OUP user menu

Are health inequalities evident at all ages? An ecological study of English mortality records

Chris Dibben, Frank Popham
DOI: http://dx.doi.org/10.1093/eurpub/cks019 39-45 First published online: 24 March 2012

Abstract

Background: Inequalities in mortality are often presumed to exist at all ages. Here we examine whether this is true. Methods: We conducted an ecological study of mortality in England for 1997–99 using routinely collected public records. We used a (smoothed) single year of age rate of mortality for males and females by the multiple deprivation quintile of their ward of residence, for all and specific causes to assess if inequalities varied by age. Results: For most ages, a greater mortality risk exists for those living in the most deprived compared with the least deprived quintile of wards. However, during late adolescence there is equality. The equalization occurs at the age of 17–19 years. There is a longer period of 10 years for females and 20 years for males of convergence and then divergence centred on late adolescence and young adulthood. The equalization is driven principally through a heightened exposure to the risk of land transport accidents and a resulting higher than expected rate of mortality for the least deprived rather than a decrease in risk for the most deprived. Conclusions: It seems likely that an increase in risk taking associated with late adolescence combined with exposure to a relatively dangerous transport environment leads to large increases in risk for young people in the least deprived parts of England. As a result mortality inequality disappears at this age.

Introduction

While health inequalities between socioeconomic groups have been viewed as reasonably constant across age groups1—a number of authors2–6 have shown that this might not actually be the case in early adolescence (12–14 years of age). It has been hypothesized that during this period, socioeconomic factors are overridden by the influence of ‘youth culture’, school and peer groups.3 Thus, early adolescence is a transition period of ‘equalization’. It is transitory because the class differential re-emerges in later adolescence and early adulthood.

There are a number of other notable features of adolescence that may be important when examining inequalities. It is a period of high-risk taking7,8 when people experiment more with activities that can potentially damage their health than at any other time of their life. This appears as a clear feature on age-specific mortality rates with the resulting rise in deaths described as the ‘accident hump’.9 Although this risk taking has been well documented, its social patterning has been less well explored and importantly, its health consequences by socioeconomic status are not established. Here, we explore whether there is any equalization in adolescence in mortality by socioeconomic deprivation and whether this is because of equalization in risk-associated causes. Earlier work on the question of equalization has tended to focus on morbidity, non-fatal injuries and intentional self-harm rather than mortality10–12 and has produced limited evidence of equalization. When mortality has been examined, deaths due to accidents have sometimes been excluded because, it is argued, they will not be preceded by poor health5,13 and again there has been limited evidence of equalization.14–18 The use of mortality has also been criticized because of its rarity but it is the ultimate negative health outcome and is well measured in contrast to morbidity that is much harder to adequately capture. Recent studies suggest that mortality inequalities are relatively persistent in adolescence6,19–21 but are limited, as is earlier research, in that they group ages together when single ages in adolescence may see changes in risk because they are associated with specific transitions (for example to obtaining a driving license). In this analysis, we use single year of age. Finally, some studies in the UK have only considered a limited number of causes of death,19,20 whereas we extend the analysis to consider all deaths by cause.

Methods

The central mortality rate was calculated by single year of age, sex and deprivation quintile. The 2001 Index of Multiple Deprivation (IMD), the official UK government measure, was used to quantify area deprivation at the ward level (areas of an average 5500 people).22,23 It has six dimensions measuring deprivation in: income, employment, health and disability, education, skills and training, housing and geographical access to services. An overall deprivation score is derived by combining each dimension after standardizing them and weighting according to importance. All 8414 wards in England were then ordered by their overall multiple deprivation score and assigned to deprivation quintiles. The proportion of households in each ward without regular access to a car or van was estimated from the 2001 UK Census.

Ward-level single year of age and gender populations for the years 1997–2000 were estimated by taking 2001 5-year age and sex-bound census counts for census wards. The 5-year total populations were then allotted to single year of age groups using the distribution of single year within 5-year bounds in the local authority district in which the ward was located. The 2001 estimates were then aged back, using a cohort component method, with deaths in each single year of age added to each ward as appropriate to produce estimates for each year between 1997 and 1999.

Mortality data were obtained from the Office of National Statistics for wards across England for the years 1997–99. These data included gender, age of death and cause of death.

