17 August, 2011 by ehauke
Unless you’ve been buried in the sand up to your ear lobes at the beach, and there were no sky-writers overhead, you can’t have missed the recent headlines concerning TV and impending doom. Apparently watching television will shorten our lives in a more drastic fashion than established culprits like smoking. Apparently.
Read on to find out whether this is really true*.
Want to Live Longer? Try Turning Off Your TV
TIME magazine – questions – always a favourite in ‘science’ headlines
An hour of TV slashes lifespan by 22 minutes
The Times of India – inflammatory language
Time Watching TV May Shorten Life, Study Says.
ABC News – qualified by ‘so the study told me, honest!’
Killer Box? Too Much TV Can Shorten Your Life
The Hindustan Times – naming and shaming TV as a murderer
Six Hours TV a Day Takes Years Off Your Life
The Independent – decided to go numerical, although stopped short of specifying the shortening of life
Watching TV as Hazardous As Smoking
The Scotsman – went as far as to equate watching TV with the known horrors of smoking
So the headlines tell the story. Or do they?
Well let’s take a look at the original bit of science that has prompted this hysterical outcry in the papers.
Seven scientists published a paper titled Television Viewing Time and Reduced Life Expectancy: A life table analysis in a journal called the British Journal of Sports Medicine. A strange home for this research perhaps as getting ill by inactively watching TV is surely the exact opposite to getting injured or ill through active sport. But moving on.
Now. The researchers started off with the hypothesis that watching more TV might be associated with poorer health and even greater risk of death. They looked at data that had been collected by the Australian Bureau of Statistics and The Australian Diabetes, Obesity and Lifestyle Study – which was a national observational study that began in 1999. The authors of the study then used some clever maths to ‘model‘ the impact of television viewing on life expectancy. Now modelling in this case is a bit like making an educated guess. There was not enough prior evidence of an association to make this model really watertight.
And then let’s take a brief look at their results. Now, all the results show the same pattern in the statistical analysis, so I’m just going to pick one particular figure as an example of why we should’t be too concerned with this study.
Why bother with complicated sums and analysis and not just tell us what they found? Well, when scientists or researchers do their sums at the end of a study, they’re trying to make sense of the results that they’ve come up with. And there are lots of statistical tools that they must apply to the their results in order to show how accurate their numbers are. And how likely it is that those numbers are actually representative of the world at large. For example, it’s no good saying that in a study of 5 people 3 had red hair. We need to know if this makes it likely that 3 out of every 5 people in a particular country or even the world have red hair. So statistics help us to understand how relevant and accurate scientists’ results are.
So in this study, the scientists looked at whether watching TV decreased lifespan. But do their numbers add up, are they accurate, and will the results stand for the general population of the world, or just the people that they looked at in their research?
Now they applied a test called the confidence interval or uncertainty interval to their results. This gives a range of accuracy. So if 3 out of 5 people have red hair, the confidence interval might be 2.9 to 3.1.
Now this would mean that the researchers were 95% sure that if there is any inaccuracy in their results, the actual answer would be between 2.9 and 3.1. So their result of 3 is pretty accurate. And all this determined by some complicated maths.
But what if their 95% confidence interval had been 0.1 to 4.9. They were 95% sure that between 0.1 and 4.9 out of 5 people might have red hair? Obviously this would mean that they really haven’t got a clue how many people have red hair. This would warn us that the statistic they were telling us – that 3 out of 5 people have red hair might not be very accurate or relevant when applied to the greater population.
So a wide confidence interval means less valid data. So what were the results like in this TV death study? Well, as I said, all their results showed the same type of confidence interval, so I’ll just give one as an example.
They found that watching 6 hours TV per day would shorten your lifespan by 4.8 years. But. And this is a big but. The uncertainty interval or confidence interval is really wide (95% CI 11 days to 10.4 years). This means that the researchers can only be 95% certain that the actual figure is between 11 days and 10.4 years. A massive difference. Imagine they’d instead said that watching six hours of TV a day could shorten your life by 11 days. That would have been a very different story. And it still could be, according to the confidence interval.
So this data has some BIG problems.
Causality and Responsibility
And there is another issue with us all going about our lives believing that watching TV is knocking time off our lives. Or even that if we don’t watch any TV today we can get away with a crafty cigarette tomorrow.
There is no evidence that it is watching TV that actually shortens our lives – there is no evidence that it is the TV watching itself that causes earlier death.
All the numbers show is that if you are a person that watches TV, you might have a shorter life.
That could be because people who watch more TV eat more unhealthy TV meals. It might be because people who watch more TV do less exercise. It may be because people who watch TV do so because they’ve got a dodgy heart and they need to take more rest.
There is a good article on the BBC website about this. Paraphrasing Fergus Walsh –
And we need to be very careful when we infer that something like TV is as bad as cigarettes. It might make smokers think that smoking is not as unhealthy as we know it is. That has been proven with narrow confidence intervals and clear evidence of cause and effect.
So when we see a startling headline, we need to think whether people are trying to sell papers (or advertising) or whether there really is a stark message to be learnt. If you even just click on the link to the original science paper – the authors say in the first paragraph, that they are ‘estimating the impact of TV watching on life expectancy’.
I could estimate a lot of things. But I don’t. And perhaps they should’t either.
* This article expresses my own opinion and understanding of statistics (which as I mentioned is rather limited). Please don’t be too harsh, but feel free to correct any mistakes in the comments. Also the original research expresses the intervals as Uncertainty Intervals but I prefer Confidence Intervals as a terminology.