Context Is Everything

It was one of those Hang on a minute moments.

Action for Safety sounds like a good idea, using celebrities to promote safer driving. The poster campaign features slogans about not driving while tired, looking out for pedestrians and so on. Celebrities include sports people and minor European royals.

The posters all state prominently that 3,500 people will die on the roads today. That’s 1,277,500 a year.

That seemed like a lot so I looked it up.

According to the latest government figures for the year ending June 2018, around 1770 people died on UK roads, about 4.85 a day. That’s quite a lot fewer.

Who is behind this campaign?

The organisation behind it is FIA (Fédération Internationale de l'Automobile), whose mission is ‘to bring coherent governance and safety to motor sport’. They also aim to ‘ensure that safe, affordable and clean systems of transport are available to all. The promotion of safe and sustainable forms of mobility has in turn led the FIA to commit to global sustainability initiatives and also to found its own major response to road safety concerns, FIA Action for Road Safety. This worldwide campaign, in support of the UN’s Decade of Action for Road Safety, aims to reduce fatalities on the roads by five million before 2020.’

All very honourable aims. But where do the stats come from? According to their fact sheet ‘Road crashes kill 1.3 million people every year’.

The source of this statistic is the World Health Organisation. Their site says that 1.25 million are estimated to have died in 2013 globally in road deaths.

So the FIA stats are not that far off. But the posters are very misleading, nowhere on them does it says these are global figures. They may be going for shock tactics to make people drive more carefully but is this a legitimate tactic? Is it ethical to be so misleading? It’s certainly unprofessional and amateurish.

If that was the UK figure it would be truly shocking but on a global scale is that significant? How does it compare with other causes of death?

According to the WHO, road injuries came eighth on the list of global killers in 2016, lagging a long way behind heart disease and stroke.

Three quarters of victims were men and boys, which would explain the prevalence of footballers and racing drivers on the posters as these are traditionally assumed to be male heroes and role models (but doesn’t explain why a princess of Monaco features…).

The distribution of deaths isn’t even globally either; some countries have much higher per capita rates than others. The worse death rates are in several African countries and Thailand. The UK has one of the lowest rates (and falling) while Monaco has the lowest of all (which might explain the princess on the poster).

That still makes them a pretty serious cause of death and a worthy campaign subject. But making the stats relevant for each country would make the campaign more relevant. Nearly five people a day every day is still too many.

The last part of the stated mission is 'This worldwide campaign, in support of the UN’s Decade of Action for Road Safety, aims to reduce fatalities on the roads by five million before 2020'. That's half a million out of 1.3 million a year for the decade of the campaign. Is that realistic in any way?

Although this is just a small example, it’s a useful skeptical reminder that without context statistics are meaningless. If even the good guys are misusing stats then it’s even harder to know who to trust, who to take seriously and who to dismiss. 

So yes, drive carefully and soberly and wide awake but keep the same attitude towards any stats you encounter along the way.

How Old Is Your Heart?

Public Health England (PHE) in collaboration with the British Heart Foundation (BHF) has launched a heart check-up to identify ‘your heart age compared to your real age’.  Is it a valuable public health intervention or a waste of time and money?

It begins by asking age, gender, ethnicity and postcode.

Then: Do you have cardiovascular disease? No.

Do you smoke? No.

But I did. I quit two years, three months and thirteen days ago. Not that I’m still counting. But this does make a big difference. My heart is ‘older’ than it would be if I’d never smoked, I know that. It can take up to ten years for risk levels to return to those of a non-smoker.

Next it asks for height, weight and a few questions on medical history. Pretty standard.

Then it asks what your cholesterol level and blood pressure are. How many people know that?

And that’s it, end of test. It tells me that my heart age is four years older than my real age and says this:

 Your risk of having a heart attack or stroke within the next 10 years is 5.8%

Your heart age is an estimate because you don't know all your numbers. We've based your result on the national average.

If you have high cholesterol and blood pressure, your heart age could be as high as 69.

Make an appointment with your doctor, nurse or pharmacist to get your blood pressure and cholesterol tested.

There are no questions about alcohol and drug consumption, diet and exercise, all things which the NHS and PHE normally tell us make a big difference to heart health. I’ve lived in central London all my adult life and we keep being told how damaging pollution is to health and longevity so it might be a good idea to include a question on that.

The NHS is already struggling, does it really need a few million people making appointments to get their BP and cholesterol checked? Or do they expect people to rush out and buy self-testing kits? And then going to the GP when they get a high result. GPs have not surprisingly expressed concern about this test adding to their already extreme workload.

