Journalism should take a zero-tolerance approach to publishing false or unverifiable claims
I often ask friends – when you read an article in mainstream media about a topic in which you have some expertise (about health if you are a doctor for example), how often do you notice that it contains incorrect information? The majority of answers fall at the frequently/very frequently end of the Likert scale we’d be looking at if I were polling rather than chatting. I’m not talking here about serpent-headed aliens, microchip-containing vaccines or stolen elections. But I am talking about mundane examples of misrepresentation through partial presentation of the facts and fabrication.
I give illustrations from the Guardian newspaper, not because it’s a major culprit but because it isn’t. If the problem is present even in the best, it’s present everywhere. I am a long-time reader of the Guardian and subscriber to its online edition. I value its balanced coverage and regard it as standing head and shoulders above all other daily newspapers in the UK for its reliability and lack of bias. But at times I am left wondering, even in this newspaper, about a particular piece – is this true? How would I know?
On the surface the examples I will give may seem like minor infringements, but unreliable reporting in any part of the paper can lead to lack of trust in the reporting of every part of the paper; and we are storing up trouble for the future if journalists following examples such as these come to believe that writing a good story takes precedence over writing an entirely accurate one. There is a fairly simple solution to the problem but before considering it, a few examples.
An article in January 2023 described a survey which was said to have “…found that one in five LGBTQ+ people and more than a third of trans people in the UK have been subjected to attempted conversion…”. As part of an online survey, respondents were asked whether they had ever experienced someone taking any action (my italics) to try to change, cure or suppress their sexual orientation or gender identity. Describing the findings, the phrase “subjected to” appeared in the article headline, in the final sentence and three times in the text. There was no link to the survey report but when I found one it revealed that the campaigning group commissioning the survey has a particular take on what “subjected to” means.
“There must be no “consent” loophole… Conversion practices are abuse and it is not possible to consent to abuse …The definition of conversion practices should include religious practices…”. So examples of what respondents were “subjected to” included “I saw a counsellor…” and “My partner ended our relationship because of God and then the people from church prayed for us to become straight.” For sure, there were quotes about much more unpleasant experiences but even there the reframing was unusual: being beaten up because you’re gay is wrong, but it’s a stretch to call it a conversion practice. There was no indication of a typology of practices or the prevalence of various practices – anything and everything goes towards the headline figure. This strikes me as a long way from what most people understand by the sort of conversion therapy that might be banned by legislation, but you wouldn’t know it from the way the survey was reported.
An article in June last year headed “Brain damage claim leads to new row over electroshock therapy” reported that electro-convulsive therapy (ECT) “…is now the focus of a huge row – which erupted last week – over claims that it can trigger brain damage, that guidelines covering its use are weak and that it is used disproportionately on women and the elderly.” Again there was no reported evidence of a huge row; just a link to a 5 year old Guardian article retailing the same criticisms from the same source as described in the 2022 article. The bust-up seems to have been imagined into life to act as a hook for the otherwise non-story.
Something from the pandemic. An article from January 2021 reported that the “Prince’s Trust happiness and confidence survey produces worst findings in its history”. Three accompanying comments linked the findings to the impact of the pandemic. The findings as reported were literally true (just) but a reading of the whole report gives quite a different picture. In 2021 just 56% of respondents said they were happy about, and 64% said they were confident about, their emotional health. Certainly the lowest on record but the corresponding figures for 2018 were 57% and 65%. In 2021 56% said there were always or often anxious. Again, the highest on record but the figures for the preceding years 2018-2020 were 53%, 54% and 55%. The really big changes have come since 2010 when more than 70% said they were happy and confident about their emotional health and fewer than 20% said they felt anxious or depressed all or most of the time. So a study that shows a decade-long decline in the emotional health of young people is reframed as a story about the impact of the pandemic by the simple expedient of not reporting most of its findings.
