Communication: engaging a wide range of audiences and minimising misuse

What we said in October 2021:

When data and statistics are clearly presented, they are valued by the public. Statistics producers should apply the lessons they have learned about how to improve public communication to other statistics.

What we found during October 2021-October 2022:

Meeting the needs of a varied range of users using different mediums

The final stage in ensuring that health and social care statistics serve the public good is good communication. We found that, as a result of the increased public appetite for data and statistics about COVID-19, producers have a greater appreciation for the need to communicate with, and meet the needs of, a range of different users. This includes members of the public who are accessing statistics for general interest, to make decisions about their lives or to enhance their understanding of an issue of high public interest. This was often a big change for producers, many of whom were used to communicating with other audiences (usually other parts of government or experts, such as academics). As a result of developing this understanding of the varied users of statistics, some producers have identified that one product does not necessarily suit all. Three themes emerged from our evidence, focussed on three different mediums for communicating statistics: dashboards, reports, and social media, specifically Twitter.

Dashboards

It is clear that dashboards were a hugely popular way of communicating statistics to the public during the pandemic. We have already discussed the astonishing number of hits that the UK COVID-19 dashboard experienced during peak times. There were also extraordinary levels of engagement with dashboards produced in other UK nations – Public Health Scotland’s COVID-19 dashboard received over 120,000 daily views at its peak in September 2021 and the most up to date figures available from the Northern Ireland Department of Health indicate that its COVID-19 dashboard had experienced over 1.5 million unique viewers worldwide as of May 2021.

We heard from both users and producers about the benefits of dashboards. For example:

• Dashboards can be more engaging for many users, who like to interact with the statistics and see them presented in a visual way (as opposed to tabular or textual forms)
• Dashboards can support the release of underlying data, letting others access them and add value through their own analysis
• While upfront resource is required for developing dashboards, and resource will be needed to maintain and improve them, many producers have found that using dashboards results in more efficient processes, allowing them to use analytical resource for other priorities

In light of the success of COVID-19 dashboards, lots of producers are planning to introduce dashboards to other statistics. For example, Public Health Scotland is consulting with its users, proposing a general move for its statistics away from reports towards dashboards and open data. The Department of Health in Northern Ireland recently published a new dashboard on cancer waiting times in place of its traditional pdf report. The Department has been working with its users to gather and respond to feedback on this new dashboard, and is planning to take a similar approach for other statistics in future.

Despite the benefits described above, we caution producers against automatically introducing dashboards or replacing existing statistics with dashboards without carefully thinking about the value they will add and their design. A dashboard in itself is not necessarily the answer to good communication of statistics; it needs to be well-designed and the reason for choosing a dashboard over other products should be clear and guided by user needs. Some attributes of the UK COVID-19 dashboard which we consider made it so successful are that:

• it responded to user needs as they changed throughout the pandemic – introducing new topics and measures over time, and engaging with users to obtain and respond to feedback
• it allows users to download the underlying data (we heard that it is particularly frustrating for some users if this option is not available on a dashboard)
• it includes a simple summary describing the latest headline figures
• it provides a wealth of information about the data to support appropriate use

The multidisciplinary team which developed and now maintain the UK COVID-19 dashboard was instrumental in this success. This included statisticians and other experts on the data and its interpretation, as well as expertise outside the analytical community such as web developers and user researchers.

There are particular challenges with dashboards, which we are encouraged to find that producers are aware of and considering how to address. Accessibility is a key challenge when producing dashboards. In the UK, all content on public sector websites must legally meet accessibility criterion. This means that dashboards must be perceivable, operable, understandable, and robust, but dashboards built using popular ‘off-the-shelf’ software packages such as Power BI and Tableau can be difficult to make accessible. Even the UK COVID-19 dashboard is not fully accessible – though it is good to see an accessibility statement published which provides transparency on what steps the UK Health Security Agency (UKHSA) has taken to make the dashboard as accessible as possible and which areas are not fully accessible.

It can also be challenging to present sufficient insight, context and guidance on appropriate use alongside statistics in a dashboard. It is not yet clear how best to achieve this, and there is unlikely to be just one right answer, so sharing learning and good practice in this area will be important. The recently updated dashboard guidance from the Analysis Function Central Team is a helpful resource for producers, providing guidance on how to decide if a dashboard is the right tool for users. It also includes advice on important design elements to think about and information on accessibility. We would like to see the Analysis Function Team supplement this guidance with user research on what makes an effective dashboard, practical support on how to build one and guidance on user testing.

Statistical reports

For some users we spoke to, reports are still an important part of statistical publications. This is because they provide the opportunity for producers to include insightful commentary and background information about the statistics. Users told us that some statistical reports can be quite technical and therefore it is helpful when a summary with key points is included. As with dashboards, some users want to access the data as well as a report and can find themselves having to manually extract this information. Therefore, producers should always consider how to provide the underlying data, whether as a data table or open data, in addition to a report.

