We've moved on from reading tea leaves to predict death. An interview with Andrea Ganna (PhD), who worked in a team that produced a Lancet publication this week, revealed what the best predictors of death really are and told us that his efforts should not be misunderstood.
The "HowLongHaveYouGot.com Online Test"
Longevity is clearly a subject harder to discuss than an apple pie recipe, but our morbid intrigue has pulled the subject into our conversations. The IT Crowd, a British black humour sitcom, has an episode where they discover a longevity calculator (or how they call it: HowLongHaveYouGot.com). Here is a cut from the show where Roy (Chris O'Dowd) responds to an online death day calculator (Episode: "Deathday", The IT Crowd, Channel 4).
The last time I exercised was...NEVER
Roy is depicted as the incarnation of a stereotypically unhealthy man and of course does not want to know the calculator prediction.
Moss (played by Richard Ayoade) tells him any way.
Roy, it is Thursday...
Predicting death is great for the young and healthy, but do we really want to know our destiny when we know we are unhealthy?
Now There is a Real Test Predicting Longevity
We have perhaps gotten a little too close for comfort to a death day calculator, thanks to a team of researchers at the Karolinska Institute in Stockholm who published an article in the Lancet last week. nuviun interviewed one of the researchers, Dr. Andrea Ganna at Karolinska Institute.
Ubble was based on the UK Biobank Data by looking at how well more than 600 variables predicted actual death in the following five years. UK Biobank recruited 500,000 people aged between 40-69 years in 2007-2010 from across the UK and invited them to take part in a health monitoring project.
Of the 498,103 UK Biobank participants between the aged of 37–73 years, 8,532 died in less than 5 years. From knowing who died and who didn’t the researchers identified 655 different factors and analysed how well they predicted death.
Karolinska Institute ranked hundreds of predictors of how likely death will occur in the coming 5 years. Click on the tab to see the differences between men and women. For example number of cars is a strong predictor for men, but not for women. Importantly, there is no causal relationship.
UbbLE - a new calculator based on the Lancet publication research - predicts death in the next five years for anyone between 40 and 70 years old living in the UK.
The calculator also gives back a so called "UbblE-age".If your UbblE-age is lower than your actual age, you have a lower five-year mortality risk than the average person of your age in the UK.
Interview with the author & publisher of the online longevity calculator, Dr Andrea Ganna from Karolinska Institutet:
Statistician, David Spiegelhalter, Talks About Calculator Validity & How the Results Should be Interpreted
David Spiegelhalter, a Winton Professor for the Public Understanding of Risk in the Statistical Laboratory at the University of Cambridge, wrote to nuviun in an email and informed that he had done the test on a BBC radio show (check it out: minute 50:34 to 57:18). Spiegelhalter has a 2% statistical chance to die in the next 5 years and says it compares well to the 5% of his age-group (of 61) that are statistically expected to pass away in the next 5 years. Spiegelhalter's UbblE-age of 50 makes him 11 years younger than his actual age. Spiegelhalter said,
The test is a bit misleading
Most people in his age group will die of an existing health problem. The prediction of death for the next five years boils down to a single factor: Are you already ill? This doesn't mean that all these additional questions about walking pace aren't valid, they are hints. These hints are predicting whether or not the person is in poor health, which they will most probably already know.
BBC Reporter's questions David Spiegelhalter:
With this online thing, is this just common sense, wouldn't we know this ourselves anyway?
Spiegelhalter's criticism for the test is that people on the healthy end of the spectrum will always get a younger age, and could lead to a false reassurance, as the test doesn't look at one's diet or exercise. In addition it doesn't really tell you what to do. Spiegelhalter also published an online risk calculator for the UK public freely to us, the JBS3 Risk Calculator.
It aims to empower patients to make appropriate decisions about their lifestyle and drug treatments based on a better understanding of their personal cardiovascular disease (CVD) risks, while it also supports doctors to advice patients better on what to do and how to reduce the risk (Spiegelhalter's complete overview here).
In his own UbblE online test, Spiegelhalter received a "UbblE-age" of 50, while being 61 years old (while in the picture above one could think he is much younger).
Now what? Spiegelhalter says that some of the questions are just a proxy for example if you got any money or not (improving your chances to be able to take better care of oneself). Buying another car on the premise to become healthier doesn't make one live longer.
Set of "Strange" Questions in the Ubble Test
If you take the test yourself, you may stumble across rather surprising questions.
The researchers used the cause of death information and 655 of UK Biobank's measurements to calculate the C-index. The C-Index is a measurement of the risk that calculates how well each variable can predict death within five year. The higher the C-index, the more accurate its prediction ability.
The study suggests that health professionals should be supported with predictors to identify high-risk individuals and guide public policy. But pulling down death predictions to the individual level can leave people feeling uneasy about the implications. What is the impact of knowing? Does this burden or relieve us from our fate? Does this help us or hinder us to change our lifestyle? Is the fear factor useful? There are lots of questions which are still unanswered.
As a matter of fact, people like Maev Kennedy, writer at the Guardian do worry about these questions as her piece from today shows.
McCandless's Recipe for a Longer Life:“Be Married Happy Go Lucky Outdoor-loving sex-mad Hippy Party Girl in Senior Management With a Cat”
Below, there is a data visualisation looking at the validity of McCandless and his team's data on longevity and what it is influenced by.
The scatterplot looks at how credible the sources are (from little credibility on the far right of x to very credible on the very left of the x-axis) and how many years are expected to be gained or lost by the different predictors (y-axis). Blue, pink and grey circles describing genders. If you hover over the bubbles, you can click through to the original data source collected by McCandless. Stop smoking is one of the best advice.
As Spiegelhalter said, these are proxies to predicting your health outcomes. Technology however might become much better at predicting deaths accurately, after overcoming hurdles mentioned by Dr. Ganna in our interview.
Companies such as Human Longevity are using the quantified self movement to pull data from Genomics to Microbiomics and applying it to predictive analytics. The question still remains about what to do with the realisations that Roy and Moss in the show IT Crowd were toying with.
Edit: Mitzi Laszlo