Research and analysis

Methods, data and definitions

Published 11 September 2018

Absolute and relative measures of inequality

Both absolute and relative measures of inequality are presented in this report.

Absolute inequality shows the magnitude of difference between subgroups of the population (most simply calculated by subtracting the value for one group from another), whereas relative inequality shows the proportional difference between subgroups (most simply calculated by dividing the value for one group by another).

For example, if 30% of people smoke in Group A and 20% smoke in Group B then the absolute inequality between them is 10 percentage points and the relative inequality is 1.5, therefore the prevalence of smoking in Group A is 1.5 times higher than Group B.

Both absolute and relative measures are important indicators of inequality. However, they can lead to differing conclusions about the direction of change in inequality over time, depending on the trend in the indicator overall. Each measure has advantages and disadvantages but used together they can provide a more complete picture of inequality.

Adult Psychiatric Morbidity Survey (APMS)

The Adult Psychiatric Morbidity Survey (APMS) series provides data on the prevalence of both treated and untreated psychiatric disorders in the English adult population (aged 16 and over). The survey has been carried out on four occasions, in 1993, 2000, 2007 and 2014. The most recent survey was conducted by NatCen Social Research, in collaboration with the University of Leicester, for NHS Digital.

All the APMS surveys have used largely consistent methods and in 2014 the sample size was around 7,500 adults.

APMS assesses psychiatric morbidity using actual diagnostic criteria for a range of disorders: common mental disorders, post-traumatic stress disorder, psychotic disorder, autism spectrum disorder, personality disorder, attention-deficit/hyperactivity disorder, bipolar disorder, alcohol dependence, drug use and dependence, suicidal thoughts, suicide attempts and self-harm, and comorbidity in mental and physical illness.

Age-standardised mortality rates

Age-standardised rates adjust for differences in the age structure of populations and allow comparisons to be made between geographical areas and through time^{[footnote 1]}. The direct method applied across this report uses the age-standardised rate for a particular condition which would have occurred if the observed age-specific rates for the condition had applied in a given standard population. The European Standard population 2013 is used.

Age-standardised years of life lost (YLL)

Age-standardised years of life lost (YLL) is a measure of premature death within a group of people. YLL are calculated by comparing the age at death with the aspirational life expectancy at that age. YLL are the years that a person who died at a certain age ideally would have lived according to the standard life table. For all countries and time periods, the same ideal standard life expectancy is applied. This allows results to be compared between locations and over time.

Breastfeeding at 6 to 8 weeks

This is the percentage of infants that are totally or partially breastfed at age 6 to 8 weeks. Totally breastfed is defined as infants who are exclusively receiving breast milk at 6 to 8 weeks of age - that is, they are not receiving formula milk, any other liquids or food. Partially breastfed is defined as infants who are currently receiving breast milk at 6 to 8 weeks of age and who are also receiving formula milk or any other liquids or food. Not at all breastfed is defined as infants who are not currently receiving any breast milk at 6 to 8 weeks of age.

To calculate the prevalence of breastfeeding, the numerator is the count of the number of infants recorded as being totally breastfed at 6 to 8 weeks and the number of infants recorded as being partially breastfed. The denominator is the total number of infants due a 6 to 8 weeks check.

The indicator is based on observation and is therefore susceptible to measurement bias. Infants whose breastfeeding status at 6 to 8 weeks after birth is unknown are included in the denominator. This will result in an underestimate of the percentage of infants who are breastfeeding.

Dental decay in children

This is defined as children who are not free from obvious dental decay (having one or more teeth that were decayed to dentinal level, extracted or filled because of caries).

Data for individual local authorities are not included if the authority did not take part in the survey or if the number of children examined was too small (less than 30) for a robust estimate.

Deprivation deciles

Deprivation deciles have been constructed using the Index of Multiple Deprivation (IMD) scores at lower super output area (LSOA) level where possible, and if not at district and unitary (UA) authority level or county and UA authority level. The choice of decile methodology is clearly stated in the text. Unless otherwise specified in the text, Index of Multiple Deprivation 2015 (IMD2015) scores have been used.

LSOAs are small geographic areas produced by the ONS to enable reporting of small area statistics in England and Wales. There are 32,844 LSOAs in England, each having a population of approximately 1,500.

To create LSOA deprivation deciles for use in chart presentation and also in the calculation of the slope and relative index of inequality measures, LSOAs within England were ranked from most to least deprived and then organised into ten categories with approximately equal numbers of LSOAs in each.

