Numerical convergence order: Rate expressed as power of step size h at which error of numerical method decreases as h reduced empirical 0.99 Euler first-order 4.06 RK4 fourth-order validated against LSODA max discrepancy 8.88e-4.

Short Conclusion

SEIR-SEI model with R0 analytically derived via next-generation matrix evaluated at 1.788 under holoendemic baseline indicating persistence solved via RK4 validated against LSODA discrepancy 8.88e-4 convergence orders Euler 0.99 RK4 4.06 confirming accuracy. DFE instability confirmed via Jacobian eigenvalue 0.0299 positive consistent with R0>1 endemic equilibrium 43.3 infectious humans 76.8 infectious mosquitoes located via fsolve. Sensitivity analysis identified biting rate 1.000 and mosquito death rate -0.708 as most influential providing quantitative justification for vector control. Intervention scenarios ITN 60 percent reduces R0 to 0.983 below unity combined ITN IRS 80 percent reduces to 0.536 effectively eliminating transmission. Seasonal regression fitted to simulated 24-month incidence reflecting NMEP surveillance achieved R2 0.985 used for six-month forecast. Study concludes combined vector-control strategies most mathematically robust route to malaria elimination and recommends intervention planning informed by sensitivity-ranked parameters rather than uniform allocation. Methodology contributes reproducible data-validated tool for public health decision support in malaria-endemic regions. Further forecasting frameworks in malaria modeling project topics on SCHOLARNESTHUB.

Frequently Asked Questions

Q: What is SEIR-SEI malaria model?

A: Coupled compartmental system seven ODEs human Susceptible-Exposed-Infectious-Recovered and mosquito Susceptible-Exposed-Infectious with forces of infection linking populations capturing latency in both host and vector.

Q: What is basic reproduction number R0 in this study?

A: Expected secondary infections from single infectious in susceptible population derived via next-generation matrix evaluated numerically at 1.788 baseline holoendemic indicating sustained persistence threshold unity.

Q: How was numerical accuracy validated?

A: RK4 fourth-order scheme implemented in Python validated against independent adaptive-step LSODA solver maximum absolute discrepancy 8.88e-4 convergence study empirical orders Euler 0.99 RK4 4.06 consistent with theory.

Q: Is disease-free equilibrium stable?

A: Under baseline R0 1.788 >1 DFE unstable confirmed via Jacobian eigenvalue criterion one eigenvalue positive real part 0.0299; endemic equilibrium located 43.3 infectious humans 76.8 infectious mosquitoes per cohort via Newton-Krylov fsolve.

Q: Which parameters most influence R0?

A: Normalized forward sensitivity analysis identified mosquito biting rate sensitivity index 1.000 greatest positive influence and mosquito natural death rate -0.708 greatest negative magnitude justifying vector-control interventions.

Q: What ITN/IRS coverage needed to eliminate transmission in model?

A: ITN 60 percent alone reduces R0 to 0.983 just below unity combined ITN and IRS 80 percent reduces to 0.536 effectively eliminating sustained transmission.

Q: What forecasting component was used?

A: Seasonal regression model fitted to simulated 24-month monthly incidence series reflecting Nigeria Malaria Elimination Programme surveillance patterns achieving R2 0.985 used to forecast six-month horizon.

Q: What are limitations of deterministic SEIR-SEI model?

A: Parameters from literature not primary field data deterministic no stochastic effects near threshold assumes homogeneous mixing no spatial age behavioural heterogeneity forecasting fitted to simulated not primary facility-level NMEP data.

Q: Why is sensitivity analysis important for malaria control?

A: Ranks parameters by proportional influence on R0 enabling allocation of limited resources toward most effective interventions biting rate and mosquito mortality rather than uniform allocation.

Q: How does this model help Nigerian public health decision-making?

A: Provides reproducible computationally validated tool linking intervention coverage levels to R0 reduction and incidence forecasts supporting resource allocation drug procurement timing of vector-control campaigns toward national elimination goal.