Slowdown in Mortality Improvement in the Past Decade: A US/UK Comparison

Abstract

Objectives

To investigate the slowdown in mortality improvement in the United States, United Kingdom, and comparator countries observed in the first decade of the twenty-first century and critically evaluate proposed explanations.

Methods

Change-point analysis to identify the year of change in comparison of national mortality trends and linear spline models in the investigation of subnational differences using data from the Human Mortality Database, Global Burden of Disease cause-specific data, and, for the United Kingdom, national statistics data. Consideration of the impact of using different methods to estimate overall mortality is also concluded together with a review of methodological assumptions made in previous studies.

Results

The results confirm the slowdown in mortality improvement observed in the early twenty-first century but indicate that proposed explanations for this are inadequate on a range of counts.

Discussion

Mortality improvement slowed down in the early twenty-first century but the explanations advanced, such as opioid use in the United States or influenza epidemics and austerity programs in the United Kingdom, seem unlikely to account for this. Further research considering longer-term life course and cohort influences is needed.

Throughout most of the twentieth century, today’s high-income countries experienced average increases in life expectancy at birth of about 3 years per decade. This improvement was initially driven particularly by declines in infant and child mortality, later by declines in mortality at “working ages” and latterly by declines at older ages, where now most deaths occur. There was an interruption to this pattern of improvement in the mid-twentieth century with some stagnation in midlife male mortality. This was interpreted by some as indicating a “ceiling” on life expectancy gains once the prevention or treatment of infectious diseases was largely achieved, leaving a residue of harder to challenge chronic disease (Bourgeois-Pichat, 1952; Ouellette et al., 2014). However, mortality declines subsequently accelerated and more recent interpretations have attributed the prior slight stagnation to health-related experiences of the cohorts involved, including exposure to harmful working conditions, fatty diet, and importantly smoking (Luy & Wegner-Siegmundt, 2015).

The past decade (2010–2019), however, saw substantial reductions in the rate of mortality improvement in the United Kingdom, United States, and most other industrialized countries, prompting intense debate especially in the United States and United Kingdom (a 2018 survey of European statistical offices indicated that recent slowdown in mortality improvement was of limited interest in other European countries; Murphy et al., 2019). The observation that disadvantaged groups appeared to have been most adversely affected (Public Health England [PHE], 2018) fueled debate about widening socioeconomic and geographic inequalities and, in the United Kingdom, the impact of government austerity policies (Hiam et al., 2017a). The causes of death proximally associated with change, or lack of it, in overall death rates are also debated. In the United States, increases in mortality from synthetic/prescription opioid poisoning and other drugs have received particular attention (Case & Deaton, 2020).

The main focus of this article is on the change in mortality improvement between the first and second decades of the twenty-first century, with particular attention to UK/UK differences. As context, we initially take a slightly longer-term view and consider mortality trends since 1990 for a broader group of 19 comparator countries: the main high-income established market economies of Western Europe (15 countries including the United Kingdom), together with the United States, Canada, Japan, and Australia. We also illustrate the impact of choice of summary mortality measure on trends and changes in them. We then consider in more detail trends around the period of mortality slowdown in the United Kingdom (or England), the United States, and selected other high-income countries. The later sections of the article focus on examining specific explanations advanced to account for the mortality slowdown in the United Kingdom; a UK/US comparison of the role of changes in specific causes of death, and differences between more and less deprived areas. Finally, we consider these trends in relation to mortality in the coronavirus disease 2019 (COVID-19) pandemic and concerns about inequalities in both exposure and adverse outcomes.

Method

We present information on mortality trends and differentials using data from national statistical offices available in the Human Mortality Database (HMD) or, for one subanalysis, from the UK Office for National Statistics (ONS), and cause-specific data produced by the Global Burden of Disease (GBD) program. We use change-point analysis to identify the year of change in comparison of national mortality trends and linear spline models in the investigation of subnational differences. In most cases, we use standardized death rates (SDRs) but include a comparison of results using SDRs and life expectancy at birth and consider reasons for differences in trends identified using these measures.

