Scientific advice and evidence in emergencies

Memorandum submitted by MRC Centre for Outbreak
Analysis and Modelling, Imperial College London (SAGE 46)

Apologies for the delay in responding to your letter of 28th October. My responses to the two follow-up questions follow:

1. Data-sharing between countries during the pandemic. There is little tradition of real-time sharing of epidemiological data by public health agencies across the world. In my view this has a number of causes: (a) concern about data protection (inadvertent release of data on individuals); (b) limited resources to document data to a level which makes sharing useful – the often quoted (and somewhat justified) fear is that raw data is complex to interpret and misinterpretation might cause confusion; (c) fears of organisational or individual reputational damage should errors or limitations of data be highlighted; (d) a desire by organisations to lead on publications in the scientific and/or medical literature which might be compromised by early release of raw data. Together these factors tend to mean that raw data is only shared with trusted collaborators under strict data access agreements. This tends to preclude the more routine sharing of data at an institutional level (e.g. between HPA and CDC). I would suggest that if this is to be a priority for future crises then effort needs to be put into building inter-agency links now – not just at the level of high level agreements, but at the level of staff exchanges and other long-term confidence building measures. Where high-level agreements might be useful is in allowing high-level briefing documents (e.g. SitReps) to be shared between HPA/DH and CDC in Atlanta (and perhaps other key EU counterparts).

2. Sir Liam’s comments on modelling. It is unclear what is being referred to here. My best guess is the confusion over the ’65,000’ deaths figure which was issued as a reasonable worst case for NHS planning in July. These were often reported as predictions:

http://www.guardian.co.uk/world/2009/jul/16/swine-flu-pandemic-warning-helpline http://www.timesonline.co.uk/tol/life_and_style/health/article6716477.ece http://news.bbc.co.uk/1/hi/8154419.stm

The release of these reasonable worst case figures unfortunately coincided with my group publishing a paper on challenges in estimating severity in a pandemic. While this paper was accurately reported by the BBC (http://news.bbc.co.uk/1/hi/uk/8150952.stm ), other reports focussed on the fact that Sir Liam’s official mortality figures (29 deaths up to that week in July) were likely to be an underestimate (e.g. http://www.heraldscotland.com/deaths-from-swine-flu-could-be-higher-than-reported-1.914626 ). This certainly caused a degree of confusion and perhaps embarrassment within DH. But overall I don’t think any of this really reflected a problem with modelling – the uncertainty around eventual mortality was very large at that time. However, it did reflect the less than optimal coordination between the SAGE group (and its SPI-M subgroup) and the CMO’s office. Had the CMO attended SAGE meetings (and been represented at SPI-M) then I suspect the 65,000 figure might have been presented differently and less emphasis might have been placed by the CMO on reporting weekly cases and deaths as precise numbers. I understand that Sir Liam rejected sitting on SAGE for fear that it would compromise his ‘independence’. I have to say that I fail to understand this argument – I suspect that the academic scientists and clinicians on SAGE are perceived as rather more independent than the CMO, and the entire purpose of SAGE was to give the best possible independent scientific advice to government. As one of the key consumers of such advice, it would have been preferable that the CMO was an ex-officio member (or indeed, co-chair). While Sir Liam was briefed on SAGE deliberations and saw resulting documents, this would not have informed him of the nuances of the scientific discussions that took place – such as the issues surrounding estimation and presentation of data on pandemic severity.

Regarding the actual ‘best guess’ and ‘reasonable worst case’ estimates of severity produced at different times during the pandemic by SPI-M, I believe Dr Peter Grove at DH has compiled a chronology of these. The committee may wish to request these from Prof David Harper.

I hope this addresses the committee’s queries. If I can assist with anything else, please get in touch.

Professor Neil M Ferguson

Director

29 November 2010