There is a great interest in understanding what might predict life expectancy in the general population. We already know several risk factors that have a substantial impact on mortality. For example, smoking, lack of exercise and increased body mass index have been shown to increase the risk of dying earlier.
Most previous studies have investigated one risk factor at a time – and usually in a limited number of individuals. However, in the first study of its kind, published in The Lancet, we compared many potential risk factors in a large number of individuals and come up with a tool that can be used by anyone living in the UK and aged between 40 and 70 to calculate their risk of dying within the next five years. In addition users can work out their “Ubble age” – a single prediction score that uses information about you to match your risk profile to the age of the average person of your gender in the UK.
You can calculate your five-year risk of dying as well as your Ubble age using our interactive website, developed in collaboration with Sense About Science. After filling in a dozen or so simple questions, the tool is able to calculate your risk and your Ubble age.
We examined and compared more than 655 measurements of demographics, health, and lifestyle factors in around 500,000 people from the UK Biobank who were aged between 40 and 70 years old. We then used a statistical survival model to work out the probability of these measurements predicting death from any cause, and also six specific causes, for men and women.
We analysed how well each measurement predicted mortality using something called the C-index. The higher the C-index, the more accurate the measurement is in discriminating between those who will die and those who won’t in the next five years. A C-index of 50-60% is considered a poor predictor, 60-70% moderate, 70-80% good, 80-90% very good and over 90% is considered excellent.
We further selected the most predictive measurements, 13 for men and 11 for women, and combined them into the prediction score. For example we selected questions such as: “Do you smoke now?” or “In general how would you rate your overall health?”. The prediction score is also used to calculate the Ubble age, which is simply the age of an average individual from the UK of same gender that matches the predicted risk. So for example, if you’re a 53-year-old woman and your estimated risk of dying in five years is 2.4%, the most similar risk in the UK life tables is of a 56-year-old woman. So your Ubble age would be 56.
We found that the variables that most accurately predicted death from all causes within five years did not need to be measured by physical examination, but could be reported by individuals through a questionnaire.
For example, asking people to rate their overall health (self-reported health) and to describe their usual walking pace were two of the strongest predictors in both men and women for different causes of death. It turns out that walking pace is more predictive of five-year mortality (a C-index of 72%) than body mass index (C-index of 68%).
We also found that among individuals who did not have any known major disease, measurements of smoking habits were the strongest predictors of mortality within five years.
Overall, our study allows comparison of a large number of predictors of overall and cause-specific mortality in a way that has never been done before. As well as being open to use by people to improve their health awareness, health professionals and organisations can also use it to identify individuals at high-risk.
Our website was evaluated by several audit groups, including scientists and representatives from patient organisations and other stakeholders as well as individuals from the general public.
However, there are some limitations in our study. The UK Biobank participants are not representative of the entire UK population, as they are more affluent and have lower mortality. We addressed this partly by recalibration of the prediction score using data from the UK census. We also only looked at the risk of dying within five years, and it is unknown how the prediction would look like beyond this period.
Finally, the tool is a predictor and cannot claim causality. Most of the strongest predictors of mortality are correlated, not proven as causally related. For example, a slow walking pace is a strong predictor of dying, but walking faster won’t necessarily lead to a reduced risk of dying. Smoking and dying, however, is one of the few things that we know are causally related.
Erik Ingelsson is Professor of Molecular Epidemiology at Uppsala University. Andrea Ganna is Research Fellow at Karolinska Institute. This article was originally published on The Conversation. Read the original article. Image by Mark van Laere under Creative Commons license.