Unraveling the role of rat and flea population dynamics on the seasonality of plague epidemics in Madagascar

Abstract

Plague continues to pose a public health problem in multiple regions of the world, including Madagascar, where it is characterized by a pronounced seasonal pattern. The drivers of plague seasonality remain poorly understood. Using a deterministic compartmental model, calibrated to rat and flea capture data, serological data collected in active rural foci, and human plague surveillance data, we analyzed the effects of seasonal rat and flea population dynamics on plague transmission. The models that incorporated seasonal fluctuations in rat and flea populations provided better predictive performances than those that did not. We found that a simpler mass-action model also performed well. Driven by these seasonal changes, the effective reproduction number (R_e) between rats peaks at 1.45 [95% credible interval (CI): 1.41, 1.48] in October and falls to 0.6 (95% CI: 0.57, 0.63) in March. We estimated that 0.5% (95% CI: 0.2%, 0.9%) of rats are infected annually, indicating that plague is not the main driver of rat population changes. Using our model, we evaluated intervention strategies and found that targeting both rats and their fleas at the start of the epidemic season (July-September) was the most effective approach for reducing human plague cases. Such an approach contrasts with the reactive strategy currently employed in Madagascar. Our findings highlight the role of flea and rat populations in plague seasonality and identify strategies that could be deployed in Madagascar to better control plague epidemics.

Keywords: Madagascar; modeling; plague; public health; seasonality.

Fig. 1.

Plague in Madagascar. (A) Plague transmission cycle that shows the interactions between rats, vector fleas, and humans. The cycle includes susceptible rats infested with uninfected fleas; rats infected by infected fleas; rats that die from plague, releasing infected fleas into the environment; and recovered rats, from which infected fleas die off, though infestation by fleas (infected or uninfected) may persist. (B) Map highlighting the regions of Madagascar where plague is endemic (dotted outlines). The study site in the Ankazobe District is marked in yellow. (C) Average monthly number of human plague cases in Madagascar between 2018 and 2023.

Fig. 2.

Model calibration. Comparison of data and model predictions for (A) the number of collected fleas, (B) the number of collected rats, (C) the number of collected plague seropositive rats, (D) the flea index (mean number of fleas per rat), (E) average monthly number of confirmed human plague cases between 2018 and 2023. For Figures A–D, black dots represent the median values of data aggregated temporally proximate capture days, with vertical lines indicating the range between the minimum and the maximum of observed values across those days. For Figure E, black dots represent the monthly average number of confirmed human plague cases in the data. The models include no seasonality (Model 1 – purple), seasonality in the rat population (Model 2 – yellow), seasonality in the flea population (Model 3 – green), seasonality in both rat and flea populations (Model 4 – red), and mass-action model with seasonality in both rat and flea populations (Model 5 – blue).

Fig. 3.

Plague epidemic in the rat population. (A) Effective reproduction number (Re) among rats, with the blue solid line showing the estimated R_e over time, and the black dashed horizontal line representing the epidemic threshold (R_e = 1). (B) Cumulative proportion of infected rats throughout a plague season from July 1 to June 30, assuming a probability of death upon infection of 0.3 (green dashed line), 0.5 (blue solid line, representing the baseline scenario), and 0.7 (red dashed line).

Fig. 4.

Impact of the control of rat and flea populations on the number of human plague cases. The figures show the proportion of human cases averted over one season (July 1 to June 30) as a function of the month of intervention when (A) only rats are targeted, (B) only fleas are targeted, and (C) both rats and the fleas on those rats are simultaneously targeted. Reduction levels are represented by gray dots (20% population reduction), red triangles (50% population reduction), and yellow rectangles (80% population reduction). Vertical bars indicate the 95% credible intervals.