In a piece from January this year promoting assisted dying and entitled “Today, 17 people will likely die in unimaginable pain…” regular contributor Polly Toynbee writes, after a warm up about torture chambers, excruciating pain, horror and humiliation, that “On average 17 people a day die in terrible pain that can’t be relieved by even the best palliative care.” The claim is based upon a review undertaken by the Office for Health Economics which, like the research it is reviewing, refers nowhere to the severity of pain but only to “unrelieved pain” much of which, it would be clear to anybody familiar with the clinical scenarios, will not match the descriptions offered. Toynbee’s account of unimaginable pain in end-of-life care comes in fact from her own imagination.
Much of this would be avoided if journalists put a bit more work in – didn’t just recycle press releases and did some of their own fact-checking, aided by basic critical appraisal skills. How would we know if they were doing that? Online encyclopedia Wikipedia, in facing its own questioning about reliability, has developed a policy it describes as Verifiability, not truth. “Verifiability” means that material must have been published previously by a reliable source, cited by the writer and consulted by them. Sources must be appropriate, must be used carefully, and must be balanced relative to other sources.
Citing reliable sources, with a clear statement that the journalist has consulted them, gives readers the chance to check for themselves that the most appropriate authorities have been used, and used well. In fact none of the four examples I give here would be compliant with such a policy. If respectable and respected mainstream media are to maintain their reputation for trustworthiness they need to demonstrate how they manage reliability in their reporting and not just assert that they do. An explicit, and explicitly followed, verifiability policy would be a good start.
Usually I ignore the accompanying pictures when I am reading a newspaper article about health. Just recently however, I have noticed something about the pictures that accompany articles about Long COVID.
Here’s a typical example – a picture of somebody, and it’s nearly always a young woman, reclining on a sofa looking exhausted. Of course we don’t need a photo to help us understand the idea of being exhausted, in the way that we might want an illustration for an article about recent floods or trouble at a street demonstration. So I took to wondering what was going on.
The use of such pictures may represent no more than a desire to break up blocks of text. But there are a couple of features that drew me to thinking of how similar images have been used to convey something about other conditions.
First, here’s a rather typical image from an article about the 19th Century malady neurasthenia: a woman reclines and does not look, if truth be told, especially discomforted by her experience.
From a different era and with another typical feature (hand on forehead) I didn’t have to look far to find this illustration (left) from an article about premenstrual syndrome.
And just to counter the idea that these things are entirely gendered here’s a picture (below) with another typical feature – it’s the back of the hand on the forehead – from an article about man flu’.
Perhaps what I’m seeing is no more than a lazy use of stock images but regardless of motive it occurs to me that there’s more to be said. The visual rhetoric hints at connections that the written text may not – between Long COVID and a group of conditions that have over the years shared some degree of being dismissed, of dubious provenance and melodramatic in presentation. And by implication, faintly ridiculous.
I don’t think this representation of illness does any favours to the discussion about the nature of Long COVID – not just to the question of possible underlying pathophysiology but also to the question of whether some cases are psychosomatic (functional) in part or whole.
My solution would be twofold: time to drop the use of the term Long COVID for the trouble it causes by bundling together so many disparate conditions. And when writing about the sequelae of COVID infection it’s time to drop the use of pictures of tired people resting, hand-on-forehead, on the sofa.
How common is Long Covid? Top of the range, with estimates of nearly 40% of people who’ve had Covid still being symptomatic 12 weeks later, come the REACT-2 study and a recently-published analysis from a US database. In the basement come estimates of 0% from a follow up study of children in Melbourne who had recovered from mild or asymptomatic infection, 2.3% at 12 weeks (the Zoe survey), 4% from a study of schoolchildren in Zurich, and 5% for symptoms at 12-16 weeks from a recent ONS survey. And other estimates in between…
Why is there so much variation? There are the usual suspects – results may be influenced by differences in the samples in terms of age, gender, BMI, vaccination status and so on. Many of the symptoms reported are non-specific (such as fatigue, aches and pains, difficulties with concentration) and endorsement is likely to depend upon exactly how they are asked about and whether an attribution to Covid infection is required. Response rates varied greatly – only 13% in a substantial UCL UCL study of young people and 20% in an ONS study of school children. Some attempt to adjust is possible but response bias and confounding can’t be entirely discounted.
