Table 2.
Summary of models of malaria transmission
| EMOD DTK | Imperial | MORU | OpenMalaria | |
|---|---|---|---|---|
| Institutional home | Institute for Disease Modelling | Imperial College London | Mahidol Oxford Tropical Medicine Research Unit | Swiss Tropical and Public Health Institute |
| Type of model and references | Individual-based stochastic microsimulation16, 17 | Individual-based stochastic microsimulations of malaria in human beings linked to a stochastic compartmental model for mosquitoes11 | Deterministic compartmental model described by differential equations,18 including drug action on each stage of the infection | Single-location individual-based simulation of malaria in human beings14 linked to deterministic model of malaria in mosquitoes19 |
| How infections are tracked | Tracks parasite densities of different surface-antigen types | Tracks membership of categories of infection (symptomatic, asymptomatic, submicroscopic, treated) | Tracks membership of categories of infection | Tracks parasite densities corresponding to different infection events |
| Relationship between entomological innoculation rate and prevalence | Immunity is acquired through cumulative exposure to different antigenic determinants,20 with heterogeneity in individual biting rates included | Immunity is acquired through cumulative exposure to mosquito bites, with heterogeneity in individual biting rates included | Subdivides population into non-immune and immune classes | Submodels of infection of human beings14 and of blood-stage parasite densities, with main immune effects controlling parasite densities21 |
| Duration of infections | Infection duration based on malaria therapy20 and cross-sectional survey data22 | Infection duration based on fitting to asexual parasite prevalence data by age, transmission intensity, seasonality | Infection duration based on malaria therapy data and data from endemic areas | Infection duration based on malaria therapy data21 |
| Effect of mass drug administration or case management | Reduces blood-stage parasite densities according to age-specific and dose-specific pharmacokinetics and pharmacodynamics,22 with corresponding clearance and prophylactic effects | Truncates infections and has subsequent prophylactic effect based on fitting pharmacokinetic and pharmacodynamic models to field studies | Post-treatment prophylactic period derived from field studies of time to next infection | Truncates infections, and has subsequent prophylactic effect based on pharmacokinetic and pharmacodynamic studies |
| Validation against trials of mass drug administration or mass screening and treatment | Assessed against MACEPA trial of mass screening and treatment in southern Zambia23 | Assessed against a controlled trial24 of mass drug administration in Burkina Faso (model slightly optimistic about effect vs data), and the MACEPA trial of mass screening and treatment in southern Zambia (model matched data) | Fitted to a trial of mass drug administration in Cambodia25 | Fitted to the data of the Garki project (Matsari),26 and assessed against the MACEPA trial of mass screening and treatment in southern Zambia27 |
| Infectiousness to mosquitoes | A function of mature gametocyte and cytokine densities20, 22 | Related to asexual parasite dynamics and lagged to allow for development of gametocytes | Infected individuals have a constant infectiousness | Lagged function of asexual parasite density28 |
| Heterogeneity in exposure | Age-dependent biting29 and configurable distribution of household variability (the latter disabled in this analysis) | Included | Not included | Included |
| Initial state | .. | Back-calculating required mosquito density to achieve given initial prevalence at an approximate steady state in the presence of treatment and long-lasting insecticide-treated nets | Set transmission rate to achieve given initial prevalence at an approximate steady state in the presence of treatment | Back-calculating required mosquito density to achieve given initial prevalence at an approximate steady state in the presence of treatment |
| Source of seasonality pattern | Rainfall and imputed temperature29 driving larval habitat model fitted to clinical incidence patterns in Sinazongwe and Gwembe districts, Zambia | Rainfall data from Zambia combined with larval and adult mosquito model | Same entomological innoculation rate input as Imperial model | Based on pattern for southern Zambia29 |
| Age-structured model | Yes | Yes | No | Yes |
| Simulation of correlated rounds of intervention | Yes | Yes | No | Yes |
All the models are extensible to include other functionality (eg, different drugs, effects of drug resistance, effect on drug resistance, vector bionomics and details of vector control, different initial conditions, other concomitant interventions). A detailed comparison of EMOD DTK, Imperial, and OpenMalaria, including references to the data to which they are fitted, is available elsewhere.15 DTK=Disease Transmission Kernel. MORU=Mahidol Oxford Tropical Medicine Research Unit. MACEPA=Malaria Control and Elimination Partnership in Africa.