[Skip to content]

Sign up for our daily newsletter
The Actuary The magazine of the Institute & Faculty of Actuaries

Critical illness looking for trends

Over three-quarters of a million critical ill-
ness (CI) policies are sold each year,
many of them with premiums guaranteed for ten years or more. Pricing this business poses an enormous challenge, both in estimating current morbidity, and in predicting future trends. The insured experience available for analysis is limited, and still likely to fall within a select period. Current pricing approaches therefore focus on adjusting general population data to be relevant to an insured population.
Much has been written about the variation of mortality by socio-economic class. The mortality gap between the most affluent lives and the most deprived lives has widened over time. In the absence of any government initiatives, this widening of the mortality differential is expected to continue, with the more affluent lives continuing to benefit to a greater extent from changes in lifestyle and medical advances than will the more deprived lives. Should we expect the same relationship between the general population and the insured population for critical illness?

Deprivation category
Carstairs and Morris developed a method of allocating people to socio-economic groups at postcode level. This classification is based on four indicators judged to represent material disadvantage in the population, overcrowding, male unemployment, social class, and car ownership. The deprivation score is then distributed over seven categories from 1 (the most affluent), to 7 (the most deprived). We have taken deprivation categories 1 to 3 as a proxy for the insured population.
Deprivation categories are a more effective measure of social level than an occupation-based social class. However, currently only Scottish data are available in sufficient detail to analyse trends.
We may expect more affluent lives to have a lower incidence of CI conditions, owing to the lower prevalence of smoking, better diets, and higher standards of education. We use the term ‘positive correlation’ to describe this expected higher incidence among the most deprived lives. However, more affluent lives tend to have more routine medical check-ups or regular screening, which may detect the illness earlier. This may lead to a reduction in the mortality rate for insured lives, but an increase in the incidence rate for critical conditions.

Heart attacks
As expected, we see a strong positive correlation between the incidence of heart attacks and deprivation category. In fact the rate of first incidence of heart attack in the most deprived groups (classes 6 and 7) is over twice the rate in the most affluent groups (classes 1 and 2), as can be seen in figure 1. The large difference is principally the result of smoking and diet.
Despite the widening mortality gap by social class, male CI incidence rates improved faster for the general population than for the insured group (see table 1). We believe this difference is partly because of the greater likelihood of the insured group’s seeking medical help. This is supported by the mortality trends, where improvement has been greater in the insured population. Greater awareness of symptoms and improvements in medical intervention are increasing survival rates from a first heart attack.

Again, there is a strong positive correlation between stroke and deprivation category. First incidence of stroke in the most deprived groups is three to four times the rate in the most affluent groups (figure 1) an even wider gap than that seen for heart attack incidence. Strokes are subject to broadly the same risk factors as heart attacks.
Stroke incidence has increased for males and females in the past ten years. The detection of strokes has improved greatly, particularly after the introduction of better diagnostic techniques, including magnetic resonance imaging (MRI) in the late 1980s. The reduction in heart attack incidence is also believed to have affected stroke incidence. Increases in operations such as coronary artery bypass grafts and angioplasty have helped to reduce heart attack levels, but do not counter arterial disease in other organs, such as the brain. Finally, medical treatment of strokes has improved greatly, with a corresponding impact on mortality rates.

The link between cancer incidence and deprivation category is less straightforward. If we analyse incidence data by the site of the primary tumour, we find that some cancers, such as lung cancer, have a higher incidence among higher deprivation classes (ie there is a positive correlation). However, others, such as breast cancer, show a negative correlation. Some cancers, such as colorectal cancer, have no clear link (table 2).

Smoking-related cancers
Cancers where smoking is believed to be a risk factor show a strong positive correlation with deprivation class. Analysis of smoker prevalence from General Household Survey data shows the same correlation.
More affluent classes have experienced a much sharper reduction in smoker prevalence over the past 30 years. When considering the impact of this on future cancer trends we need to remember both the cohort effects (for example cigarette smoking) and the impact of temporal effects (such as air pollution and cigarette composition) on cancer incidence.
The male incidence of lung cancer is approximately 40% higher in the general population, ages 4059, than in the insured population. However, the trend since 1982 has been similar across deprivation classes. The risk of lung cancer is associated with prolonged smoking, and we would expect the change in variation in smoker prevalence to emerge over the next 20 years in the lung cancer incidence levels.

Breast cancer
Here we see a higher first incidence in the insured population.
This could be for two reasons: an underlying difference in prevalence, and the impact of early identification through screening. We can reason that an insured population is more aware of the need for regular screening, is better educated, and has better access to screening. Some evidence suggests women in the less affluent deprivation classes are likely to seek treatment at a later stage of disease development.
However, in the graph of incidence by deprivation class (figure 2) we can see this is unlikely to be the only explanation. Before the breast-screening programme was introduced nationally in 1988, there was still a higher incidence in the insured population. The risk of breast cancer is believed to be lower for women who have children at a younger age, women who have several children, and women who breast-feed. From these we can speculate that women in the insured population may have a higher risk than the general population.
This difference is not seen in the mortality statistics: the mortality rates for the two groups have been similar over the past 18 years, and they have also moved in tandem both decreasing by around one-third. This decrease is partly attributed to screening, but also to the improvement in available treatments, in particular the use of tamoxifen.

Prostate cancer
For prostate cancer we also see higher incidence in the insured population. Again, this is most likely because of the increased awareness of symptoms, together with the range of diagnostic tests available. Although there is no formal screening programme, many medicals will now include tests such as a digital rectal examination (DRE) as standard.
As shown in figure 2, increased screening seems to be an important explanation for the higher incidence among the insured classes. However, mortality rates in the insured population have also been higher for much of the period, with a steadily increasing trend in both populations. Risk factors for prostate cancer are still not understood, but from international comparisons there appears to be a greater risk with higher fat consumption.

Colorectal cancer
The final site we will consider is colorectal cancer. Here the main risk factor is believed to be a diet high in animal fat content and low in fibre content. We would expect this risk factor to be lower in the insured population. However, the data show no conclusive variation by deprivation class, and in fact, female insured incidence has been higher in most years since 1992, possibly owing to increased use of screening.
The only major variation in mortality rates is between males and females, with females showing a strong downward trend, but male mortality rates broadly level over the past 18 years.
The significant difference between the mortality of insured lives and the population is well documented, with clear explanations. When pricing CI, the differential is a useful pricing aid in the absence of insured experience, but this study illustrates that the mortality differential must be used with caution. For many illnesses we see a mirror of the mortality differential. However, the dominant reason for this is the respective levels of smoking, already a rating factor for insured lives. For many major illnesses, cancer in particular, the mortality differential may not be applicable, and in fact populations with lower mortality may experience higher CI incidence rates.