Trends in Mortality 1990–2018

Choice of measure and reference period

There is no dispute that mortality improvement has stalled across high-income countries since around 2010, see Supplementary Appendix Figure A, but assessment of the magnitude of this change depends on the specific indicators used and the period of analysis. Many studies have failed to make a clear distinction between short-term fluctuations and long-term trends; this is important as choice of the base years for any comparison has a considerable impact on conclusions about the extent of change. Choice of mortality indicator is also important.

Studies discussing the slowdown in mortality improvement tend to concentrate on either life expectancy at birth (e₀) or (directly) SDRs, both often disaggregated to assess the contribution of separate sex, age, and cause of death components. Reasons for a particular choice of indicator are rarely elaborated so we initially address this issue.

Both e₀ and SDRs are calculated using age-specific (and usually sex-specific) mortality rates, although SDRs also require a choice of standard population. The weights of the standard population (size of age and sex subgroups) should ideally reflect the distribution of the populations being analyzed. Most European analyses use the 2013 European standard, ESP 2013 (European Commission, 2013), which has a broadly similar distribution to the current European and US populations.

Reasons for differing results using different indicators

Life expectancy at birth, an indicator of survival in a hypothetical population, is intuitively easy to interpret and contextualize. While usually defined as the total life table population person-years divided by the life table radix (the starting number of “newborns” at age 0), it is also the mean age at death in the hypothetical population. The overall life table death rate (LTDR) is the radix of the life table divided by the total person-years experienced, that is, the reciprocal of life expectancy at birth (a corresponding result holds for further life expectancy at other ages). Thus, the LTDR is a weighted average of the life table mortality rates (m_x or µ _x depending on discrete or continuous formulation) with weights of life table L_x or l_x values as appropriate, although the life table indicator is usually presented as its reciprocal, e₀, to aid interpretation. Therefore, both SDRs and life tables provide SDRs as their final outputs. Examples of standard weights in Supplementary Appendix Figure B show that the weights of the European 2013 standard and typical life tables are similar, as would be expected, whereas global standards such as the GBD standard give much lower weight to mortality at older ages.

The main reason for the different trends in Supplementary Appendix Figures A1 and A2 is that the LTDR weights are not fixed, but change due to changes in mortality rates, although in a complex nonlinear way. Life table calculations give more weight to mortality at younger than at older ages. An additional random death at a young age will reduce the total population person-years more than a death at an older age, whereas a death at any age has effectively the same impact on the SDR value. It is anomalous that crude death rates are rejected as useful indicators of underlying mortality trends on grounds that the age-specific weights change, but similar objections are not raised to the use of life table measures.

SDRs and life table estimates therefore give different weights to deaths at different ages. In this article, we concentrate on standardized and age-specific death rates that use an explicit set of weights, are widely available, and make interpretation of cause-specific trends more straightforward (Case & Deaton, 2017, p. 400).

Results

Main Trends in Mortality in Selected High-Income Countries 2000–2019

Table 1 presents results from change-point analyses of SDR trends in 2000–2019 in 19 main Established Market Economies ranked by annual mortality rate change in the period after the changepoint; also shown graphically in Supplementary Appendix Figure C. A slowdown in mortality improvement around 2010 is apparent in all these countries, although the magnitude, statistical significance, and timing varied. The seven countries with the lowest final rates of mortality improvement also had the seven latest changepoints, in 2011 or later. The magnitude of slowdown in these seven countries, measured as the change between the initial and final values, was greater on average than in the other 12 counties and was particularly pronounced in the Netherlands and the United Kingdom, which along with the United States and Canada, also had the lowest subsequent rates of improvement. Results using proportionate rather than absolute changes in SDR showed a similar pattern.

Table 1.