Is it good public health policy to frighten people with an unrealistic heart age? Who is the mythical ‘average’ person the figures are based on? I’m considerably taller, lighter and fitter than the average woman, for example (and an ex-smoker, yes, I know) so they can't use her to scare me like some sort of Baba Yaga.

There’s also the question of efficacy. NICE has already rejected the use of lifetime risk scores because of a lack of evidence, which means they will scare people, overload GPs - and not have any effect. It’s not even a case of the end justifying the means.

The people behind the project are defending it in various parts of the media by saying it will raise awareness and start a conversation. As an awareness raiser it fails because all it will do is frighten people into going to the GP, or frighten them and then they’ll carry on as normal, or make them think about making some changes and then give up after a few weeks because that’s how we operate. No imminent threat, no motivation.  Will there be any follow-up to see if people have taken action? Doubtful.

If it’s a way of collecting information about the population, then the holes in it make that information of very limited value – unless all PHE wants to know is how many of us don’t know our BP and cholesterol. This is not a compulsory test. The people doing it are self-selecting, opting in, and the reasons they do this could skew the results. Is it idle curiosity, concern, fear or some other reason? Is it mostly the group known as the ‘worried well’? People on very low incomes and many old people will be excluded because of lack of online access.

Heart disease is a serious and growing problem that takes long-term investment and lifestyle management to tackle.  According to the British Heart Foundation:

Heart and circulatory diseases cause more than a quarter (26%) of all deaths in the UK - over 150,000 deaths each year, an average of 420 people each day or one death every three minutes.

But hang on a minute… it also says:

Since the BHF was established the annual number of deaths from heart and circulatory diseases in the UK has fallen by more than half.

In 1961, more than half of all deaths in the UK were attributed to CVD (320,000 deaths).

Since 1961 the UK death rate from heart and circulatory diseases has declined by more than three quarters.

Would it not perhaps be worth looking at why rates have fallen, what else people are dying of, what interventions have worked in the past, where limited resources could be most usefully targeted? For example, obesity has risen since 1961, smoking rates have fallen and life expectancy has risen so people are more likely to have multiple comorbidities that accumulate with age – more than one potentially life-limiting condition at once.

As it happens, I did have my BP checked this week as part of an ongoing treatment for something not heart-related. I didn’t put it in the test because I wanted to see what results came up. The nursing assistant who checked it said he’d tried to take the test but it didn’t work because he was too young. He agreed with me about the flaws in the design. (My BP is great, thanks for asking).

If you do want to know about the state of your heart and its future, you’d be better off doing this:

Equalities Are Not a Numbers Game

Various statistics are being bandied around at the moment about how many people support or oppose equal marriage for lesbian and gay people. They're intended to persuade us of the moral worth of the arguments for and against. Not surprisingly, both sides are citing polls that support their position. There are also contradictory statistics doing the rounds about how many support or oppose abortion, or assisted dying, or how many Christians there are in the UK.

Statistics make for good headlines and propaganda but democracy should not be confused with the alleged will of the masses. There were times when the majority of people thought it was fine to deny women the vote, have slaves or force small children to work in factories.

If any group suffers some disadvantage or is denied a privilege or right given to others then morally it does not matter how many of them there are. Equally, if some group claims a privilege denied others, the numbers are irrelevant. Otherwise, when should we start caring? When one person is affected, or a hundred, or a million? Without a moral argument, numbers don't help.

The egalitarian will seek to redress the balance either by making an opportunity available to everyone (for instance, universal enfranchisement) or by removing a privilege from those who have it (for example religious groups claiming exemption from equality laws).

Any group losing its unfair privilege will of course complain, claim unfairness, cite some spurious historical precedent or scriptural justification or play the numbers game.

There are times when numbers can be useful, for example to show that a problem is widespread. However, statistics in isolation, without a moral argument, should be not given undue weight when equalities are the issue. They may be an indicator but sometimes what they indicate is that the egalitarian will have an uphill struggle. Opposing groups throw numbers at each other as if they were the killer blow. But might is not right, not on its own. The wisdom of crowds is often not very wise at all.

The pragmatist may be forced to play the numbers games with politicians who are aware that numbers may equal votes but there is no moral high ground in doing this and fighting fire with fire sometimes just makes a bigger blaze. Politicians often use statistics solely to give credibility to a decision they have already taken and ignore the ones that don't suit them.

Quoting percentages can be a cheap tactic, one often used in adverts - nine out of ten women agree! (and then in small print that the sample group was 119 women or some other improbable number). Or nine out of ten cats. Their effect often relies on the fact that people instinctively respond to big numbers, especially if you flash up the small print very quickly or hide your dubious methodology. It's instinctively safer to stick with the herd, particularly if you don't have a full grasp of the issues.