The population and mortality data were aggregated to the appropriate quintile of deprivation and mortality was expressed as a rate per 100 000 by age and sex. In order to achieve a better estimate of the underlying risk of death, kernel-weighted local polynomial smoothing was applied to the derived rates to smooth mortality across years of age. We used the lpoly programme within the Stata statistical package to fit a kernel-weighted local polynomial regression with a first degree polynomial and a Gaussian kernel whose bandwidth was chosen to be as small as possible while not introducing ‘noise’ to the resulting smoothed line. A fit for each cause of death was derived along with approximate 95% confidence intervals. For the first year of life, the actual measured mortality value is shown and no smoothed value is calculated. This was because mortality at ‘neighbouring’ ages is not a good predictor of mortality during the first year of life.

The mortality rate was calculated for all-cause mortality and the categories of mortality (major causes ICD9 1–239 390–519; external causes 800-999; other causes 240–389 520–799; land transport accidents E800–E829; drug related 292 304 305.2-9 E850–E858 E980.0–5 E962.0 and intentional self-harm E950–E959), shown, in previous research, to be significant in adolescence and early adulthood.24 For the ease of interpretation, most of the analysis contrasts the least deprived quintile with the most. Results for all quintiles are available from the lead author.

Results

Table 1 shows the characteristics of the wards in each deprivation quintile. The most deprived quintile of wards on average have close to 50% of their population living in households defined as ‘income deprived’ (in receipt of means tested benefits) and have no access to a car or van (i.e. do not own or have regular access). This is in contrast to less than 10% in wards in the least deprived quintile. On the whole, the most deprived wards tended to be geographically closer to services indicating their more urban setting.

View this table:
Table 1

Selected social and demographic characteristics of deprivation quintiles—England 1997–99

Total exposed male population over 3 yearsTotal exposed female population over 3 yearsMean percentage of ward population defined as income deprivedSDMean percentage of ward population in households with no access to a carSDMean scaled access to key services 100 most distant—1 most proximateSD
Most deprived quintile14 289 06614 875 05142845112319
214 219 14315 095 8242842993525
314 192 86715 134 5181941985029
414 211 88315 136 8801331366127
Least deprived quintile14 199 80814 973 42592956023
  • Source: 2001 Census: Standard Area Statistics (England and Wales) and Indices of Deprivation 2000.

Smoothed and unsmoothed all-cause mortality rates are presented in figure 1a and b. For males and females, the mortality rate is relatively high in the first year of life (400–700 deaths per 100 000). This then drops in early childhood, starting to rise beyond the age of 10 in an exponential form (linear on a logarithmic scale). This trend in gradually increasing risk of death is broken for most of the population groups in this study in early life. Starting at the age of 14 years, the risk increases more than would be expected (given the expectation of an exponential growth in the risk with age). Only the mortality curves of females in the most deprived wards show little or no change in gradient. The change in gradient is most distinctive for males and females living in the least deprived quintile, though males in the most deprived quintile show a slightly higher than expected increase in risk. Females in the least deprived quintiles have a risk that increases at the age of 13–14 years but by the age of 25 years, it returns to the level that would have been expected given the general trend in mortality. Males in the least deprived quintile after the relatively large increase in the risk of death at 15–18 then enter a period where their risk does not again increase until their early forties. In contrast, males in the most deprived quintile see a constant increase in risk throughout their twenties and thirties.

Figure 1

Smoothed and unsmoothed rate of mortality for males and females in England 1997–1999. (a and b) Actual and smoothed rates (bandwidth 1.5) with 95% CI. (c and d) Smoothed rates (bandwidth 1.5) for all quintiles. (e and f) Rate ratio for the most deprived quintile: least deprived quintile 95% CI

Figure 1c and d shows the rate of mortality for all quintiles of deprivation. The proportional difference in mortality between the different quintiles of deprivation remains very constant at all ages except during the age range of 15–25 where it converges.

Figure 1d and e shows the ratio of deaths in any year of life, for those living in the most deprived quintile compared with the least deprived. For both genders, the period of late- to mid-teens is a period of equalization. Relative inequalities also decline in the oldest ages (but are still evident) as the absolute rates of mortality increase. At other ages, the level of inequality is fairly constant at about twice the risk of death in females and 2.5 times in males.

