We applied the model-guided fieldwork framework to the Caribbean mongoose rabies system by parametrizing a spatially-explicit, individual-based model, and by performing an uncertainty analysis designed to identify parameters for which additional empirical data are most needed. Our analysis revealed important variation in output variables characterizing rabies dynamics, namely rabies persistence, exposure level, spatiotemporal distribution, and prevalence. Among epidemiological parameters, rabies transmission rate was the most influential, followed by rabies mortality and location, and size of the initial infection. The most influential landscape parameters included habitat-specific carrying capacities, landscape heterogeneity, and the level of resistance to dispersal associated with topography. Movement variables, including juvenile dispersal, adult fine-scale movement distances, and home range size, as well as life history traits such as age of independence, birth seasonality, and age- and sex-specific mortality were other important drivers of rabies dynamics. We discuss results in the context of mongoose ecology and its influence on disease transmission dynamics. Finally, we suggest empirical approaches and study design specificities that would provide optimal contributing data addressing the knowledge gaps identified by our approach, and would increase our potential to use epidemiological models to guide mongoose rabies control and management in the Caribbean.

The small Indian mongoose (Urva auropunctata) is the primary terrestrial wildlife rabies reservoir on at least four Caribbean islands, including Puerto Rico. In Puerto Rico, mongooses represent a risk to public health, based on direct human exposure and indirectly through the transmission of rabies virus to domestic animals. To date, the fundamental ecological relationships of space use among mongooses and between mongooses and domestic animals remain poorly understood. This study is the first to report mongoose home range estimates based on GPS telemetry, as well as concurrent space use among mongooses and free roaming domestic dogs (FRDD; Canis lupus familiaris). Mean (± SE) home range estimates from 19 mongooses in this study (145 ± 21 ha and 60 ± 14 ha for males and females, respectively) were greater than those reported in prior radiotelemetry studies in Puerto Rico. At the scale of their home range, mongooses preferentially used dry forest and shrubland areas, but tended to avoid brackish water vegetation, salt marshes, barren lands and developed areas. Home ranges from five FRDDs were highly variable in size (range 13–285 ha) and may be influenced by availability of reliable anthropogenic resources. Mongooses displayed high home range overlap (general overlap index, GOI = 82%). Home range overlap among mongooses and FRDDs was intermediate (GOI = 50%) and greater than home range overlap by FRDDs (GOI = 10%). Our results provide evidence that space use by both species presents opportunities for interspecific interaction and contact and suggests that human provisioning of dogs may play a role in limiting interactions between stray dogs and mongooses.

This study aimed to molecularly survey and evaluate the genetic diversity of Bartonella spp. in mongooses and their fleas from St. Kitts. Spleen (n = 54), blood (n = 71), and pooled flea samples, all identified as Ctenocephalides felis (n = 53), were submitted to TaqMan real-time quantitative PCR (qPCR) targeting Bartonella-nuoG fragment (84 bp). Positive samples underwent further conventional PCR assays targeting five loci (gltA, rpoB, fstZ, nuoG, and ITS), subsequent sequencing, and phylogenetic and haplotype analyses. The overall occurrence of Bartonella spp. in mongooses and fleas was 51.2% (64/125 [95% CI (42.1–60.2%)]) and 62.3% (33/53) [95% CI (47.9–75.2%)]), respectively. From samples sequenced across the five loci, 50.8% (33/65) were identified as Bartonella henselae, 26.2% (17/65) were 96.74–99.01% similar by BLAST analysis to an unidentified Bartonella sp. previously reported in Japanese badgers (Meles anakuma), and 23.1% (15/65) were co-infected with both species. Nucleotide polymorphism analysis showed low diversity amongst haplotypes but did concur with phylogenetic analysis, placing the unidentified species in a separate clade from B. henselae by multiple mutational events. Our data confirms that mongooses and Ctenocephalides felis fleas collected from them are not only potential reservoirs for B. henselae but also a novel Bartonella sp. which we propose be called ‘Candidatus Bartonella kittensis’.

The small Indian mongoose (Urva auropunctata) is a non-native invasive species across the Caribbean and a rabies reservoir on at least four islands in the region. Although previous studies reported mongoose density estimates in their non-native range, the variability in trapping designs, study seasonality, and analytical methods among studies precludes direct comparisons. This study is the first to report mongoose densities for the island of St. Kitts, West Indies. Our objective was to quantify mongoose densities across four habitats characteristic for the island. High capture and recapture rates in this study resulted in detailed estimates of spatial heterogeneity in mongoose densities, ranging from 0.53 (CI95: 0.46–0.61) mongooses/ha in suburban habitat to 5.85 (CI95: 4.42–7.76) mongooses/ha in nearby dry forest. Estimates were robust to the estimation method used (correlation among methods, r > 0.9). Female-biased sex ratios estimated from fall season versus mostly unbiased sex ratios estimated from summer season suggests seasonality in capture success resulting from differences in sex-specific activity patterns of mongooses. We found no effect of habitat characteristics, at the scale of trap placements, associated with mongoose capture success.

Agent-based models (ABMs) are computational models for simulating the actions and interactions of autonomous agents in time and space. These models allow users to simulate the complex interactions between individual agents and the landscapes they inhabit and are increasingly used in epidemiology to understand complex phenomena and make predictions. However, as the complexity of the simulated systems increases, notably when disease control interventions are considered, model flexibility and processing speed can become limiting. Here we introduce SamPy, an open-source Python library for stochastic agent-based modeling of epidemics. SamPy is a modular toolkit for model development, providing adaptable modules that capture host movement, disease dynamics, and disease control interventions. Memory optimization and design provide high computational efficiency allowing modelling of large, spatially-explicit populations of agents over extensive geographical areas. In this article, we demonstrate the high flexibility and processing speed of this new library. The version of SamPy considered in this paper is available at https://github.com/sampy-project/sampy-paper .