Estimated Changepoints for SDR Mortality Slowdown, 2000 to Latest Available Year

		Annual change per 100,000			Country rank (high to low)
Country	Changepoint	Initial	Final	Difference	Initial	Final	Difference	Significance
United Kingdom	2011.0	−27.4	−4.8	22.6	7	19	17	0
United States	2010.1	−22.0	−6.1	15.9	16	18	11	0
Netherlands	2010.3	−29.9	−6.3	23.5	5	17	18	0
Canada	2012.0	−20.1	−6.7	13.3	18	16	7	0
Spain	2013.0	−20.9	−7.3	13.6	17	15	8	0.08
Portugal	2011.0	−29.7	−9.1	20.6	6	14	15	0.01
Denmark	2014.2	−26.1	−10.1	16.0	9	13	12	0.07
Japan	2001.5	−30.3	−11.1	19.2	4	12	14	0.57
Germany	2006.5	−22.5	−11.4	11.1	15	11	5	0.02
Italy	2006.0	−25.6	−11.5	14.0	12	10	9	0.02
France	2007.2	−26.2	−11.9	14.3	8	9	10	0.01
Austria	2007.1	−25.7	−12.7	13.0	11	8	6	0.01
Switzerland	2006.3	−22.8	−13.0	9.7	14	7	4	0.03
Sweden	2005.8	−19.6	−13.9	5.6	19	6	1	0.08
Australia	2005.0	−23.3	−14.2	9.1	13	5	2	0.06
Finland	2005.7	−35.6	−15.0	20.6	2	4	16	0
Belgium	2007.0	−25.7	−16.2	9.5	10	3	3	0.06
Ireland	2007.6	−46.9	−16.4	30.5	1	2	19	0
Norway	2004.7	−32.9	−16.8	16.2	3	1	13	0.02

Notes: SDR = standardized death rate; HMD = Human Mortality Database. Source: Based on HMD, latest available year is between 2017 and 2000.

The magnitude of this discontinuity reflects the fact that the first decade of the twenty-first century was a period of particularly rapid mortality improvement in many high-income counties. For example, in England and Wales, the 2.3% p.a. SDR improvement rate (using the 2013 European Standard and official data) in 2001–2011 was the highest decadal value since records began in 1841 (using HMD data for earlier periods). (HMD values based on a single year of age data are available for 1841 to 2018. Official data from ONS, available for a much shorter period but up to 2019, show very similar patterns.) By contrast, the rate of improvement of 0.7% p.a. in 2011–2019 was lower than in any decade since 1941–1951.

The implications of this slowdown for both individuals and society partly depend on which groups are particularly affected; this may also provide insights into the plausibility of explanations advanced. We consider below first variations by age and sex and second changes by cause of death.

Trends by sex and age group

Mortality differentials between men and women increased in the first part of the twentieth century, especially linked to earlier and more intensive adoption of smoking among men (Luy & Wegner-Siegmundt, 2015; Waldron, 1986), but there has been a long-term reduction in differentials in the past half century or so, a trend that has continued through to 2019. We compare patterns before and after 2011 to identify whether the gap between women and men was increasing or not, focusing, for ease of presentation, on the United Kingdom, United States, France, and Germany.

Some commentators have claimed that it is “uncontested” that the recent slowdown has had a greater effect on women than on men in Britain and that this holds across high-income countries (Raleigh, 2019). This conclusion appears to be based on the fact that mortality increased slightly in some countries for older women, but not for older men, between 2011 and 2016. Table 2 presents the annual rates of SDR improvement before and after 2011, and the slowdown defined as the difference between these values, for both males and females in four of the largest countries we consider. Improvement values were greater for males than for females in almost all cases in each period, but the slowdown was more substantial for males than females in three cases (although the difference was small for the United Kingdom): a reduction of mortality improvement close to 2% p.a. for both males and females. The change in the annual rate of overall mortality improvement in the United States between periods 2001–2011 and 2011–2019 dropped by 1.0% p.a. for males but only by 0.6% for females. The result was a widening of US sex differentials in mortality for the first decade since the 1970s. When all 19 countries were analyzed, results showed that a male disadvantage was particularly prominent in the seven countries that experienced the lowest recent rates of improvement, six of these experienced a greater slowdown for males than for females.

Table 2.