Maybe there’s an issue of reporting threshold – the higher estimates coming from counting relatively minor experiences while the lower estimates count only more severe symptoms. Certainly there seems to be an inverse relationship between number of symptoms and prevalence, but severity (intensity, intrusiveness) is harder to gauge from readily available reports. An ONS survey from earlier this year suggested that only about 1 in 10 respondents was limited a lot by their experiences. An intriguing finding from a study of NHS general practice records suggests that GPs are recording very few cases by comparison with those reported from research surveys – in fact about 100th of the numbers, with a quarter of GPs using NHS codes for Long Covid to describe no cases at all in the study period. The headline of the Guardian’s report of the study this article was amended after publication. An earlier version said that GPs in England were “failing to recognise thousands of long Covid cases”. The headline was changed to indicate that the research concluded that “long Covid coding in primary care is low compared with early reports of long Covid prevalence”.” An acknowledgement that at least one possible explanation is that GPs are applying a clinical-severity filter not recognized by researchers.
Another way to ask about the significance of “Long Covid” symptoms is to consider the prevalence of the same sort of symptoms in the general population. For example persistent fatigue is reported by between 10 and 20% of the general population and something like 50% of people with an identified physical disorder. Not many Long Covid studies collect data from comparison groups, but one ONS survey suggested 3.4% of non-Covid respondents had the same symptoms as the Covid+ respondents and the UCL study found a figure of 16%. In both studies the Covid respondents had more symptoms than the non-Covid respondents, but between-study differences raise again the question of why so much variation – no doubt it’s down to sampling, how the questions were asked, and response rates.
So … what is the value, and what is the likely outcome, of collecting not-very-accurate information on underspecified populations of people who have had a Covid infection and who for the most part have non-specific symptoms that are common in the general population? It may be that the result will be a better understanding of the nature of longer-term problems attributable (or attributed) to Covid infection, but it isn’t clear that these surveys are the most efficient way of getting there. And there are potential pitfalls – Long Covid manifestly isn’t a single condition and it is difficult not to see, in the current approach to researching it, the process that Ian Hacking called Making Up People – the bringing into being of a new category of people with what may turn out to be a transient illness (transient as a medical category that is, not for the individual) – in ways that are not beneficial to anybody concerned.
A start would surely be to stop using the term Long Covid. It’s unlikely we’ll convince the media or pressure groups to drop it, but a move in the right direction would be if researchers and clinicians started using more specific terminology to describe exactly what it is they are studying and why.
Since early on in the COVID pandemic there have been expressions of concern about its impact on mental health and, at least as reported in the mainstream media, those concerns have been couched in frankly melodramatic terms. In May 2020 the then president of the Royal College of Psychiatrists expressed fears1 that “lockdown is storing up problems which could then lead to a tsunami of referrals.” The (by comparison) rather more muted claim that the pandemic was having a “major impact” on mental health came in the reporting of a questionnaire survey conducted later that summer2. The current president of the Royal College of Psychiatrists, who has echoed his predecessor’s use of the tsunami analogy, attracted headlines towards the end of 2020 by suggesting that the consequences for the nation’s mental health will be the greatest since the second world war3. The doom-laden tone has persisted into this year, with The Prince’s Trust declaring4 that “The pandemic has taken a “devastating toll” on young people’s mental health” and a report from Bradford in West Yorkshire suggesting5 that “The pandemic has had a deep impact on children” who are “a lost tribe in the pandemic”.
There’s a tiresomeness to this sort of coverage, with its implication – familiar to those who regret the stigmatising of mental disorders – that there is something frightening and uncontrollable going on. But melodrama has other more tangible disadvantages.
First is the lost opportunity to conduct, and demonstrate how to conduct, a balanced discussion about the difficulties and uncertainties that attend the interpretation of data. Much of the media coverage consists of anecdotes – the personal interest stories so favoured by journalists – or small-scale interview studies. Data on health service contacts are problematic because of the degree to which disruption caused by the pandemic changes the relation between population prevalence of disorders and attendance rates. So what is, for now, the best evidence probably resides in population surveys.
