First estimates of the global numbers of deaths associated with the 2009 pandemicArchived

Dawood et al Estimated global mortality associated with the first 12 months of 2009 pandemic influenza A(H1N1) virus circulation: a modelling study Lancet Infectious Diseases 2012

It is a truism that the number of individual deaths confirmed and reported as due to seasonal or pandemic influenza will represents only a proportion of the actual number of premature deaths that follow as a consequence of the virus infection.(1) For a variety of reasons such proportions are highly variable with population, place and time.(2)  Equally though much progress has been made in recent years there are no mechanisms at all for detecting and reporting influenza infections or associated disease and deaths for the majority of poorer populations in the world.  For all these reasons it should be obvious that the reported laboratory confirmed influenza deaths (globally 18,500 deaths and 2900 reported on national web-sites in the EU/EFTA countries), would be only a small proportion of the actual pandemic deaths, that those proportions would vary from place to place and that reports would simply be missing for many poorer populations. Indeed it is important to recall that this is the first influenza pandemic where reports of individual deaths have been requested.(3)  The more traditional approach is to use various estimating or modelling approaches.(2,4)

The first estimate of global pandemic deaths has now been published by Lancet Infectious Diseases and worked through for all individual counties in nine appendices by the US Centers for Disease Control and Prevention.(5)  The method is important to understand. The starting point is observed age-stratified population rates of symptomatic influenza from as many countries as data were available (17 sites in 12 countries). These were multiplied by observed case fatality ratios for symptomatic infections, using data from five high income countries. Since risk of death from lower respiratory infections varies by age group and socio-economic strata case fatality rates from better resourced country will underestimate the expected mortality in poorer countries.

To overcome this, the group devised empirical respiratory mortality multipliers drawing on WHO published data from estimated mortality from lower respiratory tract infections. This was done for three groups of countries: all African countries, non-African countries with high adult and child mortality and all other countries (multiplier = 1).  Separate estimates were made for three age-groups (0-17 years, 18-64 years and 65 years and over). Since influenza causes deaths in adults from cardiovascular causes as well as through respiratory disease (6) further multipliers were applied for the two older age-groups based on observations in the pandemic in five countries in the Americas. This led to estimated population death rates from respiratory and cardiovascular causes. These death rates were then applied to published estimated population estimates to give death totals by country.(5-appendix 9)  Finally years of life lost were estimated by applying current life expectancies. Ranges were derived by use of a Monte-Carlo simulation model by stating the 25th and 75th percentile for the estimates and it is these that are given in individual country tables (Appendix 9). As a method of validation a systematic literature review found 12 studies from 8 countries where estimates of pandemic deaths were published using a range of techniques including looking for excess observed to expected ratios, exhaustive case finding and modelling.

The consortium estimated that globally there were around 201,200 respiratory deaths (range 105,700—395,600) plus an additional 83,300 cardiovascular deaths (46 000—179,900) associated with 2009 pandemic. They expressed that as around fifteen times higher than reported laboratory-confirmed deaths. The results are presented by WHO Region in the paper and by country in the Appendices. Given the strong origins of European data and mortality estimates the estimated range of for EU/EFTA countries appear below (Table) suggesting that the 2900 reported deaths in Europe might represent between 7% and 13% of the true number of premature deaths in Europe.

ECDC Comment, 3 July 2012:

This is an important paper since so many policy makers were confused by the reported pandemic deaths and believed that these represented the actual number of premature deaths due to the pandemic.

How credible are the results  It is important to see this paper as the first word, and certainly not the last estimate of the 2009 pandemic deaths and the authors are both modest and self-critical.  Another estimate using a different methodology is coming from an international group with a WHO oversight committee (13). Cecile Viboud and one of leaders of the second enterprise provides a useful commentary on the CDC paper.(7)

The difficulty for work like this is that it has to cope with so many limitations and gaps in the data and it’s to the credit of the authors, and the journal, that they worked though these handicaps drawing on data from studies that sometimes CDC itself had sponsored and designed with national authorities.  If CDC had not undertaken this work there would have been no estimated rates of symptomatic influenza for Africa and India. Even so the data points are few and so some of the statements, like that 60% of the pandemic deaths were in Africa and SE Asia have to be especially speculative.   

The use of WHO political regions for presenting results like these is unfortunate as the titles confuse, especially for Asia and Europe. “South East Asia or SEARO” is actually the Indian sub-continent minus Pakistan plus only 3 populous SE Asian countries, “the Western Pacific (WPRO)” is dominated by China but has also six South East Asian countries in it while “Europe (EURO)” is all countries between the North Atlantic and Bering Straits above a certain latitude.

The 2009 pandemic brought special difficulties because so many of the infections were mild and this probably contributes to large variation in the observed rates of symptomatic influenza (for example from 0% to 22% in the 18 to 64 year age group). An alternative for the next pandemic will be to have seroepidemiology as a starting point.(8) The years of potential life lost (YLL) figures are especially important in emphasising that as this was a pandemic particularly affecting young people the social impact per death was greater than  with seasonal influenza affecting older people. This point has already been made by studies in the Netherlands using reported deaths and disability adjusted life years (DALYs).(9) However the estimates suffer from the usual problem for influenza that those with underlying conditions would have had less than the usual life expectancy.   

Finally these results, by indicating that reported influenza deaths represent only a small proportion of true seasonal or pandemic deaths question the value of following the actual numbers of reported deaths too closely in a pandemic.  ECDC has long argued that it is trends in and the characteristics of the reported severe cases and deaths that are the important information in the midst of the pandemic. More so than the precise numbers.(10) These allow  better descriptions of the severity of the pandemic and crucially indicate which groups are especially at risk and so needing protection. This is one of the approaches accepted by the World Health Assembly from the Finberg Committee and is included in ECDC’s annual seasonal influenza risk assessment.(11,12)  

+As reported on national EU/EFTA web-sites and observed by ECDC.  Deaths reported from France include 1 in Guyana, 9 in New Caledonia, 7 in the French Polynesia, 7 in La Réunion, 1 in Martinique, 2 in Mayotte, 5 in Guadeloupe and 312 in mainland France