Annual Rate of SDR Improvement (percent p.a.), Before and After 2011, and Slowdown Rate, by Sex, Selected Countries

	2001–2011			2011–Enda			Slowdown
Country	Male	Female	M:F difference	Male	Female	M:F difference	Male	Female	M:F difference
United States	1.88	1.56	0.32	0.91	0.96	−0.05	0.97	0.60	0.37
United Kingdom	2.73	2.20	0.54	0.66	0.20	0.46	2.08	2.00	0.08
France	2.33	2.17	0.17	1.26	0.69	0.57	1.08	1.48	−0.40
Germany	1.91	1.42	0.49	0.88	0.70	0.18	1.03	0.72	0.31

Notes: SDR = standardized death rate; HMD = Human Mortality Database. Source: Based on HMD.

^aFinal year of series; 2017 Germany, 2018 UK, France; 2019, USA.

The second main variable we consider is age. In this case, some have suggested that the slowdown in improvement affected older people more than other age groups, apart from in the United States, where trends in working-age mortality have been particularly adverse (National Academies of Sciences, Engineering, and Medicine, 2021). In general, improvements in mortality rates have tended to be lower the older the age, this gradient being especially apparent in countries such as France and Germany (Figure 1), apart from lower improvements in the age group 50–69 than at ages 70–84. The most striking exception is the rapid deterioration of mortality trends at younger ages—those younger than age 50—in the United States and to a lesser extent in the United Kingdom, countries where improvement rates at younger ages were already low in the early 2000s. Although deaths younger than age 50 account for only around 6% of all deaths in high-income countries, this will have implications for the future if these trends reflect particular characteristics or experiences of these cohorts.

Figure 1.

Annual age-specific mortality improvement (smoothed) 2000:2018. Source: Human Mortality Database using 2013 ESP.

The United States stands out in that mortality improvement since 2010 among older Americans aged 85 and older (the age group that accounts for about one third of total deaths in the high-income countries) is considerably greater than among older people elsewhere, showing almost no deterioration since 2010. However, there is some evidence that the US advantage is diminishing (Palloni & Yonker, 2016).

While there are some exceptions noted above, the key findings are that there has been some tendency for increasing convergence in annual mortality improvement trends across different age groups; that the slowdown was very widely experienced and that age groups older than age 50 show similar trends in this century. For example, the average correlation coefficient for the values for age groups, 50–69, 70–84, and 85+ with the overall value for the countries shown in Figure 1 were 0.84, 0.96, and 0.79, respectively. This suggests—but does not prove—that common period factors acting across the whole age range where deaths tend to be concentrated were influential. We show results for the complete set of countries in Supplementary Appendix Figure D.

Cause of Death Trends

Increasing, or fluctuating, risks from particular causes of death have been widely discussed in the United Kingdom and the United States as factors possibly underlying recent slowdowns in mortality improvement. In the United States, “deaths of despair,” especially associated with opioid abuse, have been emphasized (Case & Deaton, 2020). In the United Kingdom, seasonal influenza has attracted considerable attention (PHE, 2018). Here we first examine whether changes in influenza mortality can account for the slowdown in mortality improvement in the United Kingdom, as suggested by several analysts, and next compare changes in causes of death in the United Kingdom and the United States in the mortality slowdown period.

Influenza mortality in the United Kingdom

The increase in deaths in the United Kingdom (E and W) between 2014 and 2015 was the largest for nearly 50 years (Hiam et al., 2017b) and simultaneously life expectancy at birth fell between 2014 and 2015 in 20 out of the 28 EU countries. These developments have been attributed partly to an epidemic of the particularly lethal H3N2 strain of influenza in early 2015 (Ho & Hendi, 2018; Jasilionis, 2018). The UK Government review of trends in the 2011–2016 period concluded that the 2015 mortality increase was caused by influenza in the winter months of early 2015, but also that seasonal influenza had had an impact on the slowdown since 2011 (PHE, 2018). However, evidence to support the assertion of a predominant influence of influenza on underlying trends was weak. Excess winter deaths (EWDs) are calculated as the difference between deaths in the December–March period and the average of deaths in the previous and subsequent 4-month periods (ONS, 2020). The “exceptional” 2015 mortality change was mainly due to a very low number of EWDs around early 2014—the lowest ever-recorded in the EWD series—rather than a particularly high number in early 2015, which had been exceeded in 26 of the 63 winters since 1951 (Murphy, 2021). Average excess winter mortality over the whole period 2011–2016 was very low even though representation of the most vulnerable group, older people, increased over the period.

