RESEARCH REPORT Geographical variation in life expectancy at birth in England and Wales is largely explained by deprivation Laura M Woods, Bernard Rachet, Michael Riga, Noell Stone, Anjali Shah, Michel P Coleman . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . See end of article for authors��� affiliations . . . . . . . . . . . . . . . . . . . . . . . Correspondence to: Mrs L M Woods, Non- communicable Disease Epidemiology Unit, London School of Hygiene and Tropical Medicine, Keppel Street, London WC1E 7HT, UK laura.woods@ lshtm.ac.uk Accepted for publication 15 October 2004 . . . . . . . . . . . . . . . . . . . . . . . J Epidemiol Community Health 2005 59:115���120. doi: 10.1136/jech.2003.013003 Study objective: To describe the population mortality profile of England and Wales by deprivation and in each government office region (GOR) during 1998, and to quantify the influence of geography and deprivation in determining life expectancy. Design: Construction of life tables describing age specific mortality rates and life expectancy at birth from death registrations and estimated population counts. Life tables were created for (a) quintiles of income deprivation based on the income domain score of the index of multiple deprivation 2000, (b) each GOR and Wales, and (c) every combination of deprivation and geography. Setting: England and Wales. Patients/participants: Residents of England and Wales, 1998. Main results: Life expectancy at birth varies with deprivation quintile and is highest in the most affluent groups. The differences are mainly attributable to differences in mortality rates under 75 years of age. Regional life expectancies display a clear north-south gradient. Linear regression analysis shows that deprivation explains most of the geographical variation in life expectancy. Conclusions: Geographical patterns of life expectancy identified within these data for England and Wales in 1998 are mainly attributable to variations in deprivation status as defined by the IMD 2000 income domain score. Mbeen aterial poverty is the most important single determi- nant of life expectancy in Europe.1 This relation has shown in both developed and developing nations2 3 and is the focus of current health policy in the UK.4 A number of studies have shown that mortality also varies regionally in England,3 5 6 Europe,7���12 and North America,13���15 although no study has analysed the effect of deprivation and geography simultaneously. This analysis describes the patterns of age specific mortality and life expectancy in England and Wales in 1998 both by ecologically defined deprivation category and by geographical region, and quantifies the role of deprivation in explaining geographical variability in mortality. METHODS Age and period specific mortality rates were calculated and period life tables for 1998 constructed to determine life expectancy at birth. Life expectancy at birth, used in recent health policy,4 provides a useful summary of the contempor- ary age and sex specific mortality experience of any single defined population that is independent of the population age structure. It is more responsive to changes in young and middle age mortality than rate or ratio measures, because the gain in potential years of life from a reduction in mortality under, say, 60 years of age is proportionately greater than the change in the mortality rate. The two basic requirements for life table construction are the number of deaths by age and sex for a given year or period, (numerator) and an estimate of the mid-year population by age and sex (denominator). In this study the data were also classified by electoral ward (England) or electoral division (Wales) to enable construction of life tables for both (a) the nine government office regions (GOR) of England and Wales and (b) five groups of the population defined by level of material deprivation in the ward or electoral division. Data sources A number of different data sources were used (table 1). Registered deaths for 1997, 1998, and 1999 in each ward of England and Wales were obtained from the Office for National Statistics. We excluded 0.35% of male deaths and 0.20% of female deaths because of missing geographical data. Annual average numbers of deaths registered over the three year period were calculated by age group, sex, and electoral ward or electoral division. For England, the denominators for the mortality rates were derived from estimates of the electoral ward populations in three broad age bands (0���15, 16���59, and 60 years or more, both sexes combined) for 1998. These were provided by the Social Disadvantage Research Group (SDRG) at Oxford University, which apportioned ONS mid-1998 population estimates for each local authority to its constituent wards on the basis of national administrative statistics, including child benefit records and the electoral register (mean ward population 5883, standard deviation 4213). The SDRG estimates were checked for consistency against local data sources by the local authorities themselves. Full details of this procedure are reported elsewhere.16 We then used local authority populations from the 2001 census to estimate the age-sex profile of each constituent ward.17 These were the most detailed age-sex data available at the time of analysis for a year close to 1998. Each ward retained its total population count from the 1998 estimates for each age band, but within each of the three broad age bands, the 2001 age- sex structure of the local authority containing that ward was applied. This enabled us to estimate the population of each ward by sex, in the conventional age groups: under 1 year, 1 to 4 years, 5 to 9 years ��� 85 to 89 years, and 90 years or more. For Wales, the Local Government Data Unit provided counts of patients registered with general practitioners by Abbreviations: GOR, government office region IMD, index of multiple deprivation 115 www.jech.com

