In 2026, Bundibugyo virus disease (BDBV) caused a large, spatially dispersed outbreak in eastern Democratic Republic of the Congo. The authors attribute part of this dispersion to delayed clinical recognition and multi-day sample transport that slowed laboratory confirmation. They hypothesized that decentralized diagnostics, when placed strategically in frontier health zones, could function as a spatial fence: converting delayed recognition into timely PCR confirmation, isolation, and contact tracing before imported viral lineages became locally established.
To evaluate this hypothesis, the team fit a Bayesian spatial branching-process model to public situation reports through 2 August 2026. Key inputs included human movement data from Flowminder and a latent-burden anchor to align transmission potential and unobserved infections. The model was parameterized to capture overdispersed offspring distributions for viral lineages and to represent operational processes including PCR access and turnaround, clinical recognition, isolation, contact tracing, safe burial practices, and local behavior change.
They generated 2,000 posterior outbreak trees to explore counterfactual deployment strategies for decentralized diagnostics.
The simulation experiments varied three primary features:
Because the observed epidemic trajectory did not itself define the counterfactual future control trajectory, the authors assumed a national response ramp that produced a West-Africa–scale status-quo burden and then replayed the posterior trees under the intervention scenarios.
Speed of deployment was the dominant determinant of impact. Using the posterior ensemble, the authors report that 35 decentralized PCR laboratories averted a median 49% of future deaths if deployed immediately. The same number of labs resulted in a median 38% reduction with a 14-day deployment delay and an 11% reduction with a 60-day delay. The source provides corresponding 90% prediction intervals for these estimates.
Overall, faster establishment of lab capacity substantially increased the proportion of cases and deaths averted in the modeled futures.
Placement strategy influenced outcomes when deployment was not instantaneous. With a 30-day delay, model-informed placement of decentralized laboratories averted 18–33% of future deaths across scenarios with 20–100 labs. By contrast, random placement of the same number of laboratories produced much smaller impacts, averting only 1–8% of future deaths in the same delayed-deployment scenario. This contrast supports the value of model-guided selection of deployment sites ahead of likely importations.
Postmortem RDTs were simulated as an adjunct to live-case detection, designed to capture infections missed because cases died before testing. In the modeled example, adding postmortem RDTs increased the fraction of future deaths averted—for instance, with 20 decentralized PCR labs, a 14-day deployment delay, and 75% postmortem RDT coverage, deaths averted increased from 26% to 41%.
This result indicates that RDTs can partially compensate for gaps in timely live-case confirmation, particularly where delays to laboratory confirmation are expected.
The authors interpret their findings to mean that decentralized diagnostics can slow spatial spread of BDBV when three conditions are met: (1) diagnostics are placed ahead of likely importations using model guidance; (2) deployment is rapid; and (3) diagnostic placement is paired with sufficient staffing and downstream response capacities, including isolation, contact tracing, and safe-burial services. Under these conditions, conversion of delayed recognition into timely confirmation can prevent imported lineages from becoming established.
Policy implications drawn by the authors include prioritizing rapid deployment and model-informed geographic targeting of zone-level PCR capacity during spatially dispersed outbreaks. Complementary investments in postmortem RDTs and in operational response elements (staffing, isolation facilities, tracing teams, safe-burial capacity) are necessary to realize the full protective effect of decentralized diagnostics.
The study is simulation-based; details and assumptions are those encoded in the Bayesian spatial branching-process model and the specified national response ramp. The source states the posterior reproduced observed case, death, latent-burden, and spatial targets, but further methodological details, parameter priors, and sensitivity analyses are documented in the manuscript and source code. The authors declared no competing interests and reported that relevant ethical guidelines were followed. Source code for the simulations is available at the project GitHub repository: https://github.com/InstituteforDiseaseModeling/bundibugyo-decentralized-diagnostics
Practical interpretation should acknowledge that modeled effectiveness depends on the validity of movement data, burden anchoring, model structure, and the feasibility of rapid, model-guided deployment in real operational contexts.