Elsevier

Health & Place

Volume 6, Issue 4, 1 December 2000, Pages 275-285
Health & Place

Deprivation and poor health in rural areas: inequalities hidden by averages

https://doi.org/10.1016/S1353-8292(00)00009-5Get rights and content

Abstract

Poor health and social deprivation scores in 570 wards in East Anglia, UK, were much less associated in rural than in urban areas. The deprivation measure most closely related to poor health in the least accessible rural wards was male unemployment, but use of this measure did not remove the urban–rural gradient of association strength. Neither did replacing wards by smaller enumeration districts as the units of analysis. The differences between urban and rural correlations were removed by restricting the comparison to wards with the same unemployment range and combining pairs of rural wards with similar deprivation values. Apparent differences between rural and urban associations are therefore not due to the choice of deprivation indices or census areas but are artifacts of the greater internal variability, smaller average deprivation range and smaller population size of rural small areas. Deprived people with poor health in rural areas are hidden by favourable averages of health and deprivation measures and do not benefit from resource allocations based on area values.

Introduction

Poor health is strongly related to poverty in the United Kingdom (DHSS, 1980). Many studies at the local population level have demonstrated that large differences in death and illness rates between geographical areas reflect the distribution of social and economic factors. Variations in health between areas are to a large extent due to differences in the social composition of their populations (Sloggett and Joshi, 1994, Sloggett and Joshi, 1998), although there may be additional environmental and social influences working “above” the level of the individual (Duncan et al., 1998).

Government policies take social and economic factors into account in allocating resources to fund health services for local populations in particular areas. More money per capita for hospital services is allocated to districts with materially deprived populations than to more affluent districts, using socio-economic indicators derived from the national census which have been shown to predict demand for hospital care (Smith et al., 1994). General practitioners working in areas with high proportions of people likely to make heavy demands on them, identified using demographic and socio-economic measures from the national census (Jarman, 1983), receive more payment per capita than other GPs. Priority areas for local health service improvements are also identified using indicators of “social deprivation” derived from combinations of small-area census measures (Watt and Sheldon, 1993, Sims et al., 1997). The details vary but most indicators share similar ingredients, such as low status occupations, low house and car ownership levels, high unemployment, overcrowded living conditions and high proportions of vulnerable people living alone.

In general, levels of health in local populations reflect social deprivation measures, but this pattern appears to break down in rural areas. Even though 20% of the rural population of England live in absolute poverty (Cloke et al., 1994), rural death and illness rates tend to be low compared with those of urban areas (Haynes and Gale, 1999). Furthermore, the association between deprivation and poor health is much weaker in rural than in other areas. Townsend et al. (1988) first reported that the relationship between death rates and material deprivation in wards in Northern England was weaker in rural districts than in the towns or cities, and Carstairs and Morris (1990) made the same observation from Scottish data. This statistical evidence seems to contradict the experience of rural practitioners. They argue that although our knowledge of rural health need is limited, it is justifiable to assume that poverty and poor health are associated in rural areas just as they are in towns and cities (Cox, 1998).

Why should death rates and illness rates be less predictable from deprivation indicators in rural areas than in urban areas? One possible explanation is that the deprivation indicators used are urban-biased (Knox, 1985, Shucksmith, 1990). Car ownership, for example, may be a good proxy for material wealth in cities, but people in rural areas with low incomes make sacrifices in order to keep a car because alternative modes of transport do not exist. High levels of car ownership in rural areas may therefore create a misleading impression of favourable socio-economic conditions there. Urban and rural experiences of deprivation are different (Payne et al., 1996), so more sensitive indicators of rural social conditions might explain variations in health between rural communities more effectively.