The higher than expected increase in mortality, seen in the mid-teens to late-teens in both males and females, is driven largely by ‘external causes’ of death. They rise from 10 to 50 deaths per 100 000 between the ages of 10–20 years (figure 2c and d). Major and other causes of death (figure 2a, b, e and f) remain low during this period at around 10 deaths per 100 000 for all except males in the most deprived quintile. External causes can be broken down into three main components in young people: death due to drug misuse, intentional self-harm and transport accidents. Young people in the most deprived quintile are consistently at greater risk of death due to drug misuse (figure 3c and d). From the age 15 years onwards, both males and females experience a rising risk of death due to intentional self-harm (figure 3e and f). This rise in risk is almost identical whether the person is living in the least or most deprived quintile until the age of 23–24 years. Here risk stabilizes for those living in the least deprived quintile, whereas for those living in the most deprived quintiles, this risk continues to rise in their late twenties, peaking in the early thirties. At this point, for both males and females the risk is about twice that of their contemporaries in the least deprived quintile.

Figure 2

Smoothed rates of deaths grouped into three categories. (a) Male major causes (bandwidth 2.5) with 95% CI. (b) Female major causes (bandwidth 3) with 95% CI. (c) Male external causes (bandwidth 2.5) with 95% CI. (d) Female external causes (bandwidth 3) with 95% CI. (e) Male other causes of death (bandwidth 2.5) with 95% CI. (f) Female other causes of death (bandwidth 3) with 95% CI

Figure 3

Smoothed rates of external causes of deaths shown as three categories. (a) Male land transport causes (bandwidth 2.5) with 95% CI. (b) Female land transport causes (bandwidth 4) with 95% CI. (c) Male drug misuse (including cases coded under mental disorders) (bandwidth 3.5) with 95% CI. (d) Female drug misuse (including cases coded under mental disorders) (bandwidth 4) with 95% CI. (e) Female intentional self-harm (bandwidth 3) with 95% CI. (f) Female intentional self-harm (bandwidth 5) with 95% CI

In contrast, the pattern of deaths due to land transport is very different. Not only do the rates reach a peak at an earlier age, at the age of 20 years, but the relationship with deprivation is also reversed. For both males and females the risk of death is significantly higher for those living in the least deprived quintile. At the age of 20 years, it is some 40% higher than that in the most deprived quintile.

Discussion

Health inequality does appear to vary with age. Late adolescence in particular appears to be a period of mortality equalization, where the large differences in risk that exists across most of the rest of the life course are absent. This equality is driven principally by transport accidents. It occurs almost entirely due to an increase in risk experienced by those living in least deprived small areas in England, not a decrease in risk for those in the most deprived. Because the risk of death from major diseases, such as cancer and heart disease, is relatively low in young ages, external causes of death have a more substantive impact than at other times in the life course. The increase in deaths associated with land transport accidents in the less deprived parts of England, therefore, has a major impact on the pattern of mortality inequality. The rise in other ‘external’ causes of death, particularly due to intentional self-harm, is similar for those living in more and less deprived areas until their early twenties. Beyond the age of 20, a combination of a large increase in the risk of death due to drug misuse and intentional self-harm leads to the re-emergence of the health inequality ‘gap’ that existed before the period of equalization, and this persists; only declining in the highest mortality period of late old age.

Although this study was based on a very large number of cases, well-established methods and datasets, certain methodological weaknesses do exist. The population at risk for each year of age was estimated by distributing 5-year age–sex counts by the age distribution, for the period, of the local authority in which the ward was located and then aged back using information on deaths but not migration. In order to test whether these assumptions might have affected our findings, we re-ran all our analysis using the non-aged back 2001 population structure and tested the local authority distribution method for a similar population for which we had single year of age counts. There were no important differences in the results for the non-aged back population and the distribution method appeared to add only a very small amount of error that was not systematically related to area deprivation. We, therefore, conclude that our findings were robust to the population estimation method used. This study was cross-sectional in design. Studies of this type have been criticized for potentially having a numerator–denominator bias.5 Where social class has been used to measure socioeconomic position, it is difficult to correctly attribute social class to young people because they may not be in work, or if they were in that type of work it may not effectively reflect their socioeconomic position. This problem was avoided by using a small area measure of deprivation. The risk of death over the life course was measured in a period rather than cohort design. This comparison of different groups of people means that the environment that they are living within at the time of death is the same but the individuals are not from the same cohort.

