A hierarchical analysis of long-term illness and mortality in socially deprived areas

https://doi.org/10.1016/S0277-9536(00)00291-4Get rights and content

Abstract

This article is a multilevel analysis of the effects on self-reported long-term illness and mortality of the socio-economic position of the neighbourhood. Using data from the Swedish Level of Living Survey, neighbourhood social position is measured by a composite Care Need Index, (CNI) together with such indicators of individual socio-economic position as occupation and housing tenure, with adjustment for age, sex, marital status and social network. Data came from 22,236 people aged 25–74, and were collected from 1988 to 1992. The cross-sectional data were analysed using a hierarchical logistic regression model. In a second analysis, each participant was followed from the initial interview until his or her death, or until the termination of data collection (31 December 1996). A neighbourhood's low social position and an individual resident's low socio-economic position (i.e., a manual worker, or person renting a flat) were found to be associated with increased risk of long-term illness. We conclude that a neighbourhood's low socio-economic position, that is, a high score on the CNI, is a risk factor for long-standing illness above and beyond an individual's socio-economic position. The differences in mortality could be explained by the included independent individual variables.

Introduction

During the last few decades there has been increased awareness of the influence of social class on health. There is a growing body of data indicating that socio-economic differences in health are associated with individual characteristics as well as neighbourhood characteristics (i.e., compositional and contextual effects) (Duncan, Jones, & Moon, 1993; Macintyre, 1993; Duncan & Jones, 1995; Kaplan, 1996). A neighbourhood's compositional effects on health mean the aggregation of all individual characteristics, i.e., similar types of persons will have similar illness experiences no matter where they live. Purely contextual effects imply that individuals with similar characteristics will have different health status in different neighbourhoods. Susser (1994) emphasised that aggregated socio-economic characteristics of which an individual person is a part may relate to health in other ways than the individual socio-economic part. Thus, ecological research shows that the collective attributes of a population may have an impact on public health.

Another important theory that contributes to our model is Anthony Giddens's (1984) theory of structuration, which is partly based on modern time-space concepts in Swedish geography (Hägerstrand, 1982). Giddens's theory claims that the individual lives a routinised day-to-day life in different hierarchical local settings, such as family, work and school. Here the individual is exposed to interactions that exert an influence in many ways. We emphasise that the neighbourhood is such a local setting for socialisation in the context, with its conflicts, marginalisation or exclusion. However, to be able to differentiate between individual socio-economic characteristics (individual social position) and collective socio-economic characteristics (neighbourhood social position) adequate statistical methods such as multilevel and hierarchical analysis must be used. In our study the individuals’ data are level 1 and the neighbourhoods’ social position are level 2. By using multilevel technique it is possible to evaluate the contextual effect and the risk of the ecological fallacy is diminished.

An individual's social position is determined by that person's occupation, level of education, and different economic or material reference points. According to Macintyre (1993), the neighbourhood influences on health may be of different kinds such as physical environment (air pollution, water hardness, etc.), social position (socio-demographic factors, i.e., aggregated data from individuals living in the neighbourhood), socio-cultural factors of the neighbourhood, lay systems of behaviour and beliefs, availability of healthy and unhealthy environment, and service factors (health care, recreation areas, transport).

Various indices have been used to define the social position of a neighbourhood (Jarman, 1984; Townsend, Phillimore, & Beattie, 1988; Carstairs & Morris, 1991; Malmström, Sundquist, Bajekal, & Johansson, 1998). In England, for example, composite indices such as the Townsend score are used, taking the percentage of households living in non-owner-occupied housing, in overcrowded dwellings, with no car and with individuals who are unemployed (Townsend et al., 1988). Similarly, in Scotland there is the Carstairs score, which is based on the percentage of residents living in overcrowded households, households without cars, men unemployed, and all homes with a head of household in social classes IV and V (Carstairs & Morris, 1991). In Sweden there is still a scarcity of data on the combined effects of both compositional and contextual influences. The Care Need Index, (CNI) is based on eight modified (Malmström et al., 1998) Jarman indicators of socio-economic conditions (Jarman, 1984). In our study we have investigated the influences of the neighbourhoods’ social position by using CNI.

Over the past decade, several reports have demonstrated a strong association between different indicators of individual socio-economic position (e.g., occupation) and morbidity. These studies have surveyed relatively affluent countries, such as Sweden and other Scandinavian countries (Hallqvist, Lundberg, Diderichsen, & Ahlbom, 1998; Sihvonen, Kunst, Lahelma, Valkonen, & Mackenbach, 1998), and western Europe (Cavelaars et al., 1998). The data from seven west European countries demonstrated an above average prevalence of morbidity for skilled and unskilled manual men (including agricultural workers) as compared with higher- and lower-level professionals, a prevalence which was very similar between the countries (Cavelaars et al., 1998).

