Background: Low socioeconomic position is consistently associated with increased risk of premature death. The aim of this study is to measure the aggregate scale of inequality in premature mortality for the whole population of England. Methods: We used mortality records from the UK Office for National Statistics to study all 2 465 285 premature deaths (defined as those before age 75 years) in England between Jan 1, 2003, and Dec 31, 2018. Socioeconomic position was defined using deciles of the Index of Multiple Deprivation: a measure of neighbourhood income, employment, education levels, crime, health, availability of services, and local environment. We calculated the number of expected deaths by applying mortality in the least deprived decile to other deciles, within the strata of age, sex, and time. The mortality attributable to socioeconomic inequality was defined as the difference between the observed and expected deaths. We also used life table modelling to estimate years-of-life lost attributable to socioeconomic inequality. Findings: 35·6% (95% CI 35·3–35·9) of premature deaths were attributable to socioeconomic inequality, equating to 877 082 deaths, or one every 10 min. The biggest contributors were ischaemic heart disease (152 171 excess deaths), respiratory cancers (111 083) and chronic obstructive pulmonary disease (83 593). The most unequal causes of death were tuberculosis, opioid use, HIV, psychoactive drugs use, viral hepatitis, and obesity, each with more than two-thirds attributable to inequality. Inequality was greater among men and peaked in early childhood and at age 40–49 years. The proportion of deaths attributable to inequality increased during the study period, particularly for women, because mortality rates among the most deprived women (excluding cardiovascular diseases) plateaued, and for some diseases increased. A mean of 14·4 months of life before age 75 years are lost due to socioeconomic inequality. Interpretation: One in three premature deaths are attributable to socioeconomic inequality, making this our most important public health challenge. Interventions that address upstream determinants of health should be prioritised. Funding: National Institute of Health Research; Wellcome Trust.
Bibliographical noteFunding Information:
We did a cross-sectional study of 2·5 million premature deaths in England and found that one in three was attributable to neighbourhood deprivation measured by upstream determinants of health including income, employment, education, and crime. Inequality was greatest for respiratory, cardiovascular, and infectious diseases, whereas much less inequality was seen for cancers, except for cancers of the lung and mouth. Premature mortality reduced substantially during the study period but inequality persisted and increased among women because mortality due to non-cardiovascular causes plateaued or increased among the most deprived women. Our study used data from the whole population of England to provide a population-level estimate of the scale of socioeconomic inequality in premature mortality. The method that we used to calculate MASI is replicable in other settings and can be used to make international comparisons and observe changes in inequality over time. Our data can also be broken down by disease category and geographical area to inform allocation of public health resources according to the absolute scale of inequality. By using data from the whole population, the method avoids selection bias. The large proportion of premature deaths that could be avoided by eliminating socioeconomic differences in mortality is likely to exist across high-income countries, because inequalities in premature mortality are ubiquitous. Smaller cross-sectional studies (none in the UK, to our knowledge) have previously reported the proportion of premature deaths attributable to area-based deprivation and found similar values to our headline MASI value (36%), with values ranging from 22% to 42%. 5,18–22 The proportion is likely to be greater in countries with more extreme socioeconomic inequality, such as the USA, Spain, and Australia. Our cause-specific results might not be generalisable to other countries because the contribution of diseases to inequalities in mortality can differ substantially even where total inequalities are similar. 23 We aimed to measure the extent of inequality in premature mortality rather than a causal effect of deprivation on mortality. The method is not designed to show that either inequality in isolation (which might affect health through the psychological stress of low socioeconomic position, for example) or any element of the IMD (such as the quality of local education and living environments) have caused premature deaths; or that improvements in the indicators that comprise the IMD would lead to fewer premature deaths. Instead, the study shows a scenario in which everyone in England has the same mortality of the least deprived decile, without showing how that scenario would be achieved. As such, the approach is designed to show the scale rather than the causes of inequality. Nonetheless, the results of this study can help to understand some of the relations between deprivation and health. The IMD measures upstream determinants of health, and their effect on mortality is likely to be mediated by factors such as tobacco smoking, harmful alcohol consumption, hypertension, and obesity. We observed the greatest inequality for diseases with large behavioural components, such as lung cancer and liver disease, suggesting that health behaviours are important mediating pathways. Other studies have estimated the proportion of premature deaths that can be attributed to these intermediate factors. For example, a meta-analysis 11 of cohort studies in seven high-income countries estimated that an average of 29% of premature deaths were attributable to smoking. People might smoke because they live in deprived circumstances and therefore some of these deaths could also be attributed to deprivation. These attributable fractions therefore overlap to some extent, and if fractions are calculated for several variables in the same causal pathway, the total could be more than 100%. Our results also provide insight into changing inequality over time. Cardiovascular mortality reduced radically during the study period, which is part of a longer-term trend in the UK. 24 Although all-cause premature mortality reduced across deprivation groups, higher socioeconomic groups had greater proportional reductions and therefore MASI increased ( table 1 ). By contrast, absolute reductions in mortality were greatest for the most deprived groups, particularly for men. However, among the most deprived women, improvements in mortality were limited outside of cardiovascular diseases. For other diseases, such as COPD, mortality among the most deprived women actually increased. Premature mortality due to lung cancer among the most deprived men reduced between 2003–06 and 2015–18, while staying constant for the most deprived women ( appendix pp 12, 13 ). This finding might relate to differences in the history of the male and female smoking epidemics, with smoking prevalence peaking earlier and reducing faster among men such that the prevalences converged between 1960 and 1990. 