Elsevier

Social Science & Medicine

Volume 47, Issue 12, December 1998, Pages 2055-2066
Social Science & Medicine

Mortality variations in England and Wales between types of place: an analysis of the ONS longitudinal study

https://doi.org/10.1016/S0277-9536(98)00310-4Get rights and content

Abstract

This study investigates the extent to which individuals, in England and Wales, in different types of place experience differential mortality once account is taken of personal (individual and household) social circumstances. Data comes from the Longitudinal Study of England and Wales of the Office of National Statistics, the respondents being a one percent national random sample of people aged between 25 and 74 at the 1971 census, followed until the end of 1985. For males and females separately, differences in mortality are found for the 36 types of Craig–Webber classification in models which include, at the individual level, a number of demographic and socio-economic variables (women being classified by their own occupation). In general, for both males and females, the same types of place have elevated or lowered mortality. For males a (cross-level) interaction exists between the proportion in the area in professional social classes and individual social class, the effects of individual social class being larger in areas containing a higher proportion of those in professional occupations. For females mortality is negatively related to the proportion of car-ownership in the area.

Introduction

There is longstanding empirical evidence of substantial geographical differences in mortality within the United Kingdom. Indeed, Britton (1990) found a consistent and persistent difference for all-cause mortality for both men and women from the 1920s up to the period 1981–6, with the highest rates being found in the north and west of the country. Recent research by Drever and Whitehead (1995) and Charlton (1996) has also considered differences in ill-health and mortality between different types of place. A particular classification of types of place, the Craig–Webber national classification (see Webber and Craig, 1978; Webber, 1977) has been used by Fox et al. (1984), Britton et al. (1990) and Blaxter (1990) in examining mortality following the 1971 census and on morbidity data from the Health and Lifestyle Survey (Cox et al., 1987). This clusters small geographical areas into 36 types of place (Craig–Webber Types) based on socio-economic and demographic characteristics recorded at the 1971 census (see Table 1). All three studies found significant and substantial variations in health outcome between these types of place. Thus, Britton et al. (1990) using the ONS Longitudinal study, found that places characterised as ‘inner areas with low quality older housing’ had an Standardised Mortality Ratio (SMR) of 130 for males over the period 1971–1981 whereas in places identified as ‘rural established high status areas’ the equivalent ratio was only 83. A review of such place-specific studies is provided by Jones and Moon (1993), while Jones and Duncan (1995) consider the mechanisms by which such differences may be produced; Kaplan (1996) reports on similar work in the US from the Alameda County Cohort. Further research has been based on data from 1991 Census, using multilevel models; Shouls et al., 1996a, Shouls et al., 1996b constructed a socio-economic typology of Sample of Anonymous Records (SAR) areas and related this, together with individual characteristics, to limiting longterm illness and to mortality using multilevel models. Langford and Bentham (1996) related mortality at a local authority district level simultaneously to a classification of districts, ACORN (Craig, 1986), similar to the Craig–Webber classification and to ‘standard region’ controlling for social deprivation at the local authority district level. A number of calls have recently been made (e.g. Macintyre et al., 1993; Davey Smith et al., 1998) for further attention to be given to the collection, and use in such analyses, of data collected at or aggregated to the area level in conjunction with individual level data wherever possible.

A key interpretative question is the extent to which such observed place differences are ‘area’ or ‘ecological’ effects or are merely a result of different types of people living in these places. In this connection a useful distinction between ‘compositional’, ‘collective’ and ‘contextual’ effects is made by Macintyre (1997) and Macintyre and Ellaway (1996). The first refers to that notion that place differences are explainable purely in terms of differential demographic and social composition; the second relates to some form of ‘social miasma’ in which individuals conform to the behaviour of the dominant group living in an area; the third term represents the situation in which place characteristics have a direct effect. Thus, if the diet of a lower-social class person is similar to that of a higher-social class person when a place has a high proportion of the latter group, there is a collective effect. If the poor diet of the lower-classes is an outcome of inadequate retail facilities in a place, there are contextual effects. In the present study the term ‘area effects’ is used to encompass both collective, contextual and compositional effects (though we examine compositional effects specifically in one of the models used). Note that nomenclature in this research area is confusing, our terms ‘compositional’ and ‘contextual’ corresponding to Dies-Roux’s (1998) ‘derived’ and ‘integral’, our ‘compositional’ even corresponding to Susser’s (1994) ‘contextual’ (Susser, 1994, Shouls et al., 1996b). Note that Shouls et al. (1996a), Shouls et al. (1996b) also use the terms ‘contextual’ and ‘compositional’ in somewhat different senses, ‘contextual’ to refer to any effects at above the individual level and ‘compositional’ to refer to effects at the individual level only. We recommend that the term ‘compositional’ is reserved for effects at above the individual level which are due to the compositional characteristics of the higher level.