Since the total number of LSOAs in England is not exactly divisible by ten, the ‘extra’ LSOAs were allocated to deprivation deciles using a systematic method outlined in PHE’s Technical Guide: Assigning Deprivation Categories.

To create district and UA or county and UA deprivation deciles these local authorities were ranked from most to least deprived within England and then organised into ten categories with approximately equal numbers of local authorities in each. Further information can be found in PHE’s Technical Guide: Assigning Deprivation Categories.

Excess weight in children

The data presented for these indicators only includes children participating in the National Child Measurement Programme (NCMP) in state maintained schools. Any measurements taken at independent and special schools are excluded from the analysis.

Deprivation deciles are derived from the postcode of child residency; only children with valid geographical coding (postcode of residence) have been included in this analysis.

England totals include all children in state-maintained schools, with a valid height and weight measurement, including those with an unknown residency.

Children are classified as overweight (including obese) if their BMI is on or above the 85th centile of the British 1990 growth reference (UK90) according to age and sex.

Fentanyl

A powerful synthetic opioid analgesic used to treat or manage severe pain. Fentanyl can also be manufactured illegally and be misused as a recreational drug.

Forecasting

All forecasts presented in the report as PHE forecasts were produced by the expert system built into the forecasting software ForecastPro. For time series that are sufficiently long (over 20 years), the system uses either the Box-Jenkins method to fit autoregressive integrated moving average (ARIMA) models^{[footnote 2]} or Exponential Smoothing^{[footnote 3]} to fit a log-linear trend. For shorter time series it uses Exponential Smoothing. The specifications of the models are given in Table 1.

The 95% confidence intervals for the forecasts reflect the level of uncertainty around the forecasts. Uncertainty arises from two factors in the historic time series: 1) the amount of fluctuation from year to year, which will tend to result in broad confidence intervals, and 2) the consistency of any trend – any apparent changes in trend, particularly in recent years, which will tend to give rapidly widening funnel-shaped confidence intervals, as it is not clear whether the trend will be up, down or level in the future.

Table 1: forecasting model specifications

Indicator	Historic series	Model specification
Male life expectancy at birth	1981 to 2017	ARIMA (1,1,2)	AR: 0.9916, MA (1): 1.1924, MA (2): –0.4744
Female life expectancy at birth	1981 to 2017	Holt’s exponential smoothing	Smoothing weights: Level: 0.3514, Trend: 0.7448
Male all cause mortality	2001 to 2016	Holt’s exponential smoothing	Smoothing weights: Level: 0.5106, Trend: 0.6350
Female all cause mortality	2001 to 2016	Holt’s exponential smoothing	Smoothing weights: Level: 0.6052, Trend: 0.09732
Male CHD mortality*	2001 to 2016	Holt’s exponential smoothing	Smoothing weights: Level: 0.8991, Trend: 0.1175
Female CHD mortality*	2001 to 2016	Holt’s exponential smoothing	Smoothing weights: Level: 0.8079, Trend: 0.1270
Male stroke mortality*	2001 to 2016	Holt’s exponential smoothing	Smoothing weights: Level: 0.8196, Trend: 0.1138
Female stroke mortality*	2001 to 2016	Holt’s exponential smoothing	Smoothing weights: Level: 0.1259, Trend: 0.7705
Male dementia and Alzheimer’s mortality*	2001 to 2016	Holt’s exponential smoothing	Smoothing weights: Level: 0.8673, Trend: 0.2105
Female dementia and Alzheimer’s mortality*	2001 to 2016	Holt’s exponential smoothing	Smoothing weights: Level: 0.2550, Trend: 0.9998
Smoking	2010 to 2017	Holt’s exponential smoothing	Smoothing weights: Level: 0.6059, Trend: 1
High blood pressure	2010 to 2016	Holt’s exponential smoothing	Smoothing weights: Level: 0.3253, Trend: 0.2643
Obesity	2007 to 2016	Holt’s exponential smoothing	Smoothing weights: Level: 0.1485, Trend: 1
Alcohol consumption	2010 to 2016	Simple exponential smoothing	Smoothing weight: Level: 6.189×10^–8
Smoking at time of delivery	2006 to 2017	Holt’s exponential smoothing	Smoothing weights: Level: 0.8037, Trend: 0.1369
Teenage pregnancies	1998 to 2015	Holt’s exponential smoothing	Smoothing weights: Level: 0.8399, Trend: 0.9046
Excess weight in 10/11 year-olds	2006 to 2016	Holt’s exponential smoothing	Smoothing weights: Level: 0.6894, Trend: 0.09752

Note: AR refers to the AR part of autoregressive integrated moving average (ARIMA) models and MA refers to the MA part. For more information see here.

