Definitions and methods

This section provides definitions for commonly used measures of mortality and then goes on to describes the analyses carried out by Murphy et al.

Common measures of mortality

Measuring mortality presents a complex challenge. Not only is mortality itself affected by so many different interconnecting factors, but researchers have several different measures at their disposal to understand mortality. Each one can shed light on a different aspect of the overall picture of what is happening in the population.

The conclusion drawn can be influenced not only by the choice of measure but also how the change is expressed. For this reason, it is important to understand the meaning, implications and potential limitations of each measure, to be sure that the measures selected are appropriate to the question being asked.

When assessing the impact of mortality change, the figures can be looked at in one of two ways:

  • absolute change – the difference in an indicator between two points in time
  • relative change – the absolute change relative to the size of the initial value (change expressed as a percentage of the initial value).

This section sets out the indicators that are most often used when looking at life expectancy and related issues.

Counting deaths in a population

Number of deaths

This is a count of the number of deaths in a population (or population group) in a given period of time. It does not reflect the size or demographic composition of the population in any way. In the UK, the vast majority of deaths occur in older age, so the total number of deaths in a population will be strongly driven by these figures (Figure 4).

Crude death rate

This is a measure of the number of deaths per year in a given population relative to the size of the total population. It is usually expressed as deaths per 1,000 people. It adjusts only for overall size, and takes no account of the age structure of the population. This means that (as with the number of deaths, above) in the UK it is strongly driven by the older age groups, as this is the age at which most people die (Figure 4).

Figure 4: Distribution of deaths by age and sex: England and Wales, 2018

Source: ONS, Deaths registered in England and Wales, 2018.

Comparing mortality rates

As overall mortality in a population is strongly influenced by the age and sex structure of the population (Figure 4), these influences need to be understood and removed to allow appropriate comparisons between times or different populations or groups. This is done by calculating mortality rates. Two types of mortality rate are referred to in this report:

  • Age-specific mortality rate

This is the total number of deaths per year, expressed as per 100,000 population within a given age or age group, most usually expressed for men and women separately. Age-specific rates provide understanding of the different mortality experiences of different age groups, and can be used to explore how this changes over time or between different populations.

  • Age-standardised mortality rate

As the crude death rate is strongly influenced by the age distribution of a population, age-standardised mortality rates are commonly used to allow comparison between whole populations in different areas or time periods, taking account of differences in size and age distribution of those populations.

The age-standardised mortality rate is calculated as a weighted average of the age-specific mortality rates – in other words, the average is scaled to take away the effect of differing age and sex structures in different populations. This scaling is done to a fixed standard hypothetical population (such as the European Standard Population) and enables comparison of whether and how mortality would differ in two or more geographical areas or timepoints if they had the same age and sex make-up. Without this adjustment, one population could appear to have much higher overall mortality rates than another with comparable age-specific mortality rates at all ages, simply due to having a higher proportion of older adults.

Survival rates

This term describes the proportion of people in a group who are alive after a given period of time.

Life expectancy

Life expectancy is a statistical measure of the average time someone is expected to live, based on the mortality rate they subsequently experience. Life expectancy and age-standardised mortality rates are both summary measures, using the same data (age and sex-specific mortality rates). So, they are alternative ways of summarising mortality rates in a given period, and they mirror each other.

Life expectancy is used for setting the state pension age, and setting and assessing health policy, among other purposes. It is often presented at one of two ages:

  • at birth – how long someone born in a given year and place is expected to live
  • at age 65 – how long someone at this age in a given year and place might have left to live.

The UK’s life expectancy projections (both period and cohort) are currently produced by the Office for National Statistics. Life expectancy is calculated using a life table. This shows, for each age, the probability that a person will die before their next birthday. There are two types of life expectancy measure: period and cohort (see the box). They differ based only on changes in mortality over time, as explained below, and would only ever be identical if there were no changes to age-specific mortality rates over time, which is highly unlikely.

Period life expectancy

Period life expectancy represents the average number of years a person would be expected to live, based on the assumption that their likelihood of death at each age throughout life is the same as for the population at a given point in time. There is no account taken of possible future improvements in mortality at each age between cohorts.

So, period life expectancies are useful summary measures for the entire population of a given place and time. They provide an objective way of comparing trends in mortality over time, or between different populations and population subgroups. (This is how they are used in this report.)

But period measures are less useful for predicting lifespans of current or future cohorts. And, like age-standardised mortality rates, changes are influenced by deaths in older people (who are the majority of people who die). Period-based projections of life expectancy are based on assumptions about future changes in mortality and reflect the mortality rates of the entire population in each year.

Period life expectancy may also be distorted by ‘tempo effects’: a bias that may arise if period statistics (such as period life expectancy) are interpreted as a reflection of current mortality patterns, when mortality is changing during an observation period. The impact of tempo effects in the recent patterns in mortality is debated, but explored in detail in by Murphy et al.

Cohort life expectancy

In this measure, ‘cohort’ means a group of people born at the same point in time. Cohort life expectancy takes into account age-specific probabilities of death for the specific cohort calculated from observed mortality data, where available, for that cohort. It then combines it with mortality rate projections for the cohort in future years, using assumptions based on historic trends.

This can be used to estimate how much longer a person of a given age and sex, in a given place, would be expected to live. However, these measures tend to rely more heavily on assumptions about the future and predictions are unlikely to be correct.