Research (McLoone and Boddy, 1994; McCarron et al., 1994; Bryce et al., 1994; Eames, et al., 1993; Davey Smith et al., 1998) often summarises area influences on mortality through an index of area deprivation, thus only assessing the effect of a particular summarisation, though often an essential feature, of the social composition of the area, this being conceived of and restricted to operation along a single dimension. Though there are obvious advantages to such data reduction, particularly when comparisons are made over time (McLoone and Boddy, 1994; McCarron et al., 1994; Bryce et al., 1994), this does not take account of differences across a number of dimensions (Shouls et al., 1996a; Dies-Roux, 1998). The distinction between effects at a higher level due to social composition, and therefore not to independent influence by policy initiatives and those directly effected by interventions at a higher level is paralleled in the educational research literature by Willms’ (1992) distinction between Type A effects (which include both compositional and contextual effects controlling only for individual pupil characteristics) and Type B effects (contextual only, which control also for compositional factors) (Willms, 1992).

Thus, do the high rates experienced in ‘inner-areas’ remain when allowance is made for the distinct socio-economic characteristics of the individuals who live in such areas? To put it another way, do people of similar characteristics experience different health outcomes in different types of places? While previous studies using the Craig–Webber typology have allowed for age and sex differences, there has only been a limited simultaneous analysis of individual and household social variables. An analysis, by Sloggett and Joshi (1994) on data from the same source (ONS Longitudinal Study) as ours finds that

excess mortality associated with residence in areas designated as deprived…is wholly explained by the concentration in those areas of people with adverse personal or household socio-economic factors’.

Their study is effectively based on classifying England into ten types of area through a composite deprivation index of four census variables (unemployment, lack of car, non-owner occupation, and socio-economic groups IV and V). While there is a consistent and linear relation between mortality and area deprivation when only age is controlled, this effect disappears entirely for males when the results are adjusted for individual and household factors which are the counterparts of the variables used to construct the index. The results are similar for females except that the relationship does not disappear altogether but is ‘strongly attenuated’ (p.1472). For males, they find the increased risk associated with living in the North as opposed to the South (a risk larger than comparing socio-economic groups IV and V with I to III), but this is described as a

persistent effect which probably reflected socio-economic factors not captured elsewhere by the model but which might have been cultural or environmental’.

They conclude, therefore that

the evidence does not confirm any social miasma whereby the shorter life expectancy of disadvantaged people is further reduced if they live in close proximity to other disadvantaged people’.

The aim of this paper is to re-examine the question: ‘do individuals in different types of place experience differential mortality once account is taken of individual social circumstance?’ This research (as does Slogett and Joshi, 1994) uses wards as the basic area units (average population around 10,000) which are amalgamated in this study prior to further analysis into the geographically dispersed Craig–Webber types (having average population 1.4 million), and in Sloggett and Joshi (1994) into those with similar deprivation score (the SAR districts, used by Shouls et al., 1996a, Shouls et al., 1996b, are based on amalgamations of administratively defined units, again from Census, which are substantially larger than wards having average population around 200,000 in contrast to the 10,000 average for wards). Although using the same LS data source as Sloggett and Joshi, this analysis contrasts with theirs in three ways. First the Craig–Webber schema is used to define places in a more refined form than the single axis of deprivation. Second, we explore the possible complexity of the interrelationships between individual and higher-level characteristics, namely census based indicators of the social composition of the Craig–Webber Types, these being component variables of the Craig–Webber schema. The underlying argument here is that, irrespective of any overall effect of an area level characteristic, there may be some subgroups of people for whom the effect is larger or smaller. Thirdly, we allow for overdispersion of error terms. This may be either due to influential explanatory variables omitted from the model or to the influence of effects of areas of residence which are either nested within or cut across the Craig–Webber Types.

