This page reports the forecast that the Epistorm team at Northeastern University submits each week to the InsightNet BDBV Modeling Hub. We forecast the number of laboratory-confirmed cases of Bundibugyo virus disease in the DRC, nationally and for the provinces of Ituri, North Kivu and Haut-Uélé, by week of symptom onset (Sunday to Saturday). The forecast shows the final value once all reports are in and includes prospective backfilling.
Each forecast spans eight weeks: the four weeks before the reference week, during which hub data are still being reported; the reference week itself; and the three weeks after it. We report the 100 trajectories from the calibrated GLEAM model used for the forecast. We compute the median and the 50% and 90% prediction intervals from these 100 trajectories. Hover over the chart to see the values for a week, and hover over a line to highlight a specific trajectory. Use the location buttons at the top to switch between the DRC and the provinces.
The Rt tab shows, for the same 100 trajectories, the effective reproduction number by week of symptom onset: the average number of people infected by each case at that time. Rt above 1 means that cases are increasing; below 1, that they are decreasing.
Model. GLEAM is a stochastic metapopulation model in which population basins are connected by daily commuting and air travel [1]. Within each basin, epidemic progression follows a stochastic Ebola care-pathway model with community, hospital, and funeral transmission, and three health care branches (community, health-care facility, and isolation) [2–4]. Individual transmissibility is overdispersed (dispersion k = 0.1). The isolation scenarios ramp up by 1 percentage point per day and saturate at 35, 40, 45, 50, 55, 60, 65, 70 or 75% of cases isolated. In Beni (North Kivu) the saturation level is 10, 15 or 20 points lower than elsewhere, and once it is reached the isolation rate changes from week to week by up to 10 points either way. Rounds before 14 November 2026 used four or five scenarios with fixed saturation levels between 30 and 75%.
Calibration. Model runs are first selected conditioning on early cross-border onsets in Uganda and early activity in Beni. A second Bayesian selection stage is run with an acceptance tolerance of 15% weighted MAPE with respect to the increases in confirmed cases between consecutive DRC situation reports since 31 July 2026 [5] (in the round of 31 October, 5% MAPE with respect to the cumulative increase since 31 July). Confirmed cases are modeled as a multiple of cases routed to isolation and reach the reports after a delay from symptom onset, with a mean of 5 days. The share of cases isolated is measured in each run, and at most 95% of symptomatic cases are confirmed. The isolation scenarios are pooled, so each contributes in proportion to its share of accepted runs. The hub symptom onset data are not used in the fit.
Provinces. Ituri, North Kivu and Haut-Uélé are each calibrated separately on their own provincial situation reports, with the same method, tolerance and isolation scenarios as the DRC. Each province therefore has its own accepted runs and its own isolation posterior, and the provincial forecasts are not constrained to add up to the DRC forecast.
Forecast. The 100 trajectories are drawn from the accepted runs, so each isolation scenario is represented in proportion to its share. Weekly values are confirmed cases by week of symptom onset, i.e., the final counts once all reports are in. For weeks before the submitted weeks, the shown trajectories are drawn the same way from the same runs for comparison with the data; they are not part of the submission.
Effective reproduction number. Rt is computed for each of the 100 trajectories from its daily symptomatic onsets, with the method of Cori et al. [7]. The serial interval of each trajectory comes from the model: the incubation period of the trajectory plus the time from the symptom onset of an infector to transmission, which depends on the care pathway (community, health-care facility, funeral or isolation) and on the share of cases isolated in that trajectory (mean serial interval about 16 days). Each weekly value uses the 14 days ending on the Saturday of the onset week, that is, the onset week and the one before. Rt is a model output: it is computed from the model trajectories, not estimated from the surveillance data.
Data. Hub data are DRC-confirmed cases by week of symptom onset, national and by province, from the hub's target data [6] (the two most recent releases). Later releases add the most recent weeks as reports come in.
Assumptions. Isolation stays around the saturation level of each scenario over the forecast horizon, and reporting practice does not change. Detection is assumed proportional to isolation across the DRC.
- 1Balcan D, Gonçalves B, Hu H, Ramasco JJ, Colizza V, Vespignani A. Modeling the spatial spread of infectious diseases: the GLobal Epidemic and Mobility computational model. J Comput Sci. 2010;1(3):132–145.
- 2Legrand J, Grais RF, Boelle PY, Valleron AJ, Flahault A. Understanding the dynamics of Ebola epidemics. Epidemiology and Infection. 2007;135(4):610–621.
- 3Gomes MFC, Pastore y Piontti A, Rossi L, Chao D, Longini I, Halloran ME, Vespignani A. Assessing the international spreading risk associated with the 2014 West African Ebola outbreak. PLOS Currents Outbreaks. 2014.
- 4Ajelli M, Muyembe J-J, Touré A, Diallo A, Litvinova M, Merler S, et al. Vaccination strategies for Ebola in the Democratic Republic of Congo: the WHO-Ebola modeling collaboration. International Journal of Infectious Diseases. 2025;153:107779.
- 5Institut National de Santé Publique, République Démocratique du Congo. Situation reports, Maladie à Virus Bundibugyo. Available at: https://insp.cd/category/sitrep/
- 7Cori A, Ferguson NM, Fraser C, Cauchemez S. A new framework and software to estimate time-varying reproduction numbers during epidemics. American Journal of Epidemiology. 2013;178(9):1505–1512.
- 6InsightNet BDBV Modeling Hub, target data. Available at: https://github.com/InsightNet-US/BDBV-Modeling-Hub