Other suggestions have concentrated on the scale of analysis. In theory, a country can be divided into small areas in an infinite number of ways, and correlation coefficients are highly sensitive to both the number of areas and their boundaries, a manifestation of the “modifiable areal unit problem” (Openshaw, 1984, Amrhein, 1995). In practice, small areas defined in the national census are of most interest, since resource allocation policies are based on combinations of these. Local studies of urban environments in the UK often use wards (with typical populations of approximately 5000) as the geographical unit. Moore (1995) argues that wards are too large to capture rural variations, so enumeration districts (with populations of approximately 500 and assumed greater social homogeneity than wards) should be used in preference for rural health needs assessment purposes. Carr-Hill and Rice (1995), however, did not find that averages based on enumeration districts were better predictors of individual characteristics than ward values. The opposite view is that rural units should be made larger. In general, increasing the size of areas produces stronger correlations (Yule and Kendall, 1950). Phillimore and Reading (1992) demonstrated that the association between health and wealth could be strengthened by combining rural wards with similar deprivation scores to resemble more closely the larger populations of wards in urban areas. The same researchers also noted that health and deprivation inequalities are much wider in urban areas than in rural areas and that this, too, can affect the apparent relationship. Reducing the range of deprivation values in urban data to the range observed in rural wards makes urban and rural associations more similar (Haynes and Gale, 1999).

The aim of this study is to compare the strength of association between ill health and social deprivation in urban and rural small areas, and explain the difference by systematically examining the alternatives. If the difference is real, other determinants of health should be sought in rural areas. If it is due solely to artifacts of the methods used, then poor health is as much associated with poverty in rural areas as in cities. The answer might suggest the extent to which policies that allocate more resources to areas with high social deprivation scores are likely to improve health levels in rural settings.

Section snippets

Data

The units of analysis were 570 census wards in the East Anglia Region, comprising the counties of Norfolk, Suffolk and Cambridgeshire with a total population of 2 million. The region contained 576 wards at the time of the 1991 census, but one ward with a very small population was combined with another in the census and five wards were excluded from the study because they had anomalously high mortality and morbidity rates due to the presence of long-stay hospitals. Wards were divided into four

Ill health and deprivation in urban and rural wards

The first stage of the analysis was to compare the ill health and composite deprivation measures across the four geographical types of ward. Fig. 1 shows the means and typical range (indicated by twice the standard deviation) of the ill health measures in wards across geographical categories. A wide range of standardized ratios was found, with some ward values within every group differing by at least a factor of two. Mortality values were more variable than long term illness measures between

Correlation between ill health and deprivation measures

The second stage was to assess how much of the variations in ill health between wards could be explained by conventional measures of deprivation, by performing product moment correlations within each geographical category. Logarithmic transformations and polynomial expressions did not produce consistently better fits than a simple linear model so the linear model was applied.

Fig. 2a–c illustrates the determination coefficients (r2, adjusted for degrees of freedom) between the standardized

A better deprivation index for rural wards

The third stage was to identify the socio-economic indicator or combination of indicators that best explained the variations in ill health in outer rural wards. Taking the group of outer rural wards only, each of the measures in Table 1 was correlated with the standardized illness and mortality ratios in turn. Combinations of more than one predictor variable were tried using stepwise multiple regression.

Most of the 21 indicators of social deprivation were slightly associated with variations in

Enumeration districts rather than wards?

The most detailed scale of analysis possible using UK census data is at enumeration district (ED) level. Numbers of deaths in enumeration districts over a 3-year period are very small and highly subject to chance variations. The methods of this study are not suitable for a study of death rates in EDs. Limiting long term illness ratios for EDs are based on numbers an order of magnitude larger, which makes them more amenable to the same analysis as was applied to the ward data. After excluding 27

Associations within the same deprivation range

Fig. 4 combines two scatter diagrams showing the association between standardized mortality rates and unemployment rates in large urban wards and outer rural wards. From the figure, the difference in the two r2 values (r2=46% for large urban wards and r2=2% for the outer rural wards) appears to be partly due to the lack of rural wards with high deprivation values.