This research therefore adds to the relatively small corpus of work so far which examines in the UK the effects on mortality of the compositional characteristics of areas once a range of individual characteristics have been taken into account.

Section snippets

The LS data

The ONS Longitudinal Study (LS) is a record linkage study comprising data from UK Censuses from 1971 onwards, and vital events (births, deaths, cancer registrations) for 1% of the Population of England and Wales effectively randomly selected on the basis of four ‘birthdays’. The sample is updated by the addition of new births and immigrants born on these (undisclosed) dates. Census information is also included for all those living in the same household as the LS members. Confidentiality in the

Statistical methods

The statistical model used for the analysis of this data is the Logistic model as follows;logit (p)=α+Xiβ+γjwhere β represents the effects of individual characteristics, Xi, γj represent area effects, considered as fixed effects and logit (p)=log(p/(1−p)). The model was run using PROC LOGISTIC within the SAS system (SAS Institute Inc., 1989). Over-dispersion of errors is explicitly allowed for (McCullagh and Nelder, 1989) and heterogeneity statistics are reported for all models. The approach

Modelling strategy

Given the flexibility of the modelling approach adopted, the number of different models that could be fitted to the available data is enormous, even when, as here, men and women are analysed separately. The strategy adopted is one of successive elaboration of an initial model, using stepwise regression methods, with both inclusion and exclusion criteria set at the 0.01 significance level.

The initial model relates the log-odds of dying to age through terms up to and including a cubic function. A

Models including socio-economic variables at the individual level only

This model (A) includes a range of socio-economic variables (and their interactions). The results, in terms of estimated parameters on a logit scale, are shown in Table 3. For interpretation, it is crucial to recognise that all effects are shown in relation to the base category so that the log-odds of dying over the 15 year period, for a 45–49 year old owner occupier, with access to car, economically active in social classes I or II is −2.68 for females and −2.14 for males. This transforms to a

Discussion and conclusions

Differences for both females and males have been found between Craig–Webber types in mortality over a 15 year period in models which allow for a range of socio-economic variables at the individual level. These differences are significant and can be judged relatively substantial with the odds for the extremes (times 100) ranging from 87 to 138 for females and from 79 to 122 for males.

For females, between place variations are found which are of a comparable order to those between tenure and car

Acknowledgements

Russell Ecob is employed by the Medical Research Council. The Longitudinal Study data are provided by permission of the Office of National Statistics (ONS) through the LS User support group at City University and we would like to thank them for their co-operation and in particular to thank Simon Gleave for his helpful negotiation through this process. The views expressed in this paper are the authors’ alone and not necessarily those of ONS. Thanks also to Rosemary Sheerer, Department of

References (36)

  • Cox, B. D., Blaxter, M., Buckle, A. L. J. et al. (1987) In The Health and Lifestyle Survey; preliminary report of a...
  • Craig, J. A. (1986) 1981 Socio-economic classification of local and health authorities of Great Britain. OPCS Studies...
  • Smith G. Davey et al.

    Individual social class, area based deprivation, cardiovascular disease risk factors, and mortality: the Renfrew and Paisley study

    J Epidemiol. Community Health

    (1998)
  • A.V. Dies-Roux

    Bringing context back into epidemiology: variables and fallacies in multilevel analysis

    American Journal of Public Health

    (1998)
  • F. Drever et al.

    Mortality in regions and local authority districts in the 1990’s: exploring the relationship with deprivation

    Population Trends

    (1995)
  • M. Eames et al.

    Social deprivation and premature mortality: regional comparison across England

    BMJ

    (1993)
  • A.J. Fox et al.

    Approaches to studying the effect of socio-economic circumstances on geographic differences in mortality in England and Wales

    British Medical Bulletin

    (1984)
  • Goldstein, H. (1995) Multilevel Statistical Models. (2nd). Arnold,...
  • Cited by (0)

    View full text