Fig 1.
Schematic outlining the inter- and intra-island processes of livestock in the mathematical model.
To evaluate the effect of vaccine control measures against Rift Valley fever virus (RVFV), we developed a mathematical model to describe RVFV infection in livestock (cattle, sheep and goats). (A) In the model, animals moved between the four islands in the Comoros archipelago—Grande Comore (red), Mohéli (blue), Anjouan (green) and Mayotte (purple)—as governed by the livestock trade network as shown. The administrative boundary data were made available under CC BY 3.0 [22] and CC BY-IGO [23] licences. The CC BY 3.0 licensed the data for Mayotte [24], and the CC BY-IGO licensed the data for the Comoros [25], Tanzania [26], Madagascar [27] and Mozambique [28]. All presented data was unaltered. (B) The livestock population of each island was divided into compartments defined by infection status, vaccination status, and age. The diagram shows the compartments (squares) and the direction of transfer of livestock between compartments (arrows). The compartments eligible for vaccination depended on whether or not livestock were tagged (black circles) or not (white circles). The notation in each compartment defines the infection status (horizontally arranged) of the animals: susceptible to the virus (S), infected but not yet infectious (E), infectious (I) or recovered with life-long immunity to reinfection (R). The superscripts denote the vaccination status (vertically arranged) of animals: unvaccinated (U), vaccinated with developing protection ( and
), or vaccinated and partially protected from infection (W). The transitions between compartments of different age groups (diagonally arranged) are not drawn for illustrative purposes.
Table 1.
Vaccine strategies evaluated using the mathematical model. The mathematical model describing infection with and vaccination against RVFV in livestock across the archipelago was used to assess the effectiveness of six vaccination strategies. These vaccine strategies differed in their assumption on whether or not livestock were identifiable via tagging, and how the vaccines were allocated to the four islands in the Comoros archipelago.
Fig 2.
Vaccine allocation strategies across islands in the Comoros archipelago.
Vaccines were allocated to each of the four islands in the archipelago—Grande Comore (red), Mohéli (blue), Anjouan (green) and Mayotte (purple)—either proportionally to the livestock population size of each island (proportional allocation; left-most panel), or optimally to maximise the percentage of infections averted across the archipelago (globally optimal allocation; middle panel), or the percentage of infections averted on the island with the worst performance (equity-focused allocation; right-most panel). Optimally allocating vaccines based on the percentage of infections averted across the archipelago was dependent on the vaccination rate, exhibiting a shift from allocating the majority of vaccines to Anjouan for lower vaccination rates, to Grande Comore for higher vaccination rates. The vertical bars show the percentage of vaccines assigned to each island for different vaccination rates and tagging strategies.
Fig 3.
Effectiveness of different vaccine strategies against Rift Valley fever virus (RVFV) across the Comoros archipelago.
For a range of vaccination rates, allocating vaccines to the four islands in the Comoros archipelago optimally by percentage of infections averted across the archipelago (orange) and by percentage of infections averted on the worst-performing island (blue; where the worst-performing island was defined as the island with the lowest percentage of infections averted) outperformed allocating vaccines proportionally to the population size of each island (grey). Tagging animals upon vaccination (black circles) also outperformed strategies where animals were not tagged (white circles), where the difference was amplified for increased vaccination rates. The violins show the percentage of infections averted across the Comoros archipelago for different annual vaccination rates, allocation methods and tagging strategies. The points and boxplots show the median and inter-quartile range for each scenario respectively. All metrics shown were based on 25,000 model simulations.
Fig 4.
Equity in vaccine strategy effectiveness between islands in the Comoros archipelago.
Optimally allocating vaccines in terms of total infections averted across the archipelago (globally optimal allocation; orange panels) resulted in an imbalance of vaccine strategy effectiveness at the island-level at low vaccination rates. To ameliorate this, vaccines were also allocated to maximise the infections averted on the worst performing island (equity-focused allocation; blue panels), resulting in a less pronounced imbalance of strategy effectiveness between islands at low vaccination rates. Shown is the median and 95% prediction interval of the percentage of infections averted on each island when vaccinating 5% and 30% of livestock across the archipelago annually under each vaccine allocation and tagging strategy. Summary statistics were generated using 25,000 model simulations. The administrative boundary data were made available under CC BY 3.0 [22] and CC BY-IGO [23] licences. The CC BY 3.0 licensed the data for Mayotte [24], and the CC BY-IGO licensed the data for the Comoros [25]. All presented data was unaltered.