The four geographical categories all contained wards with low unemployment rates, but only urban wards included those with higher

Areas with similar populations

Average values based on large populations are less subject to chance fluctuations than values based on small populations, so the same data might show a stronger association when grouped into a few large areas than many small ones. In the study area there was a consistent difference in population size between urban and rural wards. The average population of large urban wards was 6637, it was 4278 for small urban wards, 2525 for inner rural wards and 2231 for outer rural wards; so rural wards had

More homogeneous areas with similar populations

Pairing wards simply on the basis of their population values had the disadvantage of combining some wards which had comparatively high unemployment rates with others which had low rates, to give averages in between that were not necessarily representative of either. Combining pairs of rural wards reduced the range of unemployment values to 3–8% in inner rural areas and 4–11% in outer rural areas. This problem was overcome by ranking the two sets of rural wards on the basis of their unemployment

Variations in the slope of the relationship

The argument has focused on the correlation between ill health and deprivation: the proportion of the variations in ill health that could be attributed to differences in deprivation levels. Another parameter of interest is the regression coefficient, or slope, which gives the change in ill health associated with a unit increase in deprivation. The regression coefficients between ill health and deprivation have been observed to be lower in rural wards than in other wards (Haynes and Gale, 1999),

Discussion

The study area, East Anglia, has better health than average in the UK, but nevertheless it contains local populations with very different limiting long term illness rates and death rates after adjusting for demographic structure. Mean illness and death rates in census wards were significantly lower in rural than in urban settings, but the rural wards with the highest rates were those distant from urban services. Conventional deprivation indices had the same pattern, with the worst scores in the

Acknowledgments

We thank Norfolk Health Authority for funding part of this study and an anonymous referee for comments.

References (44)

  • D. Campbell et al.

    Unemployment rates: an alternative to the Jarman index?

    British Medical Journal

    (1991)
  • R. Carr-Hill et al.

    Is enumeration district level an improvement on ward level analysis in studies of deprivation and health?

    Journal of Epidemiology and Community Health

    (1995)
  • V. Carstairs et al.

    Deprivation and mortality: an alternative to social class?

    Community Medicine

    (1989)
  • V. Carstairs et al.

    Deprivation and Health in Scotland

    (1990)
  • P.J. Cloke et al.

    Rurality in England and Wales 1981: a replication of the 1971 index

    Regional Studies

    (1986)
  • P. Cloke et al.

    Lifestyles in rural England

    (1994)
  • J. Cox

    Poverty in rural areas

    British Medical Journal

    (1998)
  • Report of the Working Group on Inequalities in Health

    (1980)
  • J. Fennel

    Health Care in Rural England

    (1992)
  • E.A. Fieldhouse et al.

    Deprived people or deprived places? Exploring the ecological fallacy in studies of deprivation with the Samples of Anonymised Records

    Environment and Planning A

    (1996)
  • C.E. Gehlke et al.

    Certain effects of grouping upon the size of the correlation coefficient in census tract material

    Journal of the American Statistical Association Supplement

    (1934)
  • R. Haynes et al.

    Unemployment rate as an updatable health needs indicator for small areas

    Journal of Public Health Medicine

    (1996)
  • Cited by (94)

    • Born into care: Associations between area-level deprivation and the rates of children entering care proceedings in Wales

      2022, Children and Youth Services Review
      Citation Excerpt :

      The literature on the associations between area deprivation and adverse outcomes for children yields mixed evidence. Several UK-wide and international studies have found positive associations between deprivation, low birth weight in children (Dibben et al., 2006), reduced physical and mental health (Bécares et al., 2012; Haynes & Gale, 2000; Lawlor et al., 2005), increased child neglect (Christoffersen, 2000; Eckenrode et al., 2014; Shook, 1999; Slack et al., 2004; Trickett et al., 1991), increased child mortality (Taylor-Robinson et al., 2019), and reduced social cohesion in families and communities (McCulloch et al., 2012, 2012). Particularly strong associations have been found between parental unemployment (Gillham et al., 1998; Slack et al., 2004) as well as area-level unemployment rates and child neglect: A Danish cohort study of children (Christoffersen, 2000) found positive associations between parental unemployment and increased incidence rates of children that were hospitalised due to neglect or abuse.

    View all citing articles on Scopus
    View full text