25 Although we are not able to fully explain why time trends in inequality differed for men and women, the results show that plateauing or worsening premature mortality for non-cardiovascular diseases among the most deprived women underlies the large increase in MASI for women. Understanding and addressing this trend should be a public health priority. Our analysis accounted for differences in age and sex between deprivation groups but we were unable to adjust for ethnicity because this characteristic is not recorded on death certificates in England. Ethnicity might be a confounder for diseases that have strong genetic or migration-related risk factors that are also associated with ethnicity. This confounder might be the case for tuberculosis, for example, where first generation migrants from countries with high incidence of tuberculosis are more likely to live in deprived areas. Analysis of census-linked data, such as the Census Longitudinal Study in the UK, 26 could be used to assess the role of ethnicity in MASI. Our analysis also does not account for selective internal migration. This factor might cause indicators to be understated if people move to wealthier areas when older or in poor health, or overstated if people move to poorer areas at these life stages, and could undermine sub-national results. Existing evidence suggests that people in poor health are more likely to move to deprived areas. Net internal immigration of people in poor health has been observed in particular in poor coastal towns in England. 27 However, internal migration appears to play a minor role in social gradients in health. 28,29 The scale of premature mortality associated with neighbourhood deprivation shows the importance of addressing upstream determinants of health. Many mechanisms through which socioeconomic deprivation might harm health have been proposed, including inability to buy adequate food, housing, or health care; coping with behaviours such as alcohol consumption and smoking; differing cultural norms relating to healthy and unhealthy behaviours; stress and feelings of worthlessness associated with low socioeconomic position, which can lead to harmful physiological changes; lower social capital in deprived communities; environmental factors such as busy, polluting roads, fast food outlets, and waste disposal sites; poor prenatal and early childhood conditions causing poor health in adulthood; and social selection—a form of reverse causality in which sickness causes poverty. 30 The relative importance of these mechanisms is debated and is beyond the scope of this study. The wide inequality in causes of death related to behavioural risk factors, and particularly alcohol, tobacco, and illicit drugs, highlights the importance of interventions that help people live healthier lifestyles. Evidence suggests that public health interventions such as smoking cessation that aim to change individual behaviour, although effective, can increase health inequalities as they are more likely to be adopted by people of higher socioeconomic position. 31 Structural interventions such as taxation and minimum unit pricing are likely to have a more progressive effect. 32 In the UK in 2018, disposable income for the richest quintile was 5·2 times the poorest quintile. 33 Fiscal and welfare policies that enable or exacerbate this inequality are perhaps the highest priority targets to address the 877 082 excess deaths that we observed between 2003 and 2018. As well as addressing income inequality, the Marmot Review of health inequalities in 2010 34 identified six policy objectives: giving every child the best start in life; enabling people to maximise their capabilities and have control; creating employment and good work; ensuring a healthy standard of living; creating healthy and sustainable places; and strengthening prevention of ill health. Our results show the high inequality in mortality among infants and young children and therefore the importance of objectives relating to early life. Every year tens of thousands of premature deaths in England alone could be avoided by closing socioeconomic inequalities in mortality. Cardiovascular mortality has reduced but stalling of improvements for other disease categories, particularly for women living in deprived areas, needs further investigation. Local public health action and national policies need to prioritise interventions that address upstream determinants of these inequalities. This online publication has been corrected. The corrected version first appeared at thelancet.com/public-health on January 3, 2020 Contributors The study was conceived by AH. DL and WJ did a literature search. AH and DL created the analysis plan. DL did data analysis. All authors contributed to drafting, revision, and approval of the Article. Declaration of interests We declare no competing interests. Acknowledgments DL is funded by National Institute for Health Research (NIHR; DRF-2018-11-ST2-016 ). RWA is supported by a Wellcome Trust Clinical Research Career Development Fellowship ( 206602/Z/17/Z ). AH is an NIHR Senior Investigator. CE is funded by an NIHR Clinical Doctoral Research Fellowship ( ICA-CDRF-2017-03-006 ). The views expressed in this Article are those of the authors and not necessarily those of the NIHR, or the Department of Health and Social Care. This work was produced using statistical data from Office for National Statistics (ONS). The use of the ONS statistical data in this work does not imply the endorsement of the ONS in relation to the interpretation or analysis of the statistical data. This work uses research datasets which might not exactly reproduce National Statistics aggregates.
© 2020 The Author(s). Published by Elsevier Ltd. This is an Open Access article under the CC BY 4.0 license