The day after the President of the Royal Statistical Society wrote to the Chief Medical Officer, Sir Liam Donaldson, about the proper collection of key H1N1 data and statistical reporting standards, the Health Protection Agency (HPA) posted some limited pearls of wisdom.

A brief explanation was offered on how HPA had arrived at last week’s headline figure of 110,000 new H1N1 cases in week 30: from Tuesday 21 to Monday 27 July 2009.
 
In particular, there is a weekly collection-scheme (how representative?) for swabs from patients who consulted their general practitioner about an influenza-like-illness (ILI) which are tested for H1N1. Positive rates for the current week are ‘estimated’ by age-group (7) and area (4) using a logistic model. "Estimation" may be code for dearth of data. One would assume at least 25 swabs per age-group per area (or 700 across England) so that, across areas, the measure of uncertainty which qualifies the percentage positive is at most 5 per cent. HPA did not reveal how many swabs from week 30 informed its headlines.
 
Week 30 was tricky for another reason. The start-up of the National Pandemic Flu Service (NPFS) on Thursday 23 July caused a major hiatus so that, in analysis terms, you would think that week 30 needed to be subdivided into: pre-start-up Tuesday+Wednesday; start-up Thursday; post-start-up Friday+Monday; and post-start-up Saturday+Sunday.
 
Like-with-like (for each age-group, say) “Tuesday+Wednesday” comparisons could then at least be made across weeks 27, 28, 29 and 30; with the full weekday versus weekend impact of the National Pandemic Flu Service assessable by how week 30 diverged from the corresponding “Friday+Monday” and “weekend” trends. But that’s NOT what HPA did.
 
Because the National Pandemic Flu Service does not cope with children under 1 year of age but applies for 15-64 year olds, then all of the difference in proportionate change between GP consultations in weeks 29 and 30 for the under 1s versus for 15-64 year olds was ascribed by HPA to the start-up of NPFS.
 
What should you look out for in week 31?
 
First, new data on how the National Pandemic Flu Service (NPFS) is operating. In particular, by age-group of patient (not caller): (i) the number of calls daily, (ii) the number for whom GP/other medical advice was recommended, (iii) the number for whom H1N1 was diagnosed, (iv) the number for whom antiviral medication was authorized, and (v) the number whose antiviral medication was collected within 24 hours of its authorisation.
 
Second, new data (by age-group and pre-specified sampling method) on H1N1 positive rates among patients given an NPFS authorization. In particular, by age-group of patient: (vi) the number of NPFS-authorised patients asked to provide a swab to be tested for H1N1, (vii) number of requested swabs which were obtained (compliance-rate), and (viii) the number of received swabs which tested H1N1-positive.
 
Sir Liam has pointed out that it is easier to collect cross-sectional data on the numbers in-hospital for H1N1 than it is to track H1N1-related admissions and discharges.
 
Thirdly, the same cross-sectional data collection can - just as easily - be used to ask hospitals about:  ix) their previous day’s new H1N1-related admissions and x) their previous day’s H1N1-related discharges.
 
Fourth, H1N1 pressure on intensive care unit (ICU) beds was already evident last week. England has only 3,000 to 4,000 ICU beds and, last week, at least one in 50 ICU beds was occupied by H1N1 patients (then 81) whereas - of some 160,000 hospital beds - fewer than 800 (that is: at most 1 in 200) were last week taken up by H1N1 patients.
 
I have raised questions about England’s unnecessary secrecy over precisely how few/many swabs  - taken in the current week from patients who consult their GP about influenza-like-illnesses (ILIs)  - are tested and reported in time to contribute to this week’s headlines on new H1N1 cases. Such secrecy is to be avoided when it comes to the numbers of H1N1 positives among swabs taken from patients who receive an NPFS-authorization.
 
And so, fifth, because England’s headline figures on new H1N1 cases have meagre underpinning – both virological and statistical – age-specific monitoring of more extreme H1N1 events is crucial  (incident hospital admissions, in-hospital cases, incident ICU admissions, in-ICU cases, and durations of both hospitalisations and ICU-stays; H1N1-related deaths; H1N1-related maternities).