This article identifies a period of life where mortality inequalities appear to be eliminated. This happens not through an improvement for the disadvantaged in society but because of a higher risk of death for the advantaged group that appears to be function of their greater access to cars (table 1).25 If the rate of access to cars measured in the 2001 census were an accurate reflection of the use of cars amongst young people in this study, then the level of risk death due to land transport accidents, associated with true exposure (i.e. a corrected at-risk population denominator), would be very similar to those in the different quintiles. The importance of transport accidents as a cause of mortality amongst young people is highlighted, but this research suggests that a policy that reduced the risk of road accidents amongst young people might actually increase inequalities if there was no action taken to also mitigate other risk factors in more deprived populations. Primary amongst these would be attempting to reduce deaths amongst those using illicit drugs. If the drug related rates are considered in the light of recent research showing that half of deaths amongst young drug misusers26,27 are not coded as drugs related (but instead as due to suicide, violence or other causes), then the scale of mortality risk associated with a drug using milieu for young adults becomes apparent and its importance as a public health issue is highlighted.

It has been argued that the ages of 10–14 is a period where health differences are virtually absent.4 An ‘equalization hypothesis’ is proposed to explain this. This hinges on the transition from childhood to adolescence that occurs around the ages of 11–12,3 a period where school, peer group and youth culture combine to ‘cut across’ socioeconomic divides.2 However, our data show a different pattern—at age 11–12, there is still considerable inequality in mortality risk. The equalization is later. It appears to be most significant beyond the age of 15–16 and continues for males well into their twenties. It seems unlikely, therefore, that an unrefined ‘equalization hypothesis’ can explain our findings. Studies on social class differentials in mortality at different ages in the 1960s and 1970s in England and Wales28 do not show the equalization in risk that we identified in this study. However, given the importance of land transport deaths in the effect identified in this article, much lower access to cars in the 1970s might explain this result. There may, therefore, be a ‘period’ effect. Other studies for different countries have on the whole not found any evidence of equalization in mortality in adolescence.6,21,29 These and earlier studies in the UK5 have frequently used the age range of 15–19 or wider and it may simply be that the effect is hidden by putting different age groups together. Our data show that this choice of age range will hide the mortality equalization because, as can be seen in figure 1e and f, it bridges the period of equalization and will, therefore, average the risk out and disguise the equalization.

Later adolescence, and particularly later male adolescence, is characterized for many by risk seeking.7 This effect may cut across social divisions because it is centrally part of child development, particularly associated with the end of puberty.30 However, in practice, the socioeconomic context within which the young person lives interacts with this heightened risk period to produce divergent mortality risks for the better and less well off. For adolescents in the least deprived parts of the country, their greater access to cars appears to translate into higher risks of mortality. Socialization, or the adoption of the patterns behaviour of those around you, is important.31 However, unlike the claim that the socialization of school, through peer groups and youth culture, leads to equalization by overriding socioeconomic influences,2 we would argue that socialization is context specific. Young people in school, college and work may be mainly interacting with socially similar individuals and groups due to geographical and social polarization. That this does not lead to a reinforcement of mortality inequalities seems to be due to unequal structural factors10 (i.e. access to a car) that make adolescent risk taking for the wealthier more hazardous.

The focus of this study is mortality in early life. In this period, there are relatively few deaths. However, early life death has a great impact on ‘years of life lost’ and, therefore, morally (loss of young people) and statistically (high impact on life expectancy measures) is an important issue. Evidence of equalization of mortality risk in late teenagehood does not mean that there is not a process operating during this period of life that produces later life inequalities in mortality. An individual’s ‘health capital’ can either be strengthened or weakened in this period without an associated increase or decrease in the risk of mortality.11

This article demonstrates mortality equalization in England for a specific time period. It may be the case that this effect is not identifiable in other countries or at other points in time in England. Lower car ownership in the 1970s might explain why studies found no evidence of equalization in mortality at this time.28 Likewise, research in Australia demonstrating significant higher rates of death from motor vehicle accidents in socially disadvantaged areas21 might reflect the different patterns of both poverty and car ownership in Australia. This emphasizes that the equalization in mortality shown in this paper is a function of child development (higher risk taking in adolescence) within a particular social context (a socioeconomic gradient in access to cars). If this social context exists in other countries, then a comparable equalization would be expected.

Acknowledgements

The work for this article was conducted when F.P. was employed as a research fellow at the University of St Andrews.

Conflicts of interest: None declared.

Key points

  • There is evidence of equalization in mortality risk at the ages of 16–18 years for people living in differently deprived areas.

  • This is caused by the ‘accident hump’, and in particular unequal rates of land transport deaths.

  • The ‘accident hump’ is very much more apparent in the mortality curve of the least multiply deprived—especially for women.

  • Between the ages of 20 and 30 years, inequality in mortality re-emerges, driven largely by differences in the risk of drug related and intentional self-harm deaths.

References

View Abstract