In spite of the existence of a political goal of equality in the UK, and the establishment of the Swedish welfare state, class distinctions were noted when mortality and long-term illness were examined in Britain and Sweden (Vågerö & Lundberg 1989). Moreover, the increase in mortality among middle-aged men in Sweden during 1965–1980 (a time of low unemployment and narrowing income differentials) was mainly the result of increasing cardiovascular mortality among industrial workers and farmers (Diderichsen & Hallqvist, 1997). However, in a Finnish study men who had an elementary education decreased their risk of limiting long-term illness between 1986 and 1994, while men with a secondary education actually increased such risk during the same time. Women's risk did not change in that period, i.e., women with an elementary education remained at high risk from 1986 to 1994 (Lahelma, Rahkonen, & Huuhka, 1997).

Scandinavian studies based on ecological data have shown an association between a deprived neighbourhood's environment and such indicators of poor health among its residents as increased admissions to psychiatric departments (Dalgard, 1980; Cullberg, Stefansson, & Wennersten, 1981; Malmström, Sundquist, Johansson, & Johansson, 1999b); a greater number of emergency room visits (Andrén and Rosenqvist, 1987; Sundquist, 1993); high CHD mortality (Malmström, Sundquist, Bajekal, & Johansson 1999a); and higher consultation rates to primary health care (Sundquist, 1993). In the northern region of England, the socio-economic patterning in mortality widened during 1981–1991 between the most affluent and most deprived fifths of wards in all age categories under 75 years (Phillimore, Beattie, & Townsend, 1994). Data from the Health and Lifestyle Survey demonstrated that for individuals with the same social position, self-reported health varied with conditions within their own neighbourhoods (Blaxter, 1990). However, another study found no independent effect of neighbourhood deprivation on mortality after individual social position was taken into account (Sloggett & Joshi, 1994). Wards located in deprived geographical districts exhibited a greater prevalence of illness than wards with the same social position located in more privileged geographical districts (Ben-Sholomo et al. 1996; Congdon, Shouls, & Curtis, 1997). Thus, the impact of small-area deprivation on health is partly defined by the wider geographical district. Congdon et al. (1997) used adequate multilevel models to separate the individual, the small area and the larger geographical district effect on health.

A study employing multilevel analyses showed that living in deprived neighbourhoods was associated with an increased prevalence of coronary heart disease and other risk factors, with these links generally persisting after adjustment for individual-level variables (Diez-Roux et al., 1997). In a survey from the Netherlands, while the age- and gender-adjusted prevalence of poor health and smoking were higher in deprived urban areas, most of the differences in poor health could be attributed to the generally lower socio-economic position of the residents in deprived areas (Reijneveld, 1998).

In the present study, we have used the CNI as a proxy for the social position of the neighbourhood (Malmström et al., 1998). By using multilevel (hierarchical) analyses, we can examine both the impact of an individual's socio-economic position and the impact of the neighbourhood on outcome factors.

We hypothesise that persons living in small, deprived, homogeneous neighbourhoods (according to CNI measurements) have a higher risk of self-reported long-term illness and increased total mortality, beyond their individual socio-economic positions.

Section snippets

Method

The present study consists of individual data from the Swedish Annual Level of Living Survey (SALLS), matched with the social position of the areas in which respondents lived — the latter being measured by CNI (Malmström et al., 1998). SALLS was a nation-wide random sample based on face-to-face interviews conducted from 1988 to 1992 (Statistics Sweden, 1996). It confined itself to individuals aged 25–74 and consisted of 22,236 interviews.

The non-response rate of SALLS was about 20%. Refusal was

Results

In Table 1 the distributions of SEI are presented. A higher percentage of junior salaried employees and a lower percentage of self-employed and farmers were women. There was no difference in poor social network between different groups. Workers were more likely to live alone than intermediate and senior salaried employees.

The percentage distribution of the independent variables in different CNI areas is shown in Table 2. The percentage of individuals living in rented flats and the percentage of

Discussion

The principal findings of the present study indicate that living in socially deprived neighbourhoods (second-level effect) contributes to increased self-reported long-term illness but not to increased mortality risk, after adjustment for age, individual SEI, housing tenure, marital status and social network. Moreover, low individual SEI was a strong risk factor for long-term illness and total mortality.

An advantage of this study is its focus on a large, well-defined, simple random sample of the

Conclusions

With the use of hierarchical analysis, we have demonstrated that the socio-economic position of a neighbourhood, the second-level effect, is generally more strongly associated with self-reported long-term illness than with mortality, and that this holds true above and beyond any single individual's socio-economic position — even after one adjusts for marital status, housing tenure and social network. However, the risk difference of total mortality in the CNI areas disappeared in the final

Acknowledgements

This study was supported by grants from the Swedish Medical Research Council (K99-27X-11651-04A), the Swedish Council for Social Research, and Axel and Margaret Johnson's Foundation to Jan Sundquist, and by a grant from the Council for Health and Health Care Research, Lund/Malmö, to Marianne Malmström.

