Projections for the activity peaks of the H1N1 pandemic in the fall

On June 11th the World Health Organization has officially raised the phase of pandemic alert to level 6. As of July 19th, 137,232 cases of the new H1N1 influenza strain have been officially confirmed in 142 different countries. Given the wide diffusion of the virus and the large number of cases the WHO has stopped keeping track of the number of cases. Since the end of May, however, most of the countries heavily affected by the virus relaxed on a regular surveillance system that could imply an underreporting of influenza cases by a factor 5 to 30 depending on various circumstances[1] . For instance estimates for Mexico underreporting ranges from one to almost 3 orders of magnitude[2] .

In this framework, the interest in forecast analysis and projections is shifted from the early local assessment of the epidemic evolution to long-term analysis on the activity peaks and impact of the pandemic in the future months. In order to do that one has to move to different methodological approaches able to capture elements such as the impact of seasonality on the virus transmissibility and the global spreading pattern.

In the past weeks we have performed a maximum likelihood analysis on the GLEaM model parameters fitting the data of the chronology of the H1N1 epidemic. This is done by generating one million simulations on the worldwide scale of the pandemic evolution and finding the set of parameters that best fit the actual evolution of the pandemic. This has allowed us to estimate the transmission potential of the disease and the seasonality features. We model the seasonality by rescaling the R0 value in the Northern and Southern hemisphere taking into account seasonal changes. The free parameter to estimate is the value of the scaling factor during summer months, amin, representing the degree of dependence of the observed swine flu epidemic on the seasonality effects (amin=1 if no seasonality is in place).

By using the seasonaility scaling, GLEaM provides an early assessment of the future unfolding of the epidemic in the different hemispheres. In the following, we report results of a worst-case scenario in which no effective containment measures are introduced. We predict a first wave of cases for countries in the Southern hemisphere that occurs between August and September in phase with the seasonal influenza pattern, and independently of the seasonality scaling factor. The situation is different in the Northern hemisphere where different seasonality parameters would progressively shift the peak of the epidemic activity in the winter months. We have analyzed the results from the model in the range of the seasonal scaling factor that best fit the pandemic evolution so far and we find the potential of an autumn/winter wave in the Northern hemisphere striking earlier than expected for seasonal influenza, with peak times starting in the second half of October.

In the following table we report the activity peaks in well-defined geographical regions of the world.

RegionEstimated activity peak time
North America[Sept 25 - Nov 09]
Western Europe[Oct 14 - Nov 21]
Lower South America[Jul 30 - Sep 06]
South Pacific[Jul 28 - Sep 17]
Activity peaks in well-defined geographical regions of the world.

The interval represents the 95% Confidence Interval for the date at which the epidemic peak should be reached in different geographical areas. The peak estimate for each geographical area is obtained from the epidemic profile summing up all subpopulations belonging to the region. The activity peak estimate for each single country can be noticeably different from the overall estimate of the corresponding geographical region as more populated areas may dominate the estimate for a given area. For instance Chile has a pandemic activity peak in the interval 1 July – 6 August, one month earlier than the average peak estimate for the Lower South America geographical area it belongs to.

It is also worth noting that the present analysis considers a worst-case scenario in which no effective containment measures are introduced. This is surely not the case in that pandemic plans and mitigation strategies are considered at the national and international level. The figure below shows the delay effect induced by the use of antiviral drugs for treatment with 30% case detection and drug administration. The left panel displays the peak times of the epidemic activity in the worst-case scenario (black) and in the scenario where antiviral treatment is considered (red), for a set of countries in the Northern hemisphere. The intervals correspond to the 95% confidence interval (CI) of the peak time. The right panel shows the incidence profiles for Spain and Germany in the worst-case scenario (black) and in the scenario where antiviral treatment is considered (red). An average delay of approximately 4 weeks is observed.

Delay effect induced by the use of antiviral drugs for treatment with 30% case detection and drug administration.

Dealy effect induced by the use of antiviral drugs for treatment with 30% case detection and drug administration.

The full details of this study are contained in the manuscript Seasonal transmission potential and activity peaks of the new influenza A(H1N1): a Monte Carlo likelihood analysis based on human mobility by Balcan et al. [3]
In the next days we will post the complete results of the analysis including the scenarios for possible mitigation measures including vaccination campaigns.

[1] Graske et al., BMJ 2009, 339: b2840; Wilson N, Baker MG, Euro Surveill. 2009, 14(26): pii=19255.
[2] Fraser C et al, Science 2009, 324: 1557-1561; Cruz-Pacheco G et al, Euro Surveill. 2009, 14(26):pii=19254.
[3] Balcan D et al, Seasonal transmission potential and activity peaks of the new influenza A(H1N1): a Monte Carlo likelihood analysis based on human mobility, BMC Medicine 2009, 7:45.

Other News