*cause-specific mortality forecasts are not included in the report, but are included in the accompanying data pack.

Good level of development in children

Children defined as having reached a good level of development at the end of the Early Years Foundation Stage (EYFS) as a percentage of all eligible children. Children are defined as having reached a good level of development if they achieve at least the expected level in the early learning goals in the prime areas of learning (personal, social and emotional development; physical development; and communication and language) and the early learning goals in the specific areas of mathematics and literacy.

This measure only includes pupils with a valid result for every achievement scale.

Global Burden of Disease study (GBD)

The Global Burden of Disease study (GBD) collects data from over 80,000 data sources from countries across the world. These are used as inputs for the GBD modelling methodology to produce comparative estimates of death and disability.

The GBD groups diseases into a 4 level cause hierarchy. Level one groups diseases at a high level: communicable diseases, non-communicable diseases, and injuries. Each subsequent level in the hierarchy presents a finer grouping of conditions that nest within the level above. For example, the second level consists of major disease or injury groups, such as musculoskeletal conditions or mental and substance use disorders. The third level, which is used for most of the analysis presented in Chapter 3, further subdivides causes into disease or injury type such as osteoarthritis, low back and neck pain, and depressive disorders.

More information on the disease hierarchy can be found in Appendix Table 3 of the GBD website.

GP Patient Survey (GPPS)

The GP Patient Survey is an independent, England wide survey run by Ipsos MORI on behalf of NHS England. The survey is annual and is sent to over a million people, providing practice-level data about patients’ experiences. The latest version saw just over 800,000 people respond. It provides data at practice level using a consistent methodology, which means it is comparable across organisations and over time. It is a long running survey, with a large sample size.

Healthy life expectancy

Healthy life expectancy at birth is an estimate of the average number of years that would be lived in a state of ‘good general health’ by babies born in a given time period, given mortality levels at each age and the level of good health at each age for that time period.

Similarly, healthy life expectancy at age 65 is an estimate of the average number of remaining years lived in ‘good general health’ from age 65, given mortality levels and the level of good health at each age beyond 65 for that time period.

The healthy life expectancy measure adds a ‘quality of life’ dimension to estimates of life expectancy by dividing it into time spent in different states of health. Health status estimates for England are based on the following survey question: ‘How is your health in general; would you say it was… very good, good, fair, bad, or very bad?’ If a respondent answered ‘very good’ or ‘good’ they were classified as having ‘good’ health. Those who answered ‘fair’, ‘bad’, or ‘very bad’ were classified as having ‘not good’ health and equate to those in ’poor’ health in the reported figures for England.

There are known limitations to data on self-assessed health status. We know that people give subjective answers to the question used to determine health status. Responses are influenced by an individual’s expectations and there are measurable differences across sociodemographic factors such as age, sex, and deprivation.

There is also likely to be a bias arising from the way respondents are selected to take part in the survey. The data are based on surveys that are not able to select people for interview who are living in institutional accommodation (for example, care homes). This may lead to an underestimate of the level of poor health.

Poor health in the EU data is defined by responses to a question in the EU Statistics on Income and Living Conditions survey (EU-SILC) which asks people: ‘How is your health in general? Is it … very good, good, fair, bad, or very bad?’ If a respondent answered ‘very good’, ‘good’ or ‘fair’ they were classified as having ‘good’ health. Those who answered ‘bad’, or ‘very bad’ were classified as having ‘not good’ health.

The EU data, therefore, uses a different definition of ‘not good’ health to that used by ONS for England estimates. In the EU data, people who say their health is ‘fair’ are regarded as having ‘good’ health but in the ONS data people who say their health is ‘fair’ are defined as having ‘not good’ health. This difference explains the very different figures quoted in Chapter 1 for the proportion of life lived in poor health in England (Table 1) and the UK (Table 2).

The EU indicator of healthy life expectancy is calculated using the same method for all countries, but Eurostat caution that the way the survey question was worded by different EU members might hamper cross-country comparisons. In addition, it has long been known that there are cultural differences in the self-reporting of health status which are difficult to resolve and are likely to affect the comparability of data for healthy life expectancy and years in poor health between countries.

