What is the death rate from HINI flu likely to be? The Government has forecast that by the end of August 2009, 100,000 people a day will be catching the disease, so it is vital to know as accurately as we can how many are likely to die of it.

But it’s no good asking the Chief Medical Officer, Sir Liam Donaldson. Radio 4’s Today programme tried doing that this morning, and was fobbed off with a series of claims about how difficult it was to make such a forecast, and how wide was the area of uncertainty.
 
Sir Liam was responding to a paper in the BMJ this week by a team of statisticians and infectious disease epidemiologists from Imperial College who estimate a death rate of one in 200 – but with wide uncertainty. But rather than mocking the imprecision of the estimate and its “huge degree of uncertainty”, Sir Liam would have been better employed remedying the deficiencies in the UK data that are the basis of the uncertainty.
 
Why, for example, did the capable Imperial College team have to rely on data from the US Centers for Disease Control, rather than the UK’s Health Protection Agency?
 
HINI is an acute infectious disease in which, depending upon the country concerned, the number of cases doubles roughly every three to seven days. But UK data on H1N1 are mired in delay because scientists – including myself- cannot publish analyses of them without having given the HPA 30 day’s notice in which to comment, before the analysis can even be submitted to a journal.
 
Having stopped virological confirmation of new H1N1 cases, the UK should have had in place a reporting system so that daily reporting of new (and cumulative) H1N1-related admissions to hospital took over. Where are these data, Sir Liam?
 
In addition, we need to know the outcome of these hospitalizations. The recent death of Dr Michael Day illustrates how crucial it is to secure a high post-mortem rate for H1N1-related deaths. The Imperial College team set out how data on hospitalisations need to be analysed to estimate death-rates in an unbiased way.
 
The only way to get a grip on what is happening with H1N1 is to use statistical science properly, not to mock its uncertainties in radio interviews. If statistical science does not underpin the Government’s estimate of 100,000 H1N1 cases per day by the end of August 2009, what does? This figure implies that 4 per cent of the UK population will have had clinically manifest H1N1.
 
How do we check if this is right before a massive UK vaccination campaign begins in the autumn? We use statistical science to design how many residual blood samples, and from whom, need to be tested to estimate, with adequate precision and demographically, the UK population’s immunity from acquired H1N1-infection – whether clinically manifest or not.
 
If statistical science does not underlie the UK’s planned locations for 500 Tamiflu collection points, what does? The plan envisages one collection point per 100,000 population - or only five for a city such as Edinburgh, which strikes me as incredibly few.
 
Statistical science tells us not only which data to collect, when, where and how; but how to validate them; how to analyse; and determines reporting standards which should include measure of uncertainty always.
 
Until Sir Liam grasps the thorny issue of England’s H1N1 data deficiencies, and supports academic freedom to publish, commentators seeking robust data on H1N1-related hospitalisations should look to the website of Health Protection Scotland.
 
(Conflict of interest: Sheila Bird serves on UK’s Scientific Pandemic Influenza Advisory Committee, which has not met since swine flu reached the UK.)