In rural sub-Saharan Africa (sSA), the burden of antimicrobial resistance (AMR) remains high. As AMR continues to rise, there is a strong need for practical, implementable surveillance to monitor and mitigate risks, as well as inform timely, evidence-based clinical decision-making. Emerging evidence points to possible community-level drivers, such as transmission between human, animal and environmental reservoirs as contributing factors, yet microbiological surveillance or opportunities for wastewater-based surveillance are often limited and insufficient in these settings. Therefore, alternative sustainable and affordable approaches are needed. We intend to build on the demonstrated potential of metagenomic profiling of pooled faecal material, which accurately predicted population-level AMR prevalence in invasive Enterobacterales infections.

We aim to validate this metagenomic pooled approach on additional populations, and to evaluate whether AMR patterns could be similarly predicted from surveillance of community One Health reservoirs. We will assemble existing data from hospital-based microbiology diagnostic laboratories in rural Burkina Faso and Kenya, and determine to what extent community-level metagenomic data, and/or faecal material of patients on hospital admission, can predict AMR in clinical isolates. We will perform community-level surveys in eight clusters per country, randomly selecting 15 households per cluster. We will systematically sample suspected environmental AMR exposure sites in and around households (soil, drinking water, latrines, chicken faeces) and collect data on community-level antibiotic use, hygiene practices, contact with domestic animals and sanitary facilities. Samples and data will be collected twice: during the dry and during the rainy season.

In addition to evaluating the accuracy of predicting resistance in clinical isolates, we will quantify community-level exposure risks. We will conduct metagenomic profiling on pooled DNA extracts from human stool samples (hospital and community-level) and from household environments. Bayesian statistical models will quantify relationships between AMR gene abundance in the environment and in human stool, and invasive bacteria identified among clinical patients, accounting for geography and seasonality. A cost-utility analysis will determine under what circumstances the use of pooled metagenomic data to inform empirical antibiotic policies would represent an efficient use of resources.

The proposed surveillance protocol is developed in partnership with local communities and local and international researchers and has received ethical approval in Kenya and Burkina Faso. It will assess whether intermittent, pooled-sample metagenomics provides a viable, low-cost and practical approach for population-level AMR surveillance in settings that—like many in rural sSA—lack systematic microbiological diagnostics and where sewage systems for wastewater-based surveillance are absent. By providing an alternative to routine microbiological-based surveillance where this proves challenging to implement, this approach may help improve treatment outcomes, contribute to equity and public health. Findings will be disseminated through peer-reviewed publications and academic conferences and will contribute to the recently proposed WHO AMR surveillance strategy, which combines survey-based approaches with routine AMR surveillance.