Longevity predictions inaccurate because actuaries ignoring data – Swiss Re

Actuaries are underestimating life expectancy because their modelling is not sophisticated enough and does not make the most of the comprehensive medical data available to them says Swiss Re.

The insurer says pension and life office actuaries need to place greater emphasis on disease-centred mortality models, and should attempt to understand the effect of the interaction of several diseases on individuals’ lifespans.

The clinical data needed to build such models exists in the UK, says Swiss Re head of life and health research and development Daniel Ryan, in the form of the General Practice Research Database (GPRD), a live database of over five million patients in the UK that provides high quality information on people with a prior history of disease But to date only Swiss Re, Towers Watson and Legal & General have paid the £50,000 fee for access to the data, says Ryan.

Analysing diseases in isolation is not sufficient, says Swiss Re, because most deaths follow an escalating pattern of disease, with different diseases impacting one another in old age. Instead, any model attempting to predict future mortality needs to track the interaction between different key diseases. To build this type of mortality model, large patient databases such as the GPRD that provide detailed information on individual diseases are required.

Swiss Re says existing models based on projecting forward historic trends do not work. It cites the example of the smoking ban and general reduction in the popularity of smoking. While this has contributed to an increase in longevity in the past, this effect cannot be repeated in the future, it argues.

Any analysis based on cause of death information might be unreliable because determining the final cause of death can be difficult, particularly for elderly people, who are more likely to have multiple diseases, it argues. Estimating future life expectancy by trends in cause of death can underestimate prospective improvements because trends in less common causes may be hidden or distorted by very common causes that are expected to be less important in the future, it says.

Ryan says: “On average UK actuaries are underestimating longevity. People have tried to address it, but they are still short. We need a deeper engagement with the available data.

“The GPRD is available to commercial organizations for £50,000 yet only three companies I am aware of have bought it to date. The reasons why tend to be budget or not having sufficient people to be able to make full use of it.”