Polls can of course be manipulated by the way in which questions are framed in order to produce the desired result. It's often in the interest of people conducting polls to ask questions without providing both sides of the argument for people to make a reasoned decision. A quick emotive response makes for better spin.

The Government is now reportedly considering whether to include a question in their consultation on equal marriage about whether people think it is a good thing. The wording of this question will be highly significant, especially because of the weight of religious lobbying influencing politicians.

Being swayed by public opinion about whether a group should be given rights others have is not true democracy. The majority of men would not have agreed to give women the vote in the early 20th century, for example. Vested interests, ignorance, protectionism, manipulation by opposing groups or just plain bigotry are always potential obstacles. Many people do not apply Rawls' veil of ignorance when deciding what they think about equalities issues.

Mistrusting the apparent will of the masses is nothing new - in 1841 Charles Mackay wrote Extraordinary Popular Delusions and the Madness of Crowds.

Unfortunately, the moral argument often doesn't make for good headlines. It can be subtle or a hard sell or require time and effort to grasp, so people with the best of intentions fall back on statistics. It can be particularly hard to sell the moral argument when the opposition is a religious group claiming that there is no morality without their moral code or that people opposing them lack morality because they are not religious - or not the right kind of religious.

There are going to be a lot more numbers in the headlines around the time of the equal marriage consultation and politicians will have an eye on winning or alienating voters, especially if religious leaders persuade them that they can sway their congregations (conveniently ignoring the fact that polls show congregations disagree with doctrine).

In the case of equal marriage, it all comes down to whether we should withhold a right from one group of people. The religious extremists argue that we should because this group is not equal, they are effectively inferior citizens to whom human rights legislation apparently doesn't apply. They use statistics to mask this prejudice.

Playing percentages may make you (temporarily) popular but Cameron needs to stick to his guns and not be swayed by numbers or doctrine. Either he believes that equalities apply to everyone or he doesn't. Either he has the courage to put that belief into policy or he doesn't. Doing the right thing is not a numbers game.

HRT Cancer Scare - here we go again

Scare stories about HRT come round every few years. This time, according to the Daily Mail, HRT can now treble the chance of dying of breast cancer. Other shock headlines include 'Study shows HRT even riskier than thought'. The Washington Post goes with 'Hormones also raise death risk of cancer' and talks about 'powerful evidence'.

Menopause is still a taboo subject that a lot of women don't feel comfortable talking about, which makes it easier for myths and misunderstandings to spread.

If you read through to the end of some of the articles they do mention the stats but the headlines are what will stick in people's minds and may be the only part of the story they see - they are what's known as the take-away. A lot of people aren't that good at interpreting stats even if they get that far.

The headlines are based on research published in the Journal of the American Medical Association (JAMA). A study of 12,788 women for 11 years looked at women taking a combination of oestrogen and progesterone in a randomized placebo controlled trial.

Rowan Chlebowski of the Los Angeles Biomedical Research Institute at Harbor-UCLA Medical Center who led the analysis commented in the Washington Post that 'the risk in the study was low and barely met the threshold for being considered statistically significant' but he was 'confident that the risk was real'.

Barely Statistically Significant Cancer Risk! is not a headline you're likely to see. Even if such a headline existed, it's the C word that readers would remember. For cancer, there is no such thing as bad publicity.

The figures are these: in the placebo group there were 0.01% deaths and in the HRT group there were 0.03% deaths a year. That's 12 and 25 women in 12,000. The research also found that combination HRT was associated with more invasive breast cancers (0.42% of cases per year in the group as opposed to 0.34% in the placebo group).



With any apparently alarming statistics, it's the explanation of absolute and relative risk that rarely gets a look-in. If a disease kills two people in ten thousand one year and four people the next then the mortality rate has increased 100%. The chance of dying is still only four in ten thousand but that doesn't make for a good headline.

During the 2008 HRT breast cancer scare, also based on research by Rowan Chlebowski, Behind The Headlines pointed out something news stories glossed over: "In addition, the authors of the study point out that the differences seen should be interpreted with caution, as they may have resulted from differences in health seeking behaviours in the two groups of women after the trial. Women who had been told that they had been taking combined HRT at the end of the trial and knew of the cancer risk may have been more likely to seek medical attention for any suspicious symptoms than women who knew they had only received placebo.

BTH also pointed out that, despite predictable scare headlines: 'the study was not directly investigating any link between breast cancer and HRT. Instead it looked at whether HRT increased the chances of detecting an abnormality on a mammogram that then required a biopsy for further investigation; this would not necessarily involve a diagnosis of breast cancer.

'The main finding of the study was that the diagnostic accuracy of mammography was decreased in women who had taken combination HRT.

'This study only investigated one type and one dosage of combination HRT'.

