by DA Shields, JR Brigstocke, JH Scurr
There are many situations where accurately predicting a person’s life expectancy, determining how long they will live (and, indeed, the quality of that life), can be extremely useful to the court in determining a financial settlement.
Can we predict with any degree of accuracy when a person will die, as such a figure will have a significant effect on the overall settlement?
In this review we have considered the historical background to life expectancy and what steps are currently being taken to improve the forecasts. By taking a more scientific approach and comparing our results with outcome, we expect to be able to determine life expectancy with an increased degree of certainty.
Courts often require estimation of life expectancy, either to calculate possible financial loss due to inability to work or early death, or to estimate duration of life to calculate quantum for care provision. Occasionally, too, a request is made to calculate loss of chance due to, for example, a delayed diagnosis of cancer with a (possible) worsening of prognosis.
Historically, life expectancy was used by the insurance industry, usually to provide pensions for widows and children, but calculating premiums had no scientific
background until 1662, when John Gaunt analysed births (christenings) and deaths (burials) in the City of London. Interest in keeping such figures remained within the insurance industry until the registration of births, marriages and deaths became compulsory in England in 1837, and the first official English life tables were published in the Registrar General’s Fifth Annual Report of 1843.
Since then such tables have been regularly produced, in the UK by the Office for National Statistics (ONS, who took over from the Government Actuary’s Department in 2006). The national period life tables are produced annually and use data from population estimates and deaths by date of registration data for a rolling period of three consecutive years. The current set of national life tables (2015) are based on the mid-year population estimates for 2012-2014. These are divided by sex, then further subdivided by country, then by local areas. Similar tables are produced by other countries such as, for example, the Central Statistics Office in Ireland.
UK Courts use the Ogden Tables (the latest Ogden Tables being the 7th edition, published in 2011) for personal injury and fatal accident cases, following the Civil Evidence Act 1995. These are produced by the Government Actuary’s Department, based on data provided by the ONS. Unfortunately the figures for projected mortality (Tables 1 & 2) lag behind the ONS tables, the 7th edition using data from the 2008-based population projections, so for mortality projections it is more accurate to use the ONS figures. Also, the Ogden tables only give figures for the whole of the UK.
The ONS tables allow one to look up the age and sex of a person and be given a remaining life expectancy. The ONS also give cohort life tables that provide mortality rates that vary over time for each age. This cohort life expectancy figure takes into account future projected mortality rates by age and for each year, so is higher than the equivalent period life expectancy. This is because the cohort figure incorporates future assumed improvements in age-specific mortality, and can be regarded as a more appropriate measure of how long a person of a given age would be expected to live on average. These tables are produced biennially based on assumptions for future mortality from the National Population Projections, produced primarily to provide an estimate of the future population of the UK and based largely on extrapolation of past trends in rates of mortality improvement (the average annual rate of improvement has been about 1.2% over the whole of the 20th century), together with expert opinion as to future mortality rates.
However, such a figure does not represent the actual likely age at death, but rather gives the median value (middle value in a sorted list of numbers), and clearly for a given age and sex there will be a range of values of years lived within this group. The distribution of deaths within the group will be roughly bell-shaped with negative skew, with a peak at the mode (the value that appears most often in a set of data), with a range either side. Currently for men the mode is 86, seven years more than the mean (average) of 79, and the median is 3 years more than the mean, so the term 'life expectancy' can be misleading. Accordingly, an attempt can be made to determine which side of the median any particular person is likely to fall, and also to calculate how far away from the median death might occur.
Unfortunately, one occasionally meets an ‘expert’ in a case who is prepared to state that there is no need for any recalculation from the median given in the ONS
tables. However, from a medical perspective it has been clear for some time that it is possible to give a potentially more accurate estimate of mortality than this in individual cases. The argument usually raised in favour of using the median is that the tables encompass the whole population, thus including such risks as diabetes and smoking, yet it is sometimes hard to get an expert to accept that if your client does not suffer from such risks, he should not be penalised for his good health (or vice versa).
It has been recognised anecdotally for a long time that certain persons are likely to live a longer or shorter period than expected (indeed, in roughly 1000 BCE King David in Psalms 90:10 noted the years of our life are 70, or even by reason of strength 80, demonstrating an understanding that we do not all live to be the same age!). From a more modern perspective we recognise changes based on socio-economic and medical data. Tables are always produced giving figures for both sexes, as it is well-recognised that mortality varies by gender, and inequalities are also evident in different geographical locales, hence the provision of data by country and county in the UK tables. Attempts are constantly being made to try to improve the data (for example, the Longevity Science
Advisory Panel - http://www.longevitypanel.co.uk/), as differences in life expectancy by socio-economic group have continued to widen. There is some overlap between medical and socio-economic risk factors (for example, increasing obesity leading to increased high blood pressure, diabetes and heart disease), but also increased research into the effects these have on life expectancy.
