Epidemiologic framework for assessing the severity of influenza epidemics and pandemics

Reed C, Biggerstaff M, Finelli L, Koonin LM, Beauvais D, Uzicanin A, et al. Novel framework for assessing epidemiologic effect of influenza epidemics and pandemics. Emerg Infect Dis [Internet]. 2013 Jan

Responses to pandemics need to be flexible and targeted

One of the most important lessons of the 2009 A(H1N1) pandemic was that the planned responses were judged to be inflexible in the face of a threat that was less than might have been expected with an A(H5N1) pandemic. There is a need to be able to modify responses according to the characteristics of the epidemics and pandemics. Devising ways of achieving this flexibility was one of the key recommendations of a number of the reviews of national and international responses, including that of WHO itself.(1) A natural consequence of this is that we need methods to rapidly assess the characteristics of any epidemic or pandemic, including that elusive concept “severity”.

Framework to assess the impact of a pandemic

In this important paper with a technical annex, Reed et al of the US Centers for Disease Control and Prevention (CDC) have created a two dimensional framework for assessing the impact of an emerging pandemic .(2)

The two dimensions chosen are the numbers of persons infected, more specifically the proportion of the population infected - the cumulative incidence, and the clinical severity of illness on average - the case fatality ratio.

As part of their work, the authors reviewed United States data from past influenza seasons and pandemics to characterise the transmissibility and severity and hence propose a formal assessment framework for assessing these two potential aspects of any novel influenza virus.The authors assessment methods are separated into two time periods. Early in a pandemic, when there is more uncertainty and estimates of incidence and case severity ratios tend to be high, the measurement of transmissibility and severity is particularly uncertain. In this period, a broad dichotomous scale is used for dividing up ranges of historic values into four categories, see Figure 2 from the paper. As more data become available, it is possible to categorise the values more precisely into a continuous scale, see Figure 3. A more refined framework is applied, which allows for more sophisticated analyses - for example, estimating impact by broad age band, see Figure 4.

It is proposed that by organising and prioritising data collection, this approach will potentially inform more evidence-based assessments and so guide decision making on objective lines.

ECDC Comment (8 January 2013):

The model is a step forward from a previous single-dimensional approach

The CDC team devised a standard method for describing the cumulative incidence and case-fatality rate of influenza epidemics, so that they can plot out epidemics and pandemics on two dimensional figures (Figure 2 and Figure 3).(2)

This is an important tool and a step forward from the previous single-dimensional quantitative approach, the so-called Hurricane Scale which relied on case fatality ratios alone. Its main limitation was the length of time: before the five point scale could be applied, any pandemic would be long past.(3)

Possible problems of the model:

TimelinessThe paper poses a series of issues and problems, which the authors acknowledge. The method proposed cannot be undertaken lightly as it relies on substantial surveillance systems operating in a timely manner with considerable standardisation and centralised analytic resources. In particular, the mechanism may be slower than most modern pandemics. The 2009 pandemic which started in April was slow to progress in the Northern Hemisphere with the summer and its school holidays slowing it down.(4) This allowed for some estimates of case fatality ratio to be made, but cumulative incidence was far less clear, partially because seroepidemiology was not planned for and slow to deliver.(5) A pandemic starting at any other time in the year might have much swifter impact and Europe could experience more what more of the uncertainty and sometimes over-estimates and over-reaction that Mexico, Australasia and South America had in 2009.

Proposed categoriesThe four categories in Figure 2 are to a certain extent arbitrary. They are based on only seven epidemics and pandemics in the USA, and perhaps they will be refined with experience. At present there are no cases in category C; furthermore,it is difficult to work out if the 1 957 pandemic is in Category B or D (compare Figure 2 and Figure 3). Also, influenza is always heterogeneous in its impact and it is not clear how the model would work in case the sites where the initial data are gathered happen to be in areas where the pandemic affects later. Is the model intended to describe where the pandemic is worst or an average And what would happen if a second or third wave was worse than the first wave,as in some European countries in 2009(6) These are not insuperable problems, but they need to be acknowledged.

Operational decisions during pandemicsFinally, we need to consider the sheer complexity of the operational decisions that are needed in a pandemic and must determine its severity. For example, in the 2009 pandemic in the US at least pregnant women were more affected than expected, and this needed to be reflected in the response. That cannot be captured in a simple model like this. What would happen, for example, if there was sudden inherent antiviral resistance or if an epidemic arrived at a holiday time when primary care is operating on out of hours capacity for some days A two dimensional approach is better but still limited. This leads to the final so-called so-what element: how the severity read across to actions. Looking at Figure 2 & Figure 3, what actions would you take if you were in Category B - close schools, operate hospital surge capacity surge, cancel public meetings, … It would be unfair to criticise Reed et al because they do not solve this issue. They do note that another team in CDC is considering response according to severity.

Application of the model in Europe

It will be very helpful if some countries in the EU consider and adapt the CDC model, preferably in several countries and in a standardised way, so that it is more likely the method can be applied wherever epidemics or a pandemic first appears. The CDC method is complimentary to the ECDC Risk Assessment approach, which is less quantitative but captures more details and operational experience, on the lines of what Fineberg recommended; it is complementary, not an alternative to the CDC approach. Both methods will be considered at the coming 3rd joint influenza meeting of the WHO Regional Office and ECDC.