Household-scale animal husbandry is pervasive across low- and middle-income countries (LMICs), delivering economic, labor and nutritional benefits while creating potential pathways for pathogen transmission between animals and people. The authors note that roughly 70% of the world’s 880 million rural poor rely on traditional extensive livestock systems. In many settings animals such as poultry, swine and ruminants are kept in close proximity to children and other household members, sometimes sharing domestic space and contributing to exposure to zoonotic pathogens implicated in diarrheal disease and other infections. The paper frames mapping household livestock ownership as an important input for infectious disease risk assessment and surveillance planning.
The analysis drew on georeferenced microdata from established population-based survey programs covering the majority of LMICs (examples include Demographic and Health Surveys, Malaria Indicator Surveys, AIDS Indicator Surveys, Multiple Indicator Cluster Surveys, and China Family Panel Studies). Household responses indicating ownership of animal taxa were categorized and georeferenced. These survey records were spatially matched to a panel of time-fixed environmental and demographic covariates (covariate sources are listed in the article’s supporting materials and Table 2). The study authors emphasize that the human-subject datasets are owned by third parties and are publicly accessible through their respective custodians; the prediction outputs from this work are available via a Dryad repository.
The authors used geostatistical methods to model and map the prevalence of household-scale livestock ownership for three taxa: poultry, swine, and ruminants. Models were fitted to the combined survey-covariate database using integrated nested Laplace approximation (INLA) frameworks and related geostatistical tools. To accommodate spatial non-stationarity—particularly evident for swine—the team implemented region-specific spatial kernels. Template Model Builder (TMB) was used in the prediction workflow, enabling efficient parameter estimation and rasterization of predicted probabilities for ownership of each taxon across broad geographic extents. Model parameter estimates were used to compute probability surfaces representing predicted household ownership prevalence.
Predicted maps revealed distinct geographic patterns for each animal taxon. Poultry ownership was predicted to be widespread across rural Central America, the Amazon basin, much of tropical Africa, and river basins and forested areas of East Asia. Swine husbandry showed the most spatially restricted footprint among the three taxa, concentrated in an undulating belt from central China through Southeast Asia into northeastern India; the authors caution that in regions with sparse survey data—particularly some Muslim-majority areas—model predictions reflect regional covariate patterns rather than fine-scale, empiric measurements. For ruminants, predicted prevalence was broad and included subequatorial Africa, Central Asia, the Gobi Desert, the Himalayas, Mongolia and northern India. The study provides GIS-compatible prediction files as supplementary outputs.
The models were evaluated using standard metrics and, according to the authors, performed impressively by most evaluation criteria. Predicted patterns were compared with available external evidence and found to align with known regional distributions of household livestock rearing. Specific numeric performance metrics and cross-validation details are reported in the article’s results and supplementary tables and figures.
Because domestic animals are reservoirs for many zoonotic pathogens (the authors cite that about 75% of emerging infectious diseases and 61% of all human pathogens are zoonotic-capable), mapping household ownership provides essential spatial context for assessing exposure risk. The predicted prevalence surfaces can help identify regions with high potential exposure to animal disease reservoirs and support targeting of surveillance, vaccination, and education programs. The authors note that these mapped risk-factor layers complement other environmental and health datasets used in One Health planning and infectious disease modeling.
The paper highlights that the underlying human-subject survey datasets are publicly available through their originating programs and that the prediction datasets produced by this analysis are publicly accessible on Dryad in GIS-compatible format. Supporting information (including covariate sources and model details) is provided in the manuscript’s supplementary files.
This research received financial support from the U.S. National Institutes of Health (NIAID and Fogarty), the European Union Horizon Europe programme, institutional funds at the University of Virginia and other supporting entities listed in the article. The authors report no competing interests.
(Article citation: Colston JM et al., PLoS One. Published July 31, 2026. DOI: 10.1371/journal.pone.0355207.)