Index of Multiple Deprivation (IMD)

The Index of Multiple Deprivation (IMD) is a measure of relative deprivation for small areas. It is one element of the English Indices of Deprivation released by the Ministry of Housing, Communities and Local Government.

Life expectancy

Life expectancy at birth is a summary measure of the population. It represents the average number of years that would be lived by babies born in a given time period according to the mortality rates for that time period. Similarly, life expectancy at age 65 is the average number of remaining years of life that a person aged 65 years would have according to the mortality levels at each age for that time period.

Life expectancy decomposition method and interpretation

The contribution of different age bands or causes of death to changes in life expectancy over time (due to changes in age or cause specific death rates) can be calculated using a method of ‘life expectancy decomposition’. In this report, the Arriaga III method has been used, as described by Ponnapalli^{[footnote 4]}. The method is based on a life table divided into 5-year age groups. The contributions of each age group are then distributed into causes of death using a method described by Preston and others^{[footnote 5]}. Contributions are distributed proportionately according to the difference in mortality between time periods by cause of death within each age group.

Contributions to changes in life expectancy over time show the amount that life expectancy has increased in the later time period due to changes in the mortality rate since the earlier time period in a given age group or cause of death, assuming all other rates remained constant. Contributions that increased life expectancy (that is, where mortality rate has reduced over time) have a positive value, while contributions that offset the life expectancy increase (that is, where mortality rate has increased over time) have a negative value.

The same decomposition method can also be used to assess the contribution of different age bands or causes of death to differences (or the gap) between areas with different levels of deprivation.

Contributions to the gap show the amount that life expectancy would increase in the most deprived area if its mortality rate for a given age group or cause of death was changed to that of the least deprived area, assuming all other rates remained constant. Contributions that widen the inequality gap (that is, where mortality rate is higher in the most deprived area) are represented with a positive value, while contributions that offset the gap (that is, where mortality rate is higher in the least deprived area) are represented with a negative value.

Life expectancy projections

ONS produces regular projections of life expectancy which are derived by estimating long term trends in mortality improvement and projecting them forward for future decades. A summary of the latest results are available from ONS.

Besides their Principal Projections of life expectancy, ONS also produce high and low life expectancy variants, which respectively assume more and less improvement in future mortality rates.

Low birthweight at term

This indicator is defined as live births with a recorded birth weight under 2,500g and a gestational age of at least 37 complete weeks as a percentage of all live births with recorded birth weight and a gestational age of at least 37 complete weeks.

The ONS has linked birth registrations with NHS birth notification records to allow reporting by gestational age and birth weight. With 99.4% of records linked successfully, completeness of this dataset is very good. However, not all births are recorded with a valid birth weight and gestational age. There may be regional variations in the completeness of these fields.

Lower Super Output Areas (LSOAs)

These are small geographic areas produced by the ONS to enable reporting of small area statistics in England and Wales. There are 32,844 LSOAs in England, each having a population of approximately 1,500.

Mean deviation

This is a measure of inequality. Where the dimension of inequality being considered contains a number of population groups which cannot be logically ordered, such as indicators by ethnic group, a summary measure called the mean deviation has been presented.

The measure shows the average of the absolute differences between each of the groups and a reference group. Values are treated as positive whether they are higher or lower than the reference group. The largest group is selected as the reference group for each indicator.

Minimum Income Standard (MIS)

‘The Minimum Income Standard (MIS) is the income that people need in order to reach a minimum socially acceptable standard of living in the United Kingdom today, based on what members of the public think. It is calculated by specifying baskets of goods and services required by different types of household in order to meet these needs and to participate in society. The research entails a sequence of detailed deliberations by groups of members of the public, informed by expert knowledge where needed. The groups work to the following definition: ‘A minimum standard of living in the UK today includes, but is more than just, food, clothes and shelter. It is about having what you need in order to have the opportunities and choices necessary to participate in society.’^{[footnote 6]}

Data is published in the source document, accompanied by text: “All rights reserved. Reproduction of this report by photocopying or electronic means for non-commercial purposes is permitted. Otherwise, no part of this report may be reproduced, adapted, stored in a retrieval system or transmitted by any means, electronic, mechanical, photocopying, or otherwise without the prior written permission of the Joseph Rowntree Foundation.” © Loughborough University 2017.

