It is important for healthcare workers to understand how health news is reported. Social media provides ways to understand who makes and shares health stories, the potential audience, and the stories themselves. Back in January 2017 Prof Chris Oliver and I prepared a research paper on this topic which we submitted to two international medical journals in February and March 2017. It was not accepted for publication – perhaps it was too early for this important topic.

I came across the paper again recently when working through files as I prepare to move job (February 2019). The timing of this analysis – just at the point that Trump acquired the keys to the White House, and just when Chris and I were trying to work out what social network analysis reports could tell us – makes this a potentially important piece of work, so Chris and I have decided to share the paper in a way made possible by social media – a blog. Download the full paper here.

I am coming to the end of an extensive programme of work (though really a hobby for nighttimes and weekends) looking at health and social media. Over the past two years this has resulted in a number of publications (BJSM with Chris Oliver and Andrew Murray; BMJ Heart with Sarah Hudson; Lancet Infectious Diseases; and in more simple form JPH), featured in cover stories in international journals (BMJ and BJS), and a number of articles in press and being prepared for submission (further details to follow). On reflection, this paper tells us something that is not available in any of these other articles, but which sparked ideas that we are continuing to explore. It is therefore timely to look at this unpublished study again.

Chris and I set out to identify the main influencers of health stories in the media, at a UK and global scale. This was at a point when I had not started toexplore the full data available from NodeXL extracts (see more recent blog for details and “how to” guides). It was also before I had started to try to put a single “number” on influence as described in this exploration of influence in healthcare tweeting for Helen Bevan in NHS England. We based the analysis on the outputs from NodeXL reports (which provide a top 10 of URLs included in tweets – a rich seam of information when looking at tweets about newspapers and medical journals) and then took these findings to map the top stories. We looked at the influencers as listed in the reports and shown on the maps. In my more recent work the focus has been on extracting complete data and producing summaries of conferences and health topics as quickly as possible. In contrast, for this paper Chris and I took time to reflect on our findings, the stories, and what they told us about the dissemination of health stories and the wider world.

This was at an early stage of #FakeNews. Trump had only just arrived at the White House. Carole Cadwalladr was yet to report on some of the clear negatives of social media, in her devastating exploration of Facebook and Cambridge Analytica, Brexit and Trump. Yet we could already see the alarm bells ringing from our social network analysis maps: in reports on Trump and environmental protection; and Brexit and the risk of UK-US trade deals to the NHS. The “big data” reports that we extracted using NodeXL were providing a barometer on health stories in the general and medical press, akin to the top health news pages in many medical journals. We were studying the positive side of social media while Carole was delving into the darker side. They are, however, two sides of the same coin.

You can read the full unpublished paper as a PDF. I think that it stands the test of time. We know that there are more powerful and succinct ways to summarise outputs from social network analysis, but this paper explains how we started out, and the ways we thought that the work might develop. In fact the automated updates that we predicted in the article would be much more difficult than we had first thought. As with any data it is important to check data for completeness and accuracy; all too often, problems accessing the Twitter API mean that a first attempt at data extraction is unsuccessful, giving only a few hours of data, or missing key data. I do not think now that automated reports are the way to go. But I also worry about the time that it would take to produce and interpret outputs on a regular basis. So while the approach of extracting and summarising lots of data from social media can work well, it is important not to rely on it exclusively.

We have advanced a long way from this article. Watch out for further conference summaries over coming weeks and months, introducing new ways of summarising the key influencers in healthcare conferences, and exploring ethically challenging topics such as conflicts of interest in conference tweeting.

The ScotPublicHealth blog will now shift in direction as I explore new avenues in my career. I start GP training in February 2019, hopefully continuing some of my Public Health work in parallel. It has been great to get back to clinical practice this year during a 3 month sabbatical on the hospital wards, and regular sessions in care of the elderly. I look forward to exploring individual and population health perspectives, and developing my interests in quality improvement further. I will post more on these topics when I feel that I have something useful to share.

Dr Graham Mackenzie, Consultant in Public Health, 31 October 2018