single year of age and sex from the National Health Service Administrative Register (NHSAR) for each electoral division (mean population 3471, standard deviation 2755). On a national basis, defining deprivation at the individual level is seldom possible. This study uses an ecological measure of deprivation as a proxy of the deprivation level of the person, as in previous work.3 12 19 The most readily available area based deprivation measure for 1998 electoral wards was the index of multiple deprivation (IMD 2000), which consists of a series of indices referring to six domains of life.16 An equivalent set of indices has been produced for electoral divisions in Wales.18 The income domain, which defines the proportion of people in receipt of any one of five different means tested benefits (income support, job seekers��� allowance, family credit, disability working allowance, and council tax benefit) was selected as the measure of deprivation, because it attempts to quantify material wealth (as against housing, educational, or health deprivation) and is thus similar to measures of material deprivation used elsewhere.20���22 Although the fifth component of the English and Welsh income domain scores are not strictly identical, the misclassification introduced into quintiles of the score by using the Welsh IMD in conjunction with the English IMD was judged to be minimal (Dibben C, personal communica- tion). A full description of how the indices were constructed can be found elsewhere.16 18 The 8444 electoral wards of England and the 865 electoral divisions of Wales were pooled and sorted in rank order according to their income domain score. The 9309 wards were then divided into five groups numbered 1 to 5, representing the most affluent 20% of wards* through to the most deprived 20%. Life table construction Life tables containing observed age and sex specific mortality rates were then constructed for each region�� i, for each deprivation quintile j, and for every combination ij of region and deprivation. These sets of death rates were smoothed and extended into complete (single year of age) life tables up to 100 years of age using a model life table method that compares observed mortality rates in the index population to that of a known mortality ������standard������ of high quality. We incorporated in this approach the overall mortality, the balance between young age and old age mortality, and the pattern observed among the young.23 The Government Actuary���s Department English Life Table for 1991 was used Table 1 Data sources used to derive deprivation and region specific life tables, England and Wales, 1998 England Wales Registered deaths* Average annual number of deaths by single year of age, sex and electoral ward or electoral division for 1997���1999 obtained directly from the Office for National Statistics. Population counts Ward level population estimates in three broad age bands: under 16, 16���64, and 65 years and over for 1998.16 Counts of patients registered with general practitioners in 1998 from the National Health Service Administrative Register, by single year of age, sex, and electoral division obtained directly from the Local Government Data Unit for Wales. Age and sex structure Proportion of population under 1 year, 1���4 years, and five year age groups up to 90 and over for each local authority in England from the 2001 census.17 Deprivation measure The income domain score of the indices of multiple deprivation 2000 for each electoral ward or electoral division.16 18 *0.35% of male and 0.20% of female deaths could not be attributed to an electoral ward or electoral division and were excluded from the analyses. 5 ��� Most deprived 4 5 ��� Most deprived ��� observed data 3 2 1 ��� Most affluent Rate ratio: quintile 5/quintile 1 100 000 10 000 1000 100 0 10 100 80 Age (y) Males Females Annual mortality rate per 100 000 (log scale) 60 40 0 20 100 80 Age (y) 60 40 0 20 2 1.6 1.2 Rate ratio Figure 1 Annual deprivation specific mortality rates per 100 000: observed (most deprived only, squares) and smoothed (all quintiles, lines) and the mortality rate ratio between the smoothed rates of the deprived and the affluent (right hand axis), by single year of age and sex, England and Wales, 1998. * For simplicity, from here on we will use the term ������ward������ to refer to the electoral wards in England and the electoral divisions in Wales. �� Henceforth the term ������region������ will be used to refer to the nine government office regions (GOR) of England and Wales. 116 Woods, Rachet, Riga, et al www.jech.com