Social media (Twitter)

As discussed in our 2022 State of the Statistical System report, there has been an increase in the use of social media, in particular Twitter, to communicate statistics over the last few years. Most producers have organisational accounts which they use to share their latest release of statistics. We have seen some innovative uses of organisational accounts to engage with users of health and social care statistics. For example, during the pandemic Public Health Scotland used its Twitter account to share videos of statisticians talking about their statistics – such as this video of the Head of Profession for Statistics discussing changes to COVID-19 statistics or this video about the effect of COVID-19 on pregnancy. Public Health Scotland found that these video tweets gained a high number of impressions (the number of times users saw the tweet) and a high engagement rate (including clicks, replies, retweets, likes and follows).

A particular phenomenon during the pandemic was the increase of “tweeting statisticians” – that is, government and non-government statisticians using personal Twitter accounts to communicate their statistics. The large followings that many of these people gained during the pandemic demonstrates that there is a public appetite for access to a reliable, expert voice on statistics (for example, during the pandemic Clare Griffiths, UK COVID-19 dashboard lead, and Meaghan Kall, an epidemiologist at the UKHSA, experienced ten- and 120-fold increases in followers respectively). Users also told us that they like this way of engaging with statisticians. Tweeting statisticians themselves told us about the many benefits they have experienced including:

• Raising awareness about and publicising statistics to a broader audience, particularly members of the public
• Summarising and explaining statistics in a succinct way
• Answering user questions and debunking common misconceptions about statistics
• Learning more about users – who they are and what they are interested in
• Personal development through following other accounts, and interacting with colleagues and other experts
• Humanising the work of statisticians – showing that there are real people behind the outputs

However, running a personal account on Twitter can be challenging. At times, Twitter can be a difficult environment and can lead to trolling or personal attacks. It can also be hard to explain complex information within short word limits and to mitigate the risk that tweets are taken out of context. Running a successful Twitter account takes time and effort and we heard that this is not always considered part of a statisticians’ or analysts’ role. There is also a burden on communications teams, who monitor tweets and may have to respond to media queries that arise from Twitter activity.

While organisations do have social media policies, we heard that these tend to focus on what not to do. There is therefore a gap in guidance about what to do in order to realise the many benefits listed above. Producers would appreciate more support, for example on handling challenging questions and comments. We would like to see guidance on running a personal Twitter account and examples of good practice provided for producers by the Analysis Function Central Team to support this engaging and effective way of communicating with the public about statistics.

Minimising the misuse of statistics

A risk to health and social care statistics serving the public good is misuse, to accidently or deliberately misinterpret what statistics means. In early 2022, we explored this issue, prompted by the misuse of UKHSA statistics on COVID-19 infection rates by vaccine status. The UKHSA published the statistics with good intentions of supporting transparency, but the statistics were misused – for example, the figures were used to support anti-vaccine misinformation in podcasts and online. Misuse also occurred based on similar analysis published by Public Health Scotland, such as misuse in a US Senate committee hearing to claim that vaccines were causing mass infections.

We consider that there are three aspects to misinformation:

• The actor: the desire by some to deliberately misinterpret statistics to support a particular agenda or narrative
• The data: if taken out of context, data can be mobilised in a way that misleads
• The statistics producer: who has responsibility for decisions about how the data are analysed and presented, providing warnings against misinterpretation, and responding quickly to address issues when it is clear that data are being misused

Much of the focus when talking about misinformation is on the first of these – the actor. Yet misinformation is not always based on complete falsehood. Sometimes people have taken small elements of data out of context and placed undue weight on them. In other cases, there may be inadvertent misinformation, where data are misunderstood.

In the case of statistics on COVID-19 infection rates by vaccine status, we found that users had to work quite hard to find official figures and that understanding them was not easy for all users due to the technical nature of the reports they were published in. We welcomed changes which both UKHSA and Public Health Scotland made to the presentation of their analyses and accompanying blogs published to guide appropriate interpretation (blogs by UKHSA and by Public Health Scotland). We also supported producers’ decisions, in light of changes to COVID-19 testing policies, to stop weekly reporting of these statistics due to uncertainties in the data resulting from policy changes and the potential for misinterpretation, particularly relating to vaccine effectiveness.

We recognise that combatting misuse is challenging and that it will not always be possible for producers to completely eliminate misuse. However, producers need to minimise the risk of misinterpretation and misuse by making it easy for people to find and understand objective, clear information. This can be achieved if producers are aware of the broader context they are working in – for example, in the case of the vaccination analysis described above, there was a heightened public interest in data and analysis during the pandemic. Producers should be agile in responding to public debates and consider the risks associated with the publication of their statistics in advance – depending on the level of risk, producers may need to take different approaches to how statistics are presented, for example it may be necessary to ‘pre-bunk’ anticipated misuse or misinterpretations. Producers should also ensure that any caveats are explained prominently alongside their analysis and consider the use of a blog, or similar explainer, to communicate technical content in a way which is easily understandable for a general user. Finally, it may be necessary for producers to collaborate across professional communities to decide on the best approach – for example, across the analytical, scientific and communications professions.