Changing cause of death patterns in the United Kingdom and the United States since 2000

Given that most deaths occur at older ages in which multimorbidity is common, disentangling cause-specific influences is difficult using deaths coded by a single underlying cause. Moreover, there are likely to be interactive influences of changes in particular causes, and risk factors for them, so an overly reductive approach may be misleading.

An additional problem in investigating the role of particular causes, or groups of causes, as drivers of change in overall mortality is that cause of death assignment may be affected by coding changes (ONS, 2014) or other factors not related to actual changes in incidence or case fatality. Most notably there has been a very large recent increase in the United Kingdom, United States, and elsewhere in the number of deaths attributed to Alzheimer’s disease or other forms of dementia. In the United States, for example, deaths attributed to Alzheimer’s disease increased by 71% between 2000 and 2013 (Alzheimer’s Association, 2016) and the increase in Britain was even greater (Murphy, 2021; Perera et al., 2016). As evidence from long-established longitudinal studies indicates that the incidence of dementia has not increased, indeed the reverse, in recent decades (Matthews et al., 2013) this likely reflects changes in certification practices. Some changes in World Health Organization recommendations on coding of chest infections (implemented in IRIS 2013) may have played a minor role in this, but increased awareness of dementia, introduction of screening programs, and campaigns to increase recognition of the condition in coding of cause of death are likely to have been more important (Alzheimer’s Association, 2016). This also has the effect of altering trends in other causes of death because deaths that formerly would have been allocated to another cause are now attributed to dementia. The timing and extent of these shifts may vary between countries so distorting comparisons. In addition, there remain significant national variation in certification and coding by cause of death (Désesquelles et al., 2012). We therefore use cause of death data provided by the GBD program, rather than data from national statistical offices, for cross-national analysis of changes in cause of death since 2000. GBD data, available up to 2019, are based on official statistics but adjusted to improve consistency across time and space (GBD 2019 Risk Factors Collaborators, 2020). In particular, these data do not show the major increase in deaths from Alzheimer’s disease and other dementias recorded by national vital statistics systems. There are limitations with GBD data; for example, they cannot be used for making very short-term comparisons, but for the long-term comparative analysis presented here they are preferred, for reasons discussed above.

GBD age-standardized mortality rates (based on the ESP 2013 standard) were improving by 17 per 100,000 p.a. in the United States and 24 in the United Kingdom in 2001–2011, but values fell to around 1 in the period 2011–2019, so the reduction of improvement after 2011 was 16 per 100,000 p.a. in the United States and 23 per 100,000 p.a. in the United Kingdom. In Table 3, we present information on the causes of death responsible for these changes in overall annual SDRs between these periods. We focus on causes (or closely linked sets of causes) that include sufficient numbers of deaths to affect overall trends between 2001–2011 and 2011–2019 by presenting data only for causes where the absolute change was at least 0.4 per 100,000. These causes account for virtually all the total change. The increase in US opioid-related deaths recorded in GBD statistics was not sufficient on its own to be included with our chosen cutoff but because of interest in this specific cause and in violent deaths, deaths from opioid use and other injuries are also shown.

Table 3.

Causes of Death Responsible for Age-Standardized Mortality Improvement: the United States and the United Kingdom 2001–2011 and 2011–2019