However, results from self-report symptom questionnaires require more cautious interpretation than they are often given. Self-reported symptoms are mainly markers of distress, and although very high scores can be an indication of mental disorder the likelihood that is so depends upon the context. Two studies that have taken repeat measures suggest that most of the distress created by conditions during lockdown resolves quite quickly 6,7. The major impact headlined in one report2 proved due to a difference between 10% (previously) and 12% (at the time of the survey) in endorsement of a question about thoughts of suicide but did not spell out how often such thoughts are an accompaniment of distress and lead neither to suicide attempts nor to suicide – suicidal thoughts in population surveys are about 1000 times more common than suicides in the same population.
A bias in thinking about attribution arises with this use of language. A few attempts have been made to link Government policy to an increase in mental health problems among the young8, but for the most part no explicit suggestions are made. Instead the resort to analogies with war or natural disaster and talk of unprecedented crisis implicitly leads to foregrounding of the pandemic as the main explanation for mental health problems. To be sure, there are stresses in the current situation but there are also longer-term forces at play9-11. Years of government austerity strategy have done great damage not just to mental health services but to community assets, employment stability and family security – the main resources that constitute resilience for the most vulnerable in society.
Mental health is a vague umbrella term that is used to cover everything from the boredom and frustration that so many of us feel for being unable to see friends or go out socially, to severe mental illness. Not all mental health problems require treatment from the mental health services. Even before the pandemic something like a third of those identified in the Adult Psychiatric Morbidity Survey12 as having depression had not been so diagnosed by a professional. A third of adults referred to the Improving Access to Psychological Therapies (IAPT) service do not attend for the first appointment – voting with their feet 13.
Of course we need better mental health services. We also need to repair the damage done by years of austerity to schools, community resources and the quality of life of the poor and disabled. And the mental health services need to be planning about exactly where their efforts are best placed – blanket statements about mental health do not help with thinking about who needs exactly what sort of assistance, either preventive or therapeutic. These are challenging tasks the public and government support for which isn’t going to be recruited by melodrama rather than reasoned analysis and careful presentation of the facts.
Fancourt D, Steptoe A, Bu F. Trajectories of anxiety and depressive symptoms during enforced isolation due to COVID-19 in England: a longitudinal observational study. The Lancet Psychiatry. 2020 Dec 9.
Shevlin, Mark and Butter, Sarah and McBride, Orla and Murphy, Jamie and Gibson-Miller, Jilly and Hartman, Todd K. and Levita, Liat and Mason, Liam and Martinez, Anton P. and McKay, Ryan and Stocks, Thomas V.A. and Bennett, Kate and Hyland, Philip and Bentall, Richard P., Modelling Changes in Anxiety-Depression and Traumatic Stress During the First Wave of the COVID-19 Pandemic in the UK: Evidence for Population Heterogeneity in Longitudinal Change. Available at SSRN: https://ssrn.com/abstract=3749211 or http://dx.doi.org/10.2139/ssrn.3749211
Townsend, E COVID-19 policies in the UK and consequences for mental health, The Lancet Psychiatry, 2020; 7(2): 1014-1015,
The coronavirus pandemic has posed some serious challenges for scientists – inherent in the novelty of the infection itself (biology and epidemiology), in the need to explain difficult research to an apprehensive public and in the desire to inform governmental policy without unduly influencing decisions that are in the end political ones. There have been some great successes in rising to the challenge – for example in the development, evaluation and delivery of vaccines supported heavily by NIHR and the NHS and in the fields of infectious disease and genetic epidemiology.
More dismaying has been the way that differences of opinion have been handled. Elements of the press have of course behaved shamefully – what else do we expect – but it has been a real disappointment to see some members of the academic and clinical communities joining in what has been at times more like a childish shouty argument than a serious discussion about a difficult subject.
Against this background, my eye was caught by the release of a document from a group called the Health Advisory and Recovery Team (HART), COVID-19: an overview of the evidence. The trailers on their website were encouraging-
“HART is a group of highly qualified UK doctors, scientists, economists, psychologists and other academic experts…Our core aim is to find the common ground between the Government and groups that are concerned about COVID-19 restrictions. The ambition is to bring all sides together and to widen the debate in order to formulate an exit strategy that benefits everyone in society…Our research has identified a need for public policy to reflect a broader and more balanced approach across a number of key areas…We must now bridge the gaps in policy areas where there is obvious disagreement. We should try to avoid bias and follow the rapidly emerging evidence to guide us. If we listen and communicate clearly and respectfully with each other we can find the common ground”.
