Projections for new H1N1 Flu spread by computational modeling
We use GLEaM, our GLobal Epidemic and Mobility modeler, to generate future projections on the global spreading of the unfolding Mexican flu epidemic. This first report presents some initial results of this ongoing effort. In the following days we will provide updated risks maps and future projections.
GLEaM is a computational modeler that integrates sociodemographic and population mobility data in spatially structured stochastic disease models to simulate the spread of epidemics at the worldwide scale. Read more about GLEaM…
Initialization and calibration of the model
We assume the epidemic originated in Mexico. We feed a first set of simulations with a cluster of infections on March 18th. We then study the predicted infected areas as of the date of April the 26th. The disease is assumed to be an ILI like disease (with both symptomatic and asymptomatic). The reproductive number is tuned according to a high transmission scenario with more than 1000 infected individuals in Mexico and a low transmission scenario in which the number of cases is halved (assuming that most of the cases investigated at the moment will result in negative to H1N1 lab test). For the sake of clarity we only provide the results for the high transmission scenario in this report. We here provide the data aggregated at the level of single countries. Finer resolution (major urban areas, states) can be provided upon request.
Model prediction as of April the 26th
The calibrated model provides the forecasts of the detection of infected cases in Mexico, USA, Canada, UK and Spain. In these countries, cases have been confirmed (as shown in red in the following figure).
The model also provide the likelihood of the occurrence of cases in the following countries:
|Country||Outbreak risk (%)||Country||Outbreak risk (%)|
In some of the listed countries there have been reports of infected individuals but they are not yet confirmed as H1N1 cases.
Risk maps three weeks ahead
This report provides risk maps for the following dates: May 3; May 10; and May 17.
The risk is quantified by the likelihood of a case detection in a given zone according to our stochastic model.
We are not pushing any anticipation beyond three weeks as the containment measures that will be put in place have to be considered (we plan to pull out new scenarios every 24/48 hours).
Worst case scenario
In the following figures we report the risk maps according to the model evolution in the worst case scenario, where no use of antivirals or other intervention measures is considered. It is important to stress once again that the evolution of the model is calibrated with the early evolution of the epidemics. Warnings and containment measures are likely to be effective and the model will have to be calibrated once more in the next days.Countries colored in black have already reported confirmed cases. We report in the map only countries with more than 1% probability of having infected individuals as of the corresponding date.
The number of expected cases and relative confidence interval in each country can be provided upon request.
While we think that this information might be useful to get possible scenarios for the spreading of the disease, we want to make clear a certain number of disclaimers which help framing the presented results in the right context.
- Data coming from different regions of the world have to be validated in most of the cases. We feed the model with what appears plausible and confirmed at the moment but the scenarios are subject to changes depending on the available data.
- We consider both a worse case and a best base scenario. Both are based on a certain number of assumptions on the data and what we find from the information gathered so far.
- Containment measures as decided by health authorities will be integrated in futures updates.
- The model is mostly used to provide first time arrival of infected individuals and provides risk maps telling the probability of infected case detection at a given date. At this early stage of the disease, the number of predicted cases is a number with large fluctuations and as such the most sensitive outcome of the model.
- The results are based on computational models and have to be considered only as an extra source of information and not as the reality of the epidemic unfolding.