Dear Iain,

Since we wrote our letter to Dr Rasmussen about the ONS COVID-19 Infection Survey ad hoc teacher analysis, we have done further work on the issues she raised.

We recognise the challenge in clearly communicating technical information to users, particularly in a context where you are striving to turn around important analysis very quickly. However, it is important to clearly explain findings such as ‘no evidence of a difference’ to avoid misinterpretation by others.

We consider that after concerns were raised about the initial publication of the ad hoc teacher analysis, ONS should have elaborated on the strengths and limitations of the analysis, what conclusions can be drawn and what these conclusions mean – for example, in relation to the impact of the rate of prevalence on the findings and the ability of the survey to determine differences between groups.

Given the broad range of users and occupations with an interest in data from the Infection Survey, we think it is fair that you are now prioritising a more comprehensive analysis of occupational risk. We have outlined below our advice regarding the development and communication of this and similar analysis in future:

• As discussed with the CIS analysis team, we encourage ONS to be open about its plans for upcoming ad hoc analysis topics. Sharing these plans as quickly and as widely as possible would provide ONS with an opportunity to engage and collaborate with expert users to improve the final analysis.
• We reiterate our previous view that ONS should be clear with users about the ability of the survey to detect differences between groups, particularly for analyses with small numbers of positive cases. ONS must clearly communicate its findings, including appropriate commentary and caveats.
• ONS should explain its decisions on methods choices to users. For example, in relation to the new occupational risk analysis, what are the criteria for an occupation being classed as high risk? And what is the reason for a two-stage approach to the analysis rather than analysing the risk of infection and death for all occupations?
• ONS should offer as much insight as possible into how interventions that were introduced for some workers at different stages of the pandemic impact the findings. It should also ensure that findings from this analysis are presented in the context of previous findings about differences in infection rates between occupations.

I would like to thank you and your team for your ongoing engagement with us as we carry out our review of the survey. We are happy to provide further support and advice as you continue to develop these statistics.

Yours sincerely

Ed Humpherson

Director General for Regulation