	Average annual rate per 100,000
	United States		United Kingdom		Difference 2001–2011 and 2011–2019		Distribution (%) of difference
Cause	2001–2011	2011–2019	2001–2011	2011–2019	United States	United Kingdom	United States	United Kingdom
Lower respiratory infections	1.0	0.0	3.1	−0.5	−1.0	−3.6	6.1	15.3
Neoplasms	3.2	0.8	2.7	−0.4	−2.4	−3.0	14.9	12.8
Tracheal, bronchus, and lung cancer	1.3	0.8	0.6	0.3	−0.5	−0.4	2.9	1.6
Breast cancer	0.4	0.1	0.7	0.1	−0.3	−0.6	1.7	2.4
Colon and rectum cancer	0.6	0.1	0.6	−0.1	−0.6	−0.7	3.7	3.0
All other neoplasms	0.9	−0.2	0.8	−0.6	−1.0	−1.4	6.6	5.9
Urinary tract infections	0.2	0.0	−0.3	0.1	−0.2	0.4	1.1	−1.7
Cardiovascular diseases	12.6	1.3	17.6	1.8	−11.3	−15.8	71.9	67.4
Ischemic heart disease	9.7	1.3	11.7	1.3	−8.4	−10.4	53.0	44.3
Stroke	2.4	0.1	4.8	0.6	−2.3	−4.1	14.7	17.7
Ischemic stroke	1.9	0.1	4.1	0.6	−1.9	−3.6	11.8	15.3
Intracerebral hemorrhage	0.4	0.0	0.5	0.0	−0.4	−0.4	2.6	1.9
Aortic aneurysm	0.3	0.0	0.7	0.1	−0.3	−0.6	2.0	2.6
All other cardiovascular diseases	−2.2	−0.3	−4.2	−0.8	1.9	3.3	−12.2	−14.2
Chronic obstructive pulmonary disease	0.0	0.1	0.4	−0.1	0.1	−0.5	−0.4	2.0
Drug use disorders	−0.5	−1.2	0.0	0.0	−0.6	0.0	4.1	0.0
Opioid use disorders	−0.4	−0.8	0.0	0.0	−0.4	0.0	2.5	0.0
All other drug use disorders	−0.1	−0.4	0.0	0.0	−0.3	0.0	1.6	0.0
Diabetes and kidney diseases	0.2	0.0	0.3	−0.2	−0.2	−0.4	1.0	1.8
Diabetes mellitus type 2	0.9	0.2	0.4	0.0	−0.7	−0.4	4.6	1.6
Chronic kidney disease	−0.8	−0.2	−0.1	−0.1	0.6	0.0	−3.8	0.1
All other diabetes and kidney diseases	0.0	0.0	0.0	0.0	0.0	0.0	0.3	0.1
Injuries	0.4	0.1	0.4	0.0	−0.3	−0.5	1.9	2.0
Transport injuries	0.5	0.1	0.3	0.1	−0.4	−0.2	2.7	0.9
All other Injuries	−0.1	0.0	0.2	−0.1	0.1	−0.3	−0.8	1.1
All causes	16.9	1.2	24.1	0.7	−15.8	−23.4	100	100
All causes included above	17.0	1.1	24.1	0.8	−15.8	−23.3	100.6	99.6
All causes not included above	0.0	0.1	0.0	−0.1	0.1	0.1	−0.6	0.4

Note: Source: Global Burden of Disease. Main categories shown in italic.

The final two columns of Table 3 show that less favorable trends in cardiovascular disease (CVD) mortality were the major reason for the slowdown, accounting for about 70% of the total change, with ischemic heart disease alone responsible for two thirds of this, even though CVD accounted for just over one third of total mortality. In the United Kingdom, CVD mortality declined by over one third in the first period, but by only 4% in the second. The slowdown in reducing cancer deaths accounted for just under 15% of the total, about half of the remaining overall change, even though cancers account for over one quarter of total mortality. Lower respiratory infections were the only other causes accounting for a substantial proportion of the slowdown, especially in the United Kingdom (about 15%). These adverse trends in particular causes could be offset by beneficial trends in others, but apart from improvements in mortality from chronic kidney diseases in the United States and a large number of small individual components of the “All other cardiovascular diseases” group, there were no examples of causes with recent mortality improvements sufficient to make a significant contribution to overall trends.

Deaths directly related to opioid abuse accounted for only 2.5% of the US slowdown using GBD data (although alternative estimates exist, see Ho, 2019) and below 0.5% in the United Kingdom. This is not to dismiss the importance of the sharp and substantial increases in mortality from this cause in the United States, because it is an indicator of other deep-rooted problems that manifest themselves across a broad range of other causes of death, including CVD.

We conclude that causes that have received considerable attention, seasonal influenza and substance abuse, were insufficient to directly account for more than a small fraction of the slowdown in mortality improvement since 2011. While this slowdown was not attributable to a small number of causes, CVD has clearly played a major role (Lopez & Adair, 2019; Mehta et al., 2020; OECD/The King’s Fund, 2020).