Neurological conditions

In this report, the term relates to deaths registered in England in adults aged 20 and over with a mention (either as the underlying cause of death or as a contributory cause of death) of at least one of the 473 ICD-10 codes used to define adult neurological conditions. The major condition groups and ICD-10 codes are detailed in Table 2.

Table 2: major condition groups and ICD-10 codes for neurological conditions

Condition	ICD-10 Code
Epilepsy	G400, G401, G402, G403, G404, G405, G406, G407, G408, G409, G410, G411, G412, G418, G419, R568
Motor neurone disease and spinal muscular atrophy	G120, G121, G122, G128, G129
Multiple sclerosis and inflammatory disorders	G35X, G360, G361, G368, G369, G370, G371, G372, G373, G374, G375, G378, G379
Neuromuscular diseases	G700, G701, G702, G708, G709, G710, G711, G712, G713, G718, G719, G720, G721, G722, G723, G724, G728, G729, G730, G731, G732, G733, G734, G735, G736, G737, M600, M601, M602, M608, M609, M620, M621, M622, M623, M624, M625, M626, M628, M629
Parkinsonism and other extrapyramidal disorders/tic disorder	R251, F950, F951, F952, F958, F959, G903, G10X, G20X, Q210, G211, G212, G213, G214, G218, G219, G22X, G230, G231, G232, G238, G239, G240, G241, G242, G243, G244, G245, G248, G249, G250, G251, G252, G253, G254, G255, G256, G258, G259
Traumatic brain and spine injury	S04, S060, S061, SO62, S063, S064, SO65, S066, S067, S068, S069, S141, S142, S143, S144, S240, S241, S242, S341, S342, S343, S344, T060, T061, T093, T094
Tumours of the nervous system	C700, C701, C709, C710, C711, C712, C713, C714, C715, C716, C717, C718, C719, C728, C729, D320, D321, D322, D330, D331, D332, D333, D334, D337, D339

Population projections

Population projections are produced by the ONS. They provide an indication of the future size and age structure of the UK and its constituent countries based on a set of assumptions of future fertility, mortality and migration. For each country, ONS produce Principal Projections, plus a number of variant projections based on alternative scenarios. Details of the methods used in their calculation are available from ONS.

From the Principal Projections, ONS makes available the estimated numbers of deaths for both calendar years and mid-year periods.

One of the variant projections assumes there will be no improvement to mortality rates in the future and that age- or sex-specific mortality rates will remain constant at the values assumed for the first year (mid-2016 to mid-2017) of the Principal Projection.

From this variant of the projections, ONS makes available the estimated numbers of deaths in each mid-year period in England, assuming no improvement in mortality rates.

Proportion of life spent in poor health

The total number of years lived in poor health divided by the total number of years lived (life expectancy), expressed as a percentage.

Quality of life score (EQ-5D)

Quality of life has been calculated using the EQ-5D score. EQ-5D is a standardised instrument for measuring health status. Health status is converted to an index score by the GP Patient Survey team from responses to Q34 on the GP Patient’s Survey, which asks respondents to describe their health status using the 5 dimensions of the EQ-5D survey instrument: mobility, self care, usual activities, pain/discomfort, and anxiety or depression.

EQ-5D can be used to assess whether health related quality of life is changing over time while controlling for potential measurable confounders (such as age, sex, long term conditions, caring responsibility). A higher score indicates a better quality of life with the best possible score being one.

Relative index of inequality (RII)

This is a measure of inequality. The relative index of inequality (RII) is a measure of the social gradient in an indicator and shows how much the indicator varies with deprivation (by deprivation decile). It takes account of inequalities across the whole range of deprivation within England and summarises this into a single number.

The relative index of inequality (RII) represents the proportional difference in the indicator across the social gradient from most to least deprived.

Where the relationship between the indicator and deprivation is not linear, an adapted version of the relative index of inequality has been used. For these indicators, the indicator values for the 10 deprivation deciles have been log transformed prior to calculation of the relative index of inequality.

Serious mental illness (SMI)

This is measured by recording the number of people in contact with secondary mental health services for specific conditions. SMI include schizophrenia-spectrum disorders, severe bipolar disorder, and severe major depression.

Slope index of inequality (SII)

This is a measure of inequality. The slope index of inequality (SII) is a measure of the social gradient in an indicator and shows how much the indicator varies with deprivation (by deprivation decile). It takes account of inequalities across the whole range of deprivation within England and summarises this into a single number. The measure assumes a linear relationship between the indicator and deprivation.

