Methods
Model Structure
The CDC employed a model to simulate outbreaks of BVD, which was adapted from prior models used during viral hemorrhagic fever incidents, like the Marburg virus disease outbreak in Ethiopia set for 2025. In this setup, each outbreak began with one infected individual, representing the original case from a zoonotic event. This initial case infected a randomly determined number of additional people, based on assumptions about the basic reproductive number (R0). Infected individuals entered the simulation at intervals selected according to the distribution of timeframes from one case to others. This branching process continued until either no infections occurred in a generation, signaling the end of the outbreak, or the simulation hit 5,000 deaths, indicating a major and rapid outbreak growth.
Time Intervals
Infection-to-symptom onset, symptom onset-to-death, and symptom onset-to-recovery intervals stayed constant across all infections within each simulated outbreak, though they varied across different outbreaks. Simulated individuals were not infectious prior to showing symptoms or after recovery, but could be post-mortem.
The parameters used were based on previously published estimates from earlier Ebola outbreaks. When available, BVD-specific estimates were preferred.
Model Calibration to Assumed Number of Deaths
For the cumulative BVD death estimates as of May 24, 2026, the model relied on publicly accessible situation reports from the DRC. It was adjusted to fit three different death counts (50, 100, and 200) to incorporate the uncertainty surrounding the total BVD deaths.
A simulated outbreak was considered valid if it aligned with the projected death toll by May 24, 2026, and if the first death occurred on or before April 24, 2026. Simulations continued until 500 met these benchmarks, which were then used to determine the outbreak start date and guided projections for intervention strategies for each model calibration.
Scenario Projections for Isolation
Four intervention scenarios were evaluated for each calibration, each representing a varying degree of isolation of symptomatic infected individuals (20% [poor], 50% [moderate], 70% [high], and 95% [extremely high]). The extremely high scenario aimed to establish a lower limit for transmission.
The interventions were expected to commence on May 24, 2026. On that date, a selected percentage of symptomatic individuals began isolation, typically after a two-day delay leading to their treatment. This same percentage was applied to those who later presented symptoms, with a similar two-day delay from their onset. Those in isolation were assumed not to contribute to further transmissions; the model suggested that deceased isolated individuals were buried according to safety protocols, avoiding washing or embalming.
Each simulation documented the total number of cases and deaths from the spillover date until August 22, or 90 days post-intervention. The proportion of simulations resulting in counts below or above the median R0 was recorded, with the effective reproductive number (Re, reflecting onward infections per infected person after considering immunity and public health measures) calculated for pre- and post-intervention periods.
This branching process model was implemented using Rust and the calibration and scenario projection pipeline utilized Python. This work underwent CDC review, classified as not research, and adhered to relevant federal laws and CDC protocols.