Social and Economic Explanations for Slowdown in Mortality Improvement in the United Kingdom and the United States

Area deprivation and mortality in the United States and England 2000–2019

In both the United Kingdom and the United States, rising levels of inequality have been proposed as an underlying reason for mortality slowdown (Hiam et al., 2017a, 2017b, 2018; Marmot et al., 2020). We examine this by comparing trends in mortality by age group and major cause group in areas classified as being above or below the median level of deprivation. In the United Kingdom, separate area deprivation indices are produced for each constituent country so we focus on England and use the 2015 Index of Multiple Deprivation (Department for Communities and Local Government, 2015) and the Census Bureau classification for the United States (Glassman, 2019). One advantage of areal-level analysis is that estimates by age and cause of death are available in GBD for 150 local authority areas in England and for 51 US states up to 2019. The questions we concentrate on are, as before, whether the period of slowdown from 2011 onward was associated with a change in trend and the extent to which British and US patterns were similar. For both countries, we present analyses for the whole population by comparing those living in areas above or below the median area deprivation score (Figure 2A–D). The size of units considered and the construction of indices differ so that they cannot be used to address the question of whether there is more or less inequality in the United States than in England, but comparisons can be made by age and cause of death between those living in more and less advantaged areas within the two countries over time.

Figure 2.

Ratio (%) mortality by areal deprivation indices: (A) United States: more to less deprived states, by age; (B) England: more to less deprived local authorities, by age; (C) United States: more to less deprived states, by cause; (D) England: more to less deprived local authorities, by cause; (E) England: linear spline model by sex, 2000–2018. Sources: (A–D) Global Burden of Disease and (E) Office for National Statistics.

Variations by age group

In both England and the United States, elevated mortality associated with living in a disadvantaged area was least for those aged 75 and older with the largest difference among individuals aged 45–64. US differentials diminished steadily at ages younger than 45 and in the 45- to 54-year-old age group until 2015. The data for England show little change in differentials since 2011, although a small increase in 2015–2019 in age groups younger than 55 is apparent. The small increase in inequality for those aged 75 and older in England contrasts with the United States, where differentials are tending to reduce and even diminish at ages 85 and older. However, overall, there is no evidence of a clear increase or decrease in these mortality differentials in either country.

Variations by cause

We present results for the main chapters in the GBD classification, together with opioid misuse deaths, which are a subset of substance abuse deaths (opioid deaths are too rare for useful analysis in England). Cause of death trends and differentials show similarities between countries. In both cases, the most substantial differentials are in substance use deaths, and in the United States, especially opioid-related deaths. However, whereas the differential between more and less deprived areas has increased in the United States, in England it has declined. In England, chronic obstructive pulmonary disease (COPD) deaths also show a large differential by level of deprivation, and COPD differentials have increased in both countries. However, apart from these few exceptions, cause-specific mortality differentials have remained largely constant.

These analyses suggest that the period of mortality slowdown improvement has not been associated with generalized increases in socioeconomic area differentials.

For England only we undertook some additional analysis, using a more granular classification of area deprivation and data from the ONS, in order to examine specifically the argument that Government austerity policies following the 2008 banking crisis (Hiam et al., 2017a, 2017b, 2018; Marmot et al., 2020) underlay the slowdown in mortality improvement. These explanations are not mutually exclusive as it is likely that there is an interaction between life course vulnerability and the impact of policy changes.