The slope index of inequality (SII) represents the absolute difference in the indicator across the social gradient from most to least deprived.

A more detailed description of the methodology used to calculate the SII can be found in the PHOF overarching indicators technical user guide.

Smoking at time of booking

This is measured by the number of mothers recorded to be smokers at the time of booking as a percentage of mothers where the smoking status is recorded.

Smoking status at the time of delivery

This is measured by the number of mothers known to be smokers at the time of delivery, as a percentage of all maternities where the smoking status is known. A maternity is defined as a pregnant woman who gives birth to one or more live or stillborn babies of at least 24 weeks gestation, where the baby is delivered by either a midwife or doctor at home or in an NHS hospital.

The indicator is based on observation and is therefore susceptible to measurement bias.

Smoking prevalence

This is defined as the percentage of adults aged 18 and over. The England average figure presented in Chapter 5 for smoking prevalence differs from the England figure presented in Chapter 3 and in the Public Health Outcomes Framework. The data are based on a household survey in which the interviewer can ask questions about household members not present at the time of interview (proxy responses). However, the sexual orientation question is not asked by proxy, and therefore the England average presented in Figure 11, Chapter 5 excludes proxy responses to the smoking question.

Teenage conceptions

This is defined as conceptions in women aged under 18 per 1,000 females aged 15 to 17. Conceptions are defined as the number of pregnancies that occur in women aged under 18 and result in either one or more live or stillbirths or a legal abortion under the Abortion Act 1967.

The date of conception is estimated using recorded gestation for abortions and stillbirths, and assuming 38 weeks gestation for live births. A woman’s age at conception is calculated as the number of complete years between her date of birth and the date she conceived. The postcode of the woman’s address at time of birth or abortion is used to determine geographical area of residence at time of conception. Only about 5% of under 18 conceptions are to girls aged 14 or under and to include younger age groups in the base population would produce misleading results. The 15 to 17 age group is effectively treated as the population at risk.

Years lived in poor health

The average number of years lived in poor health is the average life expectancy minus the average healthy life expectancy (number of years lived in good health). An increase in the average number of years lived in poor health is often referred to as ‘expansion of morbidity’, whereas a reduction is referred to as ‘compression of morbidity’^{[footnote 7]}.

Years lived with disability (YLD)

Years lived with disability (YLD) is a measure that summarises levels of poor health and disability in a given population. It combines the prevalence of each disease with a rating of how disabling the disease state is.

For a particular year, age and sex, the YLDs associated with a given disease or condition are calculated in 3 steps:

• the consequences associated with the disease are identified (known as the ‘sequelae’ for the disease)

• the prevalence of each of the associated sequelae is multiplied by a ‘disability weight’.

Disability weights are a measure of the disability a person perceives when in a particular health state. Each health state is matched to a sequel, so the disability weight represents the magnitude of health loss associated with that sequela. The weights are measured on a scale from 0 to 1, where 0 equals a state of full health and 1 equals death.

• these figures for each sequela associated with the disease or condition are then summed to give an overall estimate of the morbidity associated with the disease

This means, for example, that conditions with a low perceived disability but high prevalence are comparable with conditions with a low prevalence and high disability, in terms of overall loss of quality of life. The YLD measure is used throughout Chapter 3 to describe the non-fatal burden and is referred to as morbidity. Better health is associated with fewer years lived in disability.

References

1. ONS (2012) Age-standardised rates. Accessed 14 May 2018. ↩

2. Wikipedia (2018) Autoregressive integrated moving average. Accessed 14 August 2018. ↩

3. Wikipedia (2018) Exponential smoothing. Accessed 14 August 2018. ↩

4. Ponnapalli K (2005) A comparison of different methods for decomposition of changes in expectation of life at birth and differentials in life expectancy at birth. Demographic Research 12:141 to 172. ↩

5. Preston S, Heuveline P, Guillot M (2000) Demography: Measuring and Modelling Population Processes. Blackwell Publishing. ↩

6. Padley M, Valadez Martinez L and Hirsch D (2017) Households below a Minimum Income Standard: 2008/09 to 2015/16. Joseph Rowntree Foundation. Accessed 27 April 2018. ↩

7. Fries JF (1980) Ageing, natural death, and the compression of morbidity. N Engl J Medicine 303(3):130 to 135. ↩