I restricted my reading to the area I know most about, the consequences of the pandemic and its management for the psychological wellbeing (mental health) of the population. That is to the chapters on COVID policies and harm to children (actually, including young adults), the psychological impact of the Govts communication style and restrictive measures, and the benefits and harms of wearing masks.
On a (very) superficial glance this looks like a scientific review of the evidence; lots of expert-sounding people have authored sections and there are nearly 300 references cited over 55 pages. A more detailed read gives quite a different picture. One notable section (suggesting that wearing a mask is psychologically harmful) isn’t referenced at all. Newspaper articles are a cited source of evidence and often do no more than recycle anecdotes, and it doesn’t help that there are website links that don’t work.
Where properly conducted research is cited there seems to have been no attempt at systematic identification or reviewing of its findin. Primary research studies are either selectively chosen or selectively reported. Two examples: First, The Prince’s Trust, with a comment on the “devastating impact” of the pandemic, noted that an online survey of theirs found a quarter of young people reporting that they felt unable to cope with what was happening in the pandemic. What was hardly mentioned by the Trust’s press releases (and not at all in the HART report) is that there have been only slight changes in recent years in levels of anxiety (53% in 2018, 55% in 2019, 54% in 2020, 56% in 2021) and depression (38% in 2018 and in 2019, 39% in 2020, 41% in 2021). The same applies to reports of increasing rates of self-harm and thoughts about self-harm: “research shows that suicidal thoughts have increased dramatically” refers to a single study reporting a rise from 8.2% to 9.8%, 12.5% to 14.4% in younger adults. Bizarrely the original study isn’t referenced, but instead a synopsis of it published in a different journal.
This isn’t to say that reports of not feeling able to cope, or of ending your life, are trivial. But there is considerable uncertainty about their interpretation, which doesn’t get an airing here. For example, it isn’t clear what the link is between mental health and feelings of not being able to cope if the one changes sharply but the other doesn’t. The meaning of “suicidal thoughts” can be difficult to interpret without a lot more detail than a questionnaire can provide, and the link between such thoughts and suicidal acts or suicide itself is not straightforward. One thing we do know is that a decline in young people’s mental health and increase in self-harm rates is a decade-long phenomenon, associated with social, educational and vocational inequalities and loss of community resources, not something that can be attributed to the pandemic or its management.
The authors describe this report as a rigorously and widely researched document and we have a pretty good idea what rigour in reviewing evidence should look like. After all, David Sackett and colleagues published Evidence based medicine: what it is and what it isn’t in the BMJ in 1996, the Cochrane Collaboration began its work in 1993, and the Journal of the American Medical Association launched Its series on step-by-step critical appraisal in 1993. Set against such guidance the HART overview doesn’t come anywhere close. The parts I have reviewed are more like personal statements of opinion dressed up as science, with conclusions expressed in over-certain and melodramatic ways – crisis, devastation and catastrophe are part of the vocabulary of journalism, not critical appraisal.
Perhaps all this doesn’t matter too much? The report was written to influence a decision about managing the pandemic in the UK and that decision has now been made.
So as a single incidence it will probably disappear from view, but more generally the bias and tone of pieces like this do need to be challenged. They are bad news for several reasons. Mixing polemics with science just strengthens the arm of critics who want to represent academics as opinion-mongers who dress up their own prejudices in pseudoscience. It oversimplifies the public policy dilemmas posed by the pandemic, which are more complicated than lockdown-don’t lockdown. But perhaps most importantly it muddies the waters so it is hard to see what the real problems are; anybody sensible thinks that the pandemic is likely to have mental health consequences and we need an adult discussion about what they are and what the inevitable trade-offs are when implementing a response to the risks posed by the virus. At the moment the HART Group really isn’t contributing helpfully to that discussion.