Differentials in mortality by area deprivation in England

A widely cited finding in support of the argument for increasing inequalities in mortality in Britain is that life expectancy for females in the most deprived areas has been particularly affected (PHE, 2018; Marmot et al., 2020). As before, we compare age-standardized mortality differentials by sex between those in areas of below- and above-average area-level deprivation (Figure 2E), concentrating on the change in mortality following 2011. We fit linear spline models to rates of change in the four populations (based on logarithms of SDRs). Results show that the annual reduction of improvement following the 2011 slowdown was almost identical, just over 2% p.a., for males in both more and less deprived areas and for females in more deprived areas; females in the least deprived areas experienced the highest slowdown of 2.2% p.a. Period life expectancy declined and SDRs increased for more deprived females because this group had the lowest rate of improvement before 2011 and was becoming more detached from the other groups in that period. This trend has regrettably continued since 2011 at a similar pace. In the period of generalized slowdown in mortality improvement since 2011, the previous increase in inequalities in mortality that occurred in a time of rapid overall mortality improvement has continued, but there is no clear evidence for overall deterioration linked to the slowdown (although it is likely that some subgroups will have different experiences). These changes are, however, small when compared with those arising from the COVID-19 pandemic. Between 2018 and the 14-month period March 2000 to April 2021 that included the first two waves of the pandemic, in England, male SDR rose by 17%, 3% more than the value for females. In addition, the mortality increase was about 3% greater in more deprived than in less deprived areas.

Summary and Conclusions

The sharp downturn in mortality improvement in the past decade in many high-income countries was unexpected, especially as it followed a period when mortality had often been improving at historically high rates. The UN 2010 population projections assumed US life expectancy in 2010–2020 would increase by the same amount as observed in the previous decade (United Nations, Department of Economic and Social Affairs, Population Division, 2011). Reasons for such abrupt transitions have usually been clearly identified as due to specific major events such as wars, epidemics (HIV/AIDS in sub-Saharan Africa), or societal collapse as in the Soviet Union around 1990 (Murphy et al., 2019). No such explanation for the recent downturn in mortality improvement in North America and Western Europe is evident. Although some commentators have pointed to increasing inequality and to austerity policies, cross-national evidence in support of these arguments is weak. Multiple factors are likely to be involved, but it has been easier to reject some of the initially proposed explanations than to provide robust evidence in support of these or any other hypotheses. Recent mortality trends have likely been influenced by the interaction of historical factors and contemporary events. For example, higher rates of obesity in the United States and the United Kingdom may have made the populations more vulnerable to the recent adverse factors, such as widespread availability of opioids (US) or reductions in health budgets (UK), and contributed to their greater downturn in mortality improvement (Adams et al., 2006), but the relatively small number of annual observations and a large number of potential explanations and confounders make identification of likely causes challenging (Murphy et al., 2019).

The sharp and apparently well-defined reversal of previous trends in some countries, notably the United Kingdom and the United States, around 2010 would not be consistent with the major drivers of CVD trends as these are usually related to cumulative lifetime experiences of diet, obesity, smoking, etc. Innovations in treatment or prevention (including the rollout of statins in the early 2000s) could plausibly lead to periods of high improvement; potential mechanisms responsible for a sharp drop in rates of improvement are less obvious, but not unknown. For example, there was a substantial increase in CVD mortality in some countries of the former Soviet Union in the early 1990s subsequently linked to extreme alcohol abuse including prolonged binge drinking and use of surrogate alcohol beverages (McKee & Britton, 1998; Murphy et al., 2019).

Lack of clear understanding about the drivers of recent trends means that future trends are less predictable, especially because the emergence of COVID-19 adds a considerably greater additional layer of uncertainty. The period just before 2020 showed some recovery in mortality trends. The lowest SDR value in each of the 19 countries in Table 1 to date occurred in the period 2017–2019. Life expectancy increased in the United States for a second year in 2019, although it still remains below the level of 2014. Year-to-year changes are sensitive to short-term fluctuations due to seasonal respiratory mortality, but improvements were also evident in nonwinter months that are much less affected. Whether this would have marked the point at which mortality rates started to improve at historically typical levels of between 1% and 2% a year in the absence of the pandemic remains speculative, but the answer to this question may become more important if the infection can be controlled and the long-term health impacts of COVID-19 are limited on both those infected and on the wider population. In that case, the same question about the determinants of mortality will remain.

Funding

Funding for this research was provided by the UK Economic and Social Research Council (grant numbers ES/T014083/1, ES/L009153/1). This paper was published as part of a supplement sponsored by the University of Michigan with support from the National Institute on Aging (P30AG012846).

Conflict of Interest

None declared.

Author Contributions

M. J. Murphy was responsible for analyses presented here. M. J. Murphy and E. M. D. Grundy were equally responsible for drafting and revising the text.