Historic revisions in life expectancy

Historic estimates of both period and cohort life expectancy have been successively revised as actual changes in mortality have been measured. For example, around 1970 improvements in mortality were close to zero. This led some experts at the time to assume that the highest possible life expectancy had been achieved. This is reflected in the low improvements assumed in the 1971-based life expectancy projections (Figure 4).

We now know that this was not the case. In fact, mortality started to improve at a generally increasing pace throughout the rest of the 20th century. Later commentaries suggested this slowdown in improvements around early 1970 was mainly due to stalling of mortality improvements in cardiovascular disease, especially among older men, including those of higher working age. This may have been due to a combination of higher-fat diets, more sedentary lives and heavy smoking.

Throughout the 2000s, when mortality was improving at a faster than expected pace, cohort life expectancy tended to be underestimated. Conversely, since 2010 – when the slowdown began – it has been overestimated. These overestimations have led to subsequent downward revisions in how much longer future cohorts are expected to live, although their lifespan is still expected to be longer than that of previous cohorts.

Figure 5 illustrates historic estimates of period life expectancy, and how these have been revised as actual changes in mortality have been measured over time.

Figure 5: Successive projections of period life expectancy at birth, males: UK, 1966–2030

Source: ONS National Population Projections Accuracy Report underlying data, UK, 1966 to 2030; expectation of life, principal projection, United Kingdom, 2012-based, 2014-based and 2018-based.

Healthy life expectancy

Healthy life expectancy is an estimate of the average number of years someone would live in a state of ‘good’ general health, usually expressed at birth or at age 65. Healthy life expectancy adds a ‘quality of life’ dimension to estimates of life expectancy, dividing it into time spent in different states of health. Health status estimates are self-assessed, based on respondents’ answers to a survey question asking ‘How is your health in general?’.

This report focuses on life expectancy, but healthy life expectancy – and inequalities in this – are of critical importance as measures of population health. Inequalities in healthy life expectancy are wider than in life expectancy, meaning that with increasing levels of deprivation, people are living a greater proportion of shorter lives in poor health. This is the focus of other Health Foundation work.

Having described the most commonly used measures of mortality, we go on to look at the new analyses of mortality undertaken in the research by Murphy et al.

Murphy et al. carried out a comprehensive review of previous research into the recent trends in mortality in the UK, covering what had previously been published on the trends and potential drivers of these trends.

This review identified areas where further research would build understanding of the trends and potential drivers, and new analyses were carried out. These analyses primarily used data from the Human Mortality Database – an open-access resource providing detailed and consistent population and mortality data for 40 countries or areas. This enables long-term and international comparisons of mortality. For full detail of the research see the full research report.,

Analyses included the following:

Exploration of mortality trends in the UK

To comprehensively describe the trends in UK mortality rates and life expectancy, the report looks at these in the context of a longer-term picture, as well as in more detail in recent periods.

  • Long-term changes in age-standardised mortality rates and period life expectancy in the UK overall and constituent countries, by sex, from 1950 to 2016. In order to understand current trends, this is a relevant period to examine, beginning after the rapid transition in leading causes and ages of death that occurred in the early part of the 20th century (Figure 2), and to exclude the first and second world wars.
  • Changes in age-standardised mortality rates and period life expectancy over the 21st century, by sex, from 2000 to 2016. These analyses focused on the 21st century, 2000–2016, with the aim of better understanding the change in trend during this period. Further analyses looked at 2006–2016 to examine the decade around the change in trend and explore the differential impact of the slowdown in different age/sex groups. (This is the period used in most analyses to date.)

International comparisons of mortality trends

To provide additional insight, the report explores differences and similarities in mortality trends between the UK and other countries with similar socioeconomic structures and mortality levels.

For cross-national European comparisons, the authors mainly used the set of high-income countries largely located in western Europe that were considered to provide the most appropriate comparators for the UK. These countries are the European countries defined as ‘developed economies’ according to the MCSI Index.

They included the USA due to considerable interest in recent mortality trends there, and Australia and Japan as examples of high-income countries in other parts of the globe. The HMD provides mortality data separately for England and Wales, Scotland and Northern Ireland, treated in a consistent way with the other countries included. The following analyses were carried out:

  • Comparison of life expectancy at birth in the UK and selected comparable countries, by sex, from 2000–2016 The ranking and change in ranking over the period were looked at, plus the UK position for men and women relative to the EU average.
  • Changes in age-standardised mortality rates in selected comparable countries, by sex, from 2000–2016 Remaining with this period of particular interest during which the changes have occurred in mortality trends, these analyses were to identify similarities and differences between countries, and the time point at which trends changed.
  • Changes in age-standardised mortality rates by age for the UK and closest comparable countries (Netherlands and France), 2000–2016 To further understand similarities and differences by age group, annual percentage changes in age-standardised mortality rates for age groups were plotted.

Further analyses were then carried out to explore some of the possible causal factors identified in the literature review – particularly focusing on the UK, the Netherlands and France. These included austerity-related and other measures, including:

  • gross domestic product (GDP), government spending and spending on health care as a percentage of GDP
  • spending on health care per capita
  • self-assessed health
  • obesity prevalence
  • age-standardised mortality rates for CVD and non-CVD causes.

Other contributing factors explored included the positive contribution of the ‘golden cohort’ – born between 1925–1935 – to mortality trends, and also the possible role of tempo effects (see definition above).

Previous Next