Early work on reducing mortality was based on such methods as vaccination, reduction of scurvy in ship’s crews, and improvement in sanitation. Once the prevalence of infectious diseases began to reduce, public health experts began investigating chronic diseases such as heart disease and cancer. The first modern paper to show a significant association between smoking, lung cancer and heart disease was that of Doll and Hill in 1954 (Doll R, Hill AB. "The mortality of doctors in relation to their smoking habits". BMJ 1954;328 (7455):1529–1533), together with further follow-up papers. In 2004 this allowed a calculation of reduced life expectancy of those who smoke until 40 of 1 year, those who smoke until 50 lose 4 years, and those who smoke until age 60 lose 7 years.
Since this seminal paper, there has been a steady stream of epidemiological studies looking at life expectancy in such diseases as hypertension and heart disease (the pre-eminent study being the Framlingham Heart Study, commenced in 1948, which now allows an easy calculation of an individual’s 10-year cardiovascular risk score), diabetes and, of more recent interest, obesity. Early papers tended to look at a single risk factor, which led to difficulties in calculating a cumulative risk caused by several, often overlapping, diseases, but more recent papers have recognised this problem and analysed multiple related risk factors together - see, for example, the excellent paper by Leal, Gray and Clarke (“Development of life-expectancy tables for people with type 2 diabetes.” European Heart Journal 2009;30:834-9) that contains a series of tables that enable one to look up life expectancy based on level of diabetic control, blood pressure control and cholesterol levels for the two sexes. However, to produce an accurate figure one needs access to the whole of a patient’s medical record (including such factors as place of origin, height and weight, and information on their educational achievement, occupation and smoking and drinking history as well as their full medical history), together with constant searching of the scientific literature for newer papers on the various risk factors identified. The literature is being constantly refined as, for example, we gain a better understanding of how good control of blood pressure or diabetes impacts upon the overall survival figures. Such work is essentially a full-time job.
There are two groups who require more detailed examination and clinical expertise – neurological damage (either at birth or subsequently) and patients diagnosed with cancer. For the former, there is again an increasing body of literature as to outcome in traumatic brain or spinal cord injury, but there is, perhaps, less certainty as the development of complications such as pressure sores, chest infections or sepsis can rapidly alter the outcome, and outcome from neurological damage from stroke is again difficult to predict accurately. Brain damage in infants is the group that perhaps requires the most detailed analysis, as the greatest single proportion of damages in the UK is paid as compensation to successful claimants in brain damage at birth litigation. One author in particular, David Strauss, has been responsible for providing data in cases both of such brain injury (and in the UK by Lewis Rosenbloom – see, for example, Strauss D, Brooks J, Rosenbloom L, Shavelle R. Life expectancy in cerebral palsy: an update. Developmental Medicine & Child Neurology 2008,50:487–493) and spinal cord injury, for the latter together with such other authors as Shavelle, DeVivo and Frankel.
For cancer, there are now grading systems for every type that will give a reasonable idea of possible survival, though a difficulty sometimes arises as to what difference a late diagnosis has made – when essentially an educated guess has to be made as to what stage the cancer might have been at (and hence what chance of a cure) compared to when it was eventually diagnosed. Various factors may help (such as the histological grade of the tumour), but for the moment this area remains rather grey (and often requires further input from an oncologist and from other specialities such as pathologists and radiologists).
New risks are also being identified, for example, stress, particularly at work (with the suggestion, for example, that the cardiovascular health risk of stress is not dissimilar to the risk from cigarette smoking, according to Laura Kubzansky, Professor of Social and Behavioral Sciences at Harvard School of Public Health).
Early tables to calculate life expectancy were prepared primarily for the insurance industry, who require an analysis for the purposes of calculating premiums,
and current tables are produced for government statistical purposes. Such tables clearly have limitations with regard to life expectancy for an individual. Whilst there can be no certainty about calculation of life expectancy and there will always be exceptions to any rule, using all the available risk factors we are confident that our predictions in individual cases can be more accurate than the median quoted in the ONS tables. We are now preparing an increasing number of life expectancy reports, and are attempting to correlate our predictions with the actual date of death, to determine the accuracy of our predictions.
Accordingly, there is no reason to accept the median figure given in the ONS tables as to life expectancy, as this represents the starting point for such a calculation, not the end point. To identify the various risk factors and calculate a more appropriate figure takes time, access to considerable data regarding the client, and an active interest in the field such that the literature base can be kept up to date.