References (54)

  • M. Blaxter

    Health and lifestyles

    (1990)
  • Bresliw, N. E., & Day, N. E. (1980). Statistical methods in cancer research, vol 1. Lyon, France: International Agency...
  • R.A. Carr-Hill et al.

    Socioeconomic determinants of rates of consultation in general practice based on fourth national morbidity survey of general practices

    British Medical Journal

    (1996)
  • V. Carstairs et al.

    Deprivation and health in Scotland

    (1991)
  • A.E.J.M. Cavelaars et al.

    Morbidity differences by occupational class among men in seven European countriesAn application of the Erikson-Goldthorpe social class scheme

    International Journal of Epidemiology

    (1998)
  • P. Congdon et al.

    A multi-level perspective on small-area health and mortalityA case study of England and Wales

    International Journal of Population Geography

    (1997)
  • J. Cullberg et al.

    Psychiatry in low status dwelling areas

    Psychiatry and Social Science

    (1981)
  • Curtis, S., Jones, I.R. (1998). Is there a place for geography in the analysis of health inequality? In M. Bartley, D....
  • Dalgard, O. (1980). Bomiljø og psykisk helse (Environment and mental health). Dissertation, University of Oslo, Oslo...
  • F. Diderichsen et al.

    Trends in occupational mortality among middle-aged men in Sweden 1961–1990

    International Journal of Epidemiology

    (1997)
  • A.V. Diez-Roux et al.

    Neighborhood environments and coronary heart diseaseA multilevel analysis

    American Journal of Epidemiology

    (1997)
  • C. Duncan et al.

    Individuals and their ecologies

    Journal of Health Place

    (1995)
  • A. Giddens

    Constitution of society

    (1984)
  • M. Gissler et al.

    Social class differences in health until the age of seven years among the Finnish 1987 birth cohort

    Social Science and Medicine

    (1998)
  • T. Hägerstrand

    Diorama, path and project

    Tijdschrift voor Economische en Sociale Geografie

    (1982)
  • J. Hallqvist et al.

    Socioeconomic differences in risk of myocardial infarction 1971–1994 in SwedenTime trends, relative risks and population attributable risks

    International Journal of Epidemiology

    (1998)
  • R. Haynes et al.

    Effect of labour market conditions on reporting of limiting long term illness and permanent sickness in England and Wales

    Journal of Epidemiology and Community Health

    (1997)
  • Cited by (55)

    • Do neighborhoods affect individual mortality? A systematic review and meta-analysis of multilevel studies

      2012, Social Science and Medicine
      Citation Excerpt :

      All studies used registers to obtain mortality or cancer incidence data. Individual-level and area-level data were obtained through either surveys (Anderson et al., 1997; Backlund et al., 2007; Bosma et al., 2001; Jones et al., 2000; Malmstrom et al., 2001; Waitzman, Smith, & Stroup, 1999), censuses (Bentley et al., 2008; Blakely et al., 2003, 2006; Borrell et al., 2002; Curtis et al., 2004; Franzini & Spears, 2003; Jaffe et al., 2005a, 2005b; Jerrett et al., 2003, 2005; Marinacci et al., 2004; Naess, Leyland, et al., 2005; Naess et al., 2007; Petrelli, Gnavi, Marinacci, & Costa, 2006; Turrell et al., 2007), registers (Blomgren et al., 2004; Chaix, Rosvall, Lynch, & Merlo, 2006; Chaix, Rosvall, & Merlo, 2007a, 2007b; Dahl et al., 2006; Henriksson, Allebeck, Weitoft, & Thelle, 2006, 2007; Kravdal, 2007; Martikainen et al., 2003) or through combinations (Chaix, Lindstrom, Merlo, & Rosvall, 2008; LeClere, Rogers, & Peters, 1998; Lochner et al., 2001; Mari-Dell’Olmo et al., 2007; Robert et al., 2004; Roos et al., 2004; Sanderson et al., 2006; Veugelers et al., 2001; Webster et al., 2008; Yen & Kaplan, 1999). Studies based on surveys with participation rates lower than 80% (Malmstrom et al., 2001; Roos et al., 2004; Veugelers et al., 2001) or where certain population groups were excluded (Bentley et al., 2008; Naess, Leyland, et al., 2005; Naess et al., 2007; Turrell et al., 2007) were rated as having moderate selection bias.

    • Self-perceived health space and geographic areas in Switzerland

      2008, International Journal of Hygiene and Environmental Health
    View all citing articles on Scopus
    View full text