There are many different types and levels of dosage. For example, the combination of estradiol and dydrogesterone (oestrogen/progesterone) sold as Femoston in the UK increases the risk for women in their 50s of having a stroke from 3 in 1000 to 4 in 1000. The risk of blood clots goes up from 3 in 1000 to 7 in 1000. And the risk of breast cancer goes up from 32 in 1000 by the age of 65 to 38 in 1000 if they take it for 5 years.

The risks of HRT are real in the sense that they do exist and the latest research seems to indicate that they are higher than previously believed. Risks also have to be offset against benefits - any good GP will explain this. Some of the risks can be considerably lessened by lifestyle changes and regular check-ups. There is a good basic introduction to HRT here.

Here's another risk: in 2004, the lifetime risk of dying in a motor vehicle accident was 1 in 82 in the US and 1 in 240 in the UK. Killer Cars!

Master of all you survey

Torture numbers and they'll confess to anything
Greg Easterbrook

Since Domesday Book was compiled in 1086, large-scale data collection by survey has been part of government in this country. Can surveys based on self-reported information ever be reliable, whether it's an 11th century peasant reporting how many ducks she has to a Norman or a 21st century woman giving personal information over the phone or online? Are such surveys inherently any more reliable and honest than the 'personality' quizzes in teenage magazines? Not least among the factors influencing responses are the way questions are framed and how the respondent thinks the information will be used.

A recent survey for the Office for National Statistics (ONS) based on interviews with 450,000 people for an Integrated Household Survey (IHS) has been reported as finding that 71% of people in the UK are Christian and 20% have no religion, 1.5% are gay or bi and nearly 80% perceive themselves to be in good health. (All IHS statistics are considered experimental until assessed by the UK Statistics Authority).

The IHS asked respondents 'What is your religion, even if you are not currently practising?' with the intention of discovering 'religious affiliation - that is identification with a religion irrespective of actual practice or belief'. Not surprisingly, this has been reported as 71% are Christian.

Another survey by the ONS in 2008 found that only 22% described themselves as Christian and 45% said they had no religion.

The 2011 census will include the question 'What is your religion?'. This question was asked in the 2001 census and before that only in 1851. There has been controversy about the result of the 2001 census where 71% of people self-reported as Christian. Many of these, it is believed, identify as culturally rather than religiously Christian. Most of them rarely go anywhere near a church. According to a survey done by the Church of England, only 5% go at Christmas and 2.8% at Easter - the most important Christian festival.

Less seriously, 0.8% of people put their religion as Jedi in 2001 which offically makes them a bigger group than Sikhs, Jews or Buddhists in the UK.

The ONS deputy director said that the religion question in the next census would be 'a fabulous insight into societal changes to see how people register their religion'. Registering it is of course not the same as actively practicing it. And is 'fabulous' the best word to use for such a serious project?

Both religious and non-religious groups use these surveys in their campaigns to demand or challenge legal, financial and educational privileges, and governments use the findings to decide on funding, among other things, so they are more important than being of passing cultural interest. The widely differing findings are hardly a solid basis for policy or anything other than reflecting how variable survey findings can be.

People can and do change their minds about what they believe (although probably not so many of them in such a short space of time) but sexuality is a little less mutable.

The gay/bisexual statistic in the survey has led to headlines like 'Only one in 100 Britons is gay despite long-held myth' in the Mail.

It's inevitable that some groups will use stats to serve their own agenda and affected people will challenge them, especially if they have fought hard for equality.

One response to these findings on Facebook was '‎2,185,072 gay men and lesbians are currently registered on Gaydar in the UK - equating to 6.7 per cent of the UK population'.

More officially, in 2005, HM Treasury and the Department of Trade and Industry did a survey to help the Government analyse the financial implications of the Civil Partnerships Act (pensions, inheritance, tax benefits). They found that there were 3.6 m gay people in the UK – around 6% of the population. This figure was greeted by some gay rights activists as realistic.

If there are now only 1.5%, where have the other 4.5% gone since 2005? It should be noted that 3% of IHS respondents either said 'Don't know' or refused to answer. Again, the differences in survey results make basing any action on them a leap in the dark.

As to the 80% who 'perceived themselves to be in good health', what does this prove? Feeling well and being well are not the same thing at all for a start.

Even when stats are not used to tax the hell out of conquered Anglo Saxon peasants with very good reasons to be creative in their self-reporting, surveys are not like scientific tests. They are not reproducible in lab conditions, the methodology can be peer-reviewed but there can't be placebo questions, double blinding or a control group. At best, they can provide useful demographics, at worst they tell us nothing and can be used for propaganda. If you don't like the findings